gender and rural non-farm entrepreneurship
TRANSCRIPT
WORLD DEVELOPMENT REPORT 2012 GENDER EQUALITY AND DEVELOPMENT BACKGROUND PAPER
GENDER AND RURAL NON-FARM ENTREPRENEURSHIP
Costa, Rita, and Bob Rijkers
2011
The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the World Development Report 2012 team, the World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.
2
Gender and Rural Non-farm Entrepreneurship
Rita Costa Bob Rijkers DECWD DECMG
The World Bank The World Bank
Abstract: Using novel matched household-enterprise-community datasets from Bangladesh,
Ethiopia, Indonesia and Sri Lanka, this paper analyses gender differences in rural non-farm
entrepreneurship. Women have lower rates of non-farm entrepreneurship, except in Ethiopia.
Female-headed households which run a non-farm firm derive a larger share of their income from it,
even though female firms are smaller and less productive. Differences in output per worker are
overwhelmingly accounted for by sorting by sector and size, as well as differences in factor intensity
They are not due to, increasing returns to scale, differences in human capital or local investment
climate characteristics. By contrast, gender differences in investment and growth rates are small.
Key words: Gender, Rural, Non-farm, Firms, Bangladesh, Ethiopia, Indonesia, Sri Lanka
We would like to thank Klaus Deininger, Donald Larson, Josef Loening, Jack Molyneaux, Naotaka Sawada, and Mona Sur for their help in obtaining the data, and Reena Badiani, Mary Hallward-Driemeier, Andrew Mason and especially Carolina Sanchez Paramo for helpful discussions and critical feedback. The views expressed here are those of the authors and do not necessarily represent the views of the World Bank, its Executive Board or member countries. All errors are our own. Corresponding author: Bob Rijkers, [email protected].
3
Women Non-farm Entrepreneurs in Rural Areas
1 Introduction and motivation
The potentially deleterious effects of gender disparities on growth and poverty reduction have been
receiving progressively more policy attention, reflected, for instance, in the inclusion of the
promotion of gender parity amongst the millennium development goals and the 2012 World
Development Report on Gender Equity. Inequities in labor market opportunities are of particular
concern since labor earnings are the most important source of income for the poor in the vast
majority of developing countries (Lustig, 2000). Indeed women’s over-representation in poverty has
been related to their lack of labor market opportunities (see e.g. Buvinic and Gupta, 1997).
Moreover, labor market opportunities are an important determinant of women’s bargaining power
in household decision making, which has been shown to be positively correlated with household
spending on goods that benefit children.1
In developed countries documenting gender gaps in labor market participation, wage
employment and wages is a prominent way of measuring gender inequities in labor market
opportunities. A voluminous body of literature has demonstrated that such gaps are substantial,
even after controlling for women’s lower average educational attainment and labor market
experience (see e.g. Altonji and Blank, 1999, for a review of the literature). However, in developing
countries, earnings in the paid labor force are not the dominant source of income, especially not in
rural areas, where the vast majority of people are self-employed or working as “unpaid” workers in
family enterprises. In these settings, gender gaps in wage employment and wages and glass ceilings in
promotion prospects are less relevant (Mammen and Paxson, 2000).
While some studies have assessed gender differences in agricultural work (see e.g. Jamison
and Lau, 1982, Horrell and Krishnan, 2007, Udry, 1996, Goldstein and Udry, 2008) and
entrepreneurship in urban areas, gender-differences in off-farm entrepreneurship in rural areas have
not received much attention. This neglect is predominantly due to data-limitations (ILO, 2010), but
unfortunate because rural non-farm enterprises account for about 35-50% of rural income and
roughly a third of rural employment in developing countries (Haggblade et al, 2010) and because
1 See e.g.. Haddad, Hoddinott and Alderman, 1997, Katz and Chamorro 20002, Duflo 2000, Thomas, 1990, Quisumbing and Maluccio 2000, Attanasio and Lechece, 2002, Schady and Rosero, 2007).
4
women account for an important share of such non-farm activity (ILO, 2010). Moreover, the sector
appears to be growing (Lanjouw and Lanjouw, 2001) and rural off-farm diversification is widely
considered a potentially promising poverty alleviation strategy as the vast majority of poor people
continue to live in rural areas.
To help redress these lacunae in the literature, this paper draws on rural Investment Climate Survey
pilots Bangladesh, Ethiopia, Sri Lanka and Indonesia, unique matched household-enterprise-
community datasets recently collected by the World Bank, guided by three questions that have not
yet been satisfactorily resolved in the existing literature:
How and why does participation in rural non-farm entrepreneurship differ between men and women? In particular,
what are the roles of human capital, household characteristics, domestic responsibilities such as
childcare, the investment climate2 and the characteristics of the partner?
Is non-farm enterprise income a more important source of income for women than for men? Are female headed
households more dependent on non-farm enterprise income than male-headed ones?
How and why does non-farm enterprise performance – in terms of productivity, investment and growth - vary by
gender? To what extent are gender differences in performance driven by i) differences in endowments
in terms of access to factor inputs and human capital ii) sorting into different activities and iii)
differences in returns, either due to gender differences in returns to human and physical capital, or
differences in returns to scale and iv) differences in constraints.
The remainder of this paper is organized as follows. Section 2 selectively reviews related
literature and discusses the country context. Section 3 briefly describes the data and also presents a
bird’s eye view of the rural non-farm sector. A more detailed explanation of how our key variables
of interest are defined can be found in the appendix. Section four examines gender differences in
activity choice at the individual-level using multivariate probit models. Whether non-farm enterprise
income constitutes a more important source of income for female-headed than for male-headed
households is investigated in section five, which presents tobit models of the contribution of non-
farm enterprise income to the household budget. Differences in performance of male and female
2 The World Bank defines the investment climate as the set of location-specific factors shaping the opportunities and incentives for firms to invest productively, create jobs and expand (World Bank, 2005, p19). De facto, any factor that affects firm performance and decision making can be considered part of the investment climate. This has led some (e.g. Easterly, 2002) to criticize the concept as being devoid of any meaning. Instead, we take the view that it is important to clearly specify which aspects of the investment climate we are considering.
5
run firms are discussed in section six, which focuses both on differences in productivity, as well as
on differences in investment and growth. We summarize our findings and discuss their policy
implications in the conclusion.
2 Related Literature and Country Context
2.1 Related Literature
At the individual-level, women’s labor allocation is primarily determined by the opportunity
cost of working relative to earnings in productive employment, “unearned” income, other
household members’ characteristics (see e.g. Mammen and Paxson, 2000) and labor allocation, as
well as preferences over different types of employment, which may be dictated by cultural norms
and religious beliefs. In the Amhara region in Ethiopia, where the Ethiopian Rural Investment
Climate Survey data used in this paper were collected, the belief that the harvest will be bad if
women work on the farm is prevalent (Zwede and Associates, 2002, Bardasi et al., 2007). The
opportunity cost of working is inter alia determined by the presence of children in the household
and returns to working, which in turn depend on women’s human capital and the income-earnings
opportunities available to them. Literature from developed countries furthermore suggest that
parents’ occupation matters, as children of entrepreneurs are significantly more likely to become
entrepreneurs themselves (Parker, 2008, 2009).
Studies of gender differences in entrepreneurship in developing countries are scarce.
Existing studies are predominantly based on the World Bank’s Enterprise Surveys and typically find
that female entrepreneurship is inversely correlated with firm size;3 firms run by female
entrepreneurs are smaller in terms of employees, sales and capital stock. However, gender
differences in total factor productivity, profitability and capital-intensity become insignificant once
firm characteristics are controlled for (Bardasi and Sabbarwal, 2009), except for the very smallest
3 Gender differences in the performance of male and female entrepreneurs in developed countries are
relatively well documented, but the evidence is mixed. Some studies report evidence of female underperformance, whereas other do not find gender differences (see e.g the literature reviews in Parker, 2009 and Sabbarwal and Terrel, 2009 and the references therein).
6
firms (Bruhm, 2009).4 Thus, gender differences manifest themselves primarily in terms of scale,
rather than differences in profitability, technology or capital intensity. However, Hallward-
Driemeier and Aterido (2009a) point out that the definition of a female-firms matters; using
definitions based on decision making authority, rather than (partial) participation in ownership as is
done in the studies cited above, results in substantial gender differences, even after firm and
manager characteristics have been controlled for.
The finding that women operate smaller scale firms begs the question why. One possible
explanation is that they sort into industries which a have lower optimal scale, although this only
pushes the question another step backwards. Another salient explanation is that they lack access to
finance. Evidence from developed countries on this issue is mixed.5 Furthermore, cultural norms
may militate against women being in power or engaging in certain activities. Alternatively, successful
female firms, which tend to be larger, may be more likely to be “captured” by husbands. Women
entrepreneurs could also face different constraints. However, Bardasi and Sabarwal (2009) find little
evidence for differences once firm characteristics are conditioned on.
Male firms also appear to grow faster and to invest more, although the evidence on gender
differences in dynamic performance is mixed. Nichter and Goldmark (2009) survey the literature on
small firm growth and find that female owned firms are often, but not always, grow more quickly
than male firms.. De Mel et al. (2008, 2009) and McKenzie and Woodruff (2008) report lower
investment rates and returns to capital for women than for men in Sri Lanka, Bangladesh and
Mexico. Using manufacturing census data from Ethiopia Shiferaw (2009) shows that female-
managed firms are much more likely to survive.
Since most of these studies are based on urban enterprises, it is not clear to what extent their
conclusions generalize to rural areas, where firms tend to be smaller and firm performance is
arguably more intimately intertwined with household- and farm events, and the investment climate is
radically different (see World Bank, 2004, Deininger et al., 2007, Jin and Deininger, 2009, Rijkers et
4 In addition, Sabbarwal and Terrel (2009) find some evidence that female firms in ECA have higher returns
to scale, but the differences in returns to scale between men and women are small, and, moreover could be due to a rather restrictive production function specification. For example, differences in technology between sectors are modeled as being additive in TFP, rather than as interacting with the returns to capital, labor and material inputs. 5 Some studies suggest women indeed have more difficulty accessing finance than men (e.g. Brush 1992,
Carter, 2000), while others detect no gender differences (Blanchflower et al., 2003, Storey, 2004, Cavalluzzo and Wolken, 2005).
7
al., 2010). Despite their importance as a potential catalyst of growth and an absorber of growing
rural labor supply, little is known about the determinants of the performance of non-farm firms and
how these may vary with the gender of the manager. Studies based on household data often
document lower returns for women in off-farm employment (see e.g. Canagarajah et al., 2001), yet
evidence on gender differences in the relative importance of non-farm earnings for household income
is limited. Since the existing evidence on gender differences in rural non-farm entrepreneurship is
overwhelmingly based on household- and labor force surveys it does not allow us to disentangle the
importance of firm vis a vis household and community characteristics.
2.2 Country Context
The countries in this study vary radically in terms of the structure, size and composition of their
non-farm economy, and in their level of economic prosperity and gender parity. Ethiopia’s rural
economy is unusually fragmented and rural labor markets are very thin. In this context, self-
employment in non-farm enterprises is predominantly a means to supplement farm earnings, as well
as an important source of income for those lacking alternative options (Loening et al., 2008). By
comparison, in Indonesia, rural labor markets are well-developed, population density is high and
movements out of poverty are strongly correlated with non-farm entrepreneurship (McCulloch et
al., 2007, Priebe, 2009). Sri Lanka and Bangladesh’s rural non-farm sectors fall in between these two
starkly dissimilar cases (Headey et al., 2010), yet are interesting in their own right since Sri Lanka is
characterized by relatively high levels of human development relative to its GDP per capita, while
studying the non-farm sector Bangladesh is likely to be informative about the role of gender roles,
which place strong restrictions on women’s ability to engage in remunerated employment.
3 Data and descriptive statistics
3.1 Data
The World Bank’s Rural Investment Climate Pilot Surveys conducted in Bangladesh,
Ethiopia, Indonesia and Sri Lanka are the data used in this paper. These surveys are matched
household-enterprise-community surveys, that collect information on both enterprise and non-
enterprise owning households, the non-farm enterprises they operate, and the local investment
climate in the communities in which they are located. For the purposes of these surveys, a rural
8
nonfarm enterprise was defined as any income generating activity (trade, production, or service) not
related to primary production of crops, livestock or fisheries undertaken either within the household
or in any nonhousing units. In addition, any value addition to primary production (i.e processing)
was considered to be a rural nonfarm activity.6
The matched nature of the data is a key advantage. The surveys constitute an improvement
over traditional household surveys by collecting very detailed information on enterprise
characteristics, inputs and outputs, the local economic environment and various dimensions of
performance. For example, by virtue of containing information on the capital stock and inputs they
enable us to estimate production functions and assess how productively labor is employed in these
off-farm firms. Conversely, they complement the existing investment climate surveys, which are
typically urban-based, focused on relatively large manufacturing firms, and lack information on the
household characteristics of firm managers. Such surveys do not permit analysis of participation
patterns, since they lack information on potential entrepreneurs.
Since the surveys were very similar, they facilitate a cross-country comparison. However,
data coverage, variable definitions, and samplings frames can vary from country to country. For
example, labor input in Sri Lanka was measured in terms of number of workers only, whereas in
Ethiopia it was measured in terms of days worked per employee, while in Indonesia and Bangladesh
we have information on hours worked. To maximize comparability across surveys, we have
converted our labor inputs measure into a full-time worker equivalent. Similarly, measures of
material inputs, capital and labor were converted into USD equivalents. As another example, the
definition of what constitutes a rural area and what constitutes a rural town varies across countries.
The Appendix discusses how we defined our key variables of interest and tried to maximize
comparability across countries in more detail.
While the sampling frames for the survey vary from country to country, they typically yield a
good representation of the rural non-farm sector. The surveys in Sri Lanka and Bangladesh are
representative of all rural areas in the country, while the Ethiopia data are representative of the rural
non-farm sector in the Amhara region.7 The Indonesia data cover six different kabupaten’s8
6 Thus, in many cases the term “activity” might have been more appropriate than the term enterprise. 7 Loening et al (2008) demonstrate that the rural non-farm sector in the Amhara region is very similar to the rural non-farm sector in other regions, although its composition is slightly skewed towards manufacturing activities, whereas in other regions, trade and services enterprises dominate.
9
(“districts”) in six different provinces. In all countries but Ethiopia, relatively large firms were
oversampled to ensure they were included in the surveys. However, our survey suggests that large
off-farm firms are very rare in rural Ethiopia.9 Unfortunately, except in Bangladesh, we do not have
information on the households of the managers of such relatively large firms. For more information
on the samplings frames, the reader is referred to World Bank, 2005.
3.2 The rural investment climate
The rural investment climate is characterized by remoteness, weak infrastructure, low penetration of
commercial credit providers and localized markets (see also World Bank, 2005) and varies
substantially both across as well as within countries. Most non-farm enterprises are very small and
generate low profits, although the non-farm enterprise sector is highly heterogeneous both in
composition and performance. The characteristics of the rural business environment are also
reflected in the constraints firm managers report to be most severe. The appendix demonstrates that
both male and female managers consider a lack of markets (demand), transport, access to credit and
electricity as their most important constraints. Gender differences in self-reported constraints are
minimal, a finding which is robust to controlling for differences in firm characteristics of male and
female managed firms.
4 Which income-earning activities do men and women engage in?
4.1 The structure of rural labor markets
Rural labor markets are thin and average educational attainment is low, although it varies
across countries, with workers in Indonesia being relatively well educated and Ethiopians having had
very little education on average. Self-employment, either in agriculture or off-farm, accounts for the
bulk of employment, as demonstrated in Tables 1A and 1B, which present descriptive statistics on
8 A kabupaten is a political subdivision of a province in Indonesia. The Indonesian term kabupaten is also sometimes translated as 'district' or 'municipality'. 9 This is not an artifact of the sampling strategy. Only three enterprises in the rural dataset employ more than 10 workers. Since these enterprises are all household-based we might miss out on fully commercial enterprises owned or managed by individuals not living in these communities. However, from the community level dataset one can infer that there are not more than a dozen firms with more than 20 employees in a radius of 1-h commuting distance from the 179 surveyed communities. It thus seems safe to conclude that there are very few large firms in rural areas.
10
activity portfolios of individuals and households.10 The importance of different types of
employment, including non-farm self-employment, varies considerably across countries. Off-farm
entrepreneurship participation rates are lowest in Ethiopia and highest in Bangladesh. Running a
non-farm firm is often combined with undertaking other income earning activities. This suggest that
non-farm self-employment is often a secondary activity. This is especially interesting since
undertaking multiple activities is the exception rather than the norm since the vast majority of men
and women specialize in one activity.11 With the exception of Bangladesh, household level
participation rates are higher than individual participation rates (see table 3), yet few households rely
exclusively on non-farm enterprise income, and many engage in multiple activities.12 Taken together,
these suggest that income diversification is primarily a household, rather than an individual level
phenomenon.
