Enhancing the Efficacy of Teacher Incentives through Loss Aversion: A Field Experiment

Review of a paper by Fryer, Levitt, List and Sadoff (2012)

Improving the productivity of teachers has long been a priority in Public Policy. In recent years, there has been growing enthusiasm among policy makers for initiatives that tie teacher incentives to the achievement of their students. However, evidence as to the efficacy of such policies is mixed. Studies in developing countries have shown that linking pay to teacher performance helps reduce teacher absenteeism and dramatically improves students’ test scores. In contrast, the only two field experiments conducted in the US have shown small, if not negative, treatment effects.

The authors of this paper conducted a field experiment to test whether the psychological bias we call “loss aversion” can be exploited to design better incentives.

To be sure we’re all on the same page, let’s take a minute to define what loss aversion means. A perfectly rational person should feel the same amount of pain on losing say $5 as the joy felt on gaining $5. However, for a loss averse individual, the pain of losing is greater than the pleasure of gaining, even for the exact same amount. Studies have shown that losses may be more than twice as powerful as gains!

There is overwhelming laboratory evidence for loss aversion, but the paper in question is one of the first to demonstrate it in a field experiment. The study was conducted in Chicago Heights, Illinois during the school year 2010-2011. Chicago Heights is made up of primarily low-income minority students who struggle with low achievement rates.

Two methods of framing teacher incentives are compared:

1. The teacher gets a sum of money at the end of the academic year if his/her students perform well (gain frame; the one we’re most familiar with in this context)
2. The teacher is given the sum of money at the beginning of the year and it is withdrawn if the students do badly (loss frame)

At the beginning of the school year teachers were randomised into Treatment and Control groups. Within the treatment group, some were assigned the gain frame and others the loss frame. Further, within the “Loss” and “Gain” groups, the authors tested for heterogeneous effects for individual rewards compared to team rewards. Thus, participating teachers were randomly assigned to one of four treatment groups or a control group. Rewards were based on students’ end of the year performance on the ThinkLink Predictive Assessment, an otherwise low stakes standardised diagnostic assessment that is designed to be aligned with the high-stakes Illinois Standards Achievement Test (ISAT) taken by 3rd-8th graders in March and the Iowa Test of Basic Skills (ITBS).

They estimate intent-to-treat (ITT) effect by taking the difference between treatment and control group means. They find large impacts on both the ThinkLink as well as ISAT and ITBS scores, but only in the LOSS frame. There is no significant difference for student scores of teachers in the “Gain” treatment, in line with findings from previous studies done in the US. Also, no significant difference was found between individual and team incentives under either the Gain or the Loss treatment.

But can we really be sure that it is loss aversion that is driving these results? There could be 3 other possible explanations. These are listed below, along with the reasons why they are implausible in this case:

1. Attrition- Perhaps teachers found a way to discourage weaker students from taking the exams. The authors run a probit regression on all of their covariates, where the dependent variable is an indicator for missing any of the exams considered. They find no significant difference between the gain and loss groups. Furthermore, the ISAT scores were not directly incentivised and they found similar rates of attrition even there. So attrition is unlikely to explain the results.
2. Liquidity constraints- Teachers under the loss treatment have more money at the beginning of the year, which may enable them to spend more on resources for the classroom. However, surveys on the amount of spending reveal that this was not the case.
3. Cheating- Teachers may cheat and give their students high scores. However, they cannot manipulate the ISAT and ITBS scores, which mirror the results found on the ThinkLink tests.

Thus, by spending the same amount of money, but just changing the way the incentive was framed, they obtained dramatically better results. This suggests that there may be significant potential for exploiting loss aversion for designing Public Policy (as well as for maximising profits, of course)!