The role of human capital in economic growth revisited

Abstract

Benhabib and Spiegel (1994) examine the role of human capital in the development process empirically using a theory-driven specification rather than the standard production function approach. While they find evidence of a positive impact of human capital on income growth, their result is not robust to the inclusion of inequality as an additional covariate. Using an alternate dataset and different measures of inequality, we find robust support for the hypothesis that human capital matters even when we account for the adverse effect of income inequality on growth.

Acknowledgements

I am grateful to Thomas Osang and Daniel Millimet for helpful comments and suggestions. All remaining errors are solely mine.

Notes

¹ For a Cobb–Douglas production process with human capital, long-term growth can be expressed as: (log Y_T  − log Y ₀) = (log A_T  − log A ₀) + α(log K_T  − log K ₀) + β(log L_T  − log L ₀) + γ(log H_T  − log H ₀) + (log ϵ _T  − log ϵ₀).

² On the other hand, the traditional Kaldorian argument (Kaldor, 1956) suggests that income inequality promotes growth, because the rich have a higher propensity to save and invest than the poor.

³ In contrast, using five-year panels, Forbes (2000) finds that inequality has a positive significant effect on short-run growth.

⁴ Since it is well known that the Deininger and Squire (1998) data set represents an improvement in quality and coverage, the use of inequality indices from earlier sources are likely to make his results less reliable.

⁵ The decomposability properties of the Theil measure make it possible to fill the gaps in the Deininger and Squire data set, albeit in most case only for the limited span of the manufacturing economy. While the evolution of inequality in manufacturing earnings cannot be taken as per se indicating the larger movements of inequality in household incomes, including those outside the manufacturing sector, Conciecao and Galbraith (1998) argue on theoretical grounds that the two will rarely move in opposite directions.

⁶ Income, physical and human capital stocks, population and labour force data are taken from A new database on physical capital stock: sources, methodology and results, Rivista de Analisis Economico, 8, 37–59, by Vikram Nehru and Ashok Dhareshwar, 1993.

⁷ Income data for this project was obtained from the United Nations Industrial Development Organization (UNIDO, 1999).

⁸ Summary statistics are available upon request from the author.

⁹ In a standard growth regression using labour, physical and human capital, the latter always has a significant negative effect on growth. This finding is robust to the inclusion of initial income, the Gini coefficient, Theil's index and the covariates described in Section IV. These results are not reported in the article, but are available from the author upon request. See footnote 1.

¹⁰ BS did not report the regression results using the actual intended measure for human capital from their theoretical specification, due to lack of data. Their proxy for the measure of human capital was an average over the initial and final years only.

¹¹ See footnote 9.

¹² While the number of countries for which data on inequality is available shrinks substantively when BS add MID as a covariate in their regression, we are able to utilize a sizably larger sample even when inequality measures are included.

¹³ The correlation coefficient between Theil and POL is 0.42.