Can conditional incentives for teachers improve student learning outcomes?
Hear from two experts on the debate surrounding performance-based pay for teachers, and how to design these policies for the best results
Crédit photo : Stanford CEPA Crédit photo : Elena Gormley Photography
Prashant Loyalka Felipe Barerra-Osorio
Around the world, education planners are wondering how to get teachers to put more effort into their teaching. Whether this means doing a better job in preparing for their classes and helping struggling students, or at a more basic level simply ensuring that teachers actually show up each day to teach, there is an interest in the question of how incentives could shape teacher behaviour.
Yet the proposal of paying teachers according to the quality of their work or the performance of their students often provokes great controversy. To learn more about the issues at stake, the IIEP Learning Portal’s Catherine Honeyman caught up with two experts on this topic: Felipe Barerra-Osorio, Associate Professor of Education and Economics at the Harvard Graduate School of Education; and Prashant Loyalka, Assistant Professor at the Graduate School of Education and a Center Research Fellow in the Institute for International Studies at Stanford University.
Following is a transcript of the discussion, or listen to the audio recording below:
Felipe Barerra-Osorio: The basic issue with this idea is that education systems in general have very weak incentive structures. So for instance, teachers have automatic promotion and that determines the tier of the salary based on years of service. And nothing attaches payment to any measure of performance. So that is the issue about teacher incentives. The idea of teacher incentives is: can we trigger a response from teachers so there can be better and more effort from the teachers?
In that regard, for me the main debate has to do with the problem of the right design of incentives that trigger the best response. We know that incentives are powerful instruments to change behaviour, but the problem is that some of the time we have incentives, and the behaviour that we elicit from teachers is not the one that we want. One case is, for instance, dishonest behaviour, another case is teach-to-the-test. So in my mind, the main current debate is: what are the right designs for incentives.
Prashant Loyalka: Yes, absolutely. I think fairness is actually a very important issue when thinking about the design of performance pay. If teachers don’t think that performance pay schemes are fair, they won’t put forth the effort. I can talk about my research, which basically tests the impact of a new kind of design developed by two economists, Derek Neal and Gadi Barlevi, and basically they propose a new type of design called “pay-for-percentile”, which they tried to really make sure that teachers perceived that the incentives under which they were operating are fair. And when we test that pay-for-percentile scheme out in China, we find that it actually really leads teachers to put forth more effort than other traditional types of performance pay designs.
FBO: Basically, the idea that you need a lot of data to do this type of mechanism, it’s one of the main points that I want to have in this discussion. All the evidence that I know—experiments in India, in Kenya, in China—they have the same kids followed in time. However, the information systems that we usually have in developing countries are not of that nature. Usually, they are of the nature of cohort exams in specific grades. Let’s say in fifth grade, they take the cohort and they do an exam there. If we link teachers’ performance to those type of tests, which are cohort tests, the incentives are very weak.
Clearly one message is that we need to increase the quality of the way we are following kids in time. And that is something that I think is an important objective of a developing system: to try to start following kids in time.
[But] the problem of attaching student incentives to those type of exams is that they can corrupt the exam. So what I’m saying is: yes, it’s super important to try to put more technical space for the way we are following kids in time and the way we are linking kids to teachers, but also if we are going to put this strong accountability system [in place], we need to be very aware of the problems of inflation and corruption of the measure.
PL: But again, the design of teacher performance pay—if it’s done in the right way—can alleviate a lot of these different types of problems. Just to give one example, I think that eventually developing countries are moving towards having a longitudinal system where they track students and assess them periodically. I think that one of the suggestions behind this pay-for-percentile method proposed by Barlevy and Neal is that not only do you have a regular standardized test tied to the curriculum that’s high-stakes, but you also have a low-stakes test that measures student learning gains over time. The high-stakes test is the one that you can tie teacher incentives to; that can be harder for teachers to know year-to-year what exactly is on that test, it can be harder for them to corrupt. Whereas the low-stakes test that can accompany students as they proceed through the system can actually have things that are more transparent and open for teachers to understand.
PL: I think that there are things that—even in the United States, even though the context is very different from the context of developing countries; and even grouping developing countries as one big unit is hard because, you know I work a lot in rural China and that’s very different from Pakistan where Felipe works; we don’t have a problem there with teacher absenteeism for example—but the point is that there are some generalized aspects of the design of performance pay that I think can be very helpful in trying to elicit more teacher effort.
FBO: One of the points that I think we are trying to make is that it is critical to think about the design issues, and it is critical in both developed and developing countries. We both—both types of countries—can learn a lot on the design issues. In the United States and other developed countries, problems of absenteeism are very minimal. The problem of teacher absenteeism in developing countries is very large. It’s an important consideration. So something that we need to also start thinking [about] is: yes, there are these programs and designs are important, but also there are different margins that we can attack in different situations to solve these different problems.
PL: So I would say—you know, there are several studies coming out. I’m really excited about the fact that each of these new studies is proposing new thoughts about the design of performance pay. Some of them are replicating efforts that we’ve done in the past, or others have done in the past. So I think we’re going to have richer evidence for policy-makers to draw upon as they begin to plan how they should design performance pay.
FBO: I would like to think about two types of teachers: the ones that are already in the system, and the ones that are the future teachers. The big impact that we want to do with teacher incentives are for this stock of teachers that are already in the system. I think that the real future of education is in how can we choose better people who enter the teaching profession, and how can we train them better before they enter the system? So, what I am saying is that I think that these [incentive] programs are very important, but thinking about the future of teachers and the right training for the future teachers is even more important.