Turning to gender differences, women are much less educated than men and less likely to be
the head of their households, unless they are widows or divorcees. They are much more likely not to
report any activity, although participation rates and gender differences vary dramatically across
countries (see Table 2).13 Roughly only about half of all Bangladeshi and Sri Lankese women report
at least one activity, while in Ethiopia approximately only a quarter of women did not report any
activity. In all countries but Ethiopia, the proportion of men working in non-farm enterprises is
larger than the proportion of women. However, conditional on reporting any activity, the likelihood
of reporting being engaged in off-farm self-employment is always higher for women than for men.
Gender differences in conditional participation rates are very small in Bangladesh, where very few
women report any income earning activity.
4.2 Empirical Strategy
To examine what is driving these differences in individuals’ activity choices, reduced form trivariate
probit models for engagement in farming, non-farm self employment and wage employment are
10
The phrasing of the different questions about individual’s activity portfolios varied across countries, forcing us to come up with an arbitrary categorization (see the appendix for details). In Indonesia, information on individuals’ engagement in agricultural activities was not available. 11 Of course, it is possible that individuals engage in different types of self-employment or hold multiple wage jobs in which case they would misleadingly be classified as engaging in one activity only. 12
In conjunction with the finding that very few household rely exclusively on non-farm household income and that most individuals report engaging in one type of activity only, this strongly 13
In this paper, we consider an individual out of the labor force if she has either engaged in agricultural self-employment, not held a wage job, nor worked in a non-farm enterprise over the past twelve months The participation rates documented here may deviate somewhat from those reported in official statistics partly because of sample coverage and partly because we are counting employment in home-based enterprises as labor market participation.
11
estimated. The trivariate probit specification allows for simultaneity of activity choices, thus
accounting for the interdependence of activity choices, which has often been neglected in previous
studies of participation in the non-farm sector (two notable exceptions are de Janvry and Sadoulet,
2001 and Babatunde and Quaim, 2010). The estimable model is:
+
+
+
Where >0 if and 0 otherwise and the error terms , and are assumed to
follow a trivariate normal distribution. The dependent variables and are dummies
indicating whether or not the individual in question engages in non-farm enterprise activity, works
on the household farm or has worked for a wage respectively. Note that these categories are not
mutually exclusive.
The explanatory variables affect individuals’ relative returns and ability to participate in
different activities; is a vector of individual characteristics, notable age, education, marital status
and relation to the head of the household. is a vector of household characteristics, including the
size and composition of the household, and household assets, notably land holdings and whether or
not the household owns the house it inhabits, IC is a vector of investment climate proxies, notably,
whether or not the household uses electricity, lives in a rural town, distance to the nearest market,
prevailing wage rates, whether or not there is a credit institution present in the village. These
objective proxies correspond to the key subjective investment climate constraints identified by
managers of non-farm enterprises (see section 3.2 and the Appendix).14 Finally, is a vector that
captures partner’s educational attainment and household head’s father’s main occupation. The latter
variable is not available in the Ethiopian dataset.
4.3 Results
Results are presented in Table 6. That the decision to work in a non-farm firm cannot be
analyzed independently of the decision to work for a wage or on the family farm is evidenced by the
14 Since subjective investment climate indicators may be endogenous to performance, we prefer objective proxies instead.
12
strong correlations between the error terms of the different equations. These correlations suggest
that working on the family farm and wage work, and wage work and working for a non-farm
enterprise are substitute activities, while the decisions to work for a family farm and to work in a
non-farm firm are less strongly interdependent, perhaps because they are easier to combine.15
Although the determinants of activity choice vary by gender and across countries, there are
some interesting commonalities. We will focus mostly on non-farm entrepreneurship, the subject of
this study. Age-participation profiles in non-farm enterprise activity are strongly concave for women,
and insignificant or even convex for men. A similar, but less pronounced gender difference in age-
participation profiles is observed for self-employment in agricultural activities. Education has a
strong impact on activity choice, but its association with off-farm entrepreneurship is weak and
especially so for women. Workers with the highest levels of education are significantly more likely to
work in a wage job and significantly less likely to work on a family farm or in a family firm and this
effect is especially strong for women.16 Presumably this reflects higher returns to education in wage
jobs.
Non-farm self-employment appears especially important for women who are the heads of
their households, as they are much more likely to work in a non-farm enterprise. It is important to
bear in mind that these are conditional associations; in absolute terms, male-headed households are
more likely to run a non-farm firm than female-headed households in all countries except in
Ethiopia.17 Ceteris paribus, male heads in Ethiopia and Indonesia are also more likely to engage in
non-farm enterprise activity than other men, yet, with the exception of Ethiopia, the difference
between male heads and other men is smaller than the difference between female heads and other
women.
The association between household ownership of agricultural assets (plough and land) and
activity also varies by gender, as ownership of such assets is negatively correlated with female- but
not with male- off-farm entrepreneurship, except in Indonesia where we see the opposite pattern.
15 Note, however, that the error terms are significantly negatively correlated in Ethiopia, suggesting they are substitutes in this context. 16 Men who have completed some form of secondary education are significantly more likely to run a non-farm firm in Bangladesh and Sri Lanka. For women, no such effect is observed; if anything, women who completed secondary education are less likely to be entrepreneurs in Ethiopia, and not significantly more likely to be entrepreneurs in any of the other countries. The propensity to be an entrepreneur seems to drop with further education Indeed, both men and women who have completed tertiary education are significantly more likely to be wage employed and much less likely to be self-employed in agriculture. These effects are particularly strong for women. 17 Female headed households are typically headed by widows.
13
By contract, men’s self-employment in agriculture is always significantly positive correlated with
ownership of agricultural assets, while for women a positive conditional association is observed only
in Ethiopia and Sri Lanka.
Gender differences in the impacts of other variables are less pronounced. The association
between marriage and activity choice varies across countries and by gender but Bangladeshi women
who are married are far less likely to undertake any activity, presumably reflecting gender norms
constraining their engagement in income earning activities. Being located in a town is correlated with
an increased likelihood of working in a non-farm enterprise, both for men and women, but there are
few other investment climate variables that are consistently significant across countries. The
occupation of the father of the household head is strongly correlated with activity choice, but not
with the likelihood of running a non-farm enterprise. Individuals in households of which the head’s
father held a wage job are far less likely to work in agriculture and far more likely to hold a wage job.
Individuals whose father’s main source of income was running a non-farm firm appear less likely to
work on a farm but are not consistently much more likely to be engaged in NFE activity, which
contrast with studies of the intergenerational transmission of entrepreneurship in developed
countries (see e.g. Parker, 2009). By contrast, the impact of partner’s educational attainment varies
across countries but is generally insignificant. Last but not least, household size and composition are
not consistently significantly correlated with activity choice, ceteris paribus.
5 Contributions to Household Income
5.1. Descriptive Statistics
Non-farm enterprise earnings are an important source of income; in Indonesia female-
headed households on average derive 19% share of their pecuniary income from non-farm
enterprise activity while male-headed households derive an average of 21% of income from running
a non-farm firm as shown in Table 3 which presents descriptive household statistics.18 These
averages do not condition on participation: since female participation rates are lower, female-headed
households that run a firm typically derive a larger share of income from it. In Ethiopia, female
18 Table 3 only presents descriptive statistics for Ethiopia and Indonesia since these are the only two countries for which information on the contribution of non-farm enterprise activity to household income is available. Note that production for auto-consumption is not taken into account in these pecuniary income measures.
14
headed households on average derive 19% of their household budget from non-farm enterprise
activity, while male-headed household only derive 6% of their income from it.
Graph 1 plots the contribution of non-farm enterprise income against log household
consumption per capita for male- and female-headed households, demonstrating that the correlation
between enterprise income and income rank is weak, except for male-headed household in
Indonesia. Thus, prima facie, it does not appear to be the case that non-farm enterprises either
exacerbate or diminish income inequality. Also note that female headed households have much
lower expenditure per capita on average and tend to be smaller than male headed households.
5.2 Empirical Strategy
To assess the contribution of non-farm enterprise activity to household income a Tobit model is
estimated following Gibson and Olivia (2010):
Where the dependent variable, , is the share of income pecuniary income derived from non-
farm enterprise activity, which is assumed to be latent variables that is only observed for positive
income shares (i.e. for , is a vector of household characteristics
including characteristics of the household head and expenditure quartile, 19 is a vector of
investment climate characteristics and is an independently distributed error term with zero mean
and constant variance.
5.3 Results
Table 5 presents the results which demonstrate that the relative contribution of non-farm
enterprise activity to household income is significantly lower in Indonesia if the household head is a
man. In Ethiopia, the coefficient on a male household head is negative too, but statistically
insignificant. However, in Ethiopia non-farm enterprise income’s contribution to household
19
We include these dummies to assess whether the contribution of enterprises to household income varies with income rank. Their inclusion may create an endogeneity bias since both the share of non-farm income and income rank are functions of non-farm enterprise income. However, our results are robust to exclusion of these dummies. The results of a robustness check in which we dropped these dummies are not presented to conserve space, but available from the authors upon request.
15
income is significantly positively correlated with the share of female-household members, which is
larger in female-headed households. Thus, it seems safe to conclude that female headed households
which run a non-farm firm derive a larger share of income from non-farm enterprise activity in both
countries. In part this is due to their smaller size.
The importance of enterprise income increases with household size and being located in a
town, yet decreases with the ownership of agricultural assets and local wage rates. which presumably
proxy for the availability of alternative income earning opportunities. Household headed by
relatively highly educated individuals derive a smaller share of income from non-farm activity and
the share of income is concavely related to the age of the household head. The importance of non-
farm income does not appear to vary strongly with income levels, suggesting that the impact of non-
farm enterprise activity on inequality is limited.
6 Performance
6.1 Productivity
6.1.1 Female enterprises are smaller and less productive.
Female owned firms have significantly lower sales on average than male owned firms in all countries
surveyed. This is in part because they are significantly smaller and more likely to be household
enterprises than male run firms and in part because, with the exception of Indonesia, they tend to
have lower labor productivity, measured in terms of sales per worker. The stylized fact that the share
of firms with a female manager is inversely correlated with enterprise size, which has been
documented for large, urban-based manufacturers in Latin America (Bruhm, 2009), Africa
(Halward-Driemeier et al. forthcoming) and Eastern Europe (Sabbarwal and Terrel, 2008) also holds
in our sample of rural MSEs.
Gender differences in average output per worker are large, but vary dramatically across
countries. For example, the largest gender difference is found in Bangladesh where the US dollar
equivalent of mean log output per worker in female firms is on average $35, while it is $293 for male
run firms. In Indonesia, by contrast, there are no gender differences on average, with mean log
output per worker in female firms being equivalent to $1746 for women and $1737 for men. While
16
female firms are on average less productive, labor productivity is highly dispersed, as shown in
graphs 2A-2D which plot the density of output per worker. This dispersion is indicative both of the
high heterogeneity of the non-farm sector, which comprises a wide variety of activities, and market
power. Most firms face very few competitors, if any.
At first sight, these differences in labor productivity do not appear driven by differences in
capital intensity. Surprisingly, male enterprises do not appear to use more capital per worker than
female ones. These findings do not necessarily indicate an absence of gender differences in access
to capital;20 it is very well possible that women sort into certain sectors, which operate on a relatively
smaller scale. Since they are larger, male firms tend to use more capital in absolute terms.
Sector sorting patterns by gender are very pronounced, but vary from country to country. In
Bangladesh for instance, 95% of all female firms are engaged in manufacturing and almost none are
in trade, while 49% of all male firms are in trade and only 24% are in services. In Ethiopia too,
women are overrepresented relative to men in manufacturing, which accounts for the bulk of non-
farm enterprise activity. By contrast, in Indonesia, only 3% of all female firms are in manufacturing
compared with 14% of all male firms. In this case, women are overrepresented in the trade sector.
These broad sector categories hide sorting into different subsectors (see the appendix). Differences
in output per worker could also be due to differences in input usage and human capital. With the
exception of Indonesia, male firms use significantly more inputs per worker and male entrepreneurs
are better educated, on average, then their female counterparts.
6.1.2 Empirical Strategy
To examine to what extent differences in output per worker are due to differences in factor
inputs and human capital, sorting into different sectors, returns to scale, we estimate Cobb-Douglas
production functions, modeling output Y, as a function of capital, , labor, , material inputs,
, and Total Factor Productivity (TFP), which is in turn modeled as a function of sector, S, firm
characteristics F, characteristics of the manager, E, and a host of investment climate characteristics,
20
Interestingly, only in Bangladesh differences female firms are significantly less likely to report having a loan than male firms.
17
IC:
. Taking-logs and adding
these and an error term, v, our most general estimating equation becomes:21
Where j subscripts denote sector specific factor coefficients, as we allow capital, labor and material
inputs coefficients to vary freely across sectors. In our empirical specification, we progressively add
explanatory variables to examine to what extent these different sets of explanatory variables account
for gender differences in output. Differences in productivity could also derive from differences in
returns to scale. If returns to scale are increasing (
, the larger size of male run
firms may account for their higher average productivity. Alternatively, differences in productivity
may stem from sector selection. Note that these mechanisms may interact if returns to scale vary
across sectors. In the appendix we also present specifications where we interact all variables used to
explain total factor productivity with a dummy for a female entrepreneur, to allow for gender
differences in the returns to human capital, experience and various investment climate proxies. The
results suggest that gender differences in the impact of these variables are generally statistically
negligible.
This approach has well-known limitations. Measurement error in explanatory variables or
omitted variables may lead to biased estimates of the productivity differentials. Despite having a rich
and detailed dataset, we cannot directly control for potentially important variables such as demand
and price differences. In principle, such endogeneity problems could be remedied by means of
instrumental variable estimation but, unfortunately, credible instruments are not available in our
data. However, Loening et al. (2008) argue that endogeneity bias is likely to be small. Using
precipitation based indicators of local agricultural performance as a proxy for unobserved local
demand, they find that while better local agricultural performance is strongly correlated with firm
sales, their parameter estimates are not very sensitive to controlling for this variable, indicating
endogeneiy bias is small. Moreover, firms do not frequently adjust factor inputs, despite facing
frequent shocks, again suggesting the impact of endogeneity bias may be limited.
21 In the appendix, we also present models where we interact factor shares with the gender dummy. However, we had to
pool the data as the number of observations did not us to account for gender heterogeneity in technology by sector.
18
In interpreting the results it is also important to bear in mind that we only focus on the
direct impact of gender and the investment climate on firm productivity. However, perhaps the
most important gender and investment climate effects are indirect. For example, the investment
climate may also impact on allocative efficiency (see, e.g. Mengistae and Honorati, 2009), and
women may sort into small-scale low productivity sectors. In short, we have to be cautious when
interpreting the results from these models.
6.1.3 Results
The regressions are estimated separately for each country. Results are presented in Table 7.
The dependent variable is the log of sales in US dollars. We present five different models. The first
model only includes a gender dummy while the second model only includes a control for firm size.
Subsequent models add controls for sector (model 3), factor usage and differences in technology
across sectors (model 4) and other firm, human capital and investment climate characteristics (model
5).22
Models 1 and 2 confirm that raw performance gaps are large in Bangladesh, Ethiopia andSri
Lanka. These performance gaps are in part the result of sorting by size, (model 2), but appear to be
predominantly driven by sorting into different sectors (model 3) and material inputs usage (model 4);
once sector and factor usage are controlled for performance gaps reduce dramatically and the
explanatory power of our model shoots up, as is evidenced by the big increases in the R2s when we
add these explanatory variables. In Sri Lanka, the male-female productivity differential is entirely
explained by sorting. In Bangladesh, the gender differential reverses sign, while in Ethiopia the
productivity gap roughly halves. Parameter estimates appear reasonably well behaved, although it is
likely that the coefficients on capital are biased downwards due to measurement error. Capital- and
labor coefficients vary substantially across sectors, and within sectors across countries. The null
hypotheses of constant returns to scale is rejected only for manufacturing firms in Bangladesh and
trading firms in Indonesia, for which the parameter estimates suggest decreasing returns to scale.
Adding additional controls for human capital, firm characteristics and the investment
climate does not change results very much and adds limited explanatory power, suggesting that most
22
It would have been of interest to examine whether enterprises run by female household heads are more or less productive than other female firms. However, our data do not enable us to cleanly identify whether the enterprise manager is also the household head.
19
of the productivity differential is accounted for by sorting into different sectors and factor usage.
Gender productivity differences do not disappear altogether. The positive premium associated with
being a female manager in Bangladesh may not be well identified since we have relatively few female
firms. Alternatively, it could be due to the fact that female entrepreneurs are a highly eclectic group.
The low productivity of female-run firms in Ethiopia may in part driven by sorting into different
activities that our crude sector controls do not capture. They may also be due to gender differences
in hours worked, as we only observe differences in labor days. If women combine working in a non-
farm firm with household chores, they may work fewer hours in any given day.
