the growth and convergence of manufacturing productivity in industrial and newly industrialising...

EISEVIER Int. J. Production Economics 34 (1994) 139-149

production economics

Working paper - for discussion

The growth and convergence of manufacturing productivity in industrial and newly industrialising countries

Paul Lowea**, Elton Fernandesb

a Department qf Man~fhcturing and Engineering Systems, Brunei University. Lkbridge, h4iddlese.u. UK

bDepartment qf Engineering Production EE/COPPE/Univesidade Fed. do Rio de Janeiro, RJ, Brazil

(Received 26 August 1993; accepted for publication 11 September 1993)

Abstract

This paper focuses on the manufacturing sector of specific economies to examine “why their growth rates differ”. The technology gap approach is adopted to show the relevant role of technological change on national manufacturing productivity growth. The productivity convergence hypothesis was partially supported by the results. The results also produce empirical evidence of the importance of the inputs of the technological gap model to countries at different levels of manufacturing development. The estimated models suggest that the input coefficients might be related to the gap of a country to the leader.

1. Introduction

Economic growth is a major concern of national

policy. Comparisons between regions and countries

have naturally emerged from economic growth

studies. This paper concentrates on the “technology

gap approach” to current studies of the sources of growth.’ The development of growth analysis, through several comparative studies,’ includes a spatial dimension and in its version of the techno-

* Corresponding author.

I For an excellent historical view of economic growth ap- proaches, see Chapter 1 ‘Thinking about growth’ by Abramovitz

(1989) and for a review of theories of comparative economic

growth, see Choi (1983). ’ Such as Maddison (1987), Elias (1978). Baumol (1986), Fager-

berg (1987, 1988), etc.

logical gap model it is very useful in the formula- tion of industrial and technological policies of countries.

This paper compares the levels of manufacturing production per capita between countries. The selected sample involves 33 countries at different levels of manufacturing development. These countries are clustered into three groups: newly industrialising countries (NICs); industrial countries (ICs); and advanced industrial countries (AICS).~ Here, manufacturing productivity refers to the level

3 The classification of the countries was based on the results of

a cluster analysis of the gross domestic product and manufacturing value added per capita of the countries per year. The bench-

mark year for the classification was the position of the country in 1985 or near year available. Statistical procedures of the cluster analysis are available under request to the author. The

results follow with the respective country acronyms used in this

0925-5273/94/$07.00 0 1994 Elsevier Science B.V. All rights reserved.

SSDI 0925-5273(93)E0089-E

of manufacturing value added (MVA) per capita and size means the population of a country. The world MVA is very much concentrated in a small number of countries. In 1985 just two countries accounted for about 50% of the total MVA of the 33 countries. These are the United States and Japan with 35.19% and 15.62% of the MVA and 21.05% and 10.11% of the total population. Brazil, the largest country among the NICs, in 1985, contrib- uted 2.29% of the total MVA, while its population represented 12.09% of the total.

There is a relative stability in the country clusters, although some relevant shifts in the shares of countries to the total MVA can be observed. For example, while Japan, from 1960 to 1985, increased its share from 6.51% to 15.62%, the United King- dom dropped from 13.65% to 6.65% in the same period. Some researchers claim convergence in productivity between countries [ 1,2]. This process of convergence is mainly attributed to the diffusion of technology and the ability of countries to take advantage of the knowledge created abroad. This paper analyses the strength of such argument.

This paper develops the analyses of Fagerberg 13, 43 for the case of manufacturing output. The technology gap approach is empirically tested, with the data available for the 33 countries at different stages of manufacturing development. The analysis is divided into three stages: (1) concentration of manufacturing output; (2) the productivity convergence hypothesis; (3) the growth of MVA per capita.

2. The technological gap approach

The technological gap approach developed by Posner [S], Vernon [C;], Cornwall [7], Fagerberg

Foot note 3 (Continued) paper: AlCs - the United States (unit), Canada (cana), Switzer-

land (swit). Denmark (denm), West Germany (germ), France

(frank Japan (japa), Sweden (swed), Belgium (belg), Finland (fini), the United Kingdom (ukin). Norway (nor%>); ICs - Netherlands

(neth), Austria (aust), Ireland (irel), Israel (isrrl), Italy (ital), Aus-

tralia (&a), New Zealand (nlrea), Spain (spai), Hong Kong (hong); NlCs - South Korea (kore), Greece (gree), Mexico (mexi).

South Africa (sout), Portugal (port), Malaysia (mala). Argentine

(arge). Chile (chil). Uruguay (urug), Brazil (braz), Panama (pana).

13, 4j4 and others accentuates the crucial role of technology in the process of economic growth and trade performance. This approach is based on the fundamental principle of comparative advantage adopted in the Heckcher-Ohlin (H-O) theory of trade, which considers that a sufficient condition of trade is the inequality in the endowment of factors of production, but from a radically different per- spective. Technology in its wide sense of technical knowledge created in the country or acquired from abroad is not always freely accessed in a country because of institutional settings and the general capacity for exploiting the benefits of knowledge. This input is in the core of the explanation of the growth of output and the effort to improve the technological level of a country or region is directly proportional to the growth of output. The approach brings into the analysis of growth the different capabilities of countries to innovate and to imitate.

Fagerberg f4] argues that innovation and diffusion are conflicting forces in the process of economic development. The former helps to increase the productivity gap between countries and the latter helps to diminish this gap. This paper considers that these two forces might also act in the same direction as both help to improve the productivity of countries. For example, the cooperation between ICs might well increase the productivity gap between these and the NICs. The process of productivity convergence between countries has been studied by several scholars. The hypothesis of convergence has been tested by Abramovitz [2] and Baumol [I] for a select cluster of ICs and by Dowrick and Gemmell [S] in a world-wide analysis of capitalist economies. The former studies support the convergence hypothesis on a long-term per- spective, while Dowrick and Gemmell claim to have found “strong statistical evidence that productivity in industry is diverging within the group of

’ Cornwall [TJ studies the growth of manu~cturin~ industry

under the hypothesis that this segment of the economy repres-

ents the engine of growth while Fagerberg [3, 41 analyses are

concentrated on the growth of the aggregate output of the

economies. Both studies will be explored in some detail.

poor countries and also falling behind in relation to the world leaders”.

Cornwall [7J argues that the technological gap explanations of growth concentrate on the supply or productive aspects of manufacturing output. Cornwall’s empirical tests took the United States as the leading country and considered the growth rate from peak-to-peak, being selected over four periods from 1950 to 1970.” The models estimated by Cornwall show that as a country closes the gap its scope to take further advantage of technology from abroad decreases. One of his estimated equations is

(0.286) (0.083)

R2 = 0.80, (1)

where grn is the rate of manufacturing output growth (“!); y, is the ratio of per capita income in a country relative to the industrial leader (s) and (I/Q), is the ratio of investment to output in manufacturing. Considering (lr as an indicator of the gap, Cornwall estimated that Japan had a “gap bonus” of 8.08% for manufacturing growth6 at the begin- ning of the post-war period, while Germany had 3.74%.

The results obtained by Cornwall [7] and other early studies of the technology gap, as an important element of the explanation of growth and international trade, have encouraged new developments using the technological gap approach both in international trade and growth analyses (e.g. [3,4,991 I]). Particularly interesting for this investigation is the work developed by Fagerberg [3,4]. Fagerberg [4] assumes that:

. . . the level of production in a country Q is a multipli~dtive function of the level of knowledge diffused to the country from abroad D, the level of knowledge created in the country or ‘national technology’ N, the country’s capacity

’ For more specifications of this empirical tests, see Cornwall

[7] Chapter II.

6 Annual average rate of growth.

for exploiting the benefits of knowledge C, whether internationally or nationally created, and a constant Z

Q = ZD”N@C’ (2)

Equation (2) is a form three-factor Cobb-Doug- las production function developed by Solow and Tinbergen. It permits a change to a linear form by the use of logarithms and also it gives directly estimations of the input elasticities which are z, band T. Another interesting feature of this formula- tion is that by differentiating and dividing through by Q the equation takes a linear form in which dependent and independent variables become growth rates, by taking the following form:

y = xd + /In + zc, (3)

where q is the growth rate of output; d is the growth rate of knowledge diffused to the country from abroad; tf is the growth rate of national technological activity and c is the growth rate of national technological activity and country’s capacity to ex- ploit the benefits of knowledge.

Fagerberg [4] suggests a development of Eq. (3) to represent the increasing effect on growth relative to the size of the gap between a country and the leader which is supposed to be at the technological frontier. This measure is very similar to the gap bonus suggested by Cornwall [7], with the differ- ence that the latter applies the relation between the levels of income as an independent input of the technology gap estimator to his model (Eq. (l)), while Fagerberg specifies the effect of the diffusion of technology from abroad in the following way:

(4)

where7 d is the diffusion weighed by the gap; 1~ is the diffusion of technology from abroad; T is the total amount of a country’s knowledge and TF is the total amount of the leader country’s knowledge.

Equation (4) implies a measure of the level of diffusion of technology from abroad related to the gap. By substituting Eq. (4) into (3) Fagerberg

’ sic.

