Essays in cross-country economic growth

MISSING IMAGE

Material Information

Title:
Essays in cross-country economic growth
Physical Description:
Unknown
Creator:
Baradaran-Shoraka, Hamid-Reza, 1953-
Publication Date:

Record Information

Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 28572060
oclc - 27477159
System ID:
AA00020471:00001

Full Text












ESSAYS IN CROSS-COUNTRY ECONOMIC GROWTH


BY

HAMID-REZA BARADARAN-SHORAKA















A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


1992














ACKNOWLEDGEMENTS

This dissertation could not have been successfully

completed without the support and guidance of Professor James

D. Adams. I wish to thank him for his useful comments,

insights and faith in me which kept me on track whenever I was

faltering. I would also like to thank Professors Mark Rush and

Prakash Loungani for their invaluable comments and

suggestions. I am grateful also to Professors David Denslow

and James Seale for serving on my dissertation committee and

for their help. I am thankful also to Professors Lawrence

Kenny and William Bomberger for their comments.

I also appreciate the support of the Division of

Sponsored Research of the University of Florida in my final

year of study.

Finally, I could not have endured the Ph.D. program

without continual support from my family. Without the patience

and understanding of my wife Mansoureh, I could not have

either started or successfully concluded my doctorate. I thank

my own and my wife's parents for all the support they have

given me over the years. I also thank my brothers-in-law Ali-

Reza, Mohammad-Reza, and Ali for their encouragement

throughout my education. Last, the presence of my lovely sons

Majid, Mohammad, and Massoud kept me going during these years.















TABLE OF CONTENTS

ACKNOWLEDGEMENTS ........................................ ii
ABSTRACT .................................................. iv

CHAPTERS

1 INTRODUCTION ......................................... 1

2 OIL PRICE CHANGES AND GNP FLUCTUATIONS ................. 7

Introduction .............................................. 7
Trends in Macro Activities ........................... 10
Consumption of Oil by Different Sectors: OECD Evidence 11
Model and Methodology ................................ 12
Empirical Results .................................... 17
Conclusion ........................................... 20

3 HUMAN CAPITAL AND GOVERNMENT POLICIES ................ 30

Introduction ........................................... 30
Data and Expected Results ............................ 34
Empirical Results .................................... 37
Interpretation of the Findings ........................ 49
Conclusion ........................................... 52
Notes ................................................. 61

4 TECHNOLOGY CREATION AND TECHNOLOGY TRANSFER .......... 63

Introduction ........................................... 63
Description of the Data .............................. 68
Empirical Results .................................... 73
Interpretation of the Findings ........................ 84
Conclusion ........................................... 85
Notes ................................................. 96

5 SUMMARY AND CONCLUSIONS .............................. 97

APPENDICES

A DEFINITIONS OF VARIABLES IN CHAPTER 2 ................ 102
B DEFINITIONS AND METHOD OF CALCULATION OF MEAN YEARS OF
SCHOOLING IN CHAPTER 3 ............................... 104

REFERENCES .............................................. 108
BIOGRAPHICAL SKETCH ..................................... 112


iii














Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



ESSAYS IN CROSS-COUNTRY ECONOMIC GROWTH

By

Hamid-Reza Baradaran-Shoraka

August 1992

Chairman: Professor James D. Adams
Major Department: Economics

This dissertation consists of three essays that explore

a variety of factors affecting economic growth across

countries.

The first essay deals with fluctuations in oil prices.

These have trend effect as well as cyclical effects on growth.

Data for ten OECD countries shows that the relative share of

a country's transportation sector has a major impact on that

nation's petroleum dependency. More important, I find that the

transport sector's oil price elasticity is a key determinant

of growth.

The second essay analyzes why countries grow at different

rates from an investment perspective. From my empirical work

it appears that nations with higher initial incomes grow more

slowly than countries with initially lower incomes. This might

iv








seem to imply income convergence. However, I find that the

initial stock of human capital and its subsequent growth are

associated with faster growth which somewhat attenuates

convergence. As in recent growth models, I find that

government consumption, political instability, and higher

rates of population growth are associated with slower growth.

The third essay studies technology's contribution to

growth. In advanced economies, consistent with their

comparative advantage in R&D, domestic technology proxied by

resident patents strongly influences growth. However, my

results suggest that imitation by less advanced countries of

technologies developed in those more advanced is the principal

engine of growth in these nations. I also find that in

advanced countries, inventions per scientist and engineer

raise the rates of growth which is consistent with the

literature on invention exhaustion. Finally, I find that the

potential for enhanced division of labor through interrelated

growth of advanced and follower countries, is associated with

higher growth in the advanced countries, but that technology

acquisition in the newly industrializing countries is

associated with deceleration of growth in the advanced

countries.














CHAPTER 1

INTRODUCTION

The substance of this dissertation is contained in three

independent but related essays comprising chapters 2, 3, and

4. All deal different factors impinging on economic growth

across countries.

This dissertation considers one determinant of

cyclicality of growth, oil price shocks, and two key

determinants of long-term growth: the accumulation of capital

and improvements in technology. Capital here broadly

encompasses physical capital (machines and buildings), as well

as human capital (improvements in the quality of labor force

due to education and training). We shall see that the

accumulation of capital is an important element of growth. We

shall also find that technological advance plays a crucial

role in sustaining growth in the long run. In addition, the

dissertation studies the impact of oil price shocks on growth.

The data that I use on growth rates of real per capital

income in 114 countries for the period 1960 to 1985 is based

on Summers and Heston (1988). As a result of their work, two

facts about world economic growth have emerged. First, growth

in per capital income has occurred continually in many

countries over long periods of time. Second, this performance








2
has varied enormously across countries and over time. Also,

Maddison (1982) looked over still longer periods and finds the

same sustained growth.

How do the above facts, sustained positive growth rates

that vary systematically from country to country, bear on

theorists' attempts to explain the process of economic growth?

Classical writers like Mill and Marx speculated that standards

of living could not rise indefinitely unless advances in

technology helped augment the productivity of resources. This

proposition received support from neoclassical growth

theorists, who built models based solely on capital

accumulation. In the latter models production of output was

characterized by diminishing returns to capital and a steady

state ratio of capital per head. This suggests that net

investment per capital goes to zero in the absence of

technological progress. But the fact that investment has

continued for more than two centuries since the industrial

revolution implies that technical change has played a crucial

role in the growth process.

In the more recent literature, Paul Romer (1986, 1990)

stresses the role played by the accumulation of disembodied

knowledge, as opposed to human capital. In his 1986 paper, he

proposes a model in which growth takes place because the

production function is subject to increasing returns to scale

through knowledge spillovers. In his later paper, Romer

explicitly allows for a permanent effect on the rate of growth








3

stemming from the stock of initial human capital. His model

suggests an empirical "fanning out" of incomes among

countries: very low levels of human capital result in very

slow growth in underdeveloped economies; very high levels of

human capital cause very rapid growth in developed economies.

The same reasoning suggests that pooling of human capital

would raise world growth rates.

Lucas (1988) takes a different approach to the role of

human capital in growth models. He postulates both an internal

and external effect of human capital. The internal effect is

the impact of an individual's human capital on his private

marginal product. The external effect, which is identified

with the country's average level of human capital, contributes

to the productivity of all persons and factors of production.

This model implies strictly parallel growth paths: economies

that are initially poor will remain relatively poor, though

their long-run rate of income growth will be the same as that

of initially and permanently wealthier economies.

Both Romer's (1990) fanning out results and Lucas's

(1988) parallel growth path results ignore technology

transfer, which can result in convergence, if follower

countries are "activist" enough in their education and

technology policies.

Evenson (1984), basing his analysis on data on patented

inventions from many countries, reaches two principal

conclusions: first, the data show comparative advantage








4

patterns in invention similar to patterns observed in

production. The production of pioneering invention is

concentrated in certain firms located in countries with the

best laboratories. Large parts of industry in most countries

import inventions and concentrate on adaptive invention rather

than investing heavily in R&D. Second, the data show that

inventions per scientist and engineer have declined from the

late 1960s to 1970s in almost all of the 50 countries for

which data are available, which may suggest diminishing

returns to invention in the present era.

Grossman and Helpman (1991) make much of the fact that

countries vary greatly in their growth performances. They

point out that a reading of recent economic history suggests

two important trends. First, technological innovations are

ever more important contributors to growth. Second, nations

are becoming increasingly open and interdependent. The two are

not unrelated. Grossman and Helpman suggest that rapid

communication and close contacts among innovators in different

countries facilitate the process of invention and the spread

of new ideas.

Finally, Krugman (1979) in his famous model of product

cycle divided countries into innovating North and

noninnovating South. Innovation consists of the development of

new products, which can be produced at first only in the

North, but eventually the technology of production is

transferred to the South. This technological lag gives rise to








5

trade, with the North exporting new products and importing old

products.

The dissertation is built on these themes. The first

essay in the dissertation focuses on government policy towards

the transportation sector. I study the interaction between oil

price changes and the share of petroleum-based transportation

in 10 OECD countries, in order to determine if this has a

significant effect on the macroeconomic impacts of oil price

shocks.

The second essay sets out to test the effects of human

capital on growth. I consider how initial levels of human

capital, growth of human capital, government consumption,

investment, political instability, and public infrastructure

influence economic growth. I do so using a sample of 50

countries during the quarter century from 1960 to 1985. I have

assembled a rich data set that includes relevant variables

that have not previously been put to use.

The third essay uses patent data to study relative the

contribution of technology creation and technology transfer to

growth. Growth is fueled by both innovation and imitation.

Imitation by newly industrialized countries (NICs) of

technologies developed in the advanced countries turns out to

be the main source of growth. A related point is that primary

research in advanced countries can be imported by less

developed countries (LDCs), thereby permitting adaptive

invention, rather than expensive original research. To examine








6

these ideas, this essay develops measures of imitation and

adaptation.

It is clear that accumulation of human capital by LDCs is

essential to the successful acquisition of foreign

technologies. Therefore, a close link exists between growth of

human capital and successful technological imitation. This may

well explain another part of the variation in growth among

nations, but it is beyond the scope of this thesis.

Chapter 5 summarizes the dissertation and explores the

possibilities for further work in this rapidly advancing and

important area of research.














CHAPTER 2

OIL PRICE CHANGES AND GNP FLUCTUATIONS

Introduction

The oil price shocks of the seventies seem to dominate

the fortunes of the industrialized countries. Before 1973 most

of these countries had experienced healthy economic growth and

tolerable levels of inflation. By 1974, all these economies

were in recession, with double-digit inflation plaguing all

but West Germany and negative growth afflicting the United

States, Japan, and the United Kingdom. Inflation slowed and

growth resumed by 1978. However, the oil price shock of 1980

appears to have brought renewed stagflation.

Several writers have suggested that oil price shocks

caused the recessions of the 1970s. Hamilton (1983) suggests

that oil prices Granger-caused output and unemployment even in

the period 1947 to 1972 as well as afterwards; evidence on the

exogeneity of oil prices is provided in Hamilton (1983, 1985).

A connection between disruptions in the energy sector and the

economy is documented in Rasche and Tatom (1981), Santini

(1985) and Gisser and Goodwin (1986).

Other empirical work attempts to determine channels

through which oil shocks affect macroeconomic activity. Mork

(1988) provides a useful taxonomy of these efforts. Rasche and








8

Tatom view the shocks as aggregate supply shocks which affect

potential GNP, while Pierce and Enzler (1974) emphasize the

role of aggregate demand effects in propagating these shocks.

Gordon (1975), Mork and Hall (1981) and Mork (1985) assert the

importance of both aggregate supply and demand to the effects

of oil shocks.

Another view is motivated by the findings of Lilien

(1982), Davis (1985) and Loungani (1986a, 1986b). Lilien found

that the dispersion of employment growth across industries, a,

is positively correlated with aggregate unemployment. Davis

and Loungani subsequently demonstrated that oil price shocks

are the predominant source of movements in a. That finding is

consistent with a sectoral reallocation story oil price shocks

alter relative prices across different sectors of the economy.

The sectoral reallocation of production and employment entails

temporary unemployment and recession. Recent work by Hamilton

(1988) strengthens the theoretical underpinnings of this view.

This chapter provides cross country evidence on the

relationship between oil price changes and real GNP. The same

issue is examined in Darby (1982) and Burbidge and Harrison

(1984), but with two differences. First, previous work has

examined each nation in isolation. I use pooled data from 1960

to 1987 for a Group of Ten OECD countries (Group of Seven plus

Australia, Netherlands, and Spain). Second, I attempt to

explain why the impact of oil prices on real GNP differs

across countries.








9

The basic hypotheses that I wish to test are

(i) that there is a negative correlation between oil price

changes and real output,

(ii) that the price elasticity for the transportation sector

is smaller in absolute magnitude than for the other sectors,

and therefore

(iii) that the impact of oil price changes on macro activity

depends on the share of transportation in oil consumption.

This explanation is distinct from the sectoral

reallocation. The second and third hypotheses contain another

interpretation of oil price shocks. They have been suggested

by Danilo Santini, who presents some evidence in favor of

them. He points out there is evidence in favor of the view

that the price elasticity of transportation is relatively low:

The statistical results ... tend to confirm the
present descriptive analysis of the relative
difficulty of substituting for oil in
transportation. The estimates for both short and
long-run elasticity for transportation are
substantially smaller in absolute magnitude than
for the other sectors. (p.7)

Santini suggests the importance of share of the

transportation sector in total oil consumption in the

industrialized countries:

The fundamental arguments of this paper can be
summarized as follows:
-Because of its inability to economically and
rapidly substitute nonpetroleum fuels ...,
transportation historically has been less able to
reduce petroleum consumption than have other
sectors of industrialized economies.
-Because of transportation's fuel inflexibility,
industrialized nations that devoted a greater share
of their total petroleum consumption to








10

transportation had greater difficulty reducing oil
consumption after the 1978-1981 crude oil price
run-up. (p. 10)

The remainder of this chapter is organized as follows.

The second Section summarizes trends in macroeconomic activity

in our 10 countries, while the third Section presents data on

differences in the consumption of oil by sector in these

countries. The fourth Section specifies the regressions.

Empirical results are presented in the fifth Section, while

the last Section is a summary and conclusion.

Trends in Macro Activities

The growth rate has fallen over the last 20 years in the

advanced countries. Productivity growth in these countries

averaged 4.0 percent per year during 1960-68, 3.1 percent

during 1968-73, 1.5 percent during 1973-79, and 1.6 percent

during 1979-87. In each of the Group of Seven countries,

productivity growth during the 1970s and 1980s was about 50

percent less than that attained during 1960s. Growth in real

GDP also fell during the same years. Real GDP growth averaged

5.0 percent per year during 1960-68, 4.4 percent during 1968-

73, 2.7 percent during 1973-79, and 2.6 percent during 1979-

87. Corresponding to the reduction in growth, unemployment

climbed from an average of 2.8 percent during 1964-67 to 3.3

percent during 1968-73 and finally to 5.0 percent during 1974-

79. But there are fairly wide differences in real GDP growth

in the set of countries. Real GDP growth was 3.6 percent per

year in Japan during the years 1973 to 1979, 2.4 percent in








11

the United States, 2.3 percent in Germany, and only 1.5

percent in the United Kingdom (Table 2-1).

