Decomposition of ethnic heterogeneity on growth.
Yamamura, Eiji ; Shin, Inyong
INTRODUCTION
Since the 1990s, there has been growing interest among economic
researchers in the relationship between ethnic diversity and economic
performance (Alesina and La Ferrara, 2005). Easterly and Levine (1997)
showed a negative association between ethnic heterogeneity and economic
growth. (1) Ethnic heterogeneity is thought to influence economic growth
through several channels. First, ethnic heterogeneity has been found to
reduce investment (Mauro, 1995; Montalvo and Reynal-Querol, 2005a, b),
thus reducing public and private capital formation and economic growth.
(2) Second, ethnic heterogeneity has been shown to be positively
associated with the probability of conflict (Montalvo and Reynal-Querol,
2005a, b) and it is negatively related to trust (Dincer, 2011). (3)
Trust plays a key role in reducing transaction costs in the market (Zak
and Knack, 2001). Thus, we predict that ethnic heterogeneity impedes not
only market transactions but also information spillovers including
learning from others. This inevitably hinders economic growth. However,
the effect of ethnic heterogeneity on growth is open for discussion
because heterogeneity also appears to have a contrasting effect: social
diversity, which seems to be captured partly by ethnic heterogeneity, is
thought to promote innovation (Jacobs, 1969, 1984). If this is true,
then heterogeneity may enhance economic growth. In the analysis used in
this paper, technological progress, regarded as a proxy for innovation,
is exogenously determined. Hence, this paper does not examine the
influence of ethnic heterogeneity on growth through enhanced innovation.
Data envelopment analysis, hereafter DEA, enables us to analyze how
ethnic diversity influences economic growth. We use DEA analysis to
construct a world production frontier, and then decompose labor
productivity growth into three components: technological catch-up,
capital deepening, and technological change (Kumar and Russell, 2002).
(4) Alternatively, we have used regression analysis to examine how
initial outputs per worker influence these components (Yamamura and
Shin, 2007a, b, 2008; Yamamura, 2011).
This paper aims to examine the influence of heterogeneity on growth
and therefore provide new empirical evidence by analyzing the channels
through which ethnic heterogeneity affects growth. The key finding is
that heterogeneity has a negative effect on efficiency improvements,
which impedes growth. The rest of this paper is organized as follows:
two testable hypotheses are proposed in the next section; the subsequent
section describes the data and estimation strategy; the penultimate
section presents the estimation results; and the final section
concludes.
HYPOTHESES
The engine of economic growth seems to stem from information
spillovers (Marshall, 1920). Positive externalities brought about by
information spillovers among various firms and groups are expected to
arise if face-to-face interaction among workers occurs. (5) Information
spillovers are thought to enhance efficiency improvement, resulting in
economic growth. However, most workers, including experts, are less
likely to interact if the workforce is ethnically polarized. This is
consistent with the argument that information flows are weaker in a
heterogeneous population, which prevents individuals from learning about
their neighbors' experiences (Munshi, 2004). If this holds true,
heterogeneity has a detrimental influence on information spillovers.
Furthermore, there appears to be an additional mechanism. It is argued
that trust contributes to economic growth (eg, Beugelsdijk et al., 2004;
Beugelsdijk and van Schaik, 2005; Zak and Knack, 2001). This is in part
because trust reduces the transaction costs among agents. However,
Dincer (2011) finds that ethnic heterogeneity is negatively associated
with the level of trust. If this is true, then heterogeneity reduces
trust, and therefore increases transaction costs. Inevitably, the market
functions less well, which in turn reduces positive externalities such
as information spillovers. Accordingly, economic growth is hindered.
This argument leads us to postulate Hypothesis 1. (6)
Hypothesis 1: Racial heterogeneity impedes efficiency improvements,
which hampers economic growth.
Ethnic heterogeneity increases the number of interest groups
because the interests of each ethnic group may be different from, and in
conflict with, those of other groups. A rent-seeking model shows that
resources spent by each group to obtain political power can be
considered a social cost (Mueller, 2005). In this model, resources are
allocated to nonproductive behavior and not into productive investments.
It is also possible that ethnic heterogeneity increases the likelihood
of political conflict, creating an unstable and uncertain political
situation in a country. As a consequence, investment is reduced due to
this greater uncertainty (Montalvo and Reynal-Querol, 2005a, b). (7)
Considering the arguments above, we propose Hypothesis 2.
[FIGURE 1 OMITTED]
Hypothesis 2: Racial heterogeneity impedes capital accumulation,
which hampers economic growth.
DATA AND ESTIMATION STRATEGY
Kumar and Russell (2002) used DEA to construct a cross-country data
set by decomposing labor productivity growth into three components. They
then estimated a simple OLS regression model. In that model, the
dependent variables were the percentage changes between 1965 and 1990
for output per worker, technological change, the efficiency index, and
the capital accumulation index. Output per worker in 1965 was an
independent variable. In their estimations, both unobservable individual
and time effects were ignored. However, as suggested by Yamamura and
Shin (2007a), this can lead to estimation bias.
