|
post-autistic
economics review |
|
issue 37
contents
PAE
Review index
home page |
Can a Heterodox Economist Use
Cross-country Growth Regressions?
Matthew
McCartney (School of Oriental and African Studies, UK) © Copyright: Matthew McCartney 2006 1. Introduction This paper is concerned with how economic growth is analysed by economists. Over the last fifteen years an extremely common method has been through cross-country growth regressions. Section two shows how economic theory demonstrates that there should be a strong link between changes in economic policy and economic growth. Despite the implications of such theory empirical results using cross-country growth regressions remain disappointing. Section two demonstrates this using relevant empirical results in both a general manner and specifically those between economic growth and fiscal policy, investment, education and R+D. The third
section shows that the long-term averages typically used in cross-country
growth regressions hide an important empirical reality of growth in
developing countries. The medium-term
growth averages used by cross-country regressions conceal the periods of
stagnation, growth spurts, structural breaks, volatility and instability that
actually characterise growth in developing countries. When confronted with these ‘empirical’
problems researchers typically stop and try to confront what they perceive as
an empirical challenge. Researchers
seek for better proxies for variables such as ‘education’, refine measures of
‘trade openness’ more precisely, and perhaps most commonly seek out longer
and better data sets. Other methods
have included the use of panel data and techniques such as ‘trimmed least
squares’. Some researchers venture
further realising that there may be more than an empirical problem at work,
in particular that the theory relating growth and economic policy may be more
complex than allowed for by simple cross-country regressions. If this is the case then the emphasis on
improved data and technical refinements to the econometrics may be fruitless. Section four explores a number of theoretical reasons why cross-country regressions may be an
intrinsically poor method to isolate the link between changes in policy and
changes in economic growth rates.
Those analysed here are complementarity
among policy variables, the relation between different theories of growth, hysteresis effects and dynamics. As demonstrated in this section, responses
by researchers to these theoretical problems have been much more ad hoc. Section five demonstrates the final problem of
cross-country growth regressions that has rarely been faced by orthodox
researchers. Far from being a
positivist statistical exercise, cross-country growth regressions are bound
to an underlying neo-classical assumption – that the growth process is
universal. Each individual country in
cross-section according to this view will provide evidence that can be used
to elucidate the one underlying universal economic relation. An increase in openness for example is
hypothesised to have the same effect on growth in all countries. There is a limited amount of evidence that
can be teased out of existing cross-country growth regressions that suggests
each growth experience should be treated as potentially unique, i.e. as a
case study. The last section concludes by suggesting that heterodox
or post-autistic economists should open up the assumption of universalism to
greater scrutiny and ask why the growth process may differ across time and
space. In practical terms this would
question the neo-classical assumption of universalism and with it the
‘one-size-fits-all’ programmes of liberalisation emanating from the World
Bank, IMF and other institutions. A case study approach to economic growth
would be justified on the assumption that growth processes are not
universal. Comparative and
historically informed case studies allow researchers to question the
assumption of universality rather than be forced to assume it true a priori. 2. The Analysis of Growth and Empirical Results This section shows there is a strong theoretical link
between policy change and economic growth but also that the empirical
evidence for this link is very weak.
This is shown here in both a general sense and specifically in
relation to fiscal policy, investment, education and R+D. Policy provides the most straightforward explanation
for changes in economic growth. A
typical example is the World Bank1, which purports to show that
‘strong adjusters’ (policy) in Sub-Saharan Africa during the 1980s
experienced increased rates of economic growth. The apparent link between policy and growth
has been enhanced by theoretical developments. The original Solow
growth model predicted that policy (investment) would only impact on the
level of growth not the long-run growth rate.
Endogenous growth models by contrast were motivated by the lack of
convergence to steady-state among developing countries and the inability of
traditional models to account for cross-country differences in income and
growth rates. Romer developed an
equilibrium model of technological change in which optimising agents drove
long-run growth through the accumulation of knowledge. The creation of knowledge by one firm has a
positive external effect on the production possibilities of other firms2. Adding to capital and labour a third input
(usually education) generates externalities.
