across the sample: years teaching and previous years as principal.3 The mean teaching
experience calculated for the sample (fourteen years) is substantially higher than the
minimum mandated by the states (which averaged three years) which suggests the
measure is used by school districts as a hiring criterion.4
IV. Empirical Models of Skills Variation
Table 1 describes the heterogeneity over geography in the previous experience of
principals as well as their years teaching; table 2 illustrates regression specifications
determining the variation in skills. The skills measures used in all the empirical
specifications represent the skills level at the time the principal was hired. The spatial
variation in the two skills measures are calculated in table 1 by their sum of squares. The
unit of observation in table 1 is the individual principal, each of whom is employed by a
district that records at least four in the data.
The unit of observation for the regression models in table 2 is the school district.
The models are represented by (3) yi = α+ xiβ+ ziγ+ εi where yi is the standard
deviation in years teaching experience calculated across schools within each sampled
school district. The right-hand-side covariates are split into metropolitan area
characteristics, xi and those representing the school districts, zi.
The primary urbanization variables are total population and density of the
metropolitan area. The vector xi also includes the proportion of the adult population in
the metropolitan area with at least a master’s degree. The variable represents the share of
those in the local labor market with the minimum education qualification for principal.
3 Most of the skills variables compiled in the Schools and Staffing Survey are qualitative, indicating
whether the principal had any previous experience in specific occupations such as vice principal or school
counselor.
4 For information on state certification standards for principals, see J. Tryneski (1993). See appendix A for
summary statistics for the skills variables.