This volume on teacher norms in the developing world is an important contribution on one of the world’s biggest and more difficult challenges. 

The challenge is the learning crisis in the developing world. The “learning crisis” refers to the fact that nearly all children now attend at least some school—and yet, by a recent estimate, 94 percent of children in Sub-Saharan Africa, 89 percent of kids in South Asia, and two-thirds of all children globally do not acquire a modest standard of “foundational skills”. The authors estimate the total cumulative value of the gains from reaching universal foundational skills would be $700 trillion USD (current world GDP is only 96 trillion).¹

The only path to universal foundational skills is for children in low-performing education systems (which currently encompass, tragically, education systems in most of the developing world) to learn more per year of schooling—much, much more. Fortunately, there are high-performing education systems in the developing world that show these much higher levels of learning are possible. Data that tracks whether students have achieved mastery of “foundational numeracy” shows that 75 percent of children in Vietnam reach that level by Grade 4, compared to only 24 percent in Bangladesh, only 14 percent in Ghana, and only 9 percent in Pakistan.²

And these low levels of early grade learning persist into higher grades. Whereas 81 percent of enrolled 15-year-olds in Vietnam reach the SDG level of “minimum learning” in mathematics of PISA level 2 or above, only 10 percent do in Cambodia, and only 2.3 percent do in Zambia.³

So why is it that a resource-poor country like Vietnam can produce learning and high levels of mastery of cognitive skills among its youth, and other countries at similar levels of development resources cannot?

As I am not an educationist but first and foremost a development practitioner, let me wander a bit away from that question and even from education and then circle back to how teacher norms are an important part of the answer to that question.

And even further afield, let me start with a recent US science fiction movie, Stowaway. In that movie, some astronauts are on a spaceship travelling to Mars. The spaceship consists of two big modules that are connected by a very long tether. These two heavy halves spin rapidly around a centre point so that the centrifugal force creates an artificial gravity, pulling the contents of each half toward its floor.

The reason I bring this movie up is that it is a powerful example of the complex concept of “multiple stable equilibrium.” Suppose we call the two halves of the spaceship “red” and “green.” If astronauts in both halves of the ship dropped an object, everyone would find that the objects fell to the floor. But the objects would actually be “falling” in polar opposite directions. From the point of view of an astronaut in the “green” half, a dropped object in the “red” half would be falling up, not down.

The implication was that “falling to the floor”—and away from the “green” half of the spaceship—was a locally stable outcome. If (and this was a key plot point for the movie) an astronaut in the red half started climbing along the tether and slipped less than halfway through the trip, they would fall back to the floor of the red half. But, once the astronaut made it more than halfway, the dynamics of force would shift and, unless they resisted the forces, they would “fall” away from the red floor and towards the green floor.

Let me move from science fiction to health care in Rajasthan, and in particular to the behaviour of one kind of health care worker, the “auxiliary nurse midwife” (ANM) who has responsibilities for helping pregnant women with pre-natal care. Three brilliant economists (Banerjee, Deaton, and Duflo, all of whom now have a Nobel Prize in economics) worked together with a local NGO to examine health care and why health outcomes were so poor in Rajasthan. A major factor they identified was that these ANMs were very frequently absent from work, and that, moreover, this pattern of absence was widespread among ANMs.⁴

Two of them (Banerjee and Duflo) then worked with the local NGO, Seva Mandir, and the local government to design a new policy in order to increase attendance of the ANMs, with the goal of better services that would in turn improve healthy birth outcomes.⁵  As you might expect from a new policy designed by economists, it involved incentives: the proposed policy was that a number of new ANMs would be hired so that the local sub-centre facilities had two ANMs each, and that the newly hired ANMs would get their full salary only if their absence rate was less than 50 percent (and it is revealing that this was the goal attendance rate). But this wasn’t designed by economists alone; the NGO with long experience in the field contributed and the design also included: (i) the use of better technology to record attendance (machine time clocks, not paper records), (ii) since the ANMs often had out of sub-centre responsibilities, clarification from the Ministry that one day a week was designated as “clinic day” and the ANMs should always be present that day, and (iii) the engagement of local NGOs to “ground-truth” the machine clock stamped attendance records (so, for instance, two ANMs could not collude for one to clock the other in as present).

Since the economists were interested in knowing the causal impact of this new policy, there was a research effort that tracked a wide variety of outputs and outcomes. The study was implemented as a “randomised controlled trial” (RCT) and clinics were either in the “treatment” group, subject to the new policy, or “control” group that proceeded with business as usual.