The final specification furthermore suggest that household-based enterprises are on average
less productive than stand-alone firms (although the coefficient on the household enterprise dummy
is not statistically significant in Indonesia and Sri Lanka), that older firms are more productive
(although the coefficient on firm age is not statistically significant in Indonesia and Sri Lanka), and
that investment climate variables are not very important determinants of firm performance.
Moreover, it is remarkable how much the importance of different factors, such as schooling,
electricity usage, and location, varies across countries, underscoring how heterogeneous the non-
farm sector is across different countries.
Overall, then it seems that the evidence for gender differences in the returns to human
capital is limited. In addition, we find little evidence for gender asymmetries in the direct impact of
the investment climate on firm performance. Nor do we find evidence for gender differences in
productivity differentials associated with firm age and electrification. Moreover, differences in
returns to scale do not appear to explain gender productivity differentials; while male entrepreneurs
run larger firms we do not find evidence for increasing returns to scale. We do however, find strong
evidence for differences in productivity due to sorting into different types of activities, factor usage
and firm size.
6.2 Investment and Growth
6.2.1 Gender differences in dynamic performance are small
Gender differences in the dynamic performance of firms are small and much less pronounced than
the dramatic productivity differences discussed above. Investment and growth rates are low in
20
Ethiopia and Bangladesh, but substantial in Indonesia and Sri Lanka. Male managers are not
significantly more likely to invest than female ones in any country. In Sri Lanka, they are significantly
less likely to invest than female managers. Male firms also do not grow faster than female firms on
average. However, the absence of strong difference in absolute growth and investment rates may
obscure gender differences as male firms are very different from female firms. To assess whether or
not this is the case, we estimate growth and investment models which control for salient observable
characteristics correlated with growth and investment.
6.2.1 Empirical Strategy
Investment is modeled by means a probit for having invested using as explanatory variables
sector, firm-, manager- and investment climate characteristics. Again we progressively add
explanatory variables to assess to what extent firm and investment climate characteristics account for
gender differences in investment rates. Our most general reduced-form investment model is thus:
We use the same model to analyze growth, using as dependent variable the Haltiwanger growth
measure force as the dependent variable. The Haltiwanger growth measure is the change in
employment over the average firm-size for two periods: H = (Lt -Lt-1 )/[0.5*(Lt + Lt-1)], thus
bounding the growth rate between -2 and +2 and minimizing the impact of measurement error and
influential outliers. That is our estimable equation is:
where u is a normally distributed zero-mean error term.
The results of these regressions have to be interpreted cautiously. Suppose, for example, that
the presence of credit institutions is positively correlated with investment rates. This might indicate
that firms with better access to financial institutions are likely to invest more. However, it is also
possible that credit institutions locate in communities where the local economy is buoyant. Thus, we
have to be aware of potential reverse causality. In addition, it is important to emphasize that
estimated coefficients are conditional on firm survival.
6.2.2 Results
21
Results are presented in tables 8 and 9. We first condition on lagged size and sector to account for
sorting effects and subsequently add additional controls for firm characteristics, managerial human
capital and the investment climate.
Starting with the results for investment (table 8), controlling for lagged size and sector
(model 2) renders the investment differentials between male and female firms insignificant. Adding a
full set of controls (model 3) does not overturn these results, although male-firms in Indonesia are
some 38% less likely to invest, ceteris paribus, and this effect is significant at the 10, but not the 5%
level. Overall, investment is difficult to predict, as is evidenced by the rather low pseudo R2s and the
fact that relatively few explanatory variables are significant. Moreover, the determinants of
investment appear to vary substantially across countries; not a single explanatory variable is
statistically significant in each country.
The results of our growth regressions are presented in Table 9. Once we control for sorting
by sector and size firms run by Bangladeshi men grow faster than firms run by Bangladeshi women.
Yet controlling for sorting and manager, enterprise and investment climate characteristics does not
result in significant gender growth differentials in any of the other countries.
Our models for growth arguably have even less predictive power, as is indicated by the very
low R2s. However, the determinants of firm growth do not appear to vary as dramatically across
countries as the determinants of investment and productivity. While the parameter estimates are not
always statistically significant, they suggest that, overall, firms that started larger appear to have
grown less. This finding presumably partly reflects measurement error and survivor bias, as one may
expect firms that started large to be more likely to survive. In addition, manufacturing firms appear
to grow somewhat faster than firms engaged in services or trade. Firm growth also appears
positively correlated with electricity usage and negatively associated with firm age, though again
these patterns are not statistically significant in each country.
In the appendix we present a robustness check in which we allow all parameters to vary by
gender, but this does not overturn the results. Overall, the results suggest that gender differences in
investment and growth are not very large.
8 Conclusion
22
In spite of their increasing policy-prominence, relatively little is known about gender
inequities in rural non-agricultural labor market opportunities. This is unfortunate since such
employment accounts for a substantial share of rural income and employment, which exceed the
contributions of agriculture in some countries (Haggblade et al., 2007). Using novel matched
household-enterprise-community datasets from Bangladesh, Ethiopia, Indonesia and Sri Lanka we
attempt to redress this lacunae by documenting and analyzing gender differences in activity portfolio
choice, contributions of non-farm firms to household income and differences in productivity,
growth and investment of rural non-farm firms.
Women are much less likely to undertake income earning activities and gender differences in
non-farm entrepreneurship are large. Women have lower participation rates in non-farm enterprise
activities in Bangladesh, Indonesia and Sri Lanka, but not in Ethiopia. However, working in rural
non-farm firms appears to be very important for women who partake in income earning activities
and especially for those who are the heads of their households. Indeed, female-headed households
which run non-farm enterprises, despite being poorer on average, derive a larger share of their
pecuniary income from non-farm enterprise activity, even though enterprises run by women tend to
generate less revenue. This is of course in part due to the fact that they tend to be smaller.
Female firms are much smaller and much less productive on average, although gender
differences in productivity vary dramatically across countries. In Indonesia there are no significant
gender differences in output and productivity, while firms run by Bangladeshi men on average
produce more than ten times as much output per worker on average than female firms. In Sri Lanka
and Ethiopia gender differences in size and productivity are less dramatic, but large nonetheless.
Such differences in performance are overwhelmingly driven by sorting. Once differences in
size and sector are accounted for gender productivity differentials diminish. Differences in inputs
usage also provide part of the explanation. In fact, in Bangladesh, conditioning on sector, size and
inputs usage reverses gender productivity differentials. However, gender differences in productivity
do not fully disappear once we control for differences in human capital, firm characteristics and the
investment climate.
Gender differences in productivity are not due to returns to scale; while male owned firms
are much larger, we did not find evidence of increasing returns to scale. Nor are productivity
differentials driven by differences in capital intensity: capital labor ratios do not significantly vary by
23
gender in any of the countries considered. Moreover, our regressions do not support the hypothesis
that differences in human capital account for gender differences in firm performance, even though
male managers are on average much better educated than women who manage firms. Similarly,
differences in the local investment climate appear to have little direct impact on productivity
differences.
By contrast, gender differences in investment and growth rates are small, and do not obscure
underling underlying gender differences, as they are robust to controlling for a rich set of firm,
manager and investment climate characteristics.
Overall, our results demonstrate large gender disparities in rural off-farm labor market
opportunities and outcomes. While we have managed to rule out a large number of candidate
explanations, fully understanding why these gender differences arise requires additional research.
Collecting panel data would help us better understand the causal mechanisms underlying the
patterns documented in this paper and would permit a richer representation of the dynamics of rural
labor markets.
9 References
Altonji, J. and R. Blank. 1999. “Race and Gender in the Labor Market.” In Handbook of Labor Economics Vol. 3:
3143-3259, ed. O. Ashenfelter and D. Card, Elsevier Science, B.V. (Amsterdam).
Attanasio, Orazio and Valerie Lechene. 2002. “Tests of Income Pooling in Household Decisions”. Review of
Economic Dynamics 5(4): 720-748.
Bardasi, Elena and Shwetlena Sabarwal. 2009 “Gender, Access to Finance, and Entrepreneurial Performance
in Africa.” World Bank Mimeo.
Bardasi, E. and A. Getahun. 2007. “Gender and Entrepreneurship in Ethiopia.” Background paper to the
Ethiopia Investment Climate Assessment Washington DC: The World Bank.
Barrett, C.B., T. Reardon and P. Webb. 2001. “Nonfarm income diversification and household livelihood
strategies in rural Africa: concepts, dynamics, and policy implications.” Food Policy 26 (4): 315–331.
Brush, Candida G. 1992. “Research on Women Business Owners: Past Trends, a New Perspective and Future
Directions.” Entrepreneurship: Theory and Practice 16.
Babatunde, Raphael O. and Matin Qaim. 2010. “Impact of off-farm income on food security and nutrition in
Nigeria”. Food Policy 35 (4): 303-311.
24
Blanchflower, D.G., P. Levine and D. Zimmerman. 2003. “Discrimination in the small business credit
market.” Review of Economics and Statistics 85(4): 930-943.
Bruhm, Miriam. 2009 “Female-Owned Firms in Latin America: Characteristics, Performance, and Obstacles
to Growth.” World Bank Policy Research Working Paper No. 5122.
Buvinić, Mayra and Geeta Rao Gupta. 1997. “Female-Headed Households and Female-Maintained Families:
Are They Worth Targeting to Reduce Poverty in Developing Countries?” Economic Development and Cultural
Change 45 (2): 259-280.
Canagarajah, S., C. Newman, and R. Bhattamishra. 2001. “Non-farm Income, Gender and Inequality:
Evidence from Rural Ghana and Uganda.” Food Policy 26(4): 405–20.
Carter, Sara. 2000. “Improving the numbers and performance of women-owned businesses: some
implications for training and advisory services.” Education + Training 42(4/5): 326 – 334.
Cavalluzzo, Ken S. and John Wolken. 2002. “Small Business Loan Turndowns, Personal Wealth and
Discrimination.” Finance and Economics Discussion Series 2002-35. Washington, D.C., Board of Governors
of the Federal Reserve System, August.
de Janvry, Alain and Elisabeth Sadoulet. 2001. “Income Strategies Among Rural Households in Mexico: The
Role of Off-farm Activities.” World Development 29(3): 467-480.
de Mel, Suresh, David McKenzie and Christopher Woodruff. 2009. "Are Women More Credit Constrained?
Experimental Evidence on Gender and Microenterprise Returns," American Economic Journal: Applied Economics
1(3): 1-32
de Mel, Suresh, David McKenzie and Christopher Woodruff. 2008. "Returns to Capital in Microenterprises:
Evidence from a Field Experiment”. The Quarterly Journal of Economics 123(4): 1329-1372.
Deininger, Klaus, J. Songqing, and Mona Sur. 2007. "Sri Lanka's Rural Non-Farm Economy: Removing
Constraints to Pro-Poor Growth," World Development, 35(12): 2056-2078.
Duflo, Esther. 2000."Grandmothers and Granddaughters: Old Age Pension and Intra-household Allocation
in South Africa." NBER Working Papers 8061, National Bureau of Economic Research, Inc.
Easterly, William. 2002. “The Cartel of Good Intentions: Bureaucracy versus markets in foreign aid‟, Center
for Global Development Working Paper 4.
Goldstein, Markus and Christopher Udry. 2008. “The profits of power: land rights and agricultural
investment in Ghana”. Journal of Political Economy. 116(6): 981--1022.
Haddad, Lawrence J., John Hoddinott and Harold Alderman. 1997. Intra-Household Resource Allocation in
Developing Countries. Johns Hopkins Univ. Press, Baltimore, MD (1997).
Haggblade, S., Peter H. Hazell and Thomas Reardon. 2007. Transforming the rural nonfarm economy.
Baltimore: Johns Hopkins University Press and International Food Policy Research Institute.
Haggblade, S., Peter Hazell and Thomas Reardon. 2010. “The Rural Non-farm Economy: Prospects for
Growth and Poverty Reduction.” World Development 38(10): 1429–1441.
25
Hallward-Driemeier, Mary and Reyes Aterido. 2009a. “Whose Business Is It Anyway?” World Bank Mimeo.
Hallward-Driemeier, Mary and Aterido, Reyes, 2009b. "Comparing Apples with....Apples : how to make
(more) sense of subjective rankings of constraints to business," Policy Research Working Paper Series No
5054.
Hallward-Driemeier. "Expanding Opportunities for Women Entrepreneurs in Africa." World Bank
forthcoming.
Headey, D., D. Bezemer and P.B. Hazell. 2010. “Agricultural Employment Trends in Asia and Africa: Too
Fast or Too Slow?” World Bank Research Observer 25(1): 57-89.
Horrell, Sara and Pramila Krishnan, 2007. “Poverty and Productivity in Female-Headed Households in
Zimbabwe.” Journal of Development Studies 43 (8): 1351-1380.
ILO. 2010. Report on gender equity in rural areas.
Jamison, Dean and Lawrence J. Lau. 1982. “Farmer education and farm efficiency” World Bank Research
Publication. Published for the World Bank by Johns Hopkins University Press.
Katz, Elizabeth and Juan Sebastian Chamorro. 2002. “Gender, land rights, and the household economy in
rural Nicaragua and Honduras.” Paper prepared for USAID/BASIS CRSP. Madison, Wisconsin.
Lanjouw, Jean and Peter Lanjouw. 2001. “The rural non-farm sector: Issues and evidence from developing
countries.” Agricultural Economics 26 (1): 1–23.
Lanjouw, Peter. 2001. “Nonfarm Employment and Poverty in Rural El Salvador.” World Development 29(3):
529-547.
Lustig, Nora. 2000. “Crises and the Poor: Socially Responsible Macroeconomics.” Economia, 1(1): 1-19.
Loening, J. L., B.Rijkers and M. Söderbom. 2008. “Non-farm microenterprises and the investment climate:
Evidence from rural Ethiopia.” World Bank Policy Research working paper 4577.
Mammen, Kristin, and Christina Paxson. 2000. "Women's Work and Economic Development." Journal of
Economic Perspectives, 14(4): 141–164
McCulloch, N., J. Weisbrod and P. Timmer. 2007. “The pathways out of rural poverty in Indonesia.” World
Bank Policy Research working paper 4173.
McKenzie, David and Christopher Woodruff. 2008. “Experimental Evidence on Returns to Capital and
Access to Finance in Mexico.” World Bank Economic Review 22 (3): 457-482.
Mengistae, T. and M. Honorati. 2009. “Constraints to productivity growth and to job creation in the private
sector.” In Ethiopia investment climate assessment. Washington, DC: World Bank.
Nichter, Simeon and Lara Goldmark. 2009. “Small Firm Growth in Developing Countries” World Development
37 (September 2009): 1453–1464.
26
Parker, Simon C., 2008. "Entrepreneurship among married couples in the United States: A simultaneous
probit approach," Labour Economics, 15(3): 459-481.
Parker, Simon. 2009. The Economics of Entrepreneurship. Cambridge University Press
Priebe, J., R. Rudolf, S. Klasen, J, Weisbrod, I. Sugema, and N. Nuryartono. 2009. “Rural income dynamics in
post-crisis Indonesia.” Paper presented at the Inaugural conference of courant research center – Poverty,
equity and growth in developing and transition countries. Göttingen, Germany.
Quisumbing, Agnes R. and John A. Maluccio. 2000. “Intrahousehold allocation and gender relations: New
empirical evidence from four developing countries.” IFPRI FCND Discussion Paper No. 84.
Rijkers, Bob, Mans Soderbom and Josef L. Loening. 2010. “A Rural-Urban Comparison of Manufacturing
Enterprise Performance in Ethiopia.” World Development 38(9): 1278-1296.
Sabarwal,Shwetlena and Katherine Terrell. 2008. “Does Gender Matter for Firm Performance? Evidence
from Eastern Europe and Cetnral Asia” World Bank Policy Research Working Paper no 4705.
Schady, Norbert, and J. Rosero. 2007. “Do Cash Transfers to Women Affect the Composition of
Expenditures? Evidence on Food Engel Curves in Rural Ecuador.” World Bank Policy Research Paper 4282.
Shiferaw, Admasu. 2009. "Survival of Private Sector Manufacturing Establishments in Africa: The Role of
Productivity and Ownership “World Development. 37(3): 572-584.
Songqing Jin and Klaus Deininger. 2009. "Key Constraints for Rural Non-Farm Activity in Tanzania:
Combining Investment Climate and Household Surveys." Journal of African Economies 18(2): 319-361.
Storey, D. J. 2004. “Racial and Gender Discrimination in the Micro Firms Credit Market?: Evidence from
Trinidad and Tobago.” Small Business Economics 23(5): 401-422.
Thomas, Duncan. 1990. “Intra-Household Resource Allocation: An Inferential Approach.” The Journal of
Human Resources 25(4): 635-664.
Udry, Christopher. 1996. “Gender, Agricultural Production, and the Theory of the Household”. The Journal of
Political Economy 104(5): 1010-1046.