Due to the externality, these models yield a sub-optimal equilibrium/
market solution and in turn generate a potential role for the state. Policy is shown to effect growth through
its impact on incentives to accumulate capital and knowledge and so generate
technological change. While the link between episodes of growth and
stagnation and changes in policy seems intuitively reasonable and is
supported by economic theory there is very little empirical evidence. Levine and Renelt
took a number of variables commonly used in econometric growth analysis and
ran them in thousands of regressions with different conditioning sets of
other variables – judging them robust if they remained significantly related
to growth. Their tests excluded
variables that are only correlated with another factor that has a causal
relationship with growth, i.e. those factors with an indirect impact on
growth. They found only investment was
robustly related to economic growth3. Even those factors many would accept as
self-evidently related to economic growth, fiscal policy, investment, education
and R+D have an ambiguous empirical relation to
economic growth as revealed by conventional cross-country regression
analysis. The relevant theory and
empirical results concerning these four policy variables are analysed in
turn. 2.1. Fiscal Policy and Growth. Theory linking fiscal policy to economic growth is very
clear. The link principally revolves
around how increases in tax rates lower the return to private investment and
hence permanently lowers the rate of investment. Barro measured government
intervention as the ratio of real government consumption less spending on
education and defence to real GDP. He
found a significant negative association between this variable averaged over
1970-85 and real growth 1960-854.
More generally there is no robust relation between growth and the
ratio of total government expenditure to GDP, government consumption
expenditure, government capital formation, or government educational
expenditure5. The
coefficient in Barro becomes insignificant when the
ratio of exports to GDP is included in the conditioning set6. There are good theoretical reasons for the
relation between government expenditure and growth to become more complex
once trade openness is considered.
Openness may increase the cost of government intervention by raising
the elasticity of taxed factors7.
Rodrik finds a positive correlation between
a country’s exposure to international trade and the size of its
government. A possible explanation he
suggests is that government plays a risk-reducing role in economies exposed
to a significant amount of external risk/ openness8. There are severe empirical problems with any attempt to
quantify the role of the state through regressing the rate of growth on the
level of government expenditure. Keyensian demand management and automatic stabilisers
imply that government expenditure will increase with poor economic
performance. This will generate a
spurious negative relation between the ‘size’ of government and economic
growth. Governments also influence the
economy in many ways that do not involve expenditure, such as regulation. Tax
exemptions and fiscal transfers may have identical effects but have different
implications for the measured size of government. The demand elasticity for government services
is typically greater than one (Wagner’s Law).
The level of government expenditure would then be determined
endogenously. 2.2. Investment and Economic Growth Investment was the one factor found robustly related to
economic growth9. The average investment rate is frequently used
as an independent variable in growth regressions, though there remain severe
theoretical problems in identifying causality, and the traditional use of
instruments as a ‘solution’ is problematic.
“So many variables could be used to explain growth that it is
difficult to find variables that are not only highly correlated with the
endogenous variable but can also be plausibly excluded from the regression.”10
There is some closer empirical work on this
question. There is a positive
correlation between investment in specifically machinery and equipment and
productivity growth. The relationship
holds for countries with 1960 levels of GDP per worker greater than 25% of
the US level, between 1960 and 1985 and the result is causal, robust, strong
and statistically significant11.
There is also a positive and significant relation between the ratio of
imported to domestically produced capital goods for a large cross-country
regression between 1960 and 198512. Ultimately such a causal link remains
ambiguous. Between 1950 and 1988 the
composition of investment in the OECD shifted sharply, the share of
investment in producer durables, from 3 or 4% to more than 7% of GDP in
France, Germany, the US and the UK and in Japan from 3.5% to 9%, growth
showed no upward trend. Others have
found growth induced subsequent capital formation in a large cross-section of
countries between 1965 and 198513. 2.3. Education and Economic Growth Intuitively education has an evident link with economic
growth, but again there is no clear empirical link. Pritchett finds a robust and negative correlation between higher
school enrolment and educational attainment and total factor productivity
growth in developing countries.
Between 1960 and 1985 educational capital grew faster in sub-Saharan
Africa and South Asia than in East Asia even though the latter region grew
more rapidly14. Bils and Klenow find only a weak relation between initial
schooling and subsequent economic growth, even allowing for the indirect
effects of schooling in permitting greater technology absorption. They find the relation either spurious, the
expected return and incentive to acquire education increases in an expanding
economy when the skilled wage is growing rapidly, or reflective of omitted
variables related both to initial schooling rates and subsequent economic
growth rates15. A particular problem for regression analysis is finding
a satisfactory measurement of human capital16. A large part of investment in education
takes the form of forgone earnings by students. In addition, explicit spending on education
takes place by the individual, family and state. Not all education expenditure is intended to
generate productive human capital (for example the teaching of philosophy verses
literacy). The typical proxy used in
many cross-country regression equations is the share of the working-age
population in secondary school. This
fails to measure the quality of education, and the learning-on-the-job that
takes place in the workforce. Good explanations of missing empirical support
are possible external effects and the endogenous relation between education
and economic growth. 2.4. R+D and Economic Growth Theory and intuition suggest there is a clear link
between R+D and economic growth. Again, this link has not been
satisfactorily uncovered by empirical analysis. Between 1950 and 1988 the total number of
scientists engaged in R+D in the US increased from
200,000 to over 1,000,000. A similar
pattern was evident in Germany, France and Japan. Measured by R+D
expenditure the results are similar.
