There has been a long debate about paying the government teachers (and public sector employees, in general) as per their performance. It has been argued that problems like high absenteeism, lack of teaching when in school, and abysmal quality of teaching might be alleviated if the teacher salary is made conditional on outcomes reflecting their performance. Given the unionization among the government teachers, wider implications for payment policies in public sector, and some legitimate concerns, this policy has been opposed vociferously. Rigorous empirical evidence on this controversial question has started coming in only recently. This post summarizes results from some of the recent research papers analyzing the impact of making the teacher salary conditional on certain observable outcomes.

Duflo, Hanna & Ryan (2007) conducted a randomized experiment in the Indian state of Rajasthan where, in 57 out of 113 randomly chosen schools, a teacher’s daily attendance was verified through photographs with time and date stamps. In the comparison schools, teachers were paid a fixed rate for the month- Rs. 1,000. The teachers in the treatment schools received Rs. 50 bonus for each additional day they attended in excess of 20 days, and Rs. 50 fine for each day of the 20 days they skipped work, in a given month. The authors find that over 30 months in which the attendance was tracked, teachers in the treatment schools had an absence rate of 21% as against 42% in the comparison schools. Thus, the program halved absenteeism. Reduction in absenteeism meant the students in the treatment schools received more days of instruction, and their test scores showed significant improvement.

The schools in the sample are run by an NGO- they are not government schools, there is only one teacher and 20 students on average in the age group of 7 to 10 years. So the schools are quite small. The treatment involves monitoring and financial incentives, so it would be difficult to separate out the effect of monitoring and of the financial incentives (thought the authors estimate a structural model with some assumptions which indicate that the effects are driven by the financial incentives).

Glewwe, Ilias and Kremer (2010) analyze the effect of providing gifts (i.e. incentives in kind not cash) to all teachers and the headmasters (i.e. group incentive, not individual incentives) in 50 out of 100 government schools if the students scored well on district exams, in two Kenyan districts. Prizes were awarded based on the average performance of all students in the school who were in grades 4-8. The average performance is calculated through a formula, which depended on the fraction of enrolled students taking the exam and their scores. The results suggest that the treatment schools do significantly better on both these counts, but the effect on number of enrolled students taking exams is much stronger than the effect on test scores. These effects seem to kick in after some lag. Once the program ended (after two years), there was no difference in the treatment and comparison schools. The authors also find that the program had no effect either on teacher attendance or the manner in which teaching is done though the test prep sessions seemed to have gone up.

Muralidharan & Sundararaman (2011) analyze a large teacher performance pay program in Andhra Pradesh (AP). The study has a representative sample of 300 government schools in rural AP- 100 in control group, 100 in individual incentive treatment group and 100 in group incentive treatment group. This design allows the authors to analyse, in a single experiment, the differential effects of individual vs. group bonuses.

The incentives were based on the average improvement in test scores (math and language) of students subject to a minimum improvement of 5%. In the schools with group incentives, all the teachers received the same bonus based on average school-level improvement in test scores, while in the schools with individual incentives, bonus for teachers was based on average test score improvement of students taught by specific teacher

The authors find that at the end of two years of the program, students in the incentive (both, individual and group incentives) schools performed significantly better than those in the control schools    . Interestingly they find that students in the schools with the group incentives and the individual incentives performed equally well in the first year, but the individual incentive schools significantly outperformed the group incentive schools after two years of the program. They also find that students in the incentive schools performed better not only in math and languages, but also in science and social studies, for which there were no incentives. Their results suggest that the main mechanism for the impact of the program was not increased teacher attendance, but greater (and more effective) teaching effort conditional on being present.

The above experiments were carried out in the developing countries. A recent NBER paper (NBER working paper no. 16850) reports results from a randomized evaluation in over 200 New York City public schools. Whether a school receives incentive amount was based on whether it met the annual performance targets based on school report card scores. These scores depended on a weighted average of variety of criteria described in detail in the paper. The author finds no evidence that the teacher incentives increase student performance, attendance or graduation rates. He also doesn’t find any change in student or teacher behavior.

Thus, the experiments in the developing countries suggest that incentive payment for teachers do work. They also suggest that if there are multiple objectives and if the incentive structure is not simple, incentive payments may not work. Further, individual incentives seem to be more powerful than the group incentives. Given that 40% of the SSA budget (Centre and the states combined, for 2009-10) is spent on teacher salaries, and the teacher salaries are also the largest component of the state education budgets, it might be worth trying out this policy.