142 P. Lowe, E. Femandes~lnt. J. Production Economics 34 i 1994) 139-149

reached his theoretical model that can be represented by the following equation:

q= ap 1 -f +pn+ z(‘. ( ) f

(5)

Although, the technological gap, diffusion and innovation have been theoretically considered by Fagerberg, he did not use separate indicators of these variables in his empirical tests. It is also appreciated that the quantification of T and T, would be a major challenge.

Instead of the relation between a country and the country at the technological frontier, as used by Cornwall [7], he assumed that the growth of total output (GDP) was inversely proportional to the measure of the total set of techniques (knowledge) in use in the country (GDP per capita). One important improvement of the Fagerberg model com- pared to that of Cornwall was the weighting of diffusion by the technological gap. However, this has not been tested empirically.

Both Cornwall and Fagerberg [3, 41 have achieved results with significant statistical tests. Be- cause of the different aggregations used by the authors8 it is not possible to compare their empirical results, but as both have sought to explain why growth rates differ in terms of the technology gap theory their methodologies are related. Because of the difficulty in obtaining separate data to represent the explanatory factors of the growth phenomenon and the correlation between the independent variables, the authors simplified their assumptions to facilitate their empirical tests. This paper uses Eqs. (3) and (4) with separate indicators for each input and examines the effect of the gap over the other inputs of the technological gap model.

3. Manufacturing concentration

Manufacturing concentration in this study is expressed by the relation between the national shares

’ The former concentrated his analysis on the growth of manufacturing industry while the latter focused on the growth ofgross

domestic product.

Table 1

Socio-economic indicators

COUN %POP %MVA GPOP% GGDP% GMVA

“/”

unit 21.051

cana 2.264 swit 0.571

denm 0.457

germ 5.444

fran 4.923

japa 10.775

swed 0.745

belg 0.880

fin1 0.438

ukin 5.045

norw 0.370

ICS

neth 1.293

aust 0.674

ire1 0.317

isra 0.377

ital 5.098 atra I .406

nzea 0.290

sing 0.228 spai 3.444

hong 0.484

NlCs

kore 3.664

gree 0.887

mexi 7.043

sout 2.892

port 0.912

mala 1.399 arge 2.124

chil 1.077

urug 0.269

braz 12.096

pana 0.182

35.193 1.125 3.290 3.896 2.661 1.404 4.275 4.604 1.139 0.747 2.547 2.616

0.570 0.441 3.229 3.890

10.793 0.385 3.314 3.408

6.521 0.758 4.019 3.547

15.622 1.003 6.990 7.812 1.328 0.441 3.104 3.289 1.108 0.312 3.562 3.141 0.650 0.41 I 3.752 5.281

6.655 0.293 2.548 I.157

0.363 0.666 4.389 2.772

1.152 0.932

0.680 0.278 0.290 0.9 IO

0.304 2.816

3.647 0.519 1.001 1.124

0.202 1.273

0.168 ~

2.121 0.953 0.207 ~~

3.641 4.09 1 3.460 4.196

3.839 5.551

6.128 7.425

3.918 4.119

3.925 2.310

2.750 3.496

4.973 6.648

1.276 2.004

0.243 0.709

1.811 3.069 0.659 2.321

0.256 0.539

0.173 ~

0.588 1.583

0.247 1.877

0.064 0.689 2.290 2.530 0.018 2.532

8.261 17.417

5.214 6.1 10

5.630 6.382

3.930 4.322

4.47 I 6.722

2.073 2.34 I

2.585 2.051 0.918 0.693

7.560 5.526

6.057 6.201

COUN country; for the country’s acronyms see footnote 3.

of MVA and population to the sample total. The seven countries,’ in which most of the manufacturing production of the market economies was concentrated in 1985, accounted for 81.09% of the total MVA of the sample total, while their share of population of the sample total was 54.60%. Table 1

’ unit, cana, germ, fran, japa, ukin, ital. For acronyms see foot-

note 3.

P. Lowe, E. Fe~nande.~/~nt. J. Prod~ctj~~~ Economics 34 (1994) 139-149 143

Concentration of MVA - 1960/l 985 (The Lorenz curve)

Cumulative percentage of population

Fig. I.

shows the percentage shares of population and MVA to the sample total in 1985 (%POP, %MVA) and the annual compound rates of growth of population, GDP and MVA (GPOP%, GGDP%, GMVA%) for the period 1960-1985. Singapore, Hong Kong and Malaysia do not have data of MVA available for the 1960s. However, these countries are known to have high rates of growth between 1960 and 1985, but they started at a low level of manufacturing production and still represent a small portion of total world manufacturing. Amongst the NICs, South Korea presented a high rate of growth. This country, in 1960, had a MVA share of 0.063% of the sample total and with con- sistent growth, achieved a share of 1.276% in 1985.

Fig. 1 shows the level of concentration of the MVA by the Lorenz curvei for the 33 countries. The area between the diagonal and the curve represents the level of concentration of the manufacturing output. In Fig. 1, the diagonal represents an equality of percentages of population to MVA. The diagonal shows a perfect distribution and compfete convergence of productivity in terms of MVA per capita. Fig. 1 shows that the level of concentration had a small change from 1960 to 1985. The reduced areas near the origin and between 50% and 80% of the population suggest a slight decline in manufacturing concentration.

” The countries are in ascendening order of MVA.

4. Productivity convergence

The process of convergence of productivity has been supported by several studies [1, 2, etc.] If convergence is confirmed, how does it happen in relation to the productivity of manufacturing industry? To identify this phenomenon, an index based on the relation of the MVA share and population share of a country to the sample totals was developed. This index can be represented

by

where y is the year, c is the country, MI is the manufacturing index, %MVA is the manufacturing value added percentage to the total of the sample (33 countries) and %POP is the population percentage to the total of the sample.

Table 2 shows this index from 1960 to 1985 at intervals of five years. The growing indexes for Japan, Germany, Canada and Italy, four amongst the seven largest market economies in the world, contribute significantly to the manufacturing concentration shown in Fig. 1. The United States de- clined up to 1975 but by 1985 recovered most of its loss. Amongst the seven largest market economies, only the United Kingdom had a consistently de- clining index.

A tendency to converge would be represented for the mean of a cross-section over time tending to 100. An index of 100 signifies that a country has its manufacturing share to the total of MVA of all countries in the sample equal to its population share. From 1960 to 1985 there is a slight convergence; the mean in 1960 was 84.96 with a standard deviation of 65.41; and in 1985 it was 85.29 with a standard deviation of 57.86. All three groups: AICs, ICs and NICs have a smaller standard deviations which shows that each group is converging, but the behaviour of the mean differs between the groups. Table 3 shows the means and standard deviations of the AICs, ICs and NICs. The convergence between countries in the same group shows some balance within but the divergence between the means of the NICs and the ICs can be interpreted as a movement of production potentiality

Table -7 Table 4 Manufacturing index 1960/19X5 Weighted means 1960 and 1985

COUN 1960 1965 1970 1975 1980 19X5

AICS

unit cana

wit

denm germ

fran

japa wed

belg

fin1

ukin norw

ICS

neth

23tlSt

ire1

isra

ital

atrn

nzea

sing

spai hong

NlCs

kore

gree

mexi

sout

port

m&t

arge Cllll

urug braz

pana

168.77 168.88 I SO.58 150.X2 156.25 167.87 108.64 123.09 110.82 121.27 114.86 119.54 251.00 233.18 23 I .66 208.27 205.08 200.88 108.09 I 14.04 108.12 105.03 107.79 126.95 189.77 193.15 200.03 193.28 199.40 201.66 134.52 126.9 I 140.14 147.10 146.52 134.73 57.05 76.62 121.1X 116.73 134.09 147.47

17X.19 192.63 190.53 204.10 171.82 181.42 i26.38 120.92 122.95 329.73 127.96 128.09 91.26 98.83 I 12.84 129.00 138.60 150.X I

213.93 191.5’ 170.88 165.27 134.30 134. I J 115.17 I 17.94 11 1.79 12O.h I 99.74 99.73

82.87

77.X0 59.59

54.09

59.66 122.12

Xl.25

31.39

2.08 2.39 14.94 16.15

23.43 24.12

2X.22 30.44

12.71 15.02

36.02 36.25

44.19 43.56

47.73 36.75

19.54 15.79

8.27 11.27

X2.97

75.34 64.71

65.28

64.03

119.62

87.12

26.42

34.48

88.96

81.51 67.3 1

73.08

71.65

IOU.92

X3.39

35.32

53.70

5.41 13.65 23.70 38.64 19.85 28.04 29.45 27.87 24.75 27.52 ‘8.5X 26.16 29.98 31.71 28.67 23.1X I x.34 22.27 26.14 2X.54

6.15 9.52 12.06 12.5X 34.13 3X.47 30.05 21.96 34.59 23.78 26.92 73.32 31.53 32.0 I 34.57 24.24 I X.30 25.42 27.91 19.26 11.69 11.80 10.59 10.06

91.13 88.33 90.64 91.X9 97.71 102.56 69.55 76.X3 92.95 X4.66 77.44 Xl.94 71.38 79.64 12.17

103.23 91.55 72.43 91.09 72.57 70.X8 52.15 76.79 75.00 68.60 63.90 62.64 59.26 JO.24 43.49

COUN - country; for the country’s acronyms see footnote 3

Table 3 Mean and standard deviation of the groups 1960 and 1985

Variable Year AICS ICS NICs General ..-__.... _~__~__~

Mean 1960 145.22 71.10 23.71 X4.96

SD 55.86 26.7X IS.24 65.4 1 Mean 1985 149.44 76.53 .._ 73 ._ 76 85.19 SD 32.36 16.75 7.75 57.X6

Year -.

