light manufacturing in africa? - world...

1

DRAFT FOR DISCUSSION

The World Bank Group

Light Manufacturing in Africa? Practical Solutions to Creating Millions of Productive Jobs

With a Case Study of Ethiopia

VOLUME III

BACKGROUND PAPERS

Funded by MDTF, Japanese PHRD TF096317 and Dutch BNPP TF 09717

Development Economics, Operations and Strategy, DECOS

Africa Finance & Private Sector Development, AFTFP

July 26, 2011

2

TABLE OF CONTENT

COLLECTION OF PAPERS

BACKGROUND PAPER I

RESULTS FROM THE QUANTITATIVE FIRM SURVEY MARCEL FAFCHAMPS AND

SIMON QUINN MAY 2011 .......................................................................................................... 3

BACKGROUND PAPER II

THE BINDING CONSTRAINT ON FIRMS‘ GROWTH IN DEVELOPING COUNTRIES

HINH T. DINH, DIMITRIS A. MAVRIDIS, HOA B. NGUYEN NOVEMBER 2010 .............. 75

BACKGROUND PAPER III

ASSESSING HOW THE INVESTMENT CLIMATE AFFECTS FIRM PERFORMANCE IN

AFRICA: EVIDENCE FROM THE WORLD BANK‘S ENTERPRISE SURVEYS GEORGE

CLARK , MAY 2011 .................................................................................................................. 164

BACKGROUND PAPER IV

WAGES AND PRODUCTIVITY IN MANUFACTURING IN AFRICA: SOME STYLIZED

FACTS GEORGE CLARKE FEBRUARY 2011 ...................................................................... 191

Acknowledgments Under the guidance of Oby Ezekwesili (Vice President, AFR) and Justin Lin (Sr. VP, DEC and Chief Economist) this report was prepared by a core team consisting of Hinh T. Dinh (Team Leader and Coordinating Author), Vincent Palmade (Lead Economist and Co-Team Leader), Vandana Chandra (Sr. Economist), Frances Cossar (Junior Professional), Tugba Gurcanlar (Consultant), Ali Zafar (Sr. Economist), and Gabriela Calderon Motta (Program Assistant). The larger team responsible for the report includes, in addition to the above staff, George Clarke (Texas A&M International University), Kathleen Fitzgerald, Ying Li, Thomas Rawski (University of Pittsburgh), H. Colin Xu, Yutaka Yoshino, and Douglas Zeng (Washington); Marcel Fafchamps and Simon Quinn (Oxford University, England); Anders Isaksson (UNIDO, Austria), Mesfin Girma Bezawagaw, Nebel Kellow, Menbere Taye Tesfa (Ethiopia); Le Duy Binh and Pham Thai Hung (Vietnam); Lihong Wang (China); George Gandye, Josaphat Paul Kweka, and Michael Ndanshau (Tanzania); Tetsushi Sonobe and Aya Suzuki (The National Graduate Institute for Policy Studies (GRIPS), Japan); the Global Development Solutions team (Washington) : Yasuo Konishi, David Philipps, Glenn Surabian, Atdhe Veliu, John Weiss, Nebiye Gessese, and Christine Elbert; and Precise Consult team (Ethiopia). The work was carried out with the support and guidance of Marilou Uy (Sr. Advisor, MDM and Former Director, AFTFP), Gaiv Tata (Director, AFTFP), Shanta Devarajan (Chief Economist, AFR); Zia Qureshi (Sr. Advisor, DECOS), Greg Toulmin (Acting Country Director, Ethiopia), Ann E. Harrison (Former Director, DECVP), Asli Demirguc-Kunt (Director, DECVP), and Shahrokh Fardoust (Director, DECOS). The report also benefited from key inputs from government officials in Ethiopia, Tanzania, Zambia, China and Vietnam as well as from hundreds of private sector entrepreneurs interviewed in these five countries. In Ethiopia, we thank Minister Ato Neway Gebre-Ab (Chief Economic Adviser to the Prime Minister) and State Minister for Industry, Ato Tadesse Haile for valuable comments. In China, we thank Messrs. Gao Fu and Li Qiang of the Ministry of Industry and Information Technology (MIIT), as well as officials from Jiangxi and Zhejiang Provinces, particularly Mr. Junming Wan, Ms. Huan Ren, and the Chinese Association of Development Zones for arranging the enterprise visits and for carrying out the quantitative survey in China. In Vietnam, we thank the Vietnam Chamber of Commerce, particularly Dr. Pham Thi Thu Hang, for organizing the enterprise visits and for executing the quantitative survey. Throughout the preparation of this report, the team received valuable advice and guidance from an external advisory committee consisting of Yaw Ansu (African Center for Economic Transformation), Augusto Luis Alcorta (UNIDO), William Lewis (Founding Director, McKinsey Global Institute), Howard Pack (University of Pennsylvania), Jean-Philippe Platteau (Universite of Namure, Brussels), Kei Otsuka (The National Graduate Institute for Policy Studies (GRIPS), Japan), John Sutton (London School of Economics), Alan Gelb and Vijaya Ramachandran (Center for Global Development). The peer reviewers are Ann E. Harrison, Ioannis N. Kessides, John Murray Mcintire, David McKenzie, Brian Pinto, Vijaya Ramachandran, and Tunc Tahsin Uyanik. In addition, the team has benefited from comments by Asya Akhlaque , Jean Francois Arvis, Paul Brenton, Hai-Anh Dang, Nora Dihel , Doerte Domeland , Michael O. Engman, Thomas Farole, Gary Fine, M.

Louise Fox, Ian Gillson, Alvaro Gonzales, Michael Fuchs, Mombert Hoppe, Xiaofeng Hua, Leonardo Iacovone, Guiseppe Iarossi, Celestin Monga, Dominique Njinkeu, Paul Noumba, Gael Raballand, Ganesh Rasagam, José Guilherme Reis, Frank Sader, Marie Sheppard, Papa Demba Thiam, Pham Van Thuyet, Volker Treichel, James M. Trevino, Dileep Wagle, and Chunlin Zhang. The report was edited by a team headed by Bruce Ross-Larson at Communications Development. Earlier drafts were edited by Alison Strong and Paul Holtz. Financial support from MDTF, BNPP, and PHRD is gratefully acknowledged.

3

BACKGROUND PAPER I

Results from the Quantitative Firm Survey

Marcel Fafchamps and Simon Quinn

May 2011

Summary of the findings

This report presents the results from recent quantitative firm surveys undertaken in two East

Asian countries – China and Vietnam – and three African countries – Ethiopia, Tanzania and

Zambia. The focus of the study is on small and medium size manufacturing firms in five light

manufacturing sectors. In some countries, such as China, it proved difficult to find very small

firms in these sectors while in other countries, notably Tanzania, it was difficult to find firms

other than very small. Because many business practices are correlated with firm size, the analysis

presented corrects for size whenever necessary before making comparisons between countries.

We find that, once we control for size differences, firms in the five study countries are similar

along many dimensions, or show little systematic difference between Asia and Africa, notably:

1. Owners and managers of small and medium size firms are predominantly nationals.

2. In all countries except Tanzania the majority of sample firms have some form of business

registration. There is, however, considerable heterogeneity in registration rates across

countries for smaller firms, with a systematically lower registration probability in Africa

than in Asia.

3. Few firms pay a penalty for operating without registration or license.

4. The proportion of firms that have an account with an electricity provider is sizably less

than the proportion of firms using electricity for production.

5. We see few differences across countries in terms of perceived competition, with around

half of the firms responding that their market is moderately competitive and the rest that

it is very competitive.

6. The relationship between the number of business contacts and firm size is weak and

largely non-significant. Why entrepreneurs help each other if this does not appear to

improve performance is unclear. One possibility is that usage of social networks is

endogenous: entrepreneurs who manage to access information without recourse to social

networks tend to be more talented entrepreneurs.

7. Except for Tanzania where firms seldom innovate, there is no difference between China

and the other three countries in terms of introduction of new products and changes in

4

delivery system. When we control for size, Chinese firms are seen to innovate less, not

more, than firms in the other four countries – including Tanzania. The data nevertheless

show that innovation is associated with faster firm growth and that this association is not

stronger in China than in the other study countries.

8. In all countries, firms finance innovation and investment mostly through retained

earnings.

9. Chinese and Vietnamese manufacturers do no combine inputs from a larger number of

suppliers than African firms of the same size.

10. It is not the case that Chinese or Vietnamese manufacturers have more alternative

suppliers than firms in Africa: if anything, firms in Ethiopia and Tanzania report a larger

number of alternative suppliers, although this difference is not significant.

11. There is no significant difference across countries in terms of custom-made production.

12. The share of production workers in total employment is largest in Vietnam and lowest in

Zambia. The other three countries have a relatively similar breakdown between

production and non-production workers.

13. In all five countries at least half of the firms report having purchased or acquired

machinery, equipment, or vehicles in the past three years.

14. After controlling for size, firms in China and Ethiopia earn higher profits than firms in

Vietnam and Tanzania, with Zambia occupying an intermediate position.

15. There is substantial overlap in growth experiences across the firms in the study. It is not

the case that most firms in China are growing much more rapidly than firms in Africa or

Vietnam. Median growth in output is strong across all five countries and, if anything, is

lower in China than in the other countries except Zambia.

The data documents a number of significant differences between China and the three African

countries in the sample, with Vietnam being more like China or more like Africa depending on

the question. These findings can be summarized as follows:

1. China has more foreign-owned small and medium firms than the other countries, albeit

the owner is nearly always ethnic Chinese.

2. A smaller proportion of Chinese firms are operated by a woman.

3. Firms in China and Vietnam are much more likely to have a limited liability status than

firms in Tanzania and Zambia. But Ethiopia is more like China and Vietnam on this

issue. These differences are largely accounted by size. However, Chinese firms appear

5

less likely to enjoy liability status than their larger size would predict. The same is true

for Tanzania.

4. Chinese and Vietnamese firms have better educated owners and managers, but there is a

lot of variation among the three African countries in the study.

5. Chinese entrepreneurs are more likely to have a farming background and less likely to

have wage earning parents.

6. Entrepreneurs in China and Vietnam are less likely to have lived abroad or in other parts

of the country than those in (some of) the African countries.

7. In the two Asian countries, a majority of respondents list family members and relatives as

contributors in terms of ideas, technical expertise, and financing. The proportions are

much smaller in Africa, especially for technical expertise and financing. Chinese

respondents who are much more likely to list business acquaintances, experts and

consultants, clients, employees, and equipment suppliers as sources of assistance at start-

up. African respondents relied mostly on their own resources at startup and initiate much

smaller firms. Assistance at start-up is also a strong predictor of future firm performance,

although the correlation need not be causal.

8. In China and Vietnam, the majority of sample firms are registered for value added tax,

implying that their manufacturing output is part of the country‘s tax base.

9. The number of licenses required to operate a business is much higher in China than in the

other four surveyed countries. Tanzania is the sample country with the smallest number

of licenses but also with the smallest average firm size.

10. Corruption is more prevalent in Vietnam than in the African sample. In China the

question was deemed too sensitive to be asked, which suggests that incidence may even

be higher.

11. Except in China, the majority of studied firms experience power outages on a regular

basis. The incidence of outages appears particularly high in Vietnam, Ethiopia, and

Tanzania. When outages occur, they tend to be more frequent in the African countries in

our sample.

12. Geographical clustering is lowest in China, perhaps because firms there are larger and

thus cannot be physically located as close together.

13. Chinese firms list a smaller number of competitors in total. Asian firms also face the

lowest level of foreign competition while firms in the three African countries face the

highest, even though two of them are landlocked. One candidate explanation is that firms

in China and Vietnam may benefit from trade protection. Alternatively, Asian firms may

6

be more competitive. The proportions of respondents who cite China as source of foreign

competition is indeed large in all countries, including Vietnam.

14. The evidence suggests that producers seek to compete with (Chinese) imports primarily

on quality, less on price and design. This in turn suggest that the manufacturing imports

they compete with are low quality, low cost mass produced items that suit well the

limited budget of local consumers.

15. Membership in associations is more prevalent in China than in the other four studied

countries. Part of this difference can be attributed to variation in firm size.

16. Most respondents have friends or relatives in banks, government, or politics. Proportions

are highest in the three African countries. We cannot tell from the data whether this is a

consequence of entrepreneurship, a cause of entrepreneurship, or a channel through

which entrepreneurial success is achieved.

17. The number of business contracts mentioned by respondents is much lower in China than

in the other four countries. There is, however, some reluctance among Chinese

respondents to talk about their business contacts, so that the data may be misleading. In

the other four countries, firms typically have many contacts with other firms, and these

contacts are used for various business-related purposes.

18. Imitating other producers is a more powerful driver of product innovation in the other

four countries than in China.

19. Firms in the two Asian study countries, and particularly those in China, use a much wider

range of sources of information on the technical expertise needed to develop new

products or change the production process.

20. Asian – and particularly Chinese firms – seem to have access to a larger variety of

funding sources for innovation purposes.

21. Imported material inputs are important for manufacturing in all study countries except

China where imports represent a smaller proportion of all inputs.

22. We do not observe a significant difference between countries in the extent to which firms

extend credit to their customers. However, firms in China and Vietnam are much more

likely to purchase inputs on credit than their African counterparts. This is particularly true

among small firms which are typically most in need of external finance.

23. Chinese firms ship a substantially lower proportion of sales to customers in the same

city/district – and more to consumers in other districts – than do firms in the other study

countries. This holds even after controlling for firm size.

7

24. Chinese firms sell a much higher proportion of their output to other manufacturing firms

than African firms, with Vietnam occupying an intermediate position. This means that

gain from firm specialization are not captured in Africa.

25. Asian firms sell a substantially higher proportion of their production to their main

customer. They do not have a larger number of alternative customers to whom they

could sell. This suggests a higher degree of vertical specialization, reflected in specific

firm-to-firm relationships

26. Once we control for size, Asian firms are, on average, less likely than African firms to

engage in advertising.

27. Almost 40% of the Chinese firms in our sample export part of their output. The

corresponding figure for Vietnam is 17%. Manufacturers in the three African study

countries are much less likely to export. The destination of African exports is also

different, i.e., primarily other African countries.

28. Once we control for size, China and Tanzania have a significantly lower proportion of

permanent workers relative to the other three countries.

29. A much higher proportion of Chinese and Vietnamese firms make use of formal methods

for recruiting workers than in Zambia, Ethiopia, and Tanzania.

30. Except in China, most firms state that, without legal or regulatory restriction they would

not lay off any worker. This suggests that labor market restrictions are more binding in

China.

31. The only country in our study where a sizeable share of workers belongs to a union is

China. Housing provision to (some) workers is common in China but rare elsewhere.

32. In China and Vietnam, only a small proportion of production workers have less than 9

years of schooling. In Africa education levels are lower, with a lot of variation between

countries.

33. In Ethiopia and China, 85% to 90% report that it takes at most four weeks for new

workers to be fully trained. In contrast, in Tanzania and Zambia, less than 60% of

respondents report that new workers are trained in 4 weeks or less.

34. Asian firms are more likely to finance investment through bank loans than African firms

in our study. Small Chinese firms are also substantially more likely to have a bank

account than firms of similar size in the other four countries.

35. With respect to overdraft facilities China stands out in sharp contrast to the other four

countries: 63% of firms with a bank account have an overdraft facility whereas this

proportion is negligible in the other four countries. Chinese firms report a median annual

8

interest rate on the overdraft of 7.5% and only about 20% of them were required to

provide collateral, with an average value lower than the overdraft limit. Chinese firms are

also more likely to have a savings account.

36. Chinese and Vietnamese firms make more extensive use of credit than do firms in

Ethiopia, Tanzania or Zambia. Even after controlling for firm size, Chinese firms are

shown to face substantially lower average collateral requirements on loans.

37. Chinese firms produce significantly more per worker than Vietnamese firms which, over

much of the firm size range, produce more than Ethiopian, Tanzanian or Zambian firms.

38. Labor cost per worker is higher in China than in Vietnam and Zambia, whose costs are

higher again than in Ethiopia and Tanzania.

39. Over much of the firm size range, Vietnamese firms have more capital than their African

counterparts. Large firms in Ethiopia and Tanzania, however, appear to have more capital

invested in the firm than large Vietnamese firms. Chinese firms refused to answer

questions on capital.

40. Controlling for firm size, Chinese firms spend more on land and building than firms in

Vietnam which, in turn, spend more than in Zambia. The proportion of non-answers is

high, however, casting some doubt on these figures.

These findings are unexpected in many ways. China‘s success in manufacturing growth and

exports has struck many people‘s imagination, especially compared to what is often perceived as

a dismal manufacturing performance in Africa.

The picture that the data paints is quite different, with healthy growth rates for the African firms

in the sample. Ultimately, the main difference between China and the other countries is that

average firm size is much larger and manufacturing represents a sizeable proportion of GDP.

Hence a 14.8% growth rate in sales has a large effect on aggregate growth. This is not true in the

other countries, and particularly in the three African countries in our sample, where firms are

smaller and manufacturing only represents a minute portion of domestic GDP.

Why then is the manufacturing sector larger in China? The survey results presented here suggest

that, whatever the reasons for China‘s success relative to Africa, it is unlikely to be due to easier

regulation. If anything, China seems to have more stringent registration requirements and labor

laws. It is unlikely to be corruption, which seems to generate more anxiety among Chinese

respondents than in the other countries: if anything, faster growing firms in the sample are more

likely to report having to pay government officials to get things done. It also cannot be due to

lower labor or land costs, which in fact appear higher in China. It cannot be social networks: if

anything Chinese firms report having fewer links with banks and politicians, and fewer business

friends. It is unlikely to be entrepreneurial experience: Chinese entrepreneurs have travelled less,

9

have fewer friends abroad, and are more likely to come from an agricultural family. There also

are no strong differences across countries in the rate at which individual firms innovate and

invest.

The dimensions along which Chinese firms appear at an advantage are few. The first one is

finance: many Chinese firms seem to have access to bank finance at favorable conditions, e.g.,

low interest rates and very low collateral requirements. The second one is information about

innovations: Chinese firms are much more likely to rely on external experts than African firms at

start-up as well as subsequently when introduce new products, change their technology, or

modify their distribution system. The third one is competition: firms in China and Vietnam face

less competition from imports, which suggests some form of direct or indirect trade protection.

In contrast, manufacturing firms in Africa face stiff competition from imports, primarily from

China. The fourth is education: Asian workers and entrepreneurs in general have more schooling.

Education, however, does not predict how quickly production workers are fully operational, so it

is unclear how much of an advantage schooling is for production workers.

It is impossible from a simple cross-section survey to ascertain which of the above factors has

had a positive causal effect on China‘s manufacturing success. But the results presented here,

even though they can only document correlations and patterns, may nevertheless force us to

reconsider some of our preconceptions.

10

The Research Program Context

This report is part of a World Bank research program aimed at providing new insights into why

industrialization in Sub-Saharan Africa (SSA) in the light manufacturing sector has been slow,

while it has grown in countries with similar levels of economic development. The objective of

the research program is to inform the Bank's policy dialogue and design to foster the emergence

of SSA's light manufacturing sector. Ultimately, the aim is to help Sub-Saharan Africa diversify

from its over-dependence on unprocessed primary commodities and minerals towards a

competitive light manufacturing sector which transforms its current resource-base into higher

value added products that are consumed and eventually exported.

The research program includes five complementary modules conducted in parallel. The first four

are as follows:

1) Product mapping records what kinds of simple manufactured products are produced locally.

This activity involves market research to identify products that are 1) only imported, 2) only

produced by African firms, and 3) both imported and produced locally. Their quality, price,

markets where sold and country of origin is recorded to shed light on the products that African

firms produce competitively. This also shows who these firms compete with in the local market

for simple products.

2) Value Chain Analyses provide a microeconomic framework within which to assess, for each

selected product‘s value chain, the relative performance of countries in terms of productivity and

costs as well as to identify the main factors that cause low productivity and high costs. Value

Chain Analyses in particular highlight the critical importance of industry specific factors (e.g.

product market regulations and market failures) which tend to be overlooked by traditional cross-

cutting approaches.

3) Qualitative entrepreneur surveys cover formal and informal firms and large and small

firms. The objective is to understand what factors enable certain types of entrepreneurs to

produce new, more and better products and flourish relative to African entrepreneurs in the same

investment climate. The firm specific narratives provide deeper insights into the source of

coordination and information failures that disadvantage African entrepreneurs and constrain

industrial development in Africa. They also shed light on the complex ways in which social

networks affect entrepreneurs and enable some to innovate and compete more than others.

4) Training of entrepreneurs gauges how government interventions in the form of targeted

training programs for managers can boost firm profits by providing critical information

associated with technology, financing, marketing and managerial skills.

The fifth component is a combination of Quantitative firm surveys and randomized

experiments that cover formal and informal firms in three African countries and two East Asian

11

countries. This report presents the results from the quantitative surveys. The results from the

randomized experiments will be presented when the experiments are completed.

Investment Climate Assessments (ICAs) surveys have provided valuable information on the

investment climate and the external environment in which firms operate. They seek to identify

policy-related constraints that need to be alleviated to promote a more conducive business

environment. Many of these constraints relate to regulatory and bureaucratic hurdles, corruption,

credit and interest rates, and availability of public services such as water, electricity and roads.

The policy implication of ICAs is that the government can contribute to a more conducive

business environment by alleviating these constraints. The presumption is that firms will

respond positively to an improved investment climate.

ICA surveys are less suitable to investigate why producers facing the same investment climate in

a given country produce some products and not others. The purpose of the quantitative surveys is

to identify differences between firms and entrepreneurs which are associated with differences in

product choice, innovation, and firm performance. In particular, the surveys will examine the

role of social networks in determining the introduction of new products and the adoption of

innovations in production, marketing, and transaction technology.

The quantitative firm surveys

Quantitative surveys were conducted in five countries. Three of those – Ethiopia, Tanzania, and

Zambia – are in Sub-Saharan Africa. The other two – China and Vietnam – are in East Asia.

A. Questionnaire

An original questionnaire was developed by the World Bank, building on the expertise and

experience gained from other enterprise surveys, notably:

- The Regional Program for Enterprise Development (RPED) surveys in Africa, and their

subsequent continuation by the Centre for the Study of African Economies (Oxford

University).

- Firm and Competitiveness Surveys (FACS)

- Investment Climate Assessment (ICA) surveys

- Small enterprise surveys conducted by CSAE in Africa and South Asia

An example of the questionnaire is attached to this report.

In each of the five countries, the firm survey was conducted by a local team under the

supervision of the World Bank or the Oxford team. The local teams were as follows:

12

1. China: Ms Huan Ren

2. Ethiopia: Partner organization: Ethiopian Development Research Institute (EDRI). Team

leader: Dr. Gebehiwot Ageba

3. Tanzania: Partner organization: EDI. Team leader: Ms. Mujobu Moyo

4. Vietnam: Ms. Hang T. Pham

5. Zambia: Partner organization: RuralNet. Team leader: Mr. Stephen Tembo

B. Sampling frame and firm selection

The objective was to obtain a randomly selected representative sample of firms of a given size

range in the five target sectors. Given that the focus of the study is on small and medium-size

manufacturing firms but not on microenterprises, samples were drawn from firms in the main

urban centers. In each of the three African countries in the study, we focus on the capital city

where the overwhelming majority of small and medium size manufacuting is found.

In each country a sampling frame was constructed from firm lists obtained from the Bureau of

Statistics, Chambers of Commerce, or other similar official entities. In Tanzania, these sources

did not provide sufficient coverage of small and informal firms, so the sampling frame was

complemented by firms selected in areas with a concentration of informal firms in the target

sectors and size range.

A sample of surveyed firms was drawn by stratified random sampling from the firms listed in the

sampling frame. In each of the three African countries the sampling frame focuses on the capital

city and the area immediately adjacent to it given that most light manufacturing firms operating

in these countries can be found there. In Tanzania, a number of informal firms were added

through random walk in geographical areas with a concentration of target firms.

Sectors

In each of the five countries the study aimed to cover 250 randomly sampled manufacturing

firms, equally divided across five broadly defined sectors of activity:

1. Food processing and beverages

2. Garments

3. Leather products

4. Metal products

5. Wood products

13

We show in Table 1 the breakdown of firms according to sector. An attempt was made in all

countries to achieve an equal breakdown of the sample over the five sectors listed above.

Unfortunately, in some countries it proved difficult if not impossible to identify a large enough

sample of firms of the intended size in certain sectors – e.g., leather in China and food processing

in Tanzania.

Table 1: Breakdown of the samples by sector

Country sample

China Vietnam Ethiopia Tanzania Zambia

Food processing & Beverages 47 62 53 21 38

15.5% 20.7% 21.2% 8.0% 14.4%

Garments 71 62 48 58 65

23.4% 20.7% 19.2% 22.1% 24.7%

Leather products 26 52 49 37 42

8.6% 17.3% 19.6% 14.1% 16.0%

Metal products 85 62 46 50 61

28.1% 20.7% 18.4% 19.1% 23.2%

Wood products 74 62 54 96 57

24.4% 20.7% 21.6% 36.6% 21.7%

Total 303 300 250 262 263

% of total reported in italics

Firm size

The sample is intended to cover small and medium size firms with 2-40 paid permanent

employees excluding household members. To the extent possible, the initial intent was to divide

the sample of interviewed firms more or less equally, within each sector, between small (2-20

permanent employers) and medium (21-40 employees) firms. In Tanzania and Zambia it proved

difficult to identify a sufficient number of firms in each of the five sectors. Consequently the

lower employment limit was lowered to 1 and the upper limit was raised to 100 in an effort to

meet sample size targets. In China the survey team met considerable difficulties finding a large

enough number of small firms in the sectors of study.

It was not always possible to assess the size of sampled firms prior to conducting the survey.

Information on firm size is not always available in the sampling frame and, when available, is

not always accurate or current. Furthermore, respondents contacted to set up an interview

appointment often are reluctant to admit the true size of their workforce over the phone. There is

also considerable ambiguity as to which workers are regarded as permanent employees – e.g., a

firm with 60 workers may only list 3 of them as permanent employees. It follows that, in many

14

cases, the true size of the firm only becomes apparent when the labor force module of the

questionnaire has been completed.

Table 2 : Breakdown of sample firms by # of permanent employees

Country sample


<2 5 2 6 133 57

1.7% 0.7% 2.4% 50.8% 21.7%

2-20 69 163 221 125 176

22.8% 54.3% 88.4% 47.7% 66.9%

21-40 34 84 18 1 15

11.2% 28.0% 7.2% 0.4% 5.7%

>40 85 51 5 2 14

28.1% 17.0% 2.0% 0.8% 5.3%

Missing information 110 0 0 1 1

36.3% 0.0% 0.0% 0.4% 0.4%

Total 303 300 250 262 263


In Table 2 we show the distribution of the number of reported permanent employees in each of

the five study countries. For Vietnam, Ethiopia and Zambia, the bulk of the sample is within the

intended lower and upper bounds in terms of permanent employees, but only Vietnam

approximately reaches the intended equal split in firms with fewer and more than 20 permanent

employees. In Ethiopia and Zambia it was simply not possible to identify enough firms in the 21-

100 permanent employee range. In Tanzania the situation was even more difficult and more than

a third of surveyed firms have no permanent employees. They were nevertheless made their way

into the survey because they employ casual workers on a regular basis (see below).

In China we encountered the opposite problem, i.e., the difficulty of identifying small enough

firms: based on reported numbers of permanent employees, only a little over half of the surveyed

firms have less than 100 permanent workers. But more than a third of the surveyed firms refused

to report the number of their permanent employees. There is also a large discrepancy in the

Chinese sample between the reported number of permanent employees and the size of the regular

workforce. Firms were asked to report their number of workers in various occupations, e.g.,

management, clerical, skilled, and unskilled production workers. Summing over their answers

yields what we feel is an accurate measure of a firm‘s regular workforce, that is, the number of

positions the firm has in various occupations. From these answers we get a dramatically different

picture of the employment size of the firm. The proportion of Chinese firms with missing

15

employment data drops from 36% to 16%, suggesting that firms are less worried about

discussing their regular workforce than they are about declaring which proportion of their

workforce has a permanent contract. This casts further doubt on the accuracy of the reported

number of permanent workers as valid measure of the economic size of the firm. This is further

confirmed by noting that in terms of regular workforce, fewer than 5 percent of surveyed

Chinese firms have less than 20 employees while and 36% have more than 100. In other words,

many surveyed Chinese firms are much larger than intended.

In contrast, in terms of regular workforce, we find that in Ethiopia and Zambia 87% of surveyed

firms have between 1 and 20 workers – much more than the intended half of the sample. This

proportion rises to 94% in Tanzania. The reason for this is the dearth of medium size firms in

Sub-Saharan Africa, a reality that has sometimes been coined as the ‗missing middle‘ in the size

distribution of firms.1

The distribution of (the logarithm of the) regular workforce in the three countries is depicted in

Figure 1. We see that there are large differences in firm size between the five country samples.

The Vietnam sample follows the closest the intended firm size distribution. For China, the entire

distribution of the firm‘s workforce is shifted to the right, but with a sizeable overlap with

Vietnam. In contrast, firm size distribution is shifted to the left in the three African samples, with

some differences between them as well: the Tanzanian and Ethiopian samples are fairly tightly

distributed while the Zambian sample is more diffuse – with more medium size firms as well as

more small firms than either the Tanzania or Ethiopia samples.

1 See Fafchamps (1994) for a survey of the literature on the possible reasons for the existence of a missing middle

in the size distribution of firms.

16

Figure 1: Distribution of Regular Workforce by Country

To confirm that the differences in firm size distribution between countries as depicted in Figure 1

is not an artifact of the way employment is computed, we conduct the same exercise using total

sales expressed in US$ equivalent. The results are shown in Figure 2. The pattern apparent in

Figure 1 is by and large confirmed: the three African samples have smaller firms while China has

larger firms and Vietnam sits somewhere in between.

0

.2

.4

.6

Density

0 2 4 6 8 Log employment

China Vietnam

Ethiopia Tanzania

Zambia

17

Figure 2: Distribution of Log Sales by Country

The bottom line is that firm size varies dramatically between countries in our sample. This makes

direct comparison across countries perilous because many of the practices and outcomes we wish

to study are known to vary systematically with firm size, although not always in a proportional

or even monotonic manner. To correct for this, we conduct most of our analysis using non-

parametric regression to net out the effect of firm size from comparisons between countries. This

is achieved as follows.

Suppose that we wish to compare variable Y across the five countries. We regress Y non-

parametrically on firm size separately for each country. As measure of firm size we use the

logarithm of total employment reported by occupational category because it is less likely to be

missing and is less biased by respondents‘ reluctance to talk about their respect of labor laws,

especially in China. We then plot on a graph the non-parametric regression lines for the five

study countries, with the log of employment on the x axis and the predicted of Y is on the y axis.

If the regression lines for two countries are aligned – i.e. the predicted values of Y, controlling

for firm size, overlap between two countries over their common support – then we conclude that

Y does not differ between them, once we control for firm size. This simple approach is used

throughout this report.

One important caveat needs to be borne in mind when interpreting these graphs: the non-

parametric regression lines only represent sample averages and hence are only estimates of the

true values. Estimates have a sampling distribution, which means that small differences between

0

.1

.2

.3

.4

5 10 15 20 Log sales

China Vietnam

Ethiopia Tanzania Zambia

Density

18

them are typically not statistically significant. We therefore focus on large differences only and

tend to disregard small differences that are artifacts of small sample size.

We begin by providing a general picture of the characteristics of the firms in our samples. We

then turn to their business practices before turning to innovation and social capital.

Respondent characteristics

Survey interviews were conducted through face to face interviews with a representative of the

firm. Except in China, 70 to 80% of interviews were responded directly by the firm‘s top

manager, who in many cases is also the owner. In China this percentage drops to 59%, probably

reflecting the larger size of surveyed firms. Whenever the respondent is not the top manager, he

or she tends to be the deputy manager or a branch manager. Most respondents are male.

Firm characteristics

The overwhelming majority of surveyed firms are stand-alone operations, with only 8% of

surveyed firms being part of a larger enterprise and no strong difference across country samples.

In the Chinese and Vietnamese samples the median firm is six years old. Similar values are

observed for the African sample, with slightly younger firms in Ethiopia (median is 4 years old)

and slightly older firms in Zambia (median is 10 years old).

The legal status of sample firms is summarized in Table 3. Only a handful of firms in our

samples are government-owned, which is to be expected since we focus on relatively small

firms. In Vietnam a small proportion are cooperatives. The rest fall into two broad categories:

with or without limited liability. In China, Vietnam, and Ethiopia the majority of firms benefit

from a limited liability status. Very few of them, however, are publicly traded on a stock

exchange. In contrast, most sample firms in Tanzania and Zambia are either held in sole

proprietorship or in partnership and do not, as a result, enjoy the protection of limited liability.

The proportion of limited liability firms is particularly small in Tanzania, but perhaps this is

because of the small size of sample firms in that country.

Table 3: Legal status

Country sample


Sole proprietorship 105 95 66 187 162

35.5% 31.7% 26.4% 72.2% 63.8%

Partnership 7 0 6 64 34

2.4% 0.0% 2.4% 24.7% 13.4%

Limited liability company 173 187 178 7 48

58.4% 62.3% 71.2% 2.7% 18.9%

Publicly traded company 10 5 0 1 8

19

3.4% 1.7% 0.0% 0.4% 3.1%

Cooperative 0 12 0 0 0

0.0% 4.0% 0.0% 0.0% 0.0%

Government-owned 1 1 0 0 2

0.3% 0.3% 0.0% 0.0% 0.8%

Total observations 296 300 250 259 254


We suspect that not having a limited liability status exposes entrepreneurs to business risk on

their personal assets while at the same time putting firm assets at the mercy of individual

financial difficulties. We therefore expect firms with limited liability status to be more attractive

to entrepreneurs while facilitating external finance. We wish to investigate whether differences

across countries in the proportion of firms with limited liability status is due to differences in

institutional setup or whether it is simply driven by differences in firm size.

To this effect, we follow the procedure outlined in Section 2. For each country separately, we

regress non-parametrically the proportion of firms that are not partnerships or sole

proprietorships on firm size. The results of the five non-parametric regressions are then

combined into a single graph. Results are shown in Figure 3. We see that, in all countries, limited

liability status is strongly associated with size. Controlling for size easily explains differences in

likelihood of limited liability status between Ethiopia2 and Zambia on the one hand, and Vietnam

on the other. Even controlling for size, Tanzania has a lower proportion of firms enjoying a

limited liability status, suggesting something (e.g. cost, process) may hinder the acquisition of

such status. The Figure also shows that, if anything, Chinese firms are less likely to have a

limited liability status than their size would suggest. Although this is generally believed to be an

impediment to private sector development, it has not made it impossible for Chinese

manufacturing to enjoy high growth rates.

2 The declining portion of the Ethiopia curve is an artefact of small sample size and is not statistically significantly

different from horizontal.

20

Figure 3: Limited liability and firm size

In terms of ownership, most sampled firms are privately and domestically owned. In the Chinese

sample, 7% of the surveyed firms are listed as fully foreign-owned and another 6% as having

partial foreign ownership. For most of foreign-owned firms, however, the nationality of the

owner is listed as mainland Chinese (57%) or Taiwan Chinese (32%). In the other surveyed

countries, the proportion of foreign ownership is negligible.

The ownership structure of surveyed firms is fairly concentrated, as could be expected given

their relatively small size. In the five study countries except Ethiopia, the majority of surveyed

firms have a single owner. If we add multiple owner firms in which one owner has at least a 50%

share of the business, we end up with 80 to 90% of surveyed firms. Ethiopia stands out as an

exception, with only a third of the surveyed firms with a single or majority owner. Furthermore,

in Ethiopia multiple owners are several and rarely from the same family.

We now focus on the main owner of the firm, that is, the person owning the largest share of the

firm. In the Chinese sample 9% of the time this person is a woman. In the other countries, the

percentages are higher: 28% (Vietnam), 25% (Ethiopia), 11% (Tanzania) and 14% (Zambia).

Owners in Vietnam and China are 44 years old on average. Those in the study African countries

are slightly younger: 37 on average in Ethiopia, 39 in Tanzania, and 42 in Zambia.

There are strong differences in the owner‘s education level across the sample countries. In Figure

4 we present the cumulative distribution of the education level of the firm‘s main owner. Each

% firms with limited liability

0

.2

.4

.6

.8

1

0 2 4 6 8 Log(regular workers)

China Vietnam

Ethiopia Tanzania

Zambia

21

line represents a country. Each line starts at 0% and ends at 100%. The Figure is to be read as

follows. Take, for instance, Tanzania. We see that 70% of the owners have at most 6 years of

education, that is, have completed primary or less. This compares to Vietnam where less than 5%

of owners have stopped their education at the end of primary. Hence the lower the line is, the

more educated owners are. With this understanding, we see that survey firms in Vietnam and

China have the highest owner education level – nearly 90% of owners have more than partial

secondary education. Tanzania has by far the lowest: 80% of owners have at most some

secondary education. Ethiopia and Zambia occupy an intermediate position, with 50% of owners

having education levels similar to those of their Chinese counterparts.

Figure 4: Years of schooling of owner

Many entrepreneurs come from an enterprising family. Most owners have parents who were self-

employed, either in farming (70% in China, 32% to 45% in the other countries) or in business

(23% in China, 33% to 51% in the other countries). Few have a father who was working for a

wage in government or the private sector: 8% in the China sample, 17 to 24% in the other

countries. Entrepreneurs also have significant personal experience in the firm – or in other

businesses. The median owner has been running the firm since its inception. Over the entire

sample, 12% of owners own another business.

Firm owners also have varied life experiences. This is particularly true in the Africa sample

where, in each of the three countries, the majority of firm owners have at some point resided in

another part of the country. In Tanzania this is true for 87% of the sample. In contrast, only 30%

of Vietnamese owners and 9% of Chinese owners have ever resided in another part of their

country. A minority of owners have resided abroad at some point in their life: 3 to 4% in China,

Vietnam and Ethiopia, but 9% in Tanzania and 17% in Zambia. The proportion of foreign

22

nationals is however very small among firm owners in our sample – less than 3% in the sample

as a whole.

Origins and development of entrepreneurial capabilities

Given their relatively small size, the studied firms tend to be entrepreneurial in nature, and hence

have a large involvement of the current owner in the creation of the firm. In all five countries

more than 86% of main owners contributed ideas for the creation of the firm, more than 83%

participated in financing the creation of the firm, and more than 79% contributed technical

expertise at firm creation. There is little difference across countries, except that reported owner

involvement was slightly less in Zambia and, to a lesser extent, in Tanzania.

Where the country samples differ is in the involvement of people other than the owner in helping

set the firm up. Table 4 present respondents‘ answers about who contributed to the creation of

the firm. We report the proportion of respondents who listed each distinct source.

Table 4: Who contribute to the creation of the firm?


A. Contributing ideas

Family members and relatives 58.2% 75.7% 29.6% 39.7% 33.8%

Business friends and acquaintances 56.5% 9.7% 30.4% 15.3% 17.7%

Clients 25.0% 4.3% 1.6% 6.9% 1.2%

Employees 29.5% 2.0% 1.2% 1.1% 4.6%

Shareholder N.A. 0.7% 14.0% N.A. N.A.

Expert/consultant 21.6% 3.0% 2.8% 2.7% 3.1%

Equipment suppliers 12.0% 2.0% 0.4% 0.0% 1.2%

School teacher/professor 4.1% 0.0% 0.0% 1.9% 1.5%

Other 1.0% 0.0% 4.4% 5.7% 6.2%

No one else 4.8% 17.3% 32.4% 37.0% 38.8%

Number of observations 292 300 250 262 260

B. Contributed technical expertise



Employees 41.0% 3.3% 11.2% 2.3% 6.6%

Expert/consultant 21.4% 3.7% 1.2% 4.6% 9.3%

Clients 12.8% 4.7% 0.4% 7.6% 0.4%

School teacher/professor 3.1% 0.3% 1.2% 16.0% 4.2%


Other 1.0% 1.7% 9.6% 4.6% 4.7%

No one else 9.7% 18.3% 47.6% 33.2% 42.9%


23

C. Participated in financing



A non-bank financial institution 2.5% 1.0% 16.4% 0.8% 0.8%

A bank 4.3% 1.0% 2.0% 0.8% 0.8%


Other 0.4% 1.7% 5.6% 4.2% 2.3%

No one else 32.6% 30.3% 43.6% 68.7% 64.9%


There are large differences between the countries, and a clear pattern between the Asia and

Africa firms. In the two Asian countries, a majority of respondents list family members and

relatives as contributors in terms of ideas, technical expertise, and financing. The proportions

are much smaller in Africa, especially for technical expertise and financing. Does this mean that

African entrepreneurs find the ideas, expertise, and funding they need elsewhere? The answer is

a resounding no. Respondents in the three African study countries are much more likely than

those in the two Asian countries to report that no one else than themselves contributed to the

creation of the firm.

Furthermore, respondents in Asian firms are also more likely to list other sources of help. This is

particularly true among Chinese respondents who are much more likely to list business

acquaintances, experts and consultants, clients, employees, and equipment suppliers as sources

of assistance at start-up.

To investigate whether the kind of assistance entrepreneurs received at the start of business has

an effect on firm performance, we regress firm size at start on country fixed effects and dummy

variables taking value one if (1) the firm received start-up advise or finance from family and

relatives; (2) received start-up advice or finance from business acquaintances, clients, or

suppliers; (3) received advice from experts or teachers; and (4) received financing from banks or

other financial institutions. The results are shown in the first column of results in Table 5. We

see that receiving help from family and relatives is associated with a smaller start-up size while

receiving help from experts or finance from financial institutions is associated with a larger start-

up size.

It would be perilous to interpret this relationship as causal – entrepreneurs who seek advice from

experts and who manage to securing start-up funding from financial institutions may simply be

better than the average. The relationship nevertheless suggests that obtaining advice and finance

at start-up may be an important channel by which good entrepreneurship comes to fruition.

To investigate this idea further, we regress (the logarithm of the) current firm size and the firm‘s

employment growth rate since start-up on the assistance it received at start-up. These two

regressions are reported in the second and third columns of results in Table 5. We see that

24

assistance at start-up remains a strong predictor of future firm performance. In particular we note

that firms that received expert advice at start-up on average grow significantly faster than other

firms by 3 percentage points per year. Again these relationships should not necessarily be

interpreted as causal – help at start-up is likely to be correlated with entrepreneur acumen – but

they suggest that advice at start-up may be an important channel of firm performance. We also

note that the country where expert advice is reported by the largest proportion of respondents,

China, is also the country with the highest manufacturing growth rate.

Table 5: Firm size and assistance received at start of business

Firm size at start Firm size now Firm's growth rate

Received help from: Coef. t-value Coef. t-value Coef. t-value

Family and relatives -0.213 -3.08 -0.127 -2.56 0.01 1.04

Businesses, clients,

suppliers 0.065 0.90 0.125 2.42 0.01 1.12

Experts and teachers 0.165 1.73 0.283 4.26 0.03 2.05

Financial institutions 0.292 2.74 0.208 2.96 0.02 1.09

Intercept 2.052 34.77 2.667 62.58 0.07 7.24

Coefficient estimates from country fixed effect regression

Regulatory environment

Next we turn to the regulatory environment. We present in Table 6 summary information about

business registration and licenses. In all countries except Tanzania the majority of sample firms

have some form of business registration. In China and Vietnam, the majority of sample firms are

also registered for value added tax, implying that their manufacturing output is part of the

country‘s tax base. This is not true in Ethiopia and Tanzania where only a small proportion of

surveyed firms report being registered for VAT. In Zambia the proportion is larger but still less

than half of the firms.

The number of licenses and permits required to start or operate a business is often seen as a

disincentive to entrepreneurship, not only because of the associated fees, but also because it takes

up some of the entrepreneur‘s time, and this time is particularly precious in small and medium

size firms because there is less delegation of these tasks to clerical staff. We see in Table 6 that

the number of licenses is much higher in China than in the other four surveyed countries.

Tanzania is the sample country with the smallest number of licenses but also with the least

dynamic manufacturing sector and the smallest average firm size.

Respondents were asked whether in the last year they had to pay a penalty for operating without

registration or license. Except in Tanzania, the number of respondents who respond positively to

the question is very small, even in countries like Ethiopia and Zambia where the proportion of

unregistered firms is well below 100%. Even in Tanzania where the proportion of positive

25

responses nears 10%, the proportion of penalized firms is well below the proportion of

unregistered businesses among sample firms. This suggest that the enforcement of registration

and license requirements is present in the three African countries in our sample, but not so strong

as to prevent informal business operation.

Table 6: Regulatory environment

China Vietnam Ethiopia Tanzania

Zambia

N obs Mean N obs Mean N obs Mean N obs Mean N obs Mean

Business

registration 301 99.7% 300 93.3% 250 71.2% 262 40.3% 259 56.9%

VAT

registration 221 91.0% 300 82.0% 250 5.2% 262 6.5% 260 30.4%

# licences

before start-up 205 4.1 300 1.4 250 1.2 261 0.6 237 1.2

# licences renew

last year 188 3.0 300 0.3 248 0.8 261 0.3 252 1.1

Penalty for non-

registration

or lack of

license 256 1.2% 300 2.0% 249 2.4% 259 9.7% 259 3.1%

Incidence of

corruption N.A. 300 26.0% 248 19.8% 260 13.1% 246 13.0%

Finally firms were asked whether firms have to give presents to government officials to get

things done. This question was not asked in China because it was deemed too sensitive (which,

by itself, suggests that incidence is high). Responses tabulated in Table 6 indicate that, if

anything, corruption is more prevalent in Vietnam than in the African sample. Of course this

may be because high growth implies a higher opportunity cost of ‗not getting things done‘, and

thus a higher willingness of firms to pay to speed up the administrative process. This is indeed

what is suggested by regression analysis: when we regress the incidence of corruption on a

firm‘s growth rate, controlling for country fixed effects, firm size, and a business registration

dummy, we find that faster growing firms are more likely to report having to pay government

officials to get things done. Note that this result is not driven by China since incidence of

corruption was not reported by Chinese respondents.

The country comparison in Table 6 may be misleading, however, if the regulatory environment

of firms varies systematically with firm size, since we know that the countries in the study have

different average firm size. To investigate this possibility, we regress business registration non-

parametrically on firm size for each country separately. Results, presented in Figure 5, show that

firms above a certain size tend to be registered. There is, however, considerable heterogeneity in

26

registration rates across countries for smaller firms, with a systematically lower registration

probability in Africa than in Asia.

Figure 5: Registration and firm size

We also have information on electricity provision and usage. Information is summarized in Table

7. Most sampled firms use electricity for production, although this proportion drops to two thirds

of surveyed firms in Tanzania. Perhaps surprisingly, the proportion of firms that have an account

with an electricity provider is sizably less than the proportion of firms using electricity for

production. What accounts for the difference is unclear, but it may in part be due to illegal

connection to the grid – or to the provision of free electricity to some users.

Except in China, the majority of studied firms experience power outages on a regular basis. The

incidence of outages appears particularly high in Vietnam, Ethiopia, and Tanzania. When

outages occur, they tend to be more frequent in the African countries in our sample. The severity

of each incident, measured by the average duration of the outage, seems fairly comparable across

the five countries, however. If some users do indeed secure electrical power illegally, it may

explain why there are outages in the first place.

.2

.4

.6

.8

1

% Registered Firms


China Vietnam

Ethiopia Tanzania

Zambia

27

Table 7: Electricity usage and reliability

China Vietnam Ethiopia Tanzania Zambia Electricity:

N obs Mean N obs Mean N obs Mean N obs Mean N obs Mean

Used for production

286 97.9% 300 92.0% 250 87.6% 262 65.3% 263 80.2%

Account with provider

272 80.1% 300 80.0% 250 73.6% 261 40.2% 262 28.2%

Experience outages

262 35.9% 276 89.5% 217 99.1% 169 97.0% 208 70.2%

If yes:

# days per month

89 2.1 247 2.8 215 7.1 164 6.9 143 4.6

average duration

93 6.9 246 7.5 215 4.2 162 6.2 143 3.0

Competition

Respondent firms were asked to subjectively evaluate whether the market for their products is

competitive. We see few differences across countries, with around half of the firms responding

that their market is moderately competitive and the rest that it is very competitive. Zambia is the

only outlier, with 71% of respondent qualifying their market as very competitive.

To measure geographical clustering, firms were asked how many businesses operate in the same

sector within fifteen minutes walk. The average response is smallest in China – less than 6 – and

highest in Ethiopia – more than 28. The other three countries occupy an intermediate position

with 10-15 firms in the same sector within a 15 minutes walking distance radius. The reason why

geographical clustering may appear strongest in Ethiopia and lowest in China may be because

firms are larger in China and thus cannot be physically located as close together.

Respondents were asked to evaluate the number of their competitors that are enterprises with 10

workers or more. We find few differences across countries. If anything, Chinese respondents list

a smaller number of competitors. The same is true for Tanzanian respondents, but probably for a

different reason, namely, the dearth of medium to large manufacturers in the country.

The countries do, however, differ markedly in terms of competition from foreign firms.

Respondents were asked to evaluate the number of their foreign competitors. Their answers,

which are depicted in Figure 6 in the form of cumulative distribution, indicate that Chinese firms

face the lowest level of foreign competition while African firms face the highest. The proportion

of Chinese firms that state facing no foreign competition is 94%. For Ethiopia and Zambia, the

corresponding figures are 60% and 53% respectively. In Tanzania, only 18% of respondent

28

declare facing no competition from imports. Vietnam occupies an intermediate position between

China and Zambia.

Figure 6: Number of foreign competitors

The contrast between the three African countries may be partly attributed to geography: while

Dar es Salam, where the Tanzania survey took place, is a port, both Addis Ababa and Lusaka are

located within the continent in land-locked countries. The contrast between Africa and Asia,

however, cannot be given the same explanation. Vietnam is a coastal country with its two major

cities close to the sea. Similarly much of China‘s manufacturing growth is located on the Eastern

seaboard. Yet in both cases competition from imports is perceived to be low by respondent firms.

Why this is the case is unclear. It could be because the two Asian countries in our sample have

trade protection, either direct or indirect (e.g., through exchange rate policy).

Alternatively, it is possible that Chinese (and Vietnamese) firms are more competitive so that

foreign firms do not seek to commercialize their products in the country. There is some evidence

of this in the data in the sense that most African respondents list China as the major source of

competing imports. The proportions of respondents who cite China as source of foreign

competition is large: 89% in Vietnam, 93% in Ethiopia, 84% in Tanzania. The only exception is

Zambia where China is mentioned ‗only‘ by 56% of respondents. South Africa is a major source

of competing imports in Zambia, mentioned by 57% of respondents. There is a direct rail link

between the two countries, and many manufacturing imports into Zambia come through South

African ports. South Africa is also mentioned by 6% of Tanzanian respondents. In contrast, no

African country is mentioned as source of competing import by Chinese and Vietnamese

respondents – who only mention the US, Europe, or other East Asian countries.

Respondents who face foreign competition were asked to compare these imports to their own

products in terms of price, quality, and adaptation to local taste. There were too few responses in

29

China so we ignore them here. In the other three countries, 47-59% of respondents find

competitive imports more expensive than their products, but only 17-26% of them find imports

to be of better quality. In Vietnam, Tanzania and Zambia around 44-47% of respondents find

imported products to be better adapted to local tastes and fashion. This proportion rises to 85% in

Ethiopia. This suggests that domestic producers seek to compete with imports primarily on

quality, less on price and design. This in turn suggest that the manufacturing imports they

compete with are low quality, low cost mass produced items that suit well the limited budget of

local consumers.

Social capital

Membership in associations differs markedly between China and the other four studied countries.

In China, 55% of respondent do not belong to any business association. This percentage rises

over 80% in the other four countries, with little difference between them. Part of the variation in

the propensity to join a business association is due to differences in firm size. In Figure 7 we

show the result of non-parametrically regressing absence of membership on firm size. Except in

China where there is no systematic relationship between membership and size, in all the other

countries we observe a positive relation between the two. We also note that, once we control for

size, there is a difference between Ethiopia and Vietnam on the one hand, and the other three

countries in the other: controlling for size, firms in Ethiopia and Vietnam are less likely to

belong to a business association. Why this is the case is unclear. One possibility, which we

cannot test with the data at hand, is political reasons, e.g., (membership in) business associations

could be discouraged because of the fear these associations may foster opposition to the regime.

30

Figure 7: Association membership and firm size

If associations cannot serve their purpose, social networks offer a partial alternative.

Respondents were first asked whether they have friends working as bank officials, political

figures, or in government. While nearly all respondents in the other four countries answered

these questions, Chinese respondents were a lot more reluctant and a third of them refused to

respond. Chinese data on social networks is therefore to be treated with caution as it is likely to

be affected by selection bias – i.e., those most reluctant to respond to the questions are those

most keen to hide their privileged contacts in banks or government. With this caveat in mind,

answers show that, except in China, most respondents have friends or relatives in banks,

government, or politics. Proportions are highest in the three African countries, with two thirds to

four fifths of respondents answering yes to at least one of the three questions. For comparison,

the proportion is 52% in Vietnam. In China, two thirds of those who answered the questions

answered no to all three. If we were to assume that all those who refused to answer would have

answered ‗yes‘, this would imply that at most 54% of respondents know someone in banks,

government, or politics – a figure not too different from Vietnam. We can therefore conclude

with confidence that owners and managers of African small and medium firms in our study are

significantly more likely that their East Asian counterparts to know someone in banks,

government, or politics. What interpretation we should give to this finding is not immediately

clear, however.

0

.2

.4

.6

.8

% registered firms


China Vietnam


31

One possibility is that those born with connections in banks or government can hope to benefit

from them and are therefore more likely to start a business. This is the ‗crony capitalism‘

hypothesis. Another possibility is that entrepreneurs may not have started with connections but,

in order to succeed in business, they have to mingle with financial and political elites so as to

secure undue advantage through the exchange of favors. This is the ‗corruption‘ hypothesis. Yet

another possibility is that entrepreneurs belong to the burgeoning middle class and, as such,

mingle socially with financial and political elites, without necessarily seeking to obtain undue

advantage from these acquaintances. This is the ‗social class‘ hypothesis. Can we disentangle

these hypotheses with the data at hand – ignoring China since the data is potentially biased?

Not without great difficulty. First of all, we do not know whether respondents already knew

people in banks and government before starting their business.3 This makes it difficult to test the

crony capitalism hypothesis versus the corruption hypothesis. We can, however, examine

whether firms that have been in existence longer tend to have more contacts: if contacts are

accumulated as a consequence of business, older firms should have more contacts. We only find

weak evidence of this: controlling for country fixed effects, the relationship between firm age

and the likelihood of contacts with banks and government is positive but small in magnitude and

not statistically significant.

We can also indirectly test the crony capitalism and corruption hypotheses against the social

class hypothesis by observing whether better connected firms grow faster. Presumably, if

contacts are used to obtain finance and other advantages, connected firms should have a higher

growth rate. This is not what we find: there is no significant correlation between firm growth and

contacts. There is also no correlation between contacts and firm size.

From this exploration we conclude that we cannot tell from the data why African entrepreneurs

have more contacts with banks, government, and politicians – i.e., whether it is a consequence of

entrepreneurship (social class hypothesis), a cause of entrepreneurship (crony capitalism

hypothesis), or a channel through which entrepreneurial success is achieved (corruption

hypothesis).

Respondents were also asked about the contacts they have with other businesses. It is often

believed that social contacts help disseminate business-relevant information and practices. If this

is true, better connected firms should be at an advantage and thus should grow faster and be more

profitable. We therefore expect a positive correlation between business contacts and firm

performance.

From the survey data we construct three summary variables: the number of business contacts

respondents declare having; the number of contacts from whom respondents have received

3 Asking them would not suffice as their answers would likely be contaminated by recall bias: they would tend to

omit from their answer those contacts they had but did not subsequently help them, for instance; or they would wrongly think they already knew someone they only met after start-up.

32

assistance in the past; and the number of contacts to whom assistance has been provided by the

respondent. Many different forms of assistance are considered: identifying new markets or

sources of raw materials; securing external finance; recruiting workers; exchanging technological

information; obtaining and repairing machinery; etc.

Answers are tabulated in Table 8. We first note that China stands out as having a much smaller

number of declared business links than in all the other countries. This finding should be regarded

with some suspicion, however. First of all, Chinese respondents were more reluctant to answer

the questions – the non-response rate is higher than in all the other countries. Secondly, those

who responded were also much more likely to answer ‗no‘ to all questions. We therefore regard

the Chinese data on business contacts as subject to serious response bias. The data from the other

four countries is, in contrast, very consistent. Firms typically have many contacts with other

firms, and these contacts are used for various business-related purposes – either to receive help

from others, or to provide help. If anything, the average firm in the sample receives slightly more

help than it gives, but the difference is small.

Table 8: Business Contacts


Nobs Mean Nobs Mean Nobs Mean Nobs Mean Nobs Mean

Number of business contacts

245 1.9 300 12.9 250 9.2 257 11.2 226 7.6

of which:

Received help

174 0.7 282 5.1 250 2.7 260 3.1 258 3.4

Provided help

175 0.7 290 4.6 250 3.0 260 3.2 258 3.2

What kind of help are respondent firms most likely to give and receive? We present in Table 9

the proportion of respondents who report receiving or giving assistance of different kinds. The

reader will immediately notice that the data for China and Vietnam appear doubtful. In China,

the proportion of missing information is large, and those answers that were given appear low and

correlated – as if respondents were giving the same answer to all questions. This is also the

pattern we observe in Vietnam, where answers are nearly all the same. We therefore focus our

analysis on the three African countries in the study, where there is a bit more variety. We note

that business contacts are used to identify sources of raw materials, but the emphasis is more on

foreign sourcing in Tanzania – where the survey took place in the port city of Dar-Es-Salam –

than in Ethiopia and Zambia, which are landlocked countries. On the receiving side, other

answers are fairly similar, except perhaps to note that business contacts are slightly less likely to

be used for securing external finance. The giving side largely mirrors the receiving side, with

more emphasis on sourcing raw materials and less on external finance.

33

Table 9: Giving to and receiving assistance from business contacts



A. Assistance received

Identify local sources

of inputs 170 7% 261 80% 250 57% 257 26% 257 53%

Identify foreign

sources of inputs 171 8% 183 72% 250 8% 259 49% 257 32%

Identify new markets 173 9% 239 78% 250 48% 259 40% 254 49%

Secure external

finance 173 9% 226 75% 249 29% 259 46% 255 40%

Recruit qualified

workers 172 9% 243 80% 249 31% 258 48% 254 39%

Provide technological

information 173 10% 245 77% 249 38% 260 53% 256 44%

Obtain second-hand

equipment 174 10% 238 79% 249 25% 259 47% 255 44%

Repair machinery 173 10% 244 73% 249 35% N.A. 256 41%

B. Assistance given

Identify local sources

of inputs 257 32% 250 77% 250 60% 258 65% 252 57%

Identify foreign

sources of inputs 254 19% 178 74% 250 9% 257 16% 254 25%

Identify new markets 253 29% 234 76% 250 53% 258 47% 256 48%

Secure external

finance 250 18% 206 75% 250 31% 258 27% 256 30%

Recruit qualified

workers 251 33% 228 76% 250 36% 260 39% 256 36%

Provide technological

information 255 35% 223 76% 250 42% 259 47% 257 45%

Obtain second-hand

equipment 246 29% 221 77% 249 27% 260 47% 256 38%

Repair machinery 251 32% 225 77% 250 45% 259 37% 257 40%

The above analysis does not tell us whether social networks play a causal role, i.e., help firms

overcome barriers to trade and information asymmetries, or whether contact with others is

simply one possible channel through which the relevant information, good, or service is

obtained. If social networks either play a causal or channel role, we should observe that firms

with more contacts perform better and thus achieve a larger size. To investigate this possibility,

we regress the number of business contacts non-parametrically on firm size. Results are shown in

Figure 8. China is omitted because the data contains too many missing observations. Although

34

there is some evidence of a positive relationship between the number of contacts and firm size,

this relationship is weak and limited to some countries only. This could be because business

contacts do not, ultimately, play an important role in firm performance. Alternatively, it could be

that business contacts matter, but only for those firms that do not have alternative access to

information. If large enough firms get the necessary information in other ways – e.g., by hiring

consultants or specialized employees – then they have less need for business contacts. This could

potentially explain the non-monotonic relationships apparent in Figure 8 for Vietnam and

Ethiopia. More work is needed to ascertain the exact role played by social networks in business

performance. We anticipate that the social networks experiment will provide important new

insights on this issue in due course.

Figure 8: Number of contacts and firm size

We repeat the same analysis for receiving help and giving help. The results, not shown here to

save space, are not particularly interesting. There is no relationship between receiving help and

firm size except in Vietnam where the correlation is positive. In the three African countries, the

relationship is either non-monotonic or declining. For giving help, there is a bit more evidence

that large firms are more involved in assisting other entrepreneurs. This is not surprising, given

that entrepreneurs who manage to run large enterprises probably have more talent and

information, and consequently are in a better position to assist others.

5

10

15

20 Number of business contacts



Zambia

35

To summarize, the dominant impression that comes out of social network questions is that

entrepreneurs help each other out by passing relevant information to each other. While there is

some mild evidence that entrepreneurs who manage larger firms provide more information to

others, we find no evidence of a strong, unambiguous relationship between social business

networks and firm performance, as measured by size. Why entrepreneurs help each other if help

does not improve performance is unclear. One possibility is that usage of social networks to

access information is highly endogenous: some entrepreneurs managed to access the information

without recourse to social networks; these tend to be more talented entrepreneurs. Less talented

entrepreneurs seek – and obtain – information through their networks, so that for them social

networks are useful. Less talented entrepreneurs that have no network find it difficult to perform

and survive, and may not even be in the dataset because they are too small, had to close down, or

never started the business in the first place. While the above explanation can account for the

observed pattern, the data does not provide any serious support in its favor.

Innovation

The questionnaire elicited detailed information on the innovation activities of surveyed firms.

This information is summarized in Table 10. The first three rows refer to the introduction of new

products, new production process, or new product delivery system since 2008. The next four

rows cover less frequent innovation events but cover the lifetime of the firm. The bottom row

combines all forms of innovation into a single index.

The most striking finding that emerges from the Table is how similar four of the countries are.

