Anna Yalouris, from the Abdul Latif Jameel Poverty Action Lab (J-PAL), was back at the University of Cape Town’s Graduate School of Business on the 25th September to deliver a presentation on J-PAL’s research into incentivizing teachers. J-PAL conducts randomized evaluations to test the effectiveness of programs aimed at reducing poverty and is partnering with CEI's South African regional hub, the Bertha Center for Social Innovation and Entrepreneurship at the University of Cape Town, to hold a series of lectures on the evidence base for improving education.
What is the problem?
Teacher absenteeism plagues the education sector. Studies in Uganda and India have shown that on an average day, between 25 and 30% of teachers are absent. Particularly in rural areas, where schools may be manned by only a few teachers, chronic teacher absenteeism means that children automatically miss days at school.
This specific presentation focused on strategies for addressing absenteeism among service providers (e.g. nurses, teachers) in the education and health sectors. Randomized evaluations have found that similar incentives schemes have been effective at increasing service provider attendance in both the health and education sectors, where a low threat of being fired exists. The key motivation for addressing this problem is the contribution of teacher absenteeism to students not learning optimally.
Incentives are a very diverse tool. They can be incredibly effective at changing behavior and the form they take can be both negative and positive. Randomized Evaluations (REs) by J-PAL affiliates have sought to answer questions by testing the effectiveness of:
- impersonal monitoring with automatic incentives;
- monitoring by supervisors who distribute rewards and can make allowances for special circumstances;
- incentives tied to outcomes;
- community mobilization and information; and
- Intrinsic motivation and recognition
REs can measure the causal impact of a program. The possibility of selection bias is managed through ensuring that members of the groups (treatment and control) are statistically equivalent and changes over time do not invalidate results. REs provide a valid counterfactual that answers the question, “what would have happened if the program had not been in place?”
What are some of the characteristics of an incentivization programme?
J-PAL affiliates evaluated an incentive program to increase nurse attendance at rural sub-centers in Udaipur, India. The program was implemented collaboratively by a local NGO and the state and local health administration. The incentive type was punitive pay-incentives. Nurses absent for more than 50% of the time on monitored days had their pay reduced proportional to the number of absences recorded that month. Nurses with more than 50% days absent for a second month faced suspension from government service. The pay incentive was administered by the district health administration, which also had the power to excuse reasonable absences. The NGO monitored nurse attendance by installing time/date-stamping machines locked into a caddy on the wall of the sub-center. Monitored nurses were required to both sign and stamp the register three times a day, and field officers also conducted random unannounced visits to sub-centers to verify nurse presence.
What was discovered?
Initially, the threat of punitive pay-deductions led to a dramatic improvement in attendance, and nurse attendance in treatment centers was 15 percentage points higher than comparison centers. However, after 6 months, local health administration began to undermine the incentive structure. Supervisors and administrators excused all nurse absences, and nurses intentionally broke the time clock machines. As a result, 16 months after inception the incentive program had become completely ineffective and there was no difference between absenteeism in treatment and comparison centers which meant no impact on communities’ utilization of public health services. The key learning here is that incentive systems are quickly undermined from the inside if supervisors are given discretion over how and whether incentives are applied. Learn more about this study.
A similar result was experienced in Kenya when school principals were placed in the position of being able to reward teachers with bicycles, conditional on meeting certain attendance threshold for the year. In this setting, principals were more motivated to keep teachers happy and rewarded all teachers with bicycles regardless of attendance behavior, thus undermining the incentive system.
Building on this research
Learnings from these and other studies have highlighted the importance of designing impartial monitoring systems and removing supervisor discretion in administering the pay incentives. Such lessons have helped researchers design monitoring systems that can make financial incentives work. In one such study, researchers partnered with a local education NGO to evaluate a camera-monitoring scheme. Teachers were offered bonus pay for days worked above the standard 20 days per month, and a fine for days of work skipped.
To monitor attendance, teachers were given a tamper-proof camera, along with instructions to photograph themselves with the students at the start and end of each school day. The camera’s timestamp feature allowed the NGO to determine when and for how long each teacher was at school.
With teacher salaries linked to objective, impersonal monitoring, teacher absenteeism fell by 21 percentage points, equivalent to a 50% reduction in absence in treatment schools. The impacts for students were large: Students in treatment schools received 2.7 more days of instruction time per month, simply because their teachers were more likely to be at school. Test scores also rose significantly, with the result that 2.5 years into the program, children in treatment schools were 62% more likely to transfer into a private school, which requires passing a competency test. Learn more about this study.
In the case of camera monitoring in India, the enforcement of the incentive system was successful because the NGO had clear motivation to uphold the system. In order for the incentive payment to be effective, the incentive structure must also be easy for teachers to understand. In this setting, monitoring bias was also eradicated through the use of cameras, which were not subject to principal’s discretion. Taking pictures with the students was also viewed as a positive action, so teachers were more agreeable to it.
The results from these, and other randomized evaluations show that financial, as well as other types of incentives have the potential to be highly effective. However, incentives must be well-defined, easy to understand, impersonal, strictly enforced, subject to minimal supervisor discretion, and possible to monitor.
See pictures from the event below.