Assessing learning without betraying the field

West Africa has won the battle for access, or nearly so, without winning the battle for learning. Across Sub-Saharan Africa, 89% of ten-year-olds are in learning poverty, meaning they cannot read and understand a simple, age-appropriate text (World Bank, UNESCO and UNICEF, 2022). In other words, almost an entire age cohort passes through school without acquiring its most elementary skill. That single figure shifts the debate: the question is no longer only how many children are enrolled, but what they actually know how to do when they leave. And you only govern well what you measure well. Measuring learning, and measuring it without betraying the field or the girls, has become the decisive bottleneck of the region's education policies.

Learning poverty: a blind spot in enrolment statistics

For two decades, education steering was dominated by access indicators: gross enrolment ratios, numbers enrolled, attendance rates. These figures accompanied real progress, as millions more children crossed the school threshold. But a child who is present is not a child who is learning. Learning poverty was designed precisely to make that gap visible: it combines the share of children who cannot read by the end of the cycle with the share of those who remain out of school. Globally, 70% of ten-year-olds in low- and middle-income countries are affected, up from 57% before the pandemic (World Bank, 2022). In Sub-Saharan Africa, the rate reaches 89%, the highest of any region in the world, ahead of Latin America (80%) and South Asia (78%). The contrast with often flattering enrolment rates is striking: it shows that a system can welcome almost all its children and teach them almost nothing about reading.

Measuring learning: what PASEC changes in Francophone Africa

Francophone Africa does have an instrument to face squarely what children are learning: PASEC, the CONFEMEN's programme for the analysis of education systems. Every five years, it assesses reading and mathematics skills among pupils at the start and end of primary schooling using a harmonised protocol, allowing countries to be compared and their progress tracked over time. The PASEC 2019 round covered fourteen Francophone Sub-Saharan African countries. Its reading is harsh: in most of the countries assessed, the majority of pupils fail to reach, by the end of primary, the "sufficient threshold" of reading competence, the level deemed necessary to continue schooling in good conditions. In Chad, 77.8% of pupils fall below that threshold; in Niger, 69.9%. By contrast, Benin and Senegal place a clear majority of their pupils above it, proof that the outcome is not predetermined.

The Benin case is worth dwelling on, because it shows what regular measurement makes it possible to observe. Between 2014 and 2019, the share of end-of-primary pupils at the sufficient threshold in reading rose from 51.7% to 75.7% in Benin, and from 39.8% to 51.7% in mathematics (CONFEMEN, PASEC). A gain of that magnitude in five years is no anecdote: it signals that a system can recover quickly when the effort is targeted. But above all, it could neither have been documented nor celebrated without an assessment instrument repeated identically. Without repeated measurement, progress remains a hunch, never a proof. And what is not proven cannot be defended before a finance ministry or a donor.

The comparison between countries is just as instructive, provided it is handled with care. PASEC places side by side systems that are geographically close but far apart in their results, and that dispersion is itself information: it forbids attributing school failure to some regional fate, since neighbours sharing comparable climate, exam language and level of wealth obtain radically different scores. The gap between Benin and Chad in reading, from one to more than triple, is explained neither by the pupils nor by the children themselves, but by what systems do, or fail to do, with their school time. This is precisely the virtue of a harmonised assessment: by neutralising the instrument, it isolates what stems from education policy. But this comparison only makes sense if one resists the temptation of simplistic ranking. A low national score does not say that a country is lazy; it says where to look, and invites a step down, toward the school, the classroom and the pupil, where the real cause can finally be grasped.

A school system can welcome almost all its children and teach them almost nothing about reading. Only a learning assessment, not an enrolment rate, can reveal it.

Why a poorly designed measure manufactures false certainties

Measuring learning is not a neutral act: a poorly designed test does not merely lack precision, it steers decisions in the wrong direction. Three methodological traps lie in wait for assessments conducted in the West African context, and each produces its own false certainty.

• The language of the test. Assessing in French a child whose first language is Fon, Dioula or Hausa risks measuring their command of French rather than their reasoning. A mathematics test phrased in a poorly mastered language systematically underestimates real learning, and makes what is first a linguistic obstacle look like a failure to learn. Multilingualism is not a logistical detail; it is a variable of validity.
• The conditions of administration. The same test given in a quiet room and in a noisy courtyard, by a trained enumerator or an improvised one, does not measure the same thing. Without a standardised protocol and trained enumerators, the gap observed between two schools may reflect the quality of data collection more than the quality of teaching. Field rigour is a condition of comparability.
• Premature aggregation. Reducing a school to an average erases the dispersion that holds the useful information. A respectable class average may conceal a group of pupils far ahead and a half falling behind. Whoever measures only the average never sees the pupil sinking, and wrongly concludes that all is well.

These three traps share one feature: they are invisible in the final result. A false figure looks exactly like a true one. That is what makes bad assessment so dangerous: it does not flag itself as such, it imposes itself as a truth and commits budgets. Designing the instrument with rigour is therefore not an academic refinement; it is the condition for data to serve the decision rather than mislead it.

Gender: the variable that averages most often erase

Of all the dimensions that aggregation conceals, gender is the most structuring and the most neglected. Girls and boys do not live the same school: domestic chores that eat into study time, safety on the way to school, family expectations, differentiated teacher attitudes, early marriage and pregnancy. These factors weigh on attendance, on classroom attention and thus on learning, but they remain invisible as long as results are not disaggregated by sex. At regional scale, the stakes are vast: Sub-Saharan Africa is home to more than half the world's out-of-school girls, and 23% of primary-school-age girls are out of school, against 18% of boys (UNESCO). But the gap is no longer decided only at the school gate.