Tanzania is the outlier, with a much lower innovation rate in the first three rows and across the

board. Tanzania is also the study country with the smallest average firm size, the most informal

firms, and the least educated entrepreneurs. Apart from Tanzania, the other countries are

amazingly similar. In particular there is no difference between China and the other three

countries in terms of product introduction and changes in delivery system. If anything, the

innovation index at the bottom of the Table is slightly lower for China than for Vietnam,

Ethiopia, and Zambia. The only form of innovation where China shows more activity is in terms

of production process, where firms are more likely to report having made changes in the recent

past. The difference with the other three countries, however, is not large.

36

Table 10: Innovation


Introduced new products 44% 43% 46% 24% 44%

Changed production process 42% 34% 36% 15% 28%

Changed delivery system 20% 19% 22% 5% 25%

Introduced new product in country 10% 2% 9% 9% 21%

Exported new product from country 2% 2% 1% 3% 5%

Imported new raw material in country 1% 1% 1% 0% 3%

Introduced new machinery in country 7% 1% 6% 1% 5%

Any of the above 57% 58% 60% 34% 61%

These results are surprising because, given China‘s rapid growth in manufacturing output, one

would have expected the innovation rate to be higher. This is not the case. Chinese firms in the

sample are also larger, and one would expect larger firms to innovate more.

To investigate this idea further, we regress non-parametrically the innovation rate on firm size.

Results are displayed in Figure 9. We find a positive relationship between innovation and firm

size, as expected, although this relationship is stronger in some countries than in others. The

most striking result from this table is that, controlling for size, Chinese firms are seen to innovate

less, not more, than firms in the other four countries – including Tanzania. This indicates that,

contrary to common belief, African firms are no less open to innovation than their Asian

counterparts – including China.

37

Figure 9: Innovation and firm size

A possible explanation for the above puzzling results is that we have measured a kind of

innovation that is not associated with firm growth. To test this idea, we regressed firm growth

(measured as the annual rate of growth in total firm employment since start-up) on innovation

and we found a positive correlation between the two, especially with the introduction of new

products and new customer delivery systems. This confirms that, even in our sample, innovation

is associated with better firm performance. Yet average firm growth in China is higher: 14.6%

annual growth in employment growth since start-up, compared to 7.7% in Vietnam, 4.9% in

Ethiopia, 7.4 in Tanzania, and 10.6% in Zambia. This may be because the effect of innovation on

firm growth is higher in China.

To test this hypothesis, we regress firm growth on innovation and an interaction term between

innovation and a China dummy, controlling for country fixed effects. If the effect of innovation

on firm growth is stronger in China, the coefficient of the interaction term should be positive and

significant. Results, not shown here to save space, do not suggest that the association between

innovation and firm growth is stronger in China than in the other study countries.

Another issue we investigate is whether innovation in the African study countries is of the

‗wrong kind‘ and comes from the ‗wrong source‘. To this effect, we examine what prompted

innovation and what was the source of advice on innovation. In Table 11 we tabulate answers

0

.2

.4

.6

.8

1

Innovation index


China Vietnam

Ethiopia Tanzania

Zambia

38

that respondents gave when asked what prompted the firm to introduce new products. Results

show that in all countries the most common answer is that the firm responded to customer

demand. The next most common answer is that the firm noticed a gap in the market. One could

argue that this answer indicates more initiative on the part of the firm – it did not wait for

customers to suggest new products. This answer was given more often by Ethiopian and Chinese

respondents, but the difference could simply be due to differences in translation. In contrast, we

find rather more difference between China and the other four countries when it comes to

imitation: in China, imitating other producers, whether local or foreign, is very seldom given as a

reason for introducing new products; in the other four countries, it is an answer given by a small

but not negligible minority of firms. This suggests that imitation is a more powerful driver of

product innovation in these four countries than in China.

Table 11: What prompted the firm to introduce new products?



Noticed a gap

in market 124 41.9% 129 20.9% 114 50.0% 62 32.3% 112 31.3%

Responded to

customer demand 126 55.6% 129 72.9% 114 74.6% 62 66.1% 112 46.4%

Imitated other

local producers 124 0.8% 129 7.0% 114 14.0% 62 9.7% 112 2.7%

Imitated imports 124 0.8% 129 6.2% 114 7.0% 62 3.2% 112 0.9%

Survey participants were asked who helped them find customers for their newly introduced

products. In all countries the dominant answer is no one, i.e., the firm relied on its own

management and employees. Some 9% to 17% of respondent mentioned friends and relatives as

a source of help for finding new customers, except in Vietnam where 40% of respondents made

that answer. Other sources of help are in general unimportant, with some – possibly random –

variation across countries. There is no evidence that, for the firms in this study, government

agencies, research institutions, and foreign joint ventures play a bigger role in China than in the

other study countries in terms of finding new customers.

Things are different with respect to technical expertise. In Table 12 we present respondents‘

answers to the question of who provided them with the technical expertise needed to develop

new products. The most common answers are ‗no one‘ and ‗friends and relatives‘. There is

sizeable variable across countries, but nothing that appears systematic. Customers tend to be

cited more often as source of technical expertise in China and Vietnam than in the three African

countries covered by the study. Even sharper differences emerge regarding experts and

consultants, which are cited by 27.7% of Chinese respondents but only by a few respondents in

the other countries. A similar pattern is observed for research institutions, which are cited by

10.8% of Chinese respondents but hardly anyone in the other four countries. Finally, suppliers of

39

equipment and raw materials are cited in China and Vietnam, but not in Africa. In contrast,

competitors are cited by a number of African firms, much less so in Vietnam and China.

The picture that emerges from this Table is that firms in the two Asian study countries, and

particularly those in China, have access to a much wider range of sources of information on the

technical expertise needed to develop new products. Why this is the case is not entirely clear,

however. It could be that wider access is what caused Asian, and especially Chinese, firms to

grow faster. Alternatively, because the manufacturing sector has grown dramatically in China,

more institutions have been created that meet the need of entrepreneurs for technical advice. For

instance, suppliers of equipment and raw material may find that providing advice is a useful

marketing tool and gives them a competitive edge. Smaller markets for raw materials and

equipment in Africa mean less entry and less competition, and this may account for our finding.

More research is needed on this issue.

Table 12: Who provided the technical expertise to develop new products?



No one 130 35.4% 129 54.3% 114 67.5% 62 33.9% 112 31.3%

Friends and relatives 130 13.1% 129 36.4% 114 17.5% 62 22.6% 112 20.5%

Customers 130 32.3% 129 24.0% 114 7.0% 62 19.4% 112 7.1%

Other 124 0.0% 129 9.3% 114 7.0% 62 0.0% 112 25.9%

Experts and consultants 130 27.7% 129 0.8% 114 1.8% 62 1.6% 112 8.9%

Competitors 130 3.8% 129 5.4% 114 7.0% 62 17.7% 112 1.8%

Raw material supplier 130 11.5% 129 9.3% 114 0.0% 62 3.2% 112 0.9%

Research institution 130 10.8% 129 1.6% 114 0.0% 62 6.5% 112 1.8%

Equipment supplier 130 6.2% 129 10.1% 114 0.9% 62 0.0% 112 0.9%

Domestic joint venture 130 6.9% 129 0.0% 114 0.0% 62 9.7% 112 0.0%

Government agencies 130 0.0% 129 2.3% 114 6.1% 62 1.6% 112 0.9%

Foreign joint venture 130 3.1% 129 1.6% 114 0.0% 62 1.6% 112 0.0%

In Table 13 we report similar figures, but this time for changes to the production process or

delivery system. The pattern is similar to that found in Table 12. Many respondents state that

they received help from no one. Customers and friends and relatives are commonly mentioned

sources of information. That customers are listed is not surprising given that the question partly

refers to changes in the delivery system. Competitors are again cited heavily in Tanzania. If

firms are competing through innovation, this is not what we would expect. But technological

upgrading of other firms in the neighborhood may generate agglomeration externalities if it helps

attract customers. Such an externality is most likely to arise if similar firms cluster around each

other to form a magnet for customer, as in a specialized retail district. Experts and consultants

are again cited heavily in China, while suppliers of equipment and, to a lesser extent, raw

40

material tend to be cited more often in Asia. In all countries government agencies and joint

ventures with foreign firms play a minor role, at least in terms of the number of affected firms.

Of course, joint ventures may introduce new techniques of production and marketing that may

then subsequently spread locally, e.g., through friends and relatives, customers, and the like.

Table 13: Who provided the technical expertise to change the production process or deliver

system?



No one 121 29.8% 103 44.7% 89 62.9% 38 34.2% 71 29.6%

Friends and relatives 121 14.0% 103 44.7% 89 12.4% 38 13.2% 71 18.3%

Customers 121 23.1% 103 20.4% 89 10.1% 38 15.8% 71 12.7%

Competitors 121 10.7% 103 5.8% 89 15.7% 38 34.2% 71 4.2%

Experts and consultants 121 24.0% 103 0.0% 89 2.2% 38 13.2% 71 8.5%

Equipment supplier 121 13.2% 103 14.6% 89 2.2% 38 5.3% 71 0.0%

Other 121 0.0% 103 9.7% 89 0.0% 38 0.0% 71 21.1%

Domestic joint venture 121 19.0% 103 0.0% 89 0.0% 38 0.0% 71 0.0%

Raw material supplier 121 6.6% 103 4.9% 89 0.0% 38 0.0% 71 2.8%

Government agencies 121 2.5% 103 0.0% 89 4.5% 38 2.6% 71 1.4%

Research institution 121 5.8% 103 1.0% 89 0.0% 38 2.6% 71 1.4%

Foreign joint venture 121 2.5% 103 1.0% 89 0.0% 38 2.6% 71 0.0%

Next we examine whether there are differences in the way innovation adoption is financed by

firms. Firms that introduced new products, production process, or delivery system were asked

whether these changes required additional finance. The majority of respondents state that they

do, except for China where only a sizeable minority (39-44%) said the changes called for

additional finance. The corresponding numbers for Vietnam, Ethiopia, and Zambia are all above

63%. Tanzania stands in between, with slightly over half of the firms requiring additional

funding, but on a much smaller sample because the proportion of innovation adopters is much

smaller in Tanzania.

Table 14 summarizes firms‘ answers on how changes were financed. Most firms list retained

earnings as source of finance. Friends and relatives are cited twice as often by Asian than

African respondents. Banks are cited by a minority of respondents, but much more frequently in

Asian than in Africa. New capital from existing owners is also cited on average more often by

Asian respondents. In addition, Chinese respondents mention advances from customers and

credit from financial institutions as significant sources of funding, while these sources are largely

omitted by other respondents. Other possible sources are marginal.

41

Table 14: Source of funding for innovation adoption


Retained earnings 55% 87% 88% 47% 79%

Friends and relatives 16% 16% 6% 8% 8%

Bank 19% 16% 4% 11% 1%

New capital from owners 12% 9% 9% 2% 5%

Customers 14% 4% 9% 6% 3%

Financial institution 23% 1% 7% 0% 2%

Raw material supplier 4% 8% 2% 0% 0%

New owners 4% 1% 0% 0% 0%

Domestic joint venture 3% 0% 0% 0% 0%

Equipment supplier 3% 0% 0% 0% 0%

Government agency 1% 0% 0% 0% 0%

Foreign joint venture 0% 1% 0% 0% 0%

Development agency 0% 0% 0% 0% 0%

Other 0% 0% 2% 0% 5%


It is of course difficult to interpret these findings. The firms that responded to the finance

questions are self-selected since they are those that did find the necessary funding to innovate.

This by itself does not tells us anything about firms that did not innovate: we do not know

whether they could have secured the funding but chose not to innovate, or whether they did not

innovate because they did not find the funding. The fact that most innovation adoption is funded

out of retained earnings, however, suggests that firms that do not make sufficient profits may

find it difficult to innovate. If the failure to innovate makes a firm less productive and hence less

profitable, it may not be generating the retained earnings needed to innovate, hence creating a

vicious circle. We also note that Asian – and particularly Chinese firms – seem to have access to

a larger variety of funding sources. This, in turn, may facilitate innovation by firms. Finally, we

should note that aggregate productivity is sometimes best served by letting inefficient firms

disappear and fostering the entry of innovative firms. Given that the surveys deal with existing

firms, answers to innovation questions tell us little on how the innovation environment affects

the selection of entering firms.

Inputs

Ready access to inputs is important for manufacturing firm performance. The survey therefore

includes several questions about firms‘ suppliers. Figure 10 summarizes the number of regular

suppliers of material inputs for the firm against firm size, using the same non-parametric

regression method outlined earlier. As the figure shows, it is primarily differences in firm size,

not differences across countries, that explain variation in the number of suppliers. The only

42

visible difference is between large firms in China and Vietnam – where large Vietnamese firms

appear to have fewer suppliers than Chinese firms of similar size. This difference, however, is

not statistically significant. What is clear is that Chinese and Vietnamese manufacturers do no

combine inputs from a larger number of suppliers than African firms of the same size.

Figure 10: Number of regular suppliers and firm size

Even though the number of regular suppliers does not show markedly different trends across

countries, there are some differences in the source of inputs. In three of the study countries,

China, Vietnam, and Zambia, government agencies represent a negligible source of

manufacturing inputs. But in Ethiopia and Tanzania they account for 5.2% and 10.3% of inputs,

respectively. Of the 27 Tanzanian firms reporting input purchases from government agencies, 15

indicate that all their inputs come from such agencies.

Figure 11 shows the percentage of material inputs purchased from abroad, either via local

importers or directly from suppliers in other countries.4 Imported material inputs are important

for manufacturing in the study countries. The only exception is China, where imports represent a

smaller proportion of all inputs and we see no positive correlation between imports and firm size.

This may be because China is such a large player in world manufacturing. We also note that

African small firms are more likely to rely on imported inputs than Asian firms of equivalent

4 Nothing should be made of the prominent ‘kink’ in the fitted values for Vietnam at large firm sizes: it is not

statistically significant.

0

5

10

15

20

25


China Vietnam

Ethiopia Tanzania

Zambia

Number of regular suppliers

43

size, possibly for similar reasons. Few manufacturers import inputs directly, however: the vast

majority of imported inputs used by sample firms come from local importers. This is particularly

true in Africa. Almost none of the African firms list a foreign country as the country of any of

their three main suppliers.

Figure 11: Foreign inputs and firm size

This raises the issue of the ease with which surveyed firms obtain material inputs. The

questionnaire elicited information on the number of alternative suppliers from which firms could

source ‗similar raw materials‘. Figure 12 shows how responses vary by country and firm size.5

One would expect Chinese, and perhaps Vietnamese, manufacturers to have more viable

alternative suppliers than firms in Africa, given the larger size of the manufacturing sector in

these economics. This does not appear to be the case, however: if anything, firms in Ethiopia and

Tanzania report a larger number of alternative suppliers.

5 Data is missing for 161 of the 303 Chinese respondents.

0

20

40

60 Material inputs from abroad (%)


China Vietnam

Ethiopia Tanzania

Zambia

44

Figure 12: Alternative suppliers and firms size

One possible explanation is that Chinese and Vietnamese firms engage in more specialized

manufacturing activities and, for this reason, rely on more specialized production inputs (e.g.,

custom-made inputs) and so have fewer alternative suppliers. We return to this possibility in

considering firm outputs below. Alternatively, it may be that the African manufacturing firms

face greater demand fluctuations, and so are more familiar with available alternative suppliers

than are their Asian counterparts. Given that most African respondents source imported inputs

locally, we also cannot rule out the possibility that multiple local importers ultimately rely on the

same foreign supplier.

The questionnaires also asked how many of the alternative suppliers the respondent knows

personally. Responses show a gradual increase with firm size, but no systematic differences

across countries. Yet we find substantial heterogeneity across countries regarding supplier credit

between the African and Asian firms in the sample. Figure 13 shows the proportion of sales paid

after delivery, again controlling for firm size. We see that firms in China and Vietnam are much

more likely to purchase inputs on credit than their African counterparts. This is particularly true

among small firms which are typically most in need of external finance. We do, however, note in

Zambia and Tanzania (but not in Ethiopia) a tendency for supplier credit to increase with firm

size. The apparent lack of trust between supplier and client in Africa is puzzling, especially given

that there is no difference between countries in the number of alternative suppliers that

respondents know personally. But it has been noted elsewhere (e.g., Fafchamps 2004). We do

0

5

10

15

Number of alternative suppliers


China Vietnam

Ethiopia Tanzania

Zambia

45

not, however, observe significant differences between countries in the extent to which

respondent firms extend credit to their own customers. This puts African manufacturers at a

disadvantage: they do not receive credit from their suppliers, but they extend credit to their own

clients.

Figure 13: Payment after delivery and firm size

Outputs

Turning to outputs, the survey reveals substantial heterogeneity between countries. The main

contrast is between China and the other surveyed countries. Figures 14 and 15 show the

proportion of sales destined for customers in the same city/district and in other places in the same

country, respectively. Not surprisingly, we find that larger firms are less likely to sell their output

locally and more likely to sell their output in other parts of the country. The Figures also show

that Chinese firms ship a substantially lower proportion of sales to customers in the same

city/district than do firms in the other study countries. This holds even after controlling for firm

size. Conversely, Chinese firms ship a higher proportion of domestic sales to customers in

another city or district.

Why this is the case is not entirely clear but a likely explanation is that the Chinese economy is

geographically much more diversified, with many large cities and broadly distributed demand for

0

20

40

60

80

Proportion of sales paid post-delivery


China Vietnam

Ethiopia Tanzania

Zambia

46

manufactured products. In contrast, African economies are much smaller and the demand for

manufactures in Africa is concentrated in few cities with sufficient consumer purchasing power.

Figure 14: Sales (same city) and firm size

0

20

40

60

80

100


China Vietnam

Ethiopia Tanzania

Zambia

Sales to same city/district

47

Figure 15: Sales (elsewhere in the country) and firm size

We also observe substantial country differences in the type of customer to whom firm sell.

Figure 16 shows that, relative to firms in the other study countries, Chinese firms sell a much

higher proportion to other manufacturing firms. This relationship holds after controlling for firm

size. On average, Chinese respondents sell 29% of their output to other manufacturing firms. In

Vietnam the corresponding average is 10% while it is only around 1% in each of the African

countries surveyed. These results indicate that there is much less vertical specialization within

the light manufacturing sector in Africa and, to a lesser extent, in Vietnam than in China. In other

words, African firms produce light manufactures in a vertically integrated manner, with hardly

any sales to downstream manufacturers. This means that gain from specialization are not

captured. Such a feature is typically associated with less developed economies where the size of

the market is insufficient to allow the emergence of producers who specialize in a specific aspect

of the value chain.

Other sales go to final consumers, wholesalers, traders, or export agencies. After controlling for

firm size, we find no substantial differences between countries.

-20

0

20

40

60

Sales elsewhere in the same country


China Vietnam

Ethiopia Tanzania

Zambia

48

Figure 16: Sales to manufacturers and firm size

We noted earlier that Chinese and Vietnamese firms have fewer alternative suppliers than

African firms. We suggested that this may reflect the use of more specialized, and possibly

individually tailored, industrial inputs. We now consider this issue from the perspective of firm

output. Figure 17 shows the average proportion of output that each studied firm sells to its main

customer. It is striking that, even after controlling for firm size, the Asian firms sell a

substantially higher proportion of their production to their main customer. Similarly, a separate

question (omitted here for brevity) indicates that the Asian firms do not have a larger number of

alternative customers to whom they could sell.

Taken together, these results are consistent with the observation that the Chinese and Vietnamese

manufacturing sectors involve a higher degree of vertical specialization, here reflected in the

form of specific firm-to-firm corporate relationships. This observation runs contrary to many

standard models of manufacturing behavior in which firms produce for a large market of

anonymous customers. But it is consistent with industrial organization models of relational

contracting and vertical integration though long-term buyer-seller relationships. It adds further

weight to idea that the network structure of a country‘s manufacturing sector may be an

important determinant of its performance.

0

20

40

60

80


China Vietnam

Ethiopia Tanzania

Zambia

Sales to other manufacturers

49

Figure 17: Sales to main customer and firm size

Next we examine whether Asian firms engage is less advertising, as should be the case if their

output is more likely destined to another manufacturer with whom they have a long-term

relationship. Some support for this conjecture can be found in Figure 18 which shows that

advertising increases with firm size but, once we control for size, Asian firms appear, on

average, less likely than African firms to engage in advertising. This is potentially consistent

with a number of potential explanations, but it is not inconsistent with the idea that African firms

are more dependent upon demand from a large open market, whereas Asian firms are more

reliant upon particular bilateral arrangements.

20

40

60

80

Sales to main customer (%)


China Vietnam

Ethiopia Tanzania

Zambia

50

Figure 18: Advertising and firm size

To investigate this issue further, a series of questions were asked to survey participants. In

particular, respondents were also asked about: (i) the proportion of sales that are custom-made;

(ii) the number of alternative customers to whom products could be sold; (iii) how many of these

alternative customers the respondent knows personally; and (iv) how long it would take the

firm‘s customers to find alternative suppliers if the firm were to shut down. If firms sell most of

their output to a single buyer because their production is custom-made to the buyer‘s

specification, we would expect more custom-made production in China and Vietnam. This is not

what we find: there is no significant difference across countries. Similarly, if the focus on a

single buyer is due to relation-specific investment, we would expect answers to question (ii) to

be lower among Asian respondents – and answers to question (iv) to be higher. This is not what

we find: again there is no significant difference among countries, even after controlling for firm

size.

To summarize, we find some evidence of more vertical specialization in China and, to a lesser

extent, in Vietnam. But even though firms in our sample sell to a limited number of buyers,

particularly in Asia, all appear to remain fairly flexible in terms of the choice of potential buyers.

This suggests that relation-specific investments remain limited, even in China – at least among

firms in our sample.

-.5

0

.5

1

Proportion having advertised (past 6 months)


China Vietnam


51

Finally, we consider the issue of exports. Figure 19 shows the proportion of firms exporting,

across the different countries. Almost 40% of the Chinese firms in our sample export part of

their output. The corresponding figure for Vietnam is 17%. In contrast, manufacturers in our

three African study countries are much less likely to export. The differences are statistically

significant, including the difference between Vietnam and Tanzania. A similar finding obtains

after controlling for firm size. We also note that, of the five countries in our study, the two that

are landlocked – Ethiopia and Zambia – tend to export less manufactures. This is hardly

surprising given that the additional cost required to ship goods abroad reduce the

competitiveness of exports from such countries.

Figure 19: Exports

Chinese and Vietnamese exporters have also done so for longer on average: the median year of

first export in China and Vietnam is 2005 and 2004 respectively while in Ethiopia, Tanzania and

Zambia, it is 2008, 2007 and 2008 respectively. The destination of African exports is also

different. When asked about the main countries of destination for their exports, African firms cite

primarily other African countries. In contrast, Asian respondents cite both other Asian countries

and a variety of large Western economies.

How representative these results are is unclear. African countries that have been more successful

in exporting manufactures, such as Kenya (neighboring Ethiopia and Tanzania) and South Africa

(an important commercial partner of Zambia), are not included in the study. But there is no

0

.1

.2

.3

.4

Proportion of firms exporting


52

denying that Africa has not been as successful as China in exporting manufacturers outside the

continent.

Labor

We showed in Table 2 and Figure 1 that the size distribution of firms differs markedly across the

five country samples: firms in China tend to have more workers while African firms have less,

with Vietnam occupying an intermediate position in our data. We now turn to the occupational

structure of labor force in the five country samples.

Table 15 shows the share of workers in various four broad occupational categories, for the year

of the survey and the year before. Skilled and unskilled production workers are those who

physically produce output. A firm, however, cannot operate efficiently without management and

clerical workers. Management is in charge of making strategic choices but also of organizing and

monitoring production workers. Clerical workers help firms hold accounts and records and are

essential in large hierarchical organizations because of their role in processing and channeling

information essential to the organization of production.

Given that the firms we study are all in manufacturing, labor productivity would be theoretically

maximized if management and clerical workers are kept to a minimum. The extent to which

firms can minimize their management and clerical workforce, however, depends on how easy it

is to organize and monitor production workers. This, in turn, depends on the education level of

the workforce (e.g., whether they can read written instructions) and on social norms regarding

discipline and effort. The less disciplined and educated the workforce is, the larger the need for

monitoring and information processing workers, and thus the larger the share of management and

clerical workers in the total workforce.

Table 15 shows that the share of production workers in total employment is largest in Vietnam

and lowest in Zambia. Vietnam also has the largest share of skilled production workers. In

contrast, Zambia has the smallest share of production workers and the largest share of managers,

suggesting that labor management is more problematic there. The other three countries – China,

Ethiopia, and Tanzania – have a relatively similar breakdown between production and non-

production workers. There is some variation between workers reported as clerical or

management, but this variation is difficult to interpret as it may be affected by differences in

translation and local legal definitions.

Table 15: Breakdown of labor by occupational categories


Occupational

breakdown: Now

1 yr

ago Now

1 yr

ago Now

1 yr

ago Now

1 yr

ago Now

1 yr

ago

Skilled production

workers 46% 47% 62% 61% 58% 58% 50% 49% 49% 51%

53

Unskilled production

workers 27% 28% 19% 19% 15% 15% 25% 25% 17% 15%

Total production workers: 73% 75% 81% 81% 73% 73% 75% 74% 66% 66%

Management 16% 14% 4% 3% 23% 23% 15% 16% 27% 27%

Clerical and other 10% 10% 16% 16% 4% 3% 8% 8% 6% 6%

Number of observations 255 226 299 296 249 205 261 248 262 254

Next we examine the breakdown of the workforce between permanent and casual workers. We

tabulate in Table 16 the average share of permanent workers in surveyed firms‘ workforce at the

time of the survey, one and two years before, and at startup. In the absence of legal restrictions,

the breakdown between permanent and casual workers is expected to depend on the amount of

fluctuation in demand: the more variable and unpredictable demand is over time, the larger the

share of casual workers. Legal restrictions also affect firms‘ propensity to hire permanent

workers. In particular, restrictions on firing are thought to discourage firms from hiring

permanent workers.

The figures from Table 16 indicate that the proportion of permanent workers is high in Vietnam,

Ethiopia and Zambia. It is slightly lower in China, which may be due to fluctuations in demand

associated with rapid growth. Alternatively, it may be due to differences in labor laws and

regulation. Tanzania stands out as an outlier, with a much larger share of casual workers in total

workforce: in the average Tanzanian sample firm, casual workers represent two thirds of the

workforce. This may be due to differences in firm size, however: across the sample as a whole,

there is indeed a strong positive association between the share of permanent workers and firm

size.

Table 16: Share of permanent workers in total workforce



At time of survey

193 69% 299 83% 249 83% 261 37% 262 74%

One year before survey

167 70% 296 84% 205 84% 248 37% 252 75%

Two years before survey

157 70% 263 85% 169 88% 186 38% 234 82%

At the start of the firm

138 74% 300 88% 173 91% 176 36% 235 87%

To investigate this possibility, we estimate a non-parametric regression. Results are presented in

Figure 20. We see that variation in firm size cannot explain the differences between sample

countries: China and Tanzania stand out as having a significantly lower proportion of permanent

workers at all firm sizes relative to the other three countries.

54

Figure 20: Share of permanent workers and firm size

We now turn to the recruitment methods used by sample firms. Table 17 tabulates respondents‘

answers about their most common worker recruitment method. The proportion of missing

observations is very large in China, so responses should be regarded with some caution. In all

countries recruitment at the firm-gate – i.e., workers present themselves spontaneously to the

firm – is the dominant method for identifying workers. Job postings are used by about a quarter

of respondents in China and Vietnam, but by very few firms in Ethiopia and Tanzania, with

Zambia occupying an intermediate position. Around 17% of Chinese respondents rely on a

public or private placement agency. Proportions are much lower in the other four countries.

Close to a third of respondents in the three African countries mention other recruitment methods

– mostly word-of-mouth and friends and relatives. Recruitment through job postings, schools, or

placement agencies can be regarded as more formal than relying on friends and relatives or

waiting for workers to show up at the firm gate. With this definition, we see that 48% and 37%

of Chinese and Vietnamese firms, respectively, make use of formal methods for recruiting

workers. These percentages drop to 18%, 12% and 3% in Zambia, Ethiopia, and Tanzania,

respectively.

.2

.4

.6

.8

1 Share of permanent workers


China Vietnam

Ethiopia Tanzania

Zambia

55

Table 17: Recruitment method


Firm-gate 45% 51% 54% 62% 53%

Job posting 26% 26% 3% 1% 13%

Private placement agency 8% 4% 8% 1% 1%

Public placement agency 9% 2% 1% 0% 0%

School 4% 5% 0% 1% 4%

Other 7% 12% 33% 34% 28%

Number of valid observations 130 300 242 260 245

% missing observations 57% 0% 3% 1% 7%

Respondent firms were also asked what proportion of new employees is recommended by friends

and relatives of the owner or manager. Average responses are not very different across countries,

but the distribution of responses is. In the three studied African countries, the median firm stated

no hiring anyone recommended by friends and relatives. There is, however, a sizeable proportion

of firms that hire workers in this manner, and those who do tend to hire a large proportion of new

employee through this method. In contrast, firms in China and Vietnam are more likely to state

that they hire some workers in this manner, but the proportion of total employment hired in this

way remains small, e.g., 10% of new recruits. The data also show that recruitment through

friends and family is much more common among small firms. Firms were also asked whether

they ever hire production workers without a recommendation or referral. The most common

answer is yes, with little variation across countries.

From this we tentatively conclude that the three African countries in our study are less formal in

the way they recruit workers. This is partly due to the smaller size of firms. But it is also

suggestive of a less efficient operation of the labor market: if firms rely on spontaneous

applications and friends and relatives only, they are less likely to identify the workers best

qualified for the job.

Respondents were asked the proportion of migrants from other regions or countries in their

workforce. In Vietnam and Tanzania, the average response is half of the firm‘s workforce but

only 5% in Zambia. Ethiopia (19%) and China (26%) occupy an intermediate position.

In order to ascertain the extent to which labor regulation affects firm size, respondents were

asked how many workers the firm would wish to lay off or hire if there were no legal or

regulatory restrictions on hiring and firing. It is unclear how much confidence one should put in

respondents‘ answers – they may overstate their answers in an attempt to influence policy. But

the contrast between countries should be informative. Most firms state that, without legal or

regulatory restriction they would not lay off any worker: the median number of workers

56

respondents would lay off is 0 in all five countries. Except in China – where the average number

of workers is 9 – in the other four countries the average is quite low: around 1 worker per firm.

Respondents consistently report wanting to hire more workers if restrictions on hiring and firing

were removed. The average response is particularly large in China – 33 additional workers on

average, with a median of 20. In the other four countries, the average is around 3 to 5 workers,

with a median of 0, 1 or 2. From these answers we tentatively conclude that labor regulations are

most distorting in China.

There are big differences across studies countries in the provision of in-kind benefits to workers.

Table 18 shows the proportion of firms that offer housing or subsidized meals to workers. It also

reports the proportion of firms where a toilet with running water is available to workers on the

workplace, and the proportion of firms in which at least some of the workers belong to a trade

union. Housing provision is common in China but rare elsewhere. The provision of free or

subsidized meals varies across countries without any strong regional pattern. Toilets with

running water are much more common in China and Vietnam than in the three African countries

in our sample, largely reflecting different standards of sanitation and water distribution between

countries. The only country where a sizeable share of workers belongs to a union is China. In the

other four countries, unions are largely absent of manufacturing.

Table 18: In – kind benefits and unionization


Housing provided 82% 21% 8% 11% 15%

Meals provided 80% 60% 16% 37% 77%

Toilet provided 90% 88% 38% 53% 46%

Presence of union 65% 7% 2% 5% 5%


To conclude this section, we examine the education and experience of entry level production

workers. In Figure 21 we report the cumulative distribution of the average education level of a

new production worker in each of the five countries. The lower the curve, the higher the

education level of workers. We see that China and Vietnam have the lowest curve overall,

indicating that their workforce is better educated in general than production workers in the three

African countries in this study. Among the African countries, Zambia has a large proportion of

workers (37%) with no education at all. In contrast, only 5% of Tanzanian production workers

have not completed primary school, but 77% have not gone beyond primary school. In China and

Vietnam, only a small proportion of production workers have less than 9 years of schooling. This

probably reflects differences in legislation regarding compulsory schooling in past decades.

57

Figure 21: Schooling of production worker

What is unclear is whether production workers with a low education level require more training

before becoming fully productive on the factory floor. Respondents were asked how many weeks

it takes to train a new production worker. The cumulative distribution of their answers is

presented in Figure 22. We notice strong differences between countries. In Ethiopia and China,

85% and 90% of respondents, respectively, report that it takes at most four weeks for new

workers to be fully trained. Some 56% of Ethiopian respondents even state that new workers

require no training at all. In contrast, in Tanzania and Zambia, only 58% and 48% of

respondents, respectively, report that new workers are trained in 4 weeks or less. A sizeable

minority of respondents even report very long training periods, in excess of half a year. Vietnam

occupies a somewhat intermediate position: new workers are seldom regarded as adequately

trained when they begin work, but 87% of respondents estimate that after four months new

workers are trained. Firms that report hiring better educated workers tend also report shorter

training periods, but the correlation is not particularly strong, so it is unclear how much of a

saving in training schooling represents for manufacturing firms.

58

Figure 22: Weeks to train worker

Finance

The questionnaires included an extensive module on firm finance. This is an important aspect of

firm performance. In all five countries, for example, at least half of the firms surveyed reported

having purchased or acquired machinery, equipment, or vehicles at some point in the past three

years.

Table 19 summarizes the sources of finance that firms reported having used for those

acquisitions. The table shows that, for all countries, retained earnings are by far the most

important source of finance. In all countries but Ethiopia, bank borrowing is the second most

important; in Ethiopia, non-bank financial institutions appear to play a greater role than formal

banks.

59

Table 19: Source of finance for last purchase of machinery, equipment or vehicles


Internal funds / retained earnings 80% 82% 88% 80% 86%

Borrowed from a bank 23% 18% 2% 4% 3%

Borrowed from a non-bank financial

institutions 7% 2% 10% 2% 1%

Borrowed from a government agency 0% 0% 0% 0% 0%

Funds from family / friends 7% 18% 4% 3% 1%

Hire-purchase / Credit from the equipment

supplier 5% 1% 0% 0% 0%

Other 0% 1% 3% 1% 8%


This raises the question of the relationship that studied firms maintain with banks. Presumably

the first step before securing a loan from a bank is to have a current account. By observing the

movements of funds on a firm‘s account, the bank can gain valuable information on the financial

wherewithal of the firm. Not having this information should make banks more reluctant to lend.

At the same time, if firms derive no benefit – in terms of loans or otherwise – from banks, there

is little reason for them to maintain a bank account. We therefore expect that fewer African firms

have a bank account, and this is indeed what we find, especially among smaller firms. Figure 23

shows the average probability of having a bank current account controlling for firm size. We see

that smaller Chinese firms are substantially more likely to have a bank account than firms in the

other four countries. In a probit regression with a size coefficient common across countries, we

easily reject the hypothesis that African and Asian firms are equally likely to have a bank

account: Asian firms are on average 10 percentage points more likely to have an account.

60

Figure 23: Bank current accounts

We then examine whether having a bank account is associated with a higher probability of

obtaining the simplest form of bank finance, namely, an overdraft facility (e.g., line of credit).

Results are stark: among firms with a current account, a very small proportion has an overdraft

facility. In Vietnam and Ethiopia, only a handful of manufacturing firms (3 and 2, respectively)

have an overdraft facility. Proportions rise to 6% of firms with a bank account in Tanzania and

19% in Zambia, but in these two countries only a minority of firms has a bank account. China

stands out in sharp contrast to the other four countries: 63% of firms with a bank account have an

overdraft facility.

Overdraft facilities therefore appear to be a very important mechanism by which Chinese

manufacturing firms access short-term finance. Further, the terms for that finance do not

generally appear restrictive; in a series of follow-up questions, those Chinese firms with an

overdraft facility reported a median annual interest rate on the overdraft of 7.5%, and only about

20% of the firms were required to provide collateral (where the median collateral value was 65%

of the value of the overdraft facility).6

If firms cannot overdraw their bank account to deal with emergencies or take advantage of

passing investment opportunities, they need to accumulate positive balances to serve this

6 Non-response may be an issue, however: of the 160 Chinese firms reporting an overdraft, only 119 answered the

question about collateral and only 68 answered the question about the interest rate.

0

.5

1

Proportion having a bank current account


China Vietnam


61

purpose. Figure 24 shows the proportion of firms having a savings account, controlling for firm

size. Again, there is substantial heterogeneity across countries. Contrary to expectations, it is not

firms in countries other than China that accumulate balances as a substitute for bank credit; it is

again the Chinese firms. Why this is the case is unclear – one possibility being that returns on

savings are higher than subsidized interest charges on overdraft facilities. We also note that

Ethiopian firms make substantial use of savings accounts: across the entire sample, 56% of

Ethiopian firms have a savings account. This compares to 86% in the Chinese sample, but 22%

in Vietnam, 5% in Tanzania, and 28% in Zambia.

Figure 24: Savings account

Having discussed the issue of liquidity financing, we now turn to loans. Table 20 presents

responses from questions about whether the firm borrowed money between 2006 to 2010 and, if

so, from what source.

0

.2

.4

.6

.8

1

Proportion having a savings account


China Vietnam


62

Table 20: Proportion having borrowed from different sources, 2006-2010


Borrowed from a bank 33% 36% 4% 3% 3%

Borrowed from a non-bank financial institutions 15% 4% 32% 2% 4%

Borrowed from a government agency 12% 2% 0% 0% 1%

Funds from family / friends 15% 19% 18% 2% 3%

Moneylender 0% 5% 0% 0% 1%

Other 0% 0% 6% 0% 1%


The table supports the earlier finding that Chinese and Vietnamese firms make more extensive

use of credit than do firms in Ethiopia, Tanzania or Zambia. We also see that Ethiopian firms

make substantial use of non-bank financial institutions, perhaps to compensate for limited access

to bank finance. These conclusions are robust to controlling non-parametrically for firm size.

Further, non-parametric regressions show that, as one might expect, finance from family or

friends (or from moneylenders) is important only for smaller firms.

Next we examine the proportion of firms currently owing money to a lender. About a third of

surveyed firms in China, Vietnam, and Ethiopia currently owe money on a loan. The proportions

are much lower in Tanzania and Zambia: 11% and 6%, respectively. These findings are by and

large in agreement with Table 20 once we take into account the fact that loans from

moneylenders or from family and friends are for a much shorter duration – and are thus less

likely to be observed at any given point in time.

Loan terms tend to differ across study countries. Figure 25 shows the time for loan repayment,

controlling non-parametrically for firm size. The Figure refers to the most recent loan from a

bank, non-bank financial institution, or government agency. We see that, except in Ethiopia,

loan duration increases with firm size in Africa while in Asia there is no such association with

firm size. Since there are more large firms in the China and Vietnam samples, it follows that, on

average, loans to manufacturing firms in China and Vietnam are on average for a shorter time

period than in Ethiopia, Tanzania and Zambia.

63

Figure 25: Loan duration

Why loan duration increases rapidly with firm size in Africa is unclear. One possibility is that

transaction costs for borrowing are lower in China and Vietnam than in Ethiopia, Tanzania or

Zambia. If this is true, banks in China and Vietnam may be more willing to lend for

manufacturing firms because they can profitably make many short loans.

To investigate this possibility, we examine the differences across countries in terms of collateral

requirements and average interest rates. Figure 26 shows the proportion of firms providing

collateral for their most recent loan. Even after controlling for firm size, Chinese firms are shown

to face substantially lower average collateral requirements. Similarly, Chinese firms pay an

average annual interest rate of about 4.7%, compared to average annual interest rates of about

10% for Ethiopia, 14% for Vietnam and Tanzania and 21% for Zambia. These proportions

change little across firm size. These differences cannot be explained merely by differences in the

information available to lenders: the proportion of firms having their annual financial statements

certified by an external auditor increases systematically with firm size, but does not differ

significantly between countries once we control for firm size.

Not having to incur the cost of securing collateral likely reduces total lending costs for the

lender, and this may explain why Chinese banks are more likely to lend to manufacturers and

charge smaller interest rates. Another possible explanation is that Chinese banks, many of which

are still in government hands, have been instructed to lend to manufacturers at low interest

0

10

20

30

40

50

Time to repay (months)


China Vietnam

Ethiopia Tanzania

Zambia

64

without insisting too much on collateral. As long as Chinese manufacturing output is increasing

rapidly, the risk of default remains low. But should manufacturing growth fall for an extended

time, these uncollaterized loans could set the Chinese banking sector in crisis.

Figure 26: Collateral requirements

Actual loans may provide a misleading picture of firm access to credit if many firms that could

get a loan chose not to do so, either because they prefer to self-finance, or because they have not

identified a suitable investment opportunity. To investigate this possibility, firms were asked

whether they could borrow either to purchase equipment or to expand their working capital.

Answers to these two questions are tabulated in Tables 21 and 22 below. They do not differ

much between the two tables, so they are discussed together.

For China and Vietnam, we find that responses given to the two questions largely mirror actual

borrowing: a little over a third of the respondents (half for working capital in China) states that

they could borrow from a bank, and ten to twenty percent from family and friends. In Ethiopia

we find that 45% of the surveyed firms reckon that they could borrow from a non-bank financial

institution. This compares to 32% of respondents actually having borrowed from such a source in

the past. In contrast, a large proportion of respondents in Tanzania and Zambia state that they

could borrow from a bank or a financial institution, far in excess of the negligible percentage of

firms (i.e., less than 5%) that have actually done so in the past.

0

.2

.4

.6

.8

1

Proportion providing collateral


China Vietnam

Ethiopia Tanzania

Zambia

65

The reason for the discrepancy is unclear. One possibility is that firms in Tanzania and Zambia

could borrow if they wanted to but chose not to, either because they prefer to self-finance or

because they prefer not to invest. Another possibility is that respondents have erroneous

expectations. The latter explanation is a serious concern in the Tanzania sample where most

firms are quite small and unlikely to be appealing borrowers to the average commercial bank.

Table 21: Proportion that could borrow from different sources to purchase additional

machinery equipment or vehicles


A bank 36% 34% 8% 60% 32%

A non-bank financial institution 10% 1% 45% 12% 20%

A government agency 0% 1% 7% 3% 12%

Family / friends 11% 22% 4% 5% 9%


Other 0% 0% 0% 0% 7%


Table 22: Proportion that could borrow from different sources to expand their working

capital


A bank 49% 35% 7% 63% 27%

A non-bank financial institution 15% 1% 42% 15% 14%

A government agency 1% 2% 7% 3% 9%

Family / friends 10% 23% 6% 4% 11%


Other 6% 3% 1% 1% 6%


66

The surveys also asked firms about the maximum amount that a lender might allow the firm to

borrow. Figure 27 summarizes the responses; it shows that differences in firms‘ expected

maximum loan amount are explained by differences in firm size, rather than by differences in

country.

Figure 27: Maximum loan amount

Productivity

Finally, we consider measures of firm productivity and examine how different firms in the

sample choose to organize their operations, and how this relates to their output.

Because productivity analysis requires balance sheet information, respondent firms were asked to

provide some basic accounting information. Unfortunately, the proportion of non-response is

high. This is relatively common in surveys such as this because many firms are either unable or

unwilling to disclose information on their costs and profits. The results presented here are based

on available answers. In spite of this caveat, the information provided is useful for thinking about

differences in firm performance.

Figure 28 shows the relationship between reported annual profits and firm size. This variable is

particularly affected by non-response: some 55% of Chinese firms refused to answer this

6

8

10

12

14

Log (maximum loan amount)

0 2 4 6 8 Log (regular workers)

China Vietnam

Ethiopia Tanzania

Zambia

67

question, and about 70% in each of Tanzania and Zambia. Non-response rates in Vietnam and

Ethiopia were fortunately much lower – 4% and 15%, respectively. The Figure provides

suggestive evidence that, even after controlling for the number of regular workers, firms in China

and Ethiopia earn higher profits than firms in Vietnam and Tanzania, with Zambia occupying an

intermediate position.

Figure 28: Profits

The general trends of Figure 28 are confirmed when we examine the relationship between annual

sales and firm size, shown in Figure 29. This is important because response rates on this question

are higher and thus less subject to possible self-selection bias: non-response in China was about

29%, while it was about 70% and 45% in Tanzania and Zambia respectively. Figure 29 shows

that, even after controlling for firm size, Chinese firms sell significantly more than Vietnamese

firms – which, over much of the firm size range, sell more than Ethiopian, Tanzanian or Zambian

firms.

6

8

10

12

14


China Vietnam

Ethiopia Tanzania

Zambia

Log (profits)

68

Figure 29: Sales

Figures 28 and 29 imply that Chinese and Vietnamese manufacturing workers are, on average,

more productive than manufacturing workers in Ethiopia, Tanzania and Zambia. This finding

prompts several questions. First, do Chinese and Vietnamese firms pay higher labor costs in

return for having more productive workers? Second, to what extent can higher productivity be

explained by Chinese and Vietnamese firms having more valuable capital in the form of

machinery, equipment, vehicles, land and buildings?

We begin with the question about labor costs. Figure 30 shows aggregate labor costs by firm

size, including wages, salaries, bonuses, social security, etc. The graph shows that labor costs

are higher in China than in Vietnam and Zambia, whose costs are higher again than in Ethiopia

and Tanzania. These results are in line with expectations: higher labor productivity translates into

higher worker compensation.

8

10

12

14

16


China Vietnam

Ethiopia Tanzania

Zambia

Log (sales)

69

Figure 30: Labor costs

6

8

10

12

14

Log (annual labor cost)


China Vietnam

Ethiopia Tanzania

Zambia

70

The question that follows is how Chinese and Vietnamese firms are able to pay higher wages.

One obvious possibility is that they have more machinery and equipment per worker. We

examine this possibility in Figure 31. Unfortunately, the data is missing for China, firms refusing

to answer. We see that, over much of the firm size range, Vietnamese firms have more capital

than their African counterparts. Large firms in Ethiopia and Tanzania, however, appear to have

more capital invested in the firm than large Vietnamese firms.

Figure 31: Value of machinery, equipment and vehicles

What these results indicate is that Chinese firms have larger profits and sales, but also larger

labor costs. We suspect this is accounted for by a combination of larger capital investment, better

infrastructure, and agglomeration effects, but unfortunately we have no information on capital

for Chinese firms. Vietnamese firms occupy an intermediate position: small Vietnamese firms

perform better than their African counterparts – i.e., higher profits and sales – and spend more on

workers. But they do not perform significantly better than the larger African firms included in

our study – which by international standards would be regarded as medium-size firms. Although

it is perilous to extrapolate from this limited evidence, what these results suggest is that the

aggregate productivity of African manufacturing ultimately depends on the presence of larger

firms. Since research has already shown that very small firms in Africa (and elsewhere) seldom

if ever grow to be medium-size firms, what is needed is entry by entrepreneurs capable of

operating medium-size manufacturing firms successfully.

4

6

8

10

12



Zambia

Log (value of machinery etc)

71

This prompts consideration of the role of land and buildings. Like machinery, equipment and

vehicles, land is an important factor for firm production. Firms may face difficulties investing in

land if land suitable for manufacturing development is scarce in urban centers. The

questionnaires asked about (i) the resale value of land and buildings and (ii) the

purchase/acquisition of land and buildings in 2010. Unfortunately, non-response was a

particularly severe problem for questions about land and buildings. We report data only from

countries where at least 20 firms responded. This excludes China and Ethiopia for the question

on land value, and Ethiopia and Tanzania for the question on land acquisition.

In spite of these limitations, the questions provide suggestive evidence on land availability.

Figure 32 shows the reported value of land and buildings controlling for firm size. The Figure

suggests that, for the sample of firms responding to the question, the value of land is higher in

Tanzania and Vietnam than in Zambia.

Figure 32: Value of land and buildings

Figure 33 considers firms that reported acquiring land and building in 2010. Even controlling for

firm size, Chinese firms responding to the question spent significantly more on the acquisition of

land and building than did firms in Vietnam – which, in turn, appear to have spent more than

firms in Zambia (though the difference here is less significant).

There are two possible explanations for these differences. First, it may be that firms in some

countries pay more for land and buildings because land is more productive, i.e., is closer to better

4

6

8

10

12

14

Log (value of land and buildings)


China Vietnam


72

infrastructure. In effect, high prices may be driven by high firm demand. This is analogous to

the earlier discussion about the high relative costs of labor in China: in the same way that high

labor productivity may cause higher wages, high land productivity (e.g., in terms of access to

infrastructure) may drive high costs for land and buildings. Second, it may be that firms in some

countries pay more because of limitations of available land; in effect, high prices may be driven

by limited land supply.

There are reasons to believe that both supply and demand factors are responsible for the answers

given by survey respondents. Indeed manufacturing firms in China and Vietnam are likely to

have a higher demand for land and buildings because of their relatively higher productivity, but

they may also be more constrained in their access to suitable land if firms compete for the

limited supply of industrial land. Without exogenous variation in demand and supply, the

questions in the present survey cannot allow us to separate these effects. It remains an important

area for further research.

Figure 33: Purchases of land and buildings

Finally, we consider firm growth. We measure firm growth as the average change in log sales

between 2008 and 2010. We trim (i.e., drop) the top 5% and bottom 5% of the observations in

order to ensure robustness to outliers. Unfortunately many firms did not report sales in 2008,

with strong differences across countries. Growth figures must therefore be regarded as indicative

only, as they may be biased by selection bias.

4

6

8

10

12

14

Log (purchases of land and buildings)



Zambia

73

Figure 34 shows the empirical probability density across the five countries. There is substantial

overlap in growth experiences across the firms in the study, i.e., it is not the case that most firms

in China, for instance, are growing much more rapidly than firms in Africa or Vietnam.

Figure 34: Sales growth

To illustrate this more clearly, Table 23 summarizes median growth in sales in the five countries.

We see that median sales growth is strong across all five countries and, if anything, is lower in

China than in the other countries except Zambia. The main difference between China and the

other countries is that average firm size is much larger and manufacturing represents a sizeable

proportion of GDP. Hence a 14.8% growth rate in sales has a large effect on aggregate growth.

This is not true in the other countries, and particularly the three African countries in our sample,

where firms are smaller and manufacturing only represents a minute portion of domestic GDP.

0

.5

1

1.5

2

2.5

-.5 0 .5 1 1.5 Change in log(sales), 2008-2010 (trimmed)

China Vietnam

Ethiopia Tanzania

Zambia

Density

74

Table 23: Median sales growth

China Vietnam Ethiopia Tanzania Zambia Total

Median growth (%) 14.8 15.2 30.0 18.5 14.0 17.6

Nber observations 91 231 115 48 69 554

75

BACKGROUND PAPER II

The Binding Constraint on Firms’

Growth in Developing Countries Hinh T. Dinh, Dimitris A. Mavridis, Hoa B. Nguyen

7 November 2010

Abstract

Firms in developing countries face numerous and serious constraints on

their growth, ranging from corruption to lack of infrastructure to inability

to access finance. Countries lack the resources to remove all the

constraints at once and so would be better off removing the most binding

one first.

This paper uses data from World Bank Enterprise Surveys in 2006–10 to

identify the most binding constraints on firm operations in developing

countries. While each country faces a different set of constraints, these

constraints also vary by firm characteristics, especially firm size. Across

all countries, access to finance is among the most binding constraints;

other obstacles appear to matter much less. This result is robust for all

regions.

Smaller firms must rely more on their own funds to invest and would grow

significantly faster if they had greater access to external funds. As a result,

a low level of financial development skews the firm size distribution by

increasing the relative share of small firms. The results suggest that

financing constraints play a significant part in explaining the ―missing

middle‖—the failure of small firms in developing countries to grow into

medium-size or large firms.

1. Introduction

Private sector growth remains one of the main challenges facing developing countries in their

quest for development and poverty reduction. Extensive evidence shows that a favorable

business environment helps promote the growth of firms. As shown in recent research, however,

firms in developing countries face a tougher business environment than their counterparts in the

developed world.

Our aim in this paper is twofold. First, we seek to go beyond the traditional menu of constraints

on firm growth to find out which of these constraints is the most binding. As the growth

7 The authors would like to thank Alvaro Gonzalez, Anders Isaksson, Justin Yifu Lin, Vincent Palmade, and L.

Colin Xu for their valuable comments on a previous draft of this paper.

76

diagnostics approach points out, developing countries have scarce resources and therefore need

to focus on removing the most binding constraint. Second, we examine the effects of the most

binding constraint on firm growth not only across countries but also by firm characteristics.

We first explore the relationship between the business environment and firm growth as measured

by employment growth. Among 15 components of the business environment, we identify the

most binding constraint using both subjective and objective measures. Our focus is on the most

binding constraints for existing firms and, more specifically, the binding constraint that matters

the most for firm growth. The methodology follows two steps. The first is to find out which

constraints are statistically significant among all regressions after controlling for firm

characteristics and country fixed effects. The second is to identify the most binding constraint.

We find that besides informal sector competition, access to finance is the obstacle that matters

the most for growth. This result is robust for all regions and all sectors.

Our analysis contributes to the existing literature in several ways. First, using a large sample,

containing more than 39,000 firms across 98 countries, we identify the most binding constraint

on firms using subjective measures, then evaluate the importance of this constraint to firm

growth using objective measures and controlling for firm characteristics. The sample comes from

World Bank Enterprise Surveys conducted in 2006–2010 in mostly emerging and developing

countries. The surveys provide both subjective data on perceived obstacles and objective

measures of many constraints.

Second, we investigate the effect of financial access variables on firm growth by using firm-level

regressions across countries controlling for the effects of different firm sizes, firm ages, sectors,

and regions. Our results show that having access to finance in the form of a loan, sales credit, or

external finance helps micro firms the most. This finding holds not only for the full sample but

also for different regions. Sales credit is important only to micro and small firms, probably

because it substitutes for bank loans. Having a loan or overdraft facility and receiving external

finance for investment help growth for firms of all sizes across regions.

Third, we find clear evidence that a low level of financial sector development affects the firm

size distribution and therefore contributes to the phenomenon of the ―missing middle‖ in

developing countries. Firm size distribution is skewed toward small and medium-size firms—and

more so in Africa, among firms that are credit constrained, and among firms that perceive access

to finance as an obstacle. Our analysis shows that firm size and age are significantly correlated

with firm growth. Distinguishing between different types of ownership, we find that firms tend to

have higher growth if they are an exporter, are part of entities with multiple establishments, are

foreign owned, or are privately owned.

The paper is organized as follows. The next section reviews the literature and shows how this

paper relates to it. Section 3 presents an overview of the data and describes the sample used.

Section 4 examines the most binding constraint of the business environment. Section 5 examines

77

the effect of access to finance on employment growth in the full sample and by region, while

section 6 looks at determinants of access to finance. Section 7 investigates differences in the

effect of financial access variables on employment growth, and section 8 looks at the relationship

between firm size and financial constraints. The last section concludes.

2. Literature Review

Understanding the firm growth process is important for designing appropriate policies for job

creation and pro-poor growth. Many studies are devoted to understanding the determinants of

firm growth, especially employment growth. Most of these studies focus on the manufacturing

sector and large firms (Evans 1987a, 1987b; Hall 1987; Dunne, Roberts, and Samuelson 1989).

They find that firm age and firm size are important in the analysis of firm growth.

2.1 Effect of the Business Environment on Firm Growth

A number of recent studies use World Business Environment Survey data (a firm-level data set

covering 4,000 firms across 54 countries) to study the effect of the business environment on firm

growth. Using subjective, firm-level data on the business environment, some of these studies

show the importance of finance, corruption, and property rights (Batra, Kaufmann, and Stone

2003; Ayyagari, Demirgüç-Kunt, and Maksimovic 2006). Others examine the relationship

between the business environment and firm growth in individual countries or a small group

(Dollar, Hallward-Driemeier, and Mengistae 2005 in Bangladesh, China, India, and Pakistan;

Fisman and Svensson 2007 and Reinikka and Svensson 2002 in Uganda; Bigsten and Söderbom

2006 in Africa).

Other studies also assess the effect of different dimensions of the business environment on firm

growth. Some focus on the importance of access to finance for firm development and growth

using subjective data (Rajan and Zingales 1998; Galindo and Micco 2007). Others investigate the

impact of employment regulations on firm creation and growth (Djankov and others 2002;

Klapper, Laeven, and Rajan 2004).

Several papers have emphasized the importance of financing obstacles. Using firm-level data,

Demirgüç-Kunt and Maksimovic (1998) provide evidence on the importance of the financial

system and legal enforcement on firm growth. Rajan and Zingales (1998) present supporting

evidence on the role of external finance for faster growth in countries with better developed

financial systems. These papers focus on only a small set of obstacles that firms confront without

discussing the motivation for choosing that set.

It is essential to explore the relationship between the business environment and firm growth not

only across countries but across regions and by firm characteristics within countries—by firm

size, age, sector, and ownership type. In examining this relationship, the literature has focused

largely on the effect of difficulties in access to finance by firm type, particularly firm size.

Generally the finding is that smaller firms are more constrained (Love and Mylenko 2003; IDB

2007). Beck, Demirgüç-Kunt, and Maksimovic (2005), using the World Business Environment

Survey data set, include measures of corruption and property rights. Based on firms‘ perceptions

78

of potential constraints, they also find patterns across countries, with small firms benefiting the

most from greater financial and institutional development.

Aterido, Hallward-Driemeier, and Pagés (2007, 2009, 2010) analyze the effect of several aspects

of the business environment—access to finance, corruption, and regulations—on the growth of

firms. Their findings show that the business environment affects small, medium-size, and large

firms differently. The reason is that small firms are exposed to a different set of constraints than

large firms are. Access to electricity, for example, has heterogeneous effects: small and medium-

size firms are often affected by power cuts, while large and micro firms tend not to be. The main

reason is that micro firms use less energy-intensive tools and large firms are more likely to

secure their own energy supply (Gelb and others 2007). Thus infrastructure such as the

electricity grid affects the growth rate of small and medium-size firms directly, but has only an

indirect effect on the growth rate of micro and large firms. Micro firms are much more credit

constrained and must rely less on external funds to finance investment. Improving access to

finance might boost the entry rate and the growth of small firms, perhaps at the expense of larger

incumbents.

According to Dollar, Hallward-Driemeier, and Mengistae (2005), improving the business

environment is an important complement to trade policies aimed at increasing international trade

integration. Factors such as fast customs clearance times, good infrastructure, and availability of

financial services have a significant impact on the probability of a firm‘s exporting and receiving

foreign investment. Freund and Rocha (2010) provide more evidence of the link between the

business environment and international trade. Using data from Africa, they find that even though

poor trade infrastructure is one of the main obstacles to trade, most of the burden is due to heavy

―red tape,‖ bureaucratic customs practices that increase the time and cost of trade.

Gelb and others (2007) use subjective data on the business environment from 26 African

countries to show that perceived constraints are not always independent of scale. Complaints

about access to finance and land are more common among small firms, while complaints about

infrastructure and corruption are more evenly distributed. They also find that a country‘s level of

development strongly determines which constraints are present (country fixed effects are more

important than within-country variations). This finding is shared by the World Bank‘s Africa

Competitiveness Report 2009, which shows that as a country‘s income rises, its set of constraints

changes.

All these studies share a common result: business environment variables affect firms‘ growth, in

the expected direction. The results are heterogeneous by firm size, and they are robust.

2.2 Financial Development and Firm Size Distribution

A common finding in the literature is that the firm size distribution in developing countries is

skewed toward small and medium-size firms. Small firms are often credit constrained and cannot

borrow to engage in productive investments, which limits their growth and can prolong the

skewness. If lack of access to finance prevents small firms from growing, the allocation of

resources will be distorted. Capital and labor will not be able to flow to where they are most

productive, and growth will suffer.

79

Cooley and Quadrini (2001) and Cabral and Mata (2003) present different models of firms‘

growth, showing that capital constraints can cause a skewness in the firm size distribution. Their

prediction is verified empirically. Cabral and Mata (2003) find that the size distribution of firms

is skewed toward small firms and that the skewness decreases with firm age. Many subsequent

papers confirm the skewness of the firm size distribution, such as Angelini and Generale (2008),

Beck, Demirgüç-Kunt and Maksimovic (2005), and Desai, Gompers, and Lerner (2003).

Desai, Gompers, and Lerner (2003) find that in countries with less developed capital markets, the

firm size distribution is significantly more skewed. They also find that a better legal environment

favors entry (more small firms will enter) while the growth of small firms reduces the skewness.

Angelini and Generale (2008) and Beck, Demirgüç-Kunt and Maksimovic (2005) find that

capital-constrained firms grow more slowly than their counterparts.

2.3 Growth Diagnostics Approach

The growth diagnostics approach proposed by Hausmann, Rodrik, and Velasco (2005) (hereafter,

the HRV approach) provides a theoretical framework to identify the most binding constraints on

economic growth in general. This methodology recognizes that constraints on the growth of a

developing economy are numerous and that previous approaches to reforms and growth are

either unrealistic (as with wholesale reform that attempts to eliminate all obstacles at the same

time) or wrong (by hoping to do as many reforms as possible, the current prevailing approach

goes against the principle of second best).

The HRV approach is based on the theory of second best (Lipsey and Lancaster, 1956).

According to this theory, if there are many distortions in the economy, fixing any one distortion

would not necessarily lead to a better Pareto outcome. The HRV approach shows that if there are

many distortions, whether removing one growth constraint will have a positive effect on growth

depends on the interaction effects and coefficients of the other constraints. In the face of

uncertainty about these effects, Hausmann, Rodrik, and Velasco recommend a practical approach

based on removing the most binding constraint, with the ―most binding constraint‖ defined as the

one with the largest effect where issues of second-best effects are likely to be minimal.

2.4 This Study’s Contribution to the Literature

In this paper, based on the HRV approach, we investigate the most binding constraint on the

growth of firms, with the ―most binding constraint‖ defined as the one with the largest estimated

coefficient across all models and across regions and sectors. Compared with studies using the

World Business Environment Survey data set, our paper uses a much larger sample. And while

other studies use subjective firm responses as measures of the business environment at the firm

level, we also include objective measures, in part to deal with endogeneity and in part to avoid

measurement errors of perceptions at the country level.

In exploring the relationship between the business environment and firm growth, we go beyond

distinguishing effects by firm size. We look closely at the effect of financial access variables—

80

loan, credit constraint, sales credit, and external investment finance—on firm growth by firm size

and age in different sectors and regions. We combine multiple financial access variables in a

single regression in addition to evaluating the effect of each variable on employment growth

controlled for firm size, age, and other characteristics. This allows an understanding of the

impact of each dimension of finance on firm growth as firm characteristics change.

Moreover, our paper emphasizes which element of the business environment matters most for

firms, especially for small firms. And it analyzes how different financial access variables affect

the firm size distribution across regions and sectors.