Following Kumar and Russell (2002), this paper also uses DEA to
construct a panel data set for 57 countries, from 1965 to 1990, using
the Penn World Tables. (8) In Figures 1 and 2 the vertical and
horizontal axes indicate output per worker and capital stock per worker,
respectively. Figure 1 illustrates the frontier line and location of
each country in 1960, whereas Figure 2 illustrates the frontier line and
location of each country in 1990. In the Appendix, Table A1 shows the
codes for the countries used in each figure, matched with the name of
country. Compared with 1960, the frontier line for 1990 in Figure 2 has
shifted upward, suggesting significant improvement in technology.
Furthermore, in 1990 there are greater differences in output and capital
stock among countries.
[FIGURE 2 OMITTED]
To assess the effects of ethnic heterogeneity on growth, we
constructed panel data for the countries included in our DEA analysis.
With this data set, we used random-effects estimations to reduce omitted
variable bias caused by the time-invariant features of the various
countries. (9) We also incorporated year dummies into this model to
capture individually invariant time-specific effects. The estimated
function takes the following form:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)
where [Gr.sub.jT-t0] represents labor productivity growth and the
change in any of the two dependent variables (ie, Efficiency and
Capital) in country j from each base year t0 to next year T (t0 = 1965,
..., 1989 and T = 1966, ..., 1990). Technological progress, as measured
by the outward shift in the frontier, is determined exogenously by a
world production frontier. Hence, the characteristics of each country do
not influence technological progress. To put it differently,
technological progress is exogenously determined for all countries and
thus cannot be used as a dependent variable. The estimation model
basically follows the model of Kumar and Russell (2002) and therefore
Ln(Output), the log of GDP per capita, is included to capture the level
of economic development. (10) The dependent variables and Ln(Output) are
from the Penn World Tables (PWT 5.6). (11) In equation 1, [alpha]
represents the regression parameters, [epsilon] is the time-invariant
individual effect for each country, v represents the year-specific
effects, and u is an error term. As stated earlier, [epsilon] and v are
controlled. The key independent variable that captures ethnic
heterogeneity is the ethnic polarization index, which has been
extensively used to capture ethnic heterogeneity (Mauro, 1995; Easterly
and Levine, 1997). The index for some country j is defined as:
Fractionalization = 1 - [n.summation over (i=1)] [[pi].sup.2.sub.i]
where [[pi].sub.i] is the proportion of the population who profess
to belong to a given ethnic group i. Basically, this indicator can be
interpreted as measuring the probability that two randomly selected
individuals in a country will belong to different groups.
In addition to the ethnic fractionalization index, an ethnic
polarization index has been developed and used as an alternative measure
(Montalvo and Reynal-Querol, 2005a, b; Reynal-Querol, 2002). The ethnic
polarization index can be defined as:
Polarization = 1 - [n.summation over (i=1)] [(0.5 -
[[pi].sub.i]/0.5).sup.2] [[pi].sub.i]
This index measures the normalized distance of a particular
distribution of ethnic groups within a bimodal distribution. Here,
ethnic group is represented as i for country j. The index can be
calculated for each country.
To check the robustness of the estimation results, we used both
ethnic fractionalization and ethnic polarization as proxy variables for
ethnic heterogeneity. (12) Ethnic heterogeneity is expected to result in
conflict, hampering the cooperation and communication required to
enhance technology diffusion, and therefore efficiency improvements.
Proxies for ethnic heterogeneity hold time-invariant features. Hence,
their effects cannot be estimated when a fixed-effects model is used. To
examine these effects, a random-effects model is used in this paper. If
the coefficients of the proxies take the negative sign when efficiency
improvement or capital accumulation is a dependent variable, then
Hypothesis 1 or 2, as is applicable, is supported.
The other independent variables used in this model include the
values of the dependent variable in the base year, t0. Natural disasters
are considered to influence economic growth (Skidmore and Toya, 2002).
To capture this effect, the number of natural disasters that have
occurred in each sample country is included. (13) We investigate how a
natural disaster occurring in year t0 affects growth rates between t0
and t1. For instance, we examine the effect of the number of natural
disasters in 1965 on growth rates from 1965 to 1966. (14) As suggested
by Yamamura (2011), government size hinders capital accumulation and
thus hampers economic growth. Hence, government size is included as an
independent variable. Government size is measured by a country's
general government final consumption expenditure as a percentage of GDP,
sourced from the World Bank (2006). To capture the human capital effect,
the number of years of schooling is incorporated, as used by Easterly
and Levine (1997). (15)
Institutional factors appear to play an important role in
determining economic growth. A number of previous works have shown that
legal origin is profoundly associated with incentives for economic
agents, and therefore with economic performance (eg, La Porta et al.,
1997, 1998, 1999, 2008). Better-developed financial systems contribute
to growth in capital-intensive sectors (Rajan and Zingales, 1998).
Further, Levine (1998) argued that legal origin exogenously determined
the degree of financial development that promoted economic growth. La
Porta et al. (1998) asserted that French civil-law countries offer the
weakest legal protection to investors while British common-law countries
offer the strongest. French and British legal origin dummies are
incorporated to capture these effects. (16) Apart from institutional
factors, geographical factors such as latitude and land size are
incorporated as independent variables to capture the existence of
natural resources and climate.
The source of each variable is presented in Table A2. Table 1 shows
the summary statistics for each variable used in the estimation (mean
value, standard deviation, maximum value, and minimum value).
Furthermore, Table 2 presents a correlation matrix of the variables.
Table 2 shows that the proxies for ethnic heterogeneity are negatively
correlated with labor productivity growth, both efficiency improvement
and capital accumulation. Further, the correlation between them is
statistically significant at the 1% level. This is consistent with our
expectations. However, for a more detailed examination, the regression
results must also be examined. (17)
RESULTS
The estimation results of the random-effects model with year dummy
variables from 1966 to 1990 are reported in Tables 3-5. The results
based on all observations available for the estimation are exhibited in
columns (1) and (3) of each table. However, it is well known that DEA
frontier estimates are strongly influenced by outliers. Hence, we also
conducted an estimation where outliers were excluded to see how
sensitive the estimates are to their removal. As shown in Figures 1 and
2, there are countries located on the frontier, such as the United
States (USA), Luxemburg (LUX), and Hong Kong (HKG). Other countries,
such as Canada (CAN), Switzerland (CHL) and Norway (NOR), are farther
away. These countries can be considered as outliers, and, therefore,
excluded. The excluded countries are included in columns (1) and (3),
but not in columns (2) and (4) in Tables 3, 4, and 5. There were 1,312
observations used in the estimations for columns (1) and (3), whereas
1,187 were used for columns (2) and (4}. Hence, excluding outliers
reduced the sample size by 9.5 %. Table 3 presents the results when
labor-productivity growth is used as the dependent variable, Table 4
shows the results when efficiency improvement is used as the independent
variable, and Table 5 reports the results when capital accumulation is
the independent variable. The ethnic fractionalization and ethnic
polarization indexes are used as a proxy for ethnic heterogeneity in
each table. An F-test was conducted to check for unobservable individual
effects and time effects. In all columns of Tables 3-5 the results of
the F-test indicate that unobservable individual effects and time
effects exist. Hence, these effects need to be controlled for using a
fixed effects or random-effects estimation. In all columns in Tables
3-5, with the exception of column (4) of Table 3, the Hausman test does
not reject the null hypothesis that the differences in coefficients
between a fixed-effects model and a random-effects model are not
systematic. This result implies that the random-effects model is valid
and preferred.
We see in Table 3 that ethnic fractionalization yields a negative
sign in column (1) and that ethnic polarization has a negative sign in
columns (3). However, these coefficients are statistically significant
in both columns (1) and (3). This implies that ethnic heterogeneity
reduces labor-productivity growth. In addition, the number of natural
disasters shows a significant positive sign in column (1), which is
consistent with the argument of Skidmore and Toya (2002), where natural
disasters may stimulate economic growth. Other variables do not show a
significant sign in column (3), and hence they do not influence growth.
As shown in columns (2) and (4), the results do not change when outliers
are excluded. This suggests that the results for ethnic heterogeneity
are robust. With respect to Table 4, the coefficient signs for ethnic
fractionalization are negative in column (1), and that of ethnic
polarization is also negative in column (3). Both are statistically
significant. These results suggest that ethnic heterogeneity impedes
information spillover. Furthermore, the results in columns (2) and (4)
suggest that the results do not change when outliers are excluded. This
indicates that the results for columns (1) and (3) are robust. Hence,
the detrimental effect of ethnic heterogeneity on growth comes in part
from the detrimental effect of ethnic heterogeneity on information
spillover. Thus, Hypothesis 1 is strongly supported by the result.
Concerning the other variables, the results are almost identical to the
results presented in Table 3.
We see from Table 5 that the signs of the proxies for ethnic
heterogeneity are positive in column (1) and negative in columns
(2)-(4). However, they are not statistically significant. This indicates
that outliers do not influence the results for ethnic heterogeneity.
Ethnic heterogeneity thus does not affect capital accumulation, which is
not consistent with the argument that ethnic heterogeneity reduces
investment (Mauro, 1995). Hence, Hypothesis 2 is not supported by the
results. Considering Tables 3-5 jointly, we conclude that ethnic
heterogeneity impedes growth by reducing information spillovers rather
than by reducing investment. The combined effects of ethnic
heterogeneity on growth are negative, and thus, we conclude that ethnic
heterogeneity is an obstacle to, rather than an engine of, economic
growth.
CONCLUSIONS
There are conflicting views regarding the role of ethnic
heterogeneity or diversity on growth. Social heterogeneity is thought to
impede investment, reducing capital accumulation. What is more,
heterogeneity is thought to hinder information spillovers, which hampers
efficiency improvements over time. These have a detrimental effect on
economic growth. This paper attempts to examine the influence of
heterogeneity on economic growth by scrutinizing the channels through
which heterogeneity affects such growth.
We used panel data from 55 countries, from 1965 to 1989, to
decompose the effect of ethnic heterogeneity on growth. Using a
random-effects regression model with year dummies, we found that ethnic
heterogeneity has a negative effect on growth, mainly by hampering
efficiency improvements, but not by reducing capital accumulation. We
interpret these results to mean that ethnic heterogeneity hinders
cooperation and communication among individuals and that cooperation and
communication are important for technology diffusion. As a consequence,
efficiency improvement is hampered, thereby impeding economic growth. In
contrast, heterogeneity does not affect capital accumulation.