The diversity in macro performance among countries calls

for an explanation. I examine the role of transportation in

accounting for the differences.

Consumption of Oil by Different Sectors: OECD Evidence

The impact of oil price changes on macro activity

differed across countries. For instance, Japan, though

dependent on foreign oil to a much greater extent than the

U.S., did not experience a recession as severe as the U.S.

(Table 2-1).

Table 2-2 pursues the transportation cost hypothesis. It

shows that the share of the transportation sector in total oil

consumption differs in a way that is consistent with the

American and Japanese experience. For instance, between 1960

and 1987, Japan had a mean transport share of 29 percent, the

United States 58 percent, Canada 48 percent, France 34

percent, West Germany 31 percent, and the United Kingdom 44

percent. At the same time there was wide dispersion in the

average share of the industrial sector in total oil

consumption among these countries. For example, Japan's

average share was 48 percent, the U.S. 25 percent, Germany 30

percent, and the U.K. 39 percent.

Except for transportation, most sectors seem to be able

to substitute non-oil-based fuels for petroleum products. The

inflexibility of transportation makes its costs more








12

vulnerable to oil price shocks. West Germany and Japan, which

are the nations with the lowest share of oil demand accounted

for by the transportation sector, both managed to reduce total

oil consumption by a greater percentage, and did not

experience a recession in 1980 or 1981. This suggests that the

size of a nation's transportation sector may interact with the

oil price changes to raise costs and retard productivity.

Model and Methodology

Since there is no accepted structural model in this area,

I will formulate an empirical model which is consistent with

earlier work while permits a test of the hypotheses.

First I investigate the effect of oil price shocks on GNP

growth to see whether my result is consistent with pervious

empirical work. I regress real GNP growth on a constant, the

rate of change in the real price of oil, the rate of change in

the money base, and the rate of change in government

consumption. I also include lags of all the independent

variables.

Next I turn to estimation of the price elasticity of

demand in the transport and industry sectors. I accomplish

this by regressing the growth rate of transportation and

industry consumption on oil price changes and other

macroeconomic variables.

Finally I analyze the correlation between GNP growth and

the interaction of oil price changes and mean share of

transportation's consumption of oil. I examine this relation











Table 2-1
Key Macroeconomic Indicators
(Year to year percentage changes)

Average Average Average Average Average
1960-68 1968-73 1973-79 1979-87 1960-87

REAL GDP:
United States 4.5 3.2 2.4 2.6 3.2
Japan 10.2 8.7 3.6 3.8 6.5
Germany 4.1 4.9 2.3 1.4 3.0
France 5.4 5.5 2.8 1.7 3.7
United Kingdom 3.1 3.3 1.5 1.8 2.4
Italy 5.7 4.5 3.7 2.2 4.0
Canada 5.5 5.4 4.2 2.9 4.4
Total of above
countries 5.0 4.4 2.7 2.6 3.7
Total OECD 5.0 4.5 2.7 2.5 3.7

REAL GDP PER PERSON
EMPLOYED:
United States 2.6 1.0 0.0 1.0 1.2
Japan 8.5 7.6 2.9 2.9 5.4
Germany 4.2 4.1 2.9 1.5 3.1
France 4.9 4.3 2.5 1.9 3.4
United Kingdom 2.7 3.1 1.3 1.8 2.2
Italy 6.3 4.9 2.8 1.8 3.9
Canada 2.6 2.5 1.3 1.1 1.8
Total of above
countries 4.0 3.1 1.5 1.6 2.6
Total OECD 4.0 3.3 1.6 1.6 2.6


Average Average Average Average Average
1964-67 1968-73 1974-79 1980-87 1964-87


STANDARDIZED
UNEMPLOYMENT RATE:

United States 4.2 4.6 6.7 7.6 6.1
Japan 1.2 1.2 1.9 2.5 1.8
Germany 0.6 1.0 3.2 6.0 3.1
France 1.7 2.6 4.5 8.9 5.0
United Kingdom 2.5 3.3 5.0 10.5 6.0
Italy 5.1 5.7 6.6 9.5 7.1
Canada 3.9 5.4 7.2 9.7 7.0
Total of above
countries 2.8 3.3 5.0 7.0 4.9
Total OECD 2.7 3.2 4.9 7.5 5.0


Source: OECD, Historical Statistics


1960-1987, Paris, 1989.











Table 2-2


Summary of Oil Consumption by Sectors
(Percentage Share of Total Consumption)

Country/Sector 1973 1974 1975 1976 1977 1978 1979 1980

U.S.A.


Transportation
Industry
Others


JAPAN


Transportation
Industry
Others

GERMANY

Transportation
Industry
Others


25 27
33 33
42 40


FRANCE


Transportation
Industry
Others


CANADA


Transportation
Industry
Others

UNITED KINGDOM

Transportation
Industry
Others


45 48
27 25
28 27


ITALY


Transportation 27 26 29 29 32 34 35 37
Industry 41 42 36 38 38 34 35 33
Others 32 32 35 33 30 32 30 30











Table 2-2--continued


Country/Sector 1981 1982 1983 1984 1985 1986 1987


U.S.A.

Transportation
Industry
Others


JAPAN


Transportation
Industry
Others

GERMANY

Transportation
Industry
Others

FRANCE

Transportation
Industry
Others

CANADA


Transportation 55 54 55 56 57 57 57
Industry 26 23 22 23 24 25 25
Others 19 23 23 21 19 18 18

UNITED KINGDOM

Transportation 52 54 56 58 59 60 62
Industry 30 29 28 26 25 26 24
Others 18 17 16 16 16 14 14

ITALY

Transportation 38 41 41 42 44 46 47
Industry 32 29 30 27 27 26 26
Others 30 30 29 31 29 28 27


Source: Energy Balances


of OECD Countries,


IEA. Paris, 1988.








16

by estimating using a regression model which includes

macroeconomic controls, in order to test whether the effect is

a direct result of the interaction between oil price changes

and the share of transport oil consumption.

I estimated the following regressions:


DGNP = aI + bI DOP + ci DOPi + di DRM + ej DRMi + f1 DGC +

gi DGCi + ki Di + ul i = 1,2,....

DGNP = a2 + b2 MINTER + hi MINTERi + d2 DRM + oi DRMi +

f2 DGC + qi DGCj + li Di + U2 i = 1,2,...

DGNP = a5 + b5 DTINTER + xi DTINTERi + d5 DRM + Yi DRMi +

f5 DGC + zi DGCi + pi Di + u5 i = 1,2,...


where DGNP is GNP or GDP growth, DOP is relative oil-price

changes, DOPi is lagged DOP, DRM is growth of money base, DRMi

is lagged DRM, DGC is change in government consumption, DGCi

is lagged DGC, Di are country dummy variables, MINTER is the

interaction of mean share of transport oil consumption and

relative oil price changes (DOP x MEAN), MINTERi is lagged

MINTER, DTINTER is the interaction of real oil price (ROP) and

share of transport's oil consumption (SHARE), namely (ROP x

SHARE changes), DTINTERi is lagged DTINTER, and ui are the

residual components. We hypothesize that bl, b2, b5, ci, hi,

and xi should be negative. Also, we expect that the absolute

value of b3 < b4 and ri < vi, respectively. All the variables

except ROP, MEAN, and SHARE are in logarithmic form.








17

For the second hypothesis I estimated

DTTR = a3 + b3 DOP + ri DOPi + d3 DRM + si DRMj + f3 DGC +

ti DGCi + mi Di + U3 i = 1,2,...

DTIN = a4 + b4 DOP + vi DOPj + d4 DRM + j i DRMj + f4 DGC +

wj DGCi + ni Di + u4 i = 1,2,...

where DTTR is growth share of transport's oil consumption, and

DTIN is growth share of industry's oil consumption.

Empirical Results

I estimate these equations using pooled data for the

Group of Ten countries over the period 1960 to 1987. Details

of these data are given in Appendix A. In general, the data

provide strong support for the idea that an increase in the

price of oil decreases GNP growth.

First I estimate the price elasticity of demand. Tables

2-3 and 2-4 present the results for the price elasticity of

industry and transportation sectors, respectively, when the

other macroeconomic variables are included in the equation

(for the Group of Seven). As we expected the price elasticity

of the industrial sector is higher than the transportation

sector and both estimated coefficients are statistically

significant at better than a 99 percent level.

Table 2-5 and 2-6 present the results for transport and

industry's oil consumption growth within the Group of Ten. As

with Tables 2-3 and 2-4, column 1 of Tables 2-5 and 2-6 show

the relationship between the transportation sector oil

consumption and DOP and also between industry sector oil








18

consumption and DOP, respectively. Again as we expected, the

price elasticity of industry is greater than the elasticity

for the transport sector. Column 2 of Tables 2-5 and 2-6

substitute GNP growth for the other macroeconomic variables.

In this case, the oil price elasticity for industrial sector

is markedly higher than transport sector. Also, both estimated

coefficient are statistically significant at better than a

ninety-nine percent level.

Thus, the results of Tables 2-3, 2-4, 2-5 and 2-6 are in

accord with the idea that the transport sector is not

flexible, while the industrial sector is much more flexible.

In other words, whenever there is an increase in the price of

oil, the industrial sector can substitute other sources of

energy, but it is almost impossible for the transport sector

to do it.

Table 2-7 contains the basic set of estimated

relationships for the Group of Seven. Column 1 illustrates the

effect of oil price changes on growth. The coefficient on DOP

(relative oil price changes) is statistically significant at

the 94 percent level and the estimated coefficient on DOP1

(lagged DOP) is statistically significant at better than a 99

percent level. An F test rejects the null hypothesis that the

impact of DOP and DOP1 is zero with F(2, 151) = 8.91.

Column 2 changes the DOP variable to MINTER (interaction

of transport mean share and oil price changes) to allow for

the influence of differences in transport share across








19

countries. Compared to column 1 the estimated coefficient on

the first oil variable increased only slightly from 0.011 to

0.028.

Column 3 shows the combined effect of DOP and MINTER. But

the estimated coefficients on DOP and MINTER are not

statistically significant, because of multicollinearity.

Columns 4, 5, 6 and 7 of table 2-7 present analogous

results for the effect of lagged oil price shocks. We see that

there are no significant changes.

Thus, the results of Table 2-7 are congruent with

negative shock interpretation of oil price changes. However,

the transportation cost hypothesis is rejected.

To check for sensitivity of these conclusions to the set

of countries included, Table 2-8 contains a similar set of

estimated relationships between GNP growth and oil price

changes in the Group of Ten. This time, however I have added

another variable DTINTER, which measures the interaction of

real oil price and share of transport's oil consumption.

Column 1 illustrates the effect of DTINTER with three

lags and DOP with three lags again on the GNP growth. The

estimated coefficient of DOP is not statistically significant,

but the estimated coefficient on DOP1 (lagged DOP) is

statistically significant at better than a 99 percent level.

None of coefficients on DTINTER are statistically significant.

Column 2 shows the relationships between DOP and GNP

growth rate. Again the estimated coefficient on DOP is not








20

statistically significant, but DOP1 is significant at better

than 99 percent level. Also, an F test rejects the null

hypothesis that the impact of DOP and DOP1 is zero with

F(2,194) = 7.94.

Column 3 exhibits the effect on GNP growth of only

contemporaneous and lagged DTINTER. Again none of the

estimated coefficients are statistically significant.

Column 7 to 11 of Table 2-8 contain the set of estimated

relationships which are the same as those in Table 2-7 but

this time for the Group of Ten. In column 7 as with column 3

of Table 2-7 none of the estimated coefficients of DOP and

MINTER are significant. In column 8 ,contrary to column 2 of

Table 2-7, the coefficient of MINTER is not statistically

significant. However, even though the estimated coefficient on

MINTER1 is reduced -from 0.055 to 0.048- it remains

significant at better than a 99 percent level.

The results of Table 2-8 agree with the shock

interpretation of oil price changes, but reject the

transportation cost interpretation.

Conclusion

There exists a strong, negative correlation between oil

price changes and GNP growth. Darby (1982) has established

this correlation for eight major OECD countries data over the

period 1973-1975 and Mork et al.(1989) has attained similar

cross-section results for a sample of six major OECD

countries.








21

The contribution of this paper is to expand this list of

results and to offer evidence against the "direct causation"

hypothesis that a high share of transportation of total oil

consumption drags down GNP growth. Tables 2-7 and 2-8 contain

results which establish a negative correlation between the oil

price changes and GNP fluctuations for 10 major OECD

countries. However, the results in these tables do not support

the idea that a higher share of transportation sector lowers

the GNP growth rate in the face of oil price changes.

Tables 2-3, 2-4, 2-5, and 2-6 contain point estimate

results which are consistent with the view that the price

elasticity for transportation sector is lower in absolute

magnitude than other sectors. Further, supporting Santini's

descriptive arguments, the percentage of reduction in oil use

estimated to have occurred through fuel substitution is far

lower for transportation than for the industrial sector.

Finally, Table 2-9 which presents the effect on the

difference between growth of transport's and industry's oil

consumption of DOP and DOP1. This again supports the

hypothesis of a lower price elasticity of demand in the

transport sector.









22

Table 2-3

Industry Consumption of Oil Growth in Group of Seven

(dependent variable DTIN)



Variables Coefficients St. Error Prob>[T[



INTERCEP -0.011 0.029 0.7082

DOP -0.068 0.024 0.0066

DOP1 -0.123 0.025 0.0001

DRM 0.209 0.120 0.0834

DRM1 0.165 0.123 0.1825

DGC 0.204 0.211 0.3342

DGC1 -0.316 0.195 0.1078



R2 0.2524









23

Table 2-4

Growth of Transport Oil Consumption in Group of Seven

(dependent variable DTTR)



Variables Coefficients St. Error Prob>IT,



INTERCEP 0.020 0.011 0.0710

DOP -0.042 0.009 0.0001

DOPI -0.025 0.010 0.0132

DRM 0.137 0.048 0.0051

DRM1 0.086 0.050 0.0880

DGC 0.091 0.086 0.2917

DGC1 -0.054 0.080 0.5008



R2 0.2896











Table 2-5
Transport's Oil Consumption Growth in the Group of Ten
(dependent variable DTTR)

VARIABLE (1) (2) (3)

INTERCEP 0.033 -0.007 0.026
(0.017) (0.010) (0.007)
DOP -0.030 -0.025 -0.028
(0.009) (0.007) (0.009)
DOP1 -0.024 -0.006 -0.027
(0.009) (0.007) (0.009)
DOP2 -0.010 0.010
(0.012) (0.011)
DOP3 -0.025 0.006 -0.026
(0.012) (0.011) (0.011)
DRM 0.068 0.086
(0.042) (0.037)
DRM1 0.046
(0.043)
DRM2 -0.026 -0.014
(0.044) (0.040)
DRM3 -0.082
(0.044)
DGC 0.101 0.095
(0.077) (0.062)
DGC1 0.055
(0.082)
DGC2 0.005
(0.083)
DGC3 -0.034
(0.071)
DGNP 0.668
(0.113)
DGNP1 0.302
(0.118)
DGNP2 0.284
(0.118)
DGNP3 0.178
(0.103)
R2 0.21 0.42 0.17


NOTE: Figures in parentheses represent the standard error.











Table 2-6


Oil Price Changes and Industry's Oil Consumption Growth
in the Group of Ten (dependent variable DTIN)

VARIABLE (1) (2) (3)


INTERCEP

DOP

DOP1

DOP2

DOP3

DRM

DRM1

DRM2

DRM3

DGC

DGC1

DGC2

DGC3

DGNP


DGNP1

DGNP2

DGNP3


0.130
(0.048)
-0.034
(0.026)
-0.090
(0.026)
-0.032
(0.034)
-0.021
(0.034)
0.017
(0.116)
0.086
(0.119)
-0.043
(0.121)
-0.100
(0.123)
-0.277
(0.213)
0.209
(0.226)
0.082
(0.231)
-0.173
(0.197)


-0.078
(0.026)
-0.054
(0.019)
-0.051
(0.020)
0.016
(0.029)
0.047
(0.028)


0.017
(0.019)
-0.043
(0.025)
-0.100
(0.025)


-0.028
(0.031)
0.150
(0.105)


0.091
(0.112)


-0.114
(0.174)


1.812
(0.295)
0.987
(0.309)
0.224
(0.309)
0.920
(0.269)


0.19


0.46


0.12


NOTE: Figures in parentheses represent the standard error.











Table 2-7


GNP Growth and Oil Price Changes in the Group of Seven
(dependent variable DGNP)

VARIABLE (1) (2) (3) (4) (5) (6) (7)


INTERCEP


0.021 0.022 0.022
(0.006) (0.006) (0.006)


0.025 0.025
(0.006) (0.006)


0.029 0.024
(0.006) (0.006)


-0.011 0.001 0.003
(0.005) (0.023) (0.023)

-0.022 -0.008 -0.008
(0.006) (0.023) (0.023)

0.082 0.085 0.084 0.096 0.097 0.095 0.087
(0.029) (0.028) (0.029) (0.028) (0.028) (0.029) (0.029)


0.045 0.045 0.045
(0.030) (0.030) (0.030)

0.042 0.045 0.045
(0.052) (0.052) (0.052)


0.053 0.039
(0.031) (0.030)

0.0006 0.015
(0.052) (0.050)


-0.093 -0.090 -0.058 -0.060 -0.131 -0.088
(0.047) (0.049) (0.042) (0.040) (0.048) (0.048)

-0.028 -0.032 -0.028 -0.019 -0.022
(0.014) (0.057) (0.057) (0.013) (0.014)


-0.055 -0.035 -0.032 -0.052
(0.015) (0.057) (0.058) (0.014)


0.33


0.33


0.31


0.31


-0.053
(0.015)


0.27


0.31


NOTE: Figures in parentheses represent the standard error.


DOP


DOP1


DRM


DRMl


DGC


DGC1


-0.083
(0.048)


MINTER


MINTER1


0.33











Table 2-8


GNP Growth and Oil Price Changes in the Group of Ten
(dependent variable DGNP)

VARIABLE (1) (2) (3) (4) (5) (6)


INTERCEP

DOP

DOP1

DOP2

DOP3

DTINTER

DTINTER1

DTINTER2

DTINTER3

DRM

DRM1

DRM2

DRM3

DGC

DGC1

DGC2

DGC3


0.036
(0.010)
-0.008
(0.005)
-0.021
(0.005)
-0.003
(0.007)
-0.013
(0.007)
0.0001
(0.0001)
-0.00004
(0.0001)
-0.0002
(0.0003)
-0.0001
(0.0003)
0.057
(0.023)
0.028
(0.024)
-0.040
(0.025)
0.016
(0.025)
0.034
(0.043)
-0.067
(0.046)
-0.033
(0.047)
0.036
(0.040)


0.037
(0.009)
-0.005
(0.005)
-0.021
(0.005)
-0.005
(0.007)
-0.014
(0.007)








0.058
(0.023)
0.023
(0.024)
-0.042
(0.024)
-0.017
(0.025)
0.038
(0.043)
-0.064
(0.046)
-0.037
(0.046)
0.033
(0.040)


0.051 0.030
(0.009) (0.003)
-0.010
(0.005)
-0.024
(0.005)
-0.009
(0.006)


0.0001
(0.0001)
-0.0001
(0.0001)
-0.0001
(0.0003)
-0.0002
(0.0003)
0.071
(0.024)
0.032
(0.025)
-0.046
(0.025)
0.017
(0.025)
-0.009
(0.043)
-0.096
(0.046)
-0.058
(0.047)
0.023
(0.041)


0.0001
(0.0001)
-0.00006
(0.0001)
-0.0003
(0.0002)


0.031
(0.004)
-0.006
(0.005)
-0.025
(0.005)


-0.015
(0.006)


0.060 0.056
(0.021) (0.021)


-0.042
(0.022)


0.040
(0.003)








0.00007
(0.0001)
-0.0002
(0.0001)
-0.0003
(0.0002)


0.083
(0.021)


-0.041 -0.034
(0.022) (0.022)


0.039 0.031
(0.035) (0.035)


-0.084
(0.033)


R2 0.34 0.33 0.27 0.27 0.27 0.21


R2 0.34


0.33 0.27 0.27


0.27 0.21











Table 2-8--continued


INTERCEP

DOP

DOP1

DOP2

DOP3

MINTER

MINTER1

MINTER2

MINTER3

DRM

DRM1

DRM2

DRM3

DGC

DGC1

DGC2

DGC3


0.040
(0.009)
-0.014
(0.021)
0.0004
(0.022)
-0.048
(0.029)
-0.056
(0.028)
0.020
(0.048)
-0.050
(0.050)
0.102
(0.066)
0.097
(0.065)
0.061
(0.024)
0.009
(0.024)
-0.047
(0.025)
0.011
(0.025)
0.058
(0.044)
-0.074
(0.046)
-0.048
(0.047)
0.045
(0.040)

0.35


0.039 0.030
(0.010) (0.003)
-0.021
(0.021)
-0.003
(0.022)
-0.047
(0.027)


-0.012
(0.011)
-0.048
(0.012)
-0.003
(0.016)
-0.024
(0.016)
0.062
(0.023)
0.021
(0.024)
-0.039
(0.024)
0.017
(0.025)
0.041
(0.043)
-0.070
(0.046)
-0.047
(0.047)
0.026
(0.040)

0.32


0.028
(0.048)
-0.050
(0.049)
0.082
(0.062)


0.035
(0.005)
0.0002
(0.020)
-0.014
(0.020)




-0.013
(0.047)
-0.017
(0.047)


0.031 0.036
(0.004) (0.008)
-0.009
(0.005)
-0.023
(0.005)




-0.012
(0.011)
-0.056
(0.011)


-0.029
(0.014)
0.062 0.053 0.058
(0.021) (0.023) (0.021)


-0.041
(0.022)


0.040
(0.035)







0.27


-0.048
(0.034)





0.24


-0.041
(0.023)


0.029
(0.035)







0.26


0.061
(0.022)
0.018
(0.023)




0.043
(0.040)
-0.097
(0.038)





0.30


NOTE: Figures in parentheses represent the standard error.


(7) (8) (9) (10) (11) (12)









29

Table 2-9

Difference Between Growth of Transport's and Industry's Oil
Consumption in Group of Seven

(dependent variable DTIS)


Variables Coefficient St. Error Prob>IT{


INTERCEP 0.019 0.025 0.4639

DOP 0.024 0.019 0.2163

DOP1 0.097 0.020 0.0001

DRM 0.010 0.102 0.9177

DRM1 -0.086 0.106 0.4165

DGC -0.241 0.189 0.2055

DGC1 0.371 0.178 0.0394


R2 0.2216














CHAPTER 3

HUMAN CAPITAL AND GOVERNMENT POLICIES

Introduction

Recent theoretical developments, which emphasize the

endogeneity of growth and thus its sensitivity to policy, have

caused economists to recognize that government policies toward

investment in human and physical capital are significant for

growth (Uzawa, 1965; Arrow, 1962; Solow, 1956; Lucas, 1988;

King and Rebelo, 1990; Romer, 1986, 1990). There now exists

some empirical research which has studied the managing of

economic growth. A good example is Barro (1991), who, using

cross-country data, examines several of the factors addressed

in the endogenous growth literature. In this paper, in line

with Barro, I use the insights obtained from recent

theoretical work to investigate empirical relationships

between per capital output growth and investments in human and

physical capital. But, in contrast to previous work, I use new

human capital constructs in order to probe the data more

deeply.

There have been two major types of growth models. Older,

exogenous growth models are characterized by the assumption of

diminishing returns, and so any permanent growth in per capital

income stems from exogenous technological change (Solow,








31

1956). Endogenous growth models generate permanent growth in

per capital income within the system by assuming constant or

increasing returns to reproducible factors of production.

These models allow us to study the effects of government

policies on investment incentives and hence long-run growth.

A brief description of this approach is useful.

Paul Romer (1986, 1990) stresses the role played by the

accumulation of disembodied knowledge, as opposed to human

capital. In his 1986 paper, he proposes a new model in which

growth takes place because the production function is subject

to increasing returns to scale through knowledge spillovers.

The growth of consumption is driven permanently above zero

because knowledge pushes the marginal product of capital

permanently above the rate of time preference.

In his later paper, Romer explicitly allows for a

permanent effect on the rate of growth stemming from the stock

of initial human capital. His model suggests a fanning out of

incomes among countries: very low levels of human capital

result in very slow growth in underdeveloped economies; very

high levels of human capital cause very rapid growth in

developed economies. The same reasoning suggests that pooling

of human capital would raise world growth rates.

Lucas (1988) takes a different approach to the role of

human capital in growth models. He postulates both an internal

and external effect of human capital. The internal effect is

the impact of an individual's human capital on his private








32

marginal product. The external effect, which is identified

with the country's average level of human capital, contributes

to the productivity of all persons and factors of production.

This model has less radical predictions for fanning out:

economies that are initially poor will remain relatively poor,

though their long-run rate of income growth will be the same

as that of initially (and permanently) wealthier economies. A

world consisting of such economies, each operating

autarkically, would exhibit uniform rates of growth across

countries and would maintain a perfectly stable ranking of

income and wealth over time.

This chapter is closely tied to the developments

summarized above, especially those in Romer (1990). For the

first time as far as I know, the role of human capital in

cross-country growth is extensively studied using new and

superior measures that enable us to separate stock and growth

rate effects of human capital. In particular, I have been able

to separate the initial level of mean years of schooling of

the total population, from growth in mean schooling over a

period of three decades.' This allows me to test several

effects of human capital hitherto relegated only to theory.

It is well-known that there are transitional dynamics as

well as steady states associated with economic growth. The

presence of a transition period is why most economists have

employed long-run cross-sections. To "average out" the

transitions and thus concentrate on the steady states, I use








33

pooled time-series cross-section data supplemented by control

variables. I control for transitional effects in these data by

using cyclical variables such as percent deviation of GDP from

rest of the world GDP, investment, government consumption, and

population growth, all of which are year specific. The data

are a sample of 50 countries during the quarter century from

1960 to 1985. This is a rich data set that combines relevant

variables many of which have not been put to use for the

purpose of studying growth.

This chapter has three main concerns. The first is the

connection between the initial stock of human capital, growth

in human capital, and economic growth. As mentioned, I find

that both measures of human capital contribute to a nation's

growth.

My second concern is with convergence of per capital

income. We have seen that there are different and often

contradictory ideas about convergence of income among nations.

My empirical results, in line with Barro (1991), confirm

convergence.

My third concern is with government policies towards

consumption and investment. Drawing on neoclassical growth

models, Barro (1990) predicts a negative relationship between

the size of transfer payments, which he calls government

consumption, and economic growth. He distinguishes between

productive and nonproductive government expenditures and their

repercussions on growth. His findings show a negative








34

correlation between government consumption and growth. I find

the same negative effect of government consumption on economic

growth.

The second section discusses the important new data on

human capital and describes all the variables. Empirical

results are reported in the third section. The fourth section

draws implications of the findings for cross-country growth.

The final section is a summary and conclusion.

Data and Expected Results

The GDP data and its components are from Summers and

Heston (1988). These data include real GDP, government

consumption, investment, and population for 130 countries

outside the Eastern Bloc. The data are annual and cover the

period 1960 to 1985. The other data come from the United

Nations, the World Bank, Banks (1990), Barro (1991), and other

sources. Sources of variables used in this paper are

documented in Table 3-1.

I employ two sets of proxies for human capital which I

believe to be improvements on any that have previously

appeared. The first is collected from various UNESCO

statistical yearbooks and serves as the basis for calculating

mean years of schooling of the total population aged 25 years

and older.2 My second variable comes from Horn and Arriagada

(1986) and consists of years of schooling for young workers

alone.











Table 3-1
Definitions and Sources of Variables


Variable Name


AFRICA


ASSASSINATIONS

GDP60


GDPDEV




GOV. CONSUMP.


Definition and Source


Dummy variable for sub-Saharan Africa

Number of assassinations per year
Source: Banks (1990)
Real GDP per capital, 1960
Source: Summers and Heston (1988)

Magnitude of the deviation of GDP of each
country from the sample mean of rest of
the world over time
Source: Summers and Heston (1988)

Ratio of real government "consumption"
expenditure to real GDP (exclusive of real
military expenditure)
Source: Summers and Heston (1988) USACDA
(various years)


HUMAN CAPITAL GROWTH Growth of average years of schooling over
(UNWEIGHTED) period of 20 years
Source: UNESCO (various years)
HUMAN CAPITAL GROWTH
(WEIGHTED) Human capital growth weighted by real
expenditure per pupil with 10 years lagged
Source: UNESCO (various years) Summers and
Heston (1988)
INITIAL HUMAN CAPITAL Mean years of schooling of total
(UNWEIGHTED) population 25 years old and over, 1960
Source: UNESCO (various years)
INITIAL HUMAN CAPITAL
(WEIGHTED) Initial Human Capital weighted by real
educational expenditure per pupil
Source: UNESCO (various years) Summers and
Heston (1988)
INTERNATIONAL AND
CIVIL WAR CASUALTIES International plus civil war battle deaths
per 10000 population
Source: Small and Singer (1982)
INTERNATIONAL AND CIVIL
WAR YEAR (DUMMY) Dummy for duration of international or
civil wars
Source: Small and Singer (1982)

INVESTMENT Ratio of real domestic investment (private
plus public) to real GDP
Source: Summers and Heston (1988)













Variable Name

LAT. AMERICA

% LITERATE

MILITARY EXPEND.
DEVIATION



MIXED



PER CAPITAL GDP GROWTH


POPULATION


REAL MILITARY EXPEND.


REAL OIL PRICE
DEVIATION


REVCOUPS


SOCIAL


TEACHER STUDENT RATIO
(PRIMARY)


TEACHER STUDENT RATIO
(SECONDARY)



URBANIZATION


% WORK FORCE IN
INDUSTRY


36

Table 3-1--continued

Definition and Source

Dummy variable for Latin America

Percent literate
Source: Banks (1990)

Deviation from predicted value of real
military expenditure over time
Source: USACDA (various years)

Dummy variable for mixed free
enterprise/socialistic economic system
Source: Barro (1991)

Growth rate of real per capital GDP
Source: Summers and Heston (1988)

Population in thousands
Source: Summers and Heston (1988)

Ratio of real military expenditure to real
GNP
Source: USACDA (various years)

Fluctuations of Saudi Arabia's real oil
price over time as a basic price for OPEC

Number of revolutions and coups per year
Source: Banks (1990)

Dummy variable for socialist economic
system
Source: Barro (1991)

Teacher student ratio in primary schools
each decade
Source: World Bank (1988)

Teacher student ratio in secondary schools
each decade
Source: World Bank (1988)

Population, cities of 20000 and over per
capital
Source: Banks (1990)

Percent work force in industry sector
Source: Banks (1990)








37

Keeping in mind the framework established by the models

already discussed, I expect a positive effect of both the

stock of human capital and its growth on growth in per capital

income. I also expect a positive relation between growth and

the investment ratio. However, I expect negative effects of

initial GDP, government consumption, political instability,

and population growth.

Empirical Results

Descriptive Statistics

Descriptive statistics are shown in Tables 3-2 and 3-3.

Table 3-2 groups the countries by 1960 per capital GDP. One can

see from this Table that middle income countries grow faster

than others. Also, one observes that it is the rich nations

that have the largest initial human capital. But it is the low

income countries that have the highest percentage growth of

human capital.3 Finally, rich countries invest more in

physical capital than poor countries in this sample.

Table 3-3 organizes the data according to growth in per

capital GDP. According to Table 3-3, the more rapidly growing

economies are the countries with lower GDP per capital in 1960.

In other words, poorer countries grow more rapidly, suggesting

convergence. We also see that rapidly growing nations invest

more than slowly growing countries, and the growth of their

human capital exceeds human capital growth elsewhere. This

suggests that the higher is the growth of human capital the

higher is the growth of income. Finally, initial years of










Table 3-2
Descriptive Statistics (1960-85)
Ordered by Levels of 1960 GDP per Capita

Variable Mean Standard Deviation

TOP INCOME GROUP:1

Average Annual per Capita 2.11 1.35
GDP Growth (%)
Average Annual Growth of 1.62 1.08
Human Capital (%)
Weighted Average Annual 5.36 2.94
Growth of Human Capital (%)
GDP Level (1960) 4814 1224
Mean Years of Schooling(1960) 6.72 2.24
Average Share of Investment(%) 25.3 5.43
Average Share of Government 13.8 3.74
Consumption in GDP (%)

MIDDLE INCOME GROUP:2

Average Annual per Capita 2.73 1.47
GDP Growth (%)
Average Annual Growth of 2.19 0.72
Human Capital (%)
Weighted Average Annual 6.10 2.96
Growth of Human Capital (%)
GDP Level (1960) 1753 579
Mean Years of Schooling(1960) 3.94 2.07
Average Share of Investment(%) 21.6 7.07
Average Share of Government 15.9 6.37
Consumption in GDP (%)

BOTTOM INCOME GROUP:3
Average Annual per Capita 2.52 1.81
GDP Growth (%)
Average Annual Growth of 3.87 1.20
Human Capital (%)
Weighted Average Annual 6.94 6.29
Growth of Human Capital (%)
GDP Level (1960) 716 196
Mean Years of Schooling(1960) 1.82 1.04
Average Share of Investment(%) 16.8 6.46
Average Share of Government 17.4 5.95
Consumption in GDP (%)

Note: The sample is divided in thirds:
(1). Income level greater than 2932.
(2). Income level between 1062 and 2838.
(3). Income level less than 1012.










Table 3-3
Descriptive Statistics (1960-85)
Ordered by Growth Rates in GDP per Capita

Variable Mean Standard Deviation

RAPID GROWTH ECONOMIES:1

Average Annual per Capita 4.18 0.82
GDP Growth (%)
Average Annual Growth of 2.95 1.58
Human Capital (%)
Weighted Average Annual 8.54 4.56
Growth of Human Capital (%)
GDP Level (1960) 1903 1389
Mean Years of Schooling(1960) 3.75 2.24
Average Share of Investment(%) 24.6 6.44
Average Share of Government 15.5 4.46
Consumption in GDP (%)

MODERATE GROWTH ECONOMIES:2

Average Annual per Capita 2.58 0.46
GDP Growth (%)
Average Annual Growth of 2.06 1.06
Human Capital (%)
Weighted Average Annual 7.01 3.45
Growth of Human Capital (%)
GDP Level (1960) 2842 2050
Mean Years of Schooling(1960) 5.27 2.91
Average Share of Investment(%) 21.3 5.53
Average Share of Government 16.2 5.67
Consumption in GDP (%)

SLOW GROWTH ECONOMIES:3
Average Annual per Capita 0.73 0.86
GDP Growth (%)
Average Annual Growth of 2.74 1.35
Human Capital (%)
Weighted Average Annual 2.85 2.84
Growth of Human Capital (%)
GDP Level (1960) 2372 2022
Mean Years of Schooling(1960) 3.28 2.53
Average Share of Investment(%) 17.8 7.92
Average Share of Government 15.5 6.63
Consumption in GDP (%)

Note: The sample is divided in thirds:
(1). Income Growth Rate Greater Than 3.27.
(2). Income Growth Rate Between 1.77 and 3.26.
(3). Income Growth Rate Less Than 1.76.








40

schooling are greatest in moderate growth countries. This may

mean that human capital prevents even faster convergence; that

is, initially high levels of human capital cause high income

countries to grow more rapidly, making it more difficult for

initially low income countries to catch up.

Economic Growth and Human Capital

I now move to an econometric analysis of the data. Tables

3-4 and 3-5 show regressions for annual growth rates of per

capital real GDP. The results pertain to the period 1960 to

1985 over a cross-section of 50 countries, the largest number

of countries for which I was able to collect data on human

capital. All the regressions are OLS until Table 3-7.

The simplest regression is 4.1, which regresses the

growth rate on the initial level of GDP. In this we see a

negative effect of the initial level of GDP, which is in

accord with convergence, but the estimated coefficient is not

highly significant. In the remaining regressions I

systematically add more variables.

First I concentrate on the level and growth of human

capital. In regression 4.2 we see that the estimated

coefficient on initial human capital is positive and highly

significant (B=0.015, t = 5.07). This result agrees with Romer

(1990), which emphasizes that there is a permanent effect of

the stock of human capital on the rate of growth. Notice that

the negative effect of initial GDP is now highly significant.

In regression 4.3 I used weighted human capital, which is the








41

interaction of human capital with average real public

educational expenditure per pupil for three decades. The

estimated coefficient for this variable has the "correct" sign

and is significant.

Next, in order to measure differences in the quality of

education across nations more carefully, I use data on

teacher-student ratios covering the three decades. Regression

4.4 shows that the teacher-student ratio for primary level has

the correct positive sign and is highly significant. This

supports the idea that the higher is the ratio the higher is

the quality of education. The teacher-student ratio for

secondary level has an "incorrect" sign and is significant. I

have no explanation for this, other than to note that advanced

countries grow more slowly, perhaps since they are not as able

to borrow "technique" from more advanced countries, and it is

the advanced countries that have the high secondary teacher-

student ratios.

Regressions 4.5 and 4.6 supplement the human capital

variables by including "political instability" variables. With

the other variables held constant, the estimated coefficients

of both the initial stock of human capital and its growth

remain positive and highly significant. The positive

coefficient on human capital growth means that the countries

where human capital grows at a faster rate through educational

investments also have higher rates of economic growth. I will











Table 3-4
Basic Regressions for per Capita GDP Growth
(T-statistics in Parentheses)

VARIABLE (4.1) (4.2) (4.3) (4.4) (4.5) (4.6)


0.046 0.116 0.046
(3.75) (6.31) (3.82)


0.133
(6.90)


0.126 0.091
(6.25) (6.37)


LOG(GDP60)

INITIAL HUMAN
CAPITAL(LOG):

UNWEIGHTED


WEIGHTED


HUMAN CAPITAL
GROWTH:

UNWEIGHTED

WEIGHTED

GDP SHARE OF :


-0.002 -0.014 -0.018 -0.016 -0.018 -0.026
(-1.75) (-5.18) (-4.46) (-5.54) (-6.64) (-5.73)


0.015
(5.07)


0.008
(4.10)


0.013
(4.31)


0.015
(3.72)


0.499
(3.09)


INVESTMENT


GOV. CONSUMP.


0.138
(7.73)
-0.160
(-6.42)


GDPDEV


0.008
(3.57)





0.152
(4.71)


0.107
(5.40)
-0.144
(-5.39)


0.440 0.450
(4.70) (4.83)


REVCOUPS


REAL OIL PRICE
DEVIATION:

TEACHER STUDENT
RATIO (1960):
PRIMARY


SECONDARY


- -0.009 -0.007
(-3.62) (-3.10)

- -0.021 -0.021
(-4.06) (-4.03)


0.467
(2.06)
-0.195
(-3.33)


0.002 0.02
1149 1149


NOB


0.01
1149


0.04
925


0.15
999


0.16
999


999


CONSTANT








43

discuss the other variables in these regressions in detail

later.

Convergence

Convergence implies that there is negative relation

between the growth of per capital GDP and the initial level of

per capital GDP. In other words, convergence indicates that

poor countries can catch up to rich countries in the long run.

The results in regressions of Tables 3-4 and 3-5 tend to

support the idea of convergence.4

In 4.1 I regress growth on the initial level of per

capital GDP. The coefficient is negative, as predicted by the

Solow model, but insignificant. When I hold constant initial

human capital in regressions 4.2 and 4.3, I observe that the

coefficient for initial per capital GDP remains negative but

becomes highly significant. This result suggests that

diminishing returns set in when initial human capital is held

constant, but not before. The result is not supportive of the

Solow model augmented to include human capital, (Mankiw et

al., 1990), because it is initial human capital that is held

constant rather than its growth.

In regressions 4.5 and 4.6 I add growth in human capital.

I find evidence that both initial human capital and its growth

contribute the growth in per capital GDP. The finding that

initial human capital matters is most reminiscent of the Romer

(1990) model in which comprehensive investment depends on the

endowed skill level. That model yields parallel or even








44

divergent growth paths, and thus does not exhibit the

diminishing returns feature of the Solow model. Thus the

"convergence" result that emerges from regressions 4.3-4.6 is

very partial in its nature. Poor countries can catch up since

the coefficient on initial income is negative, but only if

human capital is held constant, which is not necessarily true

in reality. Rather, what Table 3-4 suggests is that

convergence is weak, since rich countries have a growth

advantage conferred by the abundance of their human skills.

Government Policies Towards Investment and Consumption

One of the factors affecting economic growth is the ratio

of investment in physical capital to income. In theoretical

models, such as Becker, Murphy, and Tamura (1990), growth of

per capital income and the investment ratio tend to move

together. Holding constant the other explanatory variables

included in the regressions, these theories predict that

growth and the investment ratio will be positively correlated.

I examine this in regressions 4.5 and 4.6. The results for

these regressions show that the investment ratio is positively

related with growth and in most regressions it is highly

significant, 0.138 with a t-ratio = 7.73 in regression 4.5.

Barro (1990, 1991) found that the ratio of real

government consumption to real GDP had a negative association

with growth and investment. The argument was that government

consumption had no direct effect on private productivity, but

reduced saving and growth by lowering the after tax marginal








45

product of capital. Tables 3-4 and 3-5 do indicate a

significantly negative relation between government consumption

and growth; for instance, in regression 4.5 the estimated

coefficient is -0.160, t-value = -6.42.5

In Table 3-6, I regress the investment ratio on the other

variables. Countries with larger initial GDP invest less than

others. This is especially strong once I control for human

capital weighted by real expenditure per pupil.

In Table 3-6, I see a positive effect of government

consumption on investment. One explanation for this effect may

be that government expenditure on education, which is a part

of government consumption here, is more like public investment

than public consumption. Thus this expenditure probably

affects private-sector productivity, which would have a

positive effect on private investment.

Other Explanatory Variables

One very important determinant of economic growth is

population growth. Becker and Murphy (1990) and Mankiw et al.

(1990) discuss this issue. Mankiw et al. examine a sample of

98 countries, and find that 80 percent of the variation in

income per capital can be explained by population growth,

saving, and schooling. They find that higher population growth

lowers income per capital. In Table 3-5 I confirm this negative

impact of population growth in economic growth; see, for

example, regressions 5.3 and 5.5.








46

I have also added two variables to measure political

instability, REVCOUPS and ASSASSINATIONS. The variable

REVCOUPS is the number of revolutions plus coups per year and

the variable ASSASSINATIONS is the number of assassinations

per year. The coefficients of both variables are negative and

highly significant in all the regressions. One possibility is

that these variables can be viewed as threats to property, and

hence they detract from investment and growth.

In order to show the impact of international and civil

wars on economic growth across countries, I use four

variables: INTERNATIONAL AND CIVIL WAR YEARS, INTERNATIONAL

AND CIVIL WAR CASUALTIES, MILITARY EXPENDITURE DEVIATION, and

AVERAGE OF REAL MILITARY EXPENDITURES. I include these effects

in regressions 5.3 to 5.6. The results for the last three

variables are negative and significant in 5.4 and 5.5.

However, the result for war years in regression 5.3 indicates

that there is a highly positive relation between war years and

economic growth. A negative impact of war on economic growth

seems more reasonable since war has a destructive effect on

physical capital and much income goes for military

expenditures that may have few spillovers to the civilian

economy. We can see this in the negative effect of war years

on physical investment in regression 6.5. One reason for the

positive effect of war years on growth may be that recovery

from war is usually rapid. The idea is that a wartime

destruction of physical capital in a country stimulates more








47

investment in this type of capital as well as in human

capital. Then, per capital incomes exceed what they would have

been had the war not happened. These results agree with the

idea of Becker, Murphy and Tamura (1990) who pointed out:

"Wartime destruction of physical and human capital have

different consequences because human capital is knowledge

embodied in people. When too much knowledge is destroyed, an

economy loses the foundation for further accumulations of

knowledge -whether embodied in people or disembodied in

technologies- which is the essence of economic growth." p. S35

Next, As in Barro (1991) I constructed dummy variables

for countries with socialist and mixed economy systems. In

Table 3-5 the estimated coefficients on the socialism dummy

are negative but insignificant in regressions for growth. The

mixed systems dummy is weaker still. The general failure of

these variables suggests the division of economic systems into

these groups may be subject to error, or that holding constant

my other variables, the type of government has no effect on

growth.

Next, I considered two controls for business cycles.

First, the regressions in Tables 3-4 and 3-5 show a

significant positive relation between growth and the level of

the country's GDP deviation from rest of the world. For

example, the estimated coefficient in regression 5.6 is 0.455,

t-value = 4.9. Second, I also used real oil price deviations

to capture oil price shocks. The regressions in Tables 3-4 and








48

3-5 show that real oil price fluctuations have a negative

effect on economic growth across countries. For instance, in

regression 5.5 the coefficient is -0.021 and t-ratio = -4.12.

However, in the investment ratio regressions of Table 3-6, the

impact of oil price deviation is positive and significant.

Finally, countries in Africa and Latin America have

poorer growth performance than other countries. To examine

this, I constructed a dummy variable for Africa (which equals

one for countries in Africa) and a dummy variable in Latin

America (which equals one for countries in South and Central

America). Using these dummies and the other explanatory

variables, I found insignificant coefficients for these

dummies. This means that there is no country effect for Africa

and Latin America with respect to economic growth. However, in

Table 3-6, these variable are positive and significant on

investment regressions. A comparison of regressions in Table

3-5, shows the effect of inclusion of these continent dummies

on the estimated coefficient of GDP60 and human capital

variables is not trivial. For example, for initial weighted

human capital, inclusion of these dummies reduces the

coefficient from 0.007 (t-value = 3.02) to 0.005 (t-value =

1.86).6

Simultaneous Equations Analysis of Income Growth and Human
Capital Growth Using the 2SLS Method

In Table 3-7, I present two stage least squares estimates

of regressions for economic growth and human capital growth.








49

The difference between regressions 7.1 and 7.2 versus 7.3 and

7.4 is that in the first two regressions I use unweighted

human capital measures, while I use weighted ones in the last

two regressions.

Reassuringly, the 2SLS estimates of the parameters are

almost the same as OLS estimates presented in Tables 3-4 and

3-5. The key difference between these two sets of 2SLS

estimates lies in the statistical insignificance of the

positive effect of the investment ratio in regression 7.3.

Consider regressions 7.1 and 7.3. The estimated coefficients

for human capital and its subsequent growth are more

significant when I use weighted human capital, but the

significance of the negative effect of GDP60 is somewhat

lower. One explanation for this might be that the higher the

human capital and its growth in quality of education, the

faster a country can catch up to the more advanced countries.

Government consumption also has a negative impact on economic

growth, -0.136 with t-ratio = -3.95 in regression 7.1.

Considering the regressions for human capital growth,

equation 7.2 shows that the higher the initial human capital

the lower is its growth, -0.025 with t-ratio = -20.4. The

regression with the weighted measure, 7.4, gives similar

results. Investment again has positive and significant effects

on economic growth and human capital, particularly in 7.4.








50

Interpretation of the Findings

Table 3-8 collects the major findings of this paper by

reporting the imputed impact of the variables at their means.

The first part of the Table shows the calculated effects using

the OLS estimates from regressions 4.5, 4.6, 5.1, and 5.3. The

second part displays results using the two stage least squares

estimates from regressions 7.1 and 7.3.

The first part confirms the partial convergence finding:

For example, in regression 4.6 in Table 3-8, the highly

negative effect of initial GDP on growth is 19%. Government

consumption has also a negative and highly significant effect

on growth, for instance 1.8% in regression 4.5. Also,

regressions 5.1 and 5.3 on Table 3-8 indicate that population

growth lowers the growth across countries, though by not as

much as the above variables. However, human capital and

investment, which are both highly significant, increase

economic growth. Consider regression 4.6. Here, the effects on

growth of weighted initial human capital, growth of human

capital, and investment are 11%, 0.9%, 2.3%, and are all

highly significant. Basically, investment in both physical and

human capital has a large positive impact on growth across

countries, though the effect of investment in human capital

appears with a lag of 10 or 20 years.

The second part of Table summarizes imputations from the

2SLS regressions. The effects of variables, save for

investment, are the same as above. Indeed the only difference








51

between the two lies in the statistical insignificance of the

investment ratio in the second part. Consider regression 7.1:

Here the effects on growth of initial GDP, initial human

capital, growth of human capital, investment ratio, and

government consumption are -12.1%, 2.1%, 2.1%, 3.5%, and

-1.4%, respectively. Interestingly, in regression 7.3 the

effect of the initial level of GDP is almost precisely

canceled by the effects from the initial and growth rate of

human capital. This suggests that the poor nations can catch

up to rich nations only if they invest more in human capital.

Next I compare how the contribution to growth of the

above variables differs between the poor and rich group

countries. Looking again at means and using regression 4.5 as

my example, initial GDP lowers growth by 15% in rich countries

compared to 11% in poor countries. In addition, the positive

contributions of unweighted initial human capital and its

subsequent growth are 2.8% and 0.8% in rich nations compared

to 0.9% and 1.9% in poor nations (to compare the imputed

effects of weighted human capital and its growth I can use

regression 4.6, where the impacts are 12.9% and 0.8% in rich

and 9.9% and 1% in poor nations). The effects of investment

and government consumption are 3.5% and -2.2% in rich and 2.3%

and -2.8% in poor countries. Finally, in regression 5.3, the

negative contribution of population growth to mean economic

growth is 0.4% in rich and 0.9% in poor countries.








52

Turning to 2SLS results, in regression 7.1, initial GDP

lowers the growth by 13.5% in rich and 10.5% in poor nations.

The positive contributions of unweighted human capital and its

growth are 3.2% and 1.4% in rich and 1% and 3.3% in poor

countries. In addition, the contributions of investment and

government consumption to mean economic growth are 4.2% and

-1.9% in rich nations and 2.8% and -2.4% in poor nations. In

regression 7.3, the imputed effects of weighted human capital

and its growth are 30.6% and 2.1% in rich and 23.8% and 2.7%

in poor nations.

The patterns are by now familiar: the negative effect of

initial GDP on growth is the largest, followed by government

consumption and population growth. Initial human capital and

its subsequent growth, and investment in physical capital have

positive and significant impacts on growth.

Conclusion

Building on the theory of economic growth this study

contributes to the empirical evidence about the linkages

between growth, investment, and human capital. The correlation

between per capital growth and the initial (1960) level of per

capital GDP is substantially negative when human capital

(proxied by mean years of schooling) is held constant. This

suggests convergence. However, given the level of initial per

capital GDP, the growth rate is positively related to initial

human capital and growth of human capital. Thus, the poor

countries tend to catch up with rich countries if the poor








53

countries have high initial human capital per person (in

relation to their level of per capital GDP) or have faster

growth in their human capital per person, but not otherwise.

Also, countries with high human capital have low population

growth and high ratios of physical investment to GDP.

Per capital GDP growth is negatively related to government

consumption and positively related to the ratio of investment

to GDP. An explanation for this can be that government

consumption introduces distortions, such as high taxation,

while investment in capital stimulate growth. However, since

government consumption seems to lead to increased growth in

investment, through this indirect avenue government

consumption is positively related to economic growth.

Political instability (proxied by revolutions plus coups)

is negatively related to growth. This is probably because of

their negative impact on investment.


Notes

1. As Barro indicates: "It would be better to use proxies for
the initial stock of human capital per person rather than
variables that relate to the flow of investment in human
capital. The stock of human capital derived from formal
education depends on current and lagged values of school-
enrolment rates." Barro (1991), p. 414. Therefore, it seems to
me that in this paper I have a proper proxy for human capital
and its growth.

2. This measure of the stock of human capital per person is
the average years of schooling in the population and is based
on UNESCO data that in turn are derived from the various
national censuses. The processing of the data to derive this
statistic is nontrivial; See the Appendix B for details.

3. Two possible explanations immediately come to mind why
these low income countries, which have the highest growth rate










of human capital, do not have the highest growth of income:
the high degree of political instability in these countries
and the extremely high growth of population in these nations.
I examine this more rigorously later.

4. In tests of convergence, there are at least two pitfalls.
First, one should not select the sample on the basis of latter
success, since this tends to give convergence. This problem
plagues the sample of countries in Maddison (1982), which were
employed by Baumol (1986) to test for convergence. De Long
(1988) severely criticized this selection bias. The second
problem, which also was pointed out by De Long, is that errors
in initial GDP are accompanied by equal and opposite error in
growth. The result is a bias toward -1 in the estimated
coefficient for initial GDP. To avoid these problems (biased
sample and measurement error), I have used the growth rate
starting from 1961 instead of 1960 in the regressions reported
here. I also used 1970 as the starting point of growth rate
and obtained the same results, as reported in this paper. In
addition, the sample which has been chosen in this paper
contains a broad sample of nations from 5 continents and both
poor and rich countries. As a result, it seems that De Long's
criticisms may not have much force.

5. I also employed the government consumption variable from
Barro's (1991) data set, which is the average from 1970 to
1985 of the ratio of real government consumption (exclusive of
defence and education) to real GDP, and got the same result.

6. In addition to the UNESCO data for calculation of proxy for
human capital, I employed data from Horn and Arriagada (1986)
which gave me the human capital of young generations. Then I
repeated a few regressions with this data and got the same
signs for major variables as in the regressions with UNESCO
data. However, the estimated coefficients and t-ratios were a
little lower, which may be because of the lower number of
observations in the later data set. Another reason can be that
the younger human capital is not yet completely involved in
production process.











Table 3-5
Regressions for per Capita GDP Growth
Experiments with War Effects, Dummy Variables,
Population Growth
(T-statistics in Parentheses)


VARIABLE (5.1) (5.2) (5.3) (5.4) (5.5) (5.6)


CONSTANT


LOG(GDP60)

INITIAL HUMAN
CAPITAL(LOG)
(WEIGHTED)

HUMAN CAPITAL
GROWTH
(WEIGHTED)


GDP SHARE OF :

INVESTMENT


GOV. CONSUMP.


GDPDEV


0.116 0.119 0.114 0.112 0.112 0.114
(6.54) (6.60) (6.44) (6.32) (6.36) (6.43)

-0.024 -0.023 -0.026 -0.027 -0.027 -0.028
(-4.75) (-4.45) (-5.72) (-5.95) (-5.86) (-5.97)


0.006 0.005 0.007 0.008 0.007 0.008
(2.24) (1.86) (3.02) (3.32) (3.21) (3.39)


0.120 0.110 0.135 0.143
(3.33) (2.94) (4.14) (4.37)


0.111 0.120
(5.51) (5.64)

-0.132 -0.127
(-4.82) (-4.61)

0.450 0.447
(4.84) (4.81)


0.140 0.142
(4.30) (4.36)


0.108 0.105 0.106 0.108
(5.44) (5.29) (5.37) (5.42)


-0.138
(-5.18)

0.446
(4.81)


-0.143 -0.142
(-5.39) (-5.34)


-0.150
(-5.45)


0.454 0.457 0.455
(4.88) (4.91) (4.90)


REVCOUPS

REAL OIL PRICE
DEVIATION

POPULATION
GROWTH


CONTINENTAL DUM.:


-0.007 -0.007 -0.007 -0.007 -0.007 -0.007
(-3.00) (-2.99) (-3.13) (-2.93) (-2.92) (-2.97)


-0.021 -0.021 -0.020
(-4.04) (-4.05) (-4.02)


-0.021 -0.021 -0.021
(-4.07) (-4.12) (-4.10)


-0.218 -0.213 -0.339 -0.276
(-1.41) (-1.38) (-2.35) (-1.90)


-0.295 -0.255
(-2.05) (-1.72)


LAT. AMERICA


AFRICA


-0.004 -0.005
(-1.35) (-1.47)

-0.006 -0.005
(-1.32) (-1.02)


and











Table 3-5--continued

VARIABLE (5.1) (5.2) (5.3) (5.4) (5.5) (5.6)

COUNTRY DUMMY :


-0.016
(-1.32)


MIXED

INTERNATIONAL
AND CIVIL WAR
YEAR (DUMMY)

INTERNATIONAL
AND CIVIL WAR
CASUALTIES PER
10000 POP.

MILITARY EXPEND.
DEVIATION

AVERAGE OF REAL
MILITARY EXPEN.


0.17


- -0.0009
(-0.36)


0.010
(2.06)


-0.004
(-1.61)


- -0.550 -0.559
(-1.71) (-1.73)


-0.037
(-1.16)


0.17


0.17


0.17


0.17


0.17


NOB 999 999 999 999 999 999


SOCIAL


NOB


999 999 999 999


999 999











Table 3-6
Regressions for Investment Ratios
(T-statistics in Parentheses)

VARIABLE (6.1) (6.2) (6.3) (6.4) (6.5) (6.6)


CONSTANT


LOG(GDP60)

INITIAL HUMAN
CAPITAL(LOG):

UNWEIGHTED


WEIGHTED

HUMAN CAPITAL
GROWTH :


0.146
(4.09)

-0.011
(-2.27)


-0.137
(-6.15)

-0.085
(-12.4)


-0.165
(-7.06)

-0.087
(-11.6)


-0.177
(-7.92)


-0.134
(-6.03)


-0.135
(-6.08)


-0.091 -0.086 -0.085
(-12.7) (-12.4) (-12.3)


0.086
(12.2)


0.067 0.070 0.072 0.067 0.066
(20.6) (18.7) (20.3) (20.7) (20.5)


UNWEIGHTED


WEIGHTED

GDP SHARE OF
GOV. CONSUMP.


GDPDEV


REVCOUPS


ASSASSINATIONS

REAL OIL PRICE
DEVIATION


CONTINENTAL DUM.:


LAT. AMERICA


AFRICA


- 0.559 0.597 0.661 0.562 0.542
(11.6) (11.1) (12.7) (11.6) (11.1)

0.211 0.132 0.085 0.020 0.128 0.127
(4.80) (3.13) (1.97) (0.48) (3.05) (3.01)

0.346 0.334 0.325 0.320 0.336 0.355
(2.08) (2.26) (2.22) (2.31) (2.27) (2.40)

-0.006 0.0009 0.001 0.001 0.001 -
(-1.45) (0.24) (0.46) (0.33) (0.31)

-- -- - -- -- -0.002
(-2.20)

0.015 0.014 0.013 0.013 0.014 0.015
(1.67) (1.77) (1.69) (1.72) (1.74) (1.81)


0.009 0.013
(1.78) (2.66)

0.032 0.012
(4.37) (1.69)


1.164
(4.09)











Table 3-6--continued


VARIABLE (6.1) (6.2) (6.3) (6.4) (6.5) (6.6)


COUNTRY DUMMY :

SOCIAL


MIXED

INTERNATIONAL
AND CIVIL WAR
YEAR (DUMMY)


R2

NOB


0.196
(10.8)

0.005
(1.42)


0.31

999


0.45

999


0.46

999


0.52

999


-0.012
(-1.45)

0.45

999


0.46

999











Table 3-7
Two Stage Least Squares Estimates
Regressions for per Capita GDP Growth and
Weighted Human Capital Growth
(Asymptotic T-statistics in Parentheses)

REGRESSION (7.1) (7.2) (7.3) (7.4)
DEPENDENT PER CAPITAL HUMAN CAPITAL PER CAPITAL HUMAN CAPITAL
VARIABLE GDP GROWTH GROWTH GDP GROWTH GROWTH


CONSTANT


LOG(GDP60)

INITIAL HUMAN
CAPITAL(LOG):

UNWEIGHTED


WEIGHTED

HUMAN CAPITAL
GROWTH :

UNWEIGHTED


WEIGHTED

PER CAPITAL GDP
GROWTH

GDP SHARE OF :

INVESTMENT


GOV. CONSUMP.


GDPDEV


REVCOUPS

REAL OIL PRICE
DEVIATION


0.100
(4.16)

-0.016
(-5.64)



0.017
(2.60)


0.020
(3.41)

0.002
(2.86)


0.051
(1.86)

-0.041
(-4.58)


0.155
(6.77)

0.058
(11.3)


-0.025
(-20.4)


0.019
(3.14)


-0.042
(-16.3)


0.862
(2.71)


0.393
(3.08)


0.168
(7.17)

-0.136
(-3.95)

0.185
(1.63)

-0.009
(-4.03)

-0.018
(-3.49)


-0.004
(-0.13)


0.012
(1.72)


0.052
(1.12)

-0.141
(-3.74)

0.192
(1.65)

-0.007
(-3.24)

-0.016
(-2.95)


0.349
(2.53)


0.225
(7.11)











Table 3-7--continued

REGRESSION (7.1) (7.2) (7.3) (7.4)
DEPENDENT PER CAPITAL HUMAN CAPITAL PER CAPITAL HUMAN CAPITAL
VARIABLE GDP GROWTH GROWTH GDP GROWTH GROWTH

CONTINENTAL DUM.:


LAT. AMERICA


AFRICA


INTERNATIONAL
AND CIVIL WAR
CASUALTIES PER
10000 POP.


-0.001
(-0.51)

0.002
(0.27)




-0.005
(-1.39)


MILITARY EXPEND. -0.062
DEVIATION (-0.13)

TEACHER STUDENT
RATIO (1960):

PRIMARY -


SECONDARY -


% WORK FORCE IN
INDUSTRY


URBANIZATION


% LITERATE


-0.003
(-3.71)

0.015
(7.40)




0.0008
(1.42)


0.010
(1.75)

0.003
(0.39)




-0.008
(-1.86)

0.150
(0.32)


-0.232
(-4.22)

-0.008
(-0.60)


-0.027
(-4.52)

-0.007
(-3.37)

0.044
(11.5)


-0.032
(-8.72)

0.039
(4.33)




0.005
(2.00)


-0.348
(-1.34)

-0.031
(-0.49)


-0.048
(-1.83)

-0.042
(-4.28)

0.080
(6.11)


NOB 630 630 629 629


NOB 630


630 629


629











Table 3-8
Estimated Mean Contributions to Economic Growth
(T-statistics in Parentheses)


(OLS) (2SLS)

VARIABLE REG. REG. REG. REG. REG. REG.
(4.5) (4.6) (5.1) (5.3) (7.1) (7.3)


LOG(GDP60)

INITIAL HUMAN
CAPITAL(LOG):


-0.135 -0.195 -0.180
(-6.64) (-5.73) (-4.75)


-0.195
(-5.72)


-0.121 -0.310
(-5.64) (-4.58)


UNWEIGHTED


WEIGHTED

HUMAN CAPITAL
GROWTH:

UNWEIGHTED


WEIGHTED

GDP SHARE OF :


INVESTMENT


0.112 0.084 0.099
(3.57) (2.24) (3.02)


0.013
(3.09)


0.009 0.007 0.008
(4.71) (3.33) (4.14)


0.030
(7.73)


0.023
(5.40)


0.024
(5.51)


0.024
(5.44)


GOV. CONSUMP.


POPULATION
GROWTH


-0.018 -0.016 -0.015 -0.015 -0.014 -0.015
(-6.42) (-5.39) (-4.82) (-5.18) (-3.95) (-3.74)


-- -- -0.004 -0.006 -- -
(-1.41) (-2.35)


0.018
(3.72)


0.021
(2.60)


0.269
(3.14)


0.021
(2.71)


0.026
(3.08)


0.010
(1.12)


0.035
(7.17)


NOTE. Estimated contributions are regression coefficients times
means of variables.














CHAPTER 4

TECHNOLOGY CREATION, TECHNOLOGY TRANSFER, AND
ECONOMIC GROWTH

Introduction

Countries differ in their economic performance, and these

differences persist for long periods. Countries also differ in

the technologies they use. These two facts are related in some

undiscovered way.

The idea of this chapter is to study technology creation,

technology transfer, and their relative contributions to

countries classified by level of technology. Technology

indicators here capture the effect of original innovations as

well as imitation and adaptations through technology transfer

and licensing.

We have already indicated that there are three main

concerns in this chapter. The first is with the connection

between home-based resident, or original technology and

economic growth. This is impounded in the residual term in a

model of disembodied growth of the kind advanced by Solow

(1957).

My second concern is with the effect of borrowed

technology on the growth of less advanced countries. The

recent literature concludes that many newly industrialized

countries (NICs) have enjoyed very high growth rates at the

62








63

expense of leader countries. Also, the evidence in Evenson

(1984) suggests that imitation by less industrialized

countries of technologies developed in more advanced countries

is an important source of growth in the less developed

countries. A related point is that original research in

advanced countries can be imported by LDCs, thereby permitting

adaptive invention rather than expensive original research. To

examine the empirical significance of these ideas, this

chapter develops measures of imitation and adaptation.

Therefore, I study the link between technological progress and

skills at adapting new technologies. The ability to adapt

technology is correlated with measurable factors such as

patents and the stock of scientists and engineers (S&E).

My third concern is with the potential for enhanced

division of labor, through interrelated growth of advanced and

follower countries, and for world redistribution of output as

NICs increase their technology acquisition at the cost of

market share in the advanced countries. It is possible that

the new international division of labor and the world

redistribution of output through this process could be a

reason for the slowdown of advanced countries since 1973.

I turn next to a review of the literature. Obviously,

while there is an enormous literature on growth and

technology, this chapter will concentrate on a few key

articles.








64

The starting idea for this chapter began with Evenson

(1984). Using data on patented inventions from many countries,

he reaches two principal conclusions: First, the data strongly

support the notion of comparative advantage of advanced

countries in invention. The production of original invention

is concentrated in certain firms located in countries with the

best laboratories. Industry in "follower" countries imports

inventions and concentrate on adaptive invention rather than

investing heavily in R&D. Second, the data show that

inventions per scientist and engineer have declined from the

late 1960s to 1970s in almost all of the fifty countries for

which data are available. Evenson declared that the declining

trend supports the case for interpreting this phenomenon as

the result of exhaustion of "invention potential." However,

Griliches (1990) points to a declining propensity to patent

instead.

In the same survey paper Griliches points out that in

spite of all difficulties, patent statistics remain a unique

resource for the analysis of the process of technical change.

Also, he believes that nothing else comes close in the

quantity of available data, accessibility, and the potential

industrial, organizational, and technological detail. This

viewpoint is maintained in the present chapter.

Jovanovic and Lach (1990) suggest that variation in GNP

over countries arises because countries differ in the speed of

implementation of new technology. Their model implies that








65

slow countries end up with lower GNP. They predict slower and

more trended growth in the laggard countries. They also

suggest that diffusion lags should not affect long-run growth.

Grossman and Helpman (1991) also see the variation in

growth rates as the key fact to be explained. They point out

that a reading of recent economic history suggests two

important trends. First, technological innovations are an ever

more important contributor to growth. Second, the economies of

the world are increasingly open and interdependent. The two

trends are related. Grossman and Helpman suggest that rapid

communication and close contacts among innovators in different

countries facilitate invention and the spread of new ideas.

Also, they emphasize that the rapid changes in technology

promote trade and integration into the world economy.

Grossman and Helpman model the diffusion of technology as

the central activity in innovation and growth. They begin with

the point that technical changes has often been treated as an

exogenous process in long-run economic analysis. This would be

appropriate if advances in technology followed automatically

from fundamental scientific discoveries and if basic research

were independent of market forces. While obviously wrong from

a global perspective, they explore this perspective from the

standpoint of developing economies, for whom technical change

would be largely exogenous if knowledge diffused inevitably

from the industrialized North to the lagging South and if the








66

pace of innovation in the North were little affected by events

in the South.

Finally, Krugman (1979) in his famous model of product

cycle divided countries into innovating North and

noninnovating South. Innovation consists of the development of

new products, which can be produced at first only in North,

but eventually the technology of production transfers to

South. This technological lag gives rise to trade, with North

exporting new products and importing old products.

The present chapter is closely tied to the developments

summarized above. I use pooled time-series cross-section data

to explore issues of technology creation and transfer. To

control for transitional effects I have employed the same

control variables used in a previous chapter. Due to data

limitations the sample consists of fifty-four countries during

the period 1968-1985.

I employed two sets of proxies for created versus

borrowed technology. The first is resident patents, which are

collected from industrial property statistics (World

Intellectual Property Organization, various years). This

measures in broad outline the new research output of a

country. These are annual data for the period 1968-1985. The

second proxy is nonresident patents, which are also collected

from the same source. This is a proxy for technology licensing

and perhaps transfer. In addition, I have collected data on








67

research scientists and engineers from the UNESCO Statistical

Yearbooks (various years).

The reminder of this chapter is arranged as follows. The

second section describes the data set. Empirical results are

reported in the third section. I find that technology strongly

contributes to growth, but depending on the country--advanced

or nonadvanced--the form of the technology effect is

different. For example, in the case of advanced countries a

strongest contribution comes from original inventions. On the

other hand, in nonadvanced countries the adoption of

technology is more important. Indeed, in the NICs the

imitation of technology devised in the lead countries is the

main cause of their rapid growth. It is obvious that a corps

of research scientists and engineers is very crucial to this

process. In the rapidly advancing Pacific Rim countries the

transfer of western technology is an especially large

contributor to growth. The last section of this chapter

explores directions for continued research.

Description of the Data

The study uses 1968-1985 real GDP data, its components--

investment and government consumption--and population data are

again from Summers and Heston (1988). The measure of growth is

the log first difference of real per capital GDP. Investment

and government consumption are expressed as ratios to GDP and

serve once more as controls.








68

As already noted, main technology variables include

resident patents and nonresident patents. Resident patents

measure the number of patents granted to nationals;

nonresident patents are patents granted to foreigners by the

domestic country.

Technology efficiency is the ratio of inventive output

per unit input of R&D personnel. In particular, technology

efficiency is the ratio of patents by residents of a country

divided by research scientists and engineers in the same

country.

Table 4-1 provides an overview of the data. The data are

reported at 4 or 5 year intervals to provide a sense of

change. The full sample, which contains 54 countries, has been

divided into three groups: newly industrialized, advanced, and

rest of the world countries. Advanced countries are basically

high income western countries. For the definition of these

three groups of countries see the notes to Table 4-1. Together

these countries are the only ones for whom data on patents and

scientists and engineers are reported and thus comprise the

set of active participants in technology.

In the NICs the number of patents by residents rises

markedly, but the number of nonresident patents decreases

slightly. Since nonresident patenting is licensing of

technology from elsewhere, this suggests a decline in

international protection of intellectual property. In other

words some nations use other countries' technology without











Table 4-1
Descriptive Statistics

Resident Resident Non- Rest of
Patents Patents Resident the World
per Patents Resident
S & E Patents

Newly Industrialized
Countries (NICs)1

1968-1972 0.097 2883 3073 128551

1973-1976 0.071 3984 2960 126483

1977-1981 0.043 4646 2970 114094

1982-1985 0.031 5300 2726 117417

Advanced Countries2

1968-1972 0.160 6728 13015 124706

1973-1976 0.113 6068 11513 124399

1977-1981 0.081 4880 9521 113860

1982-1985 0.060 4761 9164 117956

Rest of the World3

1968-1972 0.036 98 540 131336

1973-1976 0.034 81 420 130386

1977-1981 0.027 70 355 118670

1982-1985 0.024 50 289 122667

Notes:
1. These countries include Korea, Japan, Philippines, Singapore,
Israel, Greece, Portugal, Spain, Turkey, Mexico, and Brazil.
2. These are Australia, Austria, Belgium, Denmark, Finland, France,
Germany, Italy, Netherlands, Norway, Sweden, Switzerland, UK,
Canada, USA.
3. Rest of the world countries are Algeria, Egypt, Kenya,
Mauritius, Morocco, Tunisia, Uganda, Zaire, Zambia, Zimbabwe,
India, Iran, Iraq, Jordan, Sri-Lanka, Cyprus, Iceland, Ireland,
Malta, El-Salvador, Guatemala, Argentina, Bolivia, Chile, Colombia,
Ecuador, Uruguay, Venezuela.








70

paying licensing fees. Another less likely possibility is that

these countries now depend on their own inventions and are

less dependent on foreign technology.

The efficiency of the NICs has also declined over time.

This is in line with Evenson's finding that inventions per

scientist and engineer have declined in almost all countries

for which data are available.

Now turn to the data for the advanced countries. A steep

decline in the number of both resident and nonresident patents

is readily apparent. Clearly the lead of the advanced

countries has diminished with time. Again, in these countries

efficiency rate has declined markedly. But there is also a

decline of the advanced countries relative to the Pacific Rim,

principally Japan. Finally, a large number of resident patents

for rest of the world betokens a large pool of the stock of

new technology available to country. This is very likely a

proxy for the externality or spillover effect of technology.

Table 4-2 reports mean and standard deviation for the

various technology indicators by our three groups of

countries. As expected the average number of scientists and

engineers engaged in R&D is much larger in advanced nations.

The indicators for NICs are greater than in other countries

but less so. We see that the ability to absorb new technology

and thus to grow has risen in these nations. Average numbers

of resident and nonresident patents in advanced countries are

higher than in the NICs and other countries. Together these










Table 4-2

Descriptive Statistics (1968-85)

Variable Mean Standard
Deviation

NEWLY INDUSTRIALIZED COUNTRIES:

Annual per Capita GDP Growth (%) 3.96 4.62
GDP Level (1960) 1627 633
Resident Patents" per Million Population 53 92
Nonresident Patentsb per Million Population 123 136
S&Ec per Million Population 653 1066
Annual Growth of Resident Patents per S&E (%) -6.7 44.9
Annual Growth of Nonresident Patents per S&E (%) -4.7 79.3

ADVANCED COUNTRIES:

Annual per Capita GDP Growth (%) 2.55 2.53
GDP Level (1960) 5069 1046
Resident Patents per Million Population 137 129
Nonresident Patents per Million Population 551 447
S&E per Million Population 1492 575
Annual Growth of Resident Patents per S&E (%) -6.0 42.0
Annual Growth of Nonresident Patents per S&E (%) -7.4 45.0

REST OF THE WORLD:

Annual per Capita GDP Growth (%) 1.50 7.32
GDP Level (1960) 1575 1258
Resident Patents per Million Population 5 10
Nonresident Patents per Million Population 45 71
S&E per Million Population 118 279
Annual Growth of Resident Patents per S&E (%) -3.9 66
Annual Growth of Nonresident Patents per S&E (%) -6.0 101


Notes:
a. Resident Patents: Grants of patents to residents.
b. Nonresident Patents: Grants of patents to nonresidents.
c. S&E: Number of scientists and engineers engaged in R&D.








72

trends suggest movement of high technology production outside

the lead countries, but still a sharp difference in

comparative advantage of advanced countries in technology.

Finally, the average annual growth of patents per scientist

and engineer is everywhere negative.

Empirical Results

Simple Correlations

The data are not especially familiar. Thus, Table 4-3

reports simple correlations between income growth and the

different technology variables for the full sample, advanced,

and follower countries. In the first panel, the advanced

nations' resident patents and rest of the world's resident

patents are incorrelated and positively correlated with growth

respectively. In addition, there is a positive correlation

between income growth in advanced countries and the average

efficiency ratio (inventions per scientist and engineer) in

the NICs. Resident and nonresident patents are highly

correlated, suggesting that home technology is associated with

licensing. The same applies to rest of the world's resident

patents and resident patents in advanced countries. This

suggests correlated increases in invention, perhaps promoted

by exchange of information.

I now turn to the full sample. The only positive and

significant correlation is between the rest of the world's

resident patents and income growth. In the follower countries'











Table 4-3
Simple Correlation Matrix
(P Values in the Second Line for each Variable)

Avg.
Income Res. Non- Res. Rest of Avg. Res.
Growth Patent res. Patent World Patent Patent
Patent per Res. per per
S&E Patents S&E S&E
NICs NICs


1. Advanced Countries:
Income 1.000 -0.050
Growth 0.0000 0.4071


Res. Patents


Nonresident
Patents

Res. Patents
per S&E

Rest of the
World Res.
Patents
Avg. Patents
per S&E in
NICs
Avg. Res.
Patents
per S&E
in NICs


0.034 0.118 0.210
0.5724 0.0516 0.0005


1.000 0.529 0.015 0.740
0.0000 0.0001 0.7945 0.0001

1.000 0.158 -0.315
0.0000 0.0090 0.0001

1.000 0.152
0.0000 0.0123


0.296 0.255
0.0001 0.0001

0.069 0.059
0.2543 0.3320

0.160 0.129
0.0082 0.0340

0.332 0.284
0.0001 0.0001


1.000 0.288 0.264
0.0000 0.0001 0.0001

1.000 0.930
0.0000 0.0001

1.000
0.0000


2. All Countries:
Income 1.000 0.027 0.032 -0.032 0.125
Growth 0.0000 0.3937 0.3132 0.3054 0.0001


Res. Patents


Nonresident
Patents

Res. Patents
per S&E

Rest of the
World Res.
Patents


1.000 0.557 -0.060 -0.632
0.0000 0.0001 0.0607 0.0001


1.000 -0.052 -0.316
0.0000 0.0996 0.0001


1.000 0.030
0.0000 0.3485


1.000
0.0000











Table 4-3--continued


Avg.
Income Res. Non- Res. Rest of Avg. Res.
Growth Patent res. Patent World Patent Patent
Patent per Res. per per
S&E Patents S&E S&E
NICs NICs


3. Follower Countriesa:

Income 1.000 0.047 0.034 -0.029 0.134
Growth 0.0000 0.2113 0.3577 0.4282 0.0003

Res. Patents 1.000 0.635 -0.032 -0.530
0.0000 0.0001 0.3887 0.0001

Nonresident 1.000 0.166 -0.323
Patents 0.0000 0.0001 0.0001

Res. Patents 1.000 0.005
per S&E 0.0000 0.8915

Rest of the 1.000
World Res. 0.0000
Patents


Note:
a. The follower countries are all the countries except the advanced
countries. See the notes to Table 4-1.








75

sample, the correlation between income growth and rest of the

world's resident patents is even stronger.

Generally these correlations indicate that there is a

chain of technology exchange among countries. First, the

advanced countries exchange home generated inventions with

each other and thereby benefit from each other. Also, new

technology benefits middle income countries and poor

countries, in their role as recipients of invention.

Full Sample Regression Results

Table 4-4 presents regression findings from the pooled

data. The dependent variable is the annual rate of growth of

per capital GDP in 54 countries, the largest number of

countries for which I was able to collect clean data on

patents. The period is 1968-1985.

I begin with simple patent variables and their

relationship to growth. In regression 4.1 I find that the

estimated coefficient on resident patents is positive and

significant, 0,062 with a t of 2.2. Since this represents

technology creation, this result suggests that new inventions

have a strong effect on economic growth. The assumption is

that patents are a good proxy for the output of inventive

activities. Turning to rest of the world's resident patents,

again the estimated coefficient is positive and even more

strongly significant, 0.111 with a t-ratio of 6.10. Very

likely access to inventions elsewhere advances the growth of

other countries through imitation. In other words, this result








76

suggests that borrowing of technology is an important factor

for growth.

Regression 4.2 is an alternative specification of

technology transfer. As expected, the coefficient on non-

resident patents is positive and significant. This result is

consistent with the idea that technology licensing is an

important engine of growth. Therefore, developing countries

are able to imitate or adapt the technologies that have been

created in advanced countries. This result is in line with

much research in this area, especially work by trade

economists. Vernon (1966) first described technology transfer

in his celebrated "product cycle" hypothesis, which led to

numerous theoretical and empirical studies. Results in the

remaining regressions in Table 4-4 confirm the importance of

technology transfer from developed to less developed countries

for the pattern of relative incomes. In regression 4.3, I have

rescaled nonresident patents by GDP level of each country to

incorporate size of each country's economy. Nonresident

patents continue to be positive and significant as expected.

However, in regression 4.4, I substitute the ratio of resident

patents to GDP in place of nonresident patents. Now there is

no impact of technology. What really matters for growth in the

full sample of countries seems to be technology transfer

rather than creation. The latter is of greater importance in

the advanced countries as we will see in the next section. In

regression 4.5, I rescale patents by population. Nonresident








77

patents per capital continue to have a strong positive effect

on per capital GDP growth. Again this suggests that for less

industrialized countries imitation rather than creation of

technology is their principal engine of growth. Finally, I

incorporate all three variables in regression 4.6. These

positive and significant results again suggest that in the

world as a whole, the world-wide stock of disembodied

technology and perhaps licensed technology are the main

sources of growth. One striking result is that the coefficient

of the rest of the world patents in the full sample compared

with advanced countries is two times greater (0.111 vs 0.051).

This confirms in yet a different way that borrowing is more

effective in all countries than in advanced countries.

I now turn to impacts of the control variables. These are

much the same as in chapter 3 except that enrollment ratios

for primary and secondary levels are substituted for years of

schooling to maintain sample size.' All the regressions in

Table 4-4 support a concept of convergence--the negative

relationship between initial GDP level and its subsequent

growth. The human capital proxies have positive impacts on

growth and are significant. Investment and government

consumption ratios are positive and negative, respectively,

and both are highly significant. In addition, the estimated

coefficient of GDP and real oil price deviations, viewed as

controls for business cycles, are positive and negative,

respectively, and both are highly significant. Finally, and as








78

before, revolutions and coups, which measure political

instability, have a negative impact.

Results for Advanced Countries

In advanced economies, due to their comparative

advantage in conducting R&D, home based technology proxied by

resident patents has a major impact on growth. At the same

time the ratio of inventions to scientists and engineers

raises the rate of growth. I also explore tentatively the

potential for enhanced division of labor, through the

interrelated growth of advanced and follower countries, and

also the potential for world redistribution of output, which

is potentially threatening to the advanced nations, as newly

industrialized countries increase their technology

acquisition. These redistributive effects could be a source of

the widely noted economic slowdown in advanced countries.

Tables 4-5 and 4-6 record the results for advanced

countries. In general the data strongly support the greater

role of technology creation in per capital GDP growth in these

countries. Table 4-5 studies the effects of invention and

inventive efficiency. In contrast, Table 4-6 reports

experiments with the division of labor and world

redistribution of output.

In 5.1 I find that the effect of domestic patents per

capital alone is positive but not significant. This might be

explained by the idea of the economists who are persuaded that

old lines of technological progress have been exhausted while








79

people are only slowly learning how to apply and exploit new

potentials of, for example, computers and telecommunications.

Also, there are large errors in the resident patent data.

Regression 5.2 shows that the contribution of relative

patenting efficiency to growth is both positive and

significant. This result for technology efficiency agrees with

Evenson's (1984) findings. Evenson reported that the ratio of

patents granted per unit of inventive input had fallen from

1964 to 1979-80 in most of the forty-four countries for which

UNESCO data were available, and suggested that this fall

contributed to a decrease in growth.

Next I add the new technology efficiency variable to

resident patents. Again the efficiency ratio is positive and

significant but this is not the case for the resident patents.

In regression 5.4, I add rest of the world resident patents to

the other technology variables. The results indicate that all

three have positive effects on growth.

In regression 5.5 the significant and positive impacts of

the home-based creative innovations and the spillover effects

of the available stock of technology in the world have

benefitted the advanced nations' growth. But the impact of

these stocks in advanced countries is weaker than in less

advanced ones. In regression 5.6 the results for nonresident

patents are slightly weaker in advanced economies. The results

in 5.5 and 5.6 are in line with the idea that within the

subset of technological countries there is an ongoing process








80

of international productivity catch-up and technology

convergence. In fact, the data show that even after 1973 the

gaps between leader and follower countries continue to

decline, although at a slower pace. It may be that the

resources in trailing countries that, until recently were

devoted mainly to borrowing and adapting existing advanced

technology, are now being increasingly devoted to pushing the

technological frontier itself. This implies that in the long

run advances in knowledge benefit all countries. Thus advanced

countries gain from the technological progress among

themselves and from trade with more productive follower

countries. Through the diffusion of knowledge and technology,

the process of catch-up and convergence may be desirable not

only for the countries catching up, but also for the old

leader.

I look next at the effect of the "standard" control

variables: initial GDP, human capital variables, investment

and government consumption ratios, GDP deviation, and oil

price shocks in Table 4-5. The signs of these variables are

mostly the same as in the full sample, though of course the

level of significance is lower in this smaller sample.2 In the

case of primary and secondary enrollment ratios, we see no

effect on growth except in regression 5.2. Not surprisingly,

there is little effect of political instability on growth in

advanced countries, since instability is rare. The business

cycle variables and especially real oil price deviations are








81
more significant in advanced countries. This is because the

industrialized countries' economies are more sensitive to oil

price shocks.

In Table 4-6, I record findings for some novel

specifications involving proxies for the division of labor and

world redistribution of output. I do this by looking at

interrelated growth of advanced and follower countries perhaps

for the first time. In this table I discuss only the

coefficients related to the growth of incomes and technology

variables since the behavior of the controls is similar to

Table 4-5.

In Table 4-6, I incorporate income growth and technology

indicators for both NICs in general and Pacific Rim3 countries

into regressions for the advanced countries. In regression 6.1

the estimated coefficient of average income growth of NICs--

which has been weighted by GDP level of these countries--is

positive and highly significant. Growth in the NICs is

positively correlated with growth in advanced nations. A

weaker version of the same result is obtained for the Pacific

Rim countries. These results agree with Abramovitz (1990). He

points to the important benefits and the possibility of

positive externalities involved in the catch-up process. Then

he explains that advanced countries like the United States can

trade with more productive and efficient partners and get the

benefits of cheap goods produced in those sectors when the

trading partners' productivity has advanced rapidly.








82

Turning to the technology variables in the regressions

6.3 through 6.6, the results show a negative impact of

acquisition of technology in NICs and Pacific Rim countries on

the growth of advanced countries. This result offers a

potential explanation for the productivity slowdown in

advanced countries. The point is that the potential for growth

is weaker, but in the leading countries. Thus the newer

industrializing countries of southern Europe, Southeast Asia,

and Latin America can obtain a larger market share at the

expense of lead countries.

Results for Follower Countries

Table 4-7 collects findings for follower countries which

are useful in comparison with the analogous results in Table

4-5. Estimated coefficients reported in Table 4-7 indicate

that for follower nations only the stock of available

technologies in the world and the transfer of technology from

advanced countries are important. In other words, these

results underscore the reliance in follower countries of

transfer of expensive technology from advanced countries

through imitation. Generally, these results agree with the

findings of Pack and Westphal (1986) that "the minor role of

invention in industrialization simply means that much

technological change consists of assimilating and adapting

foreign technology" (p.105).








83

Interpretation of the Findincrs

Table 4-8 collects the major results in the form of

imputed effects, the products of means and regression

coefficients. Group 1 reports the imputed effects from the

full sample, where the source is regressions 4.1 4.3 and

4.5. Group 2 records imputations from advanced countries,

regressions 5.2, 5.5, 5.6, 6.1, and 6.4. Finally, group 3

records imputed effects from follower countries, regressions

7.2, 7.4-7.5.

Group 1 reveals that the rest of the world's pool of new

inventions strongly affects the growth in all countries (14%).

The effects of nonresident patents as a ratio of GDP,

nonresident, nonresident per capital, and resident patents are

0.5%, 0.4%, 0.3%, and 0.2%, respectively. These results

suggest that a major cause of growth across countries really

is the spillover and externalities through access to the

world-wide new inventions. In addition, technology transfer is

the second major factor of growth.

