Economic growth and economic development 507

Introduction to Modern Economic Growth
the importance of human capital externalities, and suggested that the concentration
of economic activity in cities is partly a result of these externalities and also acts
as an engine of economic growth because it facilitates the exchange of ideas among
workers and entrepreneurs. In the growth literature, a number of well-known papers,
including Robert Lucas’ (1988) paper and Azariadis and Drazen (1990), suggest
that such technological externalities are important and play a major role in the
process of economic growth. Human capital externalities are interesting in their
own right, since if such external effects are present, the competitive price system
may be inefficient (since it will fail to internalize these externalities, particularly
if they take place across firm boundaries). Human capital externalities are also
important for our understanding of the sources of income differences across countries.
Our discussion of the contribution of physical and human capital to cross-country
income differences in Chapter 3 showed that differences in human capital are unlikely
to account for a large fraction of cross-country income differences, unless external
effects are important.
At this point, it is therefore useful to briefly review the empirical evidence on the
extent of human capital externalities. Early work in the area, in particular, the paper
by James Rauch (1993) tried to measure the extent of human capital externalities by
estimating quasi-Mincerian wage regressions, with the major difference that average
human capital of workers in the local labor market is also included on the right-hand
side. More specifically, Rauch estimated models of the following form:
0

ln Wj,m = Xj,m β + γ p Sj,m + γ e Sm ,
where Xj,m is a vector of controls, Sj,m is the years of schooling of individual j
living/working in labor market m, and Sm is the average years of schooling of workers
in labor market m. Without this last term, this equation would be similar to
the standard Mincerian wage regressions discussed above, and we would expect an
estimate of the private return to schooling γ p between 6 and 10%. When the average
years of schooling, Sm , is also included in the regression, its coefficient γ e measures

the external return to schooling in the same units. For example, if γ e is estimated
to be of the same magnitude as γ p , we would conclude that external returns to
493