Wednesday, July 25, 2012

Prospect theory and financial incentives for teachers

Daniel Kahneman and Amos Tversky formulated prospect theory to explain how real-world people evaluate losses and gains. In particular, they claimed that people show endowment effect - people value a thing more once it becomes their own - and loss aversion - people are more motivated by avoiding a loss than acquiring a similarly sized gain. These findings have been experimentally validated many times subsequently and is now widely accepted.

Roland G. Fryer, Steven D. Levitt, John List, and Sally Sadoff have a new working paper that seeks to leverage loss aversion among teachers to improve student learning outcomes. The challenge with the use of performance incentives have been that while there are experimental findings to prove that teacher quality matters for both student learning outcomes and long-term life outcomes, success with the use of financial incentives for teachers has been mixed (see this and this) for a variety of reasons. The use of financial incentives assume greater significance in view of the difficulty in using various observable screening techniques to ex-ante identify who is a good teacher. All these challenges are compounded in developing countries.

The study conducted by Fryer and Co were in nine schools of Illinois during the 2010-11 period. Teachers were tandomly selected for the experiment, and among those selected one again randomly selected group was given the standard incentives - bonuses at the end of the year linked to student achievement - while the other group teachers were given a lump sum payment (of same size as the bonus) at the beginning of the school year and informed that they would have to return some or all of it if their students did not meet performance targets. In this context, the authors write,
We demonstrate that exploiting the power of loss aversion - teachers are paid in advance and asked to give back the money if their students do not improve sufficiently - increases math test scores between 0.201 (0.076) and 0.398 (0.129) standard deviations. This is equivalent to increasing teacher quality by more than one standard deviation. A second treatment arm, identical to the loss aversion treatment but implemented in the standard fashion, yields smaller and statistically insignificant results. This suggests it is loss aversion, rather than other features of the design or population sampled, that leads to the stark differences between our findings and past research.
I have blogged here, here, and here about the use of financial incentives to teachers for improving classroom instruction quality and thereby student learning outcomes. I am inclined to believe that any teacher performance-based pay system, while unobjectionable at a theoretical level, may be very difficult to implement, both for political and administrative reasons, in the prevailing environment in countries like India. However, if financial incentives are implemented, its framing is a crucial small detail, which can dramatically improve the chances of its success in implementation. 

I see two big first level challenges with the deployment of financial incentives as suggested. The biggest evaluation problem with education is the challenges with getting reliable and cross-comparable data on learning outcomes. Given the stronger loss-aversion motivation, upfront rewards, to be taken back if the expected results do not come, are liable to nudge/force teachers into manipulating scores with much greater intensity. Further, the concept of taking back something given (even if done non-invasively by deducting it from the subsequent month's pay) is a relatively new strategy and may generate other unanticipated incentive and systemic distrotions and pushbacks.

Given the logical simplicity and attractiveness of this framing, it has uses beyond education and teachers. Further empirical studies may be required before we can draw more meaningful inferences.

1 comment:

KP said...

Dear Gulzar,

I am fundamentally against an incentive based system of payment -whether it is teaching (with marks forming basis for evaluation, mostly) or if it is for IAS officers with much broader performance parameters - particularly in cadre based public services.

I think a process of weeding out bad teachers and officers is a better disincentive -than a system that will work in ways to convert the incentive to a norm.

I find that these experiments lack an ethical basis - and the sole motivation appears to be an attempt to generate a kind of one-upmanship or a dog-eat-dog behaviour - and this is particularly ineffective among cadre based services.

Normally the entire cadre will work in unison to blunt the effect of these kind of incentives and slowly capture the distribution to become a norm with little variability in performance.

This is a quick response - I think there are depeer ethical issues and fallacies (fallacy of composition for instance) - particularly with experiments on incentive based manipulation of behaviour in government services.

Thinking of scale, these experiments appear to be out of touch with real issues of institutional behaviour.