Information spillovers play an important role in developing
countries; it enables them to catch up with more developed countries
because it is otherwise difficult for them to create new technologies on
their own (Vernon, 1966). From the findings in this paper, we derive the
argument that heterogeneity is an obstacle to economic development,
particularly for developing countries trying to catch up with developed
countries via the acquisition of new technologies.
Information spillovers that occur through interactions among
workers from various industries is thought to largely occur in urban
rather than in rural areas (Jacobs, 1969, 1984). Such spatial factors
were not considered in this study when the estimations were conducted
because we used country-level macro data. Micro-level data should be
used to more closely explore the effect of heterogeneity on information
spillovers and therefore efficiency improvement. Furthermore, the
influence of institutional factors on investment differs between private
and public investment (Baliamoune-Lutz and Ndikumana, 2008). However,
due to data limitations, our research was unable to examine how
heterogeneity influences private and public capital accumulation. These
remaining issues should be addressed in future research.
APPENDIX
Table A3: Countries used in DEA
Code Country name
ARG Argentina
AUS Australia
AUT Austria
BEL Belgium
BOL Bolivia
CAN Canada (a)
CHL Switzerland (a)
COL Colombia
DNK Denmark
DOM Dominican Republic
ECU Ecuador
FIN Finland
FRA France
DEU Germany
GRC Greece
GTM Guatemala
HND Honduras
HKG Hong Kong, China (a)
ISL Iceland
IND India
IRL Ireland
ISR Israel
ITA Italy
CIV Ivory Coast
JAM Jamaica
JPN Japan
KEN Kenya
KOR Korea, Rep.
LUX Luxembourg (a)
MDG Madagascar
MWI Malawi
MUS Mauritius
MEX Mexico
MAR Morocco
NLD Netherlands
NZL New Zealand
NGA Nigeria
NOR Norway (a)
PAN Panama
PRY Paraguay
PER Peru
PHL Philippines
PRT Portugal
SLE Sierra Leone
ESP Spain
LKA Sri Lanka
SWE Sweden
CHE Switzerland
SYR Syrian Arab Republic
TWN Taiwan, China (b)
THA Thailand
TUR Turkey
GBR United Kingdom
USA United States (a)
YUG Yugoslavia (b)
ZMB Zambia
ZWE Zimbabwe
(a) indicates countries excluded in sample used for estimation in
columns (2) and (4) of Tables 3, 4, and 5.
(b) indicates countries excluded in all columns of Tables 3, 4,
and 5.
Table A2: Source of data
Source
Ln(Output) Penn World Table 5.6
Capital Stock per capita Penn World Table 5.6
Ethnic fractionalization Website of Marta Reynal-Querol (a)
Ethnic polarization Website of Marta Reynal-Querol (a)
Number of natural disasters EM-DAT (Emergency Events Database) (b)
Government size World Bank (2006)
Land size World Bank (2006)
French legal origin Website of Andrei Shleifer (c)
British legal origin Website of Andrei Shleifer (c)
Land size World Bank (2006)
Note: With the exception of the World Bank (2006), the data were
obtained from the internet as follows:
(a) http://www.econ.upf.edu/}{{{-}{}}}{reynal/data_web.htm
(accessed on December 1, 2011).
(b) http://www.emdat.be (accessed on June 1, 2011).
(c) http://www.economics.harvard.edu/faculty/shleifer/dataset
(accessed on June 2, 2011).
REFERENCES
Alesina, A and La Ferrara, E. 2005: Ethnic diversity and economic
performance. Journal of Economic Literature 43 (3): 762-800.
Alesina, A, Baqir, R and Easterly, W. 1999: Public goods and ethnic
divisions. Quarterly Journal of Economics 114(4): 1243-1284.
Alesina, A, Devleeschauwer, A, Easterly, W, Kurlat, S and Wacziarg,
R. 2003: Fractionalization. Journal of Economic Growth 8(2): 155-194.
Baliamoune-Lutz, M and Ndikumana, L. 2008: Corruption and growth in
African countries: Exploring the investment channel. Working papers
2008-8, University of Massachusetts: Amherst, Department of Economics.
Beugelsdijk, S, de Groot, HLF and van Schaik, A.B.T.M. 2004: Trust
and economic growth: A robustness analysis. Oxford Economic Papers
56(1): 118-134.
Beugelsdijk, S and van Schaik, A.B.T.M. 2005: Social capital and
growth in European regions: An empirical test. European Journal of
Political Economy 21 (2): 301-324.
Dincer, OC. 2011: Ethnic diversity and trust. Contemporary Economic
Policy 29(2): 284-293.
Easterly, B and Levine, R. 1997: Africa's growth tragedy:
Policies and ethnic divisions. Quarterly Journal of Economics 112 (4):
1203-1250.
Henderson, JD and Russell, RR. 2005: Human capital and convergence:
A production-frontier approach. International Economic Review 46:
1167-1205.
Isaksson, A. 2011: Social divisions and institutions: Assessing
institutional parameter variation. Public Choice 147(3): 331-357.