Despite this extra R+D there has been no
permanent increase in growth in these countries17. 3. An Empirical Problem: Episodes of Growth and Stagnation in Least
Developed Countries This section shows that using long-run averages typical
of cross-country growth regressions hide an important empirical reality of
growth in contemporary developing countries. Growth averages over the
medium-term (25-30 years) conceal the periods of stagnation, growth spurts,
structural breaks, volatility and instability that actually characterise
growth in developing countries. 3.1. The Historical Experience of Developing Countries Theoretical and empirical research on growth has
focused on averages over the medium-term (25-30 years)18. A decade of ten-percent growth followed by
another of zero-percent drops into a Barro-type
regression with the same average as two decades of five-percent growth. This
problem has real implications for the analysis of growth in developing
countries. Brazil enjoyed rapid growth
between 1965 and 1980, and stagnated during the 1980s. A medium-term average doesn’t distinguish
between the average of 3.1% between 1960-92 and the importance of the
structural break. Per capita GDP in
Cote D’Ivoire increased by 3.1% p.a. between 1960 and 1980 and declined by an
average of 4.1% p.a. between 1980 and 1992.
Ignoring the structural break average growth was 0.22%, almost the
same as Senegal (0.18%) which stagnated throughout the whole period19. Growth averages over the medium-term (25-30 years)
conceal the periods of stagnation, growth spurts, structural breaks,
volatility and instability that actually characterise growth experiences in
developing countries. The overall
average is not a good summary
indicator of growth performance.
Countries show shifts in growth rates, often in clear episodes, such
as the slowdown in Latin America in the 1980s. GDP growth is not well
characterised by a single exponential trend.
For forty percent of least developed countries the R2 on
such a trend is less than 0.5, suggesting that shifts and fluctuations are
the dominant feature of GDP growth.
There are instead six distinct patterns of growth, before and after
statistically chosen structural breaks, steep hills, hills, plateaus,
mountains, plains and accelerations20. Growth is very unstable across time
periods. The correlation of per capita
growth between 1977-92 and 1960-76 across 135 countries is only 0.0821.
There are very striking instances of growth
accelerations and collapses among developing countries. There are 14 episodes of growth in Africa
between 1960 and 1996 including South Africa between 1960 and 1974 (5.1%),
Cote D-Ivoire 1960 to 1978 (9.5%), Gabon 1965 to
1976 (13.1%), and Namibia 1961 to 1979 (6.4%)22. Ten countries in Africa between 1967 and
1980 had growth of more than 6% p.a., including Gabon, Botswana, Congo,
Nigeria, Kenya and Cote D’Ivoire all which were outperforming both Malaysia
and Indonesia23. Hausmann et al conducted a very broad empirical test to
locate episodes of growth by finding the year that maximises the F-statistic
of a spline regression with a break at the relevant
year24. Countries can have
more than one acceleration. This
filter yields 83 growth accelerations, capturing most well-known episodes
such as China 1978, Argentina 1990, Mauritius 1971, Korea 1962, Indonesia
1967, Brazil 1967, Chile 1986, and Uganda 1989. The magnitude of accelerations is striking. Their definition is conditional on a growth
acceleration of at least 2% p.a.; the average acceleration though was 4.7%
p.a. There are many episodes with
acceleration of 7% or more such as Ghana 1965 (8.4%), Pakistan 1962 (7.1%),
and Argentina 1990 (9.2%). The
occurrence of an episode is quite common, of 110 countries in their sample
between 1957 and 1992 54.5% had at least one episode of growth and 20.9%
two. The occurrence is also common
across space: 21 episodes occurred in Asia, 18 in Africa, 17 in Latin
America, 12 in Europe and 10 in the Middle East and North Africa. 4. Cross-country Growth Regressions: Theoretical Problems Recent theorising on endogenous growth models is clear
that there should be a strong link between policy and growth. This section shows that any empirical link
between changes in public policy and changes in growth rates will be
difficult to isolate using traditional cross-country regression analysis for
theoretical not just empirical reasons.