I960

19x5

AICs ICS NICs

147.34 51 .I6 21.29

160.95 64.50 23.82

inside the NIC group without any qualitative increase in relation to world industrial production.

neighing the manu~cturing index by the share of the population to the totai of the sample of each country, i’ it is observed that the mean of the AICs rises substantia~iy. Table 4 shows the weighted means.

In 1960 the ICs had a performance of 35.13% of the AICs, while the NICs were at a level of 14.45%. In 1985 those features were, respectively, 40.07% and 14.80%. Although the gap between the NICs and the AICs has remained roughly constant, the gap between NICs and ICs increased during this period. Hitherto the general results support the hypothesis of convergence to the mean of each group and between AICs and ICs. However, at the country level the performance is very diverse and the AICs are not converging to the 100 which is the index that represents a share of manufacturing equal to its correspondent share of population.

5. Why growth rates differ in manufacturing

The two previous sections discussed concentration and productivity convergence of manufacturing value added. This section discusses and tests the technology gap approach to the growth of MVA per capita. The periods considered in the deter- mination of rates of growth foliow Fagerberg [3,4] with an inclusion of a period from 1953 to 1960 (the first data used in Fagerberg’s studies) and an extension to the Iast period, 197991983, to 19X5. The compound annual growth rates per country

1 ’ The share of population to total by group are: AlCs 5X.40% (l9hO)and 52.34% (1985); ICs 16.85% (1960) and 15.15% (19X5);

and NlCs 24.75% (1960) and 32.51% (1985).

P. Lowe, E. Fernandesllnr. J. Production Economies 34 (1994) 139-149 145

Table 5 Compound annual manufacturing percentage growth of coun-

tries per period

COUN 53160 60168 68173 73179 79185 53185

AlCs

unit

cana

swit

denm

germ fran

japa swed

belg fin1

ukin

norw

ICS

neth

aust

ire1

isra

ital

atra

nzea

sing

spai

hong

NICs

kore

gree mexi

sout

port

mala

arge

chil

urug

braz

pana

0.72

0.29

5.42

8.89

2.18

21.36

7.38

4.48

4.68

3.36

3.54

4.67 2.91 1.63 1.13 2.30

5.94 3.99 1.41 0.59 2.94

2.93 4.36 - 0.82 1.09 1.86

4.31 5.19 0.90 3.38 3.86

4.37 5.37 1.57 0.75 4.21

4.15 6.81 1.50 - 1.02 2.64

12.61 9.71 0.87 2.79 7.76

5.24 4.09 0.02 1.49 3.16

3.60 6.22 1.56 0.29 3.18

5.42 8.69 2.43 3.43 4.82

2.41 2.46 - 0.60 - 1.01 1.40

4.27 4.21 - 0.30 0.16 2.46

5.58

6.95

3.90

4.93

7.5 1

5.48

4.03

1.37 -

4.70 5.43

3.82 8.10

6.30 4.80

7.42 7.52

7.11 3.87

3.43 3.24

4.70 4.2 1

13.21 16.02

10.03 10.51

14.11 15.85

6.36 6.84 3.71 5.31

0.94 5.20

7.76 7.88

- 0.55 - 0.77

6.68

6.55

3.62

2.57

- 0.47

3.02

9.79

1.24 1.09 3.66

2.51 1.95 4.57

3.26 3.51 4.44

2.04 0.69 4.56

2.26 0.11 4.43

- 1.18 - 3.27 1.67

- 0.91 0.24 2.60

6.86 1.91 8.23

1.66 0.16 5.32

9.20 - 6.59 1.00

17.72 20.68 6.95 14.91

12.36 3.57 - 0.33 5.57

4.20 3.01 - 0.11 3.25

3.40 0.29 - 1.74 1.74

11.51 2.51 3.26 6.50

10.34 7.72 1.94 6.39

5.80 - 0.81 - 5.38 0.74

1.38 - 2.79 - 0.98 0.12

1.04 5.10 - 5.03 -0.16

9.74 4.23 - 4.28 3.20

4.42 - 0.50 - 1.06 3.76

COUN - country; for the country’s acronyms see footnote 3.

are calculated for inputs and output of the model for 1953-1960, 196&1968, 1968-1973, 1973-1979 and 1979-1985. Table 5 shows the manufacturing growth rates.

The growth rates have significant differences from one period to another. With the exception of a few NICs, Table 5 shows that for the passage from the period 1968173 to 1973179 there was

a sharp drop for the MVA per capita growth rates of countries. The last period, 1979185, shows a diffi- cult period for most of the NICs, which produced negative MVA growth rates.

