III. Sample and Skills Measure
The primary data source is the 1999-2000 Schools and Staffing Survey produced
by the National Center for Education Statistics. The survey collected personal
information for a sample of principals and the schools they administered. The survey
data also records the geographic location of the schools, allowing for identification of the
corresponding local school district. The constructed sample in this study includes
principals with various tenure lengths and from across grade school levels, including high
school. The sample is limited, however, to those principals hired during the 1990’s; this
was done to reduce problems arising from possible changing hiring criteria over time for
school administrators. The variation in principals’ skills within school districts could be
calculated only for those districts with multiple observations. The sample includes only
those school districts with at least four observed principals.2 This restriction eliminated
all but a handful of non-metropolitan school districts; the final sample of one hundred and
ninety-four school districts is limited to metropolitan areas.
The constructed skills measures should reflect the standards school districts use to
hire principals. Years of schooling, a natural selection criterion, suffers as a skills
indicator because the measure does not appreciably vary across the sample. Most states
mandate a master’s degree as the minimum required education for principal; eighty-seven
percent of the sampled principals had either a master’s or a professional education degree
beyond the master’s but below the level of the doctorate. The survey data includes two
continuous measures of experience in grade school education that vary substantially
2 Appendix B illustrates the distribution of the number of principals sampled across school districts in the
data used in the regression models.