3. Data

In this paper we use a newly available firm-level data set from the World Bank Enterprise

Surveys. The surveys cover more than 100,000 firms across more than 120 economies and six

regions during 2006–10. We use a sample of 39,538 firms in 98 countries for which data are

complete. The unit in the sample is the establishment; one firm may have more than one

establishment. For simplicity, we use the term firms throughout the paper, though the analysis is

based on establishment data.

Our outcome variable of interest is employment growth, measured by the number of permanent

employees. Our policy interest is in understanding the determinants that are important to the

long-term business operation and employment growth of firms.8 Because there are no data on

temporary employees collected three fiscal years before the survey fiscal year, we focus on

permanent full-time employees rather than general full-time employees.

The firm growth rate is calculated as the log difference between the current number of

employees and the number of employees three fiscal years before the survey fiscal year. The

formula for employment growth is as follows:

where is firm size, and employment growth, for firm i at time t.9 The description and

summary statistics for the employment growth variable are reported in table 1.

World Bank Enterprise Surveys are conducted to provide information on different aspects of the

business environment and the performance of firms. The core questionnaire, which contains

survey questions answered by business owners and top managers around the world, provides

both subjective and objective information on the business environment that firms confront. The

8 Like other researchers, we use employment growth rather than sales growth, for several reasons. Sales growth is more volatile and is also more prone to reporting and measurement biases, especially when survey respondents are reporting

sales realized three years before. Moreover, for tax reasons, firms may not choose to report actual sales. 9 ln(1+X) is considered approximately equal to ln(X). We therefore use ln(1+X) to compute the log of the number of employees, since some firms have zero employees in a specific year but not in both years.

81

questionnaire includes a section asking firms to rank 15 components of the business

environment, indicating which represent the biggest obstacles, and to evaluate these 15

components on a scale of 0–4 (0 being no obstacle, 1 a minor obstacle, 2 a moderate obstacle, 3 a

major obstacle, and 4 a very severe obstacle). Summary statistics for the related variables are

provided in table 1.

These subjective evaluations show the severity of obstacles across regions and countries. This

makes it possible to identify the top obstacles and examine which obstacles firms consider to be

the most important. But because the data are subjective—reflecting entrepreneurs‘ perceptions of

the impact of the business environment on firm operation, with successful entrepreneurs perhaps

likely to consider the business environment to be less restrictive—we need to control for firm

characteristics in explaining firm growth. In addition, we need to include objective measures of

business environment constraints.

The World Bank Enterprise Surveys provide a large set of objective measures of business

environment constraints. In addition to subjective information on access to finance as an

obstacle, the questionnaire also collects objective information on aspects of financial access,

allowing us to create several variables: Loan is a dummy variable indicating whether a firm has a

loan or line of credit from a financial institution or an overdraft facility. Credit constraint is a

dummy variable indicating whether an establishment did not apply for loans or lines of credit for

one or more of the following reasons: application procedures for loans or lines of credit are

complex, interest rates are not favorable, collateral requirements are too high, the size and

maturity of loans are insufficient, getting bank loans requires making informal payments, or the

establishment did not think its application would be approved.10

Sales credit is a dummy variable

indicating whether the firm has positive purchase of its material inputs or services paid for after

delivery (about 70 percent of firms in the sample have sales credit). We also include a dummy

variable indicating whether a firm has a positive share of investment financed with external

funds (this applies to 24 percent of firms in the sample).

The World Bank Enterprise Surveys also provide important information on firm characteristics,

including size, age, sector,11

export activity, and ownership as well as whether a firm is an

independent, single establishment. The sample used in this paper is stratified by size, age, sector,

region, and other firm characteristics. (Variable descriptions and distributions are reported in

tables 1 and 2.) Firms are divided into four categories by size: micro (1–10 permanent

employees), small (11–50), medium (51–200), and large (more than 200). The sample includes

mostly micro firms (39 percent of the total) and small firms (37 percent); only 16 percent are

medium-size and 7 percent large. Firms are divided into three categories by age: young (1–5

years),12

mature (6–15), and older (more than 15). Most are mature (47 percent) or older (41

10 No dummy variable is included for firms applying for new loans or lines of credit whose applications were rejected because this information is available for only 14 percent of firms in the sample. 11 The questionnaire provides information on industry, and we use this information to establish the sector variable. 12 Firms operating for less than one year are classified as young firms.

82

percent); only 11 percent are young firms. Ownership is defined as being foreign or government

if ―10 percent or more‖ of the firm is foreign or government owned; 12 percent of the firms in

the sample are foreign owned and only 2 percent are government owned. Exporter is a dummy

variable indicating that direct exports account for 10 percent or more of a firm‘s sales; 13 percent

of the sample firms are exporters.

Whether a firm has a single establishment or multiple ones matters for firm growth, especially in

the manufacturing sector (see Dunne, Roberts, and Samuelson 1989). We therefore include a

dummy variable indicating whether a firm is an independent, single establishment. Most of the

firms in the sample are single establishments (85 percent), while 14 percent are part of multi-

establishment entities. Finally, we divide the firms into three sectors: manufacturing (55

percent), sales (23 percent are in the retail and wholesale sector), and other services (20 percent).

The sample includes firms from six regions: 31 percent from Sub-Saharan Africa, 28 percent

from Latin America and the Caribbean, 27 percent from Europe and Central Asia, 11 percent

from East Asia and Pacific, and only 3 percent from the Middle East and North Africa and South

Asia.13

Table 1 provides an overview of firm growth by firm characteristics and by region. Young, small

firms experience rapid growth in their labor force. The mean growth rate for micro firms is twice

that for small firms and three times that for medium-size firms. There appears to be little growth

in employment for large firms on average. The mean growth rate for young firms is nearly twice

that for mature firms and more than three times that for older firms. On average, there is little

difference in growth rate between single, independent establishments and those that are part of

multi-establishment entities or between the manufacturing, sales, and services sectors. Firms in

Africa and Latin America grow faster than those in Europe and Central Asia and East Asia and

Pacific.

4. The Most Binding Constraint of the Business Environment

Having the managers of firms rate constraints on the firms‘ operation and growth is a useful start

for identifying important obstacles in the business environment. We analyze these obstacles not

only by using econometric tools but also by examining the importance of these obstacles across

regions and sectors.

4.1 Understanding Obstacles to Firms’ Operation

In the World Bank Enterprise Survey, as noted, firms rate 15 obstacles in their business

environment. These are access to finance, practices of competitors in the informal sector,

electricity, corruption, crime, inadequately educated workforce, labor regulations, business

13 Because of space limitations, abbreviations are sometimes used for the regions in tables: AFR for Sub-Saharan Africa, LAC for Latin America and the Caribbean, ECA for Europe and Central Asia, and EAP for East Asia and Pacific.

83

licensing and permits, political instability, tax administration, tax rates, transport, customs and

trade regulations, courts, and access to land.

A review of firm responses shows that the biggest reported obstacles differ across regions and

countries (see appendix). Using model 2 (as explained in greater detail in the next section), we

find that different sectors also confront different obstacles. For example, in the manufacturing

sector access to finance, informal sector competition, tax rates, and labor regulations matter the

most, while in the sales and services sectors only access to finance and informal sector

competition are negatively and significantly correlated with firm growth (see table 6). Estimation

results for the same model show that each country faces its own set of significant obstacles.14

So

does each region (see table 5).

Many of these obstacles are linked directly or indirectly to poor firm performance. In an ideal

world a country would address all these problems in order to improve firm performance. But

governments in developing countries have limited financial and human resources and, as argued

by the growth diagnostics approach, should therefore prioritize reform efforts to remove the most

important constraints.

The top three obstacles to firms‘ operation emerging from the survey data for our sample are

electricity, access to finance, and tax rates (figure 1). But we do not know whether these are the

top obstacles to employment growth. We therefore need to analyze which obstacles have a

significant effect on employment growth.

14 The estimation results are available upon request.

1.02

3.06

4.08

10.2

15.31

17.35

23.47

25.51

0 5 10 15 20 25 30

Licenses & permits

Inadequately educated workforce

Crime, theft & disorder

Political instability

Informal sector competition

Tax rates

Access to finance

Electricity

Figure 1. Distribution of the Top Obstacle Cited by Enterprises, All Economies

Share of economies (%)

84

Source: World Bank Enterprise Surveys (2006–10).

4.2 Identifying the Most Binding Constraint on Firms’ Growth

With figure 1 as a starting point, we set up an econometric model to investigate which of the 15

constraints is the most binding. We define a constraint as the most binding if it is statistically

significant, has a large coefficient in all estimations (models), and has the right sign—that is, has

a negative effect on employment growth. We design three models:

Model 1:

EG = b0 + b1Individual Obstacle + b2Firm Characteristics + Country Fixed Effects + e1 (1)

Model 2:

EG = b0 + b1All 15 Obstacles + b2Firm Characteristics + Country Fixed Effects + e2 (2)

Model 3:

EG = b0 + b1Only Significant Obstacle (in Model 2) + b2Firm Characteristics

+ Country Fixed Effects + e3 (3)

where EG refers to the employment growth of firm i at time t; Individual Obstacle is each

obstacle among the 15 shown in the last 15 rows of table 1; Firm Characteristics include labor

size (the number of permanent employees at the beginning of period t−3), labor size squared,

age, age squared, and indicators of whether a firm is part of a multi-establishment entity (multi),

is in manufacturing (manuf), is an exporter, is foreign owned (foreign), and is government

owned (govt).

The results suggest that access to finance and competition from the informal sector are the most

binding constraints, with statistically significant effects in all models. Columns 1–15 in table 4,

presenting the estimation results for model 1 for each obstacle, show that only access to finance

and competition from the informal sector have a significant negative effect on employment

growth. Column 16 shows the estimation results for model 2, run for all 15 obstacles together,

and column 17 presents the results for model 3, which includes all significant obstacles. Once

again we find that access to finance and competition from the informal sector are the most

binding constraints. We also examine the significance of the effect of these obstacles on firm

growth across regions and sectors to check the robustness of the findings. Tables 5 and 6 confirm

that access to finance and competition from the informal sector matter the most after controlling

for firm characteristics.

Our results demonstrate that both econometrically and economically access to finance and

competition from the informal sector matter the most for firms‘ employment growth—findings

that are in line with the starting point of the rankings of reported obstacles shown in the

85

appendix. While statistically both constraints are equally binding, the meaning of the second

constraint is ambiguous. The survey asks firms whether they see competition from the informal

sector as an obstacle. To any individual firm, competition poses a threat to survival. Yet at the

level of the economy it is competition that drives firms to improve productivity and therefore

drives growth. So it is not clear to us that competition from the informal sector should be

considered an obstacle to firms‘ operation. Moreover, this survey question is not followed by

other questions on related aspects of competition, allowing too little information to assess the

importance of informal sector competition. Therefore we do not further address this issue in the

paper.

While perception-based indicators like those applied in the analysis discussed here are useful,

quantitative indicators may give a more accurate picture of the business environment. Firm

managers within a country may have different perceptions of the same obstacle, and firm

managers in different countries and regions have different frames of reference. A problem

perceived as a moderate obstacle by one firm may be perceived as a severe obstacle by another,

even though the problem imposes a smaller cost on the second firm.

In the next three sections we use objective measures to examine the importance of access to

finance. As discussed, we cannot analyze informal sector competition because of its ambiguity

and because the data do not provide sufficient information. We leave further analysis of this

constraint with objective measures for future work, when the data are available.

5. Impact of Financial Access Variables on Employment Growth

In this section we examine the effect of financial access variables on firm employment growth,

controlling for individual firm characteristics. The model is set up with the following

specification:

EG = b0 + b1Laborsize + b2Age + b3Multi + b4Manuf + b5Exporter + b6Foreign + b7Govt +

b8FC(s) + Country Fixed Effects + e (4)

where EG refers to the employment growth of firm i at time t (the growth in the number of

permanent employees between t−3 and t) and FC denotes each of the financial access

variables—loan, credit constraint, sales credit, and external finance.

Our specification accounts for heteroskedasticity and country fixed effects. All outliers have

been removed. We also emphasize the importance of ownership structure by varying the type of

establishment: single or multiple, foreign or government owned, exporter or nonexporter. The

negative relationship between firm growth and firm size shown in table 7—along with the

supportive evidence in table 1 showing that smaller firms grow faster than larger firms—

suggests that Gibrat‘s law does not hold in this sample of firms. This finding is true across

regions and sectors. The negative and statistically significant coefficient on firm age tells us that

there is an inverse relationship between firm growth and firm age, which is consistent with

Jovanovic‘s model (1982) of disproportionate growth.

86

With other firm characteristics held constant, the rate of growth is significantly lower for

independent, single-establishment firms and government-owned firms. Exporters and foreign-

owned firms tend to have greater employment growth. In Africa and East Asia firms in the

manufacturing sector have higher employment growth than firms in the sales and services

sectors.

On average, a 1 percent increase in beginning-of-period firm size is associated with a 0.93

percent increase in end-of-period size (after three years) when the beginning-of-period age is

held constant (based on the results in table 7, column 1). Based on the analysis across regions,

we get estimated elasticities of end-of-period size with respect to beginning-of-period size of

approximately 0.9 for Africa, East Asia and Pacific, Europe and Central Asia, and Latin America

and the Caribbean (table 7, columns 6–9). With beginning-of-period size held constant, a 1

percent increase in beginning-of-period firm age is associated with a 0.07 percent decrease in

end-of-period size (table 7, column 1).

The results in table 7 show that financial access variables have a significant effect on firm

growth. Columns 1–4 indicate that with other factors held constant, having a loan or overdraft

facility increases the growth in a firm‘s number of permanent employees by 3.1 percent; being

credit constrained reduces a firm‘s employment growth by 1.9 percent; having sales credit

increases firm‘s growth by 2.6 percent; and having external investment funds increases growth

by 4.2 percent. If we include all these significant financial access variables in one model after

controlling for firm characteristics, they still have significant effects on employment growth,

though the effects are of smaller magnitude. And if we use the same model and run the

regressions in different regions, the significance and signs of the effects remain the same across

regions. These strong results show that access to finance does indeed matter for firm growth.

6. Determinants of Financial Access

In this section we estimate the probability of a firm having access to finance based on its

characteristics. We use the following model:

FC = b0 + b1Small + b2Medium + b3Large + b4Mature + b5Older + b6Multi + b7Manuf +

b8Exporter + b9Foreign + b10Govt + e (5)

where FC denotes each of the financial access variables—loan, credit constraint, sales credit, and

external finance.

We estimate this model by probit. We focus on firms of different sizes (micro, small, medium-

size, and large, with micro as the base category that is omitted from the regression) and different

ages (young, mature, and older, with young as the base category that is omitted from the

regression). The results show that a firm‘s size, age, and status as an exporter are strong

determinants of its access to finance (table 8).

87

There is a bigger difference in access to finance between micro and large firms than between

small and large firms. With the analysis controlling for firm characteristics and using country

fixed effects, micro firms are more likely to be credit constrained. With other factors held

constant, large firms are 85 percent less likely to be credit constrained than micro firms. In

addition, large firms are 97 percent more likely to have a loan or overdraft facility and 75 percent

more likely to have a share of investment financed externally than micro firms. Medium-size and

large firms are about 32 and 43 percent more likely to have sales credit than micro firms, while

small firms are about 19 percent more likely to offer sales credit than micro firms.

Older firms are 29 percent more likely to have a loan, 8 percent more likely to have sales credit,

and 20 percent less likely to be credit constrained than young firms. Mature firms are only 6

percent less likely to be credit constrained and about 10 percent more likely to have a loan than

young firms, with other factors held constant.

Other interesting results also emerge. Firms in the manufacturing sector are 20 percent more

likely to be credit constrained. With other factors held constant, firms that are exporters are 41

percent more likely to have a loan, 26 percent less likely to be credit constrained, and 20 percent

more likely to have external finance for investment than non-exporters.

7. Effect of Financial Access on Employment Growth by Firm Size and Age

In this section we investigate the effect of financial access on employment growth by firm size

and firm age first for each of the financial access variables individually and then for all the

variables combined.

7.1 Effect of Individual Financial Access Variables

To examine the effect of the financial access variables individually, we use the following model:

EG = b0 + b1FC + b2Small*FC + b3Medium*FC + b4Large*FC + b5Mature*FC +

b6Older*FC + b7Multi + b8Manuf + b9Exporter + b10Foreign + b11Govt + Country Fixed

Effects + e (6)


permanent employees between t−3 and t) and FC denotes each of the financial access

variables—loan, credit constraint, sales credit, and external finance.

Table 9 shows the effect of each of the financial access variables—loan, credit constraint, sales

credit, and external finance—in turn on employment growth. Micro firms are again the base

category that is omitted from the regression.

Among size categories, micro firms appear to benefit the most from having access to finance.

Column 1 shows that having a loan increases employment growth by 9 percent in micro firms,

88

but by only 4 percent in medium-size firms and 2 percent in large firms, with other factors held

constant. The results in columns 3 and 4 tell the same story. Micro and small firms gain the most

from finance in forms ranging from simple to more sophisticated—from having a loan or

overdraft facility to sales credit to external finance for investment. Column 2 supports the

argument that micro and small firms benefit the most from having access to finance. Being credit

constrained will make larger firms suffer more than smaller firms.

Young firms expand more than older firms with access to the same forms of finance. Having a

loan or overdraft facility increases employment growth by 9 percent for young firms, but by 6

percent for mature firms and 3 percent for older firms, with other factors held constant.

Similarly, having sales credit or external finance increases growth more for young firms than for

mature and older firms. Being credit constrained reduces firm growth as firms age. This finding

emphasizes the importance of firm age to firm growth. The effect of being credit constrained also

varies by sector, appearing to be stronger in manufacturing than in the sales or services sector.

We also look at the effect of financial access on employment growth across regions. The

estimation results by region are presented in table 10 for each financial access variable at a time.

The finding that having a loan, sales credit, or a share of investment financed externally helps

micro firms the most still holds. Indeed, this finding holds for all regions. The finding that young

firms expand more than older firms with access to the same forms of finance also holds across

regions.

7.2 Effect of Combined Financial Access Variables

In this section we look at the effect of all four financial access variables combined on

employment growth, by firm size, age, sector, and region. We use the following model:

EG = b0 + b1Small + b2Medium + b3Large + b4Mature + b5Older + b6Multi + b7Manuf +

b8Exporter + b9Foreign + b10Govt + b11Loan + b12Credit Constraint + b13Sales Credit + b14

External Finance + Country Fixed Effects + e (7)


permanent employees between t−3 and t).

Table 11 shows the estimation results for equation 7. The effects of all the financial access

variables are statistically significant and have the right signs. The results in column 1 indicate

that firm growth slows both as a firm expands its labor force and as it ages, controlling for other

firm characteristics. Columns 2–5 suggest that having a loan and having external finance are

important for firms of all sizes, though the effects are largest for small firms. The effects for

medium-size and large firms are similar in size. The effect of being credit constrained increases

with firm size.

89

Columns 6–8 show that having a loan is very important for firms of all ages, with the largest

effect on young firms. Being credit constrained has the largest effect on medium-size firms

followed by large firms. The effects of having external finance are statistically significant for

firms of all ages. In other words, trust and external finance matter to firms, regardless of their

age.

Columns 9–11 show that most forms of financial access are important to firms, no matter their

sector. Having a loan has the largest effect on employment growth for firms in the manufacturing

sector. Being credit constrained has a negative effect on firms in all sectors, with the largest

effect in the manufacturing and sales sectors. Having external finance matters in all sectors, and

the effects are of similar magnitude.

Columns 12–15 present the estimation results across regions. Having a loan and external finance

matters to firms in different regions, with the largest effects in Latin America. Being credit

constrained has a significant effect on firm employment growth only in Europe and Central Asia

and Latin America. From these results, together with the results in tables 7–10, we find that the

interaction between financial access variables and firms‘ size or age is significant in explaining

firms‘ employment growth.

8. Firm Size Distribution

Because a firm‘s size plays a significant part in determining its employment growth, we further

assess the relationship between firm size and financial constraints. The survey data allow the

creation of a variable showing which firms are credit constrained, which we identify as those that

applied for a loan and were rejected or that were discouraged from applying for a loan.15

The

data also include extensive information on the sources of firms‘ investments in fixed assets.

These sources can be external (formal or informal) or internal.16

Confirming the findings of Cabral and Mata (2003), figure 2 shows that the firm size distribution

is skewed to the right and that the skewness tends to diminish with age. The size distribution of

older firms is more symmetric than that of young firms.

As shown in figure 3, the firm size distribution is skewed more to the right in the manufacturing

and services sectors, where micro and small firms make up about two-thirds of the sample.

Figure 4 suggests that Africa has the largest share of micro and small firms while other regions

have more medium-size and large firms. This again provides evidence of the ―missing middle‖ in

Africa.

15 Our measure of credit-constrained firms comprises those that applied for a loan and were rejected and those that did not apply for one or more of the following reasons: fear of rejection, collateral requirements too high, interest rates not favorable, or a belief that the application would not be approved. 16 Formal sources are private or public banks, nonbank financial institutions, issues of new debt, and suppliers’ credit. Informal sources are friends and moneylenders. Internal sources consist of issuances of new shares and own funds.

90

Using the indicator of credit constrained created for this analysis, we split the sample into two

groups, credit-constrained and non-credit-constrained firms. As figure 5 shows, the firm size

distribution is skewed to the right for credit-constrained firms. This result is in line with the

finding that being credit constrained has a negative effect on firm growth—and, especially, that

this effect is largest for small firms. Taking this analysis further, we investigate the firm size

distribution using the survey data on firms‘ perceptions of access to finance, splitting the sample

between those perceiving it as a major or very severe obstacle and those viewing it as a minor

obstacle or no obstacle. Figure 6 shows that the size distribution of firms perceiving access to

finance as a major or very severe obstacle is skewed to the right. This result is confirmed by the

data for our sample showing that most of the firms regarding access to finance as a major or very

severe obstacle are micro or small. The size distribution for firms perceiving access to finance as

a minor obstacle or no obstacle is more symmetric.

The findings in this and previous sections suggest that a low level of financial development

results in a skewed firm size distribution, with a larger relative share of small firms. Policies

favoring the development of the financial sector should therefore have an effect on the firm size

distribution and, ultimately, favor the adoption of different technologies and a better allocation of

resources.

9. Conclusion

Using a newly available data set from the World Bank Enterprise Surveys (2006–10) for 39,538

firms across 98 countries, we investigate the binding constraints on firms‘ employment growth.

With an econometric model and subjective measures, we find that access to finance and informal

sector competition are the most binding constraints—both globally and in each region. Using

objective measures and controlling for firm characteristics, we evaluate the importance of access

to finance for firms‘ employment growth. We find that access to different forms of finance

matters. These results from our cross-country firm-level analysis suggest that governments

seeking to improve the business environment and promote firm growth should make financial

sector reforms a priority.

Objective business conditions vary systematically across firms of different sizes and ages, and

good business conditions favor smaller firms, especially micro firms. Micro and small firms gain

the most from access to finance in forms ranging from simple to more sophisticated—from a

loan or overdraft facility to sales credit to external finance for investment. This finding holds not

only globally but also for different regions. While sales credit is important only for micro and

small firms, having a loan or overdraft facility and receiving external finance for investment

promote employment growth for firms of all sizes across regions. And sales credit and external

finance matter for firms of all ages.

The firm size distribution is skewed toward smaller firms. The skewness declines with firm

age—and is more present in Africa, among firms that are credit constrained, and among those

91

that perceive access to finance as a serious obstacle. These findings call for policies favoring the

development of the financial sector, which can help small firms grow into medium-size and large

firms.

The findings have several implications for developing countries. First, because the constraints

faced by firms differ across countries and, within countries, across sectors, policies to promote

firm growth need to be tailored to each country and sector. Second, finance appears to be the

most binding constraint across sectors and countries, suggesting that reforms in this sector could

yield broad benefits—including by helping to address the problem of the ―missing middle‖ in

developing countries. Third, reforms in finance take time, and a quicker development strategy

could be to identify the binding constraints in a specific subsector and try to address them

through direct policy measures.

92

I References

Angelini, Paolo, and Andrea Generale. 2008. ―On the Evolution of Firm Size Distributions.‖

American Economic Review 98 (1): 426–38.

Aterido, Reyes, and Mary Hallward-Driemeier. 2010. ―The Impact of the Investment Climate on

Employment Growth: Does Sub-Saharan Africa Mirror Other Low-Income Regions?‖

Policy Research Working Paper 5218, World Bank, Washington, DC.

Aterido, Reyes, Mary Hallward-Driemeier, and Carmen Pagés. 2007. ―Investment Climate and

Employment Growth: The Impact of Access to Finance, Corruption and Regulations

across Firms.‖ RES Working Paper 4559, Research Department, Inter-American

Development Bank, Washington, DC.

———. 2009. ―Big Constraints to Small Firms‘ Growth? Business Environment and

Employment Growth across Firms.‖ Policy Research Working Paper 5032, World Bank,

Washington, DC.

Ayyagari, Meghana, Asli Demirgüç-Kunt, and Vojislav Maksimovic. 2006. ―How Important Are

Financing Constraints? The Role of Finance in the Business Environment.‖ Policy

Research Working Paper 3820, World Bank, Washington, DC.

Batra, Geeta, Daniel Kaufmann, and Andrew H. W. Stone. 2003. Investment Climate around the

World: Voices of the Firms from the World Business Environment Survey. Washington,

DC: World Bank.

Beck, Thorsten, Asli Demirgüç-Kunt, and Vojislav Maksimovic. 2005. ―Financial and Legal

Constraints to Growth: Does Firm Size Matter?‖ Journal of Finance 60 (1): 131–77.

Bigsten, Arne, and Mans Söderbom. 2006. ―What Have We Learned from a Decade of

Manufacturing Enterprise Surveys in Africa?‖ World Bank Research Observer 21 (2):

241–65.

Cabral, Luis M. B., and Jose Mata. 2003. ―On the Evolution of the Firm Size Distribution: Facts

and Theory.‖ American Economic Review 93 (4): 1075–90.

Cooley, Thomas F., and Vincenzo Quadrini. 2001. ―Financial Markets and Firm Dynamics.‖

American Economic Review 91 (5): 1286–310.

Demirgüç-Kunt, Asli, and Vojislav Maksimovic. 1998. ―Law, Finance and Firm Growth.‖

Journal of Finance 53: 2107–37.

93

Desai, Mihir, Paul Gompers, and Josh Lerner. 2003. ―Institutions, Capital Constraints and

Entrepreneurial Firm Dynamics: Evidence from Europe.‖ NBER Working Paper 10165,

National Bureau of Economic Research, Cambridge, MA.

Djankov, Simeon, Rafael La Porta, Florencio López-de-Silanes, and Andrei Shleifer. 2002. ―The

Regulation of Entry.‖ Quarterly Journal of Economics 117 (1): 1–37.

Dollar, David, Mary Hallward-Driemeier, and Taye Mengistae. 2005. ―Investment Climate and

Firm Performance in Developing Countries.‖ Economic Development and Cultural

Change 54 (1): 1–31.

Dunne, Timothy, Mark J. Roberts, and Larry Samuelson. 1989. ―The Growth and Failure of U.S.

Manufacturing Plants.‖ Quarterly Journal of Economics 104 (4): 671–98.

Evans, David S. 1987a. ―The Relationship between Firm Growth, Size and Age: Estimates for

100 Manufacturing Industries.‖ Journal of Industrial Economics 35: 567–81.

———. 1987b. ―Tests of Alternative Theories of Firm Growth.‖ Journal of Political Economy

95: 657–74.

Fisman, Raymond, and Jakob Svensson. 2007. ―Are Corruption and Taxation Really Harmful to

Growth? Firm-Level Evidence.‖ Journal of Development Economics 83 (1): 63–75.

Freund, Caroline, and Nadia Rocha. 2010. ―What Constrains Africa‘s Exports?‖ Policy Research

Working Paper 5184, World Bank, Washington, DC.

Galindo, Arturo José, and Alejandro Micco. 2007. ―Creditor Protection and Credit Response to

Shocks.‖ World Bank Economic Review 21 (3): 413–38.

Gelb, Alan, Vijaya Ramachandran, Manju Kedia Shah, and Ginger Turner. 2007. ―What Matters

to African Firms? The Relevance of Perceptions Data.‖ Policy Research Working Paper

4446, World Bank, Washington, DC.

Hall, Bronwyn H. 1987. ―The Relationship between Firm Size and Firm Growth in the U.S.

Manufacturing Sector.‖ Journal of Industrial Economics 35: 583–606.

Hausmann, Ricardo, Dani Rodrik, and Andrés Velasco. 2005. ―Growth Diagnostics.‖ John F.

Kennedy School of Government, Harvard University, Cambridge, MA.

IDB (Inter-American Development Bank). 2007. Living with Debt: How to Limit the Risks of

Sovereign Finance. Economic and Social Progress Report. Washington, DC: IDB.

Jovanovic, Boyan. 1982. ―Selection and Evolution of Industry.‖ Econometrica 50: 649–70.

94

Klapper, Leora, Luc Laeven, and Raghuram Rajan. 2004. ―Business Environment and Firm

Entry: Evidence from International Data.‖ NBER Working Paper 10380, National Bureau

of Economic Research, Cambridge, MA.

Lipsey, R. G., and Kelvin Lancaster. 1956. ―The General Theory of Second Best.‖ Review of

Economic Studies 24 (1): 11–32.

Love, Inessa, and Nataliya Mylenko. 2003. ―Credit Reporting and Financing Constraints.‖ Policy

Research Working Paper 3142, World Bank, Washington, DC.

Rajan, Raghuram, and Luigi Zingales. 1998. ―Financial Dependence and Growth.‖ American

Economic Review 88 (3): 559–86.

Reinikka, Ritva, and Jakob Svensson. 2002. ―Coping with Poor Public Capital.‖ Journal of

Development Economics 69 (1): 51–69.

World Bank. 2009. Africa Competitiveness Report 2009. Washington, DC: World Bank.

95

Variable Description Mean SDEmployment growth Employment growth [(lnS it - lnS i,t-3)/3] 0.052 0.127Laborsize Number of permanent employees [lnS i,t-3] 3.112 1.350Age Years of firm's operation 2.602 0.741Multi Equal to 1 if firm is independent, single establishment; 0 otherwise 0.138 0.345Manuf Equal to 1 if firm is in manufacturing or construction sector; 0 otherwise 0.555 0.497Exporter Equal to 1 if direct exports account for more than 10 percent of firm's sales; 0 otherwise? 0.130 0.336Foreign Equal to 1 if firm has 10 percent or more of foreign ownership; 0 otherwise 0.117 0.321Govt Equal to 1 if firm has 10 percent or more of government ownership; 0 otherwise 0.017 0.129Loan Equal to 1 if firm has loan, line of credit, or overdraft facility; 0 otherwise 0.573 0.495Credit constraint Equal to 1 if firm did not apply for loan for some reason; 0 otherwise 0.334 0.472Sales credit Equal to 1 if firm has positive sales paid for after delivery; 0 otherwise 0.702 0.458External finance Equal to 1 if firm has a positive amount of external funds; 0 otherwise 0.237 0.425Access to finance How much of an obstacle to firm's operation is access to finance? 1.725 1.564Informal competition How much of an obstacle to firm's operation are informal sector competitors? 1.627 1.453Labor regulations How much of an obstacle to firm's operation are labor regulations? 0.958 1.181Inadequate education How much of an obstacle to firm's operation is an inadequately educated workforce? 1.408 1.353Electricity How much of an obstacle to firm's operation is electricity? 1.843 1.526Transport How much of an obstacle to firm's operation is transport of goods, supplies, and inputs? 1.224 1.310Customs and trade How much of an obstacle to firm's operation are customs and trade regulations? 0.954 1.242Access to land How much of an obstacle to firm's operation is access to land? 1.031 1.334Courts How much of an obstacle to firm's operation are courts? 1.025 1.280Crime How much of an obstacle to firm's operation are crime, theft, and disorder? 1.423 1.382Tax rates How much of an obstacle to firm's operation are tax rates? 1.828 1.374Tax administration How much of an obstacle to firm's operation is tax administration? 1.439 1.319Licensing and permits How much of an obstacle to firm's operation are business licensing and permits? 1.095 1.238Political instability How much of an obstacle to firm's operation is political instability? 1.615 1.504Corruption How much of an obstacle to firm's operation is corruption? 1.780 1.530Source: World Bank Enterprise Surveys (2006-10).

Table 1. Variable Descriptions and Summary Statistics

96

Characteristic Frequency Percent Cumulative percentBy size

Micro 15,357 38.84 38.84 Small 14,791 37.41 76.25 Medium 6,499 16.44 92.69 Large 2,845 7.2 99.88 Unknown 46 0.12 100By age

Young 4,440 11.23 11.23 Mature 18,551 46.92 58.15 Older 16,146 40.84 98.99 Unknown 401 1.01 100By establishment number

Multi-establishment 5,397 13.65 13.65 Single establishment 33,729 85.31 98.96 Unknown 412 1.04 100By sector

Manufacturing 21,783 55.09 55.09 Sales 8,901 22.51 77.6 Services 7,845 19.84 97.44 Unknown 1,009 2.55 100By trade orientation

Nonexporter 34,405 87.02 87.02 Exporter 5,133 12.98 100By foreign ownership

Domestically owned 34,587 87.48 87.48 Foreign owned 4,579 11.58 99.06 Unknown 372 0.94 100By government ownership

Government owned 37,858 95.75 95.75 Privately owned 649 1.64 97.39 Unknown 1,031 2.61 100Total establishments 39,538Source: World Bank Enterprise Surveys (2006-10).

Table 2. Firm Characteristics by Different Groups of Controls

97

Variable 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17Access to finance -0.002*** -0.004*** -0.002***

(0.000) (0.001) (0.000)Informal sector competition -0.003*** -0.003*** -0.003***

(0.000) (0.001) (0.001)Inadequate education 0.005*** 0.007*** 0.007***

(0.001) (0.001) (0.001)Electricity 0.002*** 0.001*

(0.000) (0.001)Customs and trade 0.005*** 0.005***

(0.001) (0.001)Access to land 0.003*** 0.003***

(0.001) (0.001)Political instability -0.001 -0.002**

(0.001) (0.001)Courts -0.000 -0.001

(0.001) (0.001)Crime 0.001** 0.000

(0.001) (0.001)Tax rates -0.001* -0.001*

(0.001) (0.001)Tax administration -0.000 -0.000

(0.001) (0.001)Licensing and permits 0.001 0.001

(0.001) (0.001)Corruption -0.000 -0.001*

(0.000) (0.001)Transport 0.002*** 0.000

(0.001) (0.001)Labor regulations 0.002*** -0.001

(0.001) (0.001)Adjusted R -squared 0.123 0.122 0.124 0.121 0.125 0.122 0.121 0.117 0.121 0.122 0.122 0.122 0.122 0.121 0.121 0.130 0.127Number of observations 35,837 35,466 36,216 36,554 32,967 35,399 35,814 32,794 36,278 36,287 36,154 35,350 35,435 36,222 36,297 26,574 34,359Number of countries 96 96 96 96 96 96 96 95 96 96 96 96 96 96 96 95 96

Note: Standard errors (in parentheses) are robust to heteroskedasticity and clustered on countries. Model 1 (col. 1-15): EG = b 0 + b 1Individual Obstacle + b 2Firm Characteristics + Country Fixed Effects + e1.

* Significant at the 10 percent level. ** Significant at the 5 percent level. *** Significant at the 1 percent level. Source: Authors' estimations based on data from World Bank Enterprise Surveys (2006-10).

Model 2 (col. 16): EG = b 0 + b 1All 15 Obstacles + b 2Firm Characteristics + Country Fixed Effects + e2. The hypothesis that the coefficients for access to finance and informal sector competition differ is tested and rejected.

Dependent variable: employment growthTable 4. Effect of Business Environment Obstacles on Employment Growth

98

World AFR EAP ECA LAC

Variable (1) (2) (3) (4) (5)

Laborsize -0.022*** -0.026*** -0.029*** -0.018*** -0.019***(0.001) (0.001) (0.002) (0.001) (0.001)

Age -0.023*** -0.021*** -0.019*** -0.031*** -0.024***(0.001) (0.002) (0.004) (0.003) (0.002)

Multi 0.018*** 0.012*** 0.010 0.020*** 0.022***(0.002) (0.003) (0.007) (0.006) (0.004)

Manuf 0.003 0.012*** 0.012** -0.009** 0.005(0.002) (0.003) (0.006) (0.004) (0.003)

Exporter 0.020*** 0.020*** 0.018** 0.020*** 0.025***(0.002) (0.004) (0.007) (0.004) (0.004)

Foreign 0.011*** 0.008** 0.009 0.021*** 0.011**(0.002) (0.003) (0.006) (0.006) (0.005)

Govt -0.017*** -0.013 0.017 -0.022*** -0.020(0.005) (0.011) (0.012) (0.008) (0.017)

Access to finance -0.004*** -0.002* -0.008*** -0.004*** -0.005***(0.001) (0.001) (0.002) (0.001) (0.001)

Informal sector competition -0.003*** -0.003*** -0.008*** -0.002* -0.003***(0.001) (0.001) (0.002) (0.001) (0.001)

Inadequate education 0.007*** 0.005*** 0.008*** 0.006*** 0.010***(0.001) (0.001) (0.002) (0.001) (0.001)

Electricity 0.001* -0.001 0.004** 0.001 0.002**(0.001) (0.001) (0.002) (0.001) (0.001)

Customs and trade 0.005*** 0.004*** 0.006** 0.006*** 0.003*(0.001) (0.001) (0.003) (0.002) (0.001)

Access to land 0.003*** -0.000 0.004* 0.004*** 0.005***(0.001) (0.001) (0.002) (0.001) (0.001)

Political instability -0.002** -0.001 -0.002 -0.001 -0.003**(0.001) (0.001) (0.003) (0.001) (0.002)

Courts -0.001 -0.001 0.001 -0.004** 0.000(0.001) (0.001) (0.003) (0.002) (0.001)

Crime 0.000 0.001 0.002 -0.000 -0.001(0.001) (0.001) (0.003) (0.001) (0.001)

Tax rates -0.001* -0.002 -0.000 -0.000 -0.002(0.001) (0.001) (0.003) (0.002) (0.002)

Tax administration -0.000 0.002* -0.002 -0.002 -0.002(0.001) (0.001) (0.003) (0.002) (0.002)

Licensing and permits 0.001 -0.001 0.003 0.002 0.002(0.001) (0.001) (0.003) (0.002) (0.002)

Corruption -0.001* -0.001 -0.002 0.000 -0.002(0.001) (0.001) (0.003) (0.002) (0.001)

Transport 0.000 -0.002 0.002 0.001 0.001(0.001) (0.001) (0.002) (0.001) (0.001)

Labor regulations -0.001 -0.001 -0.001 -0.002 -0.001(0.001) (0.001) (0.003) (0.002) (0.002)

Constant 0.176*** 0.182*** 0.156*** 0.184*** 0.174***(0.004) (0.006) (0.011) (0.008) (0.008)

Adjusted R -squared 0.130 0.129 0.148 0.130 0.112Number of observations 26,574 8,600 3,079 6,596 7,592Number of countries 95 37 10 30 15

Model: EG = b 0 + b 1All 15 Obstacles + b 2Firm Characteristics + Region + Country Fixed Effects + e.

Source: Authors' estimations based on World Bank Enterprise Surveys (2006-10).

Dependent variable: employment growthTable 5. Effect of Business Environment Obstacles on Employment Growth by Region

Note: Standard errors (in parentheses) are robust to heteroskedasticity and clustered on countries. Regressions for the Middle East and North Africa and South Asia are excluded because of insufficient data.

* Significant at the 10 percent level. ** Significant at the 5 percent level. *** Significant at the 1 percent level.

99

Table 6. Effect of Business Environment Obstacles on Employment Growth by Sector

Manufacturing Sales Services

Variable (1) (2) (3)

Laborsize -0.026*** -0.014*** -0.021***(0.001) (0.002) (0.002)

Age -0.024*** -0.025*** -0.018***(0.002) (0.002) (0.003)

Multi 0.022*** 0.014*** 0.014***(0.003) (0.004) (0.005)

Exporter 0.026*** 0.017** 0.010(0.003) (0.007) (0.007)

Foreign 0.011*** 0.009* 0.012**(0.003) (0.005) (0.005)

Govt -0.009 -0.036** -0.020*(0.007) (0.015) (0.011)

Access to finance -0.004*** -0.002* -0.004***(0.001) (0.001) (0.002)

Informal sector competition -0.004*** -0.002* -0.004***(0.001) (0.001) (0.001)

Inadequate education 0.008*** 0.007*** 0.003**(0.001) (0.001) (0.002)

Electricity 0.002*** 0.001 -0.002(0.001) (0.001) (0.001)

Customs and trade 0.004*** 0.006*** 0.007***(0.001) (0.001) (0.002)

Access to land 0.003*** 0.003*** 0.001(0.001) (0.001) (0.002)

Political instability -0.002 -0.001 -0.002(0.001) (0.001) (0.002)

Courts -0.001 -0.001 -0.002(0.001) (0.002) (0.002)

Crime 0.001 -0.001 0.001(0.001) (0.001) (0.002)

Tax rates -0.002** 0.001 -0.001(0.001) (0.002) (0.002)

Tax administration 0.001 -0.002 -0.001(0.001) (0.002) (0.002)

Licensing and permits 0.000 0.000 0.002(0.001) (0.002) (0.002)

Corruption -0.001 -0.001 -0.001(0.001) (0.001) (0.002)

Transport 0.000 0.001 -0.001(0.001) (0.001) (0.002)

Labor regulations -0.002* -0.003 0.004**(0.001) (0.002) (0.002)

Constant 0.191*** 0.151*** 0.167***(0.005) (0.007) (0.009)

Adjusted R -squared 0.146 0.118 0.114Number of observations 15,322 6,014 5,237Number of countries 95 95 95Note: Standard errors (in parentheses) are robust to heteroskedasticity and clustered on countries. Model: EG = b 0 + b 1All 15 Obstacles + b 2Firm Characteristics + Region + Country Fixed Effects + e . * Significant at the 10 percent level. ** Significant at the 5 percent level. *** Significant at the 1 percent level.Source: Authors' estimations based on data from World Bank Enterprise Surveys (2006-10).

Dependent variable: employment growth

100

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Variable World World World World World AFR EAP ECA LAC

Laborsize -0.024*** -0.023*** -0.022*** -0.024*** -0.026*** -0.029*** -0.032*** -0.022*** -0.023***(0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.002) (0.001) (0.001)

Age -0.024*** -0.023*** -0.023*** -0.022*** -0.023*** -0.020*** -0.018*** -0.030*** -0.023***(0.001) (0.001) (0.001) (0.001) (0.001) (0.002) (0.003) (0.002) (0.002)

Multi 0.019*** 0.018*** 0.019*** 0.019*** 0.019*** 0.013*** 0.017*** 0.020*** 0.022***(0.002) (0.002) (0.002) (0.002) (0.002) (0.003) (0.006) (0.005) (0.004)

Manuf 0.0003 0.001 -0.000 -0.001 -0.000 0.010*** 0.010** -0.010*** -0.002(0.001) (0.001) (0.001) (0.001) (0.001) (0.002) (0.005) (0.003) (0.003)

Exporter 0.022*** 0.024*** 0.024*** 0.023*** 0.021*** 0.022*** 0.023*** 0.015*** 0.028***(0.002) (0.002) (0.002) (0.002) (0.002) (0.004) (0.007) (0.004) (0.004)

Foreign 0.013*** 0.012*** 0.013*** 0.014*** 0.014*** 0.008** 0.019*** 0.020*** 0.016***(0.002) (0.002) (0.002) (0.002) (0.002) (0.003) (0.006) (0.005) (0.005)

Govt -0.009** -0.011** -0.011** -0.010** -0.008* -0.008 0.024** -0.015** -0.016(0.005) (0.005) (0.005) (0.005) (0.005) (0.010) (0.011) (0.007) (0.020)

Loan 0.031*** 0.020*** 0.014*** 0.018*** 0.023*** 0.026***(0.002) (0.002) (0.003) (0.004) (0.003) (0.003)

Credit constraint -0.019*** -0.010*** -0.004* 0.001 -0.024*** -0.012***(0.002) (0.002) (0.002) (0.004) (0.003) (0.003)

Sales credit 0.026*** 0.009*** 0.010*** 0.006 0.012*** 0.005(0.003) (0.002) (0.002) (0.005) (0.003) (0.003)

External finance 0.042*** 0.036*** 0.021*** 0.041*** 0.036*** 0.041***(0.002) (0.002) (0.003) (0.005) (0.003) (0.003)

Constant 0.163*** 0.182*** 0.166*** 0.162*** 0.174*** 0.144*** 0.165*** 0.153***(0.003) (0.003) (0.003) (0.003) (0.005) (0.009) (0.006) (0.006)

Number of observations 34,894 35,641 36,722 36,722 34,524 10,270 3,971 9,423 9,911Adjusted R -squared 0.131 0.125 0.123 0.138 0.146 0.144 0.155 0.154 0.128

Source: Authors' estimations based on data from World Bank Enterprise Surveys (2006-10).

Table 7. Effect of Objective Financial Access Variables on Employment Growth

Note: Standard errors (in parentheses) are robust to heteroskedasticity and clustered on countries. Regressions for the Middle East and North Africa and South Asia are excluded because of insufficient data.Model: EG = b 0 + b 1Laborsize + b 2Age + b3Multi + b 4Manuf + b 5Exporter + b 6Foreign + b 7Govt + b 8FC(s) + Country Fixed Effects + e. ; * Significant at the 10 percent level. ** Significant at the 5 percent level. *** Significant at the 1 percent

level.


101

(1) (2) (3) (4)

Variable Loan Credit constraint Sales credit External finance

Small 0.467*** -0.412*** 0.187*** 0.338***(0.041) (0.043) (0.035) (0.030)

Medium 0.779*** -0.721*** 0.320*** 0.572***(0.063) (0.068) (0.049) (0.038)

Large 0.973*** -0.852*** 0.427*** 0.747***(0.072) (0.073) (0.078) (0.045)

Mature 0.097** -0.063* 0.055 0.023(0.040) (0.035) (0.055) (0.037)

Older 0.287*** -0.200*** 0.079 0.034(0.066) (0.057) (0.063) (0.048)

Multi 0.032 -0.128*** 0.023 0.007(0.051) (0.036) (0.047) (0.048)

Manuf -0.089 0.196*** -0.031 -0.015(0.062) (0.048) (0.058) (0.048)

Exporter 0.412*** -0.261*** 0.084* 0.204***(0.064) (0.053) (0.049) (0.044)

Foreign -0.101 -0.024 -0.026 -0.157***(0.070) (0.049) (0.050) (0.043)

Govt -0.361*** 0.199** 0.042 -0.118(0.119) (0.082) (0.099) (0.094)

Constant -0.325*** -0.071 0.070 -1.039***(0.104) (0.091) (0.063) (0.070)

Number of observations 34,916 35,663 36,746 36,746

* Significant at the 10 percent level. ** Significant at the 5 percent level. *** Significant at the 1 percent level.Source: Authors' estimations based on data from World Bank Enterprise Surveys (2006-10).

Table 8. Objective Financial Access Variables by Firm Characteristic

Note: Standard errors (in parentheses) are robust to heteroskedasticity and clustered on countries.

Model: FC = b 0 + b 1Small + b 2Medium + b 3Large + b 4Mature + b 5Older + b 6Multi + b 7Manuf + b 8Exporter + b 9Foreign + b 10Govt + e .

102

Table 9. Differences in Effects of Objective Financial Access Variables on Employment Growth

by Firm Size and Age


(1) (2) (3) (4)

Loan

Credit

constraint Sales credit External finance

Financial access variable (FC) 0.091*** 0.053*** 0.106*** 0.105***

(0.004) (0.004) (0.012) (0.006)

Small*FC -0.040*** -0.057*** -0.043*** -0.039***

(0.002) (0.002) (0.007) (0.004)

Medium*FC -0.051*** -0.065*** -0.057*** -0.056***

(0.003) (0.004) (0.008) (0.004)

Large*FC -0.069*** -0.078*** -0.066*** -0.073***

(0.003) (0.007) (0.009) (0.005)

Mature*FC -0.035*** -0.026*** -0.042*** -0.030***

(0.004) (0.004) (0.012) (0.006)

Older*FC -0.061*** -0.044*** -0.062*** -0.051***

(0.004) (0.004) (0.012) (0.006)

Multi 0.008*** 0.006*** 0.003 0.004**

(0.002) (0.002) (0.002) (0.002)

Manuf -0.005*** -0.006*** -0.009*** -0.009***

(0.001) (0.001) (0.001) (0.001)

Exporter 0.009*** 0.003 0.001 0.004*

(0.002) (0.002) (0.002) (0.002)

Foreign 0.003 0.003 0.000 0.001

(0.002) (0.002) (0.002) (0.002)

Govt -0.035*** -0.041*** -0.046*** -0.042***

(0.005) (0.005) (0.005) (0.005)

Constant 0.045*** 0.055*** 0.055*** 0.048***

(0.001) (0.001) (0.001) (0.001)

Adjusted R-squared 0.090 0.065 0.095 0.078

Number of observations 34,894 35,641 36,722 36,722

Number of countries 95 96 96 96

Note: Standard errors (in parentheses) are robust to heteroskedasticity and clustered on countries.

Model: EG = b0 + b1FC + b2Small*FC + b3Medium*FC + b4Large*FC + b5Mature*FC + b6Older*FC +

b7Multi + b8Manuf + b9Exporter + b10Foreign + b11Govt + Country Fixed Effects + e.



103

AFR EAP ECA LAC AFR EAP ECA LAC AFR EAP ECA LAC AFR EAP ECA LAC

Financial access (FC ) 0.063*** 0.061*** 0.108*** 0.123*** 0.044*** 0.090*** 0.044*** 0.071*** 0.056*** 0.077*** 0.109*** 0.140*** 0.070*** 0.089*** 0.108*** 0.136***(0.007) (0.011) (0.008) (0.008) (0.004) (0.012) (0.010) (0.014) (0.018) (0.034) (0.020) (0.021) (0.010) (0.017) (0.012) (0.011)

Small*FC -0.026*** -0.033*** -0.035*** -0.054*** -0.049*** -0.076*** -0.043*** -0.066*** -0.019*** -0.051*** -0.031*** -0.069*** -0.027*** -0.018 -0.033*** -0.057***(0.004) (0.007) (0.004) (0.004) (0.003) (0.008) (0.006) (0.006) (0.012) (0.031) (0.013) (0.012) (0.007) (0.014) (0.006) (0.006)

Medium*FC -0.036*** -0.045*** -0.057*** -0.057*** -0.049*** -0.082*** -0.067*** -0.064*** -0.026*** -0.057*** -0.051*** -0.086*** -0.035*** -0.039*** -0.058*** -0.071***(0.005) (0.008) (0.005) (0.005) (0.006) (0.011) (0.008) (0.010) (0.014) (0.037) (0.014) (0.015) (0.007) (0.014) (0.007) (0.007)

Large*FC -0.047*** -0.069*** -0.084*** -0.069*** -0.051** -0.111*** -0.070*** -0.065*** -0.079*** -0.024*** -0.067*** -0.087*** -0.057*** -0.061*** -0.080*** -0.083***(0.008) (0.009) (0.006) (0.006) (0.021) (0.014) (0.011) (0.016) (0.023) (0.046) (0.015) (0.016) (0.011) (0.017) (0.008) (0.008)

Mature*FC -0.039*** -0.018 -0.035*** -0.043*** -0.018*** -0.040*** -0.037*** -0.035** -0.033*** 0.012** -0.043*** -0.049*** -0.037*** -0.025 -0.024** -0.038***(0.007) (0.011) (0.008) (0.008) (0.004) (0.013) (0.010) (0.014) (0.020) (0.033) (0.019) (0.022) (0.010) (0.018) (0.012) (0.011)

Older*FC -0.062*** -0.037*** -0.059*** -0.076*** -0.037*** -0.052*** -0.058*** -0.063*** -0.036*** -0.036*** -0.069*** -0.073*** -0.048*** -0.055*** -0.049*** -0.063***(0.007) (0.011) (0.008) (0.008) (0.005) (0.012) (0.011) (0.014) (0.021) (0.031) (0.020) (0.022) (0.011) (0.018) (0.012) (0.011)

Multi 0.001 0.001 0.014*** 0.012*** 0.002 -0.001 0.008* 0.006 -0.003 -0.006 0.008*** 0.006*** -0.001 -0.005 0.010** 0.007*(0.003) (0.006) (0.005) (0.004) (0.003) (0.006) (0.005) (0.004) (0.003) (0.006) (0.004) (0.004) (0.003) (0.006) (0.004) (0.004)

Manuf 0.00002 0.001 -0.014*** -0.002 0.001 -0.001 -0.015*** -0.006** -0.003 -0.004 -0.017*** -0.007 -0.003 -0.004 -0.017*** -0.006**(0.002) (0.005) (0.003) (0.003) (0.002) (0.005) (0.003) (0.003) (0.002) (0.004) (0.003) (0.003) (0.002) (0.004) (0.003) (0.003)

Exporter 0.004 -0.007 0.011*** 0.020*** -0.002 -0.008 0.003 0.012*** -0.005* -0.014 0.004*** 0.012*** -0.003 -0.011* 0.006 0.014***(0.004) (0.006) (0.004) (0.004) (0.004) (0.006) (0.004) (0.004) (0.004) (0.006) (0.004) (0.004) (0.004) (0.006) (0.004) (0.004)

Foreign -0.003 0.003 0.017*** 0.004 -0.003 0.006 0.012** 0.004 -0.008 0.002 0.011*** 0.004 -0.007** 0.002 0.013*** 0.006(0.003) (0.006) (0.005) (0.005) (0.003) (0.006) (0.005) (0.005) (0.003) (0.005) (0.005) (0.005) (0.003) (0.005) (0.005) (0.005)

Govt -0.036*** -0.008 -0.041*** -0.031 -0.041*** -0.017 -0.048*** -0.036* -0.048*** -0.026 -0.051*** -0.036* -0.044*** -0.019* -0.047*** -0.032(0.010) (0.011) (0.006) (0.020) (0.009) (0.011) (0.006) (0.021) (0.009) (0.011) (0.006) (0.020) (0.009) (0.011) (0.006) (0.020)

Constant 0.068*** 0.024*** 0.028*** 0.033*** 0.063*** 0.025*** 0.056*** 0.058*** 0.070*** 0.030*** 0.049*** 0.053*** 0.068*** 0.026*** 0.038*** 0.042***(0.002) (0.004) (0.003) (0.004) (0.002) (0.004) (0.002) (0.003) (0.002) (0.003) (0.002) (0.002) (0.002) (0.003) (0.002) (0.003)

Adjusted R -squared 0.069 0.060 0.099 0.097 0.062 0.062 0.070 0.045 0.083 0.086 0.090 0.092 0.055 0.056 0.082 0.079Number of observations 10,358 4,100 9,527 9,949 10,636 4,076 9,907 10,067 10,878 4,431 10,237 10,203 10,878 4,431 10,237 10,203Number of countries 37 10 30 15 38 10 30 15 38 10 30 15 38 10 30 15Note: Standard errors (in parentheses) are robust to heteroskedasticity and clustered on countries. Regressions for the Middle East and North Africa and South Asia are excluded because of insufficient data.

Model: EG = b 0 + b 1FC + b 2Small*FC + b 3Medium*FC + b 4Large*FC + b 5Mature*FC + b 6Older*FC + b 7Multi + b 8Manuf + b 9Exporter + b 10Foreign + b 11Govt + Region + Country Fixed Effects + e .



Table 10. Differences in Effects of Objective Financial Access Variables on Employment Growth by Firm Size and Age across Regions

Credit constraintLoan Sales credit External finance


104

Micro Small Medium Large Young Mature OlderManufactu

ringSales Services AFR EAP ECA LAC

Variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15)

Small -0.052*** -0.082*** -0.051*** -0.044*** -0.060*** -0.037*** -0.056*** -0.046*** -0.058*** -0.042*** -0.061***(0.002) (0.006) (0.002) (0.002) (0.002) (0.003) (0.004) (0.002) (0.005) (0.003) (0.003)

Medium -0.067*** -0.113*** -0.071*** -0.054*** -0.080*** -0.041*** -0.065*** -0.056*** -0.080*** -0.065*** -0.067***(0.002) (0.010) (0.003) (0.003) (0.003) (0.005) (0.005) (0.004) (0.006) (0.004) (0.004)

Large -0.091*** -0.137*** -0.089*** -0.080*** -0.110*** -0.032*** -0.082*** -0.078*** -0.107*** -0.093*** -0.085***(0.003) (0.019) (0.005) (0.004) (0.004) (0.008) (0.008) (0.007) (0.008) (0.005) (0.006)

Mature -0.030*** -0.036*** -0.016*** -0.014 -0.017 -0.038*** -0.025*** -0.019*** -0.024*** -0.021*** -0.041*** -0.037***(0.003) (0.003) (0.005) (0.009) (0.018) (0.004) (0.004) (0.005) (0.003) (0.007) (0.006) (0.006)

Older -0.053*** -0.064*** -0.038*** -0.032*** -0.043** -0.060*** -0.047*** -0.040*** -0.046*** -0.036*** -0.065*** -0.065***(0.003) (0.003) (0.005) (0.009) (0.018) (0.004) (0.005) (0.006) (0.004) (0.007) (0.006) (0.006)

Multi 0.014*** 0.013*** 0.017*** 0.012*** 0.010* 0.009 0.015*** 0.014*** 0.015*** 0.012*** 0.014*** 0.008*** 0.009 0.018*** 0.015***(0.002) (0.004) (0.003) (0.004) (0.006) (0.008) (0.003) (0.003) (0.003) (0.004) (0.005) (0.003) (0.006) (0.005) (0.004)

Manuf -0.001 0.011*** -0.002 -0.018*** -0.032*** 0.022*** -0.001 -0.007*** 0.007*** 0.009* -0.011*** -0.001(0.001) (0.002) (0.002) (0.004) (0.006) (0.005) (0.002) (0.002) (0.002) (0.005) (0.003) (0.003)

Exporter 0.014*** 0.028*** 0.024*** 0.009** -0.001 0.035*** 0.011*** 0.015*** 0.020*** 0.017** 0.007 0.011** 0.009 0.013*** 0.022***(0.002) (0.006) (0.004) (0.004) (0.005) (0.011) (0.003) (0.003) (0.002) (0.007) (0.006) (0.004) (0.006) (0.004) (0.004)

Foreign 0.009*** 0.006 0.013*** 0.008* 0.009 0.019** 0.014*** 0.001 0.010*** 0.008* 0.011** 0.002 0.012** 0.020*** 0.010**(0.002) (0.005) (0.004) (0.004) (0.006) (0.008) (0.003) (0.003) (0.003) (0.005) (0.005) (0.003) (0.006) (0.005) (0.005)

Govt -0.023*** -0.031* -0.007 -0.020*** -0.028*** -0.030* -0.019** -0.022*** -0.015** -0.032*** -0.032*** -0.024** 0.002 -0.028*** -0.024(0.005) (0.018) (0.010) (0.007) (0.009) (0.017) (0.008) (0.006) (0.006) (0.012) (0.008) (0.010) (0.010) (0.006) (0.022)

Loan 0.019*** 0.014*** 0.023*** 0.017*** 0.015** 0.027*** 0.018*** 0.017*** 0.022*** 0.015*** 0.017*** 0.011*** 0.016*** 0.023*** 0.026***(0.002) (0.003) (0.003) (0.004) (0.007) (0.006) (0.002) (0.002) (0.002) (0.003) (0.004) (0.003) (0.004) (0.003) (0.003)

Credit constraint -0.009*** -0.003 -0.014*** -0.017*** -0.018** -0.005 -0.010*** -0.008*** -0.009*** -0.010*** -0.007* -0.002 0.003 -0.023*** -0.010***(0.002) (0.002) (0.003) (0.005) (0.008) (0.006) (0.002) (0.002) (0.002) (0.003) (0.004) (0.002) (0.004) (0.003) (0.003)

Sales credit -0.004 0.001 -0.008 -0.013** 0.004 0.011 -0.007 -0.006 -0.004 -0.007 -0.003 -0.002 0.006 -0.006 -0.003(0.003) (0.007) (0.005) (0.006) (0.008) (0.014) (0.005) (0.004) (0.004) (0.006) (0.007) (0.006) (0.012) (0.005) (0.005)

External finance 0.036*** 0.034*** 0.042*** 0.030*** 0.025*** 0.032*** 0.035*** 0.037*** 0.036*** 0.036*** 0.036*** 0.019*** 0.036*** 0.038*** 0.041***(0.002) (0.004) (0.003) (0.004) (0.006) (0.008) (0.003) (0.003) (0.002) (0.004) (0.004) (0.004) (0.006) (0.003) (0.003)

Constant 0.105*** 0.106*** 0.035*** 0.036*** 0.041** 0.102*** 0.077*** 0.048*** 0.112*** 0.087*** 0.091*** 0.107*** 0.073*** 0.109*** 0.112***(0.003) (0.003) (0.005) (0.010) (0.019) (0.005) (0.002) (0.003) (0.004) (0.005) (0.006) (0.004) (0.007) (0.006) (0.007)

Adjusted R -squared 0.135 0.107 0.100 0.097 0.105 0.130 0.115 0.100 0.151 0.116 0.130 0.119 0.122 0.143 0.135Number of observations 34,524 13,169 13,060 5,739 2,527 3,734 16,387 14,403 21,675 7,978 4,870 10,270 3,971 9,423 9,911Number of countries 95 95 95 94 89 95 95 95 95 95 95 37 10 30 15Note: Standard errors (in parentheses) are robust to heteroskedasticity and clustered on countries. Regressions for the Middle East and North Africa and South Asia are excluded because of insufficient data.

Table 11. Effect of Combined Objective Financial Access Variables on Employment Growth by Firm Characteristic and by Region

Model: EG = b 0 + b 1Small + b 2Medium + b 3Large + b 4Mature + b 5Older + b 6Multi + b 7Manuf + b 8Exporter + b 9Foreign + b 10Govt + b 11Loan + b 12Credit Constraint + b 13Sales Credit + b 14 External Finance+ Country Fixed Effects + e


By age By sector By regionBy size

World



108

Appendix

3.45

6.9

10.34

17.24

24.14

37.93

0 5 10 15 20 25 30 35 40




Electricity

Access to finance

Tax rates


Europe and Central Asia

2.56

2.56

5.13

7.69

10.26

30.77

41.03

0 10 20 30 40 50


Licenses & permits


Tax rates


Access to finance

Electricity


Sub-Saharan Africa

7.69

7.69

15.38

15.38

15.38

15.38

23.08

0 5 10 15 20 25


Corruption

Tax rates



Electricity

Access to finance


East Asia and Pacific

109

Source: World Bank Enterprise Surveys (2006–10).

