These past weeks I’ve visited several southern African nations to assist on-going evaluations of health sector pay-for-performance reforms. It’s been a whirlwind of government meetings, field trips, and periods of data crunching. We’ve made good progress and also discovered roadblocks – in other words business as usual in this line of work. One qualitative data point has stayed with me throughout these weeks, the paraphrased words of one clinic worker: “I like this new program because it makes me feel that the people in charge of the system care about us.”
This expressed sentiment is in stark contrast with a prominent viewpoint that the introduction of an incentive program is ineffective at best and counter-productive at worst. A forceful summary of this view can be found in a 2012 editorial by Steffie Woolhandler and Dan Ariely, published in the Health Affairs journal blog. Woolhandler and Ariely conclude their editorial with the warning that “Few have countenanced the possibility that P4P (pay-for-performance) may simply not work in health care”. Why may P4P not work? Their answer:
I’ll review their cited evidence, but first it’s instructive to summarize a dominant psychological theory, known as Cognitive Evaluation Theory (CET), and how it models the potential impact of incentives on individual motivation. In this framework, two broad categories of motivation spur our behavior – intrinsic and extrinsic. Our intrinsic motivation pushes us to pursue activities that provide their own reward, so motivation for these activities is not dependent on external rewards. On the other hand, the promise of external rewards (of any form – verbal, financial, in-kind) propels our behavior through extrinsic motivation.
Further, in the view of CET, we all have fundamental psychological needs for:
- Autonomy, defined as “the urge to be a causal agent of one’s own life”, and
- Competence, “the urge to master skills and control outcomes”.
Incentives affect intrinsic motivation only if they affect self-perceived autonomy and/or competence. This implies that context is immensely important – what determines the motivational impact of an incentive is the agent’s interpretation in relation to her own feelings of autonomy and competence.
If an incentive disempowers an agent by reducing her perceived autonomy, then her intrinsic motivation is reduced (as her extrinsic motivation likely increases – this is why incentives are broadly effective in many economic settings). However if an incentive provides information on ability or achievement – i.e. “you did so well that you earned the big bonus” – then it can satisfy the need for competence and hence increase intrinsic motivation. A priori an incentive can have conflicting effects if it is perceived as controlling (thus thwarting autonomy) but also informational (thus enhancing competence).
So context matters a great deal. The incentive form also matters a great deal. We can broadly think of rewards as either task-contingent or non-contingent. Non-contingent incentives are something like paying a salary without any reciprocal requirement of work (a very weak incentive). Since task non-contingent incentives do not require engagement (beginning a task), completion (of the task), or performance (doing the task well), these incentives are neither informational nor controlling and therefore should not affect intrinsic motivation. These types of incentives are also fairly rare.
More common incentive forms, and those used by P4P, are task-contingent. Yet task-contingent incentives in turn are sorted into three sub-classifications: engagement-contingent, completion-contingent, and performance-contingent. As the incentive form moves from engagement- to completion- to performance-contingency, the degree of control asserted on behavior increases (potentially limiting autonomy) but so does the informational content conveying success and competence. Therefore the effect of the introduction of incentives on intrinsic motivation (as well as ultimate performance) is unclear.
So let’s return to the editorial mentioned above. As evidence for the likely failure of P4P they cite numerous RCT studies that demonstrate:
- 56 college undergraduates given puzzles to solve were less successful when offered monetary reward than when given verbal reinforcement or no reward at all.
- Paying for blood donations actually reduced the quantity of blood donated.
- Swiss volunteers reduced the time spent in volunteer activities once they were offered payment for the same activities.
- Parents in Israel were more likely to pick their children up late from school once fines were imposed for late-pick-ups (as opposed to social disapproval).
- Both MIT students and villagers in India solved fewer complex puzzles when offered large incentives than when offered either small incentives or no incentives.
Finally Woolhandler and Ariely cite a masterful meta-analysis of 128 studies by by Deci, Koestner, and Ryan (DKR). (study 6- DKR)). This wide-ranging review concludes that incentives reduce intrinsic motivation especially when incentives are task-contingent and tangible (i.e. non-verbal).
Wow! That seems like a lot of compelling evidence. But…
Note that all of the studies involve a contrast between a non-incentivized control group and (some form of) incentivized treatment group. Indeed DKR explicitly restrict their meta-review to studies that involve an experimental group compared to a no-reward control group. Further all off these studies were done either in a laboratory or under well-controlled laboratory like conditions and virtually all of the 128 studies were conducted on children, teen-agers, or college students in the US or Europe.
There is surely much to learn, in general, from research that experimentally introduces incentives into a previously non-incentivized environment. But are there clear implications for the P4P? I’m not sure. In fact, I am skeptical.
Why? First note that health workers are already in an environment where incentives are used to control behavior (this situation is commonly known as “work”). P4P reform is only a change in the form of incentive. To be more precise, P4P reform is a movement along the contingency axis of incentive type. In developing countries, P4P typically involves a switch from engagement- to completion-contingency incentives, while in richer countries the switch is more typically from completion- to performance-contingency.
Why is this distinction important? Precisely for the theoretical reasons sketched above – there are contrasting effects on intrinsic motivation as we move along the contingency axis. It’s important to note that the DKR meta-analysis finds no difference in the negative effect of incentives on intrinsic motivation whether they take the form of engagement- or completion- or performance-contingency. If anything, the DKR meta-results suggest that performance-contingent incentives impact intrinsic motivation the least (as CET theory predicts since performance-contingency contains the most information on competence). So the best guess on the basis of this definitive meta-review is that P4P reforms introduced in the public sector will have no net effect on worker’s intrinsic motivation.
And there are other differences that get in the way of generalizability from the research cited above. Results can vary depending on whether the task is easy or difficult to measure, or if the incentivized task is interesting or boring. The activities in these studies may have been either mundane or fun, but none of them come close to the repeated activities that one would conduct in a profession like health care. Further none of the study groups are adult professionals and, yes, all are WEIRD populations with questionable generalizability to public service workers in developing countries.
I am quite open to the idea that P4P reform reduces the intrinsic motivation of health workers. It may very well do so, at least in some contexts. But these studies don’t directly touch on this issue. Furthermore, even if P4P reduces intrinsic motivation, by definition it should increase extrinsic motivation – and so the net effect of reform on worker performance can still be positive even if intrinsic motivation is reduced. From the policy stand-point, it’s the net effect we care about first and foremost.
Yes I know that there are questions about long-term sustainability. We know next to nothing about the long-term effects of P4P. It is indeed possible that the gain in extrinsic motivation is short lived while the reduction in intrinsic motivation is permanent. This would be a bad thing. But if we are open minded we should also consider the opposite hypothesis: perhaps a new incentive scheme is first seen as controlling because it is a change from business-as-usual, but then no longer seen as controlling as the scheme becomes more familiar. Also, perhaps over time there will be selection into the health work force of individuals more responsive to extrinsic motivation and more resilient to lower perceived autonomy in the work setting.
The fact is we don’t know the definitive answer to any of these questions. We will assuredly know more in the coming years given the explosion of research on this topic. Until then, any conclusion is pre-mature.