AEA RC T Registr y

The American Economic Association's registry for randomized controlled trials

In developing countries, educational systems are often designed to serve the elite. This can be true not only at the institutional level, but also within classrooms. Singular focus on competitive entrance examinations leads teachers—even in early grades—to focus on the best students at the expense of those perceived as having less potential. Performance pay schemes, as they are commonly designed, can exacerbate inequality in outcomes by strengthening incentives further for teachers to focus on students with the most potential. In this paper, we present the results from a randomized evaluation designed to test the impact of teacher performance pay (which is increasingly used to address weak incentives facing educators in developing countries) on not only the level, but also the distribution of gains in student achievement. Teachers across 216 schools were randomly assigned to one of three incentive groups or a control group. In the first incentive group (“levels”), teachers were rewarded based on levels of student achievement measured by student scores on year-end standardized exams. In the second group (“gains”), teachers were rewarded based on average gains in student achievement over the course of one school year. In the third incentive group (“pay-for-percentile”), teachers were offered an incentive scheme designed explicitly to allocate teacher effort equally across all students in the class, regardless of baseline achievement (based on the pay-for-percentile scheme proposed by Berlevy and Neal (AER, 2012)).

Another concern for policymakers in the design of teacher performance pay programs is whether the size of the incentive payout—or strength of the incentive—matters in a performance pay program. On the one hand, incentivizing teachers with potentially larger payouts may be more effective than incentivizing teachers with smaller payouts. On the other hand, incentivizing teachers with smaller payouts may be just as effective (and therefore more cost-effective). This would be the case, for example, if there were diminishing marginal returns to providing teachers with performance pay. By using a (4X2) crosscutting experimental design, our randomized experiment also tests whether the size of the payout (large payout versus small payout versus control group) impacts gains in student achievement.