Jambs, J. 1969: The economies of cities. Vintage: New York.
Jacobs, J. 1984: The wealth of nations: Principles of economic
life. Vintage: New York.
Kumar, S and Russell, RR. 2002: Technological change, technological
catch-up, and capital deepening: Relative contributions to growth and
convergence. American Economic Review 92(3): 527-548.
La Porta, R, Lopez de Silanes, F and Shleifer, A. 2008: The
economic consequence of legal origins. Journal of Economic Literature
46(2): 285-332.
La Porta, R, Lopez de Silanes, F, Shleifer, A and Vishni, R. 1997:
Legal deterrents of external finance. Journal of Finance 57(3):
1131-1150.
La Porta, R, Lopez de Silanes, F, Shleifer, A and Vishni, R. 1998:
Law and finance. Journal of Political Economy 106(6): 1113-1150.
La Porta, R, Lopez de Silanes, F, Shleifer, A and Vishni, R. 1999:
Quality of government. Journal of Law, Economics and Organization 15(1):
222-279.
Levine, R. 1998: The legal environment, banks, and long-run
economic growth. Journal of Money, Credit and Banking 30(3): 596-613.
Lind, JT. 2007: Fractionalization and the size of government.
Journal of Public Economics 91 (1-2): 51-76. Marshall, A. 1920:
Principles of Economics. Macmillan Press: London.
Mauro, P. 1995: Corruption and growth. Quarterly Journal of
Economics 110(3): 681-712.
Montalvo, JG and Reynal-Querol, M. 2003: Religious polarization and
economic development. Economics Letters 80(2): 201-210.
Montalvo, JG and Reynal-Querol, M. 2005a: Ethnic polarization,
potential conflict and civil war. American Economic Review 95 (3):
796-816.
Montalvo, JG and Reynal-Querol, M. 2005b: Ethnic diversity and
economic development. Journal of Development Economics 76(2): 293-323.
Mueller, DC. 2003: Public Choice III. Cambridge University Press:
New York.
Munshi, K. 2004: Social learning in a heterogeneous population:
Technology diffusion in the Indian Green Revolution. Journal of
Development Economics 73 (1): 185-213.
Nishimizu, M and Page, J. 1982: Total factor productivity growth,
technological progress and technical efficiency change: Dimensions of
productivity change in Yugoslavia, 1965-78. Economic Journal 92(368):
920-936.
Rajan, RG and Zingales, L. 1998: Financial development and growth.
American Economic Review 88(3): 559-586.
Reynal-Querol, M. 2002: Ethnicity, political systems and civil war.
Journal of Conflict Resolution 46(1): 29-54.
Skidmore, M and Toya, H. 2002: Do natural disasters promote
long-run growth? Economic Inquiry 40(4): 664-687.
Thornton, RA and Thompson, P. 2001: Learning from experience and
learning from others: An exploration of learning and spillovers in
wartime shipbuilding. American Economic Review 91(5): 1350-1368.
Vernon, R. 1966: International investment and international trade
in the product cycle. Quarterly Journal of Economics 80(2): 190-207.
World Bank. 2006: World Development Indicators, CD-ROM. World Bank:
Washington, D.C.
Yamamura, E. 2011: Decomposition of the effect of government size
on growth. Economics Letters 112(3): 230-232.
Yamamura, E and Shin, I. 2007a: Technological change and catch-up
and capital deepening: Relative contributions to growth and convergence.
Economics Bulletin 15(3): 1-8.
Yamamura, E and Shin, I. 2007b: Dynamics of agglomeration economies
and regional industrial structure: The case of the assembly industry of
the greater Tokyo Region, 1960-2000. Structural Change and Economic
Dynamics 18(4): 483-499.
Yamamura, E and Shin, I. 2008: The benefit of efficiency
improvement on growth and convergence: A study using Japan panel data.
Economics Letters 99(1): 209-211.
Zak, PJ and Knack, S. 2001: Trust and growth. Economic Journal 111
(470): 295-321.
EIJI YAMAMURA [1] & INYONG SHIN [2]
[1] Department of Economics, Seinan Gakuin University, 6-2-92
Sawaraku Nishijin, Fukuoka 814-8511, Japan. E-mail:
yamaei@seinan-gu.ac.jp
[2] Asia University, 5-24-10 Sakai Musashino-shi, Tokyo 180-8629,
Japan.
(1) Previous works examined the effect of religious heterogeneity
on economic development, which relate to works exploring the influence
of ethnic heterogeneity (Alesina et al., 2003; Montalvo and
Reynal-Querol, 2003).
(2) Alesina et al. (1999) used US data to show that shares of
spending on productive public goods are inversely related to a
city's (metro area's/county's) ethnic fragmentation.
(3) Heterogeneity is found to influence government size (Lind,
2007). This also possibly affects economic performance and growth.
(4) The seminal work of Nishimizu and Page (1982) attempted to
decompose total productivity growth into technological progress and
technical efficiency change.
(5) Thornton and Thompson (2001), using micro-level data on wartime
shipbuilding, suggest that learning spillovers were a significant source
of productivity growth.