6.25

6.25

6.25

6.25

6.25

25

43.75

0 10 20 30 40 50

Tax rates


Electricity


Access to finance




Latin America and the Caribbean

12.5

12.5

12.5

12.5

12.5

37.5

0 5 10 15 20 25 30 35 40

Tax rates

Licenses & permits

Electricity

Corruption

Access to finance


Share economies (%)

Middle East and North Africa

16.67

16.67

16.67

50

0 10 20 30 40 50 60



Access to finance

Electricity


South Asia

110

Table A.1 Top Constraints Cited by Enterprises, by Region and Country

Country and survey

year

Frequency

(%)

First

Most cited constraint

Frequency

(%)

Second

Most Cited Constraint

Frequency

(%)

Third

Most Cited Constraint

AFRICA

Angola, 2006 36.8 Electricity 12.5 Corruption 11.6 Access to Finance

Benin, 2009 18.2 Access to Finance 15.0 Electricity 14.6 Practices Informal Sector

Botswana, 2006 24.6 Access to Finance 11.8 Practices Informal Sector 11.0 Crime, Theft & Disorder

Burkina Faso, 2009 35.5 Access to Finance 17.7 Tax Rates 10.8 Practices Informal Sector

Burundi, 2006 41.3 Electricity 16.0 Access to Finance 14.3 Political instability

Cameroon, 2009 24.9 Practices Informal Sector 19.4 Tax Administration 16.6 Access to Finance

Cape Verde, 2009 17.1 Practices Informal Sector 13.1 Access to Finance 11.0 Electricity

Chad, 2009 29.5 Political instability 23.8 Electricity 13.5 Corruption

Congo, Rep., 2009 31.9 Electricity 15.6 Access to Finance 15.5 Political instability

Côte d'Ivoire, 2009 45.2 Access to Finance 28.0 Political instability 7.5 Corruption

Congo, Dem. Rep., 2006 46.5 Electricity 14.9 Access to Finance 9.6 Tax Rates

Eritrea, 2009 28.7 Licenses & Permits 24.1 Political instability 17.0 Access to Land

Gabon, 2009 23.4 Electricity 14.6 Transportation 10.3 Corruption

Gambia, The, 2006 54.5 Electricity 11.7 Access to Finance 6.5 Tax Rates

Ghana, 2007 48.8 Electricity 33.1 Access to Finance 6.3 Tax Rates

Guinea, 2006 64.3 Electricity 10.3 Transportation 8.3 Access to Finance

Guinea-Bissau, 2006 47.1 Electricity 20.1 Access to Finance 7.7 Political instability

Kenya, 2007 21.7 Tax Rates 13.5 Access to Finance 12.0 Practices Informal Sector

Lesotho, 2009 15.9 Access to Finance 14.7 Corruption 11.2 Tax Rates

Liberia, 2009 39.8 Access to Finance 17.4 Crime, Theft & Disorder 13.3 Electricity

Madagascar, 2009 18.6 Electricity 15.4 Practices Informal Sector 13.9 Crime, Theft & Disorder

Malawi, 2009 45.7 Access to Finance 11.4 Transportation 8.9 Practices Informal Sector

Mali, 2007 28.9 Electricity 23.5 Access to Finance 15.1 Tax Rates

Mauritania, 2006 21.6 Access to Finance 14.4 Practices Informal Sector 13.8 Electricity

Mauritius, 2009 30.2 Access to Finance 18.0 Practices Informal Sector 11.3 Electricity

Mozambique, 2007 23.2 Access to Finance 21.4 Practices Informal Sector 9.1 Electricity

Namibia, 2006 21.7 Crime, Theft & Disorder 17.6 Tax Rates 12.1 Access to Finance

Niger, 2009 21.2 Practices Informal Sector 20.3 Access to Finance 15.6 Political instability

Nigeria, 2007 63.6 Electricity 15.5 Access to Finance 7.5 Transportation

Rwanda, 2006 32.9 Electricity 27.4 Tax Rates 13.6 Access to Finance

Senegal, 2007 41.2 Electricity 12.2 Access to Finance 11.0 Access to Land

Sierra Leone, 2009 17.1 Tax Rates 14.8 Access to Finance 14.3 Electricity

South Africa, 2007 40.4 Crime, Theft & Disorder 14.7 Electricity 7.5 Access to Finance

Swaziland, 2006 25.4 Practices Informal Sector 18.5 Crime, Theft & Disorder 15.4 Tax Rates

Tanzania, 2006 73.4 Electricity 9.8 Access to Finance 4.0 Tax Rates

Togo, 2009 23.7 Access to Finance 23.3 Political instability 11.2 Practices Informal Sector

Uganda, 2006 63.6 Electricity 11.3 Tax Rates 8.5 Practices Informal Sector

111

Zambia, 2007 18.6 Tax Rates 15.3 Practices Informal Sector 14.3 Access to Finance

EAST ASIA AND

PACIFIC

Fiji, 2009 44.4 Political instability 8.8 Labor regulations 8.6 Crime, Theft & Disorder

Indonesia, 2009 47.9 Access to Finance 13.7 Practices Informal Sector 6.9 Political instability

Lao PDR, 2009 36.8 Tax Rates 21.2 Access to Finance 16.5 Inadequately educated

workforce

Micronesia, Fed. Sts.,

2009 25.2

Inadequately educated

workforce 15.8 Electricity 12.6 Transportation

Mongolia, 2009 30.3 Access to Finance 16.0 Tax Rates 10.2 Inadequately educated

workforce

Philippines, 2009 26.4 Practices Informal Sector 14.8 Access to Finance 13.0 Tax Rates

Samoa, 2009 16.9 Tax Rates 13.8 Crime, Theft & Disorder 13.8 Crime, Theft & Disorder

Timor-Leste, 2009 36.3 Electricity 12.7 Crime, Theft & Disorder 12.1 Access to Finance

Tonga, 2009 20.1 Practices Informal Sector 17.0 Corruption 15.6 Tax Rates

Vanuatu, 2009 15.7 Electricity 14.8 Access to Finance 14.3 Crime, Theft & Disorder

Vietnam, 2009 24.7 Access to Finance 19.3 Practices Informal Sector 13.3 Transportation

EUROPE AND CENTRAL ASIA

Albania, 2007 27.7 Electricity 17.6 Practices Informal Sector 11.0 Corruption

Armenia, 2009 21.8 Practices Informal Sector 16.0 Tax Rates 15.9 Political instability

Azerbaijan, 2009 23.1 Access to Finance 22.2 Tax Rates 18.2 Corruption

Belarus, 2008 25.9 Tax Rates 14.6 Licenses & Permits 14.1 Inadequately educated

workforce

Bosnia and

Herzegovina, 2009 25.1 Political instability 18.7 Tax Rates 11.4 Practices Informal Sector

Bulgaria, 2009 17.2 Access to Finance 15.2 Practices Informal Sector 13.3 Political instability

Croatia, 2007 18.3 Access to Finance 17.0 Inadequately educated

workforce 15.8 Tax Rates

Czech Republic, 2009 20.0 Access to Finance 14.2 Tax Rates 11.8 Inadequately educated

workforce

Estonia, 2009 28.8 Inadequately educated

workforce 15.9 Political instability 14.7 Practices Informal Sector

Macedonia, FYR, 2009 31.3 Practices Informal Sector 26.9 Access to Finance 6.8 Political instability

Georgia, 2008 18.0 Access to Finance 17.4 Political instability 16.4 Electricity

Hungary, 2009 38.4 Tax Rates 24.2 Political instability 14.2 Tax Administration

Kazakhstan, 2009 26.6 Tax Rates 15.2 Corruption 13.2 Access to Finance

Kosovo, 2009 33.5 Electricity 20.6 Corruption 12.8 Practices Informal Sector

Kyrgyz Republic, 2009 24.5 Electricity 19.9 Access to Finance 11.0 Practices Informal Sector

Latvia, 2009 19.1 Tax Rates 16.7 Political instability 11.3 Tax Administration

Lithuania, 2009 35.2 Tax Rates 12.0 Practices Informal Sector 11.4 Access to Finance

Moldova, 2009 19.5 Access to Finance 15.7 Inadequately educated

workforce 10.4 Access to Land

Montenegro, 2009 18.7 Electricity 17.9 Access to Finance 12.7 Practices Informal Sector

Poland, 2009 22.0 Tax Rates 15.6 Inadequately educated

workforce 13.8 Practices Informal Sector

112

Romania, 2009 27.7 Tax Rates 20.7 Inadequately educated

workforce 19.9 Access to Finance

Russian Federation, 2009 17.2 Tax Rates 16.9 Access to Finance 15.4 Inadequately educated

workforce

Serbia, 2009 20.7 Political instability 19.9 Practices Informal Sector 17.8 Access to Finance

Slovak Republic, 2009 16.2 Tax Rates 13.3 Informal Sector

Competition 12.8

Informal Sector

Competition

Slovenia, 2009 20.0 Tax Rates 19.2 Access to Finance 17.4 Practices Informal Sector

Tajikistan, 2008 24.8 Electricity 22.5 Tax Rates 17.5 Access to Finance

Turkey, 2008 25.9 Access to Finance 18.2 Tax Rates 17.5 Political instability

Ukraine, 2008 23.2 Political instability 17.5 Tax Rates 10.6 Corruption

Uzbekistan, 2008 23.6 Tax Rates 17.9 Access to Finance 9.2 Inadequately educated

workforce

LATIN AMERICA AND THE

CARIBBEAN

Argentina, 2006 16.5 Political instability 15.7 Access to Finance 15.4 Labor regulations

Bolivia, 2006 30.3 Political instability 28.1 Practices Informal Sector 8.0 Corruption

Brazil, 2009 32.8 Tax Rates 13.2 Tax Administration 12.7 Access to Finance

Chile, 2006 18.5 Practices Informal Sector 15.3 Electricity 14.3 Crime, Theft & Disorder

Colombia, 2006 34.6 Practices Informal Sector 12.9 Crime, Theft & Disorder 12.5 Tax Rates

Ecuador, 2006 28.4 Political instability 18.3 Corruption 14.2 Access to Finance

El Salvador, 2006 31.3 Crime, Theft & Disorder 15.3 Practices Informal Sector 13.3 Corruption

Guatemala, 2006 21.0 Practices Informal Sector 20.0 Crime, Theft & Disorder 10.1 Political instability

Honduras, 2006 19.2 Access to Finance 19.2 Corruption 15.6 Crime, Theft & Disorder

Mexico, 2006 19.0 Practices Informal Sector 17.9 Corruption 10.6 Tax Rates

Nicaragua, 2006 26.0 Political instability 17.3 Access to Finance 16.6 Electricity

Panama, 2006 30.6 Electricity 14.6 Tax Rates 10.8 Corruption

Paraguay, 2006 25.8 Practices Informal Sector 21.0 Access to Finance 14.9 Corruption

Peru, 2006 22.1 Practices Informal Sector 17.9 Tax Administration 17.0 Political instability

Uruguay, 2006 32.4 Practices Informal Sector 20.5 Tax Rates 12.0 Access to Finance

Venezuela, RB, 2006 29.2 Inadequately educated

workforce 27.9 Crime, Theft & Disorder 10.0 Corruption

SOUTH ASIA

Yemen, Rep., 2010 32.1 Electricity 26.6 Corruption 7.7 Political instability

Afghanistan, 2008 20.0 Crime, Theft & Disorder 17.9 Electricity 16.8 Access to Finance

Bhutan, 2009 21.7 Access to Finance 12.5 Tax Rates 10.5 Inadequately educated

workforce

Nepal, 2009 62.1 Political instability 26.5 Electricity 2.6 Labor regulations

Source: World Bank Enterprise Surveys.

113

Explaining Africa’s (Dis)advantage17

Ann E. Harrison

Justin Yifu Lin

Lixin Colin Xu

World Bank

This draft: February 9, 2011

Abstract

Africa‘s economic performance has been widely viewed with pessimism. In this paper we use firm-

level data of 89 countries to examine African firms‘ performances and the reasons behind their

disadvantages. Relative to similar-income countries, formal manufacturing firms of Africa actually

do not perform much worse but do exhibit structural problems: similar sales growth, higher labor

growth, slightly lower productivity, but with much lower export intensity and investment rates.

However, once we control for a comprehensive list of the business environment, Africa leads in

productivity and sales growth, and the conditional advantage remain reasonably robust. Moreover,

the conditional advantage of Africa is higher in low-tech than in high-tech manufacturing, in small

than in large and medium firms, and exist only in manufacturing but not in services, suggesting that

Africa may have comparative advantage in simple light industries. Overwhelmingly important for

explaining Africa‘s disadvantage in firm performance relative to some reasonably successful

developing countries are firm size, infrastructure, government expropriation, crime, and the access to

informal finance. This list largely represents the basic roles of the government. Interestingly, we find

heterogeneous effects of the business environment. The effects of the business environment tend to

be stronger in manufacturing than in services for developing countries, and party monopoly and

domestic conflicts tend to shift the sector structure from manufacturing to services. Relative to low-

17

We‘re grateful to very useful comments, discussions, help, and criticisms from Jing Cai, Hinh Dinh, Luosha Du,

Dimitris Mavridis, Vincent Palmade and Philip Keefer. Helen Yang offered superb research assistance. This study

has been financed by the Japanese PHRD TF096317, the Dutch BNPP TF 097170, along with the Africa Region of

the World Bank. The views expressed here are the authors‘ own and do not reflect those of the World Bank, its

executive directors, or its member countries.

114

tech manufacturing, high-tech manufacturing depends more on modern infrastructure, market size,

and are more sensitive to firing costs. Finally, large and medium firms enjoy higher economy of

scale, direct benefits from foreign ownership, and are hurt more by labor regulation (for firm growth)

and by corruption.

I. Introduction

In the 1960s, Africa‘s economic growth was similar to South Asia (Collier and Gunning 1999

a). However, between 1970 and 2000, the average GDP per capita growth rate was only 0.5 percent

per annum (based on WDI data), and sub-Saharan Africa is now the poorest continent. After 2000

and before the recent financial crisis, however, the continent had experienced resurgence in growth.

Growth in GDP for the continent averaged 5.9 percent annually (World Economic Forum 2009). Is

this trend sustainable? How can Africa keep growing? What are the key policies that facilitate

Africa‘s economic performance?

This paper tries to shed light on these questions using micro data. The key to sustainable

economic growth for developing countries is industry upgrading based on its comparative advantage

(Maddison 2001; Lin 2009). For the poor SSA region, this largely implies the need to develop

manufacturing, especially simple light manufacturing industries that do not require high capital-labor

ratios. To focus on the competitiveness and its determinants for manufacturing firms, we therefore

rely on manufacturing firms in the World Bank‘s Enterprise Survey, along with other cross-country

data set on politics, geography and business environment, to study formal manufacturing firms‘

performance in Africa relative to other regions,18 to study the determinants of firm performance, and

to determine the key factors behind Africa‘s disadvantage, if any, relative to other regions.19

How do African manufacturing firms do compared to other regions? Using the World Bank

Enterprise Survey of 89 countries, and comparing Africa with countries in other continents with GDP

per capita below 3000 U.S. dollars (in 2005 value) (see Table 2), manufacturing firms in Africa

exhibit lower productivity (around 5.4 log points), lower export share in sales (about half of the other

countries), and lower investment rate (as measured by investment over value added, 0.134 vs. 0.175).

The sales growth levels are similar, and the labor growth rate in Africa is actually higher. Thus the

difference is actually not large if we focus on conventional firm performance indicators such as labor

productivity, sales growth, and labor growth. However, there are substantial shortfalls in terms of

export intensity and investment rates, both of which should have predictable power for future growth.

In this paper, we look at a comprehensive set of firm performance outcomes, including static

efficiency (labor productivity), dynamic efficiency (sales growth), job creation (labor growth), export

capacity (export intensity), and investment rate. The interest in the first three outcomes is obvious.

Why do we care about export intensity? Van Biesebroeck (2005) presents evidence that export raises

18

In the rest of this paper, we use SSA and Africa interchangeably. Africa does not include the North Africa region

in this paper. 19

It is important to keep in mind that all firms in the ES that we use consist of formal firms. In a few African

countries, informal firms were also surveyed, but the number of countries is too small to merit a full scale cross-

country studies.

115

productivity in Africa in a causal sense. Similarly, Bigsten et al. (2004) and Mengistae and Patillo

(2004) document a positive association between export and productivity. Partly in light of these

pieces of evidence, many policy makers and economists have argued that increasing export is a key

for Africa to circumvent the small domestic market problem. Why do we care about investment rate?

The existing literature has found that Africa tends to have significantly lower investment rate (Bisten

et al. 1999, Gyimah-Brempong and Traynor 1999, Devarajan et al. 2001), and conventional growth

analysis suggests that investment rates are key to economic growth. This broad focus on firm

performance thus differentiates our paper from the existing studies of African firms, which tend to

focus on a single outcome at a time and miss important aspects of how African firms behave and

perform. Another distinguishing aspect of this study is that we aim to be comprehensive in including

potential explanatory variables: geography, political risks, ownership, competition, infrastructure,

crime and violence, labor flexibility, and access to formal and informal finance—this is perhaps the

most comprehensive list of the business environment in a single paper (see Xu forthcoming for a

survey of this growing literature). We aim to quantify the relative importance of these policy factors

in explaining African manufacturing performance.

The main data source we rely on is the World Bank Enterprise Survey (ES hereafter), which

include firm-level data of 89 countries. In each country, firms of all sizes and ownerships are

covered from both manufacturing and services industries. The survey questions are broad, including

detailed quantitative measures which allow us to infer firm performance such as labor productivity,

sales growth, labor growth, investment rates and export intensity. More importantly, the data asks

many detailed questions, both subjective and objective, on the business environment that a firm

faces, such as infrastructure issues, regulatory burdens, corruption, crime, and access to finance.

Aiming to have comprehensive measures of the business environment, we supplement cross-

country data with relevant ES questions. In particular, we add these country-level variables: phone

density, LPI (i.e., logistical performance indicator of the World Bank) indices on customs efficiency,

on infrastructure quality, the incidence of domestic conflicts, the transportation costs of export, the

minimum costs of starting a business (both from Doing Business), voice and accountability (from

ICRG), trade orientation, and political competition (from Database of Political Institutions).

For the business environment questions answered by firms, we deal with its potential

endogeneity in several ways (see also Dollar et al. 2006, Aterido et al. forthcoming; Xu

forthcoming). First, we mainly rely on objective measures of the business environment (Dethier et

al. 2008). Subjective answers may be based on firms‘ performance directly, and may be determined

by country-specific factors such as exposure to open media and its development history. Only for

business environment aspects that we cannot find objective measures, such as the prevalence of

crime, we rely on subjective assessment. Second, we do not directly use firms‘ answers on the

business environment. Instead, we rely on city-industry-size average of firms‘ answers to gauge

local business environment. This local measure is less subject to the reverse causality issue

associated with firm-level answers, and may proxy the actual business environment well—recent

studies of business environment suggest that there are vast variations in business environment within

116

a country (Hallward-Friemeier and Pritchett 2010, Hallward-Dreimeier et al. 2010). Third, with the

local business environment on the right-hand side, there may be local-level omitted variables that are

closely correlated with our explanatory variables. To check this possibility, we control for additional

local and country-level determinants of firm outcomes, and show that our key results remain robust.

Fourth, in our interpretations, we try to explain the measure more broadly. For instance, the local

web-use intensity likely measures all related factors such as electricity supply, utility costs, and so

on, and we just view it as an indicator of modern infrastructure (Dethier, Hirn and Straub 2008).

However, it remains true that what we find are likely a collection of correlations, and causality

cannot be inferred. To the extent that we have robust results, and we have a coherent story line when

we allow the business-environment-performance relationship to vary along several data facets, our

results are more credible.

Our empirical investigation suggests that formal African manufacturing firms show slight

disadvantage in productivity but large shortfall in export orientation and investment rate, similar

sales growth, and advantage in labor growth. So on the one hand, the manufacturing performance

differences of Africa to their similar-income counterparts are limited, and on the other hand, there are

strong structural differences in behavior—in low export capacity and investment.

Africa‘s limited disadvantage in firm performance, however, can be readily explained by the

business environment. Once we control for firm size, ownership, competition, infrastructure,

transport costs, government expropriation, labor flexibility, finance, and crime, Africa actually leads

in productivity and sales growth, and the conditional advantage remain intact with more control of

business environment factors, the consideration of multicollinearity, and using only the similar

income sample. Moreover, the conditional advantage of Africa is higher in low-tech than in high-tech

manufacturing, in small than in large and medium firms, suggesting that Africa may have

comparative advantage in simple light industries.

We then explain the firm performance differences between Africa and similar-income-but-

well-performing countries—the comparison group is to be defined precisely later. Overwhelmingly

important for explaining Africa‘s disadvantage in firm performance are firm size, infrastructure,

government expropriation, crime, and the access to informal finance. Except for informal finance,

this list largely represents the basic roles of the government: property rights protection, public goods

for trade, a safe environment, and government regulations to allow firms to grow into large firms.20

In contrast, labor flexibility and access to formal finance are less important.

Our paper adds to the literature explaining Africa‘s poor economic performance. Besides

several nice summaries of this literature (see Collier and Gunning 1999a, 1999b; Bigsten and

Soderbom 2006), there are many papers examining one aspect of firm performance (such as

investment rate, sales growth, or labor growth, or export, or productivity), often using one or several

African countries‘ firm-level data.21 However, there is no study, as far as we know, that examines all

20

This list corresponds to what Blanke and Sala-i-Martin (2009) calls ―basic factor-driven requirements‖. 21 A partial list of findings from this literature includes the following:

117

these key determinants of African development at firm level, look at both static and dynamic

performance measures as well structural indicators (such as investment rates and export capacity) in

the same paper, and with both a large Africa and non-Africa firm samples. Looking at all the key

determinants allows us to reduce the extent of omitted variables for any of the key explanatory

variable. Moreover, we can gauge the relative importance of elements of our list of the business

environment. In light of the limited financial and administrative capacity of reformers and policy

makers—capacity that is especially constrained in Africa—it is important to identify the key

constraints for growth (Hausman, Rodrik and Velesco 2005). Our paper, for instance, is able to show

that import tariffs, being landlocked, and access to formal finance do not seem to matter as much as

some of the other elements such as infrastructure, political competition, crime, and trade credit.

This paper also adds to the literature of the effects of the business environment. We obtain

several novel findings: (i) political monopoly is negatively associated with firm expansion and

First, African firms feature low investment rate, likely due to its high risks (Bisten et al. 1999, Gyimah-Brempong

and Traynor 1999, Devarajan et al. 1999). There is inconclusive evidence about whether investments in Africa are

efficient. Devarajan et al. (1999) argue for inefficiency based largely on cross-country data, while Gunning and

Mengistae (2001) argue for efficiency based on firm-level African data.

Second, exports tend to have positive effects in Africa (Van Biesebroeck 2005; Bigsten et al. 2004; Mengistae and

Patillo 2004).

Third, traditional African culture plays an important role in explaining African economic performance. In particular,

ethnic division figures prominently a significant role in explaining Africa‘s growth tragedy (Easterly and Levine

1997). For instance, strong ethno-regional identities lead a weak state and continuous state failure (Platteau 2009).

Real power in Africa likely has stayed in the hands of the head of ethnic or kinship groups, and to stay in power, the

government tends to focus on redistribution transfers to ethnic groups and ―buy‖ power base.

Fourth, firm size distribution may play a role in explaining Africa‘s economic performance (Gauthier and Gersovitz

1997). SleuWaegen and Goedhuys (2002), using firm-level data of Cote D‘Ivoire, finds that relative to western

economies, small firms tend to grow slower, but large firms tend to grow faster, and very few small firms graduate

to become large firms in due time (see also Van Biesebroeck 2005). Richmond and Klapper (2010) find that

middle-sized Cote D‘Ivoire firms face highest tax burdens. Soderbom and Teal (2004), based on panel firm-level

evidence from Ghana, find that large firms tend to face higher relative labor costs than small ones, and this partly

explains why most firms prefer to stay small.

Fifth, trade credit may play an important role in Africa. Fisman (2001), using data from five African economies,

finds that supplier credit is positively associated with capacity utilization due partly to better inventory management

and lower chance of inventory shortage. Fafchamps (2000) find that women and blacks are not disadvantaged in

access to bank loans, but they have less access to supplier credit (see also Biggs, Raturi and Srivastava 2002). Dinh

and Mavridis (2010) also find access to finance to be one of the top perceived bindings constraints. Collier and

Gunning (1999b) suggests that a lack of financial depth may be a reason behind Africa‘s growth shortfall.

Sixth, corruption has negative effects in Africa. Fisman and Svensson (2007) find that the effects of bribes are

three times larger than that of taxes in Uganda.

Seventh, the cross-country literature suggests that a lack of openness to trade may be an important reason behind

Africa‘s poor growth (Sachs and Warner 1997, Easterly and Levine 1997).

Eighth, geography may be an important reason for Africa‘s backwardness. In particular, being landlocked may

hinder trade and development (Sachs and Warner 1997).

Ninth, Africa features high indirect costs (i.e., costs beyond production floor) for doing business (Iarossi 2009; Eifert

et al. 2006).

118

investment rates; (ii) domestic conflicts have adverse effects on investment rates; (iii) trade credit

plays an important role in developing countries for productivity and investment; and (iv) crime

prevalence deters firm expansion and exports. Moreover, we find strong evidence of heterogeneous

effects of the business environment, a pattern conjectured in a survey of this new literature but not

tested in a single large cross-country firm-level data (Xu forthcoming). The effects of the business

environment in general tend to be more strongly felt in manufacturing for developing countries,

while party monopoly and domestic conflicts tend to shift sector structure from manufacturing to

services. Relative to low-tech manufacturing, high-tech manufacturing depends more on modern

infrastructure, market size, and are more sensitive to firing costs. Trade credit facilitates labor

growth for low-tech manufacturing but improves productivity and investment for high-tech

manufacturing. Finally, large and medium firms enjoy higher economy of scale, direct benefits from

foreign ownership, and are hurt more by corruption and labor regulation.

II. Measurement of the Business Environment and Our Data

The main data source of this paper is the World Bank‘s Enterprise Surveys (ES) in 89

countries. With our primary focus on the manufacturing sector, we will mainly use the manufacturing

sample of around 12000 firms. The number of countries in sub-Sahara is 33 (see the Appendix for

the list of the SSA countries).

ES nicely contains various aspects of the business environment, including the access to

finance, infrastructure, regulation burdens, taxes, corruption, and so on. This allows us to

simultaneously control for various aspects of the business environment, and reduce the extent of

omitted variable bias. The survey has both objective and subjective measures of the business

environment. In general we rely on objective measures to avoid endogeneity of firm responses. The

reason is that a firm‘s perceived obstacle can reflect both the objective level of services and the

marginal cost of bad services, which differs by firm types. Moreover, the perceived obstacle of an

aspect may be affected by the openness of information, the history of performance of this aspect, all

of which differ by countries. Indeed, Bigsten et al. (2003) find that firms in the SSA region firms

complaining loudly about infrastructure tend to be those most productive firms and exporters. For

aspects without proper objective measures such as crime, we do occasionally rely on subjective

measures.

For indicators of the business environment from ES, we do not directly use individual

answers since they may be endogenous. For example, firms answering strong constraint of

electricity may reflect two vastly different realities, one in which there is indeed a strong constraint

on the firm of electricity access and costs, and the other in which electricity has higher marginal

value, which is related to firm performance (―reverse causality‖). We thus follow the literature of the

business environment to use local average of the business environment answers as a proxy of the

local business environment (Dollar et al. 2005; Hallward-Dreimeier et al. 2006; Aterido et al.

forthcoming; Xu forthcoming). In particular, we opt to rely on a city-industry-size cell as the basic

119

unit for measuring local business environment. The business environment tends to differ vastly

across regions (Almeida and Carneiro 2009; Xu forthcoming) so we allow the city-specific

dimension. Different industries face varying needs in licensing, permits and infrastructure and may

deal with different regulators and level of competition, so we allow industry-specific business

environment. The literature also documents how firms with various sizes face different business

environment, in particular, small firms are particularly vulnerable to expropriation (Beck et al. 2005;

Cull and Xu 2005), so we allow the business environment to differ by size. To implement, we define

a firm to be small (large) if the firm has fewer (more) number of employees than the median firm of

the city-industry cell.22 Relying on the local business environment indicators mitigates the reverse

causality issue associated with using firm-level indicators--it is difficult for an individual firm to

directly alter the local business environment.

The second main sources of our data come from cross-country sources. For many aspects

that we are interested in, the variations only come from the country level, such as geography,

political system, macro policies, and so on. Since we aim to be comprehensive, we combine our ES

data on the business environment with other measures of the business environment (broadly defined)

at the country level. We now describe how we measure our key variables.

Infrastructure covers a large ground. To rely on existing data while trying to be

comprehensible, we use four main measures to capture the quality of infrastructure: (1) Telephone

density, which captures the communication component of local infrastructure; (2) the local (i.e., city-

industry-size) share of firms using websites for their businesses, which captures the extent of using

modern communication infrastructure; (3) customs efficiency (from LPI), which captures the

efficiency of the clearing process (i.e., speed, simplicity, and predictability of formalities) by boarder

control agencies, including customs; (4) the infrastructure index from the Logistics Performance

Index of the World Bank (Arvis et al. 2010), which measures the quality of trade and transport

related to infrastructure (e.g., ports, roads, railroads, information technology). There are some

overlap between (4) and (1) and (2). However, the LPI infrastructure index captures the important

dimension of roads and ports. ES also has indicators of local electricity quality, and we did not

found it to matter for any of our outcomes and thus opted to leave it out.

Geography has been suggested as an important reason for Africa‘s lack of development

(Collier and Gunning 1999; Sachs and Warner 1997). In particular, Africa has more landlocked

countries, which heightens the need for coordination with neighboring countries. Given the

significantly higher trade costs associated with borders, being landlocked necessarily impede

international trade. In addition, African countries tend to be small in population (but not necessarily

in land area). To capture the geographical elements related to international trade, we thus rely on the

following variables: (1) a dummy variable of a country being landlocked; (2) the domestic market

size, as proxied by the country population; (3) transport cost to export (in US dollars per standard

container) (from Doing Business).

22

When the city-industry-size cell has fewer than 5 observations, we replace the cell mean with the city-industry

mean.

120

Ethnic fractionalization. Africa displays strikingly higher ethnic fractionalization than

other regions, not surprisingly since Africa was the origin of the human species and throughout most

of the human history the human had been evolving only in Africa. Not surprisingly then, ethnic

fractionalization has been put forth as an important explanation for poor African performance.

Collier and Gunning (1999), when proposing a lack of social capital as an explanation for African

economic performance, mention ethnic fractionalization as potential barriers to social interactions.

Easterly and Levine (1997) suggest that ethnic fractionalization accounts for 35 percent of Africa‘s

growth shortfall, and even this may only reflect direct effects. We thus include this variable as a

potential explanation for firm performance.

A basic protection of property rights is important for existing and potential investors to be

willing to commit to investment projects without fearing being expropriated. Cross-country evidence

suggests that countries with worse property rights have lower aggregate investments, worse access to

finance, and slower economic growth (North 1990; Knack and Keefer 1995; Acemoglu, Johnson and

Robinson 2001). Acemoglu and Johnson (2005) suggests and find evidence that property rights

institutions (i.e., those related to government expropriation) tend to be more important than

contracting institutions (e.g., those safeguarding private transactions). Some new firm-level evidence

also point to the importance of property rights (Johnson, McMillan and Woodruff 2002; Cull and Xu

2005), the adverse effects of corruption (Fisman and Svensson 2007; Cai et al. forthcoming), and

potential mechanisms include better external finance (Demirguc-Kunt and Maksimovic 1998); better

asset allocation (Claessens and Laeven 2003), a higher share of large formal firms (Demirguc-Kunt

et al. 2006). To take into account of the protection of property rights on firm performance, we thus

construct two measures of government expropriation.

The first is political competition. We rely on the DPI data (Beck et al. 2001; Keefer 2007), in

particular, the number of years that the ruling party has been in power. Our reasoning is that the

more a ruling party has been in power, the more absolute power the ruling party has, and the higher

the risk of unconstrained government expropriation.23 The second measure of basic protection of

property rights is the extent of corruption. We opt to use the ES measure of corruption: the local

average of bribes to the government over sales. This is directly comparable across countries, and

avoids some of the flaws of subjective perception on corruption (Dethier et al. 2008).

A very basic need for firms to invest and grow is to enjoy a safe environment in which their

factories will not be damaged by war, by gangsters and criminals. Yet, this simple demand is not met

in many countries, which are inflicted by wars and rampant crimes due to a weak government that

23

This intuition seems to capture the reality of most developing countries well. However, some argue that several

East-Asian countries also have low political competition as measured by this variable but seem to have performed

well in the past several decades. Gehlbach and Keefer (2010a, 2010b) suggests that this may be due to much better

ruling party institutionalization in East Asian countries relative to other countries. Institutionalized ruling parties

facilitate development in autocratic countries for two main reasons. First, they tend to have broad party bases so that

the party represents broad rather than narrow constituent interest. Second, there are institutionalized rules about

removing incompetent or shirking party-leading rulers. Thus the party can credibly commit to remove bad leaders,

and leaders as a result work hard and perform well.

121

cannot control domestic disorder and contain crimes. To capture the importance of this element of

the business environment, we rely on two measures: (1) A dummy variable of experiencing domestic

conflicts (whether minor or major conflicts) in the past 10 years in a country. This is based on the

methodology of Gleditsch et al. (2002) and updates of UCDP/PRIO (2010). (2) We do not have good

measures of objective amount of crime faced by firms. However, firms are asked the degree of

obstacles that they view crime constitutes for their development. We thus construct Cons_Crime,

which is the local share of firms that view crime as a moderate or major constraint.24

A key dimension of the business environment is product market competition, which

captures the extent to which the product market is competitive. For instance, is entry easy? Are the

industries competitive? Is foreign entry easy? There is a large literature on the effects of product

market competition in developing countries (Tybout 2000, Schiantarelli 2010). In general, the

literature has found product market competition tends to be important for employment, productivity,

entry and turnover and growth (Palmade 2005; Schiantarelli 2010). However, most of the literature

does not simultaneously control for other basic country characteristics such as political risks,

infrastructure, access to finance, and so on. It would thus be interesting to examine whether product

market competition remains important once we control for other basic business environment

elements.

Given the multifaceted nature of product market competition, we employ several measures to

capture product market competition. The first is the country-industry-year import tariff, a high level

of which means lower competition for domestic producers. Second, the country-industry level

competition (―Compete_Ind‖), as measured by (1 – markupCI) (Aghion et al. 2005). MarkupCI is the

country-industry average of firm-level markup (i.e., (value added – labor costs)/sales). A higher

value implies more competition. Third, in sensitivity checks, we also include a subjective measure of

informal competition. Firms are asked the extent to which ―practices of competitors in the informal

sector‖ as a serious obstacle (on the scale of 0 to 4, with 4 being serious obstacle). The share of firms

viewing it as a moderate or severe constraint is used to proxy competition from the informal sector.

Finally, the Doing Business data set contains a proxy of entry barrier, the minimum capital required

(% of income per capita) for starting a business.

Labor market flexibility. While the cross-country literature on labor regulation has been

inconclusive (Xu forthcoming), some recent micro studies suggest that stringent labor regulations can

have serious adverse effects. For instance, Amin (2009a) and Almeida and Carneiro (2009) find that

cumbersome labor regulations are associated with smaller firm size, more informality and higher

unemployment in India and Brazil, and Amin (2009b) find that tighter labor regulations in India lead

to differences in technological choices. Harrison (2010) finds strong evidence of negative effects of

minimum wage in Indonesia. Almeida (2005) find that stringent labor regulations in Brazil lead to

lower productivity and investment. Petrin and Sivadasan (2006) find that higher firing costs lead to

increasing gaps between marginal product of labor and wages in Chile.

24

The data set also has the share of sales lost due to theft and vandalism during transportation. We have not found this variable to matter for our firm performance measures.

122

Labor market flexibility is proxied by an index of toughness for firing workers at the county

level (Fire Cost hereafter) (Botero et al. 2004). Fire Cost is an index measuring the cost of firing 20

percent of the firm‘s workers (10% are fired for redundancy and 10% without cause). The cost of

firing a worker is calculated as the sum of severance pay and any mandatory penalties established by

law or mandatory collective agreements for a worker with three years of tenure with the firm.

Access to finance. There is a large literature on the link between access to finance and

economic development (see Levine 1997). Most of this literature focuses on access to formal finance

such as bank loans and overdraft facilities. There is a growing literature that examines the impact of

access to informal finance (Fisman 2001, Cull and Xu 2005). Following this literature, we include

access to both formal and informal finance, and investigate whether they have different impacts, and

which has a larger impact. Access to formal finance is measured as city-industry-size share of firms

with access to overdraft facilities, as in Dollar et al. (2005), and access to informal finance is

measured as city-industry-size share of firms to grant trade credit (to other firms).25

III. How does Africa stack up against other countries?

To examine African manufacturing performance, we include both static and dynamic

performance. To see how Africa stacks up against firms in other regions, we look at labor

productivity (in logarithm), which is measured as sales (in constant U.S. dollar of 2005 value) over

the number of employees. To see Africa‘s relative momentum, we look at African firms‘ sales

growth and their investment rate. Since the rate of job growth is of independent interest, we also

examine employment growth. But employment growth likely does not capture as much about the

growth of the economy as sales growth, which can be accomplished by labor growth, capital growth,

and increase in productivity, so labor growth is just a component of sales growth. Moreover, a recent

study of Africa shows that African employment growth tends to concentrate on micro (and relatively

non-productive) firms, and such growth patterns may represent resources misallocation under

regulatory failures (Aterido and Hallward-Driemeier 2010).

Finally, many argue that exports are especially important for development in poor and

especially African countries due to their small domestic markets. For instance, using a panel of

manufacturing firms in 9 African countries, Van Biesebroeck (2005) finds that exporters are more

productive than non-exporters, and they increase their productivity advantage after entering the

export markets. This positive effect holds even after controlling for unobserved productivity

differences and self-selection into the export market. Consistent with common wisdom of why

export may be especially important for poor and small-market countries, he finds that scale economy

from access to a larger market is an important channel for the export effects. Similarly, Bigsten et al.

(2004) find evidence in Africa that is consistent with learning by exporting, and Mengistae and

25

We have also tried using local share of firms receiving trade credit. These two variables are, not surprisingly,

very closely correlated. This fact, coupled with the fact that the latter measure has significant fewer observations,

leads us to use the former trade credit measure in our empirical analysis.

123

Patillo (2004) document a positive association in Kenya, Ghana and Ethiopia. We thus also use

export intensity (e.g., the share of export in total sales) as another firm performance measure.

For sales (and employment) growth, we use growth during a three-year period since the data

present the data this way. Following Davis and Haltiwanger (1995), we compute sales growth rate as

(sales this year – sales three years ago)/their average. This sales growth measure is bound between

-2 and 2, and successfully contains the outlier problem.26

Since our key objective is to uncover what hinders poor Sub-Saharan African countries from

having a robust manufacturing sector, when we define ―Africa‖ in our paper, we exclude from our

Sub-Saharan Africa sample South Africa, Botswana, Mauritius and Namibia, which have GDP per

capita higher than 3000 US dollars (in 2005 value). The countries in our new African sample consist

of all SSA sample countries that have GDP per capita lower than 3000 U.S. dollars (in 2005 value) in

the time of each country‘s survey. However, in later sensitivity check, we do show that our key

conclusions about Africa do not hinge on whether we keep or drop these four successful SSA

countries.

Since to judge Africa economic performance necessarily entails comparing Africa with other

countries, we construct two comparison groups for Africa. The first comparison group, called the

average comparison group hereafter, consists of countries in non-SSA countries with per capita GDP

lower than 3000 U.S. dollars (in 2005 value). The second comparison group, the better comparison

group hereafter, consists of the better-performing top half countries of the average comparison group.

We measure country performance in the following way. We first standardize each of the five

performance measures so that each now has a mean of zero and a standard deviation of 1. We then

add up the five standardized measures to form the aggregate performance measure. Countries with

mean aggregate performances in the top half of the average comparison group then form the better

comparison group.

Differences in firm performance

For the average comparison group, Africa manufacturing has worse firm performance in 3 of 5

outcomes. Log labor productivity in Africa is lower by 5.4 log points, and is statistically significant

at the 10 percent level. Sales growth is slightly lower but statistically insignificant. Export share is

lower by 8.6 percentage points (and the difference is statistically significant), only half of the

comparison group. Investment intensity in Africa is lower by 4.1 percentage points (11.8 versus 14.9

percent), about three quarters of the level of the comparison group. However, Africa has

significantly higher labor growth rate: higher than the average comparison group by 6.8 percentage

points (or 50 percent).

26

Otherwise, one often sees growth rate of 1000% or above, while the mean growth rate is often in single-digit

percent.

124

The difference with the better comparison group is more pronounced. The labor productivity

lag is significant at 31 log points. While sale growth difference remains insignificant, the export

share is significantly lower by 14.7 percentage points, and the investment rate is significantly lower

by 10.7 percentage points. The African labor growth advantage is similar.

Thus the growth performance of manufacturing firms in African countries is not bad relative

to similar-income countries, but they exhibit much lower productivity level, export capacity and

investment level when being compared with the better comparison group.

Difference in key explanatory variables

African manufacturing firms are 3 to 5 years younger. They have higher foreign ownership, by 6 (3)

percentage points relative to the average (better) comparison group. They show higher ownership

concentration, with the largest owner claiming 82 percent of firm ownership, higher than the average

(better) comparison group by 8 (12) percentage points. They are much smaller in size, with the

number of employees smaller by 70 (105) log points than the average (better) comparison group.

Geography in Africa is characterized by a higher tendency for being landlocked and a smaller

domestic market. Africa is slightly more likely to be landlocked than the average comparison group

(but statistically insignificant), but significantly more likely to be landlocked than the better

comparison group (25% versus 17.5%). African countries also have smaller populations, with the

population of the average country lower by 60 (115) log points than those in the average (better)

comparison group. Furthermore, even with the same upper bound in income, African countries

(excluding those top 4 richest countries) still have much lower average GDP per capita at $488, about

1000 dollars lower than the average and the better comparison groups.

Africa has a worse protection of property rights. Africa displays stronger political monopoly

and thus a higher likelihood of government expropriation—the logarithm of the number of years that

the ruling party in an African country has been in power is greater by 46 (14) log points for the

average (better) comparison group. Moreover, African firms have to pay higher bribes to get things

done.27 African firms‘ average share of bribes to government (over sales), 2.9%, is significantly

higher than the average comparison group by 1.3 percentage points, and the better comparison group

by 1.9 percentage points.

African firms have to confront violence and crime to a greater extent. In terms of the

incidence of major or minor domestic conflicts in the previous 10 years, Africa has a higher tendency

than the average comparison group (50% versus 41%). Surprisingly, the better-performing group

actually has higher past history of domestic conflicts (57%). The subjective perception of crime as a

moderate or severe obstacle is actually similar between Africa and the similar income countries, but

the better comparison group experiences significantly lower perceived crime.

27

This is very similar to the fact that firms in the poorer regions in China tend to spend more on entertainment costs (for government officials) in China (Cai et al. forthcoming).

125

Africa features the highest level of ethnic diversity. The ethnic fractionalization (or

diversity) in Africa, at 0.715, is much higher than the average (better) comparison group by 0.316

(0.387).

Infrastructure in Africa is worse, an outcome expected in light of its low income level. In

particular, the telephone density in Africa is much lower, by around 200 log points relative to both

comparison groups. Africa also features much lower local website intensity at 0.13, which is 0.27

lower than the average comparison group, and 0.35 lower than the better comparison group.

Moreover, by both the custom efficiency index and the LPI infrastructure index, Africa is

significantly worse than the average and especially the better comparison groups.

Labor flexibility in Africa appears to be lower, with higher firing costs than in both

comparison groups. African firms also have worse access to finance, both formal and informal. The

local average share of firms with overdraft facilities is 23 percent, 22 (9) percentages point lower

than the average (better) comparison group. Note that the better comparison group actually has

lower access to formal finance than the average comparison group, suggesting that access to formal

finance is perhaps not crucial for firm performance for countries at low income level. Equally

important, African firms also have less access to trade credit: the local average of trade credit

(granted to other firms) is 0.27, which is 0.27 lower than the average comparison group, and 0.32

lower than the better comparison group. Note the rank order of trade credit prevalence corresponds

to the order of firm performance, which suggests that trade credit may play an important role in

explaining regional economic performance, a conjecture confirmed later.

Competition differs for different segments of African economy. Formal African firms face

lower import competition. The average import tariff is higher. At 14.6%, Africa‘s average import

tariff rate is slightly more than 5 percentage points higher than both comparison groups. Similarly,

when measured by industry-level competition, formal Africa firms (in our sample) are also slightly

less competitive, though the difference is quite minor. In contrast, one can argue that the informal

sector does feature more competition. Competition from the informal sector does impose more

threats than in elsewhere. While 69% of African firms view competition from the informal sector as

important, only 40% of firms in the two comparison groups share this view.

IV. Determinants of firm performance

We now investigate how the business environment affects firm performances, and how

Africa stacks up. The empirical specification is the follows:

(1)

where Y is the firm performance indicator, and i, c, j represent firm, country, and local levels. Y

could be log labor productivity, sales growth rate, employment growth rate, export intensity, and

investment rate. F is basic firm-level controls. For labor productivity, export share (i.e., export/sales)

and investment rate (i.e., investment/value added), we include log of the denominator to capture the

126

scale effects. For export share and investment rate, we also control for log of the number of

employees to allow for difference in behavior associated with firm size. For the two growth rate

equations, we control for initial level (i.e., initial sales and initial labor) to allow for regression

toward the mean. CB represents country-level business environment, capturing all country-level

controls such as the population, being landlocked, the infrastructure indices and so on. LB represents

local-level (i.e., city-industry-size) business environment variables. We include an error term to

capture unobserved variables and measurement errors.28 Since the error terms within a country may

be correlated due to omitted common factors or shocks, and many of our explanatory variables are

aggregate variables, we cluster the standard errors of each equation at the country level (Moulton

1990).

As discussed earlier, to avoid endogeneity associated with some of the business-environment

variables being choice variables for firms—such as using websites, bribing, and access to overdraft

facilities—we use city-industry-size average of these variables as proxies of local business

environment. Omitted local variables remain possible. To address this issue, we (i) include further

country-level controls; (ii) include other local-level variables to reduce the possibility of omitting

other local variables. In addition, we examine the robustness of our results when we further control

for country fixed effects—with significant costs. Then we cannot estimate the effects of country-

level variables such as domestic conflicts, party monopoly, phone density, and so on. Moreover,

some countries only have one or a couple of cities—on average there are only 3.4 cities per country.

Controlling for country dummies thus demands the data too much. To respect the limited cross-

location variations in our data—and to be able to estimate the average Africa effect—we thus mainly

rely on the specification without controlling for country dummies.

Table 4 reports the base results. Once controlling for firm characteristics and our list of the

business environment, the coefficient of the Africa dummy is positive and large for labor

productivity (217 log points) and sales growth rate (21 percent). Thus conditional on our key

variables, Africa does not lag behind at all for productivity and sales growth—in fact it is far ahead

of other regions. Taken at face value, if fixing the daunting list of the business environment to the

levels in other countries, Africa can lead other regions. For labor growth, export intensity, and

investment rates, the Africa dummy is insignificant. Our explanatory variables therefore do a good

job in explaining the outcomes, and whatever Africa lags can be explained by our control variables.

28 Since we have many control variables, and the coverage of both Enterprise Survey and cross-country sources

differ for some countries, dropping the observations with any variable missing would result in the loss of the

majority of the sample. For some variables, we thus resort to imputation, using the predicted value based on the

country-level urbanization level, the regional dummies (Africa, East Asia and Pacific, East Europe and Central Asia,

South Asia, Middle East and North Africa), the service industry dummy, and firm size dummies (10-20, 20-60, and

60+ employees).

127

Firm characteristics matter a great deal in explaining firm performances. First, larger firms

have higher labor productivity and export share, but lower sales and labor growth rates. Thus large

firms exhibit better capacity and static efficiency, and consistent with the findings in Cote D‘Ivoire,

large firm tend to grow slower (SleuWaegen and Goedhuys 2002). Second, younger firms tend to

have lower productivity, consistent with SleuWaegen and Goedhuys (2002) and the conjecture of

entrants learning in the initial years. Younger firms also have higher growth rates, export share and

investment rate. Entry therefore proves to be important for growth and capacity building.

Third, foreign ownership have positive effects across the board except for investment rate.

This can reflect a number of mechanisms, better access to finance and/or technological and

managerial know-how, for instance. The spillover effects of foreign ownership—as proxied by the

share of foreign ownership within a country-industry cell—is largely absent for growth, but positive

for export share. Foreign presence in an industry is, therefore, good for boosting export orientation.

This may reflect information sharing on export destination and product. Foreign presence in an

industry also reduce investment rate and labor productivity—which may reflect market stealing

effects of foreign entry (Aitken and Harrison 1999).

Fourth, a large ownership share of the largest owner is negatively related to sales and labor

growth. A plausible interpretation is that these firms cannot obtain external financing for their

expansion and have to resort to their own wealth for expansion, which lead to lower growth rates.

Finally, the industry competition measure increases investment rate but nothing else. The import

tariff level does not have any discernible effect on our five performance measures. We thus see no

strong indications that our proxies of competition matter much.

Infrastructure (phone density, average web usage, the customs efficiency, and the LPI

infrastructure index) is in general associated with higher productivity, faster labor growth, and higher

investment rate. Phone density and the custom efficiency do not matter much, but average local web

presence and the LPI infrastructure index do matter significantly.

Geography. The landlock dummy is only significantly associated with a higher investment

rate. A larger country (in terms of population) is associated with lower export intensity. This makes

sense because a larger country can enjoy a greater extent of scale economy, and firms thus face less

need to export to achieve economy of scale.

Government expropriation seems to slow down development. The number of years that the

ruling party has been in power is negatively associated with sales growth and investment rates,

reflecting the fear of entrepreneurs in expanding and investing when expropriation risks rise up.

Similarly, the local share of bribe payment in sales is negatively related to both sales and labor

growth rates (with the latter being statistically significant), signaling similar fears by firms. This

result is consistent with other studies in developing countries. Fisman and Svensson (2007), using

Uganda data, find that the negative effect of bribes is three times larger than that of taxes. Cai et al.

(forthcoming) find negative effects of corruption but not taxes in China.

128

Ethnic diversity. We find no direct effects per se. It is however useful to point out that this

variable could have affected the outcomes if it affects any of our other variables.

Crime and violence. Local areas with strong concerns over crime have lower sales growth

rate and lower export intensity. Countries that experienced domestic conflicts in the previous ten

years tend to have lower investment rates (significant) and lower sales growth and labor productivity

(insignificant). These findings suggest that a safe business environment is a prerequisite for local

development, affecting almost all aspects of firm performance. A surprising finding is that countries

experienced domestic conflicts tend to have higher export intensity. A possible reason is that

domestic conflicts destroyed domestic purchase power, and firms had to resort more to export

markets. Alternatively and perhaps more plausibly, natural resource exports may lead to higher

incidence of domestic conflicts, so the result merely reflect reverse causality.

Labor regulation. Firing costs are associated with higher labor productivity and investment

rates but lower sales and labor growth rates, likely reflect the substitution of capital for labor due to

higher labor adjustment costs. Still, given higher factor costs, firms‘ expansion likely would be

lower. So the results make good sense.

Financial access. Access to bank finance (i.e., overdraft facilities) is positively correlated

with labor productivity, yet it is also negatively correlated with export intensity. The positive

correlation of formal finance and productivity is consistent with the macro finance-growth literature

(Levine 1997). The negative correlation with export intensity suggests that formal financing do not

seem to have a robust relationship with positive economic outcomes. In sharp contrast, the

prevalence of trade credit has a stronger correlation with positive economic outcomes. It is

associated with higher labor productivity, higher investment rates, and higher export intensities.

Thus, informal finance may have played a stronger role in facilitating local development in this

sample of developing countries. This result is consistent with Fisman (2001), which uses data from

five African countries and finds that supplier credit is positively associated with capacity utilization,

even after being instrumented by supplier characteristics. Fisman provides evidence that access to

supplier credit allows for better inventory management and lowers chance of inventory shortage

Do the large Africa premiums reflect multicollinearity?

The large conditional African premiums for labor productivity and sales growth may surprise many.

Does the positive and significant coefficient for the Africa dummy merely capture the close

correlation of this dummy variable with some of the aggregate (national or local) variables? Table 2,

after all, does show that Africa differs significantly from the other regions in many important aspects.

To check whether the positive coefficient for the Africa dummy merely reflects its multicollinearity

with some of our key explanatory variables, we run a series of regressions for each of the dependent

variable, dropping one aggregate variable at one time, and see if the Africa dummy remains

significant and with similar magnitudes in the labor productivity and sales growth equations.

129

The results in Table 5 dismiss our concern about multicollinearity. For the labor productivity

equation, the Africa dummy remains significant and with similar magnitudes when dropping each of

the 14 variables we show in the table. For the sales growth equation, in 11 of the 14 cases of

dropping one variable at a time the Africa dummy retains a positive and significant coefficient, and

with similar magnitudes.

Controlling for country fixed effects

In our base specifications there may be many omitted country-level variables which may be

correlated with our key measures, leading to omitted variable bias for our firm and local level

variables. To deal with this issue, we control for country fixed effects. An advantage is that now we

control for all country-specific time-invariant variables. A disadvantage is that all country-level

variables have to be omitted from the specification. A further disadvantage is that if the main

variations of our key variables are at the country level, the country fixed effect specification would

throw the baby out with the bathwater, leaving out relationships that should be there if country-level

variations are exploited. With these caveats in mind, we present the country fixed-effects

specification (see Table 6).

The majority of key relationships remain intact, although some results become weakened in

statistical precision—not surprising since we have left out the key country-level variation. There are

several key differences with the OLS specification. First, the presence of foreign ownership in the

industry level is now positively correlated with sales growth. Thus, the spillover effects of foreign

ownership become stronger once we control for country FE. Second, the import tariff rate is

negatively correlated with labor productivity and investment rate, therefore international competition

is good for efficiency and firm expansion. International tariff rates are therefore negatively

correlated with country-level determinants of productivity and investment rate. Third, the industry-

level competition now has negative correlations with labor productivity and sales growth,

surprisingly. Fourth, the positive effects of local prevalence of web sites become even stronger.

Fifth, the effects of bribes on investment rates, negative, now becomes significant. Sixth, the positive

effects of trade credit become weaker for productivity and investment rates. Finally, the access to

formal finance becomes totally insignificant. Given our strong interest in the effects of country-level

variables and in finding about the coefficient of the Africa dummy, and the fact that the country-level

variations are likely the main source of variations for many of our key business environment

variables, our sense is that probably too much variations are lost once we control for country fixed

effects. But it is reassuring that most of our results remain similar, and a few results such as those on

corruption and tariffs become more reasonable.

Omitting other business environment variables?

As with any study with cross-sectional data, a legitimate concern is that the effects of our business

environment variables may merely capture those of omitted variables. To consider this scenario, we

examine the robustness of our key results when we include more controls of the business and macro

130

environments. In particular, we add the following key aspects: (i) openness (e.g., import and export

over GDP), (2) further aspects of political accountability and competition: voice and accountability

from ICRG, and executive competition from DPI, (3) further subjective measures of business

environment constraints, including the local share of firms viewing the following business

environment elements as moderate or severe constraint: electricity, transport, informal competition,

tax rate, land access, financial access, and labor regulation; (4) an objective measure of entry barrier,

namely, the logarithm of minimum capital to start a business (from Doing Business). Moreover, we

have also tried adding inflation rate or a dummy variable of high inflation rate to proxy for macro

risks, and have not found it matter. This is clearly a relative comprehensive list of other indicators of

the business environment.

Overall, the results about our old key variables are reasonably robust (see Table 7). Some

results lose or gain on statistical significance, but the gist of the findings is the same. Most of the

locally-perceived constraint variables are not statistically significant, and often they are wrong

signed. This is consistent with the recommendation that objective measures tend to out-perform

subjective measures of the business environment (Dethier et al. 2008). The entry-barrier measure,

the minimum capital to start a business, has a negative association with labor productivity and the

two growth measures. Thus high entry barriers deter firm efficiency and growth. Higher entry

barriers are also associated with higher export intensity.29

V. Further Considerations

In this section, we first provide some sensitivity checks, and then investigate how the effects

of the business environment differ in several key dimensions in which the business environment

might have distinct effects. We want to understand whether African firms have distinct conditional

performance advantage depending on their sizes, industries, and technologies.

Adding four rich African countries

So far our ―Africa‖ consists of SSA countries excluding the richest four countries (i.e., South Africa,

Namibia, Mauritius, and Botswana), all of which have GDP per capita exceeding 3000 U.S. dollars

(in 2005 value). Some may want to know how the whole SSA is doing. To shed light on this

curiosity, in this subsection only we define Africa to include the four rich SSA countries. The results

are in Table 8. Still, we have a sizable African advantage in labor productivity level by 144 log

points. However, Africa now features an export intensity disadvantage of 4.4 percent. The

coefficients of the other variables are very similar. Thus adding the four rich countries, while

29

Hallward-Dreimeier et al. (2010) suggest that the within-location variations in regulation discretion (as proxied by

things such as the standard deviation of managerial time spent on dealing with regulators) reduce firm performance.

We have tried this, and when we have few control variables, this variable behaves as their paper conjectures.

However, once we have our comprehensive list of variables, this variable no longer matter. This is not surprising

since this variable captures similar things as government expropriation.

131

supposedly would add to the Africa‘s conditional advantage, only reduces but far from eliminating

Africa‘s conditional advantage.

Using only the low income sample

Yet another concern is how the environmental variables affect firm performance may depend on the

level of development (Xu forthcoming). If low-income countries behave completely differently,

which is possible, for instance, in the O-Ring world featuring strong complementarity (Kremer

1993), then both our estimates of the business environment effects and the conditional African

advantages would be questionable. To ensure that the results we obtained are not driven by an

inappropriate sample, we limit our sample to the poor African countries and other countries in similar

income level (i.e., with a GDP per capita of less than 3000 USD in 2005 value). The results are in

Table 9.

Many results are quite similar as before. However, there are some differences, and we detail

them below. First, after using the low-income sample exclusively, Africa now features higher sales

growth rate (by 0.30), still higher productivity of about 200 log points but statistically insignificant

(with a t-statistic of 1.58), and lower export intensity (by 6.4 percentage points). Thus Africa still

features conditional performance advantage (though slightly weaker than before) than other similar

income countries. Africa does exhibit lower export intensity, even compared only with similar-

income countries. Second, the effect of competition becomes positive and marginally significant for

sales growth. So within the low income sample, competition has more positive effect. Third,

domestic conflicts tend to significantly increase labor growth but reduce sales growth—suggesting a

scenario in which domestic conflicts increase the growth of micro enterprises, resulting in higher

labor growth but lower sales growth due to micro enterprises‘ lack of increase in complementary

inputs.

Fourth, custom efficiency becomes significant in spurring investment. Fifth, export

transportation costs now are positive and significant for investment rate. Sixth, the negative effects of

firing costs on growth become insignificant, suggesting that firing costs in low-income countries tend

to be less binding and seriously enforced. Seventh, many of the infrastructure variables become less

significant (except the average usage of web sites for business). This is probably because

infrastructure tends to be uniformly poor for this set of countries, and lower variations lead to the

conclusion that it does not matter as much. Finally, access to formal financing is associated with a

lower investment rate in low-income countries. This suggests that the formal financing system in

these countries are likely inefficient—not surprising since the efficiency of the financial system

needs complementary institutions such as contract enforcement mechanisms, and poor countries tend

to lag behind in these complementary institutions.

Matching

132

So far we have compared the poor African countries with the rest of developing countries, or with

developing countries with similar income level. However, the set of countries with similar income

still have higher average income level. If we select observationally-similar countries with the poor

African countries, do we still observe the positive advantage of the poor African countries?

To answer this question, we adopt the matching estimate for the poor African dummy

variable. In other words, we rely on the matching estimates, for each poor African country, we find

another country in the non-poor-African countries that are the most similar in our controlled

covariates, and we obtain the average differences in our five outcomes. This would then show

whether the poor African countries are similar to other non-African countries that are observationally

similar.

Now, the matching estimates (see Table 10) show non-significant differences between the

African countries and similar countries.30 Thus, African countries are not different from other

countries with similar firm characteristics and political and business environments. This also implies

that our set of covariates of the business environment predict reasonably well firm performance such

that once our covariates are controlled for, there is no difference between sample African countries

and other similar countries.31

Manufacturing versus services

The main purpose of this paper is to examine how African manufacturing fares relative to other

regions and the determinants of manufacturing firms‘ performance. However, many may wonder

how the results may differ if we look at services. To be complete, and to understand the differences

between manufacturing and services, we estimate our base specifications for manufacturing and

services separately. The results are in Table 11. Indeed, there are significant sectoral differences.

There is conditional Africa advantage only in manufacturing. So Africa has more advantage

in manufacturing than in services (conditional on the business environment). Not surprisingly,

foreign ownership has no spillover in services but has such spillover in manufacturing. Ownership

concentration hurts growth only in manufacturing but not in services. Party monopoly reduces

investment in both manufacturing and services, but reduces sales growth in manufacturing while

increases labor growth in services. Thus party monopoly is associated with shifting a country‘s

sector structure from manufacturing to services, perhaps because assets in services are less specific

and expropriatable. Similarly, domestic conflicts also increase employment growth only in services

perhaps because services jobs tend to require less investment.

30

The matching was implemented by the one-to-one (nearest neighbor) matching method, and estimated with the

Stata‘s PSMATCH2 command. 31

The matching estimates do not automatically yield insignificant dummy treatment variables once our set of

covariates is controlled for. For instance, the dummy variable of any conflicts in the past ten years is still associated

with significantly lower labor productivity, sales growth and higher labor growth rates.

133

Firing costs lead to increase in investment rate in manufacturing only, indicating more

substitutability of capital for labor in manufacturing only. The cost of a standard container to export

port has a negative effect on export and investment rate only for manufacturing, and thus

transportation costs affects manufacturing more than services. Interestingly, services are more

affected by the availability of formal finance than manufacturing. Crime tends to have stronger

effects on services, especially on employment growth.

