Theodore R. Breton* June 18, 2011
Abstract The marginal product of human capital in Mankiw, Romer, and Weil’s  augmented Solow model measures the direct and two external effects of human capital created from schooling on national income. If thismodel is valid, its estimates of the share of this marginal product accruing to workers should be consistent with estimates of the marginal return on investment in schooling in workers’ earnings’ studies. This paper uses a new set of data for the net human capital stock to show that in 1990 the macro and micro rates are consistent across 36 countries.
JEL Codes: E13, I21, O11, O15, O41
KeyWords: Human Capital, Education, Schooling, Neoclassical Model, Economic Growth
*Thanks to Steven Davis, George Psacharopoulos, Andrew Breton, Charles Mann, and numerous anonymous referees for comments on earlier drafts of this paper.
When Mankiw Romer, and Weil  added human capital from schooling to the Solow model, it solved several empirical problems. This addition enabled the model toexplain the differences in income across countries, to provide an accurate estimate of the share of national income accruing to physical capital, and to provide a more accurate representation of the rate of income convergence in response to changes in rates of capital investment. But it also revealed a statistical relationship that many found problematic – a very large effect of human capital onnational income. In Mankiw, Romer, and Weil’s version of the Solow model (hereafter denoted the MRW model), productivity (A) varies over time, but country-specific differences affecting productivity not related to physical or human capital levels are assumed to be random:1 (1) (Y/L)it = (K/L)it α (H/L)it β (At ) 1-α-β
MRW’s empirical results, since replicated by other researchers2, indicate thatdifferences in physical capital (K) and human capital (H) can explain almost all of the variation in national income (Y) across countries, leaving little or no variation to be explained directly by other factors. In this model the levels of physical capital and human capital are the proximate determinants of national income. Other national characteristics, including institutions, policies, andculture, affect income through their effect on these capital factors. 3 Critics have argued that the MRW model is mis-specified because it does not include any of these national characteristics, but the evidence supporting the inclusion of other variables is weak. Levine and Renelt  examined the relationship between over 50 variables and
For example, hours worked per year vary acrosscountries but are not included as a control variable.
Cohen and Soto  obtain similar estimates of the model within countries between 1960 and 1990. Breton  obtains similar estimates across countries for the period 1990-2000.
Mauro  presented evidence that a measure of a country’s bureaucratic efficiency, which has since been interpreted to measure either a country’sinstitutions or its policies, affects its level of physical capital.
growth and found that over the 1960-89 period only investment in physical capital is robustly correlated with growth. Ciccone and Jarocinski  have investigated whether 67 variables are correlated with growth and found that over the 1960-96 period only education variables are robustly linked with growth. These twostudies and the capability of the MRW model to completely explain cross-country differences in income (R2 = .95 in the OLS and 2SLS analysis presented here) challenge the claim that relevant variables are omitted from the MRW model. Klenow and Rodriguez-Clare  and Hall and Jones  have contributed to the perception that variables are omitted from the model. They presented analyses that...