(6) Mauro (1995) exhibits a negative and significant correlation
between ethnic heterogeneity and institutional efficiency. Institutional
efficiency is positively associated with economic efficiency. It follows
then, with the exception of the information spillover channel, that
ethnic heterogeneity impedes efficiency improvements.
(7) A secured property right is considered to provide an incentive
to invest and therefore creates capital accumulation. Isaksson (2011)
used cross-countries data to present evidence that social division
measured in terms of ethnic fractionalization weakens the association
between property rights institutions and income. If this is true, then
heterogeneity reduces the incentive to invest even when property rights
are well secured.
(8) Kumar and Russell (2002) admitted that their method includes
the possibility of an implosion of the technological frontier. Henderson
and Russell (2005) precluded an implosion of the frontier over time. In
this paper, it is also precluded.
(9) The independent variables used in this paper were not available
for two of the 57 countries. Hence, data from just 55 countries were
used in the estimation, which are identified in Table A2, in the
Appendix.
(10) In alternative estimations, instead of the log for GDP per
capita, the initial level of technical efficiency (level of capital
stock) is included when change in technical efficiency (capital
accumulation) is examined. In these alternative estimations using a
random-effects model, the results for ethnic heterogeneity are similar
to those reported in this paper. However, the Hausman test does not
reject the null hypothesis that the results of a fixed-effects model are
systematically different from those of a random-effects model. This
suggests that the results of an estimation using a random-effects model
suffer from estimation bias. However, the effect of ethnic heterogeneity
is captured completely by country-specific effects in a fixed-effects
model. Hence, the results of an alternative specification are not
reported in this paper, but are available upon request from the author.
(11) The data are available from Center of International
Comparisons at the University of Pennsylvania,
http://pwt.econ.upenn.edu/ (accessed May 1, 2007).
(12) Data on ethnic fractionalization and polarization is available
at http://www.econ.upf.edu/ ~reynal/data_web.htm (accessed June 1,
2011).
(13) The data were obtained from the International Disaster
Database, http://www.emdat.be (accessed June 1, 2011).
(14) Wars and civil disturbances are also considered to have an
effect on growth. However, in less-developed countries, discontent
regarding governmental economic polity increases, which possibly
triggers a civil war or disturbance. That is, the occurrence of wars and
civil disturbances are thought to be influenced by economic growth.
Hence, wars and civil disturbances can be considered as endogenous
variables, resulting in estimation bias. It is for this reason that
these variables are not incorporated as independent variables.
(15) The number of years at school is available only for 1960,
1970, and 1980. Therefore, to construct the panel data, additional data
were generated by interpolation based on the assumption of constant
changes in rates to compensate for this deficiency. After 1980, the
value for 1980 is used. The data are available from
http://econ.worldbank.org/WBSITE/EXTERNAL/EXTDEC/EXT
RESEARCH/0,,contentMDK:20700002 ~ pagePK:64214825 ~ piPK:64214943 ~
theSitePK:469382,00.html (accessed June 2, 2011).
(16) In addition to legal origins, institutional factors captured
by, for instance, corruption and transparency of government appear to
influence economic growth. However, the variables used to capture them
are regarded as endogenous variables because the causality between these
factors and economic growth is ambiguous. Hence, these variables lead to
estimation bias and are not used in this paper. In contrast, legal
origin is related to historical events. Therefore, legal origin dummies
can be considered as exogenous variables and are used in this paper.
(17) Correlation between British legal origin and French legal
origin countries is -0.67 and statistically significant at the 1% level.
With the exception of British legal origin and French legal origin
countries, all other legal origin countries (German legal origin and
Scandinavian legal origin countries) are included in the sample. That
is, even if a country is not a French legal origin country, there is the
possibility that the country is not a British legal origin country.
Table 1: Basic statistics
Standard
Mean deviation
Gr(output) 1.01 0.05
Gr(efficiency) 0.99 0.04