World Bank. 2005. “A Better Investment Climate for Everyone” World Development Report 2005. Washington
DC: World Bank.
World Bank. 2004. Investment Climate Assessment Sri Lanka: Improving the Rural and Urban Investment Climate. Zwede and Associates. 2002. “Jobs, Gender and Small Enterprises in Africa: Preliminary Report, Women
Entrepreneurs in Ethiopia.” ILO Office, Addis Ababa in association with SEED, International Labour
Office, Geneva, October.
27
Tables and Graphs
Figure 1: Income Portfolios vs Household Expenditure
1A: Ethiopia-male headed households 1B Ethiopia-female headed households
1C: Indonesia-male headed households 1B Indonesia-female headed households
-1-.
50
.51
Inco
me
Sh
are
2 4 6 8 10Log of per capita consumption
CI Farm CI NFE
NFE lpoly smooth: shincnfe
CI Wage Wage
CI Other Other
Male Headed Households
Income portfolios vs log C
0.5
11
.5
Inco
me
Sh
are
5 6 7 8 9 10Log of per capita consumption
CI Farm CI NFE
NFE lpoly smooth: shincnfe
CI Wage Wage
CI Other Other
Female Headed Households
Income portfolios vs log C
-.5
0.5
11
.5
Inco
me
Sh
are
6 8 10 12 14Log of per capita consumption
CI Farm Farm
CI NFE NFE
CI Wage Wage
CI Other Other
Male Headed Households
Income portfolios vs log C
-.5
0.5
1
Inco
me
Sh
are
6 7 8 9 10 11Log of per capita consumption
CI Farm Farm
CI NFE NFE
CI Wage Wage
CI Other Other
Female Headed Households
Income portfolios vs log C
28
Figure 2: Size Distributions
Notes: - Kernel density estimates, estimated using the epanechnikov kernel, - L measured in full time equivalent workers -the measure of labour in Sri Lanka is discrete, which explains why the size distribution appears bimodal.
0.2
.4.6
.8
kd
en
sity ln
Lft
-2 0 2 4 6Ln L
Female Firms Male Firms
Size Distribution - Bangladesh
0.2
.4.6
kd
en
sity ln
Lft
-4 -2 0 2 4Ln L
Female Firms Male Firms
Size Distribution - Ethiopia
0.2
.4.6
kd
en
sity ln
Lft
-2 0 2 4 6Ln L
Female Firms Male Firms
Size Distribution - Indonesia
0.2
.4.6
.81
kd
en
sity ln
Lft
0 2 4 6Ln L
Female Firms Male Firms
Size Distribution - Sri Lanka
29
Figure 3: Productivity Distributions
Notes: - Kernel density estimates, estimated using the epanechnikov kernel - L measured in full time equivalent workers, Y in USD equivalents.
0.1
.2.3
.4
kd
en
sity
lnY
Ld
ft
0 5 10 15Ln Y/L
Female Firms Male Firms
Productivity Distribution - Bangladesh
0.1
.2.3
.4
kd
en
sity ln
YL
dft
0 5 10Ln Y/L
Female Firms Male Firms
Productivity Distribution - Ethiopia
0.1
.2.3
kd
en
sity ln
YL
dft
0 5 10 15Ln Y/L
Female Firms Male Firms
Productivity Distribution - Indonesia
0.1
.2.3
kd
en
sity ln
YL
dft
0 5 10 15Ln Y/L
Female Firms Male Firms
Productivity Distribution - Sri Lanka
30
Table 1: The Structure of the Labor Market
Table 1A: Structure of the labor market – Bangladesh, Sri Lanka, Ethiopia
Bangladesh Ethiopia Indonesia Sri Lanka Women Men Women Men Women Men Women Men OLF/Not working (i) 68.24 % 4.85 % 25.69% 7.26% NA NA 50.74% 12.52% NFE only (ii) 5.46% 23.34% 5.56% 5.12% 8.78% 14.29% 19.09% 18.72% NFE + Agriculture (iii) 2.81% 19.24% 3.66% 4.18% NA NA 4.21% 12.00% NFE + Wage(iv) 0.21% 6.50% 0.31% 0.20% 1.10% 2.66% 1.17% 4.22% Ag only (v) 17.87% 14.51% 60.37% 70.30% NA NA 8.67% 11.01% Ag+wage (vi) 0.86% 8.05% 1.73% 7.77% NA NA 1.36% 8.17% Wage only (vii) 4.37 % 19.79% 2.68% 5.17% 23.05% 44.53% 14.17% 31.38% Wage + Ag + NFE (viii) 0.12% 3.70% 0 0 NA NA 0.39% 1.98%
Total NFE(sum ii,iii,iv, viii) 8.60% 52.78% 9.53% 9.50% 9.88% 16.35% 24.86% 36.92%
Total Ag (sum iii,v vi,, viii) 21.66% 45.50% 65.76% 82.25% NA NA 14.63% 33.16% Total wage sum(iv,vi, vii, viii) 5.56% 38.04% 4.72% 13.14% 24.45% 47.19% 17.09% 45.75% Total multiple activities (sum iii, iv, vi, viii)
4.00% 37.49% 5.70% 12.15% NA NA 7.13% 26.37%
Note statistics are weighted in Bangladesh, Ethiopia and Indonesia, but not in Sri Lanka – since weights were not available for that country. Ag=working on the family farm, Wage=wage employment (a composite category covering both agricultural and non-agricultural wage employment) NFE=non-farm enterprise activity; working in a household non-farm enterprise. The categories presented in this table were constructed on the basis of different questions (please refer to variable appendix for country specifics). To distract comparable indicators from these questions, we constructed employment categories sorting people on the basis of the activities they reported engaging in without drawing a distinction between primary and secondary employment.
Table 1B: Household-level participation Rates in NFE
Household- level Participation Rates
Overall Male-headed Female-headed % of hh that are female-headed
Bangladesh 27.62% 28.61% 16.79% 8.32% Ethiopia 16.20% 12.13% 31.08% 21.46% Indonesia 23.32% 24.69% 15.23% 14.45% Sri Lanka 47.15% 48.80% 37.95% 14.78%
% of households that work exclusively rely on non-farm enterprise
Overall Male-headed Female-headed
Bangladesh 13.66% 13.93% 10.07% Ethiopia 4.86% 2.12% 15.83% Sri Lanka 13.72% 13.80% 13.25%
% of households whose only source of income is non-farm enterprise income
Overall Male-headed Female-headed
Ethiopia 4.47% 2.11% 13.11% Indonesia 2.70% 2.82% 1.99%
Note statistics are weighted in Bangladesh, Ethiopia and Indonesia, but not in Sri Lanka – since household weights were not available for that country
31
Table 2: Individual Characteristics
Indonesia Ethiopia Bangladesh Sri Lanka
Women
Women Men Women Men Women Men
Mean
Std. Dev.
Mean
Std. Dev.
Mean
Std. Dev.
Mean
Std. Dev.
Mean
Std. Dev.
Mean
Std. Dev.
Mean
Std. Dev.
Mean
Std. Dev. NFE 9.7
6% 0.30 16.
72%
0.37 9.26%
0.29 9.23%
0.29 8.52%
0.28 53.24%
0.50 25.68%
0.44 36.81%
0.48 ag
66.13%
0.47 82.59%
0.38 22.49%
0.42 46.92%
0.50 14.79%
0.36 34.22%
0.47 wage 24.
17%
0.43 47.14%
0.50 4.62%
0.21 13.29%
0.34 5.63%
0.23 38.11%
0.49 17.88%
0.38 44.96%
0.50 age 35.
64 13.1
5 35.05
12.73
32.25
12.49
32.46
12.89
35.04
12.70
34.78
12.64
36.94
12.29
37.10
12.78 primary 39.
90%
0.49 43.28%
0.50 9.26%
0.29 29.78%
0.46 29.85%
0.46 37.24%
0.48 18.69%
0.39 18.30%
0.39 secondary 34.
02%
0.47 37.78%
0.48 3.93%
0.19 6.78%
0.25 30.63%
0.46 33.50%
0.47 74.91%
0.43 77.85%
0.42 tertiary 4.4
4% 0.21 6.9
0% 0.25
1.07%
0.10 5.12%
0.22 1.62%
0.13 1.70%
0.13 head 8.4
4% 0.28 55.
44%
0.50 16.60%
0.37 63.82%
0.48 4.89%
0.22 55.71%
0.50 8.17%
0.27 55.85%
0.50 spouse 59.
52%
0.49 0.13%
0.04 62.78%
0.48 0.92%
0.10 61.02%
0.49 0.70%
0.08 56.44%
0.50 1.70%
0.13 child 26.
39%
0.44 39.05%
0.49 15.98%
0.37 27.62%
0.45 26.03%
0.44 38.81%
0.49 31.20%
0.46 40.07%
0.49 married 71.
99%
0.45 63.83%
0.48 67.91%
0.47 65.67%
0.47 79.41%
0.40 71.62%
0.45 divorcee 2.4
7% 0.16 4.7
0% 0.21 13.
07%
0.34 4.27%
0.20 1.86%
0.13 0.30%
0.06 widow 8.0
9% 0.27 0.4
0% 0.06 7.0
3% 0.26 0.7
8% 0.09 9.6
9% 0.30 0.1
7% 0.04
ln hh size 1.45
0.44 1.49
0.44 1.51
0.51 1.61
0.45 1.61
0.42 1.65
0.38 1.51
0.37 1.52
0.35 sh child 15.
97%
0.17 16.43%
0.17 27.23%
0.20 27.24%
0.20 19.43%
0.18 18.46%
0.18 12.90%
0.17 12.31%
0.17 sh elderly 5.0
3% 0.12 3.5
5% 0.10 2.6
7% 0.08 1.9
5% 0.07 4.4
6% 0.10 3.2
0% 0.08 4.0
6% 0.10 4.0
2% 0.10
ln land agric
0.27
0.45 0.27
0.46
2.67
2.43 2.59
2.36 0.24
0.33 0.24
0.33 own plough
76.35%
0.43 85.15%
0.36 own house 90.
16%
0.30 91.11%
0.28 91.78%
0.27 93.95%
0.24 92.69%
0.26 92.52%
0.26 91.54%
0.28 91.11%
0.28 electricity 95.
89%
0.20 96.13%
0.19 4.94%
0.22 3.79%
0.19 87.75%
0.33 90.02%
0.30 95.88%
0.20 95.41%
0.21 town 31.
94%
0.47 30.00%
0.46 5.78%
0.23 4.78%
0.21 11.10%
0.31 12.65%
0.33 lndistmark
et 1.14
1.02 1.10
1.03 2.04
0.81 2.08
0.78 79.87%
0.57 81.19%
0.57 1.93
0.90 1.96
0.90 bank 18.
80%
0.39 19.27%
0.39 33.57%
0.47 33.08%
0.47 24.16%
0.43 23.12%
0.42 52.10%
0.50 49.93%
0.50 localwageUSD
6.23
0.48 6.22
0.49 5.56
0.41 5.57
0.41 6.00
0.25 6.02
0.25 6.69
0.24 6.69
0.24 partner Primary
27.27%
0.45 26.35%
0.44 15.75%
0.36 3.80%
0.19 21.96%
0.41 19.89%
0.40 14.72%
0.35 12.52%
0.33 partner Secondary
16.24%
0.37 13.25%
0.34 1.36%
0.12 0.56%
0.07 17.29%
0.38 12.00%
0.33 40.99%
0.49 39.70%
0.49 partner Tertiary
3.94%
0.19 2.39%
0.15
1.73%
0.13 0.73%
0.09 1.03%
0.10 0.89%
0.09 hh head father Ag
56.39%
0.50 55.96%
0.50
64.83%
0.48 63.41%
0.48 43.64%
0.50 43.56%
0.50 hh head father NFE
9.98%
0.30 9.69%
0.30
16.65%
0.37 18.21%
0.39 26.12%
0.44 24.59%
0.43 hh head father wage
30.68%
0.46 32.10%
0.47
18.25%
0.39 18.11%
0.39 25.75%
0.44 26.96%
0.44
hh head father other
2.35%
0.15 2.74%
0.16
32
Table 3: Descriptive statistics households
Ethiopia Indonesia
Women Men Women
Women
Men Mean Std.
Dev. Mean Mean Mean Std. Dev. Mean Std. Dev.
NFE 31.08% 0.46 12.13% 0.33 15.23% 0.36
24.69% 0.43 sh inc nfe 19.16% 0.36 5.88% 0.21 18.82% 0.30
20.57% 0.34
sh inc wage 9.52% 0.27 6.10% 0.20 29.52% 0.34
38.82% 0.37 sh inc other 7.60% 0.17 3.20% 0.12 28.38% 0.29
19.16% 0.24 sh inc farm 63.72% 0.41 84.82% 0.30 23.28% 0.31
21.45% 0.31
# inc sources 1.46 0.67 1.29 0.54 2.34 0.68
2.26 0.75 only NFE inc 13.11% 0.34 2.11% 0.14 1.99% 0.14
2.82% 0.17
only agric inc 44.47% 0.50 70.51% 0.46 0.48% 0.07
1.73% 0.13 total expend pc 152.37 123.96 133.90 99.83 268.46 172.95
328.21 820.17 headage 48.80 15.92 42.59 14.80 56.12 13.11
45.38 13.81 headage2 26.34 16.28 20.33 14.37 33.20 14.62
22.50 14.03 head primary 4.47% 0.21 24.84% 0.43 39.37% 0.49
44.85% 0.50
head secondary 2.91% 0.17 2.47% 0.16 5.52% 0.23
30.27% 0.46 head tertiary 0.68% 0.08
3.79% 0.19 head married 12.70% 0.33 93.93% 0.24 8.43% 0.28
94.81% 0.22 ln hh size 0.96 0.60 1.53 0.45 0.85 0.59
1.35 0.46 sh child 18.48% 0.22 30.66% 0.20 7.93% 0.15
18.01% 0.19 sh women 49.86% 0.31 29.33% 0.14 70.06% 0.25
46.35% 0.17 sh elderly 10.79% 0.26 3.55% 0.12 18.45% 0.32
7.23% 0.18 ln land agric 0.19 0.33
0.27 0.46 own plough 33.54% 0.47 86.58% 0.34
own house 80.94% 0.39 94.77% 0.22 92.19% 0.27
89.05% 0.31 town 10.18% 0.30 3.91% 0.19 31.78% 0.47
32.07% 0.47
lndistmarket 1.83 0.95 2.09 0.75 1.14 0.98
1.09 1.04 bank 38.16% 0.49 33.67% 0.47 11.82% 0.32
21.09% 0.41 localwageUSD 0.22 0.04 0.23 0.04 6.15 0.51 6.22 0.48
Note: Statistics are weighted. Bolded coefficients numbers indicate the gender differences are statistically significant at the 5%
level.