For the most instructive figure is also the most counter-intuitive: in several West African countries, it is now girls who complete primary at a higher rate than boys. World Bank data confirm this for recent years: in Senegal, the primary completion rate reaches 71.7% for girls against 52.1% for boys; in Burkina Faso, 57.2% against 46.3%; in Côte d'Ivoire, 77.7% against 74.2%. Conversely, Benin still shows girls lagging (59.1% against 64.1% for boys), as do Niger and Mali, more mildly. The message is unambiguous: there is no single "West African gender gap", but opposite local configurations. A policy built on the assumption that "girls are always behind" would aim at the wrong target in half the countries.

This reversal calls for a cautious reading, and that is precisely where the quality of measurement matters. That girls complete primary more often does not mean they learn as much, nor that they continue into secondary: at the Sub-Saharan level, young women remain the most disadvantaged in secondary completion, and the gap widens across the cycles. A policy content with the primary completion rate would conclude that "the girls' problem is solved", at the very moment it shifts toward the quality of learning and the transition to secondary. This is the textbook case of a correct indicator that, misread, manufactures a false certainty. Parity of access is not parity of learning.

The framing: from measuring presence to measuring competence

The shift to make holds in one sentence: move from steering by inputs to steering by learning. Measuring presence answers the question "how many children are there?"; measuring competence answers the far more demanding question, "what can they do?". This change of focus has concrete implications for the craft of assessment. It requires validated instruments, that is, ones verified to measure what they claim to measure; standardised administration conditions, so that observed gaps reflect learning and not collection; sampling that covers rural and remote areas, where difficulty is often concentrated; and systematic disaggregation by sex, setting and language, without which the most politically useful gaps stay invisible. None of these requirements is optional: remove any one of them, and the resulting figure ceases to be a compass and becomes a mirage.

The cost of inaction: what a poorly measured generation costs

Failing to measure learning correctly is not a saving, it is a deferred bill. When a system does not know where its pupils fall behind, it allocates its means blindly: it trains teachers on skills already mastered, finances textbooks where the problem is the language of instruction, and leaves intact the pockets of difficulty it failed to locate. Each school year steered by guesswork produces a cohort of children who advance in age without advancing in skills, and whom the system believes, for lack of measurement, it is teaching properly. Learning poverty at 89% is precisely the accumulated balance of those years of insufficient assessment. The cost is then paid across a lifetime: a child who cannot read at ten is durably exposed to dropout, to lost future income and, for girls, to a heightened risk of early exit from the system. Here too, the real cost is not measuring, but failing to measure, and discovering it too late.

The teacher factor: a constraint that measurement must integrate

You cannot interpret learning without looking at the conditions in which it is built, starting with class size. In most West African countries, the pupil-teacher ratio in primary exceeds 35 to 40 to one (World Bank), against around twenty in the best-performing systems. An overcrowded class mechanically limits individual attention, penalises the most fragile pupils first, and amplifies the gaps that the assessment will later reveal. This is why a serious measure of learning never settles for a score: it relates it to the school context, class size, teacher qualification, language of instruction, textbook availability. Without that context, a poor result is a sterile observation; with it, it becomes an actionable diagnosis. Assessing the pupil without assessing the conditions of their schooling is like measuring the fever without seeking the cause.

What national averages hide

As with any aggregate indicator, the national learning figure is an average, and the average is the enemy of equity. Behind a national reading success rate lie considerable gaps: between a supervised urban school and an isolated rural one, between a child schooled in their first language and another confronted from the start with a second language, between the girls of one setting and the boys of another. The national score, useful for comparing countries, does not say where or for whom the system fails within a country. Yet that is where the effectiveness of a policy is decided: the same remediation budget does not produce the same effect whether it targets the school already near the threshold or the one far below it. Without fine measurement, disaggregated by area, by sex and by language, education investment proceeds blindly, and almost always funds what is visible rather than what matters.

This is the conviction that guides CRAD's work in the field of education: a learning assessment has value only if it is rigorous in design and faithful to the field in execution. That means validated instruments and, where relevant, ones adapted to the multilingualism of the classroom; enumerators trained to question pupils, teachers and head teachers without bias; strict ethical protocols, particularly with minors; and systematic disaggregation by sex that makes boy/girl gaps a result in their own right, not a line lost in an annex. This is the approach CRAD deployed for UNESCO-IIEP in primary schools across several departments of Benin, and it is the same rigour, attentive to gender, that it carried into the energy sector with the regional WOCEWA project. A measure designed this way betrays neither the complexity of the field nor the real situation of girls: it finally makes it legible for those who decide.

At bottom, West Africa's learning crisis is not first a pedagogical crisis: the region knows how to enrol, and some countries even know how to improve their results quickly, as Benin showed between 2014 and 2019. It is a crisis of measurement and steering. The systems that progress are those that dare to look at what their pupils truly know how to do, that disaggregate that view by sex and by territory, and that draw decisions from it rather than press releases. Assessing without betraying the field is exactly this: refusing the comfort of flattering averages to confront the useful truth, the one that makes it possible to act where it is needed, for the children who need it most.

← All insights