Group 2 summarizes the imputations from advanced

countries. Again the rest of the world,s technology diffusion

has the higher effect on growth among others (7%). Average

income growth in the NICs, nonresident patents, resident

patents, and inventions per scientist and engineer effects on

growth of advanced countries are 1.5%, 0.7%, 0.4%, and 0.4%,

respectively. This suggests that again exchange of inventions

is the major engine of growth among advanced nations. The








84

differences between group 1 and 2 lie in a higher effect of

rest of the world's resident patents and on the effect of

home-based technology which is almost zero in the full sample

but positive in the advanced one.

Group 3 reveals that once again the main source of

technology effect on growth of follower countries is access to

the pool of technology in the world (17%). Nonresident patents

and nonresident patents as a ratio of GDP effects are 0.74%

and 0.71%, respectively. This supports the idea that imitation

and adaptation of technology is the major source of follower

countries' growth, especially in the NICs.

Conclusion

This chapter studies technology's contribution to growth.

In advanced economies, consistent with their comparative

advantage in R&D, domestic technology proxied by resident

patents strongly influences growth. My results also suggest

that imitation by less advanced countries of technologies

developed in those more advanced nations is a main source of

growth in these nations. I also find that in advanced

countries, inventions per scientist and engineer raise the

rates of growth, consistent with the literature on invention

exhaustion. Finally, I find that the potential for enhanced

division of labor, through interrelated growth of advanced and

follower countries is associated with higher growth in the

advanced countries, but that technology acquisition in the








85
newly industrializing countries is associated with

deceleration of growth in the advanced countries.

The results of this chapter also confirm income

convergence, though conditioned on the other factors, such as

imitating of technology transferred from leading countries by

follower countries. The fact is that leading countries are

paying the price for the catching up of less advanced

countries. Through transferring technology these follower

countries enjoy fast income growth.

Finally, a possible interpretation of this chapter is

that the great impact of the spillover effects from technology

may justify government subsidies to R&D sectors in all

countries.


Notes
1. Barro (1991) has used the same proxies for human capital.
The reason I used these variables instead of my own variable
for human capital as in chapter 3 is that by using my
variables have fewer observations.

2. One exception is government spending. In the advanced
countries this has positive sign but is insignificant,
suggesting that probably there is a zero impact of government
spending on growth in advanced countries and it is in contrary
with was the negative effect of government spending in full
sample. It might be that there are better government policies
in the advanced countries.

3. Pacific Rim countries include Japan, Korea, Philippines,
and Singapore.











Table 4-4
Regressions for per Capita GDP Growth in the Full Sample
Experiments with the Effect of Technology Creation and Transfer
(T-statistics in Parentheses)

VARIABLE (4.1) (4.2) (4.3) (4.4) (4.5) (4.6)


CONSTANT


LOG(GDP60)

ENROLLMENT RATIO:
PRIMARY


SECONDARY

GDP SHARE OF:
INVESTMENT

GOVERNMENT
SPENDING


GDP DEVIATION


REVOL. & COUPS


REAL OIL PRICE
DEVIATION


0.013 0.057
(0.41) (2.06)

-0.023 -0.027
(-6.86) (-7.63)


0.172
(6.55)

-0.026
(-6.96)


0.142 0.171 0.034
(5.84) (6.31) (1.00)


-0.022
(-6.32)


-0.026
(-6.71)


-0.026
(-7.17)


0.024 0.019 0.023 0.025 0.024 0.021
(2.31) (1.87) (2.13) (2.30) (2.26) (1.99)


0.041 0.047
(3.09) (3.77)


0.027 0.024
(2.31) (2.01)


0.175 0.183 0.179 0.196
(7.27) (7.58) (7.16) (7.57)


-0.131 -0.133
(-4.99) (-5.13)


0.026 0.042
(2.17) (3.15)


0.184
(7.40)


0.181
(7.48)


-0.137 -0.135 -0.131 -0.129
(-5.15) (-4.90) (-4.94) (-4.93)


0.420 0.413 0.409 0.387 0.397 0.417
(4.01) (3.95) (3.80) (3.56) (3.68) (3.99)

-0.008 -0.008 -0.010 -0.010 -0.010 -0.008
(-2.25) (-2.18) (-2.44) (-2.61) (-2.44) (-2.14)


-0.017 -0.017
(-2.88) (-2.87)


-0.016
(-2.54)


-0.016
(-2.57)


-0.016 -0.017
(-2.55) (-2.90)


INTERNATIONAL
AND CIVIL WAR
YEAR (DUMMY) 0.012 0.015 0.016 0.016
(1.86) (2.21) (2.41) (2.36)

REST OF THE WORLD
RESIDENT PATENTS 0.111 0.094 - -
(6.10) (6.56)

RESIDENT PATENTS 0.062 - - -
(2.20)

NONRESIDENT PATENTS - 0.086 - -
(2.77)


0.016
(2.39)


0.013
(1.98)


0.107
(5.89)

0.036
(1.16)

0.069
(2.04)











Table 4-4--continued


VARIABLE (4.1) (4.2) (4.3) (4.4) (4.5) (4.6)


RATIO OF
NONRES. PATENTS/GDP - - 0.122 - - -
(2.79)

RATIO OF
RESID. PATENTS/GDP - - - -0.113 - -
(-0.75)

NONRESIDENT PATENTS
PER CAPITA - -- -- 0.012 -
(2.27)
F 18.18 18.52 15.77 14.87 15.44 17.10

R2 0.24 0.24 0.19 0.18 0.19 0.24

NOB 611 611 611 611 611 611


all the countries in Table 4-1.


Note: The full sample is











Table 4-5
Regressions for per Capita GDP Growth in the Advanced Countries
Experiments with the Effects of
Invention and Inventive Efficiency
(T-statistics in Parentheses)

VARIABLE (5.1) (5.2) (5.3) (5.4) (5.5) (5.6)


CONSTANT


0.164
(2.32)


0.130
(2.05)


0.142
(1.91)


0.075 0.089 0.066
(0.99) (1.18) (0.90)


LOG(GDP60)


ENROLLMENT RATIO:
PRIMARY


SECONDARY


GDP SHARE OF:
INVESTMENT

GOVERNMENT
SPENDING


GDP DEVIATION


REAL OIL PRICE
DEVIATION


REVOL. & COUPS


REST OF THE WORLD
RESIDENT PATENTS


RESIDENT PATENTS


-0.023 -0.020 -0.021 -0.020
(-2.82) (-2.78) (-2.65) (-2.57)


0.010
(0.87)


0.012 0.009
(1.02) (0.68)


0.019 0.025 0.021
(1.65) (2.15) (1.42)


0.133
(4.47)


-0.021 -0.019
(-2.72) (-2.54)


0.012 0.009
(0.91) (0.66)

0.020 0.010
(1.38) (0.80)


0.016
(1.24)

0.023
(1.97)


0.133 0.136 0.110 0.113 0.147
(4.51) (4.35) (3.46) (3.54) (4.36)


0.026 0.027 0.029
(0.73) (0.76) (0.80)


0.371 0.367
(4.05) (4.05)


-0.018
(-3.38)


-0.017
(-3.34)


0.367
(4.04)


0.028
(0.77)

0.379
(4.24)


0.028
(0.79)

0.380
(4.24)


0.041
(1.14)

0.372
(4.19)


-0.017 -0.017 -0.017 -0.018
(-3.35) (-3.39) (-3.43) (-3.53)


-0.001 -0.0004 -0.0004
(-0.15) (-0.06) (-0.06)


-0.001 -0.001 -0.004
(-0.22) (-0.27) (-0.67)


0.051 0.056 0.035
(3.11) (3.52) (2.58)


0.007 0.050 0.065
(0.33) (1.97) (2.75)


NONRESIDENT
PATENTS


0.064
(3.59)










Table 4-5--continued


VARIABLE (5.1) (5.2) (5.3) (5.4) (5.5) (5.6


RATIO OF RESIDENT
PATENTS / S&E - 0.034 0.033 0.022 - -
(2.41) (2.16) (1.45)

RESIDENT PATENTS
PER CAPITA 0.013 -- -- -- -- -
(0.97)


F 5.81 6.46 5.80 6.33 6.73 7.38

R2 0.14 0.15 0.15 0.18 0.17 0.19

NOB 269 269 269 269 269 269

Note: See Table 4-1 for the definition of the advanced countries;
basically these are high income western countries.











Table 4-6


Regressions for per Capita GDP Growth in the Advanced Countries
Experiments with Division of Labor and World
Redistribution of Output through Inter-related
Growth of Advanced and Follower Countries
(T-statistics in Parentheses)

VARIABLE (6.1) (6.2) (6.3) (6.4) (6.5) (6.6)


AVERAGE INCOME
GROWTH OF NICs
COUNTRIES WEIGHTED
BY GDP SHARE 6.810
(7.48)

AVERAGE INCOME
GROWTH OF PACIFIC
RIM COUNTRIES
WEIGHTED BY GDP
SHARE -


AVERAGE INCOME
GROWTH OF NON-PACIFIC
RIM COUNTRIES OF
NICs SAMPLE WEIGHTED
BY GDP SHARE -


RATIO OF RESIDENT
PATENTS/S&E -



RATIO OF PATENT/S&E -


2.800
(4.83)





0.566
(1.03)


0.025
(1.74)


0.005
(2.06)


0.029
(2.08)


0.009
(3.33)


AVERAGE PATENT/S&E
OF NICs COUNTRIES
WEIGHTED BY
GDP SHARE

AVERAGE RESIDENT
PATENTS/S&E
OF NICs COUNTRIES
WEIGHTED BY
GDP SHARE


- -0.0001 -
(-3.52)




- - -0.0008
(-2.39)











Table 4-6--continued


VARIABLE (6.1) (6.2) (6.3) (6.4) (6.5) (6.6)


AVERAGE PATENT/S&E
OF PACIFIC RIM
COUNTRIES WEIGHTED
BY GDP SHARE - - - -0.200 -
(-0.64)

AVERAGE RESIDENT
PATENTS/S&E
OF PACIFIC RIM
COUNTRIES WEIGHTED
BY GDP SHARE - -- -- -2.151
(-4.25)


F 14.80 13.49 8.85 7.24 7.19 8.91

R2 0.29 0.29 0.21 0.17 0.17 0.21


NOB 269 269 269 269 269 269


NOB


269 269 269 269


269 269











Table 4-7
Regressions for per Capita GDP Growth in the Follower Countries
Experiments with the Effect of Technology Transfer
(T-statistics in Parentheses)

VARIABLE (7.1) (7.2) (7.3) (7.4) (7.5)


CONSTANT


0.011 0.019 0.164
(0.26) (0.48) (4.77)


LOG(GDP60)


ENROLLMENT RATIO:
PRIMARY


SECONDARY


GDP SHARE OF:
INVESTMENT

GOVERNMENT
SPENDING


GDP DEVIATION



REAL OIL PRICE
DEVIATION

INTERNATIONAL
AND CIVIL WAR
YEAR (DUMMY)

REST OF THE WORLD
RESIDENT PATENTS


RESIDENT PATENTS


RATIO OF RESIDENT
PATENTS / S&E


-0.027 -0.029
(-5.72) (-5.88)


0.011 0.012
(0.73) (1.87)


0.061 0.058 0.016
(2.14) (2.64) (0.54)


0.224 0.220
(6.83) (6.84)

-0.166 -0.167
(-5.03) (-5.47)


0.199
(5.38)


-0.026 -0.030
(-5.17) (-5.71)


0.230
(5.77)

-0.033
(-6.09)


0.026 0.021 0.020
(1.70) (1.41) (1.37)


0.003 -0.011
(0.18) (-0.53)


0.224 0.203 0.217
(6.43) (6.00) (6.52)

-0.145 -0.171 -0.187
(-4.30) (-5.01) (-5.37)


0.380 0.372 0.337 0.368 0.340
(2.63) (2.58) (2.24) (2.45) (2.28)


-0.020 -0.020 -0.017
(-2.42) (-2.38) (-1.97)


-0.016
(-1.92)


-0.016
(-1.88)


0.010 0.009 0.017 0.018 0.019
(1.16) (1.07) (1.79) (1.94) (2.03)


0.136 0.138 - - -
(5.47) (6.23)

-0.008 - - - -
(-0.14)


- -0.00002 -
(-1.24)


continued next page











Table 4-7--continued


VARIABLE (7.1) (7.2) (7.3) (7.4) (7.5)


RATIO OF
NONRES. PATENTS/GDP


RATIO OF
RESID. PATENTS/GDP


NONRESIDENT PATENTS
PER CAPITAL


0.311
(2.56)


-0.255
(-0.70)


0.00009
(3.24)


14.92 15.14 11.29 12.12 12.67


0.25


NOB


413


0.24

413


0.18

413


0.19

413


0.20

413


Note: The follower countries are all the countries except the
advanced countries. See the notes to Table 4-1.











Table 4-8
Estimated Mean Contributions to Economic Growth
(T-statistics in Parentheses)

VARIABLE FULL ADVANCED FOLLOWER
SAMPLE COUNTRIES COUNTRIES


REST OF THE WORLD 0.136 0.067 0.170
RESIDENT PATENTS (6.10) (3.52) (6.23)


RESIDENT PATENTS 0.0019 0.0036
(2.20) (2.75)


NONRESIDENT PATENTS 0.0039 0.0069
(2.77) (3.59)

RATIO OF NONRESIDENT 0.0046 0.0071
PATENTS/GDP (2.79) (2.56)


NONRESIDENT PATENTS 0.0029 0.0074
PER CAPITA (2.27) (3.24)


RATIO OF RESIDENT 0.0036
PATENTS/S&E (2.41)


AVERAGE INCOME 0.0149
GROWTH OF NICs (7.48)
COUNTRIES WEIGHTED
BY GDP SHARE

AVERAGE RESIDENT -0.0068
PATENTS/S&E (-2.39)
OF NICs COUNTRIES
WEIGHTED BY
GDP SHARE



NOTE: Estimated contributions are regression coefficients times
means of variables.














CHAPTER 5

SUMMARY AND CONCLUSIONS

The goal of this dissertation has been to explore the

causes of differences in growth performance across countries,

guided by the new theories of endogenous economic growth.

Chapter 2 tests the impact of oil price shocks in advanced

countries. Chapter 3 studies the effects of human capital,

both its initial level and its rate of growth. Chapter 4

investigates the effect of technology, both in its creation

and in its transfer, on the growth of advanced and developing

nations.

Chapter 2 focuses on government policy toward the

transportation sector. The interaction between oil price

changes and the share of petroleum-based transportation in ten

OECD countries is studied. It has been shown that fluctuations

in oil-prices have trend as well as more conventional cyclical

effects on the patterns of economic growth in advanced

nations. The analysis of data for ten OECD countries

demonstrates, for example, that the relative share of a

country's transportation sector in its total consumption of

oil has a major impact on that nation's petroleum dependency.

The larger the transportation sector's share, the more

sensitive to oil-shocks a nation's economy is. More important,