Across 54 African countries, national wealth barely predicts how much stunting falls

Stunting — a child too short for their age, the physical record of chronic undernutrition and repeated infection in the first years of life — is the indicator most often used to summarise whether a country's children are growing up healthy. A study published in BMJ Global Health on 28 August asks a question the usual cross-country comparisons cannot: not which countries have the least stunting, but which convert what they have into stunting reduction most effectively [s1].

The method, and what it is for

The study is a longitudinal panel analysis covering 54 African countries from 2000 to 2022, with 1,134 country-years [s1]. It uses stochastic frontier analysis, a technique borrowed from production economics, to estimate a "health production frontier" — the maximum achievable health production score, defined as 100 minus stunting prevalence — given a country's inputs [s1].

The inputs modelled were GDP per capita, dietary adequacy and maternal education, each entered as three-year weighted moving averages [s1]. Technical inefficiency — the distance between what a country achieved and the frontier its inputs should have allowed — was itself modelled as a function of lagged diarrhoea incidence, electrification, urbanisation and political stability [s1].

The value of a frontier method is that it separates two things that ordinary regression conflates: how much a country has, and how well it uses it. The authors call the resulting gap the "nutrition conversion gap" [s1].

What it found

Performance heterogeneity across the continent was substantial, and the inefficiency term dominated: the study reports high technical inefficiency (γ = 0.976, λ = 6.33), accounting for 97.6% of observed nutritional disparities [s1]. GDP elasticity was low (β = 0.042) [s1].

The authors' conclusion is stated bluntly: nutritional progress is "profoundly decoupled from national wealth," and policy should pivot from aggregate resource accumulation towards mitigating systemic bottlenecks — specifically infection control and electrification [s1]. They argue that strengthening health system efficiency through decentralised, community-led models is a more potent route to human capital development than macroeconomic expansion [s1].

How much weight the conclusion carries

The empirical finding — that between-country differences in stunting are poorly explained by between-country differences in income, given the other inputs — is robust to how you feel about the policy recommendation. A GDP elasticity of 0.042 means that a country growing richer, holding dietary adequacy and maternal education fixed, moves very little on this outcome [s1].

The policy recommendation is a further step, and a longer one. Two cautions apply.

First, the inefficiency determinants are the study's own modelling choice. Diarrhoea incidence, electrification, urbanisation and political stability were entered as explanators of inefficiency [s1]; the finding that infection control and electrification matter is partly a consequence of having selected them. Variables not entered cannot be found to matter.

Second, "inefficiency" in a frontier model is a residual. It is whatever distance from the frontier the model does not attribute to the measured inputs. That distance contains genuine differences in how well systems work, and it also contains measurement error, omitted variables, and the ordinary noise of comparing anthropometric survey data collected on different schedules across 54 countries. Attributing 97.6% of disparities to inefficiency [s1] is a statement about model structure as much as about governments.

What the frontier cannot see

There is a determinant of child stunting that operates on a timescale no annual panel captures. A study published earlier this year pooled Demographic and Health Survey data on 91,840 children aged 6–59 months and their mothers across 19 sub-Saharan African countries, collected between 2013 and 2024 [s2]. Overall child stunting prevalence was 31.5%. Among children of stunted mothers it was 66.6%, against 32.3% among children of non-stunted mothers — an adjusted odds ratio of 3.79 (95% CI 3.25–4.43) [s2]. Regional prevalence ranged from 37.8% in Central Africa to 26.6% in Western Africa [s2]. Maternal education, household wealth and place of residence were also associated with child stunting [s2].

A mother who was stunted as a child was stunted twenty or thirty years before the panel begins. That inheritance sits inside the residual of any contemporaneous production function, and it is not inefficiency in any sense a health ministry can act on this year. The two studies are complementary rather than contradictory: one says today's spending converts poorly into today's outcomes, the other says part of today's outcome was determined a generation ago.

What to watch

Africa remains off track for the 2030 stunting targets, which is the premise the frontier study opens with [s1]. The checkable claim in its conclusion is narrow and testable: that countries improving infection control and electrification move toward the frontier faster than countries improving income. That is a prediction, and the next round of survey data will bear on it.

Sources

• Performance in reducing child stunting in Africa: a stochastic frontier analysis of 54 countries over the past two decades (2000-2022), BMJ Global Health, 28 August 2026
• Breaking the Cycle: Intergenerational Transmission of Stunting in Sub-Saharan Africa—Evidence From 19 Countries, Maternal & Child Nutrition, January 2026