As a final way to understand whether schooling or institutional outcomes come first, we
follow Barro (1997, 1999) and look at timing. If institutions come first, then lagged values of
political variables should predict improvements in education. If education is the critical input, then
lagged values of education should predict improvement in institutional outcomes. In the top Panel
of Table 12, we regress, using five-year intervals, growth in years of schooling on country fixed
effects, initial schooling, and initial measures of political institutions used in the growth literature
as well as the initial level of GDP per capita. The data show some mean reversion in schooling
(perhaps due to measurement error), a large and positive effect of initial level of income on the
growth of education, and no effect of initial political institutions, no matter how measured, on the
growth of human capital.
In the bottom panel, we look at the changes in political institutions over five-year intervals
as a function of country fixed effects, initial schooling, initial level of economic development, and
initial levels of these political institutions themselves. The results are striking. Initial levels of
schooling are a strong predictor of improving institutional outcomes over the next five years using
3 out of 4 measures, including executive constraints. Initial per capita income has no predictive
power. And, as we saw before, there is a lot of mean reversion in these measures of institutions.
As before, we do not want to take these results as dispositive. However, they are strikingly
consistent with the Lipset view that high human capital leads to institutional improvement, even
over a relatively short horizon of 5 years. Moreover, like many of the earlier findings we
presented, the results are inconsistent with the view that high assessments of political institutions
predict subsequent improvement in the years of schooling.