Thus most of the business environment variables have stronger effects in manufacturing than

in services, with a few exceptions such as crime and access to formal finance. Party monopoly tends

to shift resources from manufacturing to services and thus fundamentally alter a country‘s industrial

structure. Africa has conditional advantage only in manufacturing and not in services.

Low-tech and high-tech manufacturing

The conditional Africa advantages represent latent comparative advantage for Africa. Do the latent

Africa comparative advantages differ by the level of sophistication of particular manufacturing

industries? In addition, how does the business environment affect firm performance differently for

industries with distinct levels of sophistication? These questions are interesting since some recent

thinking of development argues for finding specific areas of comparative advantage for countries at

different stages of development (Lin 2009; Lin 2010; and Lin and Monga 2010), and industries with

varying level of technological sophistication and specific investments may need different supporting

institutions and business environments (Lin 2010, Xu forthcoming).

Since Africa is mainly located at the less sophisticated end of the technology spectrum,

Africa‘s natural comparative advantage likely would be low-tech rather than high-tech

manufacturing. We thus expect the conditional African advantage to be larger for the low-tech

manufacturing sector than for the high-tech one. Since the ES data have both low-tech and ―high-

tech‖ manufacturing, we are able to class the manufacturing sectors into ―the low-tech

manufacturing‖ (e.g., food and beverages, leather, wood processing and wood products, simple metal

products, textiles, and garments) and the ―high-tech‖ manufacturing (e.g., metal and machinery,

electronics, chemical and pharmaceutical products, non-metal and plastic, automobile and parts).32

This conjecture is confirmed (see Table 12). The conditional Africa advantage for labor

productivity is 265 log points for low-tech manufacturing, but around 100 log points lower--only 168

log points--for high-tech manufacturing. The conditional Africa advantage for sales growth is 23

percent (close to be significant, with a t-statistic of 1.63) for low-tech manufacturing, but is only 11

percent (insignificant, with a t-statistic of 0.92).

The effects of the business environment differ significantly for the low- and high-tech

manufacturing. Competition (e.g., the competition index) facilitates labor growth in low-tech

manufacturing, suggesting a scenario in which competition facilitates entry and firm expansion in

32

The data set has ―other manufacturing‖, which is not classified as either since we don‘t know its nature.

134

low-tech manufacturing. But there is no effect on export intensity or investment. In contrast, strong

competition in high-tech manufacturing is associated with a higher export intensity and investment

rate. For high-tech firms to stay viable in competitive industries, they have to boost their investment

to stay technologically competitive, and expand their markets through exporting, which is also made

possible by higher cost competitiveness associated with strong competition.

The dependence on modern infrastructure should be higher for high-tech manufacturing, and

this is born out. Log phone density and local web density are both more likely to be statistically

significant for high-tech manufacturing. Phone density does not matter at all for low-tech

manufacturing, but is positively associated with labor productivity and export intensity for high-tech

manufacturing. Similarly, local web density, only significant for labor growth for low-tech

manufacturing, is significant for high-tech manufacturing in both growth and investment.

Country size has positive effects only for labor productivity of high-tech manufacturing.

This suggests that a large domestic market may be needed to generate a sufficiently large market for

sophisticated products. This is consistent with the Engle effect, which implies that labor productivity

does not hinge on market size that much for low-end products: small local markets are usually

sufficient for producing such products, and there is usually little economy of scale for such products.

Party monopoly has particularly adverse effects of low-tech industries, and has no effects for

high-tech industries. Similarly, corruption is associated with a more adverse effect on low-tech than

on high-tech manufacturing. The negative magnitude on labor growth is larger for low-tech

manufacturing (-0.96 vs -0.55), and it has negative and significant effects on investment rates only

for low-tech manufacturing. Our interpretation is that firms in high-tech industries are larger and

have more resources to strike deals with government officials for favorable protection (Cai et al.

forthcoming; Hallward-Dreimeier et al. 2010; Hallward-Dreimeier and Pritchett 2010), and that

government expropriation thus has less pronounced effects for favored firms.

Domestic conflicts are associated with higher export share only for low-tech industries. A

plausible interpretation is that low-tech industries are more affiliated with natural resources, which

has a strong correlation with domestic conflicts. Crime has much more severe negative effects on

high-tech than for low-tech manufacturing. It has negative and significant effects on productivity,

growth rates and investment rates for high-tech manufacturing, but only negative effect on export

intensity for low-tech industries.

Firing costs are associated with more adverse effects for high-tech than for low-tech

manufacturing industries--stronger negative effects on the two growth measures, and no positive

effects on labor productivity for the high-tech industries. This makes good sense. Employees of

high-tech industries need more firm-specific investment and training for learning about the industries

and are therefore more expensive. As a result, such firms are more reluctant to hire when hiring

mistakes are more costly due to higher firing costs.

The effects of trade credit differ between the two sectors. It improves export intensity for

both sectors. For low tech manufacturing, the effects of trade credit center on the two growth rates,

135

so low-tech manufacturing firms are able to expand more with access to trade credit; but they do not

have higher labor productivity or investment rates. Trade credits therefore help low-tech

manufacturing to expand firm size without increasing productivity or physical investment. For high-

tech manufacturing, trade credit improves productivity and investment but not growth rates, due

perhaps to better inventory management (Fisman 2001), which improves productivity and increase

the ability to use more retained earnings for physical investment since trade credit reduces the need

for working capital.

Small and large firms

Firm size plays an important role in developing countries. Poorer countries in general and Africa in

particular are dominated by small and micro firms. Since firms with distinct sizes differ in their

ability to deal with government officials in buying protection, the effects of business environment

may differ (Beck et al. 2005, Cull and Xu 2005). We thus are keen to see how the performances of

small and large/medium firms are affected by the business environment separately, and whether the

conditional Africa advantages persist for firms of both sizes. To this end, we classify firms into

small firms (i.e., with the number of employees less than or equal to 10), and large and medium firms

or LMEs (i.e., with more than 10 employees). Relative to the two comparison groups, Africa is

swarmed with small firms. Its share of small firms, 39%, is higher than the average comparison

group by almost 20 percentage points (statistically significant), and higher than the better comparison

group by 26 percentage points.

Since Africa has many more small firms, it makes good sense that Africa‘s conditional

competitive advantage is larger for small firms (see Table 13)—its magnitude for labor productivity

is larger (2.84 vs. 1.96), that for sales growth is larger (47 vs. 20 percent), and that for labor growth is

also larger (11 percent vs. zero). But small African firms‘ conditional export share is lower by 2.5

percentage points. Thus African small firms are more domestically-oriented, but better than small

firms in other regions in what they do.

The size effects for productivity and export intensity exist only for LMEs. This finding

implies that performance and capacity would likely diverge for small firms and LMEs. The foreign

ownership effects and the spillover effects are all stronger for LMEs than for small firms, another

avenue in which LMEs could increase productivity over time that small firms cannot go.

Firing costs are more strongly and positively associated with productivity and not

significantly associated with growth for small firms than for LMEs. Only LMEs are hurt by firing

costs in terms of grow rates. This makes sense. High firing costs (usually only applied to LMEs)

lead more productive firms to stay under the radar of labor regulations and remain as small firms,

thus raising the mean productivity level. In the mean time, since they are less subject to labor

regulations, their expansion is not affected. Only for large firms the tradeoff between expansion and

productivity is apparent.

136

Corruption (i.e., bribe) has a negative and significant effect on labor growth only for LMEs.

This is not surprising since small firms do not need government inputs such as permits, licenses, or

advanced infrastructure to operate, so they are less subject to the influence of government corruption

than LMEs.

To summarize, relative to other regions, Africa‘s small firms has higher conditional

performance advantage than LMEs and are therefore more competitive internationally. LMEs enjoy

higher economy of scale, direct foreign ownership benefits, and are more hurt by corruption and

labor regulations.

VI. Explaining Africa’s Disadvantage from the Better Comparison Group

We now examine why Africa falls behind the better comparison group—by reporting βX

(XAfrica – Xbetter), where ―better‖ stands for the better comparison group, and the percentage of

outcome differences due to each element. βX refers to the coefficient of a generic variable X—the

coefficients are drawn from our base specifications in Table 4—while XAfrica and Xbetter refer to the

mean for Africa and for the better comparison group. For simplicity, only those quantitatively and

statistically significant ones are reported. The results are in Table 14, and summarized more

succinctly (in broader category) in Table 15.

Foreign ownership (including both the direct and the spillover effects) plays a mixed role for

Africa. In a correlation sense, its difference with the better comparison group in foreign ownership

explains 18% of its labor productivity disadvantage (relative to the better comparison group), 4% of

the investment rate deficit of Africa. On the positive side, Africa‘s high foreign ownership and its

spillover effects reduce the African disadvantage in sales growth by 14%, and reduce its

disadvantage in export intensity by 4%. Ownership concentration has important explanatory power

for African manufacturing performance. Its higher value explains 47% of its sales growth deficit (to

the better comparison group), and reduces Africa‘s labor growth advantage by 12%. Younger firm

age in Africa explains 11% of Africa‘s productivity deficit, 6% of its investment rate deficit. It also

contributes to 26% of Africa‘s employment growth premium, and reduces 23% of its sales growth

deficit. Industry-level competition has a marginal explanatory power for the investment rate deficit

of Africa: only 4%.

Infrastructure (e.g., adding up the effects of phone density, web density, the LPI

infrastructure index, and customs) is a major hindrance for Africa, contributing to 277% of the

African manufacturing‘s disadvantage in labor productivity, -137% of the African advantage in labor

growth, and 36% of the African manufacturing deficit in investment rate.

The risks of government expropriation exert large negative effects. Party monopoly proves to

be an important negative factor for Africa, contributing to 188% of the African disadvantage in sales

growth, and 9% of the African investment rate deficit. Similarly, corruption reduces 19% of the

African advantage in employment growth.

137

Firing costs involve a clear tradeoff. One the one hand, relative high firing costs in Africa

explains 27% of Africa‘s sales growth deficit. On the other hand, high firing costs are also

associated with a reduction in the disadvantage in labor productivity for Africa.

Both formal and informal finance contribute to Africa‘s disadvantage, especially informal

finance. For formal finance, it explains 30% of the African manufacturing disadvantage in labor

productivity. For informal finance, it explains 178% of the African disadvantage in labor

productivity, 31% of its disadvantage in export share, and 29% of its disadvantage in investment rate.

Thus deficits in trade credit prevalence in Africa turn out to be quite important.

Export transport costs explain 15% of Africa manufacturing disadvantage in export share,

and 13% of the investment rate deficit. Crime proves to be a hindering factor as well, explaining

80% of its disadvantage in sales growth, and 4% of its disadvantage in export share.

Current or initial sizes contributes to 75% of Africa‘s disadvantage in labor productivity,

44% of its disadvantage in export share, and 16% of its disadvantage in investment rate. However,

smaller size in Africa also means higher growth rate due to mean convergence—it contributes to

280% of the reduction in Africa manufacturing disadvantage in sales growth, and 125% of its

advantage in labor growth.

VII. Conclusions

In this paper we try to understand how Africa stack up against other regions in firm

performances, whether Africa‘s business environment can explain its performance shortfalls, and

what the key factors are for explaining Africa‘s performance shortfalls. Using a comprehensive firm-

level data set, we find that formal African manufacturing firms show slight disadvantage in

productivity, similar sales growth rate, and advantage in labor growth; they also have a large shortfall

in export orientation and investment rate. So the manufacturing performance differences for Africa

are structural in the sense that they are due mainly to export capacity and low investment. African

manufacturing firms also tend to be much smaller than in similar income countries.

However, once we control for key differences in firm size, ownership, competition, infrastructure,

transport costs, government expropriation, labor flexibility, finance, and crime, Africa manufacturing

actually has conditional advantage in productivity and sales growth, and the conditional advantage

remain intact with more control of business environment factors, the dealing of multicollinearity,

using only the similar income sample, within a certain size category, and within a technology class.

Moreover, the conditional advantage of Africa is higher in small firms than in large and medium

firms, in low-tech than in high-tech manufacturing, in manufacturing than in services.

Overwhelmingly important for explaining Africa‘s disadvantage in firm performance relative to

better-performing similar income countries are firm size, infrastructure, government expropriation,

access to informal finance, and crime. Except for informal finance, this list represents the basic roles

of the government: property rights protection, public goods for trade (i.e., infrastructure), a safe

138

environment, and a regulatory environment that does not hinder firm growth. Tariff rates, labor

flexibility and access to formal finance are less important.

The findings in this paper have direct policy implications. The first implication is quite

optimistic for Africa. The results indicate that conditional on fixing the business environment (that

we have listed), Africa actually has productivity and sales growth advantages—at least no

disadvantage. So there is no inherent Africa curse that hinders its development. This is consistent

with Africa‘s growth record before the 1970s and the growth record in the past decade. Second, our

results show that Africa‘s comparative advantage is in low-tech manufacturing (instead of high-tech

manufacturing or services). Third, since the key business environment variables that explain the

African shortfall to the better-performing similar countries are mostly related to the functions of the

state—such as infrastructure, government expropriation, crime, firm size (which has everything to do

with government regulations)—the top priority for improving African competitiveness should start

with reforming the state.

Reforming the state is of course the most difficult reforms. Some have argued that African

states tend to be weak, and have to constantly please narrow-based ethnic or kinship networks or

clientelist demand to stay in power (Platteau 2009; Keefer 2010). Such states are therefore difficult

to implement strong reforms; nor do they have strong incentives for such reforms. How to

implement such reforms would therefore be major challenges and topics for future research.

139

Reference

Acemouglu, Daron, and Simon Johnson. 2005. "Unbundling Institutions." Journal of Political Economy

113(5), 949–95.

Acemouglu, Daron, Simon Johnson, and James Robinson. 2001. "The Colonial Origins of Comparative

Development: An Empirical Investigation.‖ American Economic Review 91, 1369–401.

Aghion, P., N. Bloom, R. Blundell, R Griffith, P. Howitt. 2005. ―Competition and Innovation: An

Inverted-U Relationship.‖ Quarterly Journal of Economics 120, 701-728.

Almeida, Rita. 2005. "Enforcement of Regulation, Informal Labor and Firm Performance." IZA

(Institute for the study of labor) Discussion Paper 1759.

Almeida, Rita, and Pedro Carneiro. 2009. ―Enforcement of Labor Regulation and Firm Size.‖ Journal of

Comparative Economics 37, 28–46.

Amin, Muhammad. 2009a. ―Labor Regulation and Employment in India‘s Retail Stores.‖ Journal of

Comparative Economics 37, 47–61.

———. 2009b. ―Are Labor Regulations Driving Computer Usage in India‘s Retail Stores?‖ Economics

Letter 102, 45–8.

Arvis, Jean-Francois, Monica Alina Mustra, Lauri Ojala, Ben Shepherd, and Daniel Saslavsky. 2010.

Connecting to Compete 2010: The Logistics Performance Index and Its Indicators. World Bank.

Aitken, Brian, Ann Harrison. 1999. ―Do Domestic Firms Benefit from Direct Foreign Investment?

Evidence from Venezuela.‖ American Economic Review 89, 605-618.

Aterido, Reyes and Hallward-Dreimeier, Mary. 2010. ―The Impact of the Investment Climate on

Employment Growth‖ World Bank Policy Research Working Paper 5218.

Aterido, R. Hallward-Driemeier, M., and Pages, C. Forthcoming. ―Big Constraints to Small Firms‘

Growth.‖ Economic Development and Cultural Change.

Beck, Thorsten, George Clarke, Alberto Groff, Philip Keefer and Patrick Walsh. 2001. ―New Tools and

New Tests in Comparative Political Economy: the Database of Political Institutions‖. World Bank

Economic Review 15 (September): 165-176.

Beck, Thorsten, Asli Demirguc-Kunt, and Vojislav Maksimovic. 2005. "Financial and Legal Constraints

to Growth: Does Firm Size Matter?" Journal of Finance LX, 137–77.

Blanke, Jennifer, and Xavier Sala-i-Martin. 2009. ―Examining Africa‘s Competitiveness.‖ In The

Africa Competitiveness Report 2009, World Economic Forum.

Bigsten, A.,Collier, P., Dercon, S., Fafchamps, M., Gauthier, B., Gunning, J.W., Oduro, A., Oostendorp,

R., Pattillo, C., S–derbom, M., Teal, F., and Zeufa. A, 2003. "Risk Sharing in Labor Markets,"

World Bank Economic Review, Oxford University Press 17(3), 349-366.

Bigsten, A., P. Collier, S. Dercon, M. Fafchamps, B.Gauthier, J.W. Gunning, J. Habarurema, A.

Isaksson, A. Oduro, R. Oostendorp, C. Pattillo, M. Söderbom, F.Teal, A. Zeufack .1999.

―Exports of African Manufactures: Macro Policy and Firm Behaviour‖. Journal of International

Trade and Development, 8, 1, 53-71.

Bigsten, A., Collier, P., Dercon, S., Fafchamps, M., Gauthier, B., Gunning, J.W., Oduro, A.,

Oostendorp, R.,Pattillo, C., Soderbom, M., Teal, F., Zeufack, A., 2004. Do African

manufacturing firms learn from exporting. Journal of Development Studies 40 (3), 115– 141.

Bigsten, Arne, and Mans Soderbom. 2006. ―What Have We Learned from a Decade of Manufacturing

Enterprise Surveys in Africa?‖ World Bank Research Observer 21, 241–65.

Botero, Juan, Simeon Djankov, Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Shleifer. 2004.

"The Regulation of Labor." Quarterly Journal of Economics 119(4), 1339-1382.

140

Cai, Hongbin, Hanming Fang, and Lixin Colin Xu. Forthcoming. ―Eat, Drink, Firms, Government: An

Investigation of Corruption from Entertainment and Travel Costs of Chinese Firms.‖ Journal of

Law and Economics.

Claessens, Stijn, and Luc Laeven. 2003. "Financial Development, Property Rights and Growth." Journal

of Finance 58(6), 2401–36.

Collier, Paul, and Jan Willem Gunning. 1999a. ―Why Has Africa Grown Slowly?‖ Journal of Economic

Perspectives 12(3): 3-22.

_____. 1999b. ―Explaining African Economic Performance‖ Journal of Economic Literature 37: 64-111

Cull, Robert, and Lixin Colin Xu. 2005. ―Institutions, Ownership, and Finance: The Determinants of

Investment among Chinese Firms.‖ Journal of Financial Economics 77, 117–46.

Davis, Steven and Haltiwanger, John. 1995. "Employer Size and The Wage Structure in U.S.

Manufacturing," NBER Working Papers 5393.

Demirguc-Kunt, Asli, and Vojislav Maksimovic. 1998. "Law, Finance, and Firm Growth." Journal of

Finance 53, 2107–37.

Demirguc-Kunt, Asli, Inessa Love, and Vojislav Maksimovic. 2006. ―Business Environment and the

Incorporation Decision.‖ Journal of Banking and Finance 30, 2967–93.

Dethier, Jean-Jacques, Maximilian Hirn, and Stephane Straub. 2008. ―Explaining Enterprise

Performance in Developing Countries with Business Climate Survey Data.‖ World Bank Policy

Research Working Paper 4792.

Devarajan, Shantayanan, William R. Easterly, and Howard Pack. 2001. ―Is Investment in Africa

Too Low or Too High?‖ Journal of African Economies 10: 81–108.

Dinh, Hinh, Mavridis, Dimitris and Nguyen, Hoa B. ―The Binding Constriant on Firm‘s Growth in

Developing Countries‖. World Bank Policy Research Working Paper 5485.

Djankov, Simeon. 2009. ―The Regulation of Entry: A Survey.‖ World Bank Research Observer 24,

183–203.

Dollar, David, Mary Hallward-Driemeier, Taye Mengistae. 2005. ―Investment Climate and Firm

Performance in Developing Economies.‖ Economic Development and Cultural Change 54(1),

1-31.

———. 2006. ―Investment Climate and International Integration.‖ World Development 34(9), 1498–

516.

Easterly, William, Ross Levine. 1997. ―Africa's Growth Tragedy: Policies and Ethnic Divisions.

―Quarterly Journal of Economics 112(4), 1203-50.

Fafchamps, Marcel. 2000. ―Ethnicity and Credit in African Manufacutring.‖ Journal of Development

Economics 61, 205-35.

Fisman, Raymond. 1999. "Trade Credit and Productive Efficiency in Developing Economies." World

Development 29(2), 311-21.

Fisman, Raymond, and Jacob Svensson. 2007. ―Are Corruption and Taxation Really Harmful to

Growth? Firm level evidence.‖ Journal of Development Economics 83(1), 63–75.

Gauthier, Bernard, and Mark Gersovitz. 1997. ―Revenue Erosion through Evasion and

Exemptions in Cameroon, 1993.‖ Journal of Public Economics 64: 407–24.

Gehlbach, Scott, Philip Keefer. 2010. ―Private Investment and the Institutionalization of Collective

Action in Autocracies: Ruling Parties and Legislatures.‖ Mimeo, World Bank.

141

-------------------------------------. 2010. ―Investment without Democracy: Ruling Parties

Institutionalization and Credible Commitment in Autocracies.‖ Mimeo, World Bank.

Gleditsch, Nils Petter, Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg & Håvard Strand

(2002). Armed Conflict 1946–2001: A New Dataset. Journal of Peace Research 39(5): 615–

637.

Griliches, Z., and J. Hausman. 1986. ―Errors in Variables in Panel Data Find More Like This.‖ Journal

of Econometrics 31, 93–118.

Gunning, J.Willem and Mengistae, Taye. 2001. ―Determinants of African manufacturing Investment:

the Microeconomic Evidence.‖ Journal of African Economies 10 (supp. 2), 48–80

Gyimah-Brempong, Kwabena, and Thomas L. Traynor, 1999. ―Political Instability, Investment,

and Economic Growth in Sub-Saharan Africa,‖ Journal of African Economies 8(1) , 52-86.

Hallward-Driemeier, Mary, Scott Wallsten, Lixin Colin Xu. 2006. "The Investment Climate and the

Firm: Firm-Level Evidence from China" Economics of Transition, 13(1), 1-24.

Hallward-Driemeier, Mary, Gita Khun-Jush, and Lant Pritchett. 2010. ―Deals versus Rules: Uncertainty

in Policy Implication in Africa.‖ NBER (National Bureau of Economic Research) Working

Paper 16001.

Hallward-Driemeier, Mary, and Lant Pritchett. 2010. ―How Business is Done and the `Doing Business‘

Indicators: The Investment Climate when Firms have Climate Control‖. Mimeo, World Bank

and Harvard Kennedy School of Government.

Harrison, Ann. 2010. ―Multinationals and Anti-Sweatshop Activism.‖ American Economic Review

100(1), 247–73.

Hausman, R., Dani Rodrik, and Andres Velasco. 2005. ―Growth Diagnostics.‖ In J. Stiglitz and N. Serra,

eds., The Washington Consensus Reconsidered: Towards a New Global Governance. New York:

Oxford University Press.

Iarossi, Giuseppe. 2009. ―Benchmarking Africa‘s Costs and Competitiveness.‖ In The Africa

Competitiveness Report 2009, World Economic Forum.

Johnson, Simon, John McMillan, and Christopher Woodruff. 2002. ―Property Rights and Finance.‖

American Economic Review 92(5), 1335–56.

Keefer, Philip. 2007. DPI2006. Database of Political Institutions: Changes and Variable Definitions.

Development Research Group, World Bank.

-----------. 2010. ―Do Ethnic Parties Represent Ethnic Interests? Ethnic Fractionalization, Clientelism and

Collective Action in Sub-Saharan Africa.‖ Working paper, World Bank.

Knack, Stephen and Keefer, Philip.1995. "Institutions and Economic Performance: Cross- Country Tests

Using Alternative Institutional Measures" Economics and Politics 7(3):207-27

Kremer, Michael, 1993. "The O-Ring Theory of Economic Development," The Quarterly Journal of

Economics 108(3), 551-75.

Levine, Ross. 1997. ―Financial Development and Economic Growth: Views and Agenda.‖ Journal of

Economic Literature.

Leuven, E., and B. Sianesi. 2003. "PSMATCH2: Stata module to perform full Mahalanobis and

propensity score matching, common support graphing, and covariate imbalance testing".

http://ideas.repec.org/c/boc/bocode/s432001.html. Version 1.2.3.

Lin, Justin Yifu. 2009. Economic Development and Transition: Thought, Strategy, and Viability. New

York: Cambridge University Press.

142

--------------. 2010. New Structural Economics: A Framework for Rethinking Development. World Bank

Policy Research Paper no. 5197.

------------ and Monga Celestin. 2010. ―Growth Identification and Facilitation: The Role of the State in

the Dynamics of Structural Change.‖ World Bank Policy Research Working Paper 5313.

Maddison, A.. 2001. The World Economy: A Millennial Perspective, Paris, OECD.

Mengistae, Taye and Catherine Pattillo. 2004. ―Export Orientation and Productivity in Subsaharan

Africa.‖ IMF Staff Papers 51, 327-353.

Moulton, Brent. 1990. ―An Illustration of a Pitfall in Estimating the Effects of Aggregate Variables on

Micro Units.‖ Review of Economics and Statistics 72, 334-38.

North, Douglass C. 1990. Institutions, Institutional Change and Economic Performance. Cambridge: Cambridge University

Press.

I.1 Palmade, Vincent. 2005. “Industry level analysis : the way to identify the binding constraints to

economic growth.‖ World Bank Policy Research Working Paper, 3551.

Petrin, Amil, and Jagadeesh Sivadasan. 2006. "Job Security Does Affect Economic Efficiency: Theory, A New Statistic,

and Evidence from Chile," processed, University of Minnesota.

Platteau, Jean-Philippe, 2009. "Institutional obstacles to African Economic Development: State, Ethnicity,

and Custom", Journal of Economic Behavior and Organization 71(3), 669-89.

Richmond, Christine and Klapper, Leora. 2010. ―The Political Economy of Firm Size.‖ Working Paper.

Sachs, Jeffrey D., Andrew M. Warner. 1997. ―Sources of Slow Growth in African Economies.‖ Journal

of African Economies 6(3), 335-76.

Schiantarelli, Fabio. 2010. ―Product Market Regulation and Macroeconomic Performance.‖ In Loayza

and Serve eds., Business Regulation and Economic Performance, Washington D.C.: The World

Bank.

Sleuwaegen, Leo., and Goedhuys, Micheline. 2002. ―Growth of Firms in Developing Countries, Evidence

from Côte d‘Ivoire.‖ Journal of Development Economics 68 (1): 117–35.

Tybout, James. 2000. ―Manufacturing Firms in Developing Countries: How Well Do They Do and Why,‖

Journal of Economic Literature 38(1), 11-44.

UCDP/PRIO. 2010. ―UCDP/PRIO Armed Conflict Dataset Codebook.‖ www.prio.no/cscw.

Van Biesebroeck, Johannes.2005. Exporting raises productivity in sub-Saharan African manufacturing

firms, Journal of International Economics 67(2), 373-391.

World Bank. 2005. A Better Investment Climate for Everyone. World Development Report, World Bank.

------ The World Bank Enterprise Survey Database. www.enterprisesurvey.org. Washington, D.C.

United States.

------ Doing Business. www.doingbusiness.org.

World Economic Forum. 2009. The Africa Competitive Report 2009. Washington D.C.: World Bank.

Xu, Lixin Colin. 2011. "The Effects of Business Environments on Development: A Survey of New Firm-

Level Evidence,‖ World Bank Research Observer.

http://www.prio.no/cscw

http://ideas.repec.org/s/eee/inecon.html

http://www.enterprisesurvey.org/

http://www.doingbusiness.org/

143

Table 1. Definition of variables

Variable Definitions

ln(LP) Log(value added per worker). From the Enterprise Survey of the World Bank (ES

hereafter).

sale growth (Three-year) sales growth rate at the firm level, computed as (sales this year – sales

three years ago)/their average. Based on suggestions from Davis and Haltiwanger

(1992) to reduce the influence of outliers. All monetary values are in this paper is in

constant U.S. dollars in 2005. Based on ES.

L growth (Three-year) employment growth rate at the firm level, computed as above. Based on

ES.

export share Export value over sales. Based on ES.

Inv/VA The value of investment over total value added for a firm. Investment is the sum of

new purchase in equipment and land. Based on ES.

lnL The logarithm of the number of employees. Based on ES.

Firm age Firm age. Based on ES.

Foreign The share of foreign ownership. Based on ES.

Ind avg Foreign Country-industry (employment-weighted) average of foreign ownership.

Own1 The ownership share of the largest owner of the firm. From ES.

Gdppc GDP per capita in constant U.S. dollars. From WDI.

Urban The share of urban in total population in a country. From WDI.

Ln(pop) Log of the population of the country. From WDI.

Trade The ratio of the sum of the values of imports and exports over GDP. From WDI.

Ln(phone density) Log of the number of mainline phones per 10000 inhabitants in a country. From WDI.

Landlock The dummy variable indicating that a country is landlocked.

VoiceAcc The index of voice and accountability, measuring perceptions of the extent to which a

country's citizens are able to participate in selecting their government, as well as

freedom of expression, freedom of association, and a free media. From ICRG.

Ln(partyYears) Logarithm of the number of years that ruling party has been in power, from DPI (Beck

et al. 2001; Keefer 2007).

anyConflict The dummy variable of any minor or major domestic conflicts in the previous 10 years,

from Gleditsch et al. (2002) and UCDP/PRIO (2010).

EthnicFrac Ethnic fractionalization, a larger value implies more ethnic diversity, from WDI.

FireCost An index measuring the cost of firing 20 percent of the firm‘s workers (10% are fired

for redundancy and 10% without cause). The cost of firing a worker is calculated as the

sum of the notice period, severance pay, and any mandatory penalties established by

law or mandatory collective agreements for a worker with three years of tenure with the

firm. Based on Doing Business (Botero et al. 2004).

MinK A proxy of entry barrier, the minimum capital required (% of income per capita) for

starting a business, from Doing Business of World Bank.

Web The city-industry-size share of firms using website for business. Size is defined as

large and small based on whether the firm is above or below the median of the cell.

144

When the cell size falls below 5, we use the city-industry mean instead. All proxies for

business environment based on firm answers are dealt with in the same way as here.

Table 1. Definition of variables (Cont’d)

Variable Definitions

Bribe The city-industry-size average of the share of bribes to governments over sales. From

ES.

Bank The city-industry-size share of firms having access to overdraft facility, as a proxy of

access to formal finance.

TradeCredit The city-industry-size average of the share of output sold in the form of supplier credit,

as a proxy of local access to informal finance.

InformalCompete The city-industry-size share of firms viewing ―practices of competitors in the informal

sector‖ as a serious obstacle. Based on ES.

Cons_landAccess The city-industry-size share of local firms that view access to land as a moderate or

severe constraint.

Cons_crime The city-industry-size share of firms that view crime as a moderate or severe constraint.

Based on ES.

Cons_electricity The city-industry-size share of firms that view electricity access as a moderate or severe

constraint. Based on ES.

Tariff Country-industry-year level of import tariffs. From WITS.

costExport Transport cost to export (in US$ per container), from DB.

Customs Efficiency of the clearing process (i.e., speed, simplicity, and predictability of

formalities) by border control agencies, including customs. From World Bank.

Infrastructure Quality of trade and transport related to infrastructure (e.g., ports, roads, railroads,

information technology). From World Bank.

CompetitionInd The Lerner index of competition, constructed as (1 – markupCI), and markupCI is the

country-industry average of firm-level markup. Firm-level markup is computed as

(value added – labor costs)/sales.

145

Table 2. Differences between Africa and the two comparison groups

Variable t-test for mean difference

Africa Africa – the average

comparison group

Africa – the better

comparison group

ln(LP) 9.221 (0.036) -0.054 (0.036)* -0.309 (0.044)***

sale growth 0.066 (0.010) -0.008 (0.014) -0.004 (0.015)

L growth 0.188 (0.006) 0.068 (0.008)*** 0.074 (0.009) ***

export share 0.074 (0.003) -0.086 (0.005)*** -0.147 (0.006)***

Inv/VA 0.134 (0.005) -0.041 (0.008)*** -0.107 (0.010)***

lnL 2.958 (0.019) -0.698 (0.025)*** -1.045 (0.029)***

Firm age 13.116 (0.201) -4.894 (0.300)*** -3.323 (0.332)***

Foreign 0.156 (0.005) 0.058 (0.005)*** 0.027 (0.007)***

Own1 0.822 (0.004) 0.082 (0.005)*** 0.119 (0.006)***

GDP per capita 488.1 (7.680) -1061 (13.8)*** -1073 (15.4)***

Urban 0.326 (0.003) -0.213 (0.002)*** -0.208 (0.003)***

Ln(pop) 16.134 (0.016) -0.601 (0.023)*** -1.146 (0.028)***

Infla 77.865 (3.395) 66.763 (2.251)*** 65.655 (3.341)***

Trade 72.468 (0.546) -8.080 (0.577)*** -18.061 (0.721)***

Ln(phone density) 3.042 (0.014) -2.089 (0.014)*** -2.261 (0.018)***

Landlock 0.249 (0.006) 0.005 (0.008) 0.075 (0.009) ***

VoiceAcc -0.700 (0.009) -0.293 (0.010)*** -0.235 (0.014)***

Ln(partyYears) 2.528 (0.013) 0.464 (0.023)*** 0.144 (0.024)***

anyConflict 0.504 (0.007) 0.095 (0.009)*** -0.065 (0.011)***

EthnicFrac 0.716 (0.002) 0.316 (0.003)*** 0.387 (0.003)***

Web 0.133 (0.002) -0.274 (0.003)*** -0.355 (0.004)***

Bribe 0.029 (0.0004) 0.013 (0.0005)*** 0.019 (0.0005)***

Cons_landAccess 0.357 (0.002) 0.061 (0.004)*** 0.063 (0.004)***

InformalCompete 0.693 (0.005) 0.287 (0.005)*** 0.284 (0.005)***

Cons_crime 0.385 (0.003) -0.001 (0.004) 0.128 (0.004)***

FireCosts 0.405 (0.003) 0.098 (0.003)*** 0.032 (0.004)***

tradeCredit 0.274 (0.002) -0.265 (0.003)*** -0.320 (0.004)***

Bank 0.234 (0.003) -0.218 (0.005)*** -0.092 (0.005)***

mink_sb 208.520 (4.903) 173.0 (3.45)*** 188.2 (4.76)***

Cons_electricity 0.756 (0.004) 0.230 (0.004)*** 0.335 (0.005)***

competitionInd 0.600 (0.003) -0.015 (0.002)*** -0.029 (0.003)***

Tariff 0.146 (0.001) 0.054 (0.001)*** 0.052 (0.001)***

costExport 1376.82 (13.27) 116.83 (13.30)*** 282.29 (16.31)***

Customs 2.199 (0.006) -0.168 (0.005)*** -0.232 (0.007)***

Infrastructure 2.112 (0.006) -0.295 (0.005)*** -0.319 (0.006)***

146

One-sided test. *, ** and *** means significance at the 10, 5 and 1 percent levels.

Table 3. Industry distribution for Africa, the average comparison group, and the better comparison

group

Africa

(with the four

rich countries)

Africa

(without the four rich

countries)

The average

comparison group

The better

comparison group

textile 0.022 0.018 0.066 0.056

garment 0.140 0.136 0.164 0.162

Metal & Machinery 0.084 0.065 0.105 0.143

electronics 0.009 0.006 0.026 0.047

Chemical,

pharmaceutic

al

0.051 0.040 0.103 0.091

Wood and furniture 0.015 0.019 0.023 0.033

Non-metal and plastic 0.041 0.041 0.103 0.145

Automobile and parts 0.000 0.000 0.002 0.003

Other manufacturing 0.408 0.436 0.174 0.114

Note. The manufacturing sample only.

147

Table 4. Explaining Firm Performance, OLS results

ln(LP) sale growth L growth export share inv/VA

Poor Africa 2.170** 0.210* 0.040 -0.007 -0.005

(2.053) (1.763) (1.014) (-0.252) (-0.121)

Ln(L) 0.123*** 0.063*** 0.034***

(3.990) (8.445) (5.566)

Ln(firm age) 0.083*** -0.027** -0.073*** -0.020*** -0.029***

(2.634) (-2.195) (-9.435) (-3.649) (-7.083)

foreign 0.606*** 0.142*** 0.059*** 0.135*** -0.003

(6.972) (3.754) (2.901) (5.167) (-0.243)

Country-industry avg foreign -1.246* 0.025 0.022 0.166*** -0.112***

(-1.881) (0.267) (0.631) (3.658) (-3.096)

Own1 -0.156 -0.105*** -0.064*** -0.000 0.010

(-1.286) (-3.544) (-3.172) (-0.011) (0.744)

Tariff 1.158 0.240 0.104 0.032 0.036

(1.095) (1.108) (1.116) (0.217) (0.433)

CompetitionInd -0.127 -0.033 0.034 0.034 0.158***

(-0.152) (-0.279) (0.846) (0.871) (2.700)

Ln(phone density) 0.227 0.010 -0.012 0.017 -0.010

(1.043) (0.203) (-0.687) (1.027) (-0.621)

Ln(population) 0.150 0.022 -0.002 -0.030*** 0.013

(0.868) (0.659) (-0.197) (-3.632) (1.383)

Landlock dummy 0.139 -0.057 -0.028 0.010 0.056**

(0.471) (-0.578) (-1.233) (0.580) (2.306)

Ln(number of years the ruling party in

power) -0.513 -0.117** 0.016 -0.013 -0.022*

(-1.447) (-2.088) (1.485) (-1.291) (-1.960)

Any conflict in the past 10 years -0.974 -0.113 0.012 0.044** -0.064***

(-1.621) (-1.329) (0.476) (2.383) (-2.937)

Ethnic fractionalization -0.551 0.062 0.059 -0.002 -0.002

(-0.450) (0.282) (0.790) (-0.035) (-0.050)

Firing difficulty 1.288** -0.227** -0.076** -0.013 0.063*

(1.964) (-2.021) (-2.109) (-0.435) (1.735)

Ln(CostExport) 0.185 -0.042 0.041 -0.085*** -0.051*

(0.610) (-0.666) (1.092) (-4.046) (-1.742)

Custom efficiency index -0.867 0.037 -0.035 -0.006 -0.023

(-1.416) (0.346) (-0.723) (-0.168) (-0.549)

Infrastructure index 1.485* 0.021 0.077* -0.053 -0.062

(1.774) (0.164) (1.839) (-1.155) (-1.187)

Web 0.443 0.163 0.189*** 0.026 0.109***

(0.589) (1.226) (4.602) (0.704) (2.648)

148

Table 4. Explaining Firm Performance, OLS results (Cont’d)


Bribe -0.051 -0.823 -0.696*** 0.028 -0.129

(-0.022) (-1.416) (-3.263) (0.126) (-0.611)

Trade credit 0.951** 0.096 0.035 0.149*** 0.098**

(1.999) (1.048) (0.751) (4.075) (2.134)

bank 0.563* 0.115 0.021 -0.061** 0.002

(1.780) (1.145) (0.588) (-2.191) (0.054)

cons_crime -0.849 -0.168* 0.005 -0.053* -0.063

(-1.498) (-1.918) (0.138) (-1.834) (-1.303)

Ln(sales 3 years ago) -0.073***

(-6.927)

Ln(L 3 years ago) -0.081***

(-13.861)

Ln(sales) -0.000

(-0.032)

Ln(value added) -0.020***

(-3.979)

N 12,197 10,363 12,125 12,289 9,602

adjusted R squared 0.307 0.117 0.112 0.248 0.050

Note. *, **, and *** represent statistical significance at the 10, 5 and 1 percent levels. Standard errors

are clustered at the country level.

149

Table 5. The coefficient for Africa when dropping one explanatory variable at a time

ln(LP) sale Growth labor growth export share inv/VA

base line 2.170** 0.210* 0.040 -0.007 -0.005

(2.053) (1.763) (1.014) (-0.252) (-0.121)

no ln(party years) 1.873** 0.147 0.049 -0.016 -0.020

(1.967) (1.276) (1.269) (-0.611) (-0.600)

no ln(phone density) 1.979* 0.202* 0.050 -0.022 0.004

(1.801) (1.768) (1.219) (-0.690) (0.113)

no "conflicts in past 10 yrs" 2.102* 0.202 0.041 -0.001 -0.012

(1.769) (1.586) (1.032) (-0.031) (-0.286)

no ethnic fractionalization 2.078** 0.220** 0.049 -0.008 -0.005

(2.342) (1.980) (1.442) (-0.288) (-0.136)

no fire costs 2.116** 0.205* 0.038 -0.008 -0.006

(1.983) (1.916) (1.005) (-0.281) (-0.142)

no ln(export transport costs) 2.178** 0.209* 0.043 -0.010 -0.006

(2.058) (1.723) (1.104) (-0.322) (-0.164)

no custom efficiency 2.245** 0.208* 0.043 -0.007 -0.003

(2.113) (1.732) (1.097) (-0.236) (-0.080)

no infrastructure index 2.254** 0.212* 0.045 -0.009 -0.006

(2.085) (1.770) (1.120) (-0.281) (-0.148)

no local web intensity 2.229** 0.227* 0.063 -0.003 0.006

(2.275) (1.932) (1.479) (-0.097) (0.142)

no bribe 2.170** 0.205* 0.037 -0.007 -0.005

(2.051) (1.703) (0.910) (-0.249) (-0.129)

no trade credit 2.146** 0.207* 0.039 -0.013 -0.006

(2.036) (1.717) (0.961) (-0.424) (-0.141)

no bank 2.165** 0.215* 0.040 -0.007 -0.006

(2.038) (1.759) (1.015) (-0.226) (-0.148)

no crime 2.288** 0.235* 0.039 -0.000 0.003

(2.041) (1.956) (0.969) (-0.000) (0.069)

no ln(population) 2.055* 0.196 0.042 0.007 -0.018

(1.958) (1.601) (1.098) (0.252) (-0.428)

Note. Each cell reports the coefficient or standard error of the Africa dummy for a particular equation.

*, **, and *** represent statistical significance at the 10, 5 and 1 percent levels.

Standard errors are clustered at the country level.

150

Table 6. Explaining Firm performance: adding country fixed effects

OLS FE OLS FE OLS FE OLS FE OLS FE

ln(LP) ln(LP) sale

growth

sale

growth L growth L growth

export

share

export

share inv/VA inv/VA

Ln(L) 0.123*** 0.154*** 0.063*** 0.061*** 0.034*** 0.047***

(3.990) (5.866) (8.445) (9.655) (5.566) (6.203)

Ln(firm age) 0.083*** 0.071*** -0.027** -0.019* -0.073*** -0.078*** -0.020*** -0.022*** -0.029*** -0.018***

(2.634) (3.365) (-2.195) (-1.824) (-9.435) (-9.217) (-3.649) (-4.295) (-7.083) (-3.767)

foreign 0.606*** 0.584*** 0.142*** 0.190*** 0.059*** 0.067*** 0.135*** 0.125*** -0.003 0.011

(6.972) (11.195) (3.754) (7.574) (2.901) (3.885) (5.167) (5.005) (-0.243) (1.011)

Ind avg foreign -1.246* -0.117 0.025 0.075* 0.022 0.033 0.166*** 0.196*** -0.112*** -0.054

(-1.881) (-0.903) (0.267) (1.655) (0.631) (0.942) (3.658) (4.869) (-3.096) (-1.620)

Own1 -0.156 -0.138** -0.105*** -0.114*** -0.064*** -0.062*** -0.000 0.009 0.010 0.009

(-1.286) (-2.157) (-3.544) (-4.450) (-3.172) (-3.646) (-0.011) (0.611) (0.744) (0.705)

Tariff 1.158 -1.130*** 0.240 -0.163 0.104 0.005 0.032 0.064 0.036 -0.119**

(1.095) (-2.614) (1.108) (-1.370) (1.116) (0.082) (0.217) (0.419) (0.433) (-2.200)

CompetitionInd -0.127 -0.858*** -0.033 -0.138* 0.034 -0.036 0.034 0.042 0.158*** 0.131***

(-0.152) (-5.655) (-0.279) (-1.821) (0.846) (-1.163) (0.871) (1.311) (2.700) (3.764)

Web 0.443 0.537*** 0.163 0.127** 0.189*** 0.110*** 0.026 -0.037 0.109*** 0.011

(0.589) (3.748) (1.226) (2.426) (4.602) (3.579) (0.704) (-1.033) (2.648) (0.337)

Bribe -0.051 0.999 -0.823 -0.085 -0.696*** -0.567*** 0.028 -0.094 -0.129 -0.284*

(-0.022) (0.707) (-1.416) (-0.199) (-3.263) (-3.204) (0.126) (-0.492) (-0.611) (-1.837)

Trade credit 0.951** -0.009 0.096 0.058 0.035 0.061 0.149*** 0.083* 0.098** 0.040

(1.999) (-0.049) (1.048) (0.822) (0.751) (1.344) (4.075) (1.944) (2.134) (1.101)

bank 0.563* 0.236 0.115 0.020 0.021 0.021 -0.061** -0.014 0.002 -0.014

(1.780) (1.568) (1.145) (0.430) (0.588) (0.584) (-2.191) (-0.334) (0.054) (-0.431)

cons_crime -0.849 0.123 -0.168* -0.074 0.005 -0.038 -0.053* -0.004 -0.063 0.039

(-1.498) (0.667) (-1.918) (-1.314) (0.138) (-0.920) (-1.834) (-0.102) (-1.303) (1.188)

Ln(sales t-3) -0.073*** -0.093***

(-6.927) (-16.231)

Ln(L t-3) -0.081*** -0.079***

(-13.861) (-14.736)

Ln(sales) -0.000 0.003

(-0.032) (1.049)

Ln(value added) -0.020*** -0.036***

(-3.979) (-5.884)

country FE yes yes yes yes yes

N 12,197 13,140 10,363 11,123 12,125 12,982 12,289 13,229 9,602 10,432

adjusted R squared 0.307 0.565 0.117 0.196 0.112 0.131 0.248 0.276 0.050 0.084

Note. *, **, and *** represent statistical significance at the 10, 5 and 1 percent levels. Standard errors

are clustered at the country level.

151

For the OLS specification, we control the same set of variables as in the previous table. For brevity, we

do not report results for country-level variables in the OLS specification.

152

Table 7. Sensitivity check with more controls

Without additional controls With additional controls

ln(LP) sale

growth

L

growth

export

share inv/VA ln(LP)

sale

growth

L

growth

export

share inv/VA

Africa 2.170** 0.210* 0.040 -0.007 -0.005 2.741** 0.221* 0.109*** -0.030 0.032

(2.053) (1.763) (1.014) (-0.252) (-0.121) (2.563) (1.686) (2.663) (-1.020) (0.682)

Foreign 0.606**

* 0.142*** 0.059*** 0.135*** -0.003

0.626***

0.148*** 0.061*** 0.131*** 0.007

(6.972) (3.754) (2.901) (5.167) (-0.243) (7.364) (4.229) (3.030) (4.783) (0.594)

Country-

industry -1.246* 0.025 0.022 0.166***

-

0.112*** -0.941* 0.072 0.055 0.173*** -0.080*

avg

foreign (-1.881) (0.267) (0.631) (3.658) (-3.096) (-1.749) (1.000) (1.546) (4.211) (-1.862)

own1 -0.156 -

0.105*** -

0.064*** -0.000 0.010 -0.109

-0.108***

-0.059***

0.002 0.017

(-1.286) (-3.544) (-3.172) (-0.011) (0.744) (-1.069) (-3.909) (-2.916) (0.108) (1.295)

Tariff 1.158 0.240 0.104 0.032 0.036 0.582 0.221 0.072 0.044 -0.060

(1.095) (1.108) (1.116) (0.217) (0.433) (0.777) (1.084) (0.893) (0.339) (-0.878)

CompetitionIn

d -0.127 -0.033 0.034 0.034 0.158*** -0.614 -0.010 0.039 0.078** 0.126***

(-0.152) (-0.279) (0.846) (0.871) (2.700) (-0.956) (-0.086) (0.925) (2.038) (2.608)

Ln(phoneDen) 0.227 0.010 -0.012 0.017 -0.010 0.367 0.070 0.027 0.022 0.013

(1.043) (0.203) (-0.687) (1.027) (-0.621) (1.519) (1.370) (1.479) (1.256) (0.625)

Landlock 0.139 -0.057 -0.028 0.010 0.056** -0.101 -0.097 -0.048** 0.025 0.057***

(0.471) (-0.578) (-1.233) (0.580) (2.306) (-0.435) (-1.402) (-2.266) (1.413) (3.174)

Ln(party

years) -0.513 -0.117** 0.016 -0.013 -0.022* -0.663*

-

0.117*** 0.004 -0.012

-

0.033***

(-1.447) (-2.088) (1.485) (-1.291) (-1.960) (-1.852) (-2.661) (0.366) (-1.166) (-3.103)

anyConflict -0.974 -0.113 0.012 0.044** -

0.064*** -0.660 -0.153* -0.011 0.015

-0.096***

(-1.621) (-1.329) (0.476) (2.383) (-2.937) (-1.241) (-1.953) (-0.418) (0.979) (-5.388)

Ethnic Frac -0.551 0.062 0.059 -0.002 -0.002 -1.028 -0.009 0.055 0.020 -0.000

(-0.450) (0.282) (0.790) (-0.035) (-0.050) (-0.944) (-0.046) (0.844) (0.560) (-0.005)

fireCosts 1.288** -0.227** -0.076** -0.013 0.063* 0.713 -0.077 -0.043 -0.005 0.071**

(1.964) (-2.021) (-2.109) (-0.435) (1.735) (1.538) (-0.632) (-1.280) (-0.218) (2.140)

lnCostExport 0.185 -0.042 0.041 -

0.085*** -0.051* 0.116 -0.066 0.040

-

0.078*** -0.056**

(0.610) (-0.666) (1.092) (-4.046) (-1.742) (0.486) (-1.251) (1.344) (-3.900) (-2.061)

customs -0.867 0.037 -0.035 -0.006 -0.023 -1.201** -0.056 -0.024 0.051 -0.030

(-1.416) (0.346) (-0.723) (-0.168) (-0.549) (-2.414) (-0.544) (-0.466) (1.472) (-0.749)

infrastructure 1.485* 0.021 0.077* -0.053 -0.062 2.218**

* 0.065 0.109*** -0.063 0.020

(1.774) (0.164) (1.839) (-1.155) (-1.187) (2.604) (0.461) (2.876) (-1.458) (0.416)

Web 0.443 0.163 0.189*** 0.026 0.109*** 0.286 0.128 0.150*** -0.019 0.040

(0.589) (1.226) (4.602) (0.704) (2.648) (0.452) (1.433) (4.383) (-0.499) (1.045)

Bribe -0.051 -0.823 -

0.696*** 0.028 -0.129 2.045 -0.199

-0.689***

-0.089 -0.268

(-0.022) (-1.416) (-3.263) (0.126) (-0.611) (0.956) (-0.374) (-3.427) (-0.535) (-1.222)

Trade credit 0.951** 0.096 0.035 0.149*** 0.098** 0.323 0.102 -0.004 0.158*** 0.105**

153

(1.999) (1.048) (0.751) (4.075) (2.134) (0.865) (1.025) (-0.084) (3.781) (2.402)

Bank 0.563* 0.115 0.021 -0.061** 0.002 0.489 -0.062 -0.024 -

0.068*** 0.019

(1.780) (1.145) (0.588) (-2.191) (0.054) (1.223) (-0.659) (-0.868) (-2.884) (0.603)

Table 7. Sensitivity check with more controls (Cont’d)

Without additional controls With additional controls

ln(LP)

sale

growt

h

L

growt

h

export

share

inv/V

A ln(LP)

sale

growth

L

growth

export

share inv/VA

cons_crime -0.849 -

0.168* 0.005

-

0.053* -0.063 -1.301**

-

0.304*** -0.031 -0.042 -0.060

(-

1.498)

(-

1.918)

(0.138

)

(-

1.834)

(-

1.303) (-2.565) (-3.655) (-0.891) (-1.109) (-1.393)

Trade/GDP 0.006 -

0.004***

-

0.001*** -0.000 -0.000

(1.121) (-2.992) (-2.815) (-0.298) (-0.420)

VoiceAcc 0.377 -0.032 -

0.071***

-

0.081***

-

0.077***

(1.497) (-0.461) (-3.093) (-4.744) (-3.781)

Exe -

0.409*** -0.015 -0.007 0.022*** -0.009

competition (-4.111) (-0.366) (-0.871) (3.735) (-0.844)

cons_electricit

y 0.268 -0.039 -0.028 -0.007 -0.009

(0.579) (-0.468) (-0.577) (-0.234) (-0.261)

cons_transport -0.040 0.026 0.006 0.007 -0.013

(-0.134) (0.347) (0.122) (0.208) (-0.329)

cons_infComp -0.091 -0.095 -0.011 -0.073** -0.080**

(-0.210) (-1.193) (-0.331) (-2.260) (-1.997)

cons_taxrate 0.363 0.082 -0.021 -0.008 0.068*

(0.637) (0.813) (-0.751) (-0.276) (1.866)

cons_landAcc 0.420 0.202* 0.064* -0.061** 0.140***

(0.980) (1.884) (1.750) (-2.037) (3.476)

cons_finAcc 1.069 0.026 -0.024 -0.016 -0.017

(1.354) (0.234) (-0.572) (-0.632) (-0.518)

cons_Lregu 0.005 -0.087 0.100 0.137*** -0.023

(0.012) (-0.749) (1.624) (4.058) (-0.564)

Ln(MinK) -

0.213*** -0.024* -0.007* 0.015*** -0.000

(-3.534) (-1.705) (-1.819) (3.922) (-0.011)

N 12,197 10,363 12,125 12,289 9,602 11,340 9,679 11,229 11,432 9,003

Adjusted R

squ. 0.307 0.117 0.112 0.248 0.050 0.362 0.136 0.120 0.260 0.060

*, **, and *** represent statistical significance at the 10, 5 and 1 percent levels. Intercept not reported.

Standard errors are clustered at the country level.

We also control for log size in various ways, log population, log firm age. For simplicity, we do not report them

here.

154

Table 8. Explaining Firm Performance: Africa now includes the four rich countries


Africa 1.438* 0.142 0.044 -0.044* 0.003

(1.900) (1.597) (1.629) (-1.692) (0.080)

foreign 0.593*** 0.135*** 0.057*** 0.136*** -0.003

(8.574) (3.514) (2.842) (5.172) (-0.250)

Ind avg foreign -1.411* 0.012 0.017 0.172*** -0.113***

(-1.919) (0.122) (0.514) (3.797) (-3.072)

own1 -0.097 -0.098*** -0.063*** -0.000 0.009

(-0.757) (-3.245) (-3.137) (-0.017) (0.728)

Tariff 0.537 0.182 0.088 0.045 0.036

(0.607) (0.887) (0.938) (0.295) (0.419)

Competition_Ind -0.301 -0.053 0.028 0.041 0.156***

(-0.415) (-0.461) (0.700) (1.069) (2.581)

Ln(phone density) -0.015 -0.013 -0.014 0.012 -0.009

(-0.054) (-0.244) (-0.748) (0.689) (-0.555)

Ln(partyYears) -0.580 -0.124** 0.013 -0.007 -0.023**

(-1.431) (-2.128) (1.107) (-0.752) (-1.966)

anyConflict -0.864 -0.102 0.014 0.042** -0.064***

(-1.467) (-1.195) (0.575) (2.188) (-2.744)

fireCosts 1.208* -0.237** -0.078** -0.013 0.063*

(1.800) (-2.100) (-2.109) (-0.430) (1.723)

Ln(CostExport) -0.090 -0.070 0.033 -0.075*** -0.052

(-0.247) (-1.050) (0.865) (-3.229) (-1.521)

Customs -1.329* -0.011 -0.046 0.002 -0.023

(-1.848) (-0.104) (-0.985) (0.046) (-0.526)

Infrastructure 1.244* -0.001 0.068 -0.041 -0.063

(1.725) (-0.010) (1.582) (-0.932) (-1.193)

Web 0.598 0.174 0.188*** 0.033 0.108***

(0.832) (1.302) (4.240) (0.854) (2.722)

Bribe 2.160 -0.581 -0.647*** -0.010 -0.127

(0.771) (-0.999) (-2.999) (-0.045) (-0.581)

Trade credit 1.218** 0.116 0.044 0.138*** 0.099**

(2.548) (1.313) (0.912) (3.882) (2.280)

Bank 0.533* 0.113 0.021 -0.062** 0.002

(1.668) (1.148) (0.594) (-2.239) (0.061)

Cons_crime -1.096 -0.194** 0.002 -0.056** -0.062

(-1.581) (-2.092) (0.065) (-2.067) (-1.289)

N 12,197 10,363 12,125 12,289 9,602

Adjusted R squared 0.286 0.116 0.112 0.249 0.050

Note. *, **, and *** represent statistical significance at the 10, 5 and 1 percent levels. Intercept not reported.

Standard errors are clustered at the country level. We also control for log size in various ways, log population,

155

ethnic fractionalization, landlock, log firm age, the same as in previous tables. For simplicity, we do not report them

here.

Table 9. Explaining firm performance, using only the low-income sample


Africa 2.054 0.304* 0.047 -0.064*** -0.049

(1.582) (1.832) (1.473) (-2.742) (-1.111)

Foreign 0.586*** 0.110** 0.059*** 0.156*** -0.002

(6.518) (2.328) (2.663) (4.091) (-0.135)

Country-industry avg foreign -1.242 -0.049 -0.001 0.143*** -0.059

(-1.337) (-0.468) (-0.019) (3.181) (-1.191)

own1 -0.194 -0.106** -0.028 -0.001 0.012

(-1.474) (-2.514) (-0.952) (-0.063) (0.688)

Tariff 1.004 0.114 0.146 0.216* -0.066

(0.658) (0.396) (1.287) (1.767) (-0.800)

CompetitionInd 0.207 0.253* 0.054 0.042 0.199***

(0.271) (1.655) (1.020) (0.864) (3.541)

Ln(phoneDensity) 0.253 -0.013 0.001 0.006 -0.013

(0.979) (-0.232) (0.053) (0.508) (-0.780)

Ln(party years) -0.846* -0.193*** 0.008 -0.004 -0.025*

(-1.792) (-2.822) (0.686) (-0.584) (-1.936)

AnyConflict -1.297 -0.213* 0.038* 0.004 -0.102***

(-1.433) (-1.685) (1.732) (0.238) (-4.843)

fireCosts 2.918*** -0.069 -0.016 0.052** 0.205***

(2.602) (-0.474) (-0.487) (2.464) (3.898)

Ln(CostExport) 0.628 -0.057 -0.032* -0.031 0.044*

(1.511) (-0.559) (-1.650) (-1.419) (1.712)

Customs 0.839 0.356 -0.055 0.018 0.104*

(0.602) (1.396) (-1.000) (0.527) (1.740)

Infrastructure 0.160 -0.285 0.042 0.018 -0.062

(0.096) (-1.349) (0.770) (0.311) (-0.968)

Web -0.149 0.226* 0.212*** -0.017 0.050

(-0.180) (1.795) (5.271) (-0.457) (1.254)

Bribe -0.125 -0.771 -0.607*** -0.118 -0.310

(-0.063) (-1.366) (-3.062) (-0.788) (-1.473)

Trade credit 0.515 0.198 0.031 0.146*** 0.085*

(0.929) (1.448) (0.618) (3.616) (1.801)

Bank 0.206 0.259* 0.048 -0.044 -0.080*

(0.441) (1.804) (1.243) (-1.476) (-1.834)

cons_crime -1.316** -0.166* -0.072 -0.022 -0.023

(-2.268) (-1.943) (-1.603) (-0.688) (-0.458)

N 6,913 5,966 6,887 6,998 5,499

156

Adjusted R squared 0.290 0.133 0.133 0.291 0.054


Standard errors are clustered at the country level. We also control for log size in various ways, log population,

landlock, ethnic fractionalization, log firm age, the same as in previous tables. For simplicity, we do not report them

here.

