Abstract
Do political regimes have a significant effect on economic growth? This study builds on the new neoclassical growth model to identify economic determinants of growth, and explicitly tests for the influence of political variables on economic performance for the 1990s. The results suggest that democracies and bureaucracies significantly outperform autocracies. Economic growth is also promoted by increased protection of property rights, and higher investment in education. Moreover, technology has become a requirement for efficient production, and hence, is crucial in promoting growth. Countries can therefore increase the level of economic growth by increasing the levels of education and technology in the economy, and establishing codified laws to foster property rights.
Acknowledgements
We would like to thank Brenda E. Ellis and Richard Saunders for their valuable assistance in locating pertinent literature and editors of this journal for their valuable suggestions. The usual caveat applies.
Notes
1 Teles (Citation2005) provides a nice discussion of the applicability of these models to industrialized vs. developing countries.
2 A good introduction to the field of institutional economics that this area of research builds on is presented in North (Citation1990).
3 Kurzman et al . (Citation2002) summarizes the varying results while compiling 47 studies. The 19 studies showed a positive relationship between democracy and growth, six concluded that democracy and growth were negatively related, while 10 concluded that no relationship existed between democracy and growth. Seven studies found positive but insignificant results, two found negative and insignificant relationships, 2 found both positive and negative results, while one found an inverted U effect.
4 Sloan and Tedin's (Citation1987) study of 20 Latin American countries over the period 1960 to 1979 concluded that bureaucratic dictatorships fared better than democracies, but traditional autocracies fared worse.
5 Ramsey's RESET (Regression specification error test) procedure (1969), employed on the above specification(s), indicate that the proposed specification(s) do not suffer from possible omitted variable bias. Please see the bottom of for the numerical details of this test procedure.
6 This technique, by subjecting the observations to two transformations, one designed to remove autocorrelation and the other to remove heteroskedasticity, comes up with a disturbance term (ε
it
) that is asymptotically nonautoregressive and homoskedastic. To find consistent estimates, OLS is applied to obtain the regression residuals and then these are used to perform transformations so that the error term is asymptotically nonautoregressive and homoskedastic [for details see Kmenta (Citation1997), pp. 618–622]. The particular characteristics of this model are as follows: E() =
(heteroskedasticity); E(ϵ
it
ϵ
jt
) = 0 [(i ≠ j–cross-sectional independence)], where ε
it
= ρ
i
ϵ
i, t−1 + uit
(as far as autocorrelation is concerned ‘ρ
i
’ is assumed to be different across cross-sectional units and uit
is the classical error) uit
∼ N(0,
), ε
it
∼ N(0, [
/1 − ρ2]), and E(ε
i
,
t−1
ujt
) = 0 for all i, j.
7 The effect of technology is captured by usage of computers per 1000 people. However, one may argue that this variable may also pick up the effect of level of development. Other proxies such as R and D spending, number of science graduates etc., could not be employed due to lack of available data. The list of the countries used in the empirical analysis is presented in the Appendix.
8 GDP per capita is the total gross domestic product divided by midyear population.
9 Poor and developing countries are defined as countries with per capita GDP <$3000.