Eq. (3) represents the basic technological gap model to explain why growth rates differ between countries. This paper tests the basic production function in the growth model, i.e. MVA per capita growth explained by the growth of the investment (component of GDP) per capita R&D expenditure per capita and patent applications of non-residents per million. It also tests various hypotheses about these inputs. Taking and rearranging Eq. (3) the basic growth model might be written as

qp = zc, + pn, + cd,, (7)

where qp is the geometric annual growth rate of manufacturing value added per capita, cp is the geometric annual growth rate of investment as share of GDP (US $1980) per capita, rrp is the geometric annual growth rate of R&D expenditure (US $1980) per capita and d, is the geometric annual growth rate of patents application by non- residents per million persons.

This investigation tests the Fagerberg [4] hypothesis about the reducing effect of diffusion. Fagerberg’s Eq. (4) means that when a country diminishes the gap to the leader the diffusion effect of a country’s overall growth decreases; otherwise, a large gap can be seen as an opportunity to take advantage of diffusion. The gap, in this paper, is represented by the relation between the GDP per capita of the United States to that of a country at the start date from which the growth rate of the manufacturing output for a period was calculated. Then, the multiplicative effect of the gap can be assessed by considering diffusion weighed by multi- plying this input rate of growth by the gap of the period minus unity, which in the case of the United States will become zero. This multiplicative effect of the gap over diffusion might be interpreted as a rec- ognition that the diffusion elasticity of the output growth depend on the country’s gap.

The concept of weighing by the gap was also applied to innovation and investment. Investment and R&D expenditure per capita of AICs in relation to their NICs counterparts indicate that there is also a gap effect for these variables. This suggests

that innovation has an increasing elasticity from NICs to AICs while investment has a decreasing elasticity. Therefore, the innovation proxy is multiplied by the inverse of the gap while investment is multiplied by the gap. These hypotheses are different from that for the diffusion proxy because they do not subtract a unit from the gap or inverse of the gap which nullifies the effect of the variable for the leader. Consequently. diffusion, innovation and investment are also tested weighed by the gap as follows:“2

d,, = d, (gap - 1). (8)

cpw = cp kW3 (10)

where ii,,, is the diffusion per capita weighed by the gap. npw is the innovation per capita weighed by the

gap, cpw is the investment per capita weighed by the gap and gap is the GDP per capita of the US divided by the GDP per capita of a country.

Table 5 shows a sharp decline of the rates of growth after 1973. This decline suggests the use of a dummy variable to absorb the circumstantial effect of the first oil crisis. The dummy variable is defined as

073 = 1 from 1950 to 1973, otherwise = 0.

The variables are introduced into the basic technological gap model step by step. Table 6 shows the results of the regression equations for the explanation of the growth of the countries’ MVA per capita. The input coefficients in Table 6 are significant at a 1% level (one-tail test), except for Model 4 where the innovation proxy is significant at a 14% level (one-tail test) and Model 5 where the innovation proxy is significant at a 2.5% level (one- tail test). Model 3 tests the Fagerberg hypothesis (Eq. (8)) by the consideration of gap weighed diffusion minus unity as a component of the model. The level of explanation of the equation (Model 3) in- creases from a R2 of 0.64 (Model 1) to 0.72. This

I2 Eq. (8) curresponds to Fagerberg’s hypothesis Ey. (4)

result supports the catch-up hypothesis, confirming that a laggard country might have a bonus in its growth given by its gap to the leader. Amongst the five models tested, Mode1 3 presents the best results. However, in this model the innovation and investment elasticities are constant and independent from the gap. One might argue that in the same way as a laggard country could gain from diffusion. a leader country could take advantage from innovation. Also investment could have a higher positive effect on productivity in NICs than in AICs. These two additional hypotheses, embodied in Eqs. (9) and (lo), were aggregated into the model step by step to produce Model 5. in this model input elasticities change proportionally to the country gaps. There is a kind of input trade-off. While a country diminishes its productivity gap to the leader, its input elasticities also change: investment and diffusion have decreasing elasticities and innovation has an increasing elasticity.

From the total sample (I 57 cases), 70 cases are NICs with an average gap of 3.937,56 cases are ICs with an average gap of 1.692 and 3 1 cases are AICs with an average gap of 1.236. Except the United States which is the leader country, the average gap of all large countries in the sample are above the average gap of their respective groups. For example, Brazil was 5.917, Italy was 2.193 and the United Kingdom was 1.522.

The results confirm that innovation contributes more than diffusion to MVA per capita growth for ICs and AICs, while diffusion contributes more to the MVA per capita growth than innovation in NICs. However, the level of explanation of the growth Mode1 5, expressed by the R2, is lower than previous models, Nevertheless, this model gives a good idea of the structure of a technological gap model for countries at different levels of industrialization. Table 7 shows the Model 5 coefficients, considering the average gap for each group of countries.

The transformation of the coefficients by the gaps suggests that the weight of the inputs of the technologi~l gap model depend on the level of industrialization of a country. The low coefficients of the diffusion indicator can be interpreted as a result of restrictions in the international diffusion of technology. The restrictions are in the sense that

P. Lowe. E. Fernandesllnt. J. Production Economics 34 11994) 139-149 147

Table 6

Growth model results

Variables and Model 1 Model 2 Model 3 Model 4 Model 5

statistics

c P 0.4643 1 0.42338 0.41236 0.41836 (12.36) (10.89) (12.01) (11.87)

% 0. I2448 0.12150 0.09527

(3.864) (3.875) (3.313)

d, 0.06778 0.05287

(3.377) (2.628)

c PW 0.08050

(10.41)

npw 0.09 123 0.14935

( 1.499) (2.306)

d,, 0.02550 0.02856 0.02080

(6.463) (7.134) (4.655)

D73 1.51924 1.66161 I .7069 I 2.7 1305

(3.096) (3.845) (3.844) (6.130)

const. 1.33928 0.65526 0.61090 0.88997 0.67288

(4.151) (1.707) (1.795) (2.482) (1.765)

RZ 0.6406 1 0.66194 0.72277 0.70707 0.67067

RSE 2.70820 2.63525 2.38641 2.45304 2.60100

RSE-Regression standard error. Between parentheses the t-statistics.

Table 7

Model 5 coefficients ~ average gaps

Cluster d, gap

NICs 0.31693 0.03793 0.06109 3.937

ICS 0.13621 0.08827 0.01439 1.692

AICs 0.09950 0.12083 0.0049 I 1.236

NICs and ICs are not acquiring competitive tech- nologies to improve their productivity.

6. Conclusions

The theoretical framework explaining the differences of productivity between countries based on Schumpeterian logic has been mainly tested by the studies of “why growth rates differ”. These studies do not clearly separate the input indicators of the technology gap approach in their statistical tests. Fagerberg [3, 43 is an exception, but his definition still leaves room for a more detailed specification, particularly for diffusion and the gap.

This paper explored several features of the con- trast of MVA per capita between countries. The shares of the countries in the sample show a high concentration of the MVA in the AICs. However, good the prospects for investment and diffusion, expressed by their coefficients, for the NICs, the unbalance of inputs between countries does not help the achievement of a more balanced MVA per capita. Investment plays a fundamental role for all countries; however, it is particularly important for the NICs. The effect of innovation on the MVA per capita growth in the AICs reflects their better exploitation of R&D expenditure. Conversely, some NICs are having di~culty in following the technological dynamic of the AICs. Amongst the NICs the manufacturing performances were quite diverse: Asian and European countries have shown signs of catching up; some Latin American states have fallen behind, while others stay under- developed.

The growth model tests support the assertion about the crucial role of technology in the process of economic growth based on the Schumpeterian logic. The assumption that technology is not a free

148 P. LOW. E. Femandes/lnt. J. Production Economics 34 (1994) 139-149

good and that it depends on a country’s effort to acquire it has been confirmed throughout this investigation. The individual relevance of each input was established for three groups of countries each representing a different level of industrial development. For the AICs the growth model caught the vital role of innovation on the MVA per capita. The results support Fagerberg’s hypothesis about the effect of the gap on diffusion. The link between the inputs suggests that although a country can em- phasise one aspect of its manufacturing development, if it wishes to maximise its overall economic performance it must have a reasonable balance between inputs to achieve stable development.

The convergence hypothesis was partially supported by the results of this investigation. The results particularly support the convergence between ICs and AICs and within clusters themselves. The negative performance of the Latin American NICs is neutralizing the effect of the manufacturing productivity growth of the European and Asian NICs on the mean of their group manufacturing index.

This investigation confirms that an open economy is very important for achieving a high level of manufacturing productivity. It is no less important for the innovative effort of a country, particularly for countries in the NIC and IC groups. However, the assimilation of knowledge from abroad will contribute to a more efficient use of the internal resources. Also, the effort in innovation has its component in the imitation capability of a country. The high level of the productivity of investment is a very encouraging feature for the NICs and, at a lower level, for the ICs.