33
Table 4A: Modeling Selection
NFE Agriculture Wage
Bangladesh Ethiopia Sri Lanka Bangladesh Ethiopia Sri Lanka Bangladesh Ethiopia Sri Lanka
women men women men women men women men women men women men women men women men women men
coef/se coef/se coef/se coef/se coef/se coef/se coef/se coef/se coef/se coef/se coef/se coef/se coef/se coef/se coef/se coef/se coef/se coef/se age 0.086** -0.012 0.065*** 0.030 0.068*** 0.030 0.113*** 0.006 -0.003 -0.041 0.100*** 0.041 0.060 -0.029 0.116*** 0.051** 0.088*** 0.132***
(0.040) (0.044) (0.024) (0.028) (0.024) (0.022) (0.043) (0.043) (0.018) (0.027) (0.029) (0.027) (0.042) (0.040) (0.026) (0.024) (0.029) (0.023)
agesq -0.001*** 0.000 -0.078*** -0.046 -0.001*** -0.000 -0.002*** -0.000 -0.018 0.048 -0.001*** -0.000 -0.001 0.000 -0.168*** -0.059** -0.001*** -0.002***
(0.001) (0.001) (0.029) (0.034) (0.000) (0.000) (0.001) (0.001) (0.023) (0.033) (0.000) (0.000) (0.001) (0.000) (0.035) (0.029) (0.000) (0.000)
primary 0.116 0.261 -0.197 0.243** 0.281 0.096 0.350* -0.236 0.259* 0.018 0.342 0.057 0.154 0.146 -0.108 0.061 0.114 0.464*
(0.161) (0.192) (0.160) (0.096) (0.200) (0.292) (0.187) (0.193) (0.134) (0.107) (0.250) (0.288) (0.273) (0.195) (0.184) (0.087) (0.227) (0.254)
secondary 0.089 0.469*** -0.583** 0.313 0.145 0.576** 0.126 -0.492** -0.858*** -0.714*** 0.168 -0.186 -0.009 -0.210 1.031*** 0.315* 0.218 0.390
(0.196) (0.176) (0.285) (0.217) (0.193) (0.284) (0.237) (0.196) (0.309) (0.172) (0.246) (0.281) (0.289) (0.152) (0.249) (0.176) (0.219) (0.244)
tertiary -0.201 0.406 -0.133 -0.277 -2.551*** -1.760*** -0.059 -0.946** 1.783*** 0.960*** 1.278*** 1.395***
(0.593) (0.346) (0.376) (0.424) (0.475) (0.386) (0.556) (0.440) (0.428) (0.337) (0.365) (0.387)
head 1.016*** 0.288 0.552** 0.750*** 0.725*** 0.378 0.360 0.402 0.340* 1.681*** 1.312*** 0.634* 0.392 -0.359 -0.111 -1.229*** -0.231 0.343
(0.335) (0.257) (0.237) (0.270) (0.244) (0.292) (0.357) (0.263) (0.199) (0.213) (0.375) (0.350) (0.443) (0.271) (0.259) (0.213) (0.252) (0.276)
spouse 0.313 -0.402 -0.045 -0.761 0.531* -1.185** 0.434 -1.353*** 0.155 1.306*** 0.718* -0.781 -0.431 -0.342 -0.920*** -1.710*** -0.137 1.512***
(0.354) (0.469) (0.237) (0.519) (0.295) (0.486) (0.320) (0.493) (0.179) (0.461) (0.401) (0.498) (0.434) (0.460) (0.235) (0.395) (0.322) (0.432)
child 0.304 0.033 -0.159 -0.088 0.141 -0.110 0.022 -0.605** 0.041 1.657*** 0.521 0.119 0.353 -0.325 -0.141 -1.549*** 0.013 0.436*
(0.341) (0.201) (0.214) (0.272) (0.235) (0.262) (0.444) (0.272) (0.181) (0.161) (0.381) (0.312) (0.415) (0.230) (0.229) (0.166) (0.274) (0.252)
currently married -0.531* 0.111 -0.324 0.095
-1.062*** 0.382* 0.249 0.316
-0.687** 0.211 0.180 -0.163
(0.280) (0.195) (0.215) (0.189)
(0.304) (0.213) (0.167) (0.194)
(0.329) (0.192) (0.202) (0.156)
divorcee -0.102 -6.715*** 0.218 -0.105
-0.934 -6.293*** 0.329** 0.001
-0.769 6.339*** -0.001 0.024
(0.420) (0.482) (0.193) (0.245)
(0.655) (0.446) (0.160) (0.200)
(0.521) (0.501) (0.172) (0.203)
widowed 0.255 -0.728 -0.298 0.580*
0.067 -1.379 0.769*** -0.027
0.145 -1.837*** 0.113 -1.922***
(0.536) (0.633) (0.228) (0.345)
(0.430) (0.968) (0.202) (0.469)
(0.531) (0.446) (0.246) (0.441)
ln hh size -0.080 -0.064 -0.005 0.412*** 0.046 0.175 -0.223 0.145 0.036 -0.203 -0.064 -0.431*** -0.076 -0.175 0.059 -0.197 -0.125 -0.073
(0.213) (0.187) (0.109) (0.124) (0.129) (0.131) (0.225) (0.212) (0.107) (0.136) (0.132) (0.133) (0.285) (0.183) (0.136) (0.125) (0.129) (0.118)
sh child 0.243 0.067 0.222 -0.535** -0.486* 0.118 -0.152 0.488 -0.492** 0.428 0.601** 0.382 -1.387** -0.285 0.027 0.429* -0.902*** 0.048
(0.415) (0.423) (0.256) (0.259) (0.271) (0.273) (0.566) (0.396) (0.220) (0.282) (0.285) (0.285) (0.609) (0.405) (0.280) (0.249) (0.297) (0.258)
sh elderly 0.557 0.400 0.264 0.773 -0.782 -0.314 0.923 1.757* -0.379 1.748** 0.931* 0.266 -0.841 -0.933 -0.348 -0.756 -0.347 0.235
(0.896) (0.732) (0.567) (0.752) (0.481) (0.422) (0.773) (0.970) (0.394) (0.753) (0.498) (0.432) (1.159) (0.719) (0.637) (0.654) (0.470) (0.421)
ln land agric -0.058** -0.007 -0.270* -0.062 0.034 0.275*** 0.546*** 0.619*** -0.100** -0.102*** -0.034 -0.322***
(0.029) (0.028) (0.162) (0.117) (0.037) (0.033) (0.136) (0.123) (0.042) (0.028) (0.135) (0.119)
own plough
-0.699*** -0.551***
0.526*** 1.044***
-0.557*** -0.304**
(0.121) (0.131)
(0.098) (0.116)
(0.137) (0.129)
own house 0.182 -0.173 -0.150 0.094 0.400** 0.362** -0.048 0.825*** 0.464*** 0.004 0.394* 0.711*** -0.658* -0.098 -0.120 0.219 -0.883*** -0.231*
(0.227) (0.218) (0.146) (0.162) (0.173) (0.152) (0.280) (0.233) (0.146) (0.148) (0.204) (0.175) (0.353) (0.209) (0.144) (0.189) (0.144) (0.140)
uses electricity -0.110 0.182 0.065 -0.244* 0.234 0.342* 0.509** 0.043 -0.514** -0.593*** -0.494** -0.516** 0.431 -0.331* 0.110 -0.087 0.161 0.196
(0.211) (0.194) (0.140) (0.146) (0.240)
(0.198) (0.240) (0.213) (0.204) (0.217) (0.226) (0.201) (0.412) (0.195) (0.264) (0.151) (0.221) (0.190) town 0.483*** 0.832*** 0.459** 0.582***
0.336 -0.450*** -1.653*** -0.570***
-0.244 -0.787*** -0.094 0.466**
(0.168) (0.141) (0.181) (0.190)
(0.231) (0.145) (0.179) (0.194)
(0.268) (0.161) (0.231) (0.195) lndistmarket 0.045 -0.079 -0.294*** -0.269*** -0.046 0.027 0.068 -0.198* 0.017 0.278*** -0.133** 0.167*** 0.086 -0.067 0.064 -0.004 0.092* -0.113**
(0.112) (0.097) (0.071) (0.066) (0.050) (0.050) (0.122) (0.113) (0.052) (0.062) (0.056) (0.049) (0.247) (0.096) (0.063) (0.051) (0.049) (0.046)
bank 0.052 0.376** 0.099 -0.113 0.183** 0.280*** 0.046 -0.030 0.035 -0.007 -0.265*** -0.209** -0.401 -0.226* 0.113 -0.050 0.054 -0.028
(0.180) (0.147) (0.099) (0.098) (0.085) (0.081) (0.211) (0.135) (0.070) (0.100) (0.095) (0.087) (0.253) (0.137) (0.103) (0.079) (0.088) (0.077)
localwage -0.488* -0.430* -0.128 -0.101 0.195 0.232 -0.802** -0.873*** -0.202** 0.024 -0.095 -0.440** -0.851* 0.009 -0.052 0.155* 0.118 -0.235
(0.260) (0.251) (0.102) (0.106) (0.191) (0.183) (0.329) (0.262) (0.093) (0.104) (0.219) (0.196) (0.504) (0.237) (0.101) (0.094) (0.203) (0.170)
partner primary 0.074 -0.219 0.403*** 0.050 -0.202 0.209 -0.158 -0.094 0.015 -0.034 0.052 0.007 0.736* -0.035 0.108 0.029 -0.340 0.169
(0.180) (0.217) (0.133) (0.169) (0.226) (0.197) (0.206) (0.183) (0.089) (0.177) (0.252) (0.209) (0.428) (0.242) (0.157) (0.150) (0.278) (0.197)
34
partner secondary 0.073 -0.316 0.205 -0.323 -0.052 0.279 0.122 -0.172 -0.148 0.057 0.095 -0.002 0.916** 0.228 0.109 0.685* -0.265 -0.186
(0.247) (0.203) (0.390) (0.412) (0.217) (0.182) (0.252) (0.206) (0.244) (0.252) (0.240) (0.194) (0.452) (0.207) (0.287) (0.392) (0.270) (0.183)
partner tertiary 0.217 -1.418** -0.652 0.567 -0.837 0.497 -0.443 0.205 1.260** 0.376 0.380 -0.278
(0.449) (0.679) (0.479) (0.501) (0.524) (0.559) (0.709) (0.430) (0.530) (0.507) (0.488) (0.404)
hh head father NFE 0.194 0.178 0.124 0.299*** -0.336 -0.193 -0.546*** -0.585*** 0.047 0.026 -0.017 -0.045
(0.151) (0.171) (0.103) (0.102) (0.217) (0.177) (0.133) (0.107) (0.260) (0.162) (0.112) (0.096) hh head father Wage
-0.146 -0.211 0.040 0.037 -0.558** -0.452*** -0.321*** -0.498*** 0.737*** 0.827*** 0.241** 0.241** (0.175) (0.149) (0.109) (0.102) (0.235) (0.173) (0.115) (0.105) (0.262) (0.181) (0.106) (0.096)
hh head father Other
0.053 -0.386
-0.525 -0.333
0.057 0.245
(0.299) (0.283)
(0.353) (0.352)
(0.253) (0.237) _cons -0.327 1.776 -1.293** -1.860*** -5.096** -5.556*** 1.291 2.384 0.054 -1.061** -2.349 3.415 2.052 1.634 -2.791*** -0.823* -2.863 0.097
(1.337) (1.329) (0.509) (0.566) (2.154) (2.064) (1.711) (1.531) (0.430) (0.500) (2.543) (2.220) (2.479) (1.290) (0.462) (0.457) (2.262) (1.914)
/atrho21 0.047 -0.015 -0.138*** -0.708*** 0.134** 0.053
(0.090) (0.070) (0.050) (0.069) (0.056) (0.050)
/atrho31 -0.063 -0.680*** -0.324*** -0.253*** -0.345*** -0.823***
(0.101) (0.112) (0.065) (0.048) (0.059) (0.063)
/atrho32 -0.227* -0.218** -0.189*** -0.178*** -0.021 -0.243***
(0.119) (0.102) (0.061) (0.050) (0.058) (0.051)
Observations 3,515 4,158 2,777 2,232 1,359 1,350
Logpseudolikelihood -2556690 -4676842 -2532 -1992 -1774.19 -2160.4
Wald chi2 411.88 2754.38 1171.76 886.95 355.30 681.97
35
Table 4B: Indonesia bivariate probit
NFE Work
women men women men
coef/se coef/se coef/se coef/se
age 0.123*** 0.015
0.089** 0.126***
(0.033) (0.034)
(0.041) (0.031)
agesq -0.001*** -0.000
-0.001** -0.002***
(0.000) (0.000)
(0.001) (0.000)
primary -0.326* -0.324*
0.023 0.020
(0.171) (0.186)
(0.203) (0.184)
secondary -0.072 -0.187
0.039 -0.075
(0.172) (0.195)
(0.229) (0.217)
tertiary 0.103 -0.204
1.126*** 0.351
(0.268) (0.248)
(0.274) (0.241)
head 1.218*** 0.484*
-0.225 0.997***
(0.314) (0.265)
(0.298) (0.370)
spouse -0.163 -0.233
0.580* 0.802
(0.292) (0.518)
(0.332) (0.681)
child -0.038 0.130
-0.266 0.800***
(0.235) (0.268)
(0.295) (0.300)
currently married 0.339 0.657**
0.151 -0.149
(0.215) (0.311)
(0.229) (0.236)
divorcee -0.175 -0.379
1.059** 0.297
(0.418) (0.391)
(0.422) (0.367)
widowed -0.039 1.029**
0.531 -1.040*
(0.315) (0.508)
(0.363) (0.544)
ln hh size 0.219* 0.030
-0.140 -0.099
(0.120) (0.141)
(0.152) (0.178)
sh child -0.420 -0.165
-0.826** 0.445
(0.291) (0.390)
(0.386) (0.406)
sh elderly -0.784* -0.599
0.475 0.970*
(0.477) (0.583)
(0.590) (0.565)
ln land agric 0.068 -0.196*
-0.262** -0.451***
(0.092) (0.102)
(0.122) (0.115)
own house -0.074 -0.045
0.174 0.030
(0.138) (0.137)
(0.176) (0.154)
uses electricity 0.311 0.567**
-0.540* -0.171
(0.207) (0.241)
(0.299) (0.281)
town 0.251** 0.183
0.243* 0.091
(0.120) (0.116)
(0.126) (0.128)
lndistmarket -0.126** 0.034
-0.119* -0.232***
(0.052) (0.048)
(0.062) (0.064)
bank 0.080 0.265**
0.248* -0.120
(0.107) (0.124)
(0.129) (0.135)
localwage -0.046 -0.075
-0.193 0.114
(0.128) (0.141)
(0.135) (0.164)
partner primary 0.434** 0.013
-0.404* 0.101
(0.204) (0.170)
(0.244) (0.206)
partner secondary 0.404** -0.075
-0.620** -0.094
(0.204) (0.200)
(0.258) (0.217)
partner tertiary 0.348 0.013
-0.447 0.139
(0.272) (0.274)
(0.327) (0.300)
hh head father - NFE 0.274** 0.458***
-0.244 -0.179
(0.127) (0.153)
(0.170) (0.156)
hh head father - Wage -0.145 -0.220
0.599*** 0.484***
(0.156) (0.136)
(0.145) (0.152)
_cons -3.757* -1.443
1.000 -4.160
(1.963) (2.261) (2.188) (2.702)
/athrho -0.413*** -0.726***
(0.076) (0.090)
Number of observations 3,629 3609 Log pseudolikelihood -1062811 -1432344
Wald chi2 274.74(52) 395.52 (52)
note: *** p<0.01, ** p<0.05, * p<0.1
36
Table 5: Tobit Models of Share of Pecuniary Income accounted for by non-farm enterprise
income
Ethiopia Indonesia
inctobit inctobit
coef/se coef/se
Head is Male -0.143 -0.136**
(0.166) (0.058)
sh child -0.197 0.040
(0.252) (0.072)
ln hh size 0.486*** 0.108***
(0.117) (0.028)
sh women 0.628** -0.084
(0.256) (0.061)
sh elderly -0.437 -0.116
(0.444) (0.081)
2nd
expenditure quartile -0.058 0.058*
(0.107) (0.031)
3nd
expenditure quartile -0.005 -0.040
(0.109) (0.032)
top expenditure quartile -0.111 0.028
(0.119) (0.037)
head age 0.034* 0.010*
(0.019) (0.006)
headage2 -0.046** -0.009
(0.021) (0.006)
head primary 0.213** -0.012
(0.097) (0.029)
head secondary -0.636*** -0.008
(0.232) (0.036)
head tertiary -0.235***
(0.072)
head Married -0.150 0.083
(0.155) (0.052)
ln land agric -0.116***
(0.027)
own plough -0.969***
(0.113)
own house -0.280** -0.075**
(0.132) (0.038)
town 0.444*** 0.136***
(0.169) (0.024)
lndistmarket -0.295*** -0.010
(0.054) (0.011)
bank -0.083 -0.002
(0.088) (0.030)
localwage -0.283*** -0.164***
(0.102) (0.025)
_cons -0.244 1.453***
(0.500) (0.285)
/sigma 1.125*** 0.477***
(0.060) (0.009)
Number of observations 2,475 2,641 Pseudo R2 0.1397 0.0308
note: *** p<0.01, ** p<0.05, * p<0.1
37
Table 6: Descriptive Statistics – Enterprises
Bangladesh Ethiopia Indonesia Sri Lanka
Women
Men
Women Men Women
Men
Women
Men