Now, it might seem that this was like testing gravity: of course if half of their salary is at risk this will create a powerful incentive for the ANMs to show up. In fact, one of the researchers, Esther Duflo, had already done an earlier experiment in the same state of India, Rajasthan, with the same NGO, Seva Mandir, that showed that, in the schools run by the NGO in rural areas of the state, a new policy of requiring teachers to take a date-time stamped picture of themselves at the beginning and end of each school day had increased teacher attendance and thereby significantly increased student learning. The title of the paper reporting on the results was: “Incentives work”.⁶

When it came to the new nurse midwife policy, the officially recorded absence of ANMs fell significantly in the treatment clinics that had “high-powered” incentives for attendance, and hence most of the ANMs received their full pay. But, fortunately, the study did not just rely on the official data. It actually measured the physical presence of the ANMs at the clinic on the designated clinic day—and presence also fell significantly, such that after a year of implementation of the new policy the actual presence of the ANMs was much lower than it had been before the policy. How could it be that both absence and presence fell? Well, there was a third category called “exempt” whereby an ANM could have an official document from an authority in the Ministry stating that, on that day, they were not physically present but that this would not be recorded as an official “absence” that would count against their absence tally for wage purposes.

The title of the researchers’ paper detailing the findings reflects what they learned from that experiment: “Putting a Band-Aid on a Corpse”. That is, they did not give up on the notion that “incentives work”—the incentives did work. But the incentives didn’t work to create attendance; rather, the incentives worked to induce the ANMs working inside the public sector health care system to pay the official(s) responsible for declaring them “exempt” from duty so that they would not be penalised for not being present. The incentives worked to increase the fraction of all ANM days recorded as “exempt”, which ballooned from only about 10 percent before incentives to over 50 percent of all ANM days.⁷

From science fiction and experiments about attendance of ANMs, let’s start to circle back to teacher norms with another RCT experiment, this time about education.

In the Indian state of Madhya Pradesh, the government was implementing a new initiative to improve learning in the state, which was to introduce “school improvement plans” in which each school would assess its own conditions and challenges and then formulate a school-specific plan to address those challenges. The idea was that rather than being dictated “top-down”, the improvement plan would work “bottom-up” and the school supervision process would be reoriented to help the schools implement their plan. Again, the government agreed with researchers to implement this in a randomised way so that some districts of the state were “treatment” districts while others delayed the implementation so that they could be compared as “control” (business as usual) districts.

The RCT (Muralidharan and Singh, 2020) revealed three very interesting things.⁸

One, the study showed that schools in the SIP areas actually did produce school improvement plans and that these school improvement plans were different across schools and appeared to accurately reflect the observed differences across schools. So, if there had not been a study tracking not just the SIP “outputs” of the production of plans but also actual “outcomes” of changed behaviour, it would have been easy to declare the SIP initiative a success.

Two, and but, the study also showed that besides completing the SIP, nothing else happened. Student learning in the “treatment” SIP schools was no higher than in the “control” schools. And this lack of learning impact was not a “surprise” as nothing else measurable changed in the SIP schools—not teacher attendance, not supervision visits, not teacher practices, nothing. The authors conclude: “Our results illustrate how ostensibly well-designed programmes, that appear effective based on administrative measures of compliance, may be ineffective in practice.”

Three, a side effect of doing an experimental study to measure, among other things, learning outcomes, was that the study implemented essentially the same assessment of student learning that was done as part of routine school administration, but well invigilated.  This showed that the administrative data vastly over-represented what students actually knew, as the administrative data showed 64 percent correct responses for math, whereas the re-test that worked to prevent cheating, collusion, and manipulation of results found only 25.1 percent correct responses on the same questions.⁹  If one believed the administrative data, one would be completely misled about learning performance, thinking that students could answer about two thirds of questions correctly rather than knowing the catastrophic fact that only one quarter of questions were answered correctly. But, on a deeper level, this also reveals that those responsible for the assessment and its reporting were engaged in a massive deception, year after year.