Table 10. Matching estimates of Africa, landlock, anyCon10 (updated)

Using the full sample of countries

lnLP sale growth L growth Export share inv/VA

Africa 1.864 (2.153) 1.012 (0.489) 1.209 (0.652) 0.067 (0.352) 0.096 (0.346)

landlock -0.888 (0.202) -0.077 (0.066) 0.095 (0.045) 0.028 (0.013) 0.005 (0.082)

anyConflicts -1.313 (0.164) -0.281 (0.104) 0.064 (0.054) 0.018 (0.025) -0.124 (0.112)

Using the similar-income sample and the African-Poor sample

lnLP sale growthrow L growth Export share inv/VA

Africa 1.869 (1.198) 0.465 (0.535) -0.560 (0.512) 0.067 (0.426) 0.019 (0.393)

157

Table 11. By manufacturing and services

Manufacturing Services

ln(LP) sale

growth L growth

export

share inv/VA ln(LP)

sale

growth

L

growth

export

share inv/VA

Africa 2.170** 0.210* 0.040 -0.007 -0.005 0.989 0.131 -0.000 0.004 -0.026

(2.053) (1.763) (1.014) (-0.252) (-0.121) (1.570) (1.346) (-0.016) (0.282) (-0.947)

Foreign 0.606*** 0.142*** 0.059*** 0.135*** -0.003 0.755*** 0.138*** 0.048** 0.053*** 0.012

(6.972) (3.754) (2.901) (5.167) (-0.243) (8.140) (3.594) (2.472) (4.581) (0.886)

Ind avg foreign -1.246* 0.025 0.022 0.166*** -0.112*** -0.488 -0.054 0.077 -0.003 0.004

(-1.881) (0.267) (0.631) (3.658) (-3.096) (-0.815) (-0.405) (1.194) (-0.137) (0.128)

own1 -0.156 -0.105*** -0.064*** -0.000 0.010 -0.041 0.015 -0.026 0.001 0.017

(-1.286) (-3.544) (-3.172) (-0.011) (0.744) (-0.330) (0.391) (-1.143) (0.193) (1.506)

Tariff 1.158 0.240 0.104 0.032 0.036 -0.221 -0.696 -0.703** -0.044 -0.657**

(1.095) (1.108) (1.116) (0.217) (0.433) (-0.075) (-1.071) (-2.069) (-0.305) (-1.974)

CompetitionInd -0.127 -0.033 0.034 0.034 0.158*** -0.505 0.085 0.046 0.024 0.073

(-0.152) (-0.279) (0.846) (0.871) (2.700) (-0.895) (0.718) (0.773) (1.249) (1.489)

Ln(phoneDensity) 0.227 0.010 -0.012 0.017 -0.010 0.353** 0.010 -0.028** -0.000 -0.008

(1.043) (0.203) (-0.687) (1.027) (-0.621) (2.253) (0.254) (-2.212) (-0.076) (-0.681)

Ln(partyYears) -0.513 -0.117** 0.016 -0.013 -0.022* -0.245 -0.045 0.020** -0.001 -0.016**

(-1.447) (-2.088) (1.485) (-1.291) (-1.960) (-1.154) (-1.121) (2.001) (-0.302) (-2.151)

anyConflict -0.974 -0.113 0.012 0.044** -0.064*** -0.699 -0.049 0.049** 0.012 -0.059***

(-1.621) (-1.329) (0.476) (2.383) (-2.937) (-1.327) (-0.741) (2.003) (1.295) (-3.745)

fireCosts 1.288** -0.227** -0.076** -0.013 0.063* 1.184** -0.051 -0.099** -0.002 0.031

(1.964) (-2.021) (-2.109) (-0.435) (1.735) (2.259) (-0.419) (-2.095) (-0.129) (1.056)

Ln(CostExport) 0.185 -0.042 0.041 -0.085*** -0.051* -0.088 -0.023 0.020 -0.008 0.002

(0.610) (-0.666) (1.092) (-4.046) (-1.742) (-0.339) (-0.402) (0.684) (-0.785) (0.153)

infrastructure 1.485* 0.021 0.077* -0.053 -0.062 1.066 0.229** 0.122** 0.010 -0.016

(1.774) (0.164) (1.839) (-1.155) (-1.187) (1.334) (1.964) (2.133) (0.535) (-0.413)

Web 0.443 0.163 0.189*** 0.026 0.109*** 0.482 0.168 0.136*** 0.016 -0.064**

(0.589) (1.226) (4.602) (0.704) (2.648) (0.866) (1.578) (2.637) (0.949) (-2.105)

Bribe -0.051 -0.823 -0.696*** 0.028 -0.129 -1.144 -1.326 -0.134 0.129 -0.331*

(-0.022) (-1.416) (-3.263) (0.126) (-0.611) (-0.388) (-1.633) (-0.349) (0.943) (-1.691)

Trade credit 0.951** 0.096 0.035 0.149*** 0.098** 0.899** -0.026 -0.036 0.063** 0.030

(1.999) (1.048) (0.751) (4.075) (2.134) (2.168) (-0.230) (-0.624) (2.543) (0.924)

bank 0.563* 0.115 0.021 -0.061** 0.002 0.578** 0.205** 0.099*** 0.002 0.014

(1.780) (1.145) (0.588) (-2.191) (0.054) (2.185) (2.206) (2.784) (0.078) (0.457)

cons_crime -0.849 -0.168* 0.005 -0.053* -0.063 -0.777 -0.182* -0.074* 0.004 -0.060*

(-1.498) (-1.918) (0.138) (-1.834) (-1.303) (-1.131) (-1.865) (-1.799) (0.240) (-1.785)

N 12,197 10,363 12,125 12,289 9,602 8,018 6,712 7,925 8,111 5,390

Adjusted R squared 0.307 0.117 0.112 0.248 0.050 0.245 0.091 0.092 0.050 0.033


Standard errors are clustered at the country level. As in previous tables, we also control for current or initial size,

158

log population, and log firm age. The results on size or initial level variables, ethnic fractionalization, custom

efficiency and landlock are similar for the two sectors, and we don‘t report them.

Table 12. Only light manufacturing industries

light advanced

lnLPs saleG Lgrow exportSh inv_VA lnLPs saleG Lgrow exportSh inv_VA

Africa 2.649** 0.229 0.040 -0.080** 0.021 1.684** 0.113 0.051 0.111*** 0.043

(2.381) (1.626) (0.888) (-2.037) (0.378) (2.427) (0.922) (0.793) (2.694) (0.836)

Tariff 1.443 0.156 0.136 -0.177* 0.156 3.629 1.937*** 0.352 -0.467* 0.393*

(1.248) (0.653) (0.930) (-1.677) (1.520) (1.588) (3.550) (1.304) (-1.787) (1.727)

Competition_ind 0.905 0.006 0.219** 0.105 0.063 0.035 -0.169 -0.029 0.149** 0.186**

(0.536) (0.026) (2.205) (1.011) (0.311) (0.054) (-0.829) (-0.299) (2.114) (2.315)

Ln(phoneDen) -0.026 -0.022 -0.023 -0.004 -0.005 0.429*** 0.060 -0.010 0.076*** 0.007

(-0.092) (-0.345) (-0.865) (-0.153) (-0.211) (2.651) (1.042) (-0.344) (3.460) (0.307)

web -0.211 0.119 0.159*** -0.023 0.074 0.640 0.317*** 0.192*** -0.002 0.135***

(-0.226) (0.865) (2.939) (-0.479) (1.165) (1.381) (2.896) (4.069) (-0.036) (3.050)

Ln(population) 0.149 0.014 -0.008 -0.029*** 0.013 0.233* 0.015 0.003 -0.022** 0.014

(0.819) (0.403) (-0.705) (-2.814) (1.160) (1.645) (0.410) (0.181) (-2.066) (0.990)

landlock 0.227 -0.033 -0.024 0.040* 0.097*** 0.264 -0.151* -0.032 0.011 0.016

(0.659) (-0.314) (-0.963) (1.805) (3.116) (0.977) (-1.935) (-0.796) (0.365) (0.517)

Ln(partyYears) -0.845** -0.136** -0.003 0.002 -0.031* -0.308 -0.066 0.017 -0.012 -0.002

(-2.128) (-2.342) (-0.260) (0.158) (-1.873) (-1.493) (-1.357) (0.941) (-1.068) (-0.169)

anyConflict -1.193* -0.082 0.013 0.073*** -0.064** -0.666* -0.022 -0.004 0.012 -0.068**

(-1.957) (-1.011) (0.488) (3.587) (-2.268) (-1.868) (-0.286) (-0.115) (0.462) (-2.261)

fireCosts 1.446** -0.222* -0.059 -0.035 0.087 0.483 -0.318*** -0.093* 0.018 0.047

(2.326) (-1.811) (-1.451) (-1.051) (1.511) (1.116) (-2.857) (-1.719) (0.566) (1.143)

Ln(CostExport) 0.140 -0.056 0.026 -0.118*** -0.050 -0.102 -0.092 0.038 -0.063* -0.049

(0.443) (-0.929) (0.748) (-5.107) (-1.458) (-0.392) (-1.328) (0.899) (-1.875) (-1.318)

customs -0.635 0.106 -0.054 0.038 -0.036 -1.021* -0.016 -0.068 -0.076 -0.072

(-0.952) (0.715) (-0.789) (0.867) (-0.754) (-1.857) (-0.115) (-0.981) (-1.644) (-1.401)

infrastructure 1.807** 0.013 0.079* -0.076 -0.043 1.115** -0.007 0.104 -0.056 -0.037

(2.047) (0.095) (1.690) (-1.282) (-0.621) (2.020) (-0.048) (1.432) (-1.159) (-0.608)

Bribe 3.832 0.054 -0.956*** -0.057 -0.612* -1.821 -0.725 -0.548* 0.001 -0.038

(1.015) (0.055) (-2.871) (-0.206) (-1.800) (-1.042) (-1.017) (-1.803) (0.004) (-0.176)

Bank 0.650* 0.077 -0.021 -0.034 0.022 0.459 0.121 0.071 -0.037 0.015

(1.735) (0.715) (-0.479) (-1.146) (0.518) (1.458) (1.069) (1.482) (-0.746) (0.394)

Trade Credit 0.730 0.163* 0.134** 0.173*** 0.076 1.527*** -0.141 -0.014 0.182*** 0.110**

(1.229) (1.806) (2.466) (3.100) (1.223) (3.669) (-1.261) (-0.173) (3.162) (2.059)

cons_crime -0.623 -0.046 0.073 -0.067* -0.026 -1.026* -0.364*** -0.100* -0.053 -0.167***

(-1.119) (-0.481) (1.637) (-1.706) (-0.382) (-1.929) (-3.291) (-1.937) (-1.167) (-2.593)

N 5,778 4,878 5,802 5,833 4,521 3,525 3,069 3,587 3,549 2,682

Adjusted R

squared 0.328 0.104 0.115 0.293 0.043 0.315 0.141 0.104 0.255 0.086

159


Standard errors are clustered at the country level. As in previous tables, we also control for current or initial size,

ethnic fractionalization, foreign ownership, ownership concentration, and log firm age. Qualitative results on them

for the 2 sectors are similar.

160

Table 13. Determinants of firm performance: by small firms and LMEs

Small LMEs

lnLPs saleG Lgrow exportSh inv_VA lnLPs saleG Lgrow exportSh inv_VA

Africa 2.840** 0.475** 0.109*** -0.027** -0.002 1.963** 0.198* 0.001 -0.002 0.003

(1.967) (2.147) (5.178) (-2.014) (-0.056) (2.168) (1.686) (0.030) (-0.073) (0.061)

Ln(L) 0.072 0.012 0.042*** 0.112*** 0.070*** 0.033***

(0.592) (1.282) (2.719) (3.257) (8.234) (4.747)

Ln(firm age) 0.042 -0.066*** -0.026*** 0.002 -0.030*** 0.103*** -0.023* -0.069*** -0.026*** -0.026***

(0.699) (-3.175) (-4.029) (0.659) (-2.903) (3.411) (-1.926) (-8.131) (-4.104) (-6.643)

Foreign 0.871*** 0.104 0.010 0.077** -0.019 0.616*** 0.143*** 0.066*** 0.135*** 0.001

(3.503) (1.000) (0.260) (1.963) (-0.803) (8.807) (4.005) (3.431) (5.434) (0.110)

Ind avg foreign -0.291 0.154 0.031 0.038 -0.139*** -1.384** -0.035 -0.003 0.179*** -0.105**

(-0.483) (1.012) (0.919) (1.325) (-2.677) (-2.177) (-0.363) (-0.081) (3.562) (-2.281)

own1 -0.435*** -0.100 -0.042** 0.009 -0.021 -0.074 -0.079*** -0.049** -0.007 0.020

(-3.463) (-1.591) (-2.166) (0.667) (-0.639) (-0.629) (-2.785) (-2.280) (-0.381) (1.531)

competitionInd -0.447 0.058 0.131*** 0.004 0.225*** -0.048 -0.132 -0.058 0.057 0.144**

(-0.447) (0.315) (2.609) (0.135) (3.636) (-0.067) (-1.014) (-1.363) (1.222) (2.164)

landlock 0.193 -0.062 0.023 0.012 0.059** 0.133 -0.059 -0.064*** 0.014 0.060**

(0.411) (-0.447) (1.265) (1.249) (2.161) (0.547) (-0.637) (-2.707) (0.664) (2.343)

Ln(partyYears) -0.833 -0.186** 0.019 -0.007 -0.027 -0.428 -0.115** 0.020* -0.013 -0.024*

(-1.633) (-2.469) (1.479) (-1.236) (-1.598) (-1.445) (-2.167) (1.889) (-1.157) (-1.874)

anyConflict -1.499* -0.275** 0.022* 0.014* -0.040* -0.801 -0.101 0.016 0.051** -0.074***

(-1.686) (-2.450) (1.650) (1.879) (-1.849) (-1.618) (-1.206) (0.609) (2.482) (-3.218)

fireCosts 2.502** -0.109 0.012 -0.027* 0.050 0.925* -0.217* -0.083** -0.005 0.065

(2.440) (-0.675) (0.404) (-1.723) (1.220) (1.659) (-1.908) (-2.087) (-0.132) (1.571)

Ln(CostExport) 0.818** 0.217** 0.032* -0.038*** -0.012 0.051 -0.070 0.047 -0.096*** -0.059*

(2.454) (2.153) (1.807) (-2.987) (-0.335) (0.168) (-1.050) (1.169) (-4.211) (-1.940)

infrastructure 1.768 0.398* 0.066* -0.089*** -0.036 1.439** 0.026 0.112** -0.044 -0.060

(1.438) (1.676) (1.724) (-4.538) (-0.528) (1.978) (0.215) (2.519) (-0.885) (-1.029)

web 0.889 0.281 0.112*** 0.035 0.175*** 0.329 0.107 0.153*** 0.028 0.080*

(0.841) (1.173) (3.583) (1.330) (2.602) (0.525) (0.891) (3.695) (0.678) (1.920)

bribe -1.558 0.317 -0.003 0.044 0.262 0.241 -0.937 -0.717*** 0.053 -0.238

(-0.434) (0.262) (-0.012) (0.282) (0.691) (0.111) (-1.619) (-2.864) (0.221) (-1.113)

trade credit 0.932** 0.264* 0.047 0.066** 0.088 0.926* 0.069 0.020 0.164*** 0.093*

(2.029) (1.729) (1.115) (2.476) (1.540) (1.950) (0.775) (0.386) (4.203) (1.936)

Bank 0.992 0.074 -0.028 -0.071*** -0.007 0.504* 0.137 0.042 -0.056* 0.011

(1.577) (0.436) (-1.005) (-2.993) (-0.158) (1.785) (1.377) (1.030) (-1.821) (0.334)

cons_crime -1.008* -0.165 -0.017 0.033 -0.013 -0.762 -0.192** -0.027 -0.056* -0.082

(-1.748) (-1.349) (-0.479) (1.534) (-0.226) (-1.383) (-2.118) (-0.757) (-1.768) (-1.559)

N 2,761 2,218 2,657 2,793 2,287 9,786 8,436 9,822 9,853 7,593

Adjusted R

squared 0.335 0.180 0.696 0.057 0.042 0.304 0.124 0.143 0.248 0.050

161


Standard errors are clustered at the country level. For some columns, we also control for lnS3, lnL3, lnV. We also

control for log population, ethnic fractionalization, tariff levels, customs, and log firm age.

Table 14. Accounting for performance differences between the Poor African countries

and the top half performers of the similar-income sample

∆ln(LP) =

-0.171

∆sale growth =

- 0.027

∆L growth =

0.067

∆export share =

-0.151

∆inv/VA=

-0.107

β*∆X

% explained β*∆X

% explained β*∆X

% explained β*∆X

% explained β*∆X

% explained

Foreign 0.016 -9.6 0.004 -14.2

Ind Avg

Foreign -0.047 27.8 0.006 -4.2 -0.004 4.0

Ln(age) -0.020 11.4 0.006 -23.3 0.017 25.8 0.005 0.007 -6.4

own1 -0.013 46.5 -0.008 -11.5

competitionInd -0.005 4.3

infrastructure -0.474 277.4 -0.025 -36.6

Web -0.067 -100.3 -0.039 36.3

Ln(partyYears) -0.051 188.5 -0.010 9.0

Bribe -0.013 -18.8

cons_crime -0.022 80.0 -0.006 4.1

fireCosts 0.042 -24.4 -0.007 27.2 -0.002

Bank -0.052 30.6

Trade credit -0.305 178.4 -0.045 30.1 -0.031 29.4

Ln(CostExport) -0.022 14.6 -0.013 12.5

lnL -0.128 75.0 -0.066 43.7 -0.036 33.2

lnS3 0.075 -279.3

lnL3 0.084 125.7

lnV 0.019 -17.4

Africa 2.170 -1269.2 0.210 -779.4

Note. We do not report the effects for some variables (e.g., country-industry-year tariff level, whether there is any conflicts in the

past 10 years, phone density, ethnic fractionalization, the custom efficiency index, the dummy of being landlocked) either

because they are statistically insignificant or they proves to be quantitatively trivial. We also do not report the results for log

population (explaining -23% of the African export share disadvantage). We also do not report cells that the coefficient is

significant but the magnitude explains less than 4% of the outcome difference.

,

162

Table 15. Summary of how the African performance differentials can be explained

Factors

∆ln(LP)

= -0.171

∆sale growth

= - 0.027

∆L growth

= 0.067

∆export share

= -0.151

∆inv/VA

= -0.107

Foreign ownership 18% -14% - 4% 4%

Firm age 11% - 23% 26% -6%

Ownership concentration 47% - 12%.

Industry-level Competition 4%

Infrastructure (phone density, web

density, the LPI infrastructure index): 277% -137% 36%

Trade transport costs 15% 13%

Party monopoly 188% 9%

Corruption -19%

Firing difficulty - 24% 27%

Formal finance 30%

Informal finance 178% 31% 29%

Crime 80% 4%

Current or initial size 75% -280% 125% 44% 16

163

Appendix. The list of countries for each sample

(The numbers in the table are GDP per capita in 2005 U.S. dollars)

Africa

Angola2006 767.4 Gambia2006 332.7 Mozambique2007 307.8

Benin2009 353.8 Ghana2007 290.3 Namibia2006 2460.2

Burundi2006 109.2 Guinea2006 141.6 Niger2009 170.5

Cameroon2006 692.0 GuineaBissau2006 396.5 Rwanda2006 250.2

Cameroon2009 692.0 Ivory Coast2009 530.1 Senegal2007 522.3

CapeVerde2009 1553.8 Lesotho2009 501.5 Sierra Leone2009 254.1

Chad2009 285.2 Liberia2009 144.4 Swaziland2006 1463.2

Congo2009 1156.4 Madagascar2009 260.2 Tanzania2006 316.3

DRC2006 85.8 Malawi2009 148.4 Togo2009 248.5

Eritrea2009 151.0 Mauritania2006 429.5 Uganda 283.1

The average comparison group (i.e., with GDP per capita < 3000 USD)

Albania2007 1541.0 Georgia2008 1079.9 Romania2009 2595.5

Armenia2009 1425.2 Guatemala2006 1749.2 Russia2009 2866.3

Azerbaijan2009 1945.6 Honduras2006 1244.9 Samoa2009 1800.0

Belarus2008 2067.6 Kazakhstan2009 2332.2 Tajikistan2008 217.2

Bolivia2006 1039.3 Kyrgyz Rep.2009 352.1 Timor Leste2009 299.7

Bosnia and

Herzegovina200

2041.4 LaoPDR2009 450.0 Tonga2009 1659.7

Bulgaria2009 2412.6 Mongolia2009 683.1 Ukraine2008 1037.3

Colombia2006 2955.2 Nepal2009 245.1 Vanuatu2009 1288.0

Ecuador2006 1515.7 Nicaragua2006 818.4 Vietnam2009 617.1

ElSalvador2006 2359.4 Paraguay2006 1346.5 Yemen2010 561.3

Fiji2009 2190.4 Peru2006 2228.3 Uzbekistan2008 725.4

Fyr Macedonia2009 2075.8 Philippines2009 1201.7

The better comparison group (i.e., with GDP per capita < 3000 USD, and top half in firm performance)

Albania2007 1541.0 Georgia2008 1079.9 Romania2009 2595.5

Armenia2009 1425.2 LaoPDR2009 450.0 Russia2009 2866.3

Belarus2008 2067.6 Mongolia2009 683.1 Samoa2009 1800.0

Bosnia & Herz.200 2041.4 Peru2006 2228.3 Tajikistan2008 217.2

Bulgaria2009 2412.6 Philippines2009 1201.7 Vietnam2009 617.1

Fyr Macedonia2009 2075.8

Note. The “Africa‖ in the table excludes South Africa, Botswana, Mauritius, Namibia.

164

BACKGROUND PAPER III

Assessing how the investment climate affects firm performance in Africa:

Evidence from the World Bank’s Enterprise Surveys

George Clark , May 2011

How do different aspects of the investment climate affect the performance of light manufacturing firms in

Africa? Many studies have tried to address this issue. This section assesses the different approaches that

different studies have used to try to answer this question and to try to summarize the main results from the

different studies. It focuses on studies and Investment Climate Assessments (ICAs) from the Africa

region and cross-country studies that include countries from the region.

Broadly speaking the main approaches are:

1. Asking firm managers what they see as the biggest barriers to their firms operations and

growth.

2. Analyzing the correlation between firm performance and different subjective and

objective measures of the investment climate.

This note will discuss these different approaches and some of the evidence obtained from studies that use

the different approaches. Most of the discussion will look at evidence from studies that have used

Enterprise Survey data—including Investment Climate Assessments—but some comparisons will be

made with evidence from other sources. When discussing Investment Climate Assessments, this report

will mostly focus on the most recent investment climate assessment for Ethiopia, Zambia, and Tanzania

(Regional Program on Enterprise Development, 2009b; Regional Program on Enterprise Development,

2009d; World Bank, 2009c).

I.2 Perceptions about the investment climate

Perhaps the most common approach to assessing what the binding constraints are to firm performance and

growth in a country is to ask managers what they see as the biggest obstacles. The World Bank

Enterprise Surveys include two types of questions about managers‘ perceptions. First, managers are

asked to rank a series of investment climate constraints on a five-point scale ranging from ‗no obstacle‘ to

a ‗very severe obstacle‘. Typically, firms are asked to rate about 15-25 different areas on this scale in

most Enterprise Surveys.1 Second, managers are asked which constraints are the most serious constraints

for the firm. For the second type of question, managers usually either note only the most serious obstacle

or the top three obstacles.

Although studies have used both measures of perceptions, most have focused on the first type of

questions (‗ratings‘) rather than the second (‗rankings‘).2 For this reason, this section will primarily focus

on ratings.

Rankings of constraints based upon perception data

Perception are often used to identify binding constraints in a country. For the ‗ratings‘ measure, this is

usually done by calculating the percent of firms that rated each constraint as a ‗major‘ or ‗very severe‘

165

obstacle(the top two ratings on the five-point scale) and ranking constraints based upon the percent of

firms that said that each was a serious problem. Dani Rodrik writes:

―These surveys are used increasingly to diagnose the main constraints facing firms and

to identify policy reform priorities. If, for example, firms in country A complain most

about the cost of finance while in B they complain about a skill shortage, this is taken as

an indication country is constrained by poor access to finance while country B is

constrained by poor human capital‖ 3

Although few investment climate assessments rely solely on this to identify constraints, most use this

information along with other information from the survey to identify constraints. Figure 35 below shows

the percent of firm managers that said that each investment climate constraint was a major or very severe

obstacle in Tanzania in 2006. Firms were most likely to say that electricity was a serious problem—close

to 90 percent of firms said it was a serious problem. This reflected the power crisis that was ongoing in

the country at the time of the survey—firms reported on average losing power over 20 days per month.

Firms were also concerned about access to finance and tax rates.

Figure 35: Firms managers in Tanzania were most concerned about electricity, access to

finance, and tax rates.

Source: Regional program on enterprise development (2009b).

Not surprisingly, the top concerns vary across countries. For example, firms in Zambia were most likely

to say that access to finance, competition from informal firms, and tax rates were major or very severe

problems (Regional Program on Enterprise Development, 2009d) . Relatively few Zambian firms were

concerned about electricity. The top concerns of Ethiopian firms were similar to Zambian firms-- access

to finance, competition from informal firms, and tax rates (World Bank, 2009c).

Figure 36 shows the number of countries where firm managers were most likely to say that said each of

the 15 constraints was the biggest constraint that they faced. By far the most common concerns are

0%

25%

50%

75%

100%

% o

f fi

rms

Percent of firms saying each area of investment climate is major or very severe problem in Tanzania

166

electricity (top constraint in 16 countries with available data) and access to finance (top constraint in 11

countries). In the other 11 countries, however, five additional constraints ranked as the top constraint—

emphasizing that there is considerable heterogeneity in perceptions across the continent.

Despite the differences, there are similarities. Gelb and others (2006a) find that there are some consistent

patterns across different types of firms in the poorest countries in Africa tend to be most concerned about

basic services and stability—macroeconomic stability, electricity and access to finance typically ranked

among the top concerns. As income increases, firms tend to become more concerned about the quality of

governance and the capability of the state—corruption, tax rates, tax administration and regulation

become increasingly binding. For example, the two countries where crime ranked as the top constraint—

South Africa and Namibia—are both middle-income countries.4

Figure 36: Firm managers in Sub-Saharan Africa said that electricity and access to finance

were the biggest problems

Data source: Dinh and others (Dinh and others, 2011)

Note: Biggest constraint is based upon ranking (biggest constraint) not ratings (how serious a constraint)

Differences across different types of firms

As well as differing across countries, constraints also differ across different types of firms. Most

Investment Climate Assessments include some analysis of how perceptions differ for different types of

firms.5 Typically, perceptions are broken down for groups in broad sectors of the economy (e.g., retail

trade, manufacturing and services), by size, by export status and sometimes by other firm characteristics.6

Because of the nature of the Investment Climate Assessments—they are aimed at policy makers and

donors in the respective countries not academics—few ICAs note whether the observed differences could

be due to either sampling variation (i.e., few note whether the differences are statistically significant).

Moreover, few control for other differences when comparing groups of firms (i.e., few analyze

differences in a regression framework). This could be important, for example, if large firms tend to be in

0

5

10

15

20

Electricity Access to Finance

Competition with

informal firms

Tax rates Crime Poltical instability

Business licensing

Nu

mb

er

of

cou

ntr

ies

Number of countries in Sub-Saharan Africa firm managers reported each area is biggest constraints

167

certain sub-sectors of the economy or in certain locations. Differences in perceptions due to sector of

operations might then be interpreted as differences due to size.

Some investment climate assessments and several cross-country studies have, however, included

econometric analyses of differences in perceptions between different groups of firms. The remainder of

this section will focus on statistically significant differences observed in these studies.

Size. Most studies that have looked at how different types of firms have different perceptions have

included controls for firm size—either a series of dummy variables or a continuous variables (e.g.,

number of workers). Results from two cross-country studies and several investment climate assessments

for Africa are shown in Table 24.

One of the most robust results is that smaller firms tend to be more concerned about access to finance

than large firms. In the two cross-country studies—one that looked at all countries with available data

(Hallward-Driemeier and Aterido, 2009) and one that focused only on countries in Africa (Gelb and

others, 2006a) small firms were significantly more likely to say that access to finance was a serious

problem than large firms were. 7 The same was true in two of the five investment climate assessments for

countries in Africa that have comparable regressions (Botswana and Tanzania). In the remaining three

countries, the coefficients on firm size are statistically significant.

Another robust result is that larger firms tend to be more concerned about labor regulation than small

firms. Once again this was true for both cross-country studies and for four of the five investment climate

assessments with available data. Large firms were also more concerned about other customs and trade

regulation in the larger cross-country study and in the ICA for Swaziland, although the difference was not

statistically significant in the other studies.

In three of the five investment climate assessments (Tanzania, Uganda and Namibia), larger firms appear

to be more concerned about electricity than smaller firms. This was not the case in the two remaining

ICAs or in the two cross-country studies. This result is, however, interesting given that two of the

countries with significant results were going through major power crises as the Enterprise Survey was

ongoing with firms suffering outages most working days (Regional Program on Enterprise Development,

2009b; Regional Program on Enterprise Development, 2009c). This might suggest that at least in cases

where there is a systematic crisis in a low income economy, large firms are less able to cope with it than

smaller, less capital intensive, firms.

Finally, in both cross-country studies, managers of large firms were more concerned about worker

education and skills than managers of small firms were. In the individual country studies, however, there

was little evidence that this was the case. This could be because there are differences across countries or

because samples in the individual country studies are too small to consistently find statistically significant

results.

For the other constraints, there was less consistency. For some constraints (e.g., tax rates, tax

administration, and telecommunications), coefficients were not statistically significant in most cases—

indicating that perceptions do not appear to be different for large and small firms. In other cases, results

in different studies give contradictory results. For example, large firms were less concerned about crime

in the larger cross-country study, but more concerned in three of the five individual countries within

168

Africa. These results suggest that the results might differ between countries—in some countries, small

firms might be more affected by problems related to specific areas of the investment climate in some

cases and less affected than others.

A final important point to make is that the results show differences in perceptions between different

groups of firms—but do not necessarily show that binding constraints would be different for different

groups of firms. A good example of this is power in Uganda and Tanzania. In both countries, large firms

were more concerned about power than small firms were. However, even though small firms were less

concerned about power than large firms, it was still the area of the investment climate that concerned

small firms the most. That is more small firms said that power was a serious problem than said the same

about any other area of the investment climate. That is, if firm perceptions were used to assess binding

constraints, one would probably conclude that even though power was a lesser constraint for small firms

than for large firms, it was probably the binding constraint for both. Given the extent of the power crises

in these two countries, this is probably not surprising.

Sector. The studies listed above pool data for firms across all areas of the economy. That is, they include

firms in all areas of manufacturing, in retail trade and in other service sectors. An interesting question is

whether constraints differ by sector. In practice, this is a difficult question to answer. Although both

cross-country studies include sets of sector dummies, neither report the coefficients results from the

studies discussed above are shown. Moreover, the investment climate assessments generally only include

broad sector dummies (manufacturing, retail, other) making it difficult to assess how firms in different

There are few consistent patterns across the different investment climate assessments in terms of

perceptions across broad sectors. The most consistent result is that manufacturing firms tend to be more

concerned about access to finance than service firms (see Table 25). This was true for three of the five

countries (Tanzania, Namibia, and Swaziland). There were no other consistent patterns across countries.

There are several possible reasons for this. It could simply reflect that sample sizes tend to be small—

making it difficult to find statistically significant results. But it could also reflect that there is a lot of

heterogeneity among firms in these broad sectors meaning that the data is noisy. A small tailor‘s shop

producing clothes for the local market would be very different from a large cement factory that employs

hundreds or thousands of workers and is therefore likely to have very different views about the

investment climate.

This suggests that it would be useful to provide more detailed breakdowns in terms of sectors. The focus

on broad sectors, however, probably reflects the small number of firms in most Enterprise Surveys in

Africa in even broad (i.e., two-digit ISIC) sub-sectors of manufacturing. Most of the Enterprise Surveys

are small—typically containing fewer than 200 manufacturing enterprises (see Table 26). Moreover, the

small size of the manufacturing sectors means that even the two oversampled sectors (food and garments)

typically include less than 50 firms. In practice, this will make it difficult to find statistically significant

differences at the sub-sector level—especially since even though subsectors are narrower than the broad

manufacturing sector, even the subsectors remain broad (e.g., they are basically at a 2-digit ISIC level).

Even within these sectors, there is a lot of heterogeneity in terms of size, products, and other features

between firms within the same broad sub-sector.

169

This problem could be reduced by expanding sample sizes for the Enterprise Surveys. This, however,

would require greater resources. An alternative approach would be to focus the surveys on more

homogeneous subsamples of firms. McKenzie (2011) notes that there are only about 100 manufacturing

firms with more than 100 employees in either Tanzania or Uganda. Getting robust subsamples for

medium-sized and large firms in any sub-sector of manufacturing would therefore be challenging for

these countries. This emphasizes the need for alternative approaches for large and medium-sized firms

when analyzing differences across subsectors (in terms of perceptions and enterprise performance)

Concerns about perception data

Although perception data is widely used to rank constraints and identify biding constraints, many people

have suggested that there are serious problems with doing this. Broadly speaking, these criticisms can be

grouped into three broad concerns. The first concern is that firm managers might not be able to provide

consistent and reliable information about specific investment climate problems.8 The second is that it

might not be possible to aggregate responses of individual managers in a reliable and consistent way. The

final concern is that even if enterprise managers can provide consistent information and this can be

aggregated that managers‘ concerns might not be useful for identifying constraints to private sector

development and economic growth

The most serious of the three concerns is the first. If managers‘ perceptions about specific area of the

investment climate reflect something other than the actual constraints in that area, it will be difficult to

rank constraints using this information.9

Some evidence suggests that this is a concern. In particular, Clarke (forthcoming) found results

consistent with the idea that managers‘ perceptions about specific areas of the investment climate reflect

their overall level of confidence about the economy—not just their perceptions about that specific area of

the investment climate. When the World Bank‘s 2007-2008 Enterprise Survey was being carried out, a

major electricity crisis hit South Africa. The crisis resulted in many more managers saying that power

was a serious constraint on enterprise operations—the share rose from about 10 percent of managers

before the crisis to close to 50 percent after the crisis. But managers also become more concerned about

most other areas of the investment climate after the power crisis that were unrelated to the power crisis.

Clarke (forthcoming) argued that this suggests that responses to questions about specific areas of the

investment climate do not just reflect concern about that area of the investment climate. They also reflect

overall business confidence.10

Although this suggests reason for concern—especially when using perception data for cross-country and

cross-time comparisons—it is not clear that this would completely invalidate using perception-based data

to identify binding constraints at a specific point in time. That is, although managers might become more

concerned about all areas of the investment climate during a crisis, if their concern increases to similar

degrees for all constraints, it might not affect relative rankings of constraints (including identification of

the top constraints). This appears to be the case after the power crisis in South Africa. Clarke

(forthcoming) notes that the relative rankings for constraints other than electricity remained similar after

the crisis. That is, other than for electricity, a researcher using perception data to identify constraints

would have identified similar constraints before and after the crisis.

170

Another broad response to these concerns is to note that managers perceptions about specific areas of the

investment climate appear to be correlated with objective measures of constraints. Hallward-Driemeier

and Alterido (2009), for example, find that many subjective measures of the investment climate from the

Enterprise Surveys are significantly correlated with objective measures of the investment climate.11

At

the cross-country level, however, although the correlations are often statistically significant, they are not

always high.

Even if managers can rank or rate constraints correctly, there is an additional problem of aggregating

perceptions across firms. Constraints affect different firms to different degrees and perception-based data

cannot be aggregated as easily as objective data (for example, costs measured in local currency). This

makes it difficult to rank obstacles. For example, it is not clear whether an issue that one firm considers

a very serious problem and another firm considers a minor problem, is more or less of a problem on

aggregate than one that both consider a moderately serious problem. Or is a problem that one says is the

biggest problem and another firm says is the seventh largest problem a greater or lesser constraint than

one that both rank as the third greatest constraint?

There is some evidence that this is a concern. For example, as noted above, the Enterprise Surveys asks

two questions related to perceptions. In Ethiopia, the top three constraints based upon the percent of

firms saying each was a major problem were in order competition from informal firms, access to finance,

and tax rates. Based upon the percent saying that each was the biggest problem they faced, the top four

(in order) were access to finance, access to land, and competition from informal firms (World Bank,

2009c). The top constraints are similar—but not identical.12

Overall, this suggests some reason for

caution when using these rankings to rate constraints and identify binding constraints.

If firm managers can accurately answer questions about constraints and these can be aggregated in a

meaningful way, then it should be possible to use their responses to measure at least what the managers

see as the major problems they face. But this does not mean that this provides useful information on what

the main constraints are to private sector development. A third question is whether the perceptions of the

enterprise managers interviewed in the survey reflect what the biggest constraints really are in the

country. That is, these rankings might represent what the main problems that the firm manager believe

they face, but these beliefs might not describe what the true barriers are to broader private sector

development or economic growth.

One reason why this might be the case is that the views of managers of interviewed managers might not

reflect the views of non-interviewed firms. In some cases, the omissions are due to conscious survey

design—most surveys only cover part of the economy. For example, the Executive Opinion Survey used

in the Global Competitiveness report primarily focuses on the large businesses that account the bulk of

employment in countries covered in that survey.13

Similarly, the World Bank‘s Enterprise Surveys are

conducted in the main cities in each country—usually between 3 and 5 locations—and only cover firms in

the manufacturing and service sectors with over five full-time employees.14

So, for example, rural firms

and firms in primary production (agriculture and mining) are excluded from the Enterprise Survey. If the

concerns of small, medium-sized and large manufacturing and service firms do not reflect the concerns of

microenterprises, rural firms, farms or mines, the results will need to be interpreted in this light.

In addition to not reflecting the concerns of firms that are excluded intentionally from the survey (e.g.,

firms in mining or agriculture or microenterprises), the survey might also not reflect the views of other

171

omitted firms. One group of unintentionally omitted enterprises is potential new entrants. These firms

might have different concerns about the investment climate than managers of existing firms. For

example, managers of existing enterprises that have already completed registration procedures might not

be concerned about entry costs even if they remain high. It is important, therefore, to think about how

constraints might affect new and potential entrants as well as how they affect the managers of existing

firms interviewed during the survey.

More broadly, another omitted group is firms that are unable to operate in a country due to problems in

the investment climate. For example, in countries where the cost or reliability of power supply is

particularly binding, firms that rely upon constant and cheap power might simply be unable to operate.

Similarly, if the ports and custom facilities are particularly poor, very few firms might operate in export-

oriented industries. Or if transportation costs are especially high or transportation infrastructure

particularly poor in some areas, firms that produce perishable, fragile, or heavy goods might not be able to

survive. Since you can only interview firms that exist—and by definition these are firms that have

managed to overcome the binding constraints—surveys of existing firms may underestimate the barriers

due to particularly binding constraints. Hausmann and Velasco (2005) illustrate this point with an analogy

to camel and hippos. They note that the few animals that you find in the Sahara will be camels, which

have adapted to life in the desert, rather than hippos, which depend heavily upon water. Asking the

camels about problems associated with life in the desert might not adequately represent the views of the

missing hippos

Problems of omitted firms will potentially affect both cross-country rankings if the industrial structure is

different in different countries and within country rankings if the omitted firms have different views on

certain areas of the investment climate. As discussed above, most evidence suggests that different types

of firms face different constraints, potentially making these omissions important.

A separate problem unrelated to possible differences between the perceptions of interviewed firms and

non-interviewed firms is the problem that enterprise managers‘ interests might not always be consistent

with society‘s interests. Most managers would like subsidized credit or to be charged less for electricity

or water if they believed that the cost of providing these services would be borne by someone else. They

would also prefer that the burden of taxes falls on others rather than themselves. And most would be

happy to face less competition even if the cost to society outweighed the benefits to their firm. It is

important, therefore, to think about how policy changes will affect other stakeholders (e.g., workers and

taxpayers) before adopting programs to reduce constraints.

These concerns emphasize that it is important to keep in mind the limits of perception-based data. In

particular, it is important to keep in mind: (i) that things other than concerns about the specific area being

asked about might affect perceptions; (ii) that asking questions and aggregating responses in different

ways might affect cross-country and cross-time comparisons and affect within country rankings; (iii) that

the perceptions of the interviewed firms do not necessarily reflect the perceptions of firms that are not

interviewed either due to conscious omissions or other reasons and (iv) that the views of the managers

might not be accurate reflections of the problem facing the economy as a whole.15

For this reason, it is

useful to use alternative source of information to try to identify constraints—at least to verify results from

perception-based analysis.

172

I.3 Relationship between constraints and firm performance

Rather than simply asking managers what they see as the major constraints they face and using these to

measure binding constraints to growth, a second and more direct approach that has been used is to try to

relate measures of investment climate constraints to firm performance. This approach is very attractive

because rather than relying only on perceptions, it allows the data to determine what are the most

important issues.

Many early papers using this approach focused on a small number of countries and one, or perhaps, a few

measures of the investment climate.16

For example, Beck and others (2005) focused on financial and

legal constraints and Fisman and Svensson (2007) focused on corruption and taxation. Other studies

focused on a broad number of different constraints, with the more recent studies pooling data for a large

number of firms in large numbers of countries (see Table 27). Most of these studies use data from low

and middle income throughout the world—although some break down results by continent, firm size, and

broad sector.

These analyses suggest several tentative conclusions (see Table 27 for summary of some recent studies):

1) There is strong evidence that investment climate constraints have a significant impact

on firm performance and growth. Most studies find correlations between objective

and subjective measures of investment climate constraints and measures of firm

performance and growth.

2) Different aspects of the investment climate affect different aspects of firm

performance to different degrees. For example, Harrison and others (2011) look at

how the investment climate affects labor productivity, sales growth, employment

growth, export share and investment share. They find very different results for the

different measures—none of the investment climate variables are correlated with

more than three measures of firm performance and most are only associated with one

or two measure of firm performance.

3) Most studies find that different aspects of the investment climate affect firm

performance differently for different groups of firms. Most evidence is for small and

large firms, although some studies breakdown results by broad sector (manufacturing,

retail, services), firm age, or broad areas of manufacturing (light manufacturing and

high-tech) For example, Aterido and Hallward-Driemeier (2010) find that small firms

are more affected by bureaucratic burden and medium and large firms are more

affected by the reliability of infrastructure.

4) Most, but not all, studies fund that different aspects of the investment climate often

appear to affect firm performance differently in different regions.

5) There is almost no evidence on how different aspects of the investment climate affect

firms in narrow sub-sectors of manufacturing (e.g., with breakdowns at the 4-digit or

even 2-digit ISIC level).

Although most studies find some investment climate constraints affect firm performance, there is

considerably less consensus on what areas are the most important and to whom. This is true even though

most of the studies use overlapping sources of data (e.g., from the same Enterprise Surveys) and even

when studies are focusing on the same measures of firm performance. Different studies find different

variables to be the most binding constraints and the reasons for these differences are not always clear.

For example, Dinh and others (2011) find that access to finance is the most binding constraint on firm

173

growth and that it appears to be most important for the small enterprises that are most excluded from

financial markets. In contrast, Harrison and others (2011) do not find any correlation between access to

formal credit or trade credit and their measures of sales and employment growth for either small or large

firms. They do, however, find that access to finance is associated with other measures of firm

performance. Aterido and Hallward-Driemeier (2010) find that employment growth is faster for small

firms in Sub-Saharan Africa when access to finance is more difficult—perhaps because they face less

competition from larger firms when access to finance is more difficult.17

Some results appear to be more common than others – although there is probably no variable that is

consistently significant or insignificant in all studies. Infrastructure in general and transportation and

trade infrastructure is often positively associated with measures of firm performance. There is less

evidence that corruption is significantly associated with firm performance—although Fisman and

Svensson (2007) find a negative correlation between corruption and firm growth, other studies have not

found similar results. Finally, although fewer studies look at the effect of regulation, some studies have

found that burdensome regulation (and labor regulation in particular) is associated with slower firm

growth.

Why are there disagreements on the main factors that affect firm

performance?

Given the lack of consensus on the main factors that affect firm performance, it is useful to consider why

this is the case. This subsection discusses some of these issues briefly. For further information, several

papers have discussed some of these issues in more depth in the context of estimating total factor

productivity (TFP) or technical efficiency. See, in particular, Escribano and Guasch (2005) and Dollar

and others (2005). 18

Although some of the criticisms in the papers above apply specifically to estimates

of TFP, many are relevant for other measures of firm performance.

Variations across countries. Most analyses pool large numbers of countries into single regressions. This

implicitly assumes that investment climate constraints affect firms in all countries equally. So, for

example, improving access to finance (e.g., increasing the share of firms that receive formal credit by ten

percentage points) would have the same effect in all countries irrespective of level of development or

other constraints. There are at least three reasons why this might not be the case.

1) The effect of investment climate improvements might depend upon the level of

development. For example, increasing the share of firms that have access to formal credit

might be very different in a countries where only 5 or 10 percent of firms have access and

where 70-80 percent have access. Although in theory this could be dealt with by

allowing for non-linear effects of IC constraints, this would exacerbate the problems of

many variable discussed below.

2) Improvements in any investment climate constraints might depend upon other

idiosyncratic features of the country. For example, increasing access to formal credit will

have very different effects in countries where informal institutions have evolved to make

up for problems with formal credit markets than on other countries.

3) Interactions between different constraints, might mean that changes in the investment

climate have very different effects in different countries. As Hausman and others (2005;

2008) discuss the effect of removing a constraint will depend upon what the other

174

constraints are. That is, moving from a second-best to first-best solution might not

improve performance when other distortions are in place. For example, increasing access

to formal credit is likely to have very different effects in countries where corruption is a

serious problem (and so improved credit is distributed as political favors) to other

countries.

Variations across different types of firms. Similar arguments could be made across different groups of

firms. Although existing studies have looked at some broad groups of firms (mostly defined based upon

firm size), little work has been done on specific subsectors. As discussed in the previous section, this is

probably because samples tend to be small in given subsectors—especially in African Enterprise Surveys

(see Table 26).

Endogeneity. Most of the investment climate constraints are potentially endogenous. That is, although

firm performance is affected by investment climate constraints, firm performance might also affect

constraints. Well managed—and therefore better performing—firms might be better able to get access to

finance, be better able to cope with infrastructure problems, and be better able to negotiate bureaucratic

hurdles. In practice, it is difficult or impossible to find instruments for all investment climate variables

that could be included in firm performance regressions.

The usual solution to the endogeneity problem is therefore to instrument or replace the firms‘ own

constraints with the average constraints by firms in the same city, sector and region.19

Aterido and others

(2011) show that controlling for endogeneity can have a large effect on results. They find that access to

finance, corruption, and power have a far more modest impact on firm growth after controlling for

endogeneity.

Using sector-city-size averages is not, however, a perfect solution to the endogeneity problem. In

particular, as noted by Levinsohn (2008), Verhoogen (2008), and Xu (forthcoming) discuss this will only

be valid if there are no other factors that vary consistently across regions that affect firm growth. For

example, Levinsohn (2008) argues that local government capacity is very difficult to control for in any

firm performance regression but is nearly certainly correlated with both firm performance and most

measures of the investment climate. If nothing else this argues for controlling for as many investment

climate variables as possible in the regressions—this will greatly increase the likelihood of controlling for

regional variation in the investment climate.

The large number of possible investment climate variables. The final issue is the large number of areas of

the investment climate that might affect firm performance (e.g., finance, electricity, corruption,

competition, crime, bureaucratic burden, regulation, taxes, and many others) and the large number of

ways to measure each constraint. Indeed, few of the studies listed above measure the different constraints

in the same way. For example, Aterido and Hallward-Driemeier (2011) use share of investment financed

with bank financing as their main variable, Dollar and others (2005) and Harrison and others (2011) use

whether firm has overdraft facility, and Dinh and others (2011) use whether access to finance is a serious

constraint.20

Some studies also use investment climate measures from sources other than the Enterprise

Survey data.

This large number of potential variables—at least in the 100s—leads to practical problems and might

explain the different results in different studies. The variables are potentially multicollinear and so

175

including hundreds of variables might not be possible. Moreover, allowing for non-linear effects or

interactions would amplify the problem considerably. In practice, it is difficult to assess how the various

studies arrived at their measures of the investment climate. This makes it difficult to assess why different

studies are finding different results.

I.4 Complementary Approaches

In addition to the broad approaches discussed above, several other approaches have been used to look at

how the investment climate affects firm performance. This section discusses two of these alternative

approaches: indirect costs and allocative efficiency.

Indirect costs.

For the most part, studies of firm performance that use Enterprise Survey data, such as the Investment

Climate Assessments for Africa and the studies listed above that calculate productivity measures such as

labor productivity and total factor productivity, measure output using sales data. This information is then

adjusted accounting for the use of certain inputs. For example, when calculating labor productivity, sales

are adjusted by the value of raw material and intermediate inputs used in production and is then divided

by the number of workers used. Total factor productivity is adjusted similarly, but also takes capital use

into account.

As Eifert and others (2008) point out these measures of productivity miss many costs that affect firm

profitability. For example, they note that these measures of firm performance omit the cost of

transportation, communications, security, corruption, and many other things. Following, Eifert and others

(2008), we will refer to these as indirect costs. That is, rather than including various measures of

investment climate constraints in a regression and seeing how they affect output or growth , Eifert and

others (2008) simply adjusts the measure of the firms‘ output so that it takes into account both the direct

costs of raw materials and intermediate inputs and the indirect costs listed above. They call this ‗net

value added‘.

Using data from 17 Enterprise Surveys from the early part of the decade, they show that many of these

indirect costs are far higher in Sub-Saharan Africa than in other regions. They note that transportation

and communication costs, in particular, appear to be particularly high in Sub-Saharan Africa. After

taking some of these additional costs into account, they show that African firms appear far less productive

compared to firms in other region than when they look only at conventional measures of productivity.

It is important to note that it is not possible to rank all investment climate related constraints using

measures of indirect costs as defined by Eifert and others (2008). That is, this approach would only

measure how investment climate problems affect costs—and some problems in the investment climate

affect firm performance in other ways. For example, many of the investment climate assessments that

have been completed in Africa emphasize the high cost that unreliable power imposed on firms within the

region. To the extent that this forces firms to by expensive fuel for generators, this will affect indirect

costs as they are calculated in Eifert and others (2008). However, power outages might also affect

output—either if the firm loses production during outages or if it uses less efficient production processes

that are less vulnerable to power outages. Although the measures of output used in Eifert and others

(2008) take the loss of output into account when calculating net value added, it would not be possible to

calculate the relative cost of outages (i.e., compared to transportation costs) without collecting additional

176

information. That is, it would be difficult to use an indirect cost approach to assess the relative impact of

investment climate constraints when they affect output (or growth opportunities) rather than costs.

Allocative efficiency and dispersion of productivity

The productivity analyses discussed above focus on either average measures of productivity and

performance (e.g., if means or OLS are used) or median measures (e.g., if medians or quantile regression

analysis are used). Several recent Investment Climate Assessment have noted that this approach

implicitly ignores the large variation in firm performance within and across countries (World Bank,

2009d; World Bank, 2010).

In practice, there are large gaps between the least and most productive in most countries—including in

Africa. Table 28 shows labor productivity for the median firms in countries in Sub-Saharan Africa and

for firms at the 25th and 75

th percentiles in terms of productivity.

21 In most countries, there is a large gap

between the least and most productive firms. For example, in Tanzania, the firm at the 25th percentile in

terms of labor productivity produces about $1,250 of output for each worker. The firm at the 75th

percentile produces about $9,050 per worker—about 7 times more. A similar gap can be shown for

microenterprises. For microenterprises in the manufacturing sector in Zambia, microenterprises at the

75th percentile produced about $2,000 of output per worker—about four times more than the

microenterprise at the 25th percentile (Clarke and others, 2010; Conway and Shah, 2010). Although the

comparisons above use labor productivity—and therefore do not take capital use into account—similar

dispersions can be shown for total factor productivity.

World Bank (2009d; 2010) notes that average productivity weighted by firm size, a, in a country with N

manufacturing enterprises can be written as:22

(1)

Where letters with bars above them represent unweighted means in terms of productivity, si represents

the market share of firm i; and ai represents the productivity of firm i.

Formula (1) above shows that there are two ways that aggregate productivity could be increased. One

way is to increase average productivity, . As individual firms become more productive (e.g., by

adopting new technologies or better management practices). The other would be to reallocate market

share, si, to firms that are more productive. That is, aggregate productivity will increase as the most

productive firms gain market share even if no individual firm becomes more productive. We refer to the

second way of increasing productivity as increasing allocative efficiency (i.e., ensuring that the most

productive firms in an industry are the ones that produce each good).

All else equal, competition should encourage allocative efficiency. That is, in competitive markets, we

would expect the most efficient firms to capture market share from less efficient firms and for efficient

new entrants to drive out inefficient incumbents.

World Bank (2009d; 2010) present comparisons of allocative efficiency—the extent to which the most

productive firms achieve dominant market share—for a collection of developing countries. The countries

from Sub-Saharan Africa for which they calculate allocative efficiency—Nigeria, Kenya, Zambia, and

177

South Africa—all perform poorly in terms of allocative efficiency. Of the 12 countries, all four African

economies rank in the bottom 5—only Argentina performs worse. As discussed above, this probably

reflects that low levels of competition in most countries in the region—poorly performing firms are not

forced to exit and better performing firms do not capture market share from their less competitive rivals.

An important question is why productivity varies so much between firms in the same country. There are

several possible factors that might lead to differences in productivity among firms in the same country

and industry. One factor is that differences in productivity might reflect differences in entrepreneurial or

management ability between different managers and owners in the industry. Better managed firms will

tend to be more productive than less well managed firms. Another factor is differences in the investment

climate faced by firms in the same countries and industry. Many studies have found large differences in

terms of the investment climate even within countries. A final factor is exogenous productivity shocks

that temporarily affect firm sales or performance. Sales might fall due to an unanticipated demand shock.

To the extent that the firm is unable to quickly adjust inputs (e.g., workers employed or capital used),

these shocks will affect measured productivity.

I.5 Conclusions

This paper discusses the various approaches that studies have used to try to identify the binding

investment climate constraints on firm performance. It focuses on two approaches: analyzing what firm

managers say that the major constraints that they face and analyzing the correlations between investment

climate constraints and measures of firm performance. The paper discusses some papers that have used

these approaches and the weaknesses and strengths of the approaches. It also discusses two related and

complementary approaches: analysis of indirect costs and allocative efficiency.

The main findings are:

1) The most common concerns among firm managers in Africa are basic concerns about

service and input provision: access to finance and power rank as the top concerns in over

two-thirds of countries in the region

2) Managers of small firms in the region appear to have different perceptions about the

investment climate than managers of large firms. In particular, access to finance tends to

be a greater concern for managers of small firms and labor regulation and worker

education are greater concerns for large firms

3) But there are significant differences in perceptions across countries and firm size and

other variables appear to affect perceptions differently in different countries. This

emphasizes the need for additional work.

4) There is strong evidence that investment climate constraints have a significant impact on

firm performance and growth. Most studies find correlations between objective and

subjective measures of investment climate constraints and measures of firm performance

and growth.

5) Most studies find that different aspects of the investment climate affect firm performance

differently for different groups of firms. Most evidence is for small and large firms,

although some studies breakdown results by broad sector (manufacturing, retail,

services), firm age, or broad areas of manufacturing (light manufacturing and high-tech).

6) There is less consensus on what areas of the investment climate are most binding overall.

Some results do, however, appear to be more common than others – although there is

178

probably no variable that is consistently significant or insignificant in all studies.

Improvements in infrastructure in general, and transportation and trade infrastructure in

particular, are often positively associated with measures of firm performance. In

contrast, corruption is not usually significantly associated with firm performance.

Finally, although fewer studies look at the effect of regulation, some studies have found

that burdensome regulation (and labor regulation in particular) is associated with slower

firm growth.

7) Most recent studies relating firm performance to IC variables pool data from many

countries and subsectors. In practice, given the large number of IC variables that could

affect performance and that sector-city-size averages of IC variables are often used to

control for endogeneity reduces degrees of freedom, this is probably necessary.

However, to the extent that binding constraints differ across countries—something that

would be consistent with the analysis of perceptions data—this approach obscures

differences between countries.

8) There is considerable heterogeneity in firm performance in Africa – higher than in other

regions. This probably partly reflects the importance of entrepreneurial and management

skills.

9) The high level of heterogeneity in firm performance—and the resulting low level of

allocative efficiency—probably at least partly reflects the low level of competition in

many countries. Inefficient firms are not driven from the market and entrepreneurs find

entry difficult.

10) Little work—either using perceptions or relating IC variables with firm performance—

has been done at a disaggregated sector level. Although several studies have compared

manufacturing and service firms, there have been few studies that have looked at

subsectors of manufacturing (e.g., at 2-digit or 4-digit ISIC levels)

179

I.6 Tables

Table 24: Differences in perceptions by firm size (positive sign means large firms are more concerned)

Hallward-Driemeier

and Alterido

Gelb and

Others

Tanzania

ICA Uganda ICA Botswana ICA

Namibia

ICA

Swaziland

ICA

Sample

105 countries

(including Africa)

26 countries in

Africa Tanzania Uganda Botswana Namibia Swaziland

Tax Rate non-linear ---

Political instability + --- -

+ +

Tax Administration

Worker education + +

+

Customs and trade regulation +

+

Labor Regulation + + +

+ + +

Access to finance - - -

- Corruption

+

+

Informality - ---

Crime - --- +

+ +

Electricity

+ +

+ Transport + ---

+

Telecom

---

+

Access to Land --- ---

-

Business licensing --- ---

-

Source: Hallward-Driemeier and Alterido (2009), Table 4; Gelb and Others (Gelb and others, 2006a), Table 3a; Regional Program on Enterprise Development

(2008a; 2008b; 2008c; 2009b; 2009c)

Note: Positive sign means that large firms were more concerned about that area of the investment climate. Negative sign means that small firms were more

concerned. Blank means that coefficient on size variable(s) were statistically insignificant; --- means that that variable was not included in the analysis.

180

Table 25: Differences in perceptions by sector (positive sign means manufacturers are more concerned)

Source Tanzania ICA Uganda ICA Botswana ICA Namibia ICA Swaziland ICA

Sample Tanzania Uganda Botswana Namibia Swaziland

Tax Rate

- -


Tax Administration

-

Worker education

+

Customs and trade regulation + -

-

Labor Regulation

Access to finance +

+ +

Corruption

Informality

+

Crime

- -

Electricity

Transport

Telecom

Access to Land

+

Business licensing

-

Source: Regional Program on Enterprise Development (2008a; 2008b; 2008c; 2009b; 2009c). The cross-country studies in Table 24 did not report coefficients

on sector dummies

Note: Positive sign means that manufacturing firms were more concerned about that area of the investment climate than retailers. Negative sign means that

retailers were more concerned. Blank means that coefficient on size variable(s) were statistically insignificant.

181

Table 26: Number of manufacturing observations, by sub-sector of manufacturing

Sector Total Leather Textiles Garments Food Metals and

machinery Electronics Chemicals

Wood and

furniture

Non-metallic

materials

Other

manufacturing

Angola 271

2 17 75 1

1

3 172

Benin 50

1

10 6

7 17 5 4

Botswana 145

2 25 12 3

7

4 92

Burkina Faso (2006) 51

14 1 1

35

Burkina Faso (2009) 98

5 4 19 29

7

8 26

Burundi 139

24 19

11

1 84

Cameroon (2006) 119

2

31 3

11

1 71

Cameroon (2009) 116

3 12 30 15 2 4

7 43

Cape Verde (2006) 47

12 2

1

32

Cape Verde (2009) 65

2 7 22 7 2 3 19 2 1

Chad 46

2 1 17 11

2 9 3 1

Congo 38

14 12

3

2 7

Congo, DR 192

18 56 1 1 11

4 101

Eritrea 95 5

2 38 4 1 4 11 18 12

Gabon 37

4 8 5 5

2

1 12

Gambia 62

3

59

Ghana 292

1 115 76 25 1 7

14 53

Guinea 137

6 35 27

3

7 59

Guinea Bissau 81

3 10 1

67

Cote d'Ivoire 193

7 43 29 23 3 14

18 56

Kenya 396

29 82 110 39

26 31 12 67

Lesotho 63

7 16 10 8 4 2

6 10

Liberia 73

1 7 6 2 2 1

2 52

Madagascar 204

15 53 46 8

12

4 66

Malawi (2005) 160

11

56 3 1 18

1 70

Malawi (2009) 70 1 6 1 21 10 1 7 9 9 5

Mali 301

2 138 96 27 1 11

8 18

Mauritania 128

1 4 27

3

1 92

Mauritius 183

26 30 70 11

3

8 35

Mozambique 341

52 89 83 3 7

20 87

Namibia 152

1 4 18 3

2

3 121

Niger 48 1 4 6 12 5

3 11 3 3

182

Sector Total Leather Textiles Garments Food Metals and

machinery Electronics Chemicals

Wood and

furniture

Non-metallic

materials

Other

manufacturing

Nigeria 948

14 223 302 104 8 35 136 15 111

Rwanda 68

1 5 21

7

1 33

Senegal 259

1 42 83 21 3 11

12 86

Sierra Leone 68

1 21 14 4

3

5 20

South Africa 680

9 105 114 145 22 83

30 172

Swaziland 106

5 15 14 3

4

65

Tanzania 286

3 51 70 4 1 15

8 134

Togo 28 1

6 7 3 3 3 1 4

Uganda 334

4 6 90 5

8

12 209

Zambia 304

6 57 114 36 1 10

19 61

Source: Enterprise Surveys

183

Table 27: Summary of recent papers of firm performance on investment climate

constraints using Enterprise Survey data

Study Findings

Aterido, Hallward-

Driemeier and Pages

(2009; 2011)

Main Findings:

1. Significant differences in how constraints affect depending on firm size

2. Business regulations appear to stunt the growth of small firms, but not medium-

sized or large firms.

3. Access to finance has a large impact on growth of small, medium, and large

enterprises (especially medium and large), but no impact of microenterprises.

4. Poor quality infrastructure appears to slow growth for medium-sized and large

firms, but not small or microenterprises

5. Less robust evidence of a strong impact of corruption (electricity and

transportation)

Dependent Variable: Employment growth

Main IC Variables: Finance, Bureaucratic Burden, Corruption, Infrastructure (Electricity

and Transportation).

Breakdowns: Size, country type (corrupt/non-corrupt, developed/undeveloped financial

markets; strong/weak rule of law); age of firm

Sample Size: 56,000 firms in 90 countries

Aterido and

Hallward-Driemeier

(2010)

Main Findings:

1. Investment climate constraints appear to have a different effect on firm growth in

countries in Sub-Saharan Africa than in other low income economies

2. Whereas access to finance is generally favorable for firm growth in low economies,

it is negatively or associated with firm growth in Sub-Saharan Africa

3. Access to finance is most negatively associated with growth for micro and small

enterprises in Sub-Saharan Africa

4. For larger enterprises in Sub-Saharan Africa, access to finance is not significantly

associated with firm growth

5. For large enterprises in Sub-Saharan Africa, firm growth is slower when power

infrastructure is unreliable

6. But small and microenterprises grow more quickly in Sub-Saharan Africa when

power infrastructure is unreliable—perhaps because of reduced competition from

efficient large firms

7. Business regulations appear to stunt the growth of small firms in Sub-Saharan

Africa, but not medium-sized or large firms


Main IC Variables: Finance, Bureaucratic Burden, Corruption, Infrastructure.

Breakdowns: Size, Region (Africa vs. other low-income economies)

Sample Size: 60,000 firms in Sub-Saharan Africa and other low and lower middle income

economies

184

Study Findings

Dinh, Mavridis, and

Nguyen (2011) Main Findings

1. Access to finance is the most binding constraint on firm growth—a result robust

across regions

2. Because small firms rely more heavily on internal financing, they would gain more

from reducing this constraint

3. Competition with the informal sector also appears to be a binding constraints.

4. Most other constraints are not robustly correlated with growth – and some have

counterintuitive signs


Main IC Variables included: Finance (objective and subjective measures); competition

with informal firms, worker education, regulatory burden, infrastructure, political instability,

tax rates, access to land, courts, and crime.

Breakdowns: By region (continent); size, broad sector (manufacturing, services, retail

trade); age.

Sample Size: 39,538 firms in 98 countries

Dollar, Hallward-

Driemeier and

Mengistae (2005)

Main Findings:

1. The investment climate variables are consistently associated with the various

measures of firm performance.

2. Firm productivity and profitability is higher when power outages are less common

and customs delays are smaller.

3. Access to finance is positively associated with sales growth

4. They report that corruption and governance do not appear to be associated with

their measures of firm performance

Dependent Variable: Total factor productivity, wages, profitability, sales growth.

Main IC Variables included: Infrastructure (telecom, power, and transport); finance;

barriers to trade.

