Economic growth and economic development 508

Introduction to Modern Economic Growth
schooling are as important as private returns (which would correspond to very large
externalities).
Rauch estimated significant external returns, with the magnitude of the external
returns often exceeding the private returns. External returns of this magnitude
would imply that human capital differences could play a much more important
role as a proximate source of cross-country differences in income per capita than
implied by the computations in Chapter 3. However, Rauch’s regressions exploited
differences in average schooling levels across cities, which could reflect many factors
that also directly affect wages. For example, wages are much higher in New York
City than Ames, Iowa, but this is not only the result of the higher average education
of New Yorkers. A more convincing estimate of external returns necessitates a source
of exogenous variation in average schooling.
Acemoglu and Angrist (2000) exploited differences in average schooling levels
across states and cohorts resulting from changes in compulsory schooling and child
labor laws. These laws appear to have had a large effect on schooling, especially at
the high school margin. Exploiting changes in average schooling in state labor markets driven by these law changes, Acemoglu and Angrist estimate external returns
to schooling that are typically around 1 or 2 percent and statistically insignificant
(as compared to private returns of about 10%). These results suggest that there
are relatively small human capital externalities in local labor markets. This result
is confirmed by a study by Duflo (2004) using Indonesian data and by Ciccone and
Perri (2006). Moretti (2002) also estimates human capital externalities, and he finds
larger effects. This may be because he focuses on college graduation, but also partly
reflects the fact that the source of variation that he exploits, changes in age composition and the presence of land-grant colleges, may have other effects on average
earnings in area. Overall, the evidence appears to suggest that local human capital
externalities are not very large, and calibration exercises as those in Chapter 3 that
ignore these externalities are unlikely to lead to significant downward bias in the
contribution of human capital to cross-country income differences.
The qualification “local” in the above discussion has to be emphasized, however.
The estimates discussed above focus on local externalities originally emphasized by
Jacobs. Nevertheless, if a few very talented scientists and engineers, or other very

494