These theoretical problems include complementarity
among policy variables, the relation between different theories of growth, hysteresis effects, and dynamics. 4.1. Complementarity among Policies Policy variables typically enter the right hand side of
regressions separately without diagnostic tests allowing for any but very
limited interaction among them. Theory
does suggest complementarity
is important. For example, investment
may be causally related to growth only in the presence of strong property
rights, reforms causally linked to growth only if considered credible or if
correctly sequenced. There is some
limited empirical support for the importance of complementarity
between policies. Mosley finds that complementarity between inflation, openness and the
government share to be minimal but when corrected for sequencing the
coefficient(s) increases and becomes significant25. Econometrics has coped with this problem in an ad-hoc
manner, splitting country samples by region or income level to look for
changes in the strength and direction of causal relations or including
occasional ad-hoc interaction effects between two variables. It would in theory be possible to add all
possible interaction effects by adding multiplicative relations in a
regression between all combinations of variables and adding a welter of dummy
variables for all possible structural breaks and geographical regions. The resulting loss of degrees of freedom
would then render the regression all but meaningless. 4.2. The Relation between Different Theories of Growth There are numerous cross-country econometric studies
finding some indicator of national policy to be linked to economic
growth. There is however no clear
consensus on what policy variables to include in cross-country regression
analysis. Economic theory rarely generates a complete listing of variables to
be held constant when trying to gauge the impact on the relation between the
dependent and independent variable.
Mauro for example adds measures for corruption and Knack and Keefer
likewise add proxies for trust to standard Barro-type
regressions26. There is no
means to compare the merits of these two approaches and the relationship
between these and other theories remains confusing. A causal relation between two variables
(e.g. trade and growth) does not imply the falsity of another (e.g. democracy
and growth). Levine and Renelt find “statistical relationships between long-run
average growth rates and almost every particular policy indicator considered
by the profession are fragile: small alterations in the ‘other’ explanatory
variables overturn past results”27. 4.3. Growth and Hysteresis Effects Hysteresis effects are likely
to exist in the process of economic growth.
Hysteresis implies that a temporary economic
shock can have a permanent impact on future growth. The implication being that growth is not a
linear process and regression analysis will have trouble capturing this
effect. There may be virtuous and
vicious circles at work in growth connected with threshold effects28. If a country has a critical initial mass of
human and physical capital, growth will be virtuous, capital accumulation
attracting yet more capital. Green revolution technology for example
depends on the availability of both seeds and fertilisers through access to
adequate credit (and hence collateral).
Households with enough collateral can invest in the necessary skills
and technology to get the virtuous circle going. The option is not open for poor households
without collateral. With an exogenous
shock there is the potential for hysteresis
effects. A disaster can wipe out the
liquid assets of a household and leave it in a poverty trap, unable
thereafter to invest in green revolution technology. Potential poverty traps make households and
an entire economy more vulnerable to shocks.
If a country were near the critical mass level of capital, then a
terms-of-trade shock that rendered part of that country’s capital stock
useless might shift that country from strong growth to decline. The same shock may have little effect on a
country far from the threshold. This
implies there may be no primary causal factor but an interlocking circular
process with feedback. Neat
econometric models with fixed coefficients will by definition then be
impossible to find. 4.5. Cross-Country Growth Regressions and Dynamics Theories of cyclical and adjustment dynamics of output
are not well developed within growth theories. Reliable data sets for many traditional
growth determinants (inflation, government expenditure, tariffs, inequality
etc) have typically run for twenty-five plus years. Averages over this sample length are too
short for history and too long to model macroeconomic policy changes and
short-run dynamics. In cross-sectional
regression analysis it is not clear whether variables affect long-term
growth, the steady-state, or both.
Some growth effects are contemporaneous (macroeconomic and cyclical
factors), some take several years (transitional dynamics due to changed
investment incentives), others even decades (incentives effecting the rate of
technical change). Some
right-hand-side variables may have output/ growth effects at all three
horizons - cyclical, transitional and steady-state. There is no reason to assume these are of
the same magnitude or even the same sign29. What little ad hoc empirical work has been
carried out finds regression parameters commonly are unstable over time. Knack and Keefer for example look at the
relation between social characteristics and growth and find that social
variables have different signs on growth before and after 198030. With such findings common attempted
explanations are notable only by their absence. 5. Universalism: A Problem of Neo-classical Economics? In order to run large cross-country regressions
researchers are tightly constrained to the assumption of universalism. Conventional growth analysis assumes growth
parameters are identical across countries.