Breakdowns: None

Sample Size: About 5,000 firms in Bangladesh, china, India, and Pakistan

Fisman and

Svensson (2007)

Main Findings:

1. Both tax rates and bribes are negatively associated with firm growth.

2. But bribes are more strongly negatively associated with firm growth

Dependent Variable: Sales Growth

Main IC Variables included: Tax rates; Corruption

Breakdowns: None

Sample Size:120 firms, Uganda

185

Study Findings

Harrison, Lin, and

Xu (2011)

Main Findings:

1. Weakness in various areas of the investment climate appear to explain most—if not

all—of the difference between countries in Africa and countries in other regions in

terms of firm performance

2. Many investment climate variables—and some from all groups—are significantly

correlated with some measures of firm performance (see Table 4);

3. Few measures of the investment care consistently correlated with multiple measures

(see Table 4)

4. Access to trade credit appears to be robustly correlated with multiple measures of

firm performance—labor productivity, export share and investment share.

5. The presence of political monopoly, crime, labor regulation, the quality of

transportation infrastructure, and availability of web services are correlated with

multiple performance measures.

6. Other areas of the investment climate such as the competition, corruption, and

access to formal finance are not strongly and robustly correlated with multiple

measures of performance (i.e., are significantly correlated with one or fewer

performance measures).

7. There are significantly differences across groups of firms (low tech compared to

high tech manufacturing; small compared to large and services compared to

manufacturing).

Dependent Variable: Value-added per worker (labor productivity); sales growth;

employment growth; exports as share of output; investment (share of value added)

Main IC Variables included: Competition; political stability; infrastructure

(transportation, power, telecommunications); finance; crime; and regulatory burden.

Breakdowns: Size, sector (manufacturing and services); sub-sector (light and high tech

manufacturing)

Sample Size: 12,000 manufacturing firms in 89 countries

Note: IC Variables are listed by type of variable. Actual variables included differed between studies.

186

Table 28: Labor productivity (value-added per worker) for firms at median, 25th

and 75th

percentiles

25

th percentile Median 75

th percentile

Ratio (75th

to 25th

percentile

Angola $3,377 $5,038 $7,648 2.3

Botswana $3,651 $8,966 $19,112 5.2

Burkina Faso $2,121 $6,454 $12,704 6.0

Burundi $858 $1,722 $3,346 3.9

Cameroon $2,499 $5,017 $12,526 5.0

Cape Verde $1,652 $4,226 $7,395 4.5

Congo $3,176 $8,885 $28,905 9.1

Congo, DR $1,381 $2,158 $4,027 2.9

Cote d'Ivoire $678 $1,315 $2,820 4.2

Eritrea $1,263 $2,925 $8,269 6.5

Gambia $496 $1,097 $2,915 5.9

Ghana $545 $1,093 $2,958 5.4

Guinea $692 $1,382 $2,961 4.3

Guinea Bissau $1,018 $1,719 $2,460 2.4

Kenya $1,944 $5,354 $12,981 6.7

Liberia $243 $545 $10,507 43.2

Madagascar $679 $1,690 $3,296 4.9

Malawi $1,516 $2,965 $7,362 4.9

Mali $1,260 $1,922 $3,497 2.8

Mauritania $2,417 $3,687 $6,761 2.8

Mauritius $4,196 $11,205 $16,430 3.9

Mozambique $1,010 $2,151 $4,173 4.1

Namibia $7,273 $14,967 $30,613 4.2

Nigeria $1,138 $2,263 $4,779 4.2

Rwanda $1,613 $3,125 $6,563 4.1

Senegal $1,845 $3,472 $9,090 4.9

Sierra Leone $751 $1,725 $7,435 9.9

South Africa $9,684 $18,682 $35,401 3.7

Swaziland $4,546 $8,398 $14,818 3.3

Tanzania $1,247 $2,613 $9,050 7.3

Uganda $1,099 $2,240 $5,615 5.1

Zambia $1,974 $4,035 $7,726 3.9

Source: Authors‘ calculation based upon data from the World Bank‘s Enterprise Surveys.

I.7 References

Aterido, Reyes, and Mary Hallward-Driemeier. 2010. "The Impact of Investment Climate on

Employment Growth: Does Sub-Saharan Africa Mirror Other Low-Income Regions." Policy

Research Working Paper 5218. World Bank, Washington DC. Available on line at

http://go.worldbank.org/KDAS6GU2L0.

Aterido, Reyes, Mary Hallward-Driemeier, and Carmen Pages. 2009. "Big Constraints to Small Firms'

Growth: Business Environment and Employment Growth across Firms." Policy Research

Working Paper 5032. World Bank, Washington DC. Available on line at

http://go.worldbank.org/SE95ERXVV0.

--------. 2011. "Big Constraints to Small Firms' Growth: Business Environment and Employment Growth

Across Firms." Economic Development and Cultural Change 59(3):609–647.Available online at

http://www.jstor.org/stable/10.1086/658349.

Azfar, Omar, and Peter Murrell. 2009. "Identifying Reticent Respondents: Assessing the Quality of

Survey Data on Corruption and Values." Economic Development and Cultural Change

57(2):387–411.Available online at http://dx.doi.org/doi:10.1086/592840.

Beck, Thorsten, Asli Demirgüç-Kunt, and Vojislav Maksimovic. 2005. "Financial and Legal Constraints

to Firm Growth: Does Firm Size Matter?" Journal of Finance 60(1):137–177.

Bertrand, Marianne, and Sendhil Mullainathan. 2001. "Do People Mean What They Say? Implications for

Subjective Survey Data." American Economic Review: Papers and Proceedings 91(2):67–72.

Clarke, George R. G. 2009. "Beyond Tariffs and Quotas: Why Do African Manufacturers Not Export

More?" Emerging Markets Finance and Trade 45(2):44–64.

--------. Forthcoming. "Are Managers' Perceptions about Constraints Reliable? Evidence from a Natural

Experiment in South Africa." Journal of Globalization and Development.

--------. Forthcoming. "How Petty is Petty Corruption? Evidence from Firm Surveys in Africa." World

Development. Available on line at http://dx.doi.org/doi:10.1016/j.worlddev.2010.11.002 .

Clarke, George R. G., Robert Cull, and Maria Soledad Martinez Peria. 2006. "Foreign Bank Participation

and Access to Credit Across Firms in Developing Countries." Journal of Comparative Economics

34(4):774–796.Available online at http://dx.doi.org/doi:10.1016/j.jce.2006.08.001 .

Clarke, George R. G., James Habyarimana, Michael Ingram, David Kaplan, and Vijaya Ramachandran.

2007. An Assessment of the Investment Climate in South Africa. Washington DC: World Bank.

Clarke, George R. G., James Habyarimana, David Kaplan, and Vijaya Ramachandran. 2008. "Why Isn't

South Africa Growing Faster? Microeconomic Evidence From a Firm Survey." Journal of

International Development 20(7):837–868.

http://go.worldbank.org/KDAS6GU2L0

http://go.worldbank.org/SE95ERXVV0

http://www.jstor.org/stable/10.1086/658349

http://dx.doi.org/doi:10.1086/592840

http://dx.doi.org/doi:10.1016/j.worlddev.2010.11.002

http://dx.doi.org/doi:10.1016/j.jce.2006.08.001

188

Clarke, George R. G., Manju Kedia Shah, and Roland V. Pearson. 2010. "The Business Landscape for

MSMEs and Large Enterprises in Zambia." World Bank: Washington DC.

Conway, Patrick, and Manju Shah. 2010. "Who's Productive in Zambia's Private Sector." World Bank:

Lusaka, Zambia.

Dinh, Hinh T., Dimitris A. Mauvridis, and Hoa B. Nguyen. 2011. "The Binding Constraint on Firms'

Growth in Developing Countries." Policy Research Working Paper 5485. World Bank,

Washington DC. Available on line at http://go.worldbank.org/LGDC8TDE00.

Dollar, David, Mary Hallward-Dreimeier, and Taye Mengistae. 2005. "Investment Climate and Firm

Performance in Developing Countries." Economic Development and Cultural Change 54(1):1–

31.Available online at http://dx.doi.org/10.1086/431262.

Eifert, Benn, Alan Gelb, and Vijaya Ramachandran. 2008. "The Cost of Doing Business in Africa:

Evidence From the World Bank's Investment Climate Surveys." World Development 36(9):1531–

1546.

Escribano, Alvaro, and J. Luis Guasch. 2005. "Assessing the Impact of the Investment Climate on

Productivity Using Firm-Level Data: Methodology and the Cases of Guatemala, Honduras and

Nicaragua." Policy Research Working Paper 3621. World Bank, World Bank, Washington DC.

Available on line at http://go.worldbank.org/F4W5VBGDR0.

Fisman, Raymond, and Jakob Svensson. 2007. "Are Corruption and Taxes Really Harmful to Growth?

Firm Level Evidence." Journal of Development Economics 83(1):1–37.Available online at

http://dx.doi.org/doi:10.1016/j.jdeveco.2005.09.009 .

Gelb, Alan, Vijaya Ramachandran, Manju Kedia Shah, and Ginger Turner. 2006a. "What Matters to

African Firms? The Relevance of Perceptions Data." World Bank: Washington DC. Processed.

--------. 2006b. "What Matters to African Firms? The Relevance of Perceptions Data." World Bank:

Washington DC. Processed.

Hallward-Driemeier, Mary, and Reyes Aterido. 2009. "Comparing Apples with...Apples: How to Make

(More) Sense of Subjective Rankings of Constraints to Business." Policy Research Working

Paper Paper 5054. World Bank, Washington DC.

Harrison, Ann E., Justin Yifu Lin, and Lixin Colin Xu. 2011. "Explaining Africa's (Dis)advantage."

World Bank: Washington DC. Processed.

Hausmann, Ricardo, Bailey Klinger, and Rodrigo Wagner. 2008. "Doing Growth Diagnostics in Practice:

A Mindbook." CID Working Paper 177., Center for International Development, Harvard

University, Cambridge MA. Available on line at http://www.cid.harvard.edu/cidwp/177.html.

Hausmann, Ricardo, Dani Rodrik, and Andres Velasco. 2005. "Growth Diagnostics." Harvard University:

Cambridge, MA. Processed.

http://go.worldbank.org/LGDC8TDE00

http://dx.doi.org/10.1086/431262

http://go.worldbank.org/F4W5VBGDR0

http://dx.doi.org/doi:10.1016/j.jdeveco.2005.09.009

http://www.cid.harvard.edu/cidwp/177.html

189

Hausmann, Ricardo, and Andres Velasco. 2005. "Slow Growth in Latin America: Common Outcomes,

Common Causes?" Kennedy School of Government: Boston MA. Processed.

Hellman, Joel, Geraint Jones, Daniel Kaufmann, and Mark Schankerman. 1999. "Measuring Governance

and State Capture: The Role of Bureaucrats and Firms in Shaping the Business Environment."

EBRD Working Paper 51. European Bank for Reconstruction and Development, London, UK.

Iarossi, Giuseppe. 2009. An Assessment of the Investment Climate in Kenya. Washington DC: World

Bank.

Iarossi, Giuseppe, and George R. G. Clarke, eds. 2011. Nigeria 2011: An Assessment of the Investment

Climate in 26 States. Washington DC: World Bank.

Iarossi, Giuseppe, Peter Mousley, and Ismail Radwan. 2009. An Assessment of the Investment Climate in

Nigeria. Washington DC: World Bank.

Jensen, Nathan M., Quan Li, and Aminur Rahman. 2008. "Heard Melodies Are Sweet, But those Unheard

Are Sweeter : Understanding Corruption Using Cross-National Firm-Level Surveys." Policy

Research Working Paper 4413., World Bank, Washington DC. Available on line at

http://go.worldbank.org/FVMFX9KVJ0.

Levinsohn, James. 2008. "Comments on Methodology Used by Escribano, Guasch, and Co-Authors in

Analyzing World Bank Investment Climate Surveys." University of Michigan: Ann Arbor, MI.

Processed.

McKenzie, David. 2011. "How Can We Learn Whether Firm Policies Are Working in Africa? Challenges

(and Solutions?) for Experiments and Structural Models." Policy Research Working Paper 5632.

World Bank, Washington DC. Available on line at http://go.worldbank.org/O6FCOLQZD0.

Pakes, Ariel. 2008. "Comments on Methodology Used in Analyzing World Bank Investment Climate

Survey." Harvard University: Cambridge, MA. Processed.

Porter, Michael E, Klaus Schwab, Xavier Sala-i-Martin, and Augusto Lopez-Claros. 2005. The Global

Competitiveness Report 2004-2005. Geneva, Switzerland: World Economic Forum.

Regional Program on Enterprise Development, Africa Private Sector Group. 2008a. "An Assessment of

the Investment Climate in Botswana." World Bank: Washington DC.

--------. 2008b. "An Assessment of the Investment Climate in Namibia." World Bank: Washington DC.

--------. 2008c. "An Assessment of the Investment Climate in Swaziland." World Bank: Washington DC.

--------. 2009a. "An Assessment of the Investment Climate in Mauritius." World Bank: Washington DC.

--------. 2009b. "An Assessment of the Investment Climate in Tanzania." World Bank: Washington DC.

--------. 2009c. "An Assessment of the Investment Climate in Uganda." World Bank: Washington DC.

http://go.worldbank.org/FVMFX9KVJ0

http://go.worldbank.org/O6FCOLQZD0

190

--------. 2009d. "An Assessment of the Investment Climate in Zambia." World Bank: Washington DC.

Rodrik, Dani, 2007. The (Mis)Use of ICAs. Retrieved 11-8-2009, from Dani Rodrik web site:

http://rodrik.typepad.com/dani_rodriks_weblog/2007/12/the-misuse-of-i.html.

Schwab, Klaus. 2009. The Global Competitiveness Report 2009-2010. Geneva, Switzerland: World

Economic Forum.

Svensson, Jakob. 2003. "Who Must Pay Bribes and How Much? Evidence From a Cross Section of

Firms." Quarterly Journal of Economics 118(1):207–230.

Verhoogen, Eric. 2008. "Comments on Escribano-Guasch methodology for Estimating the Effects of

Investment Climate on Productivity." Columbia University, New York, NY. Processed.

World Bank. 2009a. "An Assessment of Zambia's Private Sector." World Bank: Washington DC.

--------. 2009b. "Enterprise Survey and Indicator Surveys." World Bank: Washington DC. Available on

line at http://www.enterprisesurveys.org/documents/Sampling_Note.pdf.

--------. 2009c. "Toward the Competitive Frontier: Strategies for Improving Ethiopia's Investment

Climate." World Bank: Washington DC. Available on line at

http://go.worldbank.org/HDWRW0SY80.

--------. 2009d. "Zambia Second Investment Climate Assessment: Business Environment Issues in

Diversifying Growth." Wold Bank: Washington DC.

--------. 2010. "South Africa Second Investment Climate Assessment: Business Environment Issues in

Shared Growth." World Bank: Washington DC. Available on line at

http://go.worldbank.org/M5RMVWHJ30.

Xu, Lixin Colin. Forthcoming. "The Effects of Business Environments on Development: Surveying New

Firm-Level Evidence." World Bank Research Observer.

http://rodrik.typepad.com/dani_rodriks_weblog/2007/12/the-misuse-of-i.html

http://www.enterprisesurveys.org/documents/Sampling_Note.pdf

http://go.worldbank.org/HDWRW0SY80

http://go.worldbank.org/M5RMVWHJ30

191

BACKGROUND PAPER IV

Wages and Productivity in Manufacturing in Africa:

Some Stylized Facts

George Clarke

February 2011

I.1 Introduction

Although there is no shortage of unskilled workers in most low-income economies in Sub-Saharan Africa,

few countries in the region have successfully entered export-oriented labor-intensive manufacturing. One

possible reason for this is that labor costs for formal firms are relatively high in Africa compared to other

countries at similar levels of development. As shown in this note, although formal sector wages are low

in absolute terms, they are higher on average in Africa than in other countries with similar levels of per

capita GDP. It is plausible that this might affect the export competitiveness of firms in the region. This

note explores this question in greater depth, comparing labor costs and productivity in Sub-Saharan

Africa with labor costs and productivity in other regions using data from the World Bank‘s Enterprise

Surveys.

This study focuses on the formal manufacturing sectors in Sub-Saharan Africa and other regions of the

world.xxiii

The focus on manufacturing is appropriate because performance measures are more easily

compared for firms in the same sectors of the economy. It also follows from data restrictions.

Accounting data is only collected for manufacturing firms in most Enterprise Survey meaning that the

productivity measures that the paper focuses on are not available for other sectors of the economy (e.g.,

retail trade or other service sectors).

Exporting in Africa

Few countries in Sub-Saharan Africa have been successful in export-oriented manufacturing. In part, this

reflects that few countries have a large manufacturing base. On average, manufacturing accounts for only

about 13 percent of GDP between 2005 and 2009 for countries in the region—lower than for developing

countries in any other region except North Africa and the Middle East (see Table 30). For example,

manufacturing accounts for about 31 percent of GDP on average for developing countries in East Asia.

Moreover, other than in a handful of middle-income economies in Southern Africa (South Africa,

Lesotho, Swaziland), Mauritius, and Cote d‘Ivoire, manufacturing accounted for less than 15 percent of

GDP in all countries. And it accounted for less than 10 percent of GDP in most of these countries.

As a result, countries in Sub-Saharan Africa mostly rely on narrow ranges of primary commodities in

most countries (Collier, 1998). A recent study noted that in the late 1990s, 39 of 47 of African countries

depended on two primary commodities for over half of their export earnings (Morrissey and Filatotchev,

2000). As a result, countries in the region are highly vulnerable to terms-of-trade shocks. Diversifying

exports away from primary commodities into labor-intensive manufacturing, which currently accounts for

only a relatively modest share of GDP and an even more modest share of exports, could reduce this

vulnerability.

192

Evidence from the Enterprise Surveys

Data from the enterprise surveys confirms that manufacturing enterprises in most, but not all, countries in

Sub-Saharan Africa are heavily focused on external markets. In most countries, less than one in five in

manufacturing enterprises export any part of their output (see Figure 37). Given the small size of the

manufacturing sectors in most countries, this means that manufacturing exports remain unimportant in

most economies in the region.

Figure 37: Percent of firms that export, by region

Source: Author‘s calculations based upon data from World Bank‘s Enterprise Surveys

Note: See Table 33 for additional notes on data construction. East Asia is China, Indonesia, Philippines, Thailand and

Vietnam. Africa is Sub-Saharan Africa only. Data are for all Enterprise Surveys conducted since 2006 with at least 50

firms. Countries with GDP over $8,000 are excluded for presentational purposes. Fitted values is line from log-log

regression

Although manufacturing enterprises in many African countries have been relatively unsuccessful in

export markets, there are significant differences between countries. For example, both macroeconomic

data and the firm-level data used in this paper suggest that manufacturing enterprises in Kenya, where 47

percent of firms export some part of their output, are more successful than enterprises in most other

countries in the region.

Even in successful manufacturing countries, however, many enterprises only export to neighboring

countries within Sub-Saharan Africa rather than to Western Europe or other industrialized economies.

For example, the most important export destinations for firms in the 2003 Enterprise Survey in Kenya

were Uganda and Tanzania, with 74 percent and 61 percent of exporters exporting to these countries. In

comparison, only 8 percent of exporters exported to the United Kingdom, the biggest export market

among industrialized countries.

0%

10%

20%

30%

40%

50%

60%

70%

80%

$0 $1,000 $2,000 $3,000 $4,000 $5,000 $6,000 $7,000 $8,000

Val

ue

ad

de

d p

er

wo

rke

r (U

S$)

GDP per capita (2005 PPP $)

Non-Africa Fitted Values Africa East Asia

193

Although the most recent Enterprise Surveys do not collect information on the destination of exports,

information was collected on destination in earlier surveys. In the Enterprise Surveys conducted between

2002 and 2004, enterprises in all countries other than Ethiopia were more likely to export to neighboring

countries than they were to export to more distant European markets (see Table 31). This pattern appears

to be true for both landlocked countries (e.g., Uganda and Zambia) and countries with access to the sea

(e.g., Tanzania and Kenya).xxiv

In contrast to Africa, the most important markets for Chinese goods are overseas (i.e., China does not

share a land border with its most important markets). Although the most popular destination for ‗exports‘

from China was Hong Kong, the three next most important destinations were the United States (32% of

exporters rated it among their 3 most important destinations), Japan (31%) and Germany (10%).

Benefits of Exporting

Boosting exports is an important goal. In addition to reducing vulnerability to shocks, increasing exports

might boost income by increasing economic growth.xxv

Exporters tend to be more efficient than non-

exporters—something that holds for the enterprises in this study.xxvi

If the enterprise-level correlation

between exporting and efficiency is due to exporting improving productivity, increasing exports might

increase income.

There has been considerable debate over whether this is the case—the correlation could also simply be

due to productive enterprises self-selecting into exporting. Under the first hypothesis, the learning-by-

exporting hypothesis, the discipline of competing in international markets encourages enterprises to

improve their productivity and exposes them to foreign technologies and modes of production. Under the

second hypothesis, the self-selectivity hypothesis, only firms that are already efficient are able to export.

Although inefficient firms are protected from international competition in domestic markets by natural

barriers (e.g., high transportation costs) and policy barriers (e.g., government tariffs and quotas) to trade,

they are unable to enter international markets. It is important to note that the two hypotheses are not

mutually exclusive. Even if more productive enterprises self-selected into exporting, it would still be

possible that exporting results in further productivity improvements.

Although there is no definitive answer as to which hypothesis better explains the higher productivity of

exporters, recent enterprise-level studies have found evidence consistent with the learning-by-exporting

hypothesis in Africa.xxvii

Using enterprise-level data from the mid-1990s for Cameroon, Ghana, Kenya,

and Zimbabwe, Bigsten et al. (2004), who use simultaneous equations estimation to control for reverse

causation, find that exporting results in efficiency gains.xxviii

In addition, Mengistae and Pattillo (2004)

find that direct exporters and firms that export outside of Africa are more productive than other exporters,

which they interpret as consistent with the learning-by-exporting hypothesis. If exporting resulted in

productivity improvements, policies that promote exports—or at least remove biases that discourage

exports—might improve productivity and ultimately result in higher wages and income.

This paper uses enterprise-level data from over 30 countries to explore different factors that affect export

performance, focusing mostly on the role of wage costs. Throughout the paper comparisons will be made

to manufacturing firms in other regions and in several successful manufacturing economies in East Asia.

194

I.2 Data

The firm-level surveys used in this paper were conducted as part of the World Bank‘s Enterprise Surveys

project. Because performance measures are more easily compared for firms in the same sectors of the

economy and because many of the most important measures of productivity are collected only for

manufacturing firms, the analysis is restricted to manufacturing firms. This restriction is also necessary

because accounting data is generally only collected for manufacturing firms in the Enterprise Survey

program.

Another important restriction is that the analysis omits microenterprises and informal enterprises.

Although separate microenterprise surveys, which include many informal enterprises, were conducted in

most countries in Sub-Saharan Africa similar surveys have not been conducted in countries in other

regions. In summary, this paper focuses on formal manufacturing firms with over 5 employees. Given

that these are the firms that are most likely to be involved in exporting, this focus seems appropriate.

Although, as discussed below, the Enterprise Survey data is carefully collected using a standard and

rigorous approach, it is important to note that the data has some limitations. One particular concern is that

firm managers might not report accurate data to interviewers. One problem might be that firm manager

underreport their sales because they are concerned about the tax authorities using the data to target them.

That is, although managers are assured that the data are confidential, they might underreport due to a

sense of concern (‗better safe than sorry‘). On the other hand, other pressures might lead them to over-

report their performance. In particular, managers are often concerned about their competitors gaining

information on their performance. Recent studies using Enterprise Survey data for Nigeria suggests that

managers who appear deceptive (‗reticent‘) appear to over- not underreport labor productivity.xxix

Surveys

The surveys provide representative samples of each country‘s private sector. The surveys collect

information on standard accounting measures of firm performance and many areas of each country‘s

investment climate. Summaries of the broad survey results are included on the Enterprise Survey website

(www.enterprisesurveys.org). More detailed summaries of Enterprise Surveys have been completed for

the World Bank‘s Investment Climate Assessment program.

Since 2006, the surveys have used almost identical questionnaires and identical sampling methodologies.

They were conducted in between two and five cities in each country and covered firms with more than 5

employees in the following industries (according to ISIC, revision 3.1): all manufacturing sectors (group

D), construction (group F), retail and wholesale services (sub-groups 52 and 51 of group G), hotels and

restaurants (group H), transport, storage, and communications (group I), and computer and related

activities (sub-group 72 of group K). The surveys—including the corruption questions—were delivered

to the managing director or his or her direct representative. Accountants or human resource managers

sometimes provided information on company accounts and labor practices.

The samples were stratified random samples—stratified by industry, location, and, when information was

available, by firm size. Sample sizes were set within each stratum to allow certain variables to be

calculated to specified levels of precision. Firms were randomly selected from each group. Weights were

constructed to re-weight the population to take account of varying probabilities of selection between

strata.

http://www.enterprisesurveys.org/

195

The sampling frames were constructed from lists of registered enterprises in each country. Lists of firms

were obtained from government agencies in each country—usually the Bureaus of Statistic, Business

Registrar, or Ministry of Trade and Industry. These lists were then verified and updated to get complete

and up-to-date sampling frames. The survey methodology is described in more detail and the survey data

are available for download on the World Bank‘s Enterprise Surveys website

(www.enterprisesurveys.org).

Although Enterprise Surveys were conducted before 2006, the analysis is restricted to this period because

data collection was less uniform before 2006. Before 2006, there was considerable heterogeneity across

countries and regions in terms of sectors covered, questionnaire format, coverage of microenterprises with

less than 5 employees, and sampling methodology. Moreover, the Enterprise Surveys conducted before

2006 were not generally representative of the formal economies in this region and weights are not

available that would allow for the computation of population averages.

Productivity Data

The study focuses on several measures of firm productivity. These are calculated in a uniform way in all

countries with available Enterprise Survey data from between 2006 and 2009. The measure of firm

performance are:

Labor Productivity. Value-added per worker is the basic measure of labor productivity used in this paper.

It is the value of the goods and services that the firm produces less the cost of the raw materials (such as

iron or wood) and intermediate inputs (such as engine parts or textiles) used to produce the output divided

by the number of full-time workers in the firm.xxx

Firms that produce more output with less raw material

and fewer workers have higher labor productivity. Differences in labor productivity can be the result of

differences in technology, differences in organizational structure, differences in worker education or

skills, differences in management ability, or differences in capital use. Because labor productivity does

not take the use of capital (i.e., machinery and equipment) into account, it will generally be higher in

firms that use capital in place of labor (i.e., firms that are capital intensive).

Labor costs per worker. The cost of labor is the cost of wages, salaries, bonuses, other benefits, and

social payments for workers at the firm divided by the number of workers. The data is taken from the

firms‘ accounts. It therefore includes wages and salaries paid to all workers and managers – not just

production workers.

Average monthly wage for production workers. The average monthly wage for production workers. This

information will generally not come from company accounts. Instead it reflect the manager‘s views about

wages paid to typical production workers. Although it should generally provide information very similar

to the measure ‗labor costs per worker‘, it provides a useful robustness check since it will not generally

come from the source of information (i.e., not from company accounts).

Unit labor costs. This measure is labor costs as a percent of value-added. Although it is an

approximation to true unit labor costs (i.e., it measures output in dollar terms rather than as physical

measure of production), it can be calculated using information from the Enterprise Survey. It is a better

measure of labor costs than labor cost per worker in that it makes it easier to assess the net impact of labor

costs on competitiveness by taking differences in productivity into account when assessing labor costs.

Unit labor costs are higher when higher labor costs are not fully reflected in higher productivity. When


196

unit labor costs are higher (i.e., when labor costs are higher compared to productivity), all else equal,

firms will find it more difficult to compete on international markets than when they are lower. Although

unit labor costs are not the only factor that affect competitiveness—for example, they do not take the cost

of capital or capital intensity into account—they are a better measure of competitiveness than labor costs

alone.

Capital intensity. This is the amount of capital that the firm uses per workers. Firms that use more

capital per worker are more capital intensive. Two measures of capital per worker are used: the book

value (i.e., the depreciated amount of capital from the company‘s accounts) and the sales value (i.e., the

manager‘s view on how much they could sell machinery and equipment for given its current condition).

Although capital intensity provide some context for results for labor productivity, it is important to note

that it is more difficult to measure capital than it is to measure labor (e.g., it is relatively easy to measure

wages and number of workers). Because most machinery is long-lived and provides services over a long

period of time, it is difficult to measure its contribution to output in a single year. As capital ages, it

becomes less productive (i.e., it depreciates in value) and will eventually stop producing anything, either

by breaking or becoming obsolete. Although accounting rules for depreciating machinery and equipment

exist, these often bear little resemblance to true rates of economic depreciation—and can vary across

countries. The book value of capital (i.e., the value of capital included in company accounts) is therefore

not an especially accurate measure of the value of capital—especially for small firms that do not keep

detailed audited accounts. The sales value is similarly problematic in that managers‘ might not know the

value of the equipment in its current condition and when markets for machinery are thin, its sales value

might not reflect its true economic value.

I.3 Firm Performance in Sub-Saharan Africa

Among other things, whether firms can export will depend upon how productive they are and their labor

costs. Using data from the World Bank‘s Enterprise Surveys, this section discusses some stylized facts

about firm productivity and labor costs in Sub-Saharan Africa. As noted earlier, the analysis focuses on

formal firms in the manufacturing sector. As discussed below, formal firms account for only a small

share of employment in many countries in Sub-Saharan Africa. But these firms are likely to account for

most manufacturing exports. Although microenterprises are important in terms of employment, most will

need to enter the formal sector and expand to enter export markets. It is costly to enter export markets

and, combined with economies of scale, this means that microenterprises are unlikely to be able to do so

without expanding.xxxi

Both labor costs and labor productivity are relatively low in Sub-Saharan Africa (see Table 32). In the

average country in Sub-Saharan Africa, labor productivity is $4,734 per worker and per worker labor

costs are $1,464 per worker. Labor costs are particularly low in the low and lower middle income

economics – value added is $3,316 per worker and labor costs are $1,059 per worker.

This is lower than in any region other than South Asia where Enterprise Surveys have conducted.

Moreover, it is important to note that the two countries where surveys have been conducted in South Asia

(Afghanistan and Nepal) are not necessarily representative of the region as a whole. In comparison, labor

productivity is $6,713 per worker in the average country in East Asia with a strong manufacturing base

and per worker labor costs are $1,629.

197

Labor costs and labor productivity are both low in Sub-Saharan Africa. Labor costs, however, appear

relatively lower than labor productivity. As a result, unit labor costs are not particularly high in Sub-

Saharan – 33 percent of value added overall and 34 per cent for low and lower middle-income economies.

This is lower than in Europe and Central Asia (38 percent), Latin America (37 percent), or South Asia (40

percent). It is, however, slightly higher than in the manufacturing economies in East Asia (28 percent).

Firm productivity in Sub-Saharan Africa

As noted above, labor productivity is relatively low in Sub-Saharan Africa compared to other regions. It

is important to note that labor productivity is affected by many things internal and external to the firm. At

the firm level, labor productivity – defined as value-added per workers—will be affect by the capital

intensity of the firm and the human capital of workers. These, in turn, will be affect by the availability of

external financing and the availability of skilled and educated workers. Labor productivity will also be

affected by other country-level characteristics such as physical infrastructure, the institutional

environment, and the quality of government regulation. Previous studies have found that many African

countries compare unfavorably with other countries in terms of both the institutional environment and the

quality of physical infrastructure.xxxii

Although per capita GDP is not a perfect proxy for these external factors that affect firm performance,

many of them appear to vary systematically with income. In addition to the quality of infrastructure and

education, institutional quality also appear to vary systematically across countries. That is, corruption

tends to be higher in low-income countries, the rule of law less well protected and government efficiency

lower.xxxiii

To partly control for these differences, Figure 38 shows value-added per worker plotted against GDP per

capita. Although the fit is far from perfect, value-added per worker tends to be lower in poor countries.

Taking this into account, however, value-added per worker does not appear to be consistently lower in

Sub-Saharan Africa than in other regions. In fact, more countries lie above the regression line than below

the regression line. This indicates that, all else equal, value added per worker is higher in these countries

than would be expected given their relative income levels. Possible reasons for high measured labor

productivity are discussed below.

Figure 38: Value added per worker for firms in Africa and other regions

198


Note: See Table 33 for additional notes on data construction. East Asia is China, Indonesia, Philippines, Thailand and


firms. Countries with GDP over $8,000 are excluded for presentational purposes. Fitted values is line from log-log

regression

It is also interesting to consider some successful exporters of manufactured goods from East Asia—

China, Indonesia, the Philippines, Thailand and Vietnam.xxxiv

Per capita GDP and value-added per

worker is in higher in all of these countries than in most countries in Sub-Saharan Africa. Interestingly,

these countries do not appear to consistently lie above or below the line. China lies significantly above

the line—value-added per worker is higher than would be expected given productivity levels. Indonesia

lies significantly below the line—value added per worker is lower than would be expected given per

capital income. The other three countries lie very close to the fitted line—labor productivity is about at

the level that would be expected given their income levels.

Labor Costs in Sub-Saharan Africa

Although labor costs are lower in Sub-Saharan Africa than in most other regions, this could at least in part

reflect things that affect labor productivity either directly or indirectly. One important factor that is likely

to affect wages and labor productivity is the quality and quantity of human capital. Better educated

workers will generally be more productive and will therefore command higher wages. Low wages and

low labor productivity might therefore reflect the quality or quantity of education.

But this is not the only thing that might affect wages. To the extent that problems in the investment

climate reduce the marginal productivity of labor, we would expect wages to be lower. So, for example,

if weak infrastructure or institutions reduce the marginal productivity of labor then wages will generally

be lower in countries with poor infrastructure and weak infrastructure than in other countries. In the same

way as in the previous sub-section, it is useful to therefore compare wages taking per capita income into

$0

$2,500

$5,000

$7,500

$10,000

$0 $1,000 $2,000 $3,000 $4,000 $5,000 $6,000 $7,000 $8,000

Val

ue

ad

de

d p

er

wo

rke

r (U

S$)



199

account. As discussed above, although per capita income is not a perfect proxy for the missing variables,

income is highly correlated with the quality of the business environment.

Figure 39: Labor costs per worker for firms in Africa and other regions


Note: See Table 33 for additional notes on data construction. East Asia is China, Indonesia, Philippines, Thailand, and


firms. Countries with GDP over $8,000 are excluded for presentational purposes. Fitted values is from log-log

regression

Figure 38 shows labor costs per worker plotted against per capita GDP. Not unsurprisingly, the cost of

labor, like labor productivity, is generally higher in countries where income is higher. This could reflect

higher levels of human capital, better quality institutions, or better quality public infrastructure.

For the most part, as with labor productivity, labor costs appear to be relatively high in most countries in

Sub-Saharan Africa. Of the 31 countries in Africa with available data, labor costs were higher than would

be predicted based upon per capita income alone in 19 countries. For many of the remaining countries,

labor costs were relatively close to predicted values. This provides some evidence that labor costs are

relatively high for formal firms in the manufacturing sector—at least relative to per capita income.

Although labor cost per worker gives some indication of labor costs, differences in labor costs can reflect

differences in things such as worker education and worker skills. That is, labor costs might be low

because the cost of labor is low or might be low because workers are poorly educated or unskilled and,

hence, are less productive. Because wages and productivity are both relatively high relative to other

countries at similar income levels, firms could potentially remain competitive despite high labor costs.

$0

$1,000

$2,000

$3,000

$4,000

$5,000

$0 $1,000 $2,000 $3,000 $4,000 $5,000 $6,000 $7,000 $8,000

Lab

or

cost

s p

er

wo

rke

r (U

S$)



200

One concern is that wage measures in the previous analysis come primarily from enterprises‘ income and

balance sheets. If workers and production are kept off the books, these measures might not give an

accurate measure of firm performance or wages. Although managers might also provide inaccurate data

on the second question as well, it is not clear that the bias will be the same.

Another concern is that high wages might not reflect high wages among production workers in general

but rather might reflect high wages among highly skilled workers, line supervisors and managers. If there

are shortages of skilled workers in Africa relative to other regions, high labor costs might reflect high

wages for these workers rather than high wages among production workers.

For these reasons, it is useful to look at a second measure of wages. In a separate question on the

Enterprise Survey, firm managers are asked about the wages that they pay production workers. This

provides a useful robustness check for the previous results because the measure does not come directly

from the firm‘s accounts.

In practice, the results are similar when we focus on this measure of labor costs rather than on labor costs

from the firms‘ balance sheets (see Figure 40). In 18 of the 26 countries in Sub-Saharan Africa where this

data was available, monthly wages were higher than would be expected given income levels in these

countries. In comparison, among the successful exporters from East Asia with available data, monthly

wages were lower than would be expected given their income levels in all three countries with available

data.

This suggests that the high labor costs observed in Africa relative to income are not due to high wages

among a small set of highly skilled workers and managers. Rather wages for production workers also

appear to be higher than would be expected given income levels in most countries in the region.

Figure 40: Ave monthly wages for production workers for firms in Africa and other

regions

201


Note: See Table 33 for additional notes on data construction. East Asia is China, Indonesia, and Thailand. Africa is

Sub-Saharan Africa only. Data are for all Enterprise Surveys conducted since 2006 with at least 50 firms. Countries

with GDP over $8,000 are excluded for presentational purposes. Fitted values is line from log-log regression

Unit Labor Costs in Sub-Saharan Africa

Although productivity and wage costs appear low in Sub-Saharan Africa, the analysis above suggests that

relative to what would be expected given per capita income in these countries, both appear to be relatively

high. That is, given observed levels of per capita income, labor productivity and labor costs are higher in

most countries than would be expected.

In practice, firms can remain competitive in international markets if labor costs are high when

productivity is high. That is, if labor costs are high because workers are highly productive (e.g., because

they are highly skilled or highly educated), then firms can remain competitive while paying workers high

wages. The ratio of value added to labor costs – which is similar to unit labor costs – gives some idea

whether this is the case. Although, as discussed above, it is not a perfect measure of competitiveness – it

does not take capital use into account, it is a better measure of competitiveness than labor costs alone.

Figure 41 shows unit labor costs for firms in Africa and other regions. Unlike labor productivity and

labor costs, there is not a strong relationship between income and unit labor costs. This suggests that for

the most part higher wages reflect higher levels of productivity – unit labor costs are not consistently

higher or lower among higher income countries.

Unit labor costs in Africa seem to be a little higher than would be predicted based on per capita income

alone. Of the 31 countries in Sub-Saharan Africa with available data, 18 have higher unit labor costs than

would be predicted with based upon per capita income alone. In contrast with the notable exception of

Indonesia, which has relatively high unit labor costs, the East Asian countries that have been relatively

$0

$100

$200

$300

$400

$500

$0 $1,000 $2,000 $3,000 $4,000 $5,000 $6,000 $7,000 $8,000

Ave

. Mo

nth

ly W

age

Pro

du

ctio

n w

ork

ers

(U

S$)



202

successful in manufacturing have unit labor costs that are lower than would be predicted based upon their

income alone.

Figure 41: Unit Labor costs for firms in Africa and other regions


Note: See Table 33 for additional notes on data construction. East Asia is China, Indonesia, the Philippines, Thailand

and Vietnam. Africa is Sub-Saharan Africa only. Data are for all Enterprise Surveys conducted since 2006 with at

least 50 firms. Countries with GDP over $8,000 are excluded for presentational purposes. Fitted values is line from

log-log regression

In summary, at least relative to countries in East Asia that have been relatively successful in

manufacturing, unit labor costs appear to be relatively high in Sub-Saharan Africa. That is, although both

labor cost and labor productivity are relatively high in many countries in the region, labor costs are still

relatively high compared to labor productivity.

I.4 Econometric Analysis

Although the analysis above provides a good starting point, it is useful to do a more formal analysis of the

data. This will allow us to break down the difference measures of performance across regions and to see

whether the differences between countries in Sub-Saharan Africa and other regions are statistically

significant.

Model

To see whether the differences between firms in Sub-Saharan Africa and firms in East Asia and the

Pacific and other regions are statistically significant after controlling for difference in income, we

estimate models of the following form:

0%

20%

40%

60%

80%

100%

$0 $1,000 $2,000 $3,000 $4,000 $5,000 $6,000 $7,000 $8,000

Un

it la

bo

r co

sts

(US$

)



203

The regression regresses various median performance measures in country j on per capital income in

country j and a vector of region dummies. The performance measures are the measures described above

including labor productivity, per worker labor costs, average wages for production workers and unit labor

costs.

The region dummy for Africa is omitted from the regressions and so the coefficients, γ, represent the

average difference in productivity or the other performance measures between the median firms in

countries in Sub-Saharan Africa and the median firms in countries in other regions. The coefficients on

the regional dummy variables can be transformed in percentages using the following formulaxxxv

(1)

Empirical Results

Table 33 shows the results from the base regression.

Per Capita Income. Consistent with earlier graphical analysis, value-added per worker, labor costs per

worker, monthly wages for production workers increase as per capita income increases. For the first two

variables, the point estimates of the coefficients are very close to 1. This suggests that labor productivity

and labor costs increase at about the same rate as per capita income. That is a 1 percent increase in per

capita income is correlated with a 0.94 percent increase in labor productivity and a 0.91 percent increase

in per worker labor costs. The coefficient on monthly wages for production workers is smaller (0.67)

suggesting that a 1 percent increase in per capital income is associated with a 0.67 percent increase in

monthly wages.

The coefficient on per capita income is statistically insignificant and small in the regression for unit labor

costs. This suggests that unit labor costs are not consistently lower or higher countries with higher per

capital income. This is also consistent with the graphical analysis, which suggested no relationship

between unit labor costs and per capita income.

In the final two regressions for the two measures of capital intensity using the book value and the

estimated sales value of the equipment in its current condition, the coefficients are also positive and

statistically significant. A 1 percent increase in per capita income is associated with a 0.79 percent

increase in capital intensity using the book value of capital and a 0.64 percent increase in capital intensity

using the sales value. This suggests that is higher in countries with higher per capita income but that

capital intensity increases more slowly than per capita income

Regional Dummies. As noted above, the omitted regional dummy is for Sub-Saharan Africa. The

percentages calculated using equation (1) above can therefore be interpreted as the average difference

between countries in that region and countries in Sub-Saharan Africa. For the most part, the coefficients

in the first three regressions are negative and, in many cases, are statistically significant. This suggests

that after taking per capita income difference into account labor productivity, labor costs, and monthly

wages for production workers are higher in Sub-Saharan Africa on average than in most other regions.

The coefficients are consistently statistically significant for the dummy variables for both sets of countries

(manufacturing intensive and others) in East Asia and the Pacific and countries in Europe and Central

Asia. In contrast, the coefficients are mostly statistically insignificant for the dummy variables indicating

that the country is in Latin America or South Asia.

204

The coefficients on the dummy variable for manufacturing intensive countries in East Asia indicate that

after income differences are taken into account labor productivity is about 50 percent lower in these

countries than in Sub-Saharan Africa, that labor costs are about 56 percent lower and that wages for

production workers are about 42 percent lower. This is broadly consistent with the graphical analysis

above.

In contrast, the coefficients on most of the dummy variables are statistically insignificant in the

regressions for unit labor costs. This suggest that unit labor costs are similar in Sub-Saharan Africa to

similar costs in other regions. The one exception is the coefficient on the dummy variable indicating that

the country in one of the manufacturing intensive countries in East Asia. For these countries, the

coefficient is positive and negative. The coefficient suggests that unit labor costs are about 20 percent

lower in these countries on average than in countries in Sub-Saharan Africa.

Finally, the coefficients on the dummy variables are mostly statistically insignificant in the regressions for

capital intensity. This suggests that after taking income differences into accounts, firms in Sub-Saharan

Africa are on average about as capital intensive as firms in other regions. The coefficient on the dummy

variable for manufacturing intensive countries in East Asia is negative and statistically significant in the

regression where capital is measured as the manager‘s estimate of the sales value but it positive and

statistically insignificant when measured as book value. This could reflect that, as discussed above,

capital is poorly measured, possibly making it difficult to find statistically significant results.

Robustness Checks

Omitting Per Capita GDP. It is important to note that high productivity and high labor costs in Africa are

relative to other countries at the same level of development. Countries in Africa are also, on average,

poorer than countries in other region. As discussed above, before controlling for income, productivity

and wages appear relatively low in Sub-Saharan Africa (see Table 32). This can also be seen by

excluding per capita income from the previous regressions. After per capita income is excluded, the

coefficients on most of the dummies become positive in the regressions for value-added per worker. The

coefficients on the dummies for Europe and Central Asia and Latin America, in particular, become

positive and statistically significant indicating that wages and productivity are higher in these regions on

average than they are in Africa. The coefficients o the dummies on the East Asia and Pacific exporters

are also positive, but are statistically insignificant.

In the regression for unit labor costs, the results are similar to the results when per capita GDP is

included. For the most part, unit labor costs do not appear to be excessively high on average in Sub-

Saharan Africa. The only region with lower unit labor costs in the manufacturing economies in East Asia.

The coefficient on the dummy variable for this region, however, becomes smaller in absolute value and its

statistical significance falls (to remain statistically significant only at an 11 percent significance level).

Pooling all countries in East Asia and Pacific. As a robustness check, we re-run the regressions pooling

all of the countries in East Asia and the Pacific into a single group. The results are similar except that the

coefficient on the dummy variable becomes statistically insignificant and slightly smaller in the second

regression (see Table 35).

Non-linear effect of per capita GDP. As a final robustness check, we include a squared term for per

capita income in the regression. This allows for a non-linear relationship between income and the

205

dependent variables. The coefficients on the dummy variables are mostly unaffected by this change. In

particular, the coefficients on labor costs, labor productivity and unit labor costs remain statistically

significant and negative in these regressions.

In summary, the results from the econometric analysis confirm many of the previous results. Most

notably, firms in Sub-Saharan Africa appear to be both relatively productive and to have relatively high

labor costs compared to firms in other regions after taking into account the lower income in the region.

The differences are largest and most statistically significant when comparing firms in Sub-Saharan Africa

with firms in East Asia and Europe and Central Asia.

For the most part, unit labor costs are no different on average in Sub-Saharan Africa than in other regions.

Unit labor costs are, however, significantly higher than in successful manufacturing intensive economies

of East Asia (China, Indonesia, Malaysia, Philippines, Thailand, and Vietnam). The point estimate

suggests that on average unit labor costs are about 20 percent lower on average in these countries than in

Sub-Saharan Africa. In this sense, it will be more difficult for firms in Sub-Saharan Africa to compete

with firms from these regions.

Sector-based Analysis

One concern about the previous results is that they do not control for sectoral differences in productivity

and wages. That is, the medians are calculated across all manufacturing firms. It is possible that the high

wages and high levels of productivity in Sub-Saharan Africa could be due to firms operating in high

productivity and high wage sectors. To control for this, as a robustness check, we perform an enterprise-

level analysis that regresses the dependent variables on the previous variables and a series of sector

dummies. Although controlling for sector is useful, it is important to note that if the high wages and

productivity were the result of sectoral differences, this would leave the question of why firms in Africa

tend to be in relatively high wage-high productivity sectors unanswered.

The results from the firm-level regressions are shown in Table 43, which contain a single region dummy

for East Asia and the Pacific, and Table 44, which contains separate dummies for East Asia and Pacific

for manufacturing and non-manufacturing intensive countries.xxxvi

The results are broadly similar to

pervious results. Value added per worker is higher on average among low-income countries in Sub-

Saharan Africa than in other regions. The differences are statistically significant in several cases.

Similar, but more highly significant results, are also visible for labor costs per worker. Per worker labor

costs are higher on average in low-income countries in Sub-Saharan Africa than in all other regions

except Latin America. The coefficient on the Latin America dummy is negative (i.e., suggesting that

wages are higher in Africa) but not statistically significant. Results are similar for monthly wages for

production workers—although the difference are less highly significant and the coefficients are positive.

The only significant coefficients are for the East Asia dummies—production wages are significantly

higher on average in Sub-Saharan Africa than in East Asia.

Unit labor costs are similar, however, in Sub-Saharan Africa to unit labor costs in other regions except

East Asia. In East Asia, unit labor costs are significantly lower than in Sub-Saharan Africa. This is true

for both manufacturing and non-manufacturing economies

206

In summary, the results after controlling for sectoral difference appear consistent with the country-level

results. Wages and productivity appear higher than in other regions in Sub-Saharan Africa after

controlling for per capita income and sectoral differences. These differences are, however, not always

statistically significant. Given that the differences were generally significant before controlling for sector,

this suggests that some of the differences between Africa and other regions might reflect differences in

sectoral composition. Wage costs and unit labor costs do, however, appear to be higher than in East Asia.

I.5 Why do firms in Sub-Saharan Africa appear so productive?

Before discussing wages in Sub-Saharan Africa, it is useful to consider possible bias related to

productivity data. In particular, it is somewhat surprising that firms in many countries in Sub-Saharan

appear to be more productive than firms in other low-income countries at similar stages of development.

As noted above, firms in Sub-Saharan Africa appear to be significantly more productive than similar

firms in other regions after accounting for differences in per capita income. Before accounting for

differences in per capita GDP, firms in Sub-Saharan Africa are less prod

This section discusses several possible reasons for this including: (i) the effect of taxes on profits; the role

of indirect costs; and low levels of competition in the region.

High Tax Rates

The standard measures of productivity used in this study do not take taxation into account. This could be

a problem and might make comparisons across countries and region difficult. When taxes are high—

especially for taxes that are not directly tied to profits—they can affect incentives to invest, open new

businesses, and expand operations.xxxvii

That is, firms will have to be relatively productive in order to

expand and even keep operating when taxes are high.

Although it is possible to calculate many performance measures that take taxes into account (e.g., after-

tax rates of return), it is not possible to calculate similar measures using data from the Enterprise Surveys

– which do not collect any information on taxes. It is therefore necessary to use evidence from other

sources.

One particularly useful measure of the burden of taxation is the total tax rate, which is calculated in the

World Bank‘s Doing Business report (2010b). The measure is better than narrower measures such as

statutory corporate tax rate because it takes into account features of the tax system that can have a

significant effect on the burden of taxation and includes taxes other than corporate income taxes.

Taxes are relatively high in Sub-Saharan Africa (see Table 37). The average tax rate on profits is 23

percent. In comparison, the average profit tax is about 10 percent in Europe and Central Asia, 12 percent

in the Middle East and Africa and 18 percent in East Asia and the Pacific.

In the classification used in the Doing Business report, the most important component in the category

‗profit tax‘ is the corporate income tax. Firms are, however, affected by other taxes including taxes on

labor, inputs, and other things. Although labor taxes are comparable in Sub-Saharan Africa to other

regions—and lower than in several regions including Eastern Europe and Central Asia and OECD

economies—other taxes are considerably higher. As a result, the total tax rate is far higher in Sub-

Saharan Africa than in other regions. The total tax rate is equal to 68 percent of profits in Sub-Saharan

207

Africa—almost twice as high as in East Asia and the Pacific and 20 percentage points higher than in the

region with the next highest total tax rate (Latin America).

In this respect, the high levels of productivity observed in Africa might be misleading. High tax rates will

make it difficult for even productive firms to remain competitive. That is, even if their before-tax profits

are relatively high, taxes will erode these profits. This is particularly true for things such as labor and

other taxes that are not paid on profits.

High Indirect Costs

For the most part, studies looking at firm productivity, such as Investment Climate Assessments, measure

firm performance by looking at revenue and subtracting the cost of intermediate inputs, raw materials and

energy and fuel. As Eifert and others (2008) point out these measures of productivity miss many costs

that affect firm profitability. For example, they note that this omits the cost of transportation,

communications, security, and many other things. Using data from 17 Enterprise Surveys from the early

part of the decade, they show that many of these costs are far higher in Sub-Saharan Africa than in other

regions. After taking some of these additional costs into account, they show that African firms appear far

less productive compared to firms in other region than when they look only at conventional measures of

productivity.

Because the Enterprise Surveys have changed significantly with respect to the information that is

collected on indirect costs, it is not possible to make similar calculations using data from the recent

surveys. However, the evidence that exists from the surveys that have now been completed in close to

100 countries since 2006 are consistent with this idea. Figure 42 shows the cost of bribes, power outages,

losses during transportation, security, and losses due to crime and theft.xxxviii

These costs appear far

higher in Sub-Saharan Africa than in most other regions. In total, the total cost of these factors are equal

to about 9 percent of sales (see Table 38). In comparison, the total cost of these factors was only 2

percent of sales for exporting countries in East Asia with available data (Indonesia, Vietnam, and the

Philippines). They are also far lower in Europe and Central Asia, upper middle-income countries in Sub-

Saharan Africa, and Latin America and the Caribbean. Although costs are higher in South Asia, the

countries in this region with available data (Afghanistan, Bangladesh, Bhutan, and Nepal) might not be

representative of other countries in the region.

Figure 42: Indirect costs as % of sales, by region

208


Note: Regional averages are unweighted averages across countries in that region with available data. Costs are only

for manufacturing firms.

Two factors that appear to be especially cost are corruption (1.6 percent of sales on average) and power

outages (4.8 percent of sales). These are far higher than in any region other than South Asia. Evidence

from outside the Enterprise Surveys also suggest that indirect costs are very high in the Sub-Saharan

Africa. As discussed below, other transportation costs (i.e., other than the costs associated with breakage

and theft during transportation) are also very high. Similarly, data from the International

Telecommunications Union suggests that broadband and telecom costs are also very high (Eifert and

others, 2008). These costs would make it very difficult for firms from Sub-Saharan Africa to compete

with firms from other regions.

Low levels of Competition

A final potential reason for the high measured productivity of firms in Sub-Saharan Africa is the low level

of competition. Competition can affect productivity in at least two ways. The first is that high levels of

competition are likely to result in higher productivity. Firms that are unable to raise their productivity to

the levels of the market leaders will find themselves unable to compete and, as entry drives prices down,

will eventually find themselves unable to stay in the market.

On the other hand, low levels of competition can make measured productivity appear artificially high. In

theory, we would like to have physical measures of output that control for quality differences. If we had

this, then labor productivity would be a good measure of actual productivity—more productive firms

would be producing more output with fewer workers. In practice, this is not what is used to measure

productivity in most empirical studies. Instead productivity is measured in monetary terms—revenue

from sales minus the cost of intermediate inputs and raw materials.xxxix

With firms producing

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

South Asia Africa - Low and Low Mid.

Inc

East Asia and Pacific - Non-

Exp.

Latin America and

Carribbean

Europe and Central Asia

Africa - Upper Mid inc.

East Asia and Pacific - Exp.

Ind

ire

ct c

ost

s (a

s %

of

Sale

s)

Bribes Power Outages Transportation Security Crime

209

heterogeneous products, this can be a problem if some have market power. That is, firms with market

power that charge high prices for their output (e.g., monopolists) would appear more productive than a

similar firm in competitive markets that have to charge lower prices even if their physical output were the

same.xl

Although it is difficult to measure the level of competition in an economy, there are several reasons to

believe that competition is limited in most countries. The first reason is that there are relatively few

modern manufacturing firms in most low-income countries in Sub-Saharan Africa. In Zambia, for

example, only about 150 manufacturing firms in the main population centers had more than 50 employees

at the time of the most recent Enterprise Survey.xli

Although most countries have large and vibrant

informal microenterprise sectors, these firms often do not compete directly with large manufacturing

firms in many industries.

There are several reasons for the small size of the formal manufacturing sector in many countries in the

region. First, as discussed above, taxes are relatively high in many countries in the region (see Table 37).

High taxes can discourage informal firms, who often fall below the tax authorities‘ radars, from

expanding and becoming formal. Only about one in fifty microenterprises in Zambia reported that they

were registered to pay taxes with the Zambia Revenue Authority (Clarke and others, 2010b). In rural

areas, less than one in one hundred reported the same.

A second reason is that the cost of registering a business is high in terms of money and the time required

to complete required procedures in Africa (see Table 39). On average, it takes 45 days and a cost equal to

95 percent of per capita income to start a formal limited liability company in Sub-Saharan Africa. In

comparisons, it takes only 16 days and costs an amount equal to 8.5 percent of per capita income to start a

business in the average country in Eastern Europe and Central Asia. Studies have found that the formal

sector tends to smaller in countries where it takes a long time to start a business.xlii

Even in small countries, competition from imported goods can be very important even when competition

from domestic firms is limited. Most countries in the region took steps towards liberalizing trade by

reducing tariffs and quotas during the 1990s and 2000s. According to data from the World Bank, the

average weighted tariff in the manufacturing sector fell from about 15-18 percent at the beginning of the

1990s to about 8 percent by 2008 (World Bank, 2010c).

Although formal barriers to trade have fallen over time, other barriers remain. Most notably the cost of

exporting and importing manufactured goods is very high in Africa. According to data from the World

Bank‘s Doing Business report it takes an average of about 32 days to complete all procedures to export a

standard manufactured product from Sub-Saharan Africa and an average of about 38 days to import a

standard manufactured product. This is as least as long as in any other region. It is also expensive. It

costs an average of about $2,000 to export a standard container from Sub-Saharan Africa to the country‘s

largest overseas trading partner and about $2,500 to import a standard container from that country. This

is far higher than in any other region—it costs less than $1,000 on average in East Asia for both exporting

and importing.

Several studies have noted that this discourages firms in Africa from exporting.xliii

In the same way, high

costs are likely to discourage imports of manufactured goods. This could reduce competition

210

significantly and, in so doing, increase unit prices. Higher unit prices, in turn, could result in higher

productivity as measured in monetary terms.

The aggregate impact of competition on measured productivity will depend upon which of the

mechanisms is strong. If lower levels of competition reduce productivity by allowing inefficient firms to

stay in the market more than they increase measured productivity by raising prices, then measured

productivity will be lower in countries with less competition. If the reverse is true, the measured

productivity will be higher in countries with less competition.

Some evidence is consistent with the idea that low levels of competition might increase measured

productivity that this could be the case in Sub-Saharan Africa. In particular, when we add a variable

representing the cost of importing to the simple productivity regressions in the previous section, the

coefficient is negative but statistically significant for the whole sample of firms (see Table 41). But when

the sample is restricted to countries in Sub-Saharan Africa, labor productivity is higher in countries with

high import costs. In part this seems to be passed on to workers in the form of higher wages--workers at

formal manufacturing firms in countries with high import costs appear to be paid more than workers at

similar firms in countries with low import costs (see Table 42).

I.6 Why are wages in the formal sector manufacturing so high?

Although high unit labor costs—at least relative to successful exporters of manufactured goods in East

Asia—might explain firms‘ difficulties in entering export markets, this seems to simply move the

explanation down another level rather than explain it. That is, the question becomes why are wages in the

formal sector relatively high in Africa? And why are they not driven downwards making firms in the

region competitive in these industries?

Labor costs and exporters

As a first exercise, we compare the wages paid by exporters and non-exporters in Africa. As discussed

above, exporters are more efficient than non-exporters in most countries in the World. Evidence from the

Enterprise Surveys and earlier studies confirm that this is true in most countries in Sub-Saharan Africa.xliv

If, as discussed above, the differences in firm performance mostly reflects high tax rates and high indirect

costs, then we would probably expect to see exporters being highly productive. That is, they face the

same challenges as non-exporters in terms of taxes and indirect costs. To be productive enough to

overcome these barriers to compete on international markets would require that they are highly

productive.

It might, however, seem inconsistent with the previous discussion of the impact of competition on wages

in Sub-Saharan Africa. That is, since exporters compete on international markets, it might seem that their

productivity and the wages that they charge should be lower than non-exporters if high measured

productivity and high wages were primarily due to low levels of competition.

In practice, exporters appear to have higher labor costs than non-exporters in most low-income countries

in Sub-Saharan Africa. In countries with at least 10 exporting firms, exporters had lower per worker

labor costs in only two countries—Lesotho and Swaziland—were wages higher for non-exporters than for

exporters. In most countries the gap is relatively large—in six countries, wages are more than twice as

high for the median exporter. What accounts for this difference and what does it suggest with respect to

wages and productivity in Africa?

211

One possibility is that wages are higher on average for exporters than for non-exporters because exporters

hire more highly skilled or better educated workers. If this is the case, then the high wages and high

productivity could be explained by high levels of human capital among exporters.

Several detailed analyses have looked at whether this is the case—are wages high among exporters

because they hire highly skilled or highly educated. One recent study that looked at this question in detail

was Chapter 5 in the recent investment climate assessment for Nigeria (Iarossi and Clarke, 2011). A first

cut of the data—looking at the difference between exporters and non-exporters—shows a large gap

between them before controlling for other things that might affect wages (see Table 47). Production

workers make about $175 per month at the median exporter and non-production workers make about

$255 per month. In comparison, production workers make about $77 per month at the median non-

exporter and non-production workers make $125 per month.

The raw data suggests that production workers in exporting firms make about 127 percent more and non-

production workers make about 104 percent more than non-exporters. A detailed firm-level analysis that

controls for differences in firms in terms of size, ownership (foreign versus domestic), firm age, skill ratio

for production workers, unionization, presence of training programs and various other differences

suggests that there difference explain a part of the difference in wages. After controlling for these things,

they find that production workers earn about 19 percent more than non-exporters and non-production

workers make about 16 percent more in exporting firms. The difference between exporters and non-

exporters remains statistically significant. Moreover, the difference for managers is even larger –

managers of exporting firms earn 46 percent more than managers of non-exporting firms.

In addition to analyzing differences at the firm-level, Chapter 5 in Iarossi and Clarke (2011) also matches

individual worker‘s wages with characteristics of the worker and characteristics of the firm. After doing

this, the difference between exporters and non-exporters becomes small and statistically insignificant. In

particular, the analysis controls for enterprise level characteristics (e.g., size, foreign ownership, and firm

age) and individual level characteristics (education, experience, years at firm, gender, union status, and

training received). Although detailed breakdowns by type of worker are not available, after controlling

for all of these differences, the difference between exporters and non-exporters becomes statistically

insignificant.

These results suggest that at least in part, the differences between exporters and non-exporters reflect

differences in other firm-level characteristics. It is, however, important to note that most exporters do not

export all, or even most, of their output. And, as discussed above, when they do export, they most export

to neighboring countries (see Table 31). It would be useful to look at those firms that export all of their

output and compare them to purely domestic firms. Unfortunately, there are very few ‗pure‘ exporters

who only sell in export markets. In fact, even though the samples are often in the hundreds (see Table

50), there are only four countries with over 10 ‗pure‘ exporters—Kenya, Lesotho, Swaziland, and

Madagascar. Interestingly, in all of these countries, median wages are similar or lower for the pure

exporters than they are for purely domestic firms.

