Matthew G. Springer, Dale Ballou, Laura Hamilton, Vi-Nhuan Le, J.R. Lockwood, Daniel F. McCaffrey, Matthew Pepper, and Brian M. Stecher, "Teacher Pay for Performance: Experimental Evidence from the Project on Incentives in Teaching" (Nashville, TN: National Center on Performance Incentives at Vanderbilt University, September 2010).
This much-anticipated study made waves even before it was released. That's because it presents some of the most methodologically sound evidence available to date on the effectiveness (or lack thereof) of monetary incentives for teachers to produce better student outcomes. But its robust design examined only one narrow question: Is offering teachers a reasonable chance to earn up to $15,000 extra dollars enough to significantly raise value-added student achievement? The Vanderbilt research team divided a group of about 300 volunteer middle school math teachers in the Metropolitan Nashville Public Schools (MNPS) into a control group, who received only a small stipend for participating, and a treatment group, who both received the stipend and were eligible for $5,000, $10,000, or $15,000 bonuses for raising the valued-added scores of their math students. For three years (2006-07 through 2008-09), analysts compared each participating teacher's students' value-added gains to state wide average gains, and then averaged those "benchmarked" scores for the teacher's class. To earn a bonus, the gains had to be at least in the 80th percentile of gains for the district, based on distributions of gains from the two years prior to POINT (2004-06). The findings were disappointing: Treatment teachers' students made no more gains than those in the control group, though district-wide scores rose over the same period, ostensibly because of looming state sanctions for poor performance. The only exception was in fifth grade, where treatment teachers' pupil scores did improve, but the boost did not carry through to sixth grade. Participating teachers (in both groups) were also queried on their impressions of the experiment. Two findings are notable: Not only did treatment teachers indicate that they did not change their behavior over the course of the intervention, neither group "bought" the idea that monetary incentives were a legitimate reason to do so. After all, that would presume that MNPS teachers were simply not working hard enough before more cash was dangled before their noses. Not examined by this study, and probably a more important question, is whether merit pay plans recruit different kinds of teachers to the classroom in the first place. On that key issue, the jury is still out.
