Panel Paper: Principal Value-Added In Wisconsin: Where Do Effective Principals Go?

Saturday, November 10, 2012 : 9:10 AM
McKeldon (Sheraton Baltimore City Center Hotel)

*Names in bold indicate Presenter

Jeffery Dean, Univ. of Arkansas

Using grade-level achievement and individual-level staffing data from Wisconsin, I develop a value-added model to estimate principal effectiveness over the years 1999-2011. Consistent with other value added models, I standardize test scores in math and reading across years and grades. I examine school and principal effects after controlling for time-varying school demographics as well as year and grade shocks. I further control for shocks or changes in the teaching workforce in a school, considered as high turnover or decreases in average experience.

School value added is a poor proxy for principal value added, so in order to distinguish the two I observe changes in school value added under different principals. Thus I only observe principal effectiveness for principals who switch schools or who are in a school where a switching principal has taught. I find that about one-quarter of Wisconsin principals are observed in networks of switchers of at least ten principals, with one network having 185 principals in and around Milwaukee. Value-added on principals can be calculated only relative to other principals in the same switching group; the individual with mean effectiveness in each group is thus assigned a null effect, and all other principal effects estimated relative to that individual. Even though estimates of principal effectiveness cannot be linked across switching networks, characteristics of highly effective and less effective principals can be aggregated and compared across these networks. Thus for all principals across switching networks, one can compare the average characteristics of highly effective principals to ineffective principals.

Having estimated principal value added for one-quarter of Wisconsin principals, I compare highly effective and significantly less effective principals, as well as principals indistinguishable from the mean individual in each group. I compare them on individual characteristics, such as sex, age, race, education, experience, and pay. I find no significant differences on any of these characteristics between highly effective and ineffective principals. I then examine the characteristics of schools where highly effective and ineffective principals are found: race and ethnicity, poverty levels, size, and achievement levels. I find that effective principals are more likely to be found in low-poverty schools.

As a final analysis, I examine the switching patterns of principals as a function of their effectiveness. Since switching principals are observed in the dataset, it is worth examining whether effective principals are more likely to switch into higher-paying jobs, into schools with lower poverty levels, or into schools with more experienced teachers.