Gr(capital) 1.01 0.02
Ln(Output) 9.19 0.87
Ethnic fractionalization 0.39 0.27
Ethnic polarization 0.48 0.26
Number of natural disasters 1.24 2.19
Government size 14.9 5.81
(Government expenditure/GDP) (%)
French legal origin 0.49 0.50
British legal origin 0.32 0.46
Latitude 0.33 0.19
Land size ([Km.sup.2]) 871,614 1,904,466
Maximum Minimum
Gr(output) 1.37 0.75
Gr(efficiency) 1.38 0.71
Gr(capital) 1.15 0.92
Ln(Output) 10.5 674
Ethnic fractionalization 0.90 0.01
Ethnic polarization 0.95 0.02
Number of natural disasters 22 0
Government size 43.4 3.84
(Government expenditure/GDP) (%)
French legal origin 1 0
British legal origin 1 0
Latitude 0.72 0.01
Land size ([Km.sup.2]) 9,158,960 1,042
Table 2: Correlation matrix
Gr Gr Gr
(output) (efficiency) (capital)
Gr(output) 1.00
Gr(efficiency) 0.85 *** 1.00
(0.00)
Gr(copital) 0.31 *** -0.11 *** 1.00
(0.00) (0.00)
Ln(Output) -0.01 0.03 -0.09 ***
(0.71) (0.15) (0.00)
Ethnic fractionalization -0.12 *** -0.09 *** -0.08 ***
(0.00) (0.00) (0.00)
Ethnic polarization -0.10 *** -0.08 *** -0.07 ***
(0.00) (0.00) (0.00)
Number of natural -0.002 0.03 -0.07 ***
disasters
(0.91) (0.13) (0.00)
Government size -0.09 *** 0.03 -0.25 ***
(0.00) (0.24) (0.00)
French legal origin -0.05 ** -0.07 *** 0.06 *
(0.03) (0.00) (0.02)
British legal origin -0.03 0.02 -0.16 ***
(0.20) (0.35) (0.00)
Latitude 0.05 ** 0.03 0.04 *
(0.02) (0.15) (0.07)
Land size -0.04 * -0.01 -0.11 ***
(0.06) (0.58) (0.00)
Ln Ethnic
(Output) fractionalization
Gr(output)
Gr(efficiency)
Gr(copital)
Ln(Output) 1.00
Ethnic fractionalization -0.43 ** 1.00
(0.00)
Ethnic polarization -0.14 *** 0.66 ***
(0.00) (0.00)
Number of natural 0.02 0.17 ***
disasters
(0.36) (0.00)
Government size 0.39 *** -0.19 ***
(0.00) (0.00)
French legal origin 0.01 0.02
(0.75) (0.36)
British legal origin -0.23 *** 0.29 ***
(0.00) (0.00)
Latitude 0.07 *** 0.03
(0.00) (0.16)
Land size 0.23 *** 0.25 ***
(0.00) (0.00)
Ethnic Number of Government
polarization natural size
disasters
Gr(output)
Gr(efficiency)
Gr(copital)
Ln(Output)
Ethnic fractionalization
Ethnic polarization 1.00
Number of natural -0.02 1.00
disasters (0.32)
Government size -0.14 *** -0.15 *** 1.00
(0.00) (0.00)
French legal origin 0.24 *** -0.04 * -0.26 ***
(0.00) (0.07) (0.00)
British legal origin 0.10 *** 0.10 *** 0.10 ***
(0.10) (0.00) (0.00)
Latitude -0.13 *** 0.14 *** -0.09 ***
(0.00) (0.00) (0.00)
Land size 0.16 *** 0.38 *** 0.06 **
(0.00) (0.00) (0.01)
French British Latitude
legal legal
origin origin
Gr(output)
Gr(efficiency)
Gr(copital)
Ln(Output)
Ethnic fractionalization
Ethnic polarization
Number of natural
disasters
Government size
French legal origin 1.00
British legal origin -0.67 *** 1.00
(0.00)
Latitude -0.06 ** -0.008 1.00
(0.01) (0.75)
Land size -0.15 *** 0.30 *** 0.18 ***
(0.00) (0.00) (0.00)
Land
size
Gr(output)
Gr(efficiency)
Gr(copital)
Ln(Output)
Ethnic fractionalization
Ethnic polarization
Number of natural
disasters
Government size
French legal origin
British legal origin
Latitude
Land size 1.00
*, **, and *** indicate significance at
the 10%, 5%, and 1% levels, respectively.
Numbers in parentheses are p-statistics.
Table 3: Determinants of labor-productivity
growth (random-effects estimates: 1965-1989)
(1) (2)
Full sample Excluding outliers
Ln(Output) -0.001 -0.002
(-0.52)- (-0.73)-
Ethnic fractionalization -0.030 *** -0.030 ***
(-3.45) (-3.19)
Ethnic polarization
Number of natural disasters 0.001 * 0.001
(1.86) (1.57)
Government size -0.001 * -0.001 *
(-1.77) (-1.58)
French legal origin -0.007 -0.007
(-1.06) (-1.03)
British legal origin -0.001 -0.001
(-0.09) (-0.23)
Latitude 0.01 0.01
(1.49) (1.12)
Land size -0.96 x [10.sup.2] -2.09 x [10.sup.2]
(-0.79) (-1.09)
Constant 1.05 *** 1.06 ***
(38.6) (36.3)
F-test (Year dummies) 3.28 2.79
p-value=0.00 p-value=0.00
F-test (Country dummies) 2.24 2.28
p-value=0.00 p-value=0.00
Housman test 33.1 25.2
p-value=0.18 p-value=0.55
Groups 55 50
Observations 1,312 1,187
(3) (4)
Full sample Excluding outliers
Ln(Output) 0.002 0.001
(0.91)- (0.44)
Ethnic fractionalization
Ethnic polarization -0.020 ** -0.018 *
(-2.28) (-1.96)
Number of natural disasters 0.001 0.001
(1.43) (0.99)
Government size -0.001 * -0.001 *
(-1.97) (-1.65)
French legal origin -0.007 -0.008
(-1.05) (-1.09)
British legal origin -0.002 -0.004
(-0.32) (-0.57)
Latitude 0.009 0.006
(0.95) (0.57)
Land size -1.70 x [10.sup.2] -2.70 x [10.sup.2]
(-1.40) (-1.36)
Constant 1.02 *** 1.03 ***
(39.8) (36.5)
F-test (Year dummies) 3.28 2.79
p-value=0.00 p-value=0.00
F-test (Country dummies) 2.43 2.49
p-value=0.00 p-value=0.00
Housman test 2.08 41.1
p-value=0.99 p-value=0.04
Groups 55 50
Observations 1,312 1,187
*, **, and *** indicate significance at the
10%, 5%, and 1% levels, respectively.
