Making teacher salaries contingent on teacher performance can theoretically lead to better student outcomes. On the other hand, it can incentivize cheating (by the teachers) and teaching to the test. Interestingly, no long-term study of teachers’ pay-for-performance schemes has been conducted until quite recently.

In this post, I’ll look into the first study that investigated the effect of merit-based teacher pay not only on student performance in high school, but also on post-secondary and labor market outcomes.

Lavy (2015) (ungated) conducted an experiment in 2001-2002 with 49 Israeli schools involving around 10,000 students. This is a large sample and a long time frame making it quite an ideal experiment.

Basically, a pay-for-performance (PFP) scheme was introduced in grades 10-12 at some of these schools’ English, Hebrew, Arabic and math classes. If teachers outperformed expectations they would receive a bonus. Bonus receivers were further subdivided into four groups, receiving an award of $7,500, $5,750, $3,500 and $1,750, respectively. Of the 629 teachers involved, 302 received awards. The students were also followed up on in 2009-2011 when they were 27-29 years old.

The author examined several life outcomes on which the PFP scheme could have an effect. These can be broadly grouped into three categories: high school outcomes, post-secondary schooling outcomes, and labor market outcomes. Let us go through each of these.

For high school outcomes, the expected effects are clear, and even if positive effects are found, that is no evidence of the PFP scheme working properly. Good high school outcomes could be the result of cheating or teaching to the test. Indeed, average matriculation scores and matriculation rates both increase for students under the PFP scheme.

Interestingly though, the increase in matriculation rates is not significant for the whole sample. It, however, is significant for the lower three quartiles indicating that at-risk students are preferentially affected by PFP interventions. The increases are sizable. In the lower three quartiles, the probability of matriculating increases by 5.5 percentage point, a 13% increase over the control group.

(A side note: matriculation refers to receiving a *Bagrut* certificate, which is necessary for entrance into higher education. On average, around 45% of the Israeli student population receives one.)

Let’s continue with post-secondary schooling outcomes. In the lower three quartiles, students taught under a PFP scheme are more likely to attend a post-secondary institution. To be more precise, they’re more likely to attend universities but their attendance at other (lower-level) post-secondary institutions (e.g. vocational/applied colleges) does not change.

For the full sample, there is no overall increase in post-secondary attendance, but there is a shift from applied/vocational institutions towards universities. Altogether, treated students are 4.6 percentage points more likely to attend universities (a 27% increase), but this effect is as high as 5.5 percentage points (45% increase) for the lower three quartiles. Treated students also stick around longer in university: years spent at university increases by 33% in the full sample, and by 63% in the lower three quartiles.

This shows that whatever the underlying reason for good high school performance was, the gains from the PFP scheme (higher matriculation rates) linger on and can affect students in the longer term. More of them go to university, and at-risk students are more likely to do post-secondary studies altogether as well. Furthermore, they are less likely to drop out (as indicated by the longer time spent at university).

Finally, what about employment outcomes? The figure below (made by me) shows the strong effect on wages. The treatment effects are always significant, and they show that students taught under PFP schemes earn 7.2% to 9.4% more at ages 27-29.

Employment and unemployment rates at ages 27-29 are not significantly affected by the PFP scheme. There is, however, evidence that treated students are less likely to be on disability. People are generally more likely to be on disability when their employment outcomes are bad, so this is further proof of the PFP scheme’s positive effects.

So does this indicate that PFP for high school teachers can even increase future earnings directly? Not really. The data indicates that almost all earnings increases are due to the better educational outcomes of treated students. Interestingly though, there is some evidence that the improvements in post-secondary educational outcomes are partially a direct result of PFP.

This means that even after taking better high school matriculation outcomes into account, PFP still has an independent positive effect on post-secondary outcomes. So it is possible that teachers go beyond teaching to the test and for instance (i) appeal to or reignite students’ natural curiosity, or (ii) equip students with more useful skills than the baseline requirements for successful high school graduation.

Bottom line: in this first experiment on pay-for-performance for high school teachers, the scheme works surprisingly well, and has persistent long-run benefits.

One caveat that may come to mind is whether this’d work on a large scale over a longer time horizon. I think it’s impossible to tell from this experiment what would happen if all teachers in all schools within a country were suddenly incentivized by cash bonuses. Also, given enough time, the teachers may learn to game the system.