What Averages Hide About Who Gets Left Behind

Why the choice of learning indicator changes how large — or how small — Africa's wealth gaps in learning appear

Household wealth is one of the strongest predictors of educational outcomes anywhere in the world, and Africa is no exception. Children from wealthier households are more likely to be enrolled in school, more likely to stay enrolled, and more likely to learn once they get there.

None of this is surprising. What is less obvious is how much the size of that gap depends on which indicator is used to measure it.

Two ways of measuring the same gap

Our report compares mathematics performance among children from the poorest and richest 20% of households across eighteen African countries, using two measures side by side: the proportion of children reaching each programme's minimum proficiency benchmark, and the average percentage of assessment items answered correctly.

Both measures tell the same broad story — children from wealthier households outperform children from poorer households in every country examined. They do not, however, agree on how large that gap actually is.

Proficiency benchmarks can mask substantial wealth inequalities in learning graph

Where overall learning is low, benchmarks understate inequality

In the Democratic Republic of the Congo and Sierra Leone, the proficiency-benchmark comparison suggests wealth barely matters: virtually no child from the poorest households reaches the benchmark, and only around one percent of children from the richest households do either. Two numbers close to zero look, at a glance, like a small gap.

Average scores tell a different story. Children from the richest households in both countries score consistently higher across the full assessment than children from the poorest households — a gap the benchmark comparison hides almost entirely, simply because so few children from either wealth group cross the threshold.

The lesson is not that either measure is wrong. It is that a single summary statistic can make substantial inequality look negligible, depending on where an education system happens to sit on the learning curve.

The scale of the gap, where it is largest

In other countries the wealth gap is stark under either measure. In Nigeria, children from the richest households answer an average of 90% of assessment items correctly, compared with 23% among children from the poorest households — a gap of 67 percentage points. Even in the countries with the smallest measured gaps — Eswatini, São Tomé and Príncipe, and Tunisia — children from the richest households still outperform those from the poorest by around 13 percentage points.

Wealth shapes who is in the room before it shapes what they know

Learning gaps are only part of the picture. Household assessment data also make it possible to see that wealth shapes school enrolment itself, long before any test is administered.

Wealth gaps in school participation remain substantial in many countries graph

In Chad, only 22% of children from the poorest households are enrolled in school, compared with 72% of children from the richest households. In Nigeria, the gap is 44% versus 96%. In several countries, enrolment among the richest children is close to universal while enrolment among the poorest remains far from it.

Because out-of-school children score substantially lower than enrolled children on every learning measure — as the previous AFLEARN Insight showed — part of the wealth gap visible in school-based data may already reflect who is missing from the room, not only what happens once children are seated at a desk.

Neither measure is wrong — but neither is complete alone

Programme-specific proficiency benchmarks answer an important question: have children reached an agreed minimum standard? Average scores answer a different one: where do children sit along the full continuum of learning? Enrolment data answer a third: who gets the opportunity to learn in the first place?

Examining all three together — rather than relying on any single indicator — gives a far more complete picture of educational inequality than any one of them can provide alone.

Why this matters for policy

Wealth-related inequality in learning is not a new finding. What open microdata add is precision: the ability to see not just that gaps exist, but how large they are, where they emerge, and whether the indicator being used to report them is masking or revealing their true scale.

For governments and donors allocating scarce resources, that precision matters. A gap that looks negligible under one indicator and severe under another calls for very different policy responses — and very different levels of urgency.

AFLEARN Insight #1 AFLEARN Insight #2 AFLEARN Insight #3 AFLEARN Insight #4 Read the report