When we consider “economic effort”, “innovation” and “diffusion”, it is important to have in mind what these factors represent in the socio- economic environment. The fundamental restruc- turing of the “socio-institutional framework” needed by a country to take advantage of the next economic cycle suggested by Perez [12] is roughly represented by the inputs of the technological gap model. The scientific and technical skills necessary for a country to recognise opportunities and to design adequate strategies as well as the social conditions stressed by Perez and Soete [13] are strongly related to the inputs of the model analysed in this investigation. Certainly, in an

operational sense, it is not only the increase of investment to GDP or the expenditure on R&D or the easier acquisition of patents from abroad that makes for an efficient development policy. The growth model represents only roughly the set of dynamic forces of an economic system. Neverthe- less, it makes a significant contribution to the de- bates on national economic policy.

Appendix

A.1. Data base

The indicators of manufacturing, socio-economic and technological activities used in this investigation were collected from United Nations Statistics. Another important source of data were the statistics published by Summers and Heston [14]. In- formation on 33 countries is available from these sources for the period 1950-1985. Because data on various items were found at different currencies and with incomplete time series, the data were transformed to 1980 US dollar and a procedure to estimate missing data adopted. Details of the estimations of missing data are available on request to the authors.

A.2. Acronyms and initials

AICs Advanced Industrialised Countries EE/UFRJ Escola de Engenharia/UFRJ COPPE Coordena@o de Programas de Pbs-

GraduaCao em Engenharia/ UFRJ COUN Country GDP Gross Domestic Product MVA Manufacturing Value Added NICs Newly Industrialising Countries R&D Research and Development ICS Industrial countries UFRJ Universidade Federal do Rio de

Janeiro us United States of America S US Dollar

A.3. Sources qf‘variables-period 19X-1985

(1) Yearbook cf Industrial Statistics, United Nations.

P. Lowe, E. Femandes/Inr. J. Producfion Economics 34 (1994) 139-149 149

(2)

(3)

(4)

(5)

International Financial Statistics, International Monetary Fund, Yearbooks and Supplements. Monthly Bulletin of Statistics, Department of International Economic and Social Affairs, Statistical Office, United Nations. Industrial Property Statistics, World Intellec- tual Property Organization, Yearbooks. Statistical Yearbook, United Nations Educa- tional, Scientific and Cultural Organization (UNESCO).

References

Cl1

c21

c31

M

c51

C61

Baumol, W.J., 1986. Productivity growth, convergence and

welfare: what the long-run data show. American Eco-

nomic Review, 76.5: 1072-1085.

Abramovitz, M., 1989. Thinking About Growth and Other

Essays on Economic Growth and Welfare. Cambridge

University Press, Cambridge.

Fagerberg, J., 1987. A technology gap approach to why

growth rates differ. Res. Policy, 16: 87-99.

Fagerberg, J., 1988, Why growth rates differ, in: G. Dosi et al. (eds.) Technological Change and Economic Theory.

pp. 432457.

Posner, M.V., 1961. International trade and technical

change. Oxford Economic Papers, 13: pp. 3233341.

Vernon, R., 1966. International investment and interna-

tional trade in the product cycle. Quart. J. Econom., 80.2:

19@207.

[7] Cornwall, J., 1977. Modern Capitalism: its Growth and

Transformation. Martin Roberson & Company Ltd., London.

[S] Dowrick, S. and Gemmell, N., 1991. Industriahsation. catching up and economic growth: A comparative study

across the world’s capitalist economies. The Econom. J.

101: 2633275.

[9] Katrak, H., 1988. R&D, international production and trade: the technological gap theory in a factor-endowment

model. The Manchester School, 56.3: 2055222.

[IO] Krugman, P., 1979. A model of innovation, technology transfer and the world distribution of income. J. Political

Economy, 87: 2533266.

[ll] Krugman, P.R., 1990. Rethinking International Trade. The MIT Press, MA.

[12] Perez, C., 1985. Microelectronics, long waves and world structural change: New perspectives for developing coun-

tries. World Development, 13.3: 441463.

[13] Perez, C. and Soete, L., 1988. Catching up in technology: Entry barriers and windows of opportunity, in: Dosi, G.,

Freeman, C., Nelson, R., Silverberg, G. & Soete, L. (eds.

1988) Technical Change and Economic Theory. Publisher

Limited, London, pp. 46&479.

[14] Summers, R. and Heston, A., 1988. A new international

comparisons of real product and prices levels estimates for

130 countries. 195C-1985. The Rev. Income Wealth, 34.1,

l-25.

[IS] Dosi, G., Freeman, C., Nelson, R., Silverberg, G. and Soete, L., 1988 (eds.). Technical Change and Economic Theory. Publisher Limited, London,

[16] Freeman, C. and Perez, C., 1988. Structural crises of ad- justment, business cycles and investment behaviour, in: G.

Dosi et al. (eds.) Technological Change and Economic

Theory. 38866.

the growth and convergence of manufacturing productivity in industrial and newly industrialising...

Documents