Economic freedom and development: new calculations and interpretations.

• Author: Weede, Erich

For some time there has been a debate about the effect of economic freedom on economic growth and development (Beach and Davis 1999: 10; de Haan and Siermann 1998; de Haan and Sturm 2000; Edwards 1998; Goldsmith 1997; Gwartney, Lawson, and Block 1996: 109; Knack and Keefer 1995; Pitlik 2002; Scully 1992; Torstensson 1994; Weede and Kampf 2002). Although there is wide agreement about the stylized fact that economically free societies are richer than other societies, there is less agreement about the impact of economic freedom on growth rates. Some writers contend that the level of economic freedom affects growth, whereas others, in particular de Haan and his associates, dispute the robustness of this claim and find only a relationship between improvements in economic freedom and growth.

The most recently published research on the effects of economic freedom on growth (Gwartney and Lawson 2004; Gwartney, Holcombe, and Lawson 2006) reaffirms that there are strong and beneficial effects of the level of economic freedom and of its improvement on growth rates. Looking at the published literature as well as at the work in progress by one of my doctoral students (Liu 2007), my impression is that there are two ways to strengthen the effects of the level of economic freedom on growth: first, choose a longer rather than shorter period of growth observation; second, and more important, use an average measure of the level of economic freedom rather than a single time point measure of economic freedom that refers only to the first year of growth observation. If one compares, say, de Haan and Sturm's (2000) study with Gwartney, Holcombe, and Lawson's (2006), then one finds that the former study uses a somewhat shorter period of growth observation, but both of them use the level of economic freedom at the beginning of the growth period to be explained. (1) Whereas de Haan and Sturm (2000) find no significant and robust effect of the level of economic freedom on growth, Gwartney, Holcombe, and Lawson (2006) arrive at the opposite conclusion: The level of economic freedom does promote growth. (2)

The purpose of this article is to discover whether one should believe in the results reported by Gwartney, Holcombe, and Lawson (2006). My approach is straightforward and simple. Neither extreme bounds analysis, nor Bayesian averaging shall be applied. But, of course, a study of robustness requires that one should not follow the example of Gwartney, Holcombe, and Lawson (2006) in every respect. I work with a research design that is similar to theirs (and inspired by it), but I do change some of their procedures. For the purposes of a robustness check, one does not necessarily need to claim superiority of one's own design. It is sufficient to claim that one's design is about as defensible or reasonable as the other one. Robust findings should be supportable by a variety of approaches.

Research Design

For a start, the period of growth observation, 1980 to 2000, is identical to Gwartney, Holcombe, and Lawson's (2006). But I expand the data set from 94 to 102 cases which, of course, is related to a different choice of control variables or other presumed determinants of growth. The sample expansion is not great, but better than nothing. Since all of us have to rely on accidental instead of random samples, (3) the effects of sample extensions tend to be unpredictable. Whereas Gwartney, Holcombe, and Lawson (2006) rely on the economic freedom ratings in 1980, I prefer to average ratings from Gwartney and Lawson (2005) for the 1980 to 1995 period. The later the measure of economic freedom within the period of growth observation, the less likely it is to affect the growth rate. (4) Whereas Gwartney, Holcombe, and Lawson (2006) generate two change in economic freedom variables, which separately refer to the 1980s and 1990s, I rely on a single change or improvement in economic freedom variable.

Like them, I use the level of economic development in 1980 as a control variable, but from a different source (Bhalla 2002). I also apply the control variables tropical location and coastal population from the same source. According to Sachs (2005), geography should matter more than institutions or policies. By contrast to Gwartney, Holcombe, and Lawson (2006), I neglect the impact of investment and replace their growth measure of human capital formation by a level measure of it. In my view (Weede and Kampf 2002; Weede 2004), all standard measures of human capital suffer from being based on the input to human capital formation, such as years of schooling or some related measure. Frequently, schooling input-based measures of human capital do not significantly affect growth rates (e.g., DeLong and Summers 1991; Hegre, Gissinger, and Gleditsch 2003; Pitlik 2002; Plumper and Martin 2003). Inspired by Pritchett (2006), one could also point to the difficulty of reconciling the divergence of growth rates between many Asian or rich countries on the one hand and many non-Asian developing economies on the other hand with globally converging levels in schooling. The World Bank (2005: 68) even admitted that "education is not translating into human capital and that the rise in per worker schooling explains only a small part of the growth in output per worker."

By contrast, the intelligence quotient (IQ) always does consistently and robustly improve economic growth rates (Lynn and Vanhanen 2002; Weede and Kampf 2002; Weede 2004; Jones and Schneider 2006). Moreover, it always outperforms standard measures of human capital by a wide margin. Since this article neither necessitates a specific assumption about the genetic and environmental components of intelligence, nor sheds light on this issue, one should regard IQs as scores on an achievement test. Although they do not necessarily say much about cognitive potentials, average IQs assess the current level of human capital availability within nations.

After these changes in the research design, one gets the results of column 1 in Table 1 where the economic growth rate (5) from 1980 to 2000 is regressed on the level of economic development to assess the opportunities of backwardness or the catch-up effect, on the national IQ to estimate the human capital effect, on coastal population and tropical location to estimate the impact of geographic advantages or disadvantages, and on average economic freedom (1980 to 1995) and the change or improvement of it (between 1980 and 2000). All of the coefficients are significant at least at the 10 percent level in two-tailed tests, which corresponds to the 5 percent level in one-tailed tests. The control variables perform as expected. The higher the level of economic development, the lower the economic growth rate in the following two decades is. The higher the IQ, the faster the economy grows. As can be seen from the standardized regression coefficients, these two effects are much stronger than the other effects. Tropical location does some harm. An ice-free coast and much of the population close to it helps. The main concern of this study, however, is economic freedom or capitalism. The average level of economic freedom is the third strongest determinant of economic growth, doing better than the two geographical variables. The economic freedom effect is significant at the three per thousand level. By contrast, the effects of improvements in economic freedom are much weaker. Only in...