Contrast that with another study from RISE that used videos of teachers to examine pedagogical practices of the more and less effective teachers in Vietnam. The study found that high-performing teachers in Vietnam were able to help students acquire metacognitive skills, learning about their own learning, by implementing in practice “five features of the learning environment that foster effective metacognitive learning include (a) engaging curriculum, (b) explicit instruction and modelling, (c) verbalizing, (d) assessment integration, and (e) consistent practice.” An almost equally striking finding is that these video observations of the classrooms that were identified as “low-performing” found teachers engaged in more or less adequate, but not excellent, teaching.¹⁰

Which brings us back to teacher norms. March and Olsen (2008) summarise their approach to a “logic of appropriateness”:

The simple behavioral proposition is that, most of the time humans take reasoned action by trying to answer three elementary questions: What kind of situation is this? What kind of person am I? and What does a person such as I do in a situation like this? (p. 690)¹¹

In some (many?) education systems, the situation is such that practices like buying exemption-from-duty certificates or mispresenting reported test scores are the routine, and hence are statistically “normal” practice. If one also assumes that most people do not regard themselves as the kind of person who consistently engages in “norm deviant” behaviour, then we have to acknowledge that many (most?) teachers in the situation of assessing student learning in Madhya Pradesh would reason that what is appropriate for people like them in situations like this would be conforming with the norm of “gaming” the assessment in various ways to improve the reported results. In contrast, the statistical norm for teachers in Vietnam is good teaching, and truly excellent teaching is common enough to be an aspirational norm.

This means that we have to expect that the expected impact of any particular policy or programme or “intervention” or education reform effort is going to be conditioned on whether we are in the “red” world or “green” world with respect to teacher norms. That is, to perhaps strain the metaphor, the “gravitational attraction” created by existing teacher norms may cause objects to fall in exactly opposite directions in different education systems.

And now circling back to the opening motivating fact that recently (but pre-COVID) two thirds of children completing their basic schooling do not acquire basic skills, we need to ask whether a mere “business as usual” or even “more of the same” continuation of the policy and programmatic efforts to improve the quality of teaching and learning is likely to work. On this, a very recent study by Le Nestour, Moscoviz, and Sandefur (2022) needs to give us pause.¹²  They use the DHS and MICS data across a large number of developing countries to track the long-term trend in a descriptive fact: “What is the likelihood an adult woman born in a given year who completed exactly 5 years of schooling can read a simple sentence (in any language of her choosing)?” (And I emphasise this is a descriptive fact and I am not super-imposing any casual interpretation of this fact, like that this fact represents “school quality”, but it is a fact in the sense that is the result of a clearly defined measurement). They find that in Sub-Saharan Africa this fell from 71.2 percent for women born in the 1950s to only 43.6 percent for women born in the 1990s (the most recent available). In contrast, in East Asia and Pacific this rose from 86.5 percent for women born in the 1950s to 89.1 percent for women born in the 1990s.

Comparing two particular countries for which we also have recent PISA results, Zambia and Vietnam, we see strikingly different trends. The percent of women who could read a sentence having completed exactly Grade 5 rose from 72 percent to 94.1 percent from the first to last available date in Vietnam. In contrast, this fell from 61.2 percent to 22.6 percent in Zambia. This means that most of the massive 71.5 percentage point difference in the most recent cohorts (94.1-22.6) is the result of a gap of only 10.8 percent growing over time due to the very different dynamics.

This is ominous as it means that nothing that has happened in Zambia over the last 50 years: none of the education reforms, none of the education projects, none of the innovations, none of the changes in teacher training, none of the global technological progress has prevented a continued deterioration in this measure of learning. (And the simplistic dismissal of the deterioration of this learning measure as a “natural” or “inevitable” consequence of the expansion of a system towards universal enrolment is belied by the fact that the fraction of women with Grade 5 or higher complete expanded by very similar amounts in East Asia and SSA—34.5 percentage points in East Asia versus 39.5 in SSA—and the trends in enrolment expansion and the measure of learning are completely uncorrelated across countries).

As one wag’s definition of “crazy” is doing the same thing again and again and expecting different results, expecting that a long-term trend of stagnation or deterioration in learning outcomes is going to turn around just as a result of more of the same types of education reforms, policies, programmes, and projects is, well, not a good bet.

The role of the foreword is not to summarise or synthesise or say in other words what this volume says. Rather, its modest role is to prod the reader forward to read the book. So: read this book, as the future of humanity depends on accelerating learning progress in the developing world and that in turn almost certainly will require a massive shift in teacher norms. This book from introduction to the body of interviews to the synthesis essays is a great place to start the thinking about how that might just happen.

To access the rest of the book, please download the full book text (PDF).