The effects of performance pay for teachers: An analysis of Arizona's Career Ladder program

This dissertation considers the theoretical and empirical effects of providing performance pay for teachers. Two facets of the performance of Arizona's Career Ladder program are analyzed empirically. The dissertation examines the program's impact on student learning and on teacher retention.; To test the program's impact on student learning, grade level percentile scores on reading and mathematics tests for the years 1997–98, 1998–99, and 1999–2000, along with the annual changes in these levels, are used as dependent variable. These data are used to test four model specifications. These progressively restrict the means through which participation in the Career Ladder program can affect student learning. Propensity score matching is employed to address the problem of selection bias. This arises from the non-random assignment of school districts' program participation status. Matching is done at both the district level and the grade level. Regression results suggest that Career Ladder participation has a positive and significant impact on student learning. Improvements in student learning are motivated both by the additional wages provided by the program and by its incentive structure.; This dissertation also analyzes the effect of merit-pay on teacher retention. Two dependent variables—Median Teacher Experience and Teachers with Less Than Four Years Experience—are used to evaluate the impact Career Ladder participation on teacher retention. Experience levels provide an indirect means of assessing retention in the absence of data on teacher quits/fires. To gauge the importance of program participation, four specifications of the model are estimated. The first two specifications include variables to control for the wage impact of the program. This isolates the effect of the program's incentive component. In the second two specifications wage controls are dropped, allowing the Career Ladder variable to pick up all changes associated with the program. Propensity score matching is used to address the selection bias problem. The technique identifies a control group that is insignificantly different from the participant group. Regression analysis is done using both the set of matched districts and complete set of non-participating districts as control groups.; Results of these tests are mixed. Participation has a positive and significant impact on Median Teacher Experience in three of the models estimated using the matched control group. The insignificant result suggests that it is the program's additional wages, not its incentive related behaviors, that affect teachers' quit/stay decisions.