Far from being a positivist statistical exercise, cross-country growth
regressions are bound to an underlying neo-classical assumption – that the growth
process is universal. Each individual
country in cross-section according to this view will provide evidence that
can be used to elucidate the one underlying universal economic relation. An increase in openness for example is
hypothesised to have the same effect on growth in all countries. There are a
small number of exceptions that for example allow the constant term to differ
across countries (controls for fixed effects) using panel data or on
occasional a dummy variable is added for regions and notable events. The over-riding assumption behind
cross-country growth regressions is a that of a universalist
growth process. Many studies explain Africa’s slower growth as a
function of different levels of explanatory variables31. They seek to explain growth as the result
of a common growth process
beginning from different levels of
the same explanatory variables.
Significant regional dummies remain common in much of this literature,
and especially so for Sub-Saharan Africa.
The usual assumption is that significant dummy variables are capturing
the influence of missing variables, which must then be unearthed. This has led researchers to propose ever
more variables in the hope that the dummy variable will be rendered
insignificant and growth will finally be ‘explained’32. The alternative
route is to relax the assumption that only the levels of explanatory
variables are different and explore the idea that the growth process in
Africa works differently. There are a
limited number of studies that suggest this latter notion may be true. The
implication being that cross-country growth regressions are an intrinsically
poor mechanism to analyse growth and each growth experience should be treated
as potentially unique i.e. as a case study. Block conducts a more flexible analysis and allows for
the slope coefficients to differ and finds openness in Sub-Saharan Africa has
a much stronger effect on growth and that growth is less responsive to fiscal
policy than his sample average33.
Brock and Durlauf find “the operation of
ethnic heterogeneity on growth is different in Africa, not just the levels of
ethnic heterogeneity. A comparison of
other regressor coefficients for Africa with those
of the rest of the world makes clear the growth observations for African countries
should not be treated as partially exchangeable with the growth rates of the
rest of the world.”34. Asiedu finds that FDI is less
responsive to openness in Africa than in other regions. For a given level of trade openness,
infrastructure and return on capital, Sub-Saharan Africa receives less FDI35. Mosley finds that inequality only has a
negative impact on growth in regions other than sub-Saharan Africa36.
5. Summary Cross-country growth regressions assume that economic growth operates according to universal laws across all economies through time and space. This is one of the key ideological foundations of neo-classical economics. There are only a few exceptions and the occasional dummy variable for regions and notable events. Discussion in this paper has demonstrated that there is evidence the growth process differs significantly between different regions and countries and over time. Finding fragility and heterogeneity of regression coefficients by region and country is only a beginning. Opening up the assumption of universalism to greater scrutiny leaves us asking why the growth process may differ. In doing so we would throw doubt on the neo-classical assumption of universalism and with it the ‘one-size-fits-all’ programmes of liberalisation emanating from the World Bank, IMF and other institutions. A case study approach to economic growth would be justified on the assumption that growth processes are not universal. The recent collection of country case studies together with an introduction drawing general lessons edited by Rodrik (2003) is a useful step in this direction37. Comparative and historically informed case studies allow researchers to question the assumption of universality rather than be forced to assume it true a priori. A heterodox or post-autistic economist should begin with case studies and only then proceed to cross-country growth regressions with all due caution. Notes 1. World
Bank ‘Adjustment in Africa: Reforms,
Results and the Road Ahead’, (Oxford University Press, New York 1994). 2. P.M.Romer, ‘Increasing Returns and Long-Run Growth’, Journal of Political Economy, 94:5
