Religion and educational mobility in Africa

The paper investigates how intergenerational mobility in education varies across religious groups—Christians, Muslims, and adherents of Traditional (folk) religions—throughout postcolonial Africa. Motivated by persistent and sizable education gaps observed since independence and the continent’s high religiosity, the study asks: (1) how large and persistent are faith-specific mobility gaps; (2) to what extent do these gaps reflect household characteristics, economic conditions, initial literacy differences, and regional features; and (3) what roles do segregation and internal migration play? The context is critical because Africa hosts some of the world’s largest Christian and Muslim populations and is experiencing substantial educational reforms and demographic change. Understanding religion-specific educational mobility is important for equitable policy design and for interpreting broader development outcomes.

The study situates itself within several literatures. Historical accounts and policy reports have noted Christian–Muslim education gaps since the colonial period, linked to missionary activity and colonial investments. Prior quantitative work on Africa has emphasized ethnicity over religion, while research on religion and economic performance has focused mostly on the Middle East and Asia. Case studies in Nigeria (including among the Yoruba) suggest religion-related schooling differences, and parallel work (Platas and others) documents sizable Christian–Muslim gaps in 11 African countries, with larger gaps in predominantly Muslim areas and roles for religious leaders and social norms. Global and regional studies of intergenerational mobility (World Bank; Chetty et al.; Alesina et al. for Africa) have not deeply examined religion, though India-focused research highlights lagging mobility among Muslims relative to low-status castes’ improvements. The paper adds a comparative, multi-country African perspective, quantifies faith-specific mobility patterