III adds state fixed effects. The log-log specification for each model generates point
elasticity estimates.
The empirical relationship between the variation in skills and urbanization
characteristics is the test for the matching efficiency of urban areas. The results indicate
metropolitan area population significantly determines the variation in teaching experience
among principals, and the direction of its estimated effect is intuitive. Models I and II
indicate that as urban population increases, principals become more homogeneous within
school districts. This is consistent with the hypothesis that larger markets increase the
chance an employer will find closer labor matches (Wheeler, 2001).
In models I and II, a ten percent increase in population decreases predicted
variation in skills by 1.1% to 1.3%. If the actual size distribution of urban areas is
considered, the evidence suggests the urbanization effect is substantial. Table 3
illustrates the predicted variation in teaching experience calculated from model I for
various urban sizes. A metropolitan area at the 20th percentile in population has an
expected skills variation approximately 32% higher than that for an area at the 80th
percentile. The efficiency gains with urban size are more pronounced for smaller
metropolitan areas. Moving from the 5th percentile to the median population decreases
predicted skills variation by 30% - a difference substantially larger than the difference in
predicted skills variation between the median and the 95th percentile.7 Urban size may
largely account for differences over geography in the matching efficiency for school
principals.
7 Models in which urban population was entered quadratically performed no better than the non-quadratic
specifications.