Roland Fryer has been a pioneer in experimenting the use of financial incentives, both individual and group-based, to prod students, parents and teachers to perform better. His latest research paper comes from a randomized control trial of teacher incentives on student achievement in over 200 New York City public schools. His conclusion runs contrary to findings from developing countries,
"I find no evidence that teacher incentives increase student performance, attendance, or graduation, nor do I find any evidence that the incentives change student or teacher behavior. If anything, teacher incentives may decrease student achievement, especially in larger schools."
Though schools had the flexibility to fashion their own incentive structure, a vast majority of schools choose to distribute incentives equally (this could have under-cut individual initiative) if teachers achieved their performance targets and also preferred group-based incentives (which may promote free-riding).
Prof Fryer examines the failure of incentives on teachers from four perspectives - incentives may not have been large enough; incentive scheme was too complex; group-based incentives may not be effective; and teachers may not know how they can improve student performance. He argues that the most probable reason for the lack of response with incentives in the NYC schools is that the incentive scheme is far too complex and provides teachers with too little agency.
He examines the results of the Project on Incentives in Teaching (POINT), a pilot initiative in Nashville, Tennessee, which finds that a larger incentive in schools was still not any more effective. He argues that group-based incentives are more likley to work when the number of teachers in each school is small (compared to New York's more than 50 in each school). When teacher numbers are small, as in developing countries, monitoring and imposing cost on those teachers who shirk their responsibility is easier.
However, Fryer points to a study by Victor Lavy in Israel on success with teacher incentives in schools with an average of 80 schools. He found that when teachers were incentivized on the average number of credit units per student, the proportion of students receiving a matriculation certificate, and the dropout rate, there was a positive and significant impact on the average number of credits and test scores.
It is not possible to draw too many conclusions from these mixed results. I have already blogged about the difficulty with implementing teachers-based financial incentives. Here are a few observations, some of which have already been dealt with in the earlier posts
1. Prof Fryer's disappointing results stands in contrast to apparent successes with incentives in developing countries. Given the much lower baseline achievements in developing countries, it is plausible that smaller incentives may be effective. In fact, the much lower level of standards and consequent higher potential for marginal improvements in developing countries means that it is possible to design simpler incentives that teachers will find easier to respond to.
2. The long-term impact of financial incentives, especially in professions like teaching and activities like studying, is largely unknown. Does extrinsic motivation crowd out (and even kill) intrinsic motivation? A student being incentivized to read a book (or write an essay) for Rs 50 for sometime is surely going to expect (only the degrees vary) some similar incentive every time. The same psychological chain of thinking works with teachers and parents. Without fully understanding its consequences, pumping for financial incentives is evidently dangerous and could have potentially disastrous consequences.
3. There is the difficulty of identifying when (and which environments) do we offer financial incentives. In most developing country environments, even small supervisory and monitoring improvements can achieve disproportionately large improvements in performance. Most often, the same investments if made in improving the capacity of the supervisory system can achieve gains that are far higher than those with cash incentives. It is like jumping high to pluck a low hanging fruit!
4. Finally, there is the issue of scaling up such incentives. It is one thing to run a pilot with 500 or 1000 schools, and a qualitatively different task to do it in an entire state. Forget the problems with quantification, gaming of the performance parameters, apples-to-apples comparison, and all other qualitative challenges associated with having a reasonably widely acceptable performance monitoring system. Implementation on scale generates the very real and almost inevitable risk of the incentive ending up as an entitlement. Teachers could start demanding incentives for even complying with their routine duties and responsibilities, for which they draw their salaries in the first place. A corollary to this is the political impossibility of withdrawing or scaling back, even after the basic and initial objectives are achieved.
5. Performance parameters become a moving target once the stakes are raised. As long as these performance parameters are merely captured and not acted upon, except in cases of the negative outliers, there is little incentive for participants to game the system. Once the stakes increase, and financial incentives do raise the stakes (atleast for some participating teachers), incentives to game the system and manipulate results get triggered off. Staying ahead of this game is difficult, even for mature systems (as New York City, with its grade inflation on standardized tests, found out recently).
6. A less controversial and surely less risky approach with incentives is to have in place a carefully designed plain vanilla approach to incentives. Such an incentive structure would either offer (directly) non-financial incentives (say, vacation pass), or appeal to intrinsic motivation (public recognition or acknowledging with certificate, teacher of the month), or selectively reward a few bright spots (say, cash incentives to the best 10 teachers). All of them carries some or all of the aforementioned dangers. However, its much less scale and scope, eliminates many of the dis-incentives and makes its effective management possible.