This post originally appeared in the World Bank Research Digest Fall 2017 edition.

Extensive research from the United States clearly establishes the importance of good teachers for students’ learning outcomes. It also shows that predicting teacher quality is difficult. Unfortunately, we don’t know whether these findings also apply to low-income countries. Filling this gap is critical: student learning in these countries is poor, and teacher salaries account for 80 percent of recurring education expenditures. Increasingly, countries are asking how to recruit and retain high-quality teachers.

In a new paper, Bau and Das investigate these issues using a unique data set on primary school students and their teachers in both public and private schools in Punjab, Pakistan, for 2004–07. The data track students through primary school, matching them to the teachers who taught them in each year. One novel aspect of the data set is that it includes scores on primary-school-level tests in Urdu, English, and math for both students and teachers.

Using student test score data, the authors calculate a value added (as a proxy for productivity) for each teacher: an estimate of the test score gains that a random student would receive if assigned to that teacher. These estimates show that moving a student from a teacher at the 5th percentile to one at the 95th would increase the student’s test scores by 0.5 standard deviations — the equivalent of more than a year of school.

So teachers clearly matter. But as in the United States, predicting their effectiveness is hard. The authors found that only teachers’ content knowledge (as measured by their scores on the primary school tests) and their first two years of experience predict their value added. Neither a bachelor’s degree nor teacher training—two factors often used in teacher recruitment—is associated with value added.

Are there teacher characteristics that predict higher salaries in Pakistan? The authors examine this issue in both the public and private sectors. Results show that having a bachelor’s degree positively predicts wages in both sectors. And in the private sector, teachers with higher value added are paid more. In the public sector, however, there is no relationship between teacher value added and wages. These findings suggest that teacher value added is at least some-what observable and that the private sector identifies and rewards better teachers with higher wages. In contrast, higher wages in the public sector are not associated with higher teacher value added.

This lack of an association between teacher productivity and wages suggests that the public sector does not reward better teachers with higher wages. But this does not mean that increasing teacher wages would not attract higher-quality applicants for teaching positions.

A change in hiring practices in Punjab makes it possible to assess how teacher wages affect the quality of applicants. In 1998, Pakistan initiated unanticipated nuclear tests, which led to international sanctions on the country and tighter budget constraints in the province. As a result, teacher hiring was largely frozen for three years. When hiring resumed, almost all new teachers were hired under temporary contracts with 35 percent lower salaries. With the caveat that the new teachers’ contract status may have increased accountability, the authors compare the productivity of teachers hired right before the nuclear tests with that of teachers hired right after hiring resumed to see whether lower salaries attracted worse teachers.

The results show no negative effect of being hired at a lower salary on teacher productivity. Indeed, when the authors account for the fact that teachers are observed at different levels of experience, contract teachers performed better than permanent ones.

The fact that reducing teacher salaries by 35 percent had no adverse effects on student outcomes might seem surprising. But comparing public and private sector teachers’ salaries helps illustrate why this is so: there is virtually no overlap, with public sector teachers making about five times as much on average as private sector teachers. This large public sector premium is not unique to Pakistan. According to a recent study using data from 52 countries, these premiums are larger in low-income countries than in high-income ones.

Altogether, these results suggest that calls to improve student learning by raising teacher wages, while keeping the same hiring and pay structure in place, are misguided. Redesigning compensation systems for government employees is complex enough as it is; one issue that policy makers need not worry about is that teacher salaries are too low to attract talented teachers.