Few issues in education are as polarising as proposals to introduce “pay for performance” for teachers. This blog is not an attempt to persuade you that such proposals are right or that they are wrong. Rather, it is a plea for us all to pause, unpack the question, and appreciate that arguments about the external validity and contextual nature of “the evidence” cut both ways.

Here is the essence of our argument. High-performing education systems already use some form of “pay for performance," in the broad sense that teachers on the payroll have to “perform” in the basics and are expected to contribute to student learning, without controversy. Only dysfunctional systems feature the reverse: “pay for no performance at all.” For instance, teachers often remain on the payroll even if they fail to even turn up to school. Precisely because of this difference, theories and evidence about one controversial variant of “pay for performance”—paying teachers bonuses based on narrow performance metrics, especially those based on student learning—that are primarily grounded in the experience of well-functioning systems such as the USA and the UK may not apply elsewhere.

“Pay for performance”: What is “pay”? What is “performance”?

The term “pay for performance” can be interpreted in (at least) three different ways.

First, “pay” could be salary and “performance” could be doing the minimal basics implied by the job description of “teacher”: reliably turning up at school, going into a classroom, and spending time on task teaching students (and minimal performance would include not doing other things like beating or abusing or soliciting bribes from students).

Second, “pay” could be everything that contributes, in a hedonic sense, to teacher well-being: salary and benefits (current and future) plus the nature of the posting (location and task assignments); and “performance” could encompass how well a teacher is doing across the range of his/her duties, however these were specified and evaluated.

Third, “pay” could be a monetary bonus incremental to existing compensation; and “performance” could be how well a teacher is doing on a set of objective metrics, assessed against pre-specified criteria.

Let’s call these three interpretations of “pay for performance”:

1. Salary for the basics.
2. Hedonic rewards for broadly-appraised performance.
3. Bonus for objectively-measured specific dimensions of performance.

The first interpretation seems common sense—why would scarce public sector resources be devoted to paying a teacher who did not perform the basics? Indeed, nearly all education systems in high income settings have, by this simple definition, quite strict pay for performance compensation schemes. But this version of pay for performance is not always achieved in low income settings. The World Bank’s Service Delivery Indicators reveal average classroom absence rates of over 50 percent in Tanzania. In unannounced visits during scheduled classroom time in 2012, more teachers were found to be absent from class than present. A survey in India in 2007 found almost a quarter of the poorer students in public schools had been “beaten or pinched” by the teacher in the previous month. Yet teachers who persistently fail to do even the basics continue to draw a salary. Surely this lack of consequences for non-performance might be a frustration to, and even perhaps reduce the intrinsic motivation of, dedicated teachers who are doing the basics, who may well feel that their professional efforts are being devalued. This obvious, perhaps borderline pedantic point, just emphasises that no one really argues for “pay for non-performance” so the argument is really about what performance is measured and how it is linked to pay.

The second interpretation also seems pretty natural—why not reward a teacher who is performing well across the board with something he/she values, like a posting close to home, an interesting assignment, or even future benefits via a promotion? While most education systems in high income settings also use pay for performance in this broad sense, this version does not always work as planned in low income settings. Iyer and Mani (2012) show that political connections (via caste) can dictate the assignment of bureaucrats in India, while Habib (2015) argues that similar extraneous factors influence the promotion of teachers in Khyber Pakhtunkhwa province, Pakistan. Again, this must surely be a frustration to teachers who work hard, love their job, and do a good job, but this is not reflected in their rewards as they lack the right political connections.

The third interpretation, or some close variant, is what economists often mean when they use the term “pay for performance”. While not a huge departure from the first two schemes, the use of objective performance metrics to trigger bonus payments for teachers is controversial, particularly if the metrics include teacher value-added constructed from student learning outcomes. Such schemes are less widely used and are not commonplace in either high or low-income settings.

What is the controversy?

The list of concerns that have been raised in relation to paying teachers bonuses for objectively-measured performance is long (and legitimately so, as it is a complicated question), but to justify our plea to pause, we focus on just two: unintended consequences and opposition from teachers.

Unintended consequences. There are at least three concerns here. Since the set of objective performance metrics will necessarily be narrow in focus—not everything can be measured on a feasible timescale/budget (or even measured “objectively” at all), there is a concern that effort will get drawn away from those important and productive tasks which do not contribute to the chosen measure. There is also a worry that the act of offering an extrinsic financial incentive to perform (some part of) the job will erode a teacher’s intrinsic incentive to teach. Further, it is feared that these extrinsic incentives will prompt teachers to engage in behaviour that “games” the measure, from teaching to the test to outright cheating.

Are these concerns sufficient to caution against paying teachers bonuses for objectively-measured performance? Wait, pause before you answer and during your pause think where, and in what contexts these objections are the most powerful.