Note: Year dummies are not reported but are included in all
estimations as independent variables. Numbers in parentheses
are z-statistics.
Table 4: Determinants of efficiency improvement
(random-effects estimates: 1965-1989)
(1) (2)
Full sample Excluding outliers
Ln(Output) -0.001 -0.001
(-0.06)- (-0.56)-
Ethnic fractionalization -0.021 *** -0.018 ***
(-3.34) (-2.80)
Ethnic polarization
Number of natural disasters 0.001 * 0.001 *
(1.78) (1.68)
Government size 0.0001 0.0002
(0.50) (0.91)
French legal origin 0.001 -0.001
(0.50) (-0.26)
British legal origin 0.001 0.004
(0.01) (0.77)
Latitude 0.01 0.01
(1.44) (0.67)
Land size -0.73 x [10.sup.2] -1.33 x [10.sup.2]
(-0.85) (-0.99)
Constant 0.99 *** 1.00 ***
(49.2) (47.6)
F-test (Year dummies) 3.43 2.78
p-value=0.00 p-value=0.00
F-test (Country dummies) 1.95 1.90
p-value=0.00 p-value=0.00
Hausman test 28.4 19.1
p-value=0.38 p-value=0.85
Groups 55 50
Observations 1,312 1,187
(3) (4)
Full sample Excluding outliers
Ln(Output) 0.002 0.001
(1.40)- (0.58)-
Ethnic fractionalization
Ethnic polarization -0.013 ** -0.010 *
(-2.26) (-1.68)
Number of natural disasters 0.001 0.001
(1.24) (1.10)
Government size 0.001 0.0002
(1.24) (0.75)
French legal origin -0.004 -0.002
(-0.08) (-0.45)
British legal origin 0.006 0.002
(1.24) (0.40)
Latitude 0.006 0.001
(0.90) (0.23)
Land size -1.19 x [10.sup.2] -1.63 x [10.sup.2]
(-1.37) (-1.22)
Constant 0.96 *** 0.98 ***
(51.4) (49.9)
F-test (Year dummies) 3.43 2.78
p-value=0.00 p-value=0.00
F-test (Country dummies) 2.06 2.00
p-value=0.00 p-value=0.00
Hausman test 36.4 30.7
p-value=0.11 p-value=0.28
Groups 55 50
Observations 1,312 1,187
*, **, and *** indicate significance at the 10%, 5%, and 1%
levels, respectively.
Note: Year dummies are not reported but are included in all
estimations as independent variables. Numbers in parentheses
are z-statistics.
Table 5: Determinants of capital accumulation
(random-effects estimates: 1965-1989)
(1) (2)
Full sample Excluding outliers
Ln(Output) 0.004 *** 0.005 ***
(2.81)- (3.07)-
Ethnic fractionalization 0.0007 -0.001
(0.11) (-0.17)
Ethnic polarization
Number of natural 0.003 0.002
disasters (1.34) (0.86)
Government size -0.001 *** -0.001 ***
(3.99) (3.92)
French legal origin -0.004 -0.002
(-0.98) (-0.48)
British legal origin -0.006 -0.003
(-1.13) (-0.54)
Latitude 0.004 0.007
(0.67) (0.95)
Land size -1.79 x [10.sup.9] ** -2.35 x [10.sup.9] *
(-2.13) (-1.79)
Constant 0.99 *** 0.98 ***
(58.9) (55.4)
F-test (Year dummies) 18.7 19.6
p-value=0.00 p-value=0.00
F-test (Country dummies) 10.8 10.8
p-value=0.00 p-value=0.00
Housman test 27.9 29.9
p-value=0.41 p-value=0.31
Groups 55 50
Observations 1,312 1,187
(3) (4)
Full sample Excluding outliers
Ln(Output) 0.004 *** 0.005 ***
(3.01)- (3.40)-
Ethnic fractionalization
Ethnic polarization -0.003 -0.005
(-0.58) (-0.79)
Number of natural 0.003 0.002
disasters (1.31) (0.82)
Government size -0.001 *** -0.001 ***
(4.00) (3.93)
French legal origin -0.003 -0.001
(-0.61) (-0.10)
British legal origin -0.004 -0.001
(-0.83) (-0.21)
Latitude 0.004 0.007
(0.61) (0.90)
Land size -1.70 x [10.sup.9] ** -2.40 x [10.sup.9] *
(-2.06) (-1.80)
Constant 0.99 *** 0.98 ***
(62.1) (57.2)
F-test (Year dummies) 18.7 19.6
p-value=0.00 p-value=0.00
F-test (Country dummies) 11.6 11.9
p-value=0.00 p-value=0.00
Housman test 26.2 26.8
p-value=0.50 p-value=0.47
Groups 55 50
Observations 1,312 1,187
*, **, and *** indicate significance at the 10%, 5% and 1%
levels, respectively.
Note: Year dummies are not reported but are included in all
estimations as independent variables. Numbers in parentheses
are z-statistics.