212

Labor costs and informality

Any explanation for high wages in the formal sector needs to take into account another interesting aspect

of labor markets in the region—the coexistence of the formal sector with a large informal sector.xlv

In

Zambia, for example, about 84 percent of workers work in the informal sector.xlvi

Moreover, in many countries in the region wages in the informal sector are far lower than in the formal

sector. Workers in large firms and the public sector earn in Ghana and Tanzania earn over twice as much

as similar workers in small firms and self-employed persons in both countries (Sandefur and others,

2010). A similar pattern can be observed in Zambia. Sales per worker and labor costs per worker are

both very low among MSMEs in Zambia. The median unregistered MSME – and, as noted above, most

MSMEs are unregistered – has monthly labour costs of less than US$30 per worker.xlvii

The median

registered MSME has monthly labor costs of about US$70 per worker. Based on data from the World

Bank‘s Enterprise Surveys, which covers manufacturing enterprises with more than five employees, the

median firm reported that monthly labor costs were about US$120 per worker.

At least, in part, the difference in wages between the formal and informal sector is due to difference

between formal and informal enterprises in terms of location or sector. Microenterprises in urban areas

are more likely to be registered than similar enterprises in rural areas. For example, about 20 percent of

urban MSMEs are registered with at least one government entity in Zambia compared to only about 6

percent of rural enterprises. Similarly, MSMEs in the manufacturing sector are far more likely to be

registered than MSMEs in the agricultural sector (12 percent compared to 4 percent).xlviii

Wages for the

most part are particularly low in areas and sectors where the level of informality is high. Whereas

monthly per worker labor costs are about $57 per month for urban MSMEs, they are only $19 per month

for rural MSMEs (see Table 49).

That said, a large gap remains large formal enterprises in Zambia, registered MSMEs in urban areas, and

unregistered MSMEs in urban areas. As noted above, average monthly per worker labor costs for the

large, formal firms in the Zambia Enterprise Survey—all located in urban areas—was about $120 per

month. For registered MSMEs in urban areas, the average was $95 per month. And for unregistered

MSMEs in urban areas, the average was $43 per month. Similarly a large difference between informal

and formal retail MSMEs is visible

Table 29: Monthly labor costs for registered and unregistered MSMEs in Zambia

213

Note: See Table 48 for notes

One plausible explanation for the remaining difference in wages between the small formal sector and a

large informal sector with low wages is that labor market rigidities might make the informal sector the

employer of last resort. Although some countries such as Uganda and Rwanda have relatively flexible

labor markets (World Bank, 2009a), this is not the case in many countries in the region (see Table 45).

Inflexible labor markets could potentially lead to rationing of high paying formal sector jobs, with

unemployed workers forced into the informal sector as they wait for formal jobs to open up. Labor

market rigidities might also make it difficult for formal firms to layoff underperforming workers,

lowering productivity.

Another possible explanation is that problems with basic education could have led to a skills mismatch,

where workers with adequate education and skills are in short supply despite the large pool of unskilled

workers in the informal economy.xlix

That is, the poorly paid workers in the informal sector might not

have the required skills and education to compete in the modern sector. Especially when combined with

problems in the investment climate that drive productivity downwards and other costs associated with

exporting upwards, this might mean that formal firms are unable to compete with exporters from regions

such as East Asia.l

Labor market rigidities and shortages of skilled labor are not, however, the only possible reasons why the

large informal sector fails to drive wages in the formal sector downwards. The failure could also be

because employment in the informal sector makes employment in the formal sector relatively

unattractive. That is, the presence of a large informal sector might be due to reservation wages for formal

jobs being relatively high. Maloney (1999; 2004) notes that this appears to be the case in Latin

America—many self-employed workers report that they would prefer to work in the informal sector

rather than in the formal sector. For example, he notes that almost two-thirds of workers that moved from

$0

$10

$20

$30

$40

$50

$60

$70

$80

$90

$100

Registered Unregistered Retail --

Registered

Retail --

Unregistered

Urban --

Registered

Urban --

Unregistered

Mo

nth

ly la

bo

r co

st (

US$

)Ave. Monthly Labor Cost for registered and unregisted firms in Zambia

214

the formal sector to the informal sector in Mexico reported that they did so by choice—mostly citing a

desire for greater independence or higher pay.

This could also be the case in some countries in Sub-Saharan Africa. In the survey of small business

owners in Zambia discussed above, less than half said that they would take a full-time job in the formal

sector employment if they were offered one. Moreover, among the owners who said they would take a

formal sector job, many would prefer to work for either the government (67 percent), a state-owned

enterprise (4 percent) or an NGO (17 percent). Only 10 percent said they would like to work for a formal

sector private firm.

So why would many people rather work in the informal sector than in the formal sector if wages are

lower? Although wages in the informal sector appear low in Sub-Saharan Africa, Maloney (2004) notes

that it is difficult to compare wages in the formal and informal sectors. Some things might make

employment in the informal sector more attractive than employment in the formal sector despite low

wages. One important factor might be that informal workers can mostly avoid taxes—meaning that

before-tax wage comparisons are difficult. Moreover, some people like to work for themselves and being

in the informal sector allows for more flexibility. In particular, they have more freedom to hire and fire

works. It is also likely that informal enterprises have lower indirect costs than formal enterprises—they

avoid the cost of dealing with license fees and other regulatory costs and might avoid much of the cost of

corruption.li

Even for informal workers who are not the owners, there are some additional benefits to being in the

informal sector. Many—if not most—workers are family members who often receive payments in kind

such as food or lodging.lii In contrast, other things might make formal employment preferable given the

same wage rates. Formal firms often pay benefits that informal firms do not and employment is more

secure. Maloney (2004) argues that these large differences—both positive and negative—make wage

comparisons between the two sectors very difficult.

I.7 Conclusion

Few countries in Africa have successfully managed to diversify their economies into export-oriented

manufacturing. Most countries have small, underdeveloped manufacturing sectors and many relatively

successful countries such as Kenya mostly export to nearby countries rather than to developed economies.

With a large pool of underemployed, low wage workers in the informal sector, this seems puzzling. That

is, it seems that many countries on the continent should be well placed to enter labor-intensive areas of

manufacturing. This note explores several possible reasons for this, focusing on the role of wages in the

formal sector.

Firm productivity is lower in Sub-Saharan Africa than in other regions. Although this would seem to

make it difficult for firms in the region to compete with firms in other regions, wages are also relatively

low. If wages were low enough, it might be possible for firms to remain competitive in international

markets despite being less productive than firms in other countries.

Because of this, it is useful to look at unit labor costs—the ratio of labor costs to value-added. Although

this is not a perfect measure of productivity—it does not, for example, take into account the use of

capital—it is better than measures such as labor costs measured in dollars because it takes into account

the quality of human capital (that is, better educated and more highly skilled workers earn more).

215

Unit labor costs are similar on average in Africa to unit labor costs in other regions. The difference

between unit labor costs in Africa and most other regions is small and statistically insignificant. They are,

however, slightly higher than in countries in East Asia that have been successful in export-oriented

manufacturing. However, the difference, although statistically significant, is not particularly large. In

this respect, it is not clear that high unit labor costs alone can explain Africa‘s poor relative performance

in export-oriented manufacturing to developed economies.

Because of this, the paper looks at several additional issues in more depth. Although, as noted above,

wages and productivity are relatively low in Africa, they are high relative to other countries at similar

levels of development. That is, formal sector wages in low-income countries in the region are higher on

average than in similar low-income countries in other regions and formal sector wages in middle-income

economies are higher on average than is similar middle-income countries in other regions.

This leads to two additional questions. First, why do firms in Africa appear relatively productive

compared to firms in other countries at similar levels of development? Second, why aren‘t wages forced

downwards?

The first question is whether the relatively high levels of productivity are real or illusionary. The note

discusses several possible explanations for the high observed levels of productivity. First, it notes that

taxes on formal firms are relatively high in Sub-Saharan Africa. Since the productivity measures do not

take taxes into account, it is plausible that less productive firms might not be able to remain profitable and

therefore have to shut down in Africa. The result should be a small manufacturing sector with only the

most productive firms surviving.

Second, it notes that other indirect costs are very high in many countries in the region. Although the

productivity measures take into account the cost of labor and intermediate inputs, they do not take into

account the costs imposed by poor infrastructure (e.g., transportation costs and communication costs),

poor governance (the ‗bribe‘ tax), or crime and insecurity. Previous papers have shown that these

additional costs tend to be high in Africa, potentially making relatively productive firms unprofitable.

Third, the paper notes that most firms operate only in domestic markets, which are small and have low

levels of competition. This could mean that labor productivity is overestimated. That is, because

revenues are used to calculate productivity rather than physical measures of output, high productivity

might reflect high prices rather than high levels of physical output. Although tariffs are high,

transportation costs and other barriers to trade mean that firms in these markets are often well protected

against international competitors.

The final issue that the paper discusses is why wages appear high relative to per capita income. The paper

notes that this appears to be because firms in the large informal sector pay far lower wages than firms in

the formal sector—something that remains true after controlling for difference in location or sector. The

paper then discusses several potential reasons for this.

216

I.8 Tables

Table 30: Manufacturing as share of GDP

Country

Average

(2005-2009)

Country

Average

(2005-2009)

Regions (developing countries only)

Sub-Saharan Africa (cntd.)

East Asia & Pacific 31.46 Gambia, The 4.97

Europe & Central Asia 18.26 Ghana 8.74

Latin America & Caribbean 17.39 Guinea 4.26

Middle East & North Africa 11.92 Guinea-Bissau ..

South Asia 16.24 Kenya 10.10

Sub-Saharan Africa (developing only) 13.18 Lesotho 19.74

East Asia (Exporters)

Liberia 13.05

China 33.03 Madagascar 14.90

Indonesia 27.93 Malawi 14.06

Malaysia 27.70 Mali 3.13

Philippines 22.15 Mauritania 5.05

Thailand 34.86 Mauritius 19.84

Vietnam 20.74 Mozambique 14.90

Sub-Saharan Africa

Namibia 14.92

Angola 4.79 Niger ..

Benin 7.51 Nigeria 2.71

Botswana 3.69 Rwanda 5.23

Burkina Faso 14.10 Sao Tome and Principe 6.37

Burundi 8.83 Senegal 13.83

Cameroon 17.00 Seychelles 12.23

Cape Verde 6.73 Sierra Leone ..

217

Central African Republic 7.52 Somalia ..

Chad 5.92 South Africa 16.89

Comoros 4.24 Sudan 6.38

Congo, Dem. Rep. 6.09 Swaziland 43.28

Congo, Rep. 3.92 Tanzania 6.87

Cote d'Ivoire 18.13 Togo 10.14

Equatorial Guinea 11.15 Uganda 7.67

Eritrea 5.90 Zambia 11.39

Ethiopia 4.61 Zimbabwe 13.52

Gabon 4.01

Source: World Bank (2010c).

Note: Averages for regions are weighted. Averages across years and unweighted averages for each country for available years for that country

Table 31: Export destinations for enterprises included in the Investment Climate Surveys

from early 2000s.

Most Important Export Destinations

(% of exporters that report destination is important)

Most important

industrialized export

destination

Ethiopia Italy (55%), United Kingdom (29%), Germany (19%) Italy (55%)

Kenya Uganda (74%), Tanzania (61%), Rwanda (19%) United Kingdom (8%)

Mali Burkina Faso (63%), Guinea (53%), Niger (38%) France (9%)

Senegal Gambia (39%), Mali (36%), Mauritania (31%) France (18%)

Tanzania Kenya (38%), Malawi (14%), Uganda (12%), United Kingdom(12%), Zambia (12%) United Kingdom (12%)

Uganda Rwanda (49%), Congo (33%), Kenya (18%) United Kingdom (16%)

Zambia Congo (38%), Malawi (22%), Germany (21%) Germany (21%)

Source: Investment Climate Surveys.

Note: Enterprises were asked to list their three most important export destinations. Countries are ranked based upon the number of enterprises that

ranked each country among the top three. Not all enterprises reported three destinations. Data were not available for Mozambique.

Table 32: Labor Costs in Africa and other regions

218

Obs.

Labor Cost per

Worker (US$)

Value-Added per

Worker (US$)

Unit Labor

Costs

Mean Median Mean Median Mean Median

Africa 37 $1,464 $887 $4,734 $3,210 33.5% 32.6%

Africa -- Low and lower middle income 32 $1,059 $873 $3,316 $2,462 33.9% 33.9%

Africa -- Upper middle income 5 $4,056 $2,818 $13,811 $14,967 30.7% 28.8%

East Asia 12 $1,733 $1,246 $6,631 $5,192 31.7% 31.9%

East Asia - non manufacturing 6 $1,837 $1,800 $6,713 $4,064 33.4% 34.6%

East Asia - manufacturing 6 $1,629 $1,246 $6,548 $5,684 30.0% 28.1%

Europe and Central Asia 30 $4,046 $2,869 $10,297 $7,741 37.7% 37.5%

Latin America and Caribbean 14 $3,241 $2,795 $8,890 $7,884 36.6% 37.2%

South Asia 2 $817 $817 $1,483 $1,483 39.9% 37.5%

Source: Authors‘ Calculations using data from World Bank Enterprise Surveys

Note: Means and medians are unweighted country-level means and medians for all countries in the region. The country level data are weighted

medians for that country. See Table 50for list of countries in each region

Table 33: Difference in median values of productivity variables by region

219

Value-

Added per

Worker

(2005 US$)

Labor costs

per worker

(2005 US$)

Monthly Wage

Production

Workers

(2005 US$)

Unit

Labor

Costs

Capital per

Worker

(2005 US$,

book value)

Capital per

Worker

(2005 US$,

sales value)

Observations 77 77 71 78 72 56

Per capita GDP (log) 0.943*** 0.910*** 0.667*** 0.039 0.792*** 0.641***

(11.50) (13.65) (8.73) (0.96) (4.79) (5.21)

Region Dummies a

Asia and Pacific Exporters -0.702** -0.840*** -0.545* -0.255* 0.383 -1.239**

(-2.59) (-3.80) (-1.90) (-1.87) (0.73) (-2.62)

Asia and Pacific Other -0.797* -0.708** -1.125** 0.134 0.538 -0.675

(-1.93) (-2.11) (-2.20) (0.65) (0.68) (-0.87)

Europe and Central Asia -0.954*** -0.745*** -0.319* 0.004 0.190 -0.177

(-4.50) (-4.32) (-1.69) (0.04) (0.45) (-0.60)

Latin America and Caribbean -0.493** -0.298 -0.033 -0.006 0.406

(-2.24) (-1.66) (-0.17) (-0.05) (0.94)

South Asia -0.613 0.061 0.263 0.098 0.873 -0.172

(-1.48) (0.18) (0.71) (0.47) (1.11) (-0.31)

Constant 1.188* 0.334 -0.333 -1.358*** 0.338 2.815***

(1.95) (0.67) (-0.58) (-4.45) (0.27) (3.03)

R-squared 0.72 0.79 0.66 0.09 0.45 0.53


Note: ***,**,* means statistically significant at 1%,5% and 10% levels. T-statistics in parentheses. All dependent variables are

weighted median values for enterprises with available data. Value added is calculated by subtracting intermediate inputs and

energy costs from sales from manufacturing. Workers include permanent and temporary full-time workers. Labor cost is the

total cost of wages, salaries, allowances, bonuses and other benefits for both production and non-production workers. Unit labor

costs are labor costs divided by value-added. Capital is the book value and sales value of machinery and equipment (i.e., the

amount the manager thinks his or her machinery and equipment would cost if sold in its current condition).

a Omitted dummy is dummy for Sub-Saharan Africa

Table 34: Difference in median values of productivity variables by region (GDP omitted)

220

Value-Added

per Worker

(2005 US$)

Labor Costs

per Worker

(2005 US$)

Monthly Wage

Production

Workers

(2005 US$)

Unit

Labor

Costs

Capital per

Worker

(2005 US$,

book value)

Capital per

Worker

(2005 US$,

sales value)


Region Dummies a

Asia and Pacific Exporters 0.373 0.198 0.069 -0.210 1.276** -0.841

(0.87) (0.50) (0.15) (-1.64) (2.26) (-1.48)

Asia and Pacific Other -0.629 -0.546 -1.233 0.141 0.667 -0.569

(-0.91) (-0.86) (-1.46) (0.68) (0.74) (-0.60)

Europe and Central Asia 0.709*** 0.880*** 0.565** 0.086 1.601*** 0.888***

(2.76) (3.73) (2.44) (1.12) (4.55) (3.40)

Latin America and Caribbean 0.854*** 1.003*** 0.616** 0.052 1.527***

(2.73) (3.49) (2.20) (0.56) (3.63)

South Asia -0.960 -0.274 -0.287 0.083 0.570 -0.492

(-1.39) (-0.43) (-0.47) (0.40) (0.63) (-0.72)

Constant 8.068*** 6.977*** 4.843*** -1.070*** 6.132*** 7.581***

(43.61) (41.08) (28.17) (-19.25) (23.07) (39.13)

R-squared 0.20 0.25 0.16 0.08 0.27 0.28


Note: ***,**,* means statistically significant at 1%,5% and 10% levels. T-statistics in parentheses.

See Table 33 for additional notes


Table 35: Difference in median values of productivity variables by region (East Asia

combined)

221

Value-Added

per Worker

(2005 US$)

Labor Costs

per Worker

(2005 US$)

Monthly Wage

Production

Workers

(2005 US$)

Unit

Labor

Costs

Capital per

Worker

(2005 US$,

book value)

Capital per

Worker

(2005 US$,

sales value)


Per capita GDP (log) 0.945*** 0.906*** 0.629*** 0.028 0.787*** 0.636***

(11.79) (13.87) (6.24) (0.68) (4.87) (5.21)

Region Dummies a

Asia and Pacific -0.728*** -0.804*** -0.839** -0.147 0.427 -1.095**

(-3.07) (-4.16) (-2.43) (-1.21) (0.93) (-2.64)

Europe and Central Asia -0.959*** -0.738*** -0.474* 0.025 0.199 -0.169

(-4.59) (-4.34) (-1.89) (0.23) (0.48) (-0.58)

Latin America and Caribbean -0.497** -0.292 -0.201 0.011 0.414

(-2.28) (-1.65) (-0.77) (0.10) (0.96)

South Asia -0.612 0.059 0.027 0.094 0.871 -0.175

(-1.49) (0.18) (0.05) (0.44) (1.11) (-0.31)

Constant 1.167* 0.362 0.168 -1.274*** 0.377 2.852***

(1.96) (0.75) (0.22) (-4.19) (0.31) (3.10)

R-squared 0.721 0.794 0.458 0.050 0.453 0.524



See Table 33 for additional notes.


Table 36: Difference in median values of productivity variables by region (squared term

included)

222

Value-Added

per Worker

(2005 US$)

Labor Costs

per Worker

(2005 US$)

Monthly Wage

Production

Workers

(2005 US$)

Unit

Labor

Costs

Capital per

Worker

(2005 US$,

book value)

Capital per

Worker

(2005 US$,

sales value)


Per capita GDP (log) -1.726** -1.618** -1.058 -0.015 -0.267 -0.365

(-2.13) (-2.53) (-1.29) (-0.03) (-0.16) (-0.28)

Per capita GDP Squared (log) 0.164*** 0.156*** 0.105** 0.003 0.065 0.062

(3.31) (3.97) (2.12) (0.13) (0.64) (0.77)

Region Dummies a

Asia and Pacific Exporters -0.492* -0.641*** -0.432 -0.251* 0.464 -1.151**

(-1.88) (-3.10) (-1.52) (-1.77) (0.86) (-2.35)

Asia and Pacific Other -0.595 -0.517* -1.001** 0.138 0.613 -0.603

(-1.52) (-1.67) (-2.00) (0.65) (0.77) (-0.77)

Europe and Central Asia -0.931*** -0.723*** -0.301 0.005 0.198 -0.179

(-4.70) (-4.61) (-1.64) (0.04) (0.46) (-0.60)

Latin America and Caribbean -0.328 -0.141 0.070 -0.002 0.470

(-1.55) (-0.84) (0.36) (-0.02) (1.05)

South Asia -0.573 0.100 0.256 0.099 0.883 -0.172

(-1.48) (0.32) (0.71) (0.47) (1.12) (-0.30)

Constant 11.741*** 10.335*** 6.575* -1.142 4.542 6.819

(3.63) (4.04) (1.99) (-0.65) (0.68) (1.29)

R-squared 0.76 0.83 0.69 0.09 0.46 0.53





Table 37: Tax rates in Sub-Saharan Africa and other regions

223

Profit tax

(%)

Labor tax

(%)

Other taxes

(%)

Total tax rate

(% profit)

East Asia & Pacific 18.3 10.3 6.8 35.4

Eastern Europe & Central Asia 9.8 22.9 8.5 41.2

Latin America & Caribbean 20.9 14.7 12.4 48

Middle East & North Africa 12 16.8 4.1 32.8

OECD 16.8 23.3 3 43

South Asia 17.8 7.8 14.2 39.9

Sub-Saharan Africa 23.1 13.5 31.5 68

Source: World Bank (2010b)

Table 38: Indirect Costs as % of Sales by Region

224

Bribes Power

Outages

Losses

During

Transport

Security Crime Total

Africa - Low and Lower Middle Income 1.6 4.8 0.9 1.0 0.5 8.9

Angola 2.6 2.9 0.5 0.9 0.2

Benin 3.1 6.8

1.2 0.0

Burkina Faso 0.0 2.3 0.1 0.7 0.0

Burundi 3.3 8.1 0.1 0.6 0.5

Cameroon 3.0 4.3 2.0 1.4 0.9

Cape Verde 0.0 2.8

0.7 0.8

Chad 1.4 1.9

0.9 1.6

Congo 2.6 13.2

2.1 0.2

Congo, DR 3.5 6.1 0.3 0.3 0.5

Cote d'Ivoire 3.6 1.8 0.3 0.3 0.2

Eritrea 0.0 0.0

0.3 0.0

Gambia 2.0 8.9 0.9 2.6 1.0

Ghana 1.5 5.3 0.8 0.5 0.2

Guinea 3.6 14.0 0.7 0.2 0.7

Guinea Bissau 4.3 1.3 0.6 0.8 0.5

Kenya 2.2 4.3 1.6 0.7 0.5

Lesotho 0.1 2.1

2.2 1.1

Liberia 0.6 0.5

2.3 1.7

Madagascar 1.8 7.9 1.0 1.1 0.4

Malawi 0.0 7.9

4.7 1.6

Mali 0.5 0.8 0.6 0.3 0.1

Mauritania 3.4 0.9 0.5 0.5 0.0

Mozambique 0.5 0.9 0.9 0.5 1.0

Niger 1.6 1.4

0.5 0.1

225

Nigeria 0.9 8.8 2.4 1.3 0.2

Rwanda 0.9 6.5 1.0 0.6 0.3

Senegal 0.8 6.2 0.6 0.5 0.1

Sierra Leone 0.3 5.4

0.9 0.8

Swaziland 0.7 1.0 0.8 0.8 0.4

Tanzania 2.3 7.7 1.4 1.4 0.6

Togo 0.0 6.3

0.9 0.2

Uganda 2.7 8.9 1.0 0.7 0.3

Zambia 0.2 2.3 0.7 1.0 0.4

Africa - Upper Middle Income 0.2 0.4 0.9 0.8 0.4 2.6

Botswana 0.2 0.3 1.3 1.0 0.5

Gabon 0.5 0.7

0.8 0.2

Mauritius 0.0 0.5 0.2 0.8 0.1

Namibia 0.1 0.0 0.8 0.6 0.6

South Africa 0.1 0.4 1.3 1.0 0.5

East Asia and Pacific - Exporters 0.2 0.7 0.5 0.4 0.1 1.9

Indonesia 0.1 0.1 0.6 0.1 0.0

Philippines 0.3 0.6 0.7 0.7 0.3

Vietnam 0.3 1.3 0.2 0.4 0.0

Table 39: Time and Cost to Start a Business in Africa

226

Procedures

(number)

Time

(days)

Cost

(% of income per capita)

Paid-in Min. Capital

(% of income per capita)

East Asia & Pacific 7.8 39 27.1 50.6


Latin America & Caribbean 9.3 56.7 36.2 4.6

Middle East & North Africa 8.1 20 38 104

OECD 5.6 13.8 5.3 15.3

South Asia 7.1 24.6 24.5 24.1

Sub-Saharan Africa 8.9 45.2 95.4 145.7


Table 40: Cost of importing and exporting

Time to export

(days)

Cost to export

(US$ per

container)

Time to import

(days)

Cost to import

(US$ per

container)



Latin America & Caribbean 18 1228.3 20.1 1487.9

Middle East & North Africa 20.4 1048.9 24.2 1229.3

OECD 10.9 1058.7 11.4 1106.3

South Asia 32.3 1511.6 32.5 1744.5



Table 41: Difference in median values of productivity variables by region (cost of

importing included)

227

Value-Added per

Worker

(2005 US$)

Labor Costs per

Worker

(2005 US$)

Monthly Wage

Production Workers

(2005 US$)

Unit Labor

Costs

Observations 77 77 71 78

Per capita GDP (log) 0.900*** 0.858*** 0.572*** 0.047

(9.82) (11.63) (4.75) (1.02)

Cost of importing (log, US$) -0.175 -0.213 -0.154 0.031

(-1.05) (-1.58) (-0.72) (0.37)

Region Dummies a

Asia and Pacific Exporters -0.861*** -1.034*** -0.836* -0.227

(-2.77) (-4.13) (-1.94) (-1.44)

Asia and Pacific Other -0.865** -0.791** -1.331* 0.146

(-2.07) (-2.35) (-1.94) (0.69)

Europe and Central Asia -0.924*** -0.708*** -0.424 -0.001

(-4.33) (-4.11) (-1.65) (-0.01)

Latin America and Caribbean -0.514** -0.323* -0.202 -0.003

(-2.33) (-1.82) (-0.76) (-0.02)

South Asia -0.572 0.111 0.044 0.091

(-1.38) (0.33) (0.09) (0.43)

Constant 2.838* 2.343* 1.776 -1.650*

(1.68) (1.72) (0.79) (-1.95)

R-squared 0.73 0.80 0.47 0.09





Table 42: Difference in median values of productivity variables for Africa (cost of

importing included)

228

Value-Added per

Worker

(2005 US$)

Labor Costs per

Worker

(2005 US$)

Monthly Wage

Production Workers

(2005 US$)

Unit Labor

Costs

Observations 31 31 26 31

Per capita GDP (log) 0.781*** 0.716*** 0.493*** -0.045

(7.38) (8.63) (3.02) (-0.59)

Cost of importing (log, US$) 0.570** 0.409** 0.612* 0.103

(2.65) (2.42) (1.87) (0.66)

Constant -1.990 -1.390 -3.520 -1.546

(-1.00) (-0.90) (-1.12) (-1.07)

R-squared 0.66 0.73 0.30 0.04




Note: Only includes countries from Sub-Saharan Africa

Table 43: Difference in median values of productivity variables by region (Firm level

regressions)

229

Value-Added per

Worker

(2005 US$)

Labor Costs per

Worker

(2005 US$)

Monthly Wage

Production Workers

(2005 US$)

Unit Labor

Costs

Observations 26340 29112 27431 27497

Sector Dummies Included Included Included Included


(8.02) (7.56) (6.68) (-0.94)

Region Dummies a

Asia and Pacific -0.164 -0.869*** -0.452** -0.758***

(-0.56) (-4.65) (-2.32) (-4.02)

Europe and Central Asia -0.663** -0.585** -0.092 -0.030

(-2.14) (-2.29) (-0.35) (-0.19)

Latin America and Caribbean -0.237 -0.249 0.087 -0.056

(-0.90) (-1.17) (0.39) (-0.40)

South Asia -0.499*** -0.210* 0.021 -0.119

(-3.22) (-1.78) (0.16) (-0.59)

Africa Middle Income 0.012 0.112 0.156 0.154

(0.03) (0.28) (0.55) (0.73)

Constant 1.237 0.820 -0.509 -0.401

(1.63) (1.06) (-0.73) (-0.87)

R-squared 0.26 0.35 0.51 0.15


Note: ***,**,* means statistically significant at 1%,5% and 10% levels. Robust t-statistics clustered at country level in

parentheses. See Table 33 for additional notes on variables. The model also includes sector dummies for textiles, garments, food

and beverage, chemicals, construction materials, wood and furniture, metal products, paper and publishing, plastics, machinery,

electronics, cars, and other manufacturing. Outliers more than 3 standard deviations from the country-level averages are

excluded.


Table 44: Difference in median values of productivity variables by region (Firm level

regressions)

230

Value-Added

per Worker

(2005 US$)

Labor Costs per

Worker

(2005 US$)

Monthly Wage

Production Workers

(2005 US$)

Unit Labor

Costs

Observations 26340 29112 27431 27497

Sector Dummies Included Included Included Included


(7.91) (7.46) (6.54) (-0.74)

Region Dummies a

Asia and Pacific Manufacturing Only -0.142 -0.890*** -0.434** -0.783***

(-0.47) (-4.58) (-2.22) (-4.05)

Asia and Pacific Other -0.547 -0.550* -0.861*** -0.347**

(-1.60) (-1.85) (-6.26) (-2.38)

Europe and Central Asia -0.640** -0.607** -0.075 -0.056

(-2.06) (-2.33) (-0.29) (-0.36)

Latin America and Caribbean -0.216 -0.269 0.102 -0.079

(-0.82) (-1.24) (0.46) (-0.56)

South Asia -0.499*** -0.209* 0.020 -0.125

(-3.26) (-1.75) (0.16) (-0.61)

Africa Middle Income 0.034 0.091 0.172 0.130

(0.10) (0.23) (0.61) (0.61)

Constant 1.318* 0.745 -0.448 -0.492

(1.74) (0.94) (-0.64) (-1.06)

R-squared 0.26 0.35 0.51 0.15


Note: ***,**,* means statistically significant at 1%,5% and 10% levels. Robust t-statistics clustered at country level in

parentheses. See Table 33 and Table 43 for additional notes on variables and sector dummies.


Table 45: Labor regulations in Sub-Saharan Africa and other regions

231

Difficulty of

Hiring Rigidity of Hours

Difficult of

Redundancy

Rigidity of

employment



Latin America & Caribbean 34.4 21.3 24.1 26.6

Middle East & North Africa 21.3 22.1 30 24.5

OECD 26.5 30.1 22.6 26.4

South Asia 27.8 10 41.3 26.3


Source: World Bank (2009a)

Note: Higher values mean more rigid regulations (on 0-100 scale)

Table 46: Labor Costs for Importers and Exporters

232

Non-Exporters Exporters Pure Exporters

Low and lower middle income Africa

Benin $874 $1,758

Burkina Faso $854 $1,286

Cameroon $1,575 $3,921

Congo, DR $794 $1,631

Cote d'Ivoire $636 $2,446

Eritrea $424 $518

Ghana $531 $551

Guinea $460 $588

Kenya $1,733 $2,096 $782

Lesotho $1,077 $441 $294

Madagascar $612 $619 $604

Malawi $748 $774

Mali $804 $1,282

Mauritania $1,577 $1,879

Mozambique $858 $1,559

Nigeria $898 $2,704

Rwanda $1,019 $1,095

Senegal $1,191 $2,256

Swaziland $2,590 $1,986 $2,080

Tanzania $957 $1,107

Togo $564 $2,549

Uganda $828 $1,113

Zambia $1,101 $1,797

Upper middle income Africa

Botswana $2,503 $3,069

233

Mauritius $2,171 $2,543 $2,645

Namibia $3,593 $6,621

South Africa $7,290 $12,161

East Asia and Pacific -- Exporters

China $1,148 $1,466 $1,302

Indonesia $520 $965 $895

Malaysia $3,725 $4,331 $4,104

Thailand $1,405 $1,951 $1,896

Vietnam $1,097 $1,108 $1,014

Source: Authors‘ Calculations using Data from World Bank Enterprise Surveys

Table 47: Differences in wages between Nigerian exporters and non-exporters

Production Non-Production Managers

Wages in 2005 US$

Non-exporter $77 $125 ---

Exporter $175 $255 ---

Percent difference between exporters and non-exporters

No controls 127% 104% ---

After controlling for enterprise-level differences a 19% 16% 46%

After controlling for enterprise and worker level differences -4%

Source: Iarossi and Clarke (2011)

a Skilled production workers only. Difference for unskilled workers is about 14 percent.

Table 48: Employment in Zambia in 2005 by formality status

234

Number % of Employed Persons

Total Employed 4,131,782 100%

Formal Private Sector 277,680 7%

Government 218,104 5%

Informal Sector 3,635,998 88%

Source: Clarke and others (2010b) using data from Ministry of Labor and Social Security (2009).

Table 49: Average monthly labor cost for MSMEs in Zambia, by registration status,

sector, and location

Ave. Monthly Labor Cost

Registered $71

Unregistered $28

Urban $57

Rural $19

Manufacturing $71

Retail $28

Other Services $43

Agriculture $20

Other $36

Retail -- Urban $57

Retail -- Rural $20

Retail -- Registered $85

Retail -- Unregistered $28

Urban -- Registered $95

235

Urban -- Unregistered $43

Source: Author‘s calculations using data from the Zambia Business Survey

Note: Data are for firms with workers that are paid in cash. Workers paid in-kind are excluded and in-kind payments are

excluded. Firms with more that 50 employees are included.

236

Table 50: List of Countries, regions, and number of observations for productivity data

Country Region Obs. Country Region Obs.

Afghanistan SA 73 Lithuania ECA 73

Albania ECA 42 Madagascar AFR 148

Angola AFR 215 Malawi AFR 55

Argentina LAC 510 Malaysia EAP-M 775

Armenia ECA 83 Mauritania AFR 79

Azerbaijan ECA 101 Mauritius AFR-UMI 129

Belarus ECA 62 Mexico LAC 1000

Benin AFR 17 Micronesia EAP 9

Bolivia LAC 258 Moldova ECA 101

Bosnia and Herzegovina ECA 84 Mongolia ECA 128

Botswana AFR-UMI 110 Montenegro ECA 22

Brazil LAC 992 Mozambique AFR 341

Bulgaria ECA 385 Namibia AFR-UMI 102

Burkina Faso AFR 47 Nepal SA 124

Burundi AFR 102 Nicaragua LAC 314

Cambodia EAP 129 Niger AFR 17

Cameroon AFR 89 Nigeria AFR 2008

Cape Verde AFR 45 Panama LAC 158

Chad AFR 21 Paraguay LAC 198

Chile LAC 528 Peru LAC 303

China EAP-M 10697 Philippines EAP-M 689

Colombia LAC 582 Poland ECA 88

Congo AFR 14 Romania ECA 91

Congo, DR AFR 149 Russia ECA 391

Cote d'Ivoire AFR 137 Rwanda AFR 58

Croatia ECA 270 Samoa EAP 17

237

Czech Republic ECA 60 Senegal AFR 259

Ecuador LAC 286 Serbia ECA 126

El Salvador LAC 410 Sierra Leone AFR 43

Eritrea AFR 53 Slovak Republic ECA 54

Estonia ECA 80 Slovenia ECA 85

FYR Macedonia ECA 79 South Africa AFR-UMI 678

Gabon AFR-UMI 24 Swaziland AFR 68

Gambia AFR 33 Tajikistan ECA 91

Georgia ECA 76 Tanzania AFR 271

Ghana AFR 292 Thailand EAP-M 1385

Guatemala LAC 297 Timor Leste EAP 38

Guinea AFR 134 Togo AFR 13

Guinea Bissau AFR 50 Turkey ECA 542

Honduras LAC 233 Uganda AFR 307

Hungary ECA 92 Ukraine ECA 279

Indonesia EAP-M 857 Uruguay LAC 219

Kazakhstan ECA 134 Uzbekistan ECA 118

Kenya AFR 395 Vanuatu EAP 8

Kosovo ECA 84 Venezuela LAC 0

Kyrgyz Republic ECA 73 Vietnam EAP-M 670

Lao PDR EAP 143 Mali AFR 301

Latvia ECA 74 Zambia AFR 304

Liberia AFR 72

Source: Author‘s calculations based upon data from World Bank Enterprise Surveys

Note: Number of observations in table are for labor productivity calculations. Other productivity measures may have slightly

more or slightly fewer observations.

AFR: Africa (low and lower middle income); AFR-UMI: Africa (upper middle income); EAP: East Asia and Pacific (non-

manufacturing); EAP-M: East Asia and Pacific (manufacturing); ECA: Europe and Central Asia; SA: South Asia;

238

I.9 References

Aw, Bee Yan, Sukkyun Chung, and Mark J. Roberts. 2000. "Productivity and Turnover in the Export

Market: Micro-Level Evidence From the Republic of Korea and Taiwan (China)." World Bank

Economic Review 14(1):65–90.

Azfar, Omar, and Peter Murrell. 2009. "Identifying Reticent Respondents: Assessing the Quality of

Survey Data on Corruption and Values." Economic Development and Cultural Change

57(2):387–411.

Bernard, Andrew B., and J. Bradford Jensen. 1999. "Exceptional Exporter Performance: Cause, Effect or

Both?" Journal of International Economics 47(1):1–25.

Bigsten, Arne, Paul Collier, Stefan Dercon, Marcel Fafchamps, Bernard Gauthier, Jan Willem Gunning,

Abena Oduro, Remco Oostedorp, Catherine Pattillo, Mans Soderbom, Francis Teal, and Albert

Zeufeck. 2004. "Do African Manufacturing Firms Learn From Exporting." Journal of

Development Studies 40(3):115–141.

Clarke, George R. G. 2009a. "Beyond Tariffs and Quotas: Why Do African Manufacturers Not Export

More?" Emerging Markets Finance and Trade 45(2):44–64.

--------. 2009b. "Regulation and Corruption in Africa: Results from an MSME Survey in Zambia." World

Bank: Washington DC. Processed.

--------. Forthcoming. "How Petty is Petty Corruption? Evidence from Firm Surveys in Africa." World

Development. Available on line at http://dx.doi.org/doi:10.1016/j.worlddev.2010.11.002 .

--------. 2011. "Reticence and Firm Performance: Evidence from Nigeria." Texas A&M International

University: Laredo, TX. Processed.

Clarke, George R. G., Manju Kedia Shah, and Roland V. Pearson. 2010a. "The Business Landscape for

MSMEs and Large Enterprises in Zambia." World Bank: Washington DC.

Clarke, George R. G., Manju Kedia Shah, Marie Sheppard, Julliet Munro, and Roland Pearson. 2010b.

"The Profile and Productivity of Zambian Businesses." World Bank: Lusaka, Zambia.

Clausen, Bianca, Aart Kraay, and Peter Murrell. 2010. "Does Respondent Reticence Affect the Results of

Corruption Surveys? Evidence from World Bank Enterprise Survey for Nigeria." Policy Research

Working Paper 5415. World Bank, Washington DC.

Clerides, Sofronis K., Saul Lach, and James R. Tybout. 1998. "Is Learning by Exporting Important?

Micro-Dynamic Evidence From Colombia, Mexico, and Morocco." Quarterly Journal of

Economics 113(3):903–947.

http://dx.doi.org/doi:10.1016/j.worlddev.2010.11.002

239

Collier, Paul. 1998. "Globalization: Implications for Africa." In Zubair Iqbal and Moshin S Khan, eds.,

Trade Reform and Regional Integration in Africa. Washington DC: International Monetary Fund,

pp. 147--181.

Conway, Patrick, and Manju Shah. 2010. "Who's Productive in Zambia's Private Sector." World Bank:

Lusaka, Zambia.

Djankov, Simeon, Caroline Freund, and Cong Pham. 2010. "Trading on Time." Review of Economics and

Statistics 92(1):166–173.

Djankov, Simeon, Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Shleifer. 2002. "The

Regulation of Entry." Quarterly Journal of Economics 117(1):1–37.

Eifert, Benn, Alan Gelb, and Vijaya Ramachandran. 2008. "The Cost of Doing Business in Africa:

Evidence From the World Bank's Investment Climate Surveys." World Development 36(9):1531–

1546.

Escribano, Alvaro, and J. Luis Guasch. 2005. "Assessing the Impact of the Investment Climate on

Productivity Using Firm-Level Data: Methodology and the Cases of Guatemala, Honduras and

Nicaragua." Policy Research Working Paper 3621. World Bank, Washington DC. Available on

line at http://go.worldbank.org/F4W5VBGDR0.

Escribano, Alvaro, J. Luis Guasch, Manuel de Orte, and Jorge Pena. 2008. "Investment Climate

Assessment Based on Demean Olley and Pakes Decompositions: Methodology and Applications

to Turkey's Investment Climate Assessment." Universidad Carlos III de Madrid, Getafe, Spain.

Escribano, Alvaro, J. Luis Guasch, Jorge Pena, and Manuel de Orte. 2005. "Investment Climate

Assessment on Productivity and Wages: Analysis Based on Firm Level Data from Selected South

East Asian Countries." World Bank: Washington DC. Available on line at

http://www.bnm.gov.my/microsites/rcicc/papers/s1.escribano.pdf. Processed.

Fajnzylber, Pablo. 2004. "Learning Before of From Exporting? Evidence from Brazilian Manufacturing

Firms." World Bank: Washington DC. Processed.

Grenier, Louise, Andrew McKay, and Oliver Morrissey. 1999. "Exporting, Ownership, and Confidence in

Tanzanian Enterprises." World Economy 22(7):995–1011.

Hallward-Driemeier, Mary, Giuseppe Iarossi, and Kenneth L. Sokoloff. 2002. "Exports and

Manufacturing Productivity in East Asia: A Comparative Analysis with Firm-Level Data."

National Bureau of Economic Research Working Paper Series 8894. Cambridge, Mass.

Hardy, Melissa A. 1993. Regression With Dummy Variables. Newbury Park, CA: Sage Publishing.

Iarossi, Giuseppe, and George R. G. Clarke, eds. 2011. Nigeria 2011: An Assessment of the Investment

Climate in 26 States. Washington DC: World Bank.

http://go.worldbank.org/F4W5VBGDR0

http://www.bnm.gov.my/microsites/rcicc/papers/s1.escribano.pdf

240

Iarossi, Giuseppe, Peter Mousley, and Ismail Radwan. 2009. An Assessment of the Investment Climate in

Nigeria. Washington DC: World Bank.

Iwanow, Tomasz, and Colin Kirkpatrick. 2010. "Trade Facilitation and Manufactured Exports: Is Africa

Different?" World Development 37(6):1039–1050.

Kraay, Aart. 1999. "Exports and Economic Performance: Evidence From a Panel of Chinese Enterprises."

Revue D'Economie Du Developpement 0(1-2):183–207.

Langbein, Laura, and Stephen Knack. 2010. "The Worldwide Governance Indicators: Six, One, or

None?" Journal of Development Studies 46(2):350–370.

Levinsohn, James. 2008. "Comments on Methodology Used by Escribano, Guasch, and Co-Authors in

Analyzing World Bank Investment Climate Surveys." University of Michigan: Ann Arbor, MI.

Processed.

Liu, Jin-Tan, Meng-Wen Tsou, and James K. Hammitt. 1999. "Export Activity and Productivity:

Evidence From the Taiwan Electronics Industry." Weltwirtschaftliches Archiv 135(4):675–91.

Maloney, William F. 1999. "Does Informality Imply Segmentation in Urban Labor Markets? Evidence

From Sectoral Transitions in Mexico." World Bank Economic Review 13(2):275–302.

--------. 2004. "Informality Revisited." World Development 32(7):1159–1174.

Mengistae, Taye, and Catherine Pattillo. 2004. "Export Orientation and Productivity in Sub-Saharan

Africa." International Monetary Fund Staff Papers 51(2):327–353.

Ministry of Labor and Social Security. 2009. "Labor Force Survey." Ministry of Labor and Social

Security: Lusaka, Zambia.

Morrissey, Oliver, and Igor Filatotchev. 2000. "Globalization and Trade: The Implications for Exports

From Marginalized Economies." Journal of Development Studies 37(2):1–12.

Pakes, Ariel. 2008. "Theory and Empirical Work on Imperfectly Competitive Markets." NBER Working

Paper 14117. National Bureau of Economic Research, Cambridge MA. Available on line at

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1149363.

Pritchett, Lant. 2001. "Where Has All the Education Gone?" World Bank Economic Review 15(3):367–

391.

Regional Program on Enterprise Development, Africa Private Sector Group. 2009. "An Assessment of the

Investment Climate in Zambia." World Bank: Washington DC.

Sandefur, Justin, Francis Teal, Paolo Falco, Andrew Kerr, and Neal Rankin. 2010. "The Returns to

Formality and Informality in Urban Africa." Center for the Study of African Economic 2010-03.

Oxford University, Oxford UK.

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1149363

241

Schneider, Friedrich. 2005. "Shadow Economies Around the World: What Do We Really Know?"

European Journal of Political Economy 21(3):598–642.

Schneider, Friedrich, and Robert Klinglmair. 2004. "Shadow Economies Around the World: What Do We

Know?" Discussion Paper 1043. Institute for the Study of Labor (IZA), Bonn, Germany.

Soderbom, Mans, and Francis Teal. 2003. "Are Manufacturing Exports the Key to Economic Success in

Africa?" Journal of African Economics 12(1):1–29.

Svensson, Jakob. 2005. "Eight Questions About Corruption." Journal of Economic Perspectives

19(3):19–42.

Tybout, James R. 2003. "Plant- and Firm-Level Evidence on "New" Trade Theories." In E. Kwan Choi

and James Harrigan, eds., Handbook of International Trade. Malden, MA: Blackwell Publishers,

pp. 389--415.

World Bank. 2003. Doing Business in 2004. Washington DC: World Bank.

--------. 2004. "Investment Climate Assessment: Enterprise Performance and Growth In Tanzania." World

Bank: Washington DC.

--------. 2009a. Doing Business 2010. Washington DC: World Bank.

--------. 2009b. "Enterprise Survey (Productivity and Investment Climate Survey) in Zambia: Sample

Survey Design." World Bank: Washington DC.

--------. 2010a. "Cape Verde: Initial Assessment of the Formal Labor Market." World Bank: Washington

DC.

--------. 2010b. Doing Business 2011. Washington DC: World Bank.

--------. 2010c. World Development Indicators. Washington, D.C.: World Bank. Available on line at

http://data.worldbank.org/data-catalog.

--------. 2010d. "Zambia Business Survey: The Profile and Productivity of Zambian Businesses." World

Bank: Lusaka, Zambia.

Yoshino, Yutaka. 2008. "Domestic constraints, firm characteristics, and geographical diversification of

firm-level manufacturing exports in Africa." Policy Research Working Paper 4575. World Bank,

Washington DC. Available on line at

http://intranet.worldbank.org/servlet/main?pagePK=64161651&theSitePK=84798&piPK=641616

52&menuPK=64166238&entityID=000158349_20080324134022.

I.10 Endnotes

http://data.worldbank.org/data-catalog

http://intranet.worldbank.org/servlet/main?pagePK=64161651&theSitePK=84798&piPK=64161652&menuPK=64166238&entityID=000158349_20080324134022

http://intranet.worldbank.org/servlet/main?pagePK=64161651&theSitePK=84798&piPK=64161652&menuPK=64166238&entityID=000158349_20080324134022

242

1 In the most recent set of surveys, managers rank 16 obstacles: (i) electricity; (ii) telecommunications; (iii)

transport; (iv) customs and trade regulation; (v) practices of competitors in the informal sector; (vi) access to land; (vii) crime, theft and disorder; (viii) access to finance; (ix) tax rates; (x) tax administration; (xi) business licensing and permits; (xii) political instability; (xiii) corruption; (xiv) courts; (xv) labor regulation; and (xvi) inadequately educated workers. The list of constraints—and how the constraints are phrased—has varied considerably across Enterprise Surveys. 2 The Zambia and Ethiopia ICAs, for example, only discuss the first measures (Regional Program on Enterprise

Development, 2009d; World Bank, 2009c), while the Tanzania ICA discusses both but focuses on the first measure (Regional Program on Enterprise Development, 2009b) 3 See Dani Rodrik’s weblog (2007)

4 During the middle of the 2007-08 Enterprise Survey in South Africa, a serious power crisis hit the country. Since

South African firms were used to cheap and reliable power (Clarke and others, 2007; 2008), this was a shock to managers. Before the crisis hit, firm managers were most likely to say that crime was a serious problem. After the crisis, they were most likely to say that electricity was a problem (Clarke, forthcoming). 5 In addition to the investment climate assessments for Ethiopia, Tanzania, and Zambia, see, for example, the

recent investment climate assessments for Kenya (Iarossi, 2009), Mauritius (Regional Program on Enterprise Development, 2009a), and Nigeria (Iarossi and others, 2009; Iarossi and Clarke, 2011). 6 For example, the South African ICA compares the perceptions of managers of white-owned and African-owned

businesses (Clarke and others, 2007). 7 Other studies have looked at that have looked only at access to finance have found similar results. In particular,

both Beck and others (2005) and Clarke and others (2006), which use data from the World Business Environment Survey (WBES), a World Bank-EBRD initiative that preceded the Enterprise Surveys also found that small firms were more likely to say that access to finance were more serious problems for small enterprises in some model specifications. 8 Bertrand and Mullainthan (2001), for example, argue that cognitive problems, the social acceptability of some

responses and wrong, non- and soft attitudes all affect the reliability of subjective survey responses. 9 A separate concern is whether perception-based data can be used to benchmark constraints across countries. If

cultural differences make complaining about the investment climate less acceptable or make people more optimistic in some countries than others, this might be difficult. Similarly, firm managers might be more willing to complain about government policy in countries where political freedom is greater. Some evidence is consistent with this. Jensen and others (2008) show that non-response patterns and lying reduce measured corruption in politically repressive environments. Similar patterns also appear for less sensitive questions. In particular, Clarke and others (2006) show that firms appear to complain more about access to finance in countries that are more free politically than in other countries after controlling for other country and firm characteristics. Another concern is that respondents in different countries—or in the same country at different times—might use different yardsticks or reference points to assess the severity of constraints. Having five days each month without power might seem manageable for a firm in a low-income country in Africa that is used to more frequent cutoffs, but might be seen as a serious constraint in a high-income economy such as the United States where firms are used to reliable power supply. If yardsticks or reference points vary across countries or across time, the cross-country or cross-time comparisons will be difficult. 10

This does not imply that response only reflect business confidence. If it did, we would expect firms to rank all concerns similarly depending on their level of overall confidence. Hallward-Driemeier and Alterido (2009) show that this is not the case—there is substantial variation in how individual respondents respond to questions on different constraints. 11

Gelb and others (2006b) find some significant correlations (e.g., for finance, power, and corruption) in surveys for Africa, but weaker correlations for measures related to regulation. Hellman and others (1999) show that perceptions about exchange rates and telecommunications infrastructure are correlated with objective data in these areas using data from Eastern Europe and Central Asia 12

In other case, the concerns appear similar using the two measures. In Tanzania, for example, the top constraints based upon the percent of firms saying each were electricity, access to finance, and tax rate. These were also the

243

top three concerns based upon the percent of firms saying that they were the biggest problem that they faced (Regional Program on Enterprise Development, 2009b). 13

The Global Competitiveness report has historically focused on larger, foreign-owned enterprises, although this focus has shifted in recent years. For example, the 2004-05 report noted that ‘survey respondents should be chief executives or members of senior management with some international perspectives’ (Porter and others, 2005, p. 200). In the 2009 report, it is noted that the ‘survey sampling guidelines’ emphasize the need to have a sample with a sufficient presence of large companies (Schwab, 2009). When the sample for the Executive Opinion Survey is compared with the Enterprise Survey sample, the comparison suggests that this remains the case. For example, in Zambia, about 37 of the 90 enterprises in the Executive Opinion Survey (40 percent) that reported their size had over 100 employees For the weighted sample for the Enterprise Survey, only about 13 percent had over 100 employees. Even the Enterprise Survey, with its restriction on firms having at least 5 employees, ignores most small firms. The Zambia Business Survey, a survey focusing on microenterprises, noted that two-thirds of MSMEs in Zambia have no employees other than the owner and 97 percent have less than 10 employees (World Bank, 2009a). 14

Based upon the ISIC 3.1 categorization, the Enterprise Survey covers all manufacturing sectors (group D), construction (group F), retail and wholesale services (sub-groups 52 and 51 of group G), hotels and restaurants (group H), transport, storage, and communications (group I), and computer and related activities (sub-group 72 of group K). Survey design is discussed on the World Bank’s Enterprise Survey website in more detail (www.enterprisesurveys.org). See also, World Bank (2009b) 15

It is also important to remember that there are concerns about objective data as well—particularly for sensitive and difficult questions. For example, some work has shown that managers appear to find it difficult to answer questions that involve calculating percentages. Clarke (forthcoming) shows that managers in Sub-Saharan Africa that report bribes as a percentage of sales report bribe payments that are between four and fifteen times higher when they report them as a percent of sales than when they report them in monetary terms. This does not appear to be due to outliers, differences between firms that report bribes in monetary terms and firms that report them as a percent of sales, and the sensitivity of the corruption question. Lying is also a problem. Azfar and Murrell (2009) show that even broad questions about corruption, including questions about ‘firm like yours’, suffer from serious problems with lying and non-response that can lead to substantial underestimates of the extent of corruption. 16

Xu (forthcoming) provides an excellent literature survey on this topic. 17

Consistent with this, Harrison and others (2011) find that bank credit is negatively associated with investment in low-income countries. 18

See also comments on the Escribano-Guasch methodology by Levinsohn (2008), Pakes (2008) and Verhoogen (2008). 19

See, for example, Aterido and others (2009; 2011); Aterido and Hallward-Driemeier (2010); Clarke (2009); Dollar and others (2005); Escribano and Guasch (2005); Fisman and Svensson (2007); Harrison and others (2011); and Svensson (2003). 20

Dinh and others (2011) also use four objective measures—mostly different from the ones used in other studies. 21

Firms at the 25th

percentile, for example, in terms of productivity have higher labor productivity (value-added per worker) than 25 percent of firms in the sample and lower labor productivity than the remaining 75 percent of firms. 22

Note that World Bank (2009d; 2010) measure productivity using total factor productivity, a measure of productivity that takes into account capital use, rather than labor productivity. xxiii

Microenterprises with five employees or less are excluded because these enterprises are not included in the World Bank’s Enterprise Survey program—the source of data for the analysis. . xxiv

Although Ethiopia appears to be an exception to this general rule—Italy, the United Kingdom and Germany are the three most important export destinations for Ethiopian enterprises—it is important to note that very few Ethiopian enterprises export. Although over 50 percent of exporters in Ethiopia export goods to their main industrial market (Italy), this represents less than 4 percent of Ethiopian enterprises. In contrast, although only about 8 percent of Kenyan exporters export to their main industrial market (the United Kingdom), since 58 percent of Kenyan enterprises export, this represents over 4 percent of Kenyan enterprises. The poor performance of


244

Ethiopian exporters in regional markets probably reflects regional difficulties that have prevented Ethiopian enterprises from developing export partnerships with firms in neighboring countries (e.g., in Eritrea, Somalia and Sudan) xxv

Söderbom and Teal (2003) find that exports—although not manufacturing exports in particular—were associated with income growth in nine countries in Sub-Saharan Africa. xxvi

Several investment climate assessments, which calculate productivity data for the firms in this study have found evidence consistent with the idea that exporters are more efficient. See, for example, World Bank (2004) for Tanzania or Iarossi and others (2009) and Iarossi and Clarke (2011) for Nigeria. These studies are available on the World Bank’s website (http://www.worldbank.org/ privatesector/ic/ic_ica.htm ). A similar relationship has been observed in many developing and developed countries. The large literature on this topic is summarized in Tybout (2003). xxvii

Results from other countries are inconclusive. Using data from Columbia, Mexico and Morocco from the 1980s and early 1990s, Clerides (1998) conclude that the evidence supports the self-selection hypothesis, while providing little support for the learning by exporting hypothesis. Bernard and Jensen (1999) and Liu et al. (1999) find similar results for the United States and Taiwan, China. Aw et al. (2000), using data from the 1980s and early 1990s, find some evidence to support the learning-by-exporting hypothesis for some industries in Taiwan, China but no evidence to support it for Korea.

Other studies, however, do find evidence consistent with learning by exporting.

Kraay (1999) and Fajnzylber (2004) finds evidence or learning among Chinese and Brazilian enterprises respectively. Finally, using data from Indonesia, the Philippines, Thailand, Malaysia and Korea, Hallward-Driemeier et al. (2002) find that exporters take concrete steps to improve productivity before they enter export markets (e.g., training employees and using foreign technology). They interpret this as suggesting that firms try to pass the threshold to enter such markets. xxviii

Bigsten et al. (2004) also find evidence consistent with the self-selectivity hypothesis. xxix

Azfar and Murrell (2009) develop a methodology using random response questions to identify reticent respondents who appear to be evasive or deceitful. They show that reticent respondents in Romania appear to underreport bribe transaction. Clausen and others (2010) repeat this analysis for firms in Nigeria and show similar results with respect to direct questions on corruption. They also show that reticent managers appear to over-report that they are ISO certified. They suggest that this is because reticent managers appear to exaggerate their firm’s performance. Using the same data as Clausen and others (2010), Clarke (2011) shows that reticent managers appear to over-report how productive and capital intensive their firms are. xxx

The number of workers is the number of permanent and temporary full-time workers. Data on part-time workers is not collected in most countries outside of Sub-Saharan Africa and so these workers are omitted to allow for reasons of comparability. In practice, for countries with data on part-time workers, including these workers does not have a large impact on relative rankings. xxxi

Clarke (2010a) note that of the 4,800 microenterprises and SMEs in the Zambia business survey only 15 exported any part of their output and only 2 exported outside of the sub-region. Even among formal manufacturing enterprises (i.e., enterprises larger than microenterprises), many studies have confirmed that large enterprises are more likely to export than small enterprises. Clerides et al. (1998) find evidence consistent with this for Colombia, Mexico and Morocco. Similarly, Grenier et al. (1999) found that large Tanzanian enterprises export more than smaller enterprises. Using data from several countries in sub-Saharan Africa from the mid-1990s, Bigsten et al. (2004) and Söderbom and Teal (2003). Clarke (2009a) found similar results using Enterprise Survey data from Ethiopia, Kenya, Mali, Mozambique, Senegal, Tanzania, Uganda and Zambia. xxxii

See, for example, Eifert and others (2008). They show that indirect costs (related to infrastructure and services) account for a relatively high share of firms’ costs in poor African countries and pose a competitive burden on African firms. xxxiii

See, for example, Svensson (2005) on corruption. Many papers have shown the strong link between other measures of institutional quality and corruption. See, for example, Langbein and Knack (2010). xxxiv

These countries were chosen because manufactured goods accounted for over 20 percent of output between 2005 and 2009 (see Table 30). xxxv

See Hardy (1993)

http://www.worldbank.org/privatesector/ic/ic_ica.htm

245

xxxvi

The previous analysis focused on medians rather than means. Medians are used because a small number of outliers tend to distort mean comparisons. Outliers could be due to a number of things including enumerator error, misreported numbers, or other events. Many firms in developing countries either do not keep detailed accounts or are unwilling to share the information with enumerators. In the analysis in this section, outliers –defined as firms where the dependent variable is more than three standard deviations from the mean—are excluded. This is used rather than robust or absolute deviation regressions because these other approaches do not allow for clustered standard errors. xxxvii

Even corporate income taxes and other taxes that are related to profits tax accounting, not economic, profits. xxxviii

Clarke (forthcoming) points out that it is very difficult to compare these costs with costs from company accounts. In particular, it appears that managers might overestimate these costs. xxxix

See, for example, Pakes (2008). A related concern is that exchange rates can also affect measured productivity. For cross-country comparisons, value-added has to be denominated in a common currency (e.g., US dollars in this paper). Because sales and intermediate inputs are denominated in local currency in the survey, cross-country comparisons of productivity are vulnerable to exchange rate fluctuations. If the exchange rate is overvalued relative to its long-run equilibrium then productivity might look artificially low. xl See, for example, the discussion by Levinsohn (2008) on the Escribano-Guasch methodology (2005; 2008; 2005).

xli The 2009 Investment Climate Assessment was based on a survey of formal enterprises of five or more

employees in Lusaka, Kitwe, Ndola and Livingstone. The list of enterprises, which was provided by the Central Statistical Office, yielded a final list of only 3 336 enterprises in manufacturing and services with over five employees in these cities. Only 449 of these enterprises had over 50 employees. Most of these, however, were in retail trade and other services. Only about 156 of these 449 enterprises were in manufacturing. At the time of the last census (2000), these cities accounted for about 20 percent of total population and probably account for a greater share of the number of large enterprises in the country. See Regional Program on Enterprise Development (2009)and World Bank (2009b) for more details. xlii

See World Bank (2003) and Djankov and others (2002). xliii

Clarke (2009a), Iwanow and Kirkpatrick (2010) and Yoshino and others (2008) show that problems with trade regulation and customs administration also make exporting difficult for firms in Africa. Djankov and others (2010) show that increasing the days to export by one day has an equivalent impact on trade as a 70KM distance addition in a gravity model of trade. xliv

See endnote xxvi for more details. xlv

Although it is difficult to compare the size of the informal sector across countries due to difficulties associated with both definitions and measurement, most evidence suggests that the size of the informal sector is larger in Sub-Saharan African than in other regions. Schneider (2005; 2004) estimates that the informal sector accounted for about 41 percent of GDP in the 24 African countries for which data were available. This is similar to Latin America, but higher than in most other regions. As in most regions, informality is generally higher in low-income countries. See, for example, Figure 1-5 in World Bank (2010a) xlvi

See World Bank (2010d) xlvii

Note that this excludes workers that are not paid in cash and excludes in-kind payments. The exclusion of in-kind payments might partially explain the exceedingly low wages in the agricultural sector. xlviii

Authors calculations based upon data from the Zambia Business Survey (ZBS). See Clarke and others {Clarke, 2010 1765 /id} for a description of the ZBS. xlix

See, for example, Pritchett (2001) for a discussion of education. l Eifert and others (2008) show that indirect costs such as those related to poor infrastructure, crime and corruption are very high in Sub-Saharan Africa. li Using data from informal enterprises in Zambia, Clarke (2009b) shows that formal microenterprises pay more in

bribes than informal enterprises. The difference appears to be, at least in part, because formal microenterprises demand more government services than informal enterprises. liiliilii

See, for example, Conway and Shah (2010).

light manufacturing in africa? - world...

Documents