Suppose the proposal is to introduce a bonus scheme into a high income setting where “salary for the basics” and “hedonic rewards for broadly-appraised performance” are functioning well and learning outcomes are already quite high. Moreover, suppose teachers who fail to do the basics do not stay on the payroll, and, due to a combination of intrinsic motivation (including internalised professional norms) and promotion prospects, the typical teacher is performing at least adequately, perhaps even well. In this context, layering a bonus scheme on top of the largely functional system may not make sense. Distorting teacher effort, reducing intrinsic motivation, and gaming the measure could plausibly make things worse. In fact, the merits of bonus schemes are debated, not just in education, but also in purely private sector firms in the rich countries for many of these same reasons. Even economists working on “agency theory” emphasise how difficult it is to align pay for performance schemes within the standard HR policies of modern firms (e.g. Roberts 2004), and bastions of capitalism like the Harvard Business Review continually debate whether narrow versions of pay for performance are worth the bother. Not at all surprisingly, therefore, evidence from impact evaluations in education systems in high income, high functioning, settings suggest that sometimes bonus schemes have led to worse outcomes (e.g. Breeding, Béteille and Evans, forthcoming 2019).

Now suppose the proposal is to introduce a scheme into a low income setting that is not functioning well and where teachers who fail to perform the basics remain on the payroll. That is, imagine a setting where, due to already eroded intrinsic motivation and professional norms and a lack of promotion prospects (or other hedonic rewards) based on broadly assessed high performance, few teachers are performing adequately and many are not even achieving minimally adequate time on task to teaching and learning activity. In this context, layering a bonus scheme on top of the existing system might make sense. All three concerns loom less large. True, getting teachers to focus on measured inputs (school presence, classroom conduct) and outcomes (student learning in tested subjects) is only part of the job of a teacher. But they are an important part of the job and these might be a natural place to start changing not just the compensation scheme but the norms. Plus, it is hard to imagine teachers being more demoralised by a scheme that rewards good teaching than by the status quo. In fact, maybe some teachers would actively like to see such a scheme introduced? Which brings us to…

Opposition among teachers. It is certainly true that in some high-income settings teacher unions have vociferously opposed the introduction of “bonus for objectively-measured performance” schemes. Surveys in the USA and UK, for instance, show only about a third to two-fifths of teachers are supportive of basing incentive payments on test-scores (Leigh, 2012).

But what about in less well-functioning systems? Take Tanzania, where as we mentioned above in 2012 the World Bank’s SDI team found that over 50 percent of teachers were failing to do the basics such as show up to class. Mbiti and Schipper (forthcoming 2019) use data from Tanzania in 2013 and 2014 on teacher attitudes towards such a bonus scheme. Specifically, during an impact evaluation conducted in 350 public primary schools, the authors asked teachers the following question: “What is your overall opinion about the idea of providing high-performing teachers with bonus payments on the basis of objective measures of student performance improvement?” On a five-point scale over 95 percent of teachers gave a favourable (4) or very favourable (5) response. Moreover, there was no difference in support between teachers who did and did not experience such an incentive scheme as part of the evaluation. One of us (Clare) asked the exact same question in a recent study in Rwanda and found that over 80 percent of teachers gave a favourable or very favourable answer. Their question originated from a study in India in 2005-2007 where the authors found over 85 percent support.

It seems sensible that teachers would oppose the introduction of bonuses based on narrow measures into well-functioning systems but—perhaps with the exact same attitudes and for the exact same reasons (e.g. the impact on intrinsic motivation)—favour them in low-functioning settings, at least if base pay remains the same. Of course, this is just a hypothesis, and the reason behind the apparent acceptability of such schemes in low-functioning settings should be probed in future research. Our point here is that teacher opposition cannot be taken for granted.

Summing up: external validity cuts both ways

In writing this blog, we hope to have persuaded you of one thing: when asked “should ‘pay for performance’ be used for teachers?” we should all pause before we leap to an answer. High-performing education systems already use some variant of “pay for performance”—in the broad sense—without controversy. In fact, the opposite, paying teachers for nothing at all, would be highly controversial in functional systems. But dysfunctional systems can feature the reverse: individuals remain on the payroll even if they fail to perform the minimal basics implied by the job description of “teacher”. Precisely because of this difference, theories and evidence about the more controversial variant of “pay for performance”—bonus for objectively-measured performance—that are grounded in the experience of well-functioning systems may not apply elsewhere.

One recent study in Rwanda (of which Clare is an author) found positive impacts on learning from the introduction of a pay for performance scheme. This does not mean the scheme that worked in Rwanda would work in India or Ghana or Guatemala. However, this modesty about external validity should cut both ways. Studies from the USA and UK should not dominate the discussion about pay for performance in settings that are obviously very different in many, many respects. The answer to any question requires a pause for thought about conditions and context rather than a presumption that there is an easy and universal answer.