(1986) p1003. 3. R.Levine and D.Renelt, ‘A
Sensitivity of Cross-Country Growth Regressions’, Amercian Economic Review, 82:4
(1992), p942-963. 4. R.J.Barro, ‘Economic Growth in a Gross Section of
Countries’, Quarterly Journal of
Economics, 106, (1991), p407-43l. 5. Levine
and Renelt (1992). 6. Barro (1991) and Levine and Renelt
(1992:951). 7. J.Slemrod, ‘What do Cross-Country Studies Teach About
Government Involvement, Prosperity and Growth’, Brookings Papers on Economic Activity, 2 (1995), p405. 8. D.Rodrik, ‘Why Do More Open Economies Have Bigger
Governments?’, Journal of Political
Economy, 106:5 (1998), p997-1032. 9. Levine
and Renelt (1992). 10. J.Temple 1999:128) ‘The New Growth Evidence’, The Journal of Economic Literature,
38, (1999), p128. 11. J.B.De Long and L.H.Summers,
‘Equipment Investment and Economic Growth’, Quarterly Journal of Economics, 106, (1991), p445-502. 12. J-W.Lee, ‘Capital Goods Imports and Long-run Growth’, Journal of Development Economics, 48,
(1995), p91-110. 13. M.Blomstrom, R.E.Lipsey and M.Zejan, ‘Is Fixed Investment the Key to Economic
Growth’, Quarterly Journal of Economics,
111 (1996), p269-276. 14. L.Pritchett, ‘Where
has all the Education Gone?’, World Bank (1999), p3. 15. M.Bils
and P.J.Klenow, ‘Does Schooling Cause Growth?’, American Economic Review, 90:5 (2000),
p1160-1183. 16. N.G.Mankiw,
D.Romer and D.N.Weil, ‘A
Contribution to the Empirics of Economic Growth’, Quarterly Journal of Economics, 107 (1992), p418-9. 17. C.I.Jones, C.I, ‘R+D Based Models of
Economic Growth’, Journal of Political
Economy, 103:4 (1995), p759-784. 18. Beginning
with Barro (1991). 19. L.Pritchett, ‘Understanding Patterns of Economic Growth:
Searching for Hills among Plateaus, Mountains, and Plains’, World Bank Economic Review, 14:2
(2000), p221-50. 20. Pritchett
(2000). 21. W.Easterly and R.Levine
(2001:195) ‘It’s Not Factor Accumulation: Stylised Facts and Growth Models’, World Bank Economic Review, 15:2
(2001) p195, see also Temple (1999:116). 22. Defined
as a ten-year plus period in which the five-year moving average of annual GDP
growth exceeds 3.5%, J-C.Berthelemy and L.Soderling, ‘The Role of Capital Accumulation,
Adjustment and Structural Change for Economic Take-Off: Empirical Evidence
from African Growth Episodes’, World
Development, 29:2 (2001), p323-343. 23. T.Mkanawire, ‘Thinking about Developmental States in
Africa’, Cambridge Journal of Economics,
25 (2001) p2989-313). 24. R.Hausmann, L.Pritchett and D.Rodrik, ‘Growth
Accelerations’, JFK School of Govt, Harvard Universitry (2004). 25. Adding
a premium for example when reforms were conducted in the correct sequence,
deflation before devaluation, liberalisation of the current before capital
account etc, P.Mosley, ‘Globalisation, Economic
Policy and Convergence’, World Economy,
23:5 (2000) p613-634. 26. P.Mauro, ‘Corruption and Economic Growth’, Quarterly
Journal of Economics, 110 (1995) p681-712). S.Knack
and P.Keefer, ‘Does Social Capital Have an Economic
Payoff: A Cross-Country Investigation’, Quarterly
Journal of Economics, 112 (1997), p1251-1288. 27. Levine
and Renelt (1992:943). 28. W.Easterly, ‘The
Elusive Quest for Growth: Economists’ Adventures and Misadvantures
in the Tropics’, (MIT Press, Cambridge, 2001), Chapter 10. 29. Temple
(1999:124). 30. Knack
and Keefer (1997). 31. W.Easterly and R.Levine,
‘Africa’s Growth Tragedy: Policies and Ethnic Divisions’, Quarterly Journal of Economics, 112
(1997), J.D.Sachs and A.M.Warner,
‘Fundamental Sources of Long-Run Growth’, American
Economic Review, 87:2, (1997) p184-8 among many others. 32. A
good example of an exhaustive attempt to remove the significant dummy
variable is P.Collier and J.Gunning
‘Explaining African Economic Performance’, Journal of Economic Literature, 37:1 (1999), p64-111. 33. S.A.Block, ‘Does Africa Grow Differently?’, Journal of Development Economics, 65,
(2001), p443-467. 34. W.A.Brock and S.N.Durlauf,
‘Growth Empirics and Reality: What Have We Learned from a Decade of Empirical
Research on Growth?’, World Bank
Economic Review 15:2, (2001), p264. 35. E.Asiedu, ‘On the Determinants of Foreign Direct
Investment in Developing Countries: Is Africa Different?’, World Development, 30:1 (2002),
p107-119. 36. Mosley
(2000). 37. D.Rodrik, ‘In
Search of Prosperity: Analytic Narratives on Economic Growth’, (Princeton
University Press, Princeton, 2003).
___________________________ SUGGESTED CITATION: |