Variable Mean SE Mean SE Mean SE Mean SE Mean SE Mean SE Mean SE Mean SE lnYd 5.42 1.26 8.16 1.48 4.52 1.78 5.45 1.83 7.74 1.48 8.10 1.66 6.35 1.54 7.27 1.61 lnLft -0.22 0.53 0.40 0.70 -0.42 0.74 -0.55 0.92 0.29 0.88 0.57 1.05 0.35 0.49 0.57 0.60 hiredL 8.47% 0.28 30.65% 0.46 1.95% 0.14 5.95% 0.24 25.27% 0.44 43.98% 0.50 33.75% 0.47 51.67% 0.50 sharepaid 4.54% 0.15 18.41% 0.30 1.46% 0.12 3.39% 0.15 22.53% 0.40 38.31% 0.46 18.75% 0.27 30.23% 0.31 1 worker 76.07% 0.43 32.26% 0.47 80.27% 0.40 72.97% 0.44 59.65% 0.49 44.27% 0.50 57.49% 0.50 37.62% 0.48 2workers 13.94% 0.35 37.72% 0.48 16.66% 0.37 21.57% 0.41 21.82% 0.41 24.46% 0.43 30.93% 0.46 38.58% 0.49 3-5 workers 9.53% 0.29 23.27% 0.42 2.69% 0.16 4.41% 0.21 15.10% 0.36 18.35% 0.39 8.69% 0.28 19.00% 0.39 5-50 workers 0.45% 0.07 6.31% 0.24 0.38% 0.06 0.40% 0.06 3.42% 0.18 12.59% 0.33 2.90% 0.17 4.54% 0.21 >50 workers 0.00% 0.00 0.44% 0.07 0.00% 0.00 0.66% 0.08 0.01% 0.01 0.33% 0.06 0.00% 0.00 0.26% 0.05 lnKd 4.98 1.35 5.32 1.74 4.58 2.68 4.14 2.68 6.64 2.55 6.91 2.79 6.72 1.69 7.25 1.71 lnYLdft 5.64 1.15 7.76 1.35 4.95 1.62 6.00 1.90 7.45 1.40 7.53 1.41 6.02 1.46 6.71 1.45 lnKLdft 5.21 1.40 4.92 1.60 4.92 2.53 4.50 2.65 6.34 2.58 6.34 2.87 6.37 1.67 6.68 1.61 lnMLdft 3.94 1.68 7.22 1.74 3.79 2.15 4.85 2.62 6.72 1.54 6.51 1.68 5.35 1.57 6.28 1.56 invest 20.05% 0.40 29.42% 0.46 16.27% 0.37 21.17% 0.41 48.34% 0.50 46.73% 0.50 59.89% 0.49 49.94% 0.50 lninv 0.75 0.97 3.10 1.63 1.89 1.66 2.81 1.77 5.81 5.27 6.75 0.19 4.53 1.55 5.53 1.60 dlnLft 0.01 0.07 0.03 0.22 5.79% 0.72 11.10% 0.53 0.08 0.29 0.10 0.40 0.01 0.09 0.00 0.11 shrank workers 0.06% 0.03 0.52% 0.07 2.53% 0.16 2.18% 0.15 1.05% 0.10 2.41% 0.15 0.31% 0.06 1.65% 0.13 growth workers 0.95% 0.10 6.38% 0.24 7.57% 0.26 10.43% 0.31 50.25% 0.50 55.02% 0.50 21.12% 0.41 13.54% 0.34 manufacturing 95.10% 0.22 27.84% 0.45 75.36% 0.43 46.34% 0.50 2.54% 0.16 13.51% 0.34 50.41% 0.50 37.76% 0.49 services 4.64% 0.21 23.62% 0.42 10.18% 0.30 12.41% 0.33 24.42% 0.43 39.58% 0.49 15.43% 0.36 23.35% 0.42 trade 0.26% 0.05 48.54% 0.50 14.47% 0.35 41.25% 0.49 73.03% 0.44 46.92% 0.50 34.16% 0.48 38.90% 0.49 hh ent 95.53% 0.21 20.61% 0.40 5.14% 0.22 3.92% 0.19 74.72% 0.44 73.50% 0.44 60.24% 0.49 34.46% 0.48 uses electricity 64.42% 0.48 75.39% 0.43 69.53% 0.46 64.11% 0.48 77.07% 0.42 73.27% 0.44 firm age 6.41 7.19 10.06 10.49 8.78 8.43 8.73 9.87 9.12 9.46 8.98 8.68 8.75 10.93 9.10 10.57 registered 2.90% 0.17 42.20% 0.49 2.30% 0.15 8.44% 0.28 2.96% 0.17 10.93% 0.31 35.38% 0.48 58.56% 0.49 months normal act 11.47 1.31 11.38 1.62 6.77 2.73 5.87 2.74 8.74 1.61 8.61 1.90
competition more20 1.47% 0.12 9.54% 0.29 9.74% 0.30 12.12% 0.33 29.51% 0.46 32.79% 0.47 competition 11to20 11.71% 0.32 12.05% 0.33 5.24% 0.22 3.99% 0.20 20.80% 0.41 8.48% 0.28 competition 5to10 30.16% 0.46 29.53% 0.46 4.99% 0.22 7.97% 0.27 19.95% 0.40 15.63% 0.36 competition less5 56.65% 0.50 48.88% 0.50 80.04% 0.40 75.92% 0.43 29.74% 0.46 43.10% 0.50 firm has loan 1.34% 0.12 10.15% 0.30 22.60% 0.42 22.26% 0.42 14.44% 0.35 13.32% 0.34 24.15% 0.43 28.58% 0.45
manager age 31.64 8.53 39.25 13.45 43.18 13.04 41.67 13.73 40.0 12.10 41.4 11.34 manager experience
3.25 5.32 5.60 7.52
manager primary 24.05% 0.43 35.69% 0.48 11.18% 0.32 42.17% 0.49 35.18% 0.48 31.55% 0.03 20.42% 0.40 11.26% 0.32 manager sec 35.87% 0.48 41.51% 0.49 2.89% 0.17 5.39% 0.23 41.66% 0.49 41.53% 0.03 76.51% 0.42 85.96% 0.35 manager tertiary 0.21% 0.05 4.09% 0.20 9.19% 0.29 16.52% 0.03 1.88% 0.14 2.51% 0.16 town 49.96% 0.50 45.72% 0.50 28.76% 0.45 21.09% 0.41 51.37% 0.50 51.40% 0.50
lndistmarket 60.51% 0.40 66.50% 0.50 1.16 1.00 1.64 0.93 0.66 0.97 0.93 1.12 1.51 0.92 1.81 1.14 bank 12.87% 0.34 42.46% 0.49 51.48% 0.50 59.89% 0.49 48.47% 0.50 44.29% 0.50 60.32% 0.49 69.53% 0.46 Localwage 6.15 0.21 6.07 0.24 5.52 0.40 5.55 0.40 6.50 0.45 6.46 0.43 6.69 0.23 6.70 0.22
Note: Statistics are weighted. The samples are confined to observations for whom the production function could be estimated. Bolded coefficients numbers indicate the gender differences are statistically significant at the 5% level.
38
Table 7: Production Functions
Production functions
Bangladesh Ethiopia Indonesia Sri Lanka
coef se coef se coef se coef se MODEL 1: Gender only manager is male 2.744*** (0.206)
0.929*** (0.232) 0.143 (0.197) 0.917*** (0.178) Observations
observations
2,480
648
1,369
1,259
R2 0.228
0.060
0.002
0.056
Adjusted R2 0.228
0.059
0.001
0.055
MODEL 2:Gender and Size manager is male 2.196*** (0.194) 0.998*** (0.226) -0.104 (0.187) 0.641*** (0.169) lnLft 0.875*** (0.094) 0.561*** (0.128) 0.833*** (0.081) 1.200*** (0.091) Observations
observations
2,480
647
1,369
1,259
R2 0.359
0.125
0.256
0.223
Adjusted R2 0.359
0.123
0.255
0.221
MODEL 3:Gender, Size and Sector manager is male 0.998*** (0.290) 0.632*** (0.206) -0.067 (0.277) 0.514*** (0.157) lnLft 1.252*** (0.105) 0.786*** (0.125) 0.790*** (0.088) 1.308*** (0.089) services 0.463* (0.258) 0.092 (0.364) -0.439* (0.222) 0.619*** (0.196) trade 1.809*** (0.237) 1.443*** (0.261) -0.012 (0.367) 1.020*** (0.132) Observations 2,480
647
1,369
1,259
R2 0.555
0.227
0.273
0.298
Adjusted R2 0.554
0.222
0.271
0.296
MODEL 4: Gender Size and Sector (+ K and M) manager is male -0.310** (0.124) 0.459*** (0.152) 0.076 (0.092) -0.010 (0.099) manuflnKd 0.074*** (0.027) 0.073* (0.042) 0.027 (0.017) -0.005 (0.052) manuflnLft 0.229*** (0.042) 0.591*** (0.166) 0.344*** (0.121) 0.348*** (0.090) manuflnMd 0.668*** (0.029) 0.384*** (0.075) 0.599*** (0.087) 0.746*** (0.054) servlnKd 0.082*** (0.013) 0.167** (0.069) 0.048*** (0.017) 0.106** (0.042) servlnLft 0.423*** (0.047) 0.642*** (0.237) 0.256*** (0.075) 0.272 (0.167) servlnMd 0.536*** (0.022) 0.261*** (0.071) 0.563*** (0.046) 0.640*** (0.056) tradelnKd 0.012 (0.009) 0.062 (0.038) 0.014 (0.013) 0.043* (0.024) TradelnLft 0.182*** (0.038) 0.559*** (0.116) 0.142*** (0.046) 0.235* (0.122) tradelnMd 0.847*** (0.023) 0.558*** (0.106) 0.752*** (0.051) 0.863*** (0.034) Services 0.839*** (0.281) 0.511 (0.533) 0.222 (0.793) 0.116 (0.532) Trade -1.097*** (0.301) -0.715 (0.710) -1.041 (0.770) -1.256** (0.529) hh ent -0.320*** (0.047)
-0.525*** (0.160) -0.143* (0.080)
F TESTS – CRS
CRS-Manufacturing
p>F 0.2321
0.7564
0.8018
0.1868
CRS-Services
p>F 0.3598
0.7991
0.0234
0.8967
CRS-Trade
p>F 0.2149
0.2266
0.0322
0.1812
Observations
observations
2,480
537
1,369
1,058
R2 0.957
0.487
0.789
0.806
Adjusted R2 0.957
0.475
0.787
0.804
MODEL 5: Full specification manager is male -0.242** (0.109) 0.606*** (0.161) 0.058 (0.072) -0.064 (0.111) manuflnKd 0.073*** (0.026) 0.064* (0.034) 0.012 (0.019) -0.019 (0.052) manuflnLft 0.224*** (0.042) 0.413*** (0.121) 0.367*** (0.116) 0.398*** (0.094) manuflnMd 0.654*** (0.029) 0.467*** (0.073) 0.552*** (0.091) 0.733*** (0.056) servlnKd 0.073*** (0.013) 0.117* (0.063) 0.041** (0.019) 0.123*** (0.041) servlnLft 0.383*** (0.040) 0.517 (0.443) 0.243*** (0.078) 0.289 (0.180) servlnMd 0.521*** (0.019) 0.261** (0.102) 0.547*** (0.039) 0.622*** (0.063) tradelnKd 0.000 (0.012) 0.054 (0.044) 0.009 (0.014) 0.025 (0.027) TradelnLft 0.166*** (0.040) 0.634*** (0.168) 0.072* (0.042) 0.139 (0.102) tradelnMd 0.814*** (0.029) 0.505*** (0.097) 0.762*** (0.036) 0.866*** (0.033) Services 0.912*** (0.283) 0.827* (0.484) -0.124 (0.768) 0.010 (0.603) Trade -0.888*** (0.311) 0.090 (0.668) -1.540** (0.690) -1.222* (0.660) hh ent -0.275*** (0.050)
-0.255*** (0.096) -0.169* (0.090)
manager age -0.002 (0.006) -0.004 (0.026) -0.004 (0.013) 0.013 (0.011) mngrage2_100 0.160 (0.617) -0.010 (0.024) -0.001 (0.015) -0.034 (0.032) manager primary 0.012 (0.044) -0.620*** (0.185) -0.112 (0.090) 0.090 (0.140) manager secondary 0.020 (0.050) -0.082 (0.396) 0.020 (0.136) 0.164 (0.167) manager tertiary 0.089 (0.055)
0.268 (0.175) 0.092 (0.191)
Electricity Usage Usage 0.122** (0.051) 0.830** (0.342) 0.051 (0.081) -0.081 (0.129) lnfirmage 0.018 (0.016) 0.101 (0.062) 0.104*** (0.035) 0.072* (0.039) Town 0.026 (0.034) 0.436* (0.251) 0.075 (0.065)
39
Lndistmarket -0.073* (0.039) 0.014 (0.096) -0.004 (0.046) 0.020 (0.048)
Bank -0.066* (0.035) 0.227 (0.213) -0.044 (0.062) 0.035 (0.083)
Localwage 0.208** (0.087) 0.183 (0.208) 0.107 (0.066) 0.317** (0.156)
F TESTS – CRS
CRS-Manufacturing
p>F 0.0926
0.6777
0.5523 0.0941 CRS-Services
p>F 0.5143
0.8228
0.0176 0.8239
CRS-Trade
p>F 0.5389
0.3095
0.0005 0.7862 Observations 1,993
476
1,229 943
R2 0.960
0.570
0.809 0.822 Adjusted R2 0.960
0.549
0.806 0.818
note: *** p<0.01, ** p<0.05, * p<0.1
40
Table 8: Investment Probit
Investment Probit - Marginal Effects - Baseline
Bangladesh Ethiopia Indonesia Sri Lanka
coef se coef se coef se coef se
MODEL 1: Gender only manager is male 0.299 (0.217) 0.183 (0.130) -0.001 (0.133) -0.252** (0.128)
Observations observations
2,480
649
1,369
1,315
Pseudo R2 0.003
0.004
0.000
0.005
MODEL 2:Gender – initial size and sector manager is male 0.021 (0.242) 0.228 (0.150) -0.274 (0.220) -0.198 (0.150)
lnLftlag 0.164** (0.076) 0.263** (0.103) 0.187 (0.123) -0.050 (0.087)
services 0.002 (0.181) 0.157 (0.326) -0.388 (0.238) -0.343** (0.162)
trade -0.452*** (0.148) -0.002 (0.320) -0.029 (0.279) -0.106 (0.134)
hh ent -0.347* (0.193) 0.228 (0.150) -0.215 (0.249) 0.085 (0.139)
Observations observations
2,301
648
776
1,128
Pseudo R2 0.037
0.032
0.040
0.016
MODEL 3: FULL SPECIFICATION manager is male -0.181 (0.259) 0.291 (0.237) -0.378* (0.204) -0.074 (0.136)
lnLftlag 0.181* (0.093) 0.142 (0.087) 0.220* (0.117) 0.075 (0.098)
services -0.034 (0.193) 0.243 (0.304) -0.610** (0.302) -0.289* (0.168)
trade -0.436*** (0.168) -0.021 (0.276) -0.182 (0.278) -0.086 (0.139)
hh ent -0.397* (0.225)
-0.071 (0.288) 0.086 (0.139)
manager age -0.017 (0.031) -0.008 (0.043) 0.122*** (0.043) -0.000 (0.016)
mngrage2_100 1.477 (3.654) -0.000 (0.043) -0.142*** (0.048) -0.003 (0.052)
manager primary -0.172 (0.231) 0.144 (0.208) -0.457 (0.330) -0.396 (0.513)
manager secondary
-0.204 (0.209) 0.294 (0.251) 0.037 (0.284) -0.375 (0.513)
manager tertiary -0.108 (0.417)
-0.426 (0.472) -0.393 (0.634)
electricity Usage 0.325 (0.206) 1.227*** (0.243) 0.144 (0.251) 0.196* (0.110)
lnfirmage 0.039 (0.075) 0.191*** (0.070) -0.029 (0.119) -0.151*** (0.054)
town 0.055 (0.148) -0.137 (0.286) 0.384 (0.248)
lndistmarket 0.430*** (0.163) -0.155 (0.152) 0.005 (0.102) -0.024 (0.092)
bank 0.240 (0.177) -0.388 (0.273) 0.224 (0.279) 0.091 (0.152)
localwage -0.482 (0.299) -0.043 (0.278) 0.626** (0.262) -1.574*** (0.334)
Observations observations
1,855
574
697 993
Pseudo R2 0.069
0.106
0.152 0.073
note: *** p<0.01, ** p<0.05, * p<0.1
41
Table 9: Growth Models
Growth Models
Bangladesh Ethiopia Indonesia Sri Lanka
coef se coef se coef se coef se
MODEL 1: Gender only manager is male 0.013 (0.010) 0.085 (0.058) 0.008 (0.041) -0.006 (0.012)
Observations observations
2,301
725
776
1,130
R2 0.000
0.009
0.000
0.001
Adjusted R2 0.000
0.008
-0.001
-0.000
MODEL 2:Gender – initial size and sector
manager is male 0.054*** (0.013) 0.075 (0.051) 0.038 (0.034) -0.004 (0.011)
lnLftlag -0.047*** (0.011) -0.132*** (0.029) -0.030 (0.045) -0.018* (0.010)
services -0.065*** (0.022) -0.057 (0.046) 0.070 (0.053) -0.019 (0.015)
trade -0.096*** (0.020) -0.041 (0.049) 0.080 (0.055) -0.013 (0.014)
hh ent -0.059*** (0.020)
-0.082 (0.058) -0.005 (0.012)
Observations observations
2,301
724
776
1,128
R2 0.046
0.116
0.026
0.015
Adjusted R2 0.044
0.112
0.020
0.010
MODEL 3: FULL SPECIFICATION
manager is male 0.055*** (0.015) 0.077 (0.063) 0.017 (0.031) -0.001 (0.012)
lnLftlag -0.046*** (0.012) -0.159*** (0.034) -0.041 (0.043) -0.020* (0.011)
services -0.055*** (0.020) -0.071* (0.040) 0.044 (0.060) -0.026 (0.018)
trade -0.095*** (0.023) -0.085 (0.057) 0.076 (0.063) -0.018 (0.016)
hh ent -0.041** (0.019)
-0.064 (0.070) 0.003 (0.012)
manager age 0.001 (0.002) 0.013* (0.008) 0.020* (0.012) -0.003** (0.001)
mngrage2_100 -0.146 (0.212) -0.012 (0.009) -0.020* (0.011) 0.006* (0.003)
manager primary 0.004 (0.013) 0.047 (0.059) -0.016 (0.038) -0.003 (0.016)
manager secondary -0.004 (0.015) 0.151** (0.075) 0.041 (0.042) 0.010 (0.018)
manager tertiary 0.055* (0.033)
0.024 (0.075) 0.015 (0.024)
electricity usage 0.024* (0.012) 0.063 (0.088) 0.126*** (0.038) 0.005 (0.012)
lnfirmage -0.014*** (0.005) -0.017 (0.018) 0.022 (0.029) 0.000 (0.005)
town -0.018 (0.011) 0.096* (0.058) 0.031 (0.029)
lndistmarket 0.016* (0.008) -0.027 (0.022) 0.015 (0.014) -0.009 (0.006)
bank 0.007 (0.008) 0.001 (0.042) 0.030 (0.036) 0.009 (0.010)
localwage 0.038** (0.016) -0.035 (0.048) 0.046* (0.027) 0.031 (0.021)
Observations observations
1,855
636
697 993
R2 0.064
0.170
0.121 0.045
Adjusted R2 0.056
0.152
0.100 0.030
note: *** p<0.01, ** p<0.05, * p<0.1
42
Appendix A: Data construction
Individual Activity Portfolios.
