Our data performs better than Barro and Lee (2001) series in the standard crosscountry
growth regressions estimated in the earlier literature. Indeed when our data is
entered into such regressions we find that it is significant, as opposed to the Barro and
Lee (2001) series. This holds even for the kind of regressions estimated by Krueger
and Lindahl (2001), who failed to find significance for the Barro and Lee (2001)
series when physical capital is included among the regressors. Moreover, our series
are also significant in more sophisticated panel data regressions that attempt to
account for endogeneity or problems identifying the effect of schooling. Indeed when
we estimate an augmented Solow production function that embeds the Mincerian
approach to human capital we find that our series are highly significant. Moreover,
the estimated long-term effect of schooling is close to the typical micro Mincerian
return. This result suggests the absence of externalities to education, which is consistent
with Acemoglu and Angrist (2001) and Ciccone and Peri (2005) among others.
But this should be a matter of further empirical investigation. The main contribution
of this paper is to make a new and reliable data set on schooling available for
a large group of countries that may prove useful for further research into human
capital