Cognitive Ability and Technology Diffusion: An Empirical Test

Recent economic research has shown that cognitive ability scores are robustly associated with good national economic performance. How much of this is due to high-ability countries doing a better job of absorbing total factor productivity from the world's technology leader? Following Benhabib and Spiegel (JME 1994, Handbook of Economic Growth 2005) who estimated the Nelson-Phelps technology diffusion model, I use the database of IQ tests assembled by Lynn and Vanhanen (2002, 2006) and find a robust relationship between national average IQ and the conditional rate of total factor productivity growth over the 1960-1995 period. In a horse race between IQ and education, national average IQ performs better as a predictor of TFP growth. The results hold even if only pre-1970 IQ scores are used. ______________________________ * Garett Jones, Department of Economics and Center for Study of Public Choice, George Mason University. E-mail: [email protected]. Recent economic research, including Hanushek and Woessmann (NBER working paper, 2007), Jones and Schneider (Journal of Economic Growth, 2006), Weede and Kampf (2002) and Ram (Economics Letters, 2006) has shown that cognitive ability scores are robustly associated with good economic performance. They almost invariably find that cognitive ability scores have vastly more predictive power than traditional schooling measures (the sole exception is Volken (2003)). The question of whether intelligence tests and other standardized tests are robust predictors of growth has apparently been settled. The present paper turns to the question of why this is so. Herein, I focus on the following question: How much of the cognitive ability/economic growth relationship is due to high-ability countries being better at absorbing ideas from the world's technology leader? Benhabib and Spiegel (JME 1994, Handbook of Economic Growth 2005) estimated the technology diffusion model of Barro and Sala-i-Martin (Journal of Economic Growth, 1997); Benhabib and Spiegel used years of education as their measure of human capital, and found a modestly robust relationship that weakened considerably when additional control variables were added. Instead, I use the database of IQ tests assembled by Lynn and Vanhanen (IQ and Global Inequality, 2006), and invariably find a robust relationship between national average IQ and the conditional rate of total factor productivity growth over the 19601995 period. In a horse race between IQ and education, national average IQ easily wins under all specifications. The results also hold even if only pre-1970 IQ scores are used. One reason to use IQ tests rather than the international math and science test scores employed by Hanushek and Kimko (AER 2000) and Barro and Lee (AER 1996) is that IQ tests are much more widely available. For instance, the former have data from 31 countries, and the latter from 23. Further, the psychological profession has worked to make such test scores comparable across time and space–indeed, a substantial number of the Lynn and Vanhanen observations come from country-wide “standardization samples” that are created when an IQ test is revised. As Jones and Schneider (2007) demonstrate, the positive relationship between IQ and year 2000 output per worker holds whether one uses verbal or visual IQ tests, whether one uses “culture reduced” or traditional IQ tests, and whether one uses pre -1980, pre-1970, or pre-1960 IQ tests. Arthur Jensen's 1998 book The g Factor, provides the best overview of the IQ literature; Ian Deary’s Intelligence: A Very Short Introduction (2001) provides a more accessible overview written by another prominent intelligence researcher. The psychology literature demonstrates that while environmental effects can explain some of the IQ gap across countries, there is also some evidence in support of genetic sources for this gap, as Jensen as well as Lynn and Vanhanen (2002) discuss. (N.B. On average, East Asian countries routinely outperform European countries on these tests designed by Europeans and their descendants). Accordingly, disentangling this nature-nurture question will likely be of increasing importance to growth economists in the future. Greg Clark’s (2007) book A Farewell to Alms, which builds on the theoretical model of Galor and Moav (2002), is one recent attempt at disentangling this important question. In the Galor/Moav framework, some combination of the agricultural revolution and private property begins a process favoring human-capital-investment genes; indeed, one section of their paper is entitled “Evolution of Intelligence and Economic Growth.” Since the agricultural revolution has been more thorough in some regions of the world than others (Galor and Moav, 2007), this could provide one possible mechanism explaining some differences in cognitive ability across countries. At the same time, environmental explanations, including cultural differences across time and space as well as health and nutrition differences, clearly play an economically significant role, and may be the sole explanation for IQ differences across countries. I have summarized this literature elsewhere (Jones and Schneider, 2006), so I will only note here that the long-term rise in IQ’s across countries known as the Flynn Effect (2007) provides some reason to believe that cognitive skills respond to changes in the environment. In all, the results presented below support the hypothesis that abstract tests of reasoning ability given to a random sample of the population can tell us more about an economy's economic potential than measuring years of schooling. Where these differences in reasoning ability come from is a matter of ongoing research in a variety of disciplines; for economists, the main lesson is that these differences appear to be quantitatively significant correlates of TFP. In the conclusion, I point to some literature that might begin to provide a micro-level explanation for this macroeconomic result.