Adding to a growing list of reports on performance pay that have come to similar conclusions, an evaluation of New York's Schoolwide Performance Bonus Program by RAND Education pronounced the program a failure, at least in its in ability to improve student achievement. The news comes as no surprise to us, as we noted some big design flaws even when the program was in its planning stage.
Beyond the inherent structural problems of performance pay, including whether incentives are applied at the individual or school level, there is the issue of relying on standardized test scores to determine teacher effectiveness. This creates incentives to teach to the test and thereby boost test scores without actually having improved student learning.
Economists Gadi Barlevy and Derek Neal have another idea. They would have districts use a randomized student assessment that is separate from the one used to evaluate student performance. At the core of their "pay for percentile" scheme appearing in an NBER Working Paper, the test used to evaluate teacher performance would change on a yearly basis to prevent pernicious coaching practices.
The process begins with a comparison group being formed for each student at the onset of the school year. After a cumulative test is given at the end of the year, each student is given a percentile score based on performance relative to the comparison group that can be totaled by class. With all teachers receiving a class score, it is possible to award bonus pay to the top performers. The premise is that by randomizing the test used to allocate incentive pay, well-rounded teaching practices would emerge to improve learning outcomes and move teachers away from preparing students for an annual assessment.
Leaving aside the issue of how to motivate students to do well on such a test, their solution begs an important question: If the proposed tests push teachers to focus more on the content than on test prep, wouldn't these also be better tests for evaluating students?