Individuals labor market activity portfolio:
Employment categories were created based on the activities individuals reported engaging. In particular, individuals were categorizes as engaging in one or more of the following activities: Self employed in NFE; Self employed in agriculture, or wage work (includes both agricultural and non-agricultural wage employment. Individuals not reporting any such activities were classified as being out of the labor force (OLF); (Retired, unemployed or housework). However, the categorizations were based on different sets of questions:
In Bangladesh and Sri Lanka, people were asked about their main occupation and whether they worked on own
farm, for a wage and in a household non-farm enterprise over the past year. This means that people could
indicate up to three activities.
In Ethiopia we have information on workers’ primary and secondary income earning activities over the past
year (but not on other income earning activities, which explains why there are no individuals who reported
having engaged in wage employment, agricultural employment and non-farm enterprise activity).
In Indonesia, people were asked whether they worked in a household non-farm enterprise over the past year
and whether they worked for a wage in the past twelve months. However, no questions were asked about
whether they worked on a family farm.
Education (primary, secondary, tertiary): Education categories were based on the years of schooling reported and
each countries’ education system, years of schooling were converted in dummies for each level of education. The few
people who indicated having completed “other types of education” were considered to have attended primary school.
Bangladesh Ethiopia Indonesia Sri Lanka
Primary “First to fifth grade” “other”
1 to 6 years of schooling “Adult literacy program” “Other literacy program” “Church/Mosque schooling”
“elementary” “First to fifth grade” “other”
Secondary “Sixth to ninth grade” “SSC\Equivalent” “HSC\Equivalent”
Above 6 years of schooling “Tech/Vocational training” “University/College”
“Junior High” “High School”
“Sixth to ninth grade”
Tertiary “University” Tertiary was not created given that the number of respondents who studied past high school was extremely small.
“Vocational school”, “Diploma I/II” “Diploma III” “Diploma IV/BA” “Masters – Phd”
“University” “Professional” “Technical college”
Note: the same education categories were used to define partner’s education and education of firm managers.
Occupation of father of household head: Categories of occupation were defined in a similar way to individual
occupations (self employment in agriculture, self employment in NFE and wage employment).
In Sri Lanka, the option other was available which led to the inclusion of the extra category “others”. No questions
about father’s occupation was included in the Indonesia survey.
43
Household variables
Household size: Number of individuals living in each household
Share of Children: Proportion of household members younger than 10 years old.
Share of Elderly: Proportion of household members older than 65 years of age.
Share women: Proportion of adult women in the household (10 to 65 years old).
House ownership: Dummy variable that indicates whether or not the household owns the house which it inhabits.
Agriculture land: Log of total land used by the household for farming. This might include both land that is owned and
land that is only controlled by the household. This variable is not available in Ethiopia.
Plough ownership: Dummy variable that takes the value one when the household owns a plough. Used only in
Ethiopia as a replacement for agriculture land owned.
Community Characteristics
Town: This is a dummy variable which takes the value one if the households live in a rural town or city. In the case of
Bangladesh it also includes peri-urban locations.
Daily wage of a male agricultural laborer in the community: The going wage rate to hire a male agricultural laborer
for one day (from the community questionnaire). In Indonesia it is an average of the wages paid in monsoon and non-
monsoon seasons.
Distance to market – Distance in km to the nearest market where locals usually travel to sell their products. In
Ethiopia, respondents were given the opportunity to answer the distance question in terms of distance covered or
travelling time; fortunately, virtually all respondents who answered these questions in terms of travelling time indicated
that their typical mode of transportation was walking, enabling us to impute distance by assuming that people on average
work 6 kilometers an hour.
Credit Institution – Dummy variable that takes the value one if there is any financial institution in the community. The
type of institution included in each survey was different. The following table shows which type of credit institution was
included in each country.
Bangladesh Ethiopia Indonesia Sri Lanka
NCB Private Bank NGO Bank Other
Bank Microfinance Community group Other
Local Commercial Bank Other Local Bank Foreign Bank NGO or business organization Government Program (KIK, KUT, etc) Cooperative Investment Funds Other
Private Commercial Bank Government Commercial Bank Rural Bank (Gramiya Bank) Samurdi/Janasaviya Bank Sanasa Bank Rural Development Bank IRDP/REAP/ABGEP Financial and leasing company Others
Factors of Production
44
Labor (Full-time Equivalent - FTE): Labor inputs were reported as the number of hours worked per year per worker
(only days in Ethiopia). This includes the time worked by paid and unpaid workers and the owner/manager. To make
them comparable we converted each of the labor input measure into a full-time worker equivalent (FTE). One FTE was
considered to work 8 hours per day, 25 days per month, 12 months per year.
Output: Measured by the total value of sales in the previous year.
Capital: the capital variable used is the sum of the value of all the assets used by the firm. This includes land, buildings,
equipment and machinery and vehicles plus the value of rental buildings and equipments.
Material inputs: material inputs are the sum of total expenditure on intermediate inputs (including, but not limited to,
costs associated with utilities usage, insurance and phone usage) and items to be resold. In Ethiopia it was computed as
the product of the number of material inputs bought times and the average price of each material input unit.
Firm variables
Size (one worker, 2-5 workers, more than 5 workers): Firm size based on the number of workers.
Sector (Manufacturing, Services, Trade): Sector in which the firm develops its activity. Sectors were defined
according to the following correspondence.
Bangladesh Ethiopia Indonesia Sri Lanka
Manufacturing Mining and quarrying Manufacturing Electricity, gas and water supply Construction
Food and beverages, brewing/distilling Manufacturing (excl: grain milling, food and beverages, distilling, wearing apparel) Grain milling
Mining and Excavation Manufacturing, including processing of agricultural products Electricity Gas and water Construction
Manufacturing non-agricultural goods Processing of Agriculture, hunting, fishing products Mining and quarrying Electricity, gas, water Construction
Trade Wholesale and retail trade (excluding repair of motor vehicles, motorcycles, personal and household good)
Retail trade via stalls and markets Retail (not stalls/mkts) Whole sale trade
Wholesale and retail trade
Retail or wholesale trading
Services Hotels and Restaurants Transport, storage and communications Financial intermediation Real estate, renting and business activities Public administration and defense; compulsory social security Education Health and social work Other community, social and personal service activities Activities of private households (ISIC Rev.
Hotels and restaurants Services(services, manufac. apparel/tailoring, rental, rec) Transport services Others (specialized services)
Car, motorcycle and household goods Repair shops Hotels, food and beverages Transportation, storage, and communication Finance Real estate, leasing and business services Health services Public and social services Catering
Services
45
3.1. – Top Level Category P) Exterritorial organizations and bodies Repair of motor vehicles, motorcycles and personal and households goods
Note: In Bangladesh and Sri Lanka firms were asked which percentage of their sales corresponded to
manufacturing, services or trade and they were considered to be of a given sector if more than 33% of sales was
generated in that sector during the past 12 months.
Manager experience: In Sri Lanka there manager’s age was not available so that variable was replaced by number of
years of experience as a manager.
Household enterprise: This variable identifies home based firms.
Electricity usage: Dummy variable that takes value one when the firm uses electricity.
Firm age: the age of the firm measured in years.
Registered: Dummy variable that takes the value one when the firm is registered.
Seasonality: Number of months of normal or high sales reported by the firm.
Number of competitors (more than 20, 11 to 20, 6 to 10, less than 5): Number of other firms working in the same
business in the area.
Loan: Whether the firm has a loan with any type of credit institution. The time period of loans to be reported was
different countries in each country. In Sri Lanka and Ethiopia firms reported loans obtained during the last 5 years, in
Bangladesh the period was the last 3 years and in Indonesia the last 12 months.
46
Appendix B: Constraints
When asked about the most important constraint to running their firm, both male and female
managers consider a lack of markets (demand), transport, access to credit and electricity as their
most important constraints, as demonstrated in Table B1. The importance of these constraints
varies across countries. For example, concerns about markets are particularly pressing in Bangladesh
and Ethiopia, which is characterized by high remoteness. Other constraints, such as with taxation,
registration procedures, labor regulation and the availability and regulation of land are only
considered the most important constraint by a very small minority of firms in each country, though
governance and public utilities are the most pressing constraints for a small group of firms. Gender
differences in perceived constraints and their ranking are not large in general, yet vary across
countries.
Table b2 demonstrates how important of a constraint a particular issue was, disaggregated
by gender and country. There are very few constraint that are always considered more of an issue by
either gender in each country. An exception is governance, which men always consider more of a
constraint than women (although the gender difference is not statistically significant in Indonesia),
perhaps because male firms tend to be larger. Registration, taxes and public infrastructure are
considered a significantly more severe obstacle by male managers in Bangladesh, Ethiopia and Sri
Lanka, but not in Indonesia, perhaps reflecting differences in firm size. Women consider access to
finance more of a constraint in Bangladesh, Ethiopia and Indonesia, while men consider it more of a
problem in Sri Lanka.
To assess to what extent gender differences in constraints might be driven by, or masked by,
firm characteristics, we estimate ordered probit models of the severity of a given problem,
controlling for firm size, capital stock, sector, human capital of the manager, investment climate
characteristics and gender.
Results are presented in Table B3A. The gender dummies are most often insignificant, and
their sign varies considerably across countries. Overall, no consistent gender patterns emerge.
Since subjective responses may be biased because of individuals’ latent optimism (or
pessimism), we also run a regression in which we use as the dependent variable the severity of the
constraint minus the average constraint severity reported by the individual in question (see
Hallward-Driemeier and Aterido, 2009b). This arguably eliminates individual “fixed” effects and
allows variation in reported constraint severity to be primarily identified off within-individual
variability. Of course, the drawback of this procedure is that variation across individuals is discarded.
The results are presented in table B3B, and are generally robust to this transformation of the
dependent variable, although it is noticeable that the coefficients associated with certain constraints
change. Interestingly, however, there is still no systematic pattern of gender differences across
countries.
In sum, gender differences in self-reported constraints are small both unconditional and
conditional on differences in firm, human capital, and investment climate characteristics.
47
Table B1: Most important constraint
Bangladesh Ethiopia Indonesia Sri Lanka
Men Women Men Women Men Women Men Women
Mean/se Mean/se Mean/se Mean/se Mean/se Mean/se Mean/se Mean/se electricity 16.99% 3.62% 4.50% 7.89% 16.65% 12.26% 27.02% 24.78%
0.38 0.19 0.21 0.27 0.37 0.33 0.44 0.43
public utilities 0.81% 0.00% 5.33% 10.00% 2.93% 5.13% 13.75% 11.38%
0.09 0.00 0.22 0.30 0.17 0.22 0.34 0.32
finance 10.24% 23.31% 31.74% 30.72% 25.07% 22.46% 19.15% 15.51%
0.30 0.43 0.47 0.46 0.43 0.42 0.39 0.36
transport 9.12% 11.13% 18.13% 11.89% 11.35% 18.72% 18.27% 19.13%
0.29 0.32 0.39 0.32 0.32 0.39 0.39 0.39
markets 42.93% 50.09% 35.56% 37.02% 28.37% 23.81% 14.02% 27.09%
0.50 0.50 0.48 0.48 0.45 0.43 0.35 0.45
labor 0.06% 0.00%
0.22% 0.39% 1.80% 1.01%
0.03 0.00
0.05 0.06 0.13 0.10
land 1.04% 0.00% 1.90% 1.41% 1.30% 0.97% 1.00% 0.40%
0.10 0.00 0.14 0.12 0.11 0.10 0.10 0.06
governance 8.15% 8.44% 0.12% 0.57% 9.60% 10.68% 1.39% 0.14%
0.27 0.28 0.03 0.08 0.29 0.31 0.12 0.04
registration 1.04% 0.00% 0.76% 0.00% 0.98% 2.23% 0.82% 0.18%
0.10 0.00 0.09 0.00 0.10 0.15 0.09 0.04
taxes 0.07% 0.00% 0.89% 0.00% 0.99% 3.27% 0.74% 0.27%
0.03 0.00 0.09 0.00 0.10 0.18 0.09 0.05
others 9.55% 3.40% 1.07% 0.50% 2.54% 0.07% 2.03% 0.10%
0.29 0.18 0.10 0.07 0.16 0.03 0.14 0.03
Table B2: Average rating of severity of constraints
Bangladesh Ethiopia Indonesia Sri Lanka
Men Women Men Women Men Women Men Women
Mean/se Mean/se Mean/se Mean/se Mean/se Mean/se Mean/se Mean/se
electricity 1.28 0.65 1.00 0.87 1.10 1.17 1.20 0.87
1.12 0.81 1.36 1.27 1.29 1.28 1.29 1.15
public utilities 0.06 0.00 1.06 1.21 0.79 1.03 1.01 0.62
0.37 0.06 1.35 1.33 1.18 1.30 1.24 1.01
finance 1.27 1.50 1.84 2.11 1.78 1.80 1.40 1.24
1.23 1.40 1.38 1.27 1.38 1.38 1.40 1.39
transport 1.36 1.38 1.81 1.33 1.26 1.32 1.17 1.13
1.18 0.95 1.37 1.40 1.33 1.35 1.36 1.32
markets 2.08 2.09 2.29 2.31 1.25 1.21 1.25 1.39
0.98 0.82 1.07 1.11 1.33 1.28 1.36 1.35
labor 0.03 0.02 0.47 0.15 0.49 0.50 0.20 0.11
0.23 0.16 1.04 0.63 0.99 0.99 0.72 0.56
land 0.14 0.11 0.52 0.65 0.49 0.46 0.11 0.04
0.51 0.55 1.10 1.14 0.99 0.99 0.54 0.28
governance 2.05 1.48 0.61 0.24 1.19 1.13 0.22 0.13
0.96 1.00 1.06 0.71 1.34 1.38 0.74 0.58
registration 0.11 0.00 0.22 0.03 0.61 0.62 0.17 0.08
0.38 0.06 0.71 0.24 1.07 1.09 0.66 0.49
taxes 0.02 0.00 0.47 0.10 0.65 0.68 0.15 0.07
0.18 0.05 1.04 0.45 1.12 1.11 0.62 0.44
Note: In each country managers were asked about the seriousness of different obstacles to investment. Constraints were
grouped by theme and the maximum value of among the different constraints in a given category was used. For example,
public utilities includes problems with water, telecommunication and postal services. Different country scales of severity of
were converted to the following common scale: 0=no problem, 1==a minor problem, 2=somewhat of a problem, 3=a major
problem.
48
Table B3: Ordered probits
A - Constraints-Ordered probit electricity finance transpot market public
utilities labor land governance registration taxes
coef/se coef/se coef/se coef/se coef/se coef/se coef/se coef/se coef/se coef/se
Bangladesh Male 0.276* -0.317 -0.024** 0.251 0.504 0.156 -0.109 0.548 1.093* 0.618
(0.208) (0.255) (0.214) (0.273) (0.320) (0.434) (0.487) (0.223) (0.463) (0.490)
Ethiopia Male 0.237 -0.247 0.300 -0.210 -0.049 1.553*** -0.065 0.685*** 0.559** 0.794**
(0.248) (0.196) (0.220) (0.270) (0.213) (0.430) (0.236) (0.204) (0.273) (0.378)
Indonesia Male -0.122 -0.058 -0.058 -0.009 -0.196* -0.010 0.071 -0.049 -0.046 0.053
(0.131) (0.115) (0.133) (0.118) (0.115) (0.166) (0.134) (0.125) (0.131) (0.143)
Sri Lanka Male 0.264* 0.086 0.148 -0.004 0.355*** 0.162 0.214 -0.082 0.150 -0.112
(0.139) (0.143) (0.131) (0.136) (0.123) (0.196) (0.207) (0.218) (0.183) (0.182)
Controls not
presented:
size, capital stock, sector, manager’s age and education, firm age, a dummy for whether or not the firm uses electricity, and investment climate variables
B – “Demeaned” Constraints - Ordered probit
The demeaned measure of severity of the constraints is the reported value of each constraint subtracted by the average constraint
severity for each individual (average of the 10 constraint categories).
The demeaned measure of severity of the constraints is the reported value of each constraint subtracted by the
average constraint severity for each individual (average of the 10 constraint categories).
electricity finance transpot market public utilities
labor land governance registration taxes
coef/se coef/se coef/se coef/se coef/se coef/se coef/se coef/se coef/se coef/se
Bangladesh Male 0.128* -0.301 -0.134** 0.059 0.000 -0.045 -0.056 0.512 -0.040* -0.071
(0.242) (0.198) (0.176) (0.270) (0.204) (0.197) (0.252) (0.191) (0.200) (0.220)
Ethiopia Male -0.391* -0.250 0.816** -0.033 -0.184 -0.215 -0.395 -0.163 -0.542* -0.168
(0.219) (0.356) (0.366) (0.271) (0.394) (0.274) (0.280) (0.279) (0.291) (0.241)
Indonesia Male -0.085 0.040 -0.012 0.039 -0.159* 0.043 0.176* 0.032 0.074 0.137
(0.116) (0.091) (0.114) (0.114) (0.085) (0.106) (0.092) (0.103) (0.096) (0.113)
Sri Lanka Male 0.073 -0.093 0.004 -0.142 0.139 -0.205* -0.176 -0.284*** -0.216* -0.276**
(0.103) (0.110) (0.094) (0.103) (0.095) (0.110) (0.124) (0.107) (0.111) (0.115)
Controls not
presented:
size, capital stock, sector, manager’s age and education, firm age, a dummy for whether or not the firm uses electricity, and investment climate variables
note: *** p<0.01, ** p<0.05, * p<0.1
49
Appendix C: Additional Robustness Checks
C1: Productivity
As an additional robustness check we examine whether there might be ender differences in the
impact of the investment climate on productivity or gender differences in the returns to human
capital and firm characteristics. To investigate this possibility we interact our total factor productivity
term with a dummy for a female entrepreneur, Taking-logs and adding these gender interactions
and an error term, v, our most general estimating equation becomes:23
Under the null hypotheses of no differences in returns, all the gender interaction terms
should be equal to zero ( . There are various alternative
hypotheses. For example, women may have lower returns to schooling. In this case we would expect
the interaction terms between the gender dummy and our proxies to be negative. Alternatively,
women may benefit less from access to finance, in which case we would expect the coefficient on
the gender*credit institution interaction to be significant.
Results are presented in Table C1. Gender interaction terms add very little explanatory
power. Moreover, with the exception of Bangladesh, where the sample of female firms is very small,
gender interaction terms are not in general statically significant, either individually or jointly.
Nonetheless, there is some evidence that female enterprises that are household based in Sri Lanka
are significantly more productive than male firms and that there is a positive correlation between
local wages and the productivity of male firms, but not that of female firms. In addition, age
productivity profiles for female entrepreneurs in Indonesia may be different from those of their
male counterparts. However some differences are to be expected with so many interactions terms
and, moreover, few differences are qualitatively similar in all countries under considerations. Overall
then, the evidence for the hypotheses that the returns to human capital and the impact of
investment climate factors varies by gender is limited.
23 In the appendix, we also present models where we interact factor shares with the gender dummy. However,
we had to pool the data as the number of observations did not us to account for gender heterogeneity in
technology by sector.
50
Table C1: Production function robustness check
Production functions – Robustness Check
Bangladesh Ethiopia Indonesia Sri Lanka
coef se coef se coef se coef se
manuflnKd 0.116*** (0.022) 0.100* (0.052) 0.021 (0.038) -0.023 (0.059) manuflnLft 0.187*** (0.027) 0.461*** (0.138) 0.368*** (0.123) 0.397*** (0.104)
manuflnMd 0.655*** (0.026) 0.436*** (0.095) 0.545*** (0.092) 0.749*** (0.057)
servlnKd 0.135*** (0.019) 0.152 (0.094) 0.068*** (0.021) 0.231*** (0.066)
servlnLft 0.340*** (0.043) 0.744* (0.435) 0.248*** (0.076) 0.173 (0.184)
servlnMd 0.501*** (0.020) 0.227** (0.105) 0.543*** (0.037) 0.593*** (0.066)
tradelnKd 0.024 (0.022) 0.059 (0.071) 0.048*** (0.015) 0.034 (0.040)
TradelnLft 0.138*** (0.038) 0.624*** (0.175) 0.047 (0.042) 0.139 (0.099)
tradelnMd 0.824*** (0.031) 0.440*** (0.115) 0.748*** (0.033) 0.865*** (0.030)
services 0.929*** (0.230) 0.904* (0.514) -0.283 (0.860) -0.539 (0.722)
trade -0.768*** (0.253) 0.283 (0.719) -1.669* (0.854) -1.208* (0.720)
hh ent -0.204*** (0.049)
-0.280*** (0.100) -0.239** (0.097)
manager age 0.001 (0.005) 0.012 (0.053) -0.027 (0.022) 0.009 (0.011)
mngrage2_100 -0.125 (0.487) -0.028 (0.048) 0.028 (0.023) -0.022 (0.030)
manager primary 0.086*** (0.033) -0.750*** (0.238) -0.172 (0.114) -0.106 (0.217)
manager secondary 0.056 (0.038) -0.780 (0.775) -0.016 (0.162) -0.091 (0.200)
manager tertiary 0.123*** (0.043)
0.204 (0.191) -0.193 (0.241)
electricity usage 0.062* (0.035) 0.566 (0.566) 0.081 (0.084) -0.165 (0.108)
lnfirmage 0.032*** (0.010) 0.012 (0.112) 0.133*** (0.045) 0.065 (0.044)
town 0.014 (0.025) 0.793** (0.372) 0.071 (0.078) lndistmarket -0.037 (0.029) 0.119 (0.171) -0.000 (0.054) 0.007 (0.051)
bank -0.032 (0.027) -0.079 (0.288) -0.035 (0.089) 0.001 (0.100)
localwage 0.081 (0.074) 0.305 (0.235) 0.099 (0.069) 0.323* (0.164)
Fem*hhe 0.155 (0.286)
0.187 (0.150) 0.318** (0.136)
Fem*mngrage -0.057 (0.049) -0.010 (0.059) 0.044 (0.033) 0.020 (0.036)
Fem*mngrage2_100
5.834 (6.368) 0.014 (0.057) -0.057 (0.035) -0.075 (0.120)
Fem*primary -0.722*** (0.208) 0.353 (0.470) 0.135 (0.187) 0.236 (0.264)
Fem*secondary -0.733** (0.319) 1.270 (0.855) -0.004 (0.218) 0.413 (0.321)
Fem*tertiary -0.771** (0.343)
0.114 (0.351) 0.653 (0.591)
Fem*Electricity Usage
0.407 (0.273) -0.102 (0.597) -0.194 (0.130) 0.236 (0.180)
Fem*lnfirmage -0.153 (0.116) 0.096 (0.147) -0.041 (0.061) 0.009 (0.072)
Fem*town 0.315* (0.170) -0.512 (0.507) -0.041 (0.133)
Fem*lndistmarket -0.244 (0.181) -0.149 (0.216) 0.013 (0.062) 0.059 (0.101)
Fem*bank -0.322** (0.136) 0.490 (0.452) 0.052 (0.145) 0.158 (0.113)
Fem*localwage 0.177** (0.081) -0.090 (0.165) -0.057 (0.050) -0.079** (0.034)
F TESTS – CRS CRS-Manufacturing
p>F 0.0859
0.9830
0.6235 0.0692
CRS-Services p>F 0.5221
0.7830
0.0349 0.9858
CRS-Trade p>F 0.6283
0.5688
0.0001 0.7171
F TESTS –Female
interactions
Schooling p>F 0.0074
0.2779
0.7016 0.5458
Age p>F 0.2079
0.8890
0.0758 0.8229
Investment
Climate
p>F 0.0033
0.7060
0.8132 0.0395
Other Firm
characteristics
p>F 0.4736
0.8029
0.0489 0.1313
All
p>F 0.0000
0.0279
0.0098 0.2597
N 1993
408
1,229 942 R2 0.966
0.534
0.817 0.827
Adjusted R2 0.966
0.495
0.812 0.821 note: *** p<0.01, ** p<0.05, * p<0.1
51
C2: Investment and Growth
Analogous robustness checks are conducted for our investment and growth models. That is, we
interact the explanatory variables with a dummy for the gender of the manager to see how their
impact varies with the gender of the manager. Thus the models become:
And
The null hypothesis is that the determinants of investment and growth do not vary with the gender
of the manager (i.e. that and
.
Results are presented in tables C2 and C3 and demonstrate that the overwhelming majority of
gender interaction terms are insignificant, and moreover, that few gender interactions consistently
have the same sign in different countries, although it is of interest to note that the interaction
between a presence of a credit institution and being managed by a female entrepreneur is
consistently negative in the investment probit, suggesting that women may benefit less from
proximity to credit providers. In addition, there appear to be gender differences in the age-growth
profiles in Bangladesh and Sri Lanka. Overall, however, gender differences are not systematic.
52
Table C2: Investment Robustness checks
Investment Robustness check
Bangladesh Ethiopia Indonesia Sri Lanka
coef se coef se coef se coef se
lnLftlag 0.205** (0.097) 0.184 (0.123) 0.251** (0.112) 0.070 (0.105) services 0.030 (0.189) 0.291 (0.376) -0.612* (0.332) -0.316* (0.187) trade -0.399** (0.175) -0.030 (0.295) -0.194 (0.329) -0.130 (0.155) hh ent -0.325 (0.230)
-0.048 (0.322) 0.041 (0.161)
manager age -0.019 (0.033) -0.042 (0.045) 0.093 (0.058) -0.006 (0.017) mngrage2_100 1.829 (3.835) 0.034 (0.045) -0.100 (0.062) 0.015 (0.053) manager primary -0.013 (0.218) 0.140 (0.221) -0.610 (0.470) 0.083 (0.653) manager secondary -0.081 (0.173) 0.308 (0.364) -0.097 (0.408) 0.153 (0.657) manager tertiary -0.033 (0.407)
-0.708 (0.610) -0.069 (0.763)
electricity usage 0.473** (0.213) 0.916*** (0.226) 0.130 (0.299) 0.198 (0.141) lnfirmage 0.000 (0.078) 0.281** (0.112) -0.161 (0.171) -0.158*** (0.058) town 0.012 (0.148) -0.340 (0.408) 0.578* (0.298)
lndistmarket 0.470*** (0.166) -0.263 (0.204) 0.060 (0.130) 0.066 (0.094) bank 0.301* (0.166) -0.212 (0.297) 0.457 (0.338) 0.136 (0.168) localwage -0.526* (0.302) 0.013 (0.278) 0.526** (0.250) -1.559*** (0.339) Fem*lnLftlag -0.844 (0.749) -0.201 (0.206) -0.035 (0.213) -0.181 (0.278) Fem* Serv -2.331 (1.748) 0.027 (0.477) -0.151 (0.709) 0.132 (0.416) Fem*trade
0.182 (0.391) -0.252 (0.704) 0.486 (0.319)
Fem*hhe -3.594 (2.259)
-0.169 (0.651) 0.140 (0.309) Fem*mngrage 0.068 (0.148) 0.117* (0.063) 0.058 (0.093) 0.067 (0.052) Fem*mngrage2_100 -12.161 (19.048) -0.115* (0.062) -0.084 (0.105) -0.179 (0.204) Fem*primary -0.945 (0.774) 0.055 (0.646) 0.804 (0.632) -1.204 (0.912) Fem*secondary -0.221 (0.842) 0.401 (0.564) 0.687 (0.636) -1.216 (0.991) Fem*tertiary
1.272 (0.935) 0.545 (1.220) Fem*Electricity Usage -1.131 (0.744) 0.778 (0.645) 0.329 (0.360) -0.012 (0.338) Fem*lnfirmage 0.708* (0.423) -0.272* (0.150) 0.371 (0.307) -0.042 (0.119) Fem*town 0.796 (0.657) 0.673 (0.588) -0.633 (0.394) Fem*lndistmarket 0.764 (0.594) 0.239 (0.176) -0.018 (0.147) -0.543*** (0.180) Fem*bank
-0.746* (0.448) -0.836* (0.492) -0.287 (0.282)
Fem*localwage 0.149 (0.379) -0.367* (0.208) -0.054 (0.178) 0.180* (0.098)
N 1,834
574
697 993
Pseudo R2 0.102
0.133
0.188 0.094
note: *** p<0.01, ** p<0.05, * p<0.1
53
Table C3: Growth Robustness Check
Growth – Robustness Check
Bangladesh Ethiopia Indonesia Sri Lanka
coef se coef se coef se coef se
lnLftlag -0.048*** (0.013) -0.141*** (0.042) -0.026 (0.051) -0.019 (0.013) serv -0.060*** (0.021) -0.113** (0.051) 0.079 (0.059) -0.025 (0.021)
trade -0.100*** (0.024) -0.099 (0.077) 0.139** (0.062) -0.014 (0.017)
hh ent -0.041** (0.020)
-0.111 (0.091) 0.004 (0.014)
manager age 0.001 (0.002) 0.013 (0.011) 0.029* (0.015) -0.004** (0.001)
mngrage2_100 -0.095 (0.223) -0.013 (0.012) -0.026* (0.015) 0.008** (0.004)
manager primary 0.005 (0.014) 0.066 (0.059) -0.010 (0.046) -0.028 (0.023)
manager secondary -0.001 (0.015) 0.156* (0.089) 0.000 (0.057) -0.019 (0.020)
manager tertiary 0.057* (0.033)
0.017 (0.094) -0.024 (0.027)
electricity usage 0.030** (0.013) 0.128 (0.095) 0.104* (0.058) 0.003 (0.013)
lnfirmage -0.015*** (0.006) -0.022 (0.024) 0.030 (0.037) 0.002 (0.006)
town -0.020* (0.012) -0.039 (0.089) 0.061 (0.044) lndistmarket 0.017* (0.009) -0.072* (0.042) 0.009 (0.020) -0.012 (0.007)
bank 0.009 (0.009) -0.038 (0.060) 0.024 (0.052) 0.006 (0.011)
localwage 0.043*** (0.016) -0.029 (0.052) 0.010 (0.033) 0.031 (0.022)
Fem*lnLftlag 0.025 (0.031) -0.036 (0.066) -0.050 (0.063) 0.000 (0.025)
Fem* Serv 0.123*** (0.046) 0.132 (0.085) -0.343** (0.172) -0.016 (0.037)
Fem*trade 0.182*** (0.066) 0.051 (0.093) -0.409** (0.166) -0.031 (0.035)
Fem*hhe 0.126*** (0.045)
0.056 (0.105) -0.009 (0.022)
Fem*mngrage 0.009 (0.006) -0.005 (0.014) -0.026 (0.018) 0.006** (0.002)
Fem*mngrage2_100
-1.191* (0.693) 0.005 (0.017) 0.020 (0.017) -0.016** (0.008)
Fem*primary 0.027 (0.039) -0.085 (0.174) -0.010 (0.066) 0.024 (0.041)
Fem*secondary -0.010 (0.032) -0.046 (0.156) 0.054 (0.083) 0.034 (0.031)
Fem*tertiary 0.077 (0.064)
0.003 (0.149) 0.068 (0.053)
Fem*Electricity Usage
-0.072*** (0.026) -0.161 (0.167) 0.058 (0.090) 0.023 (0.028)
Fem*lnfirmage -0.003 (0.014) 0.012 (0.038) 0.006 (0.071) -0.013 (0.011)
Fem*town 0.057*** (0.020) 0.245** (0.105) -0.067 (0.066) Fem*lndistmarket 0.008 (0.025) 0.076 (0.052) 0.047 (0.031) 0.013 (0.008)
Fem*bank -0.051 (0.039) 0.078 (0.098) -0.006 (0.097) 0.011 (0.024)
Fem*localwage -0.032*** (0.010) -0.031 (0.035) 0.063* (0.033) -0.004 (0.004)
N 1,855
636
697 993
R2 0.071
0.193
0.179 0.058
Adjusted R2 0.056
0.158
0.142 0.030
note: *** p<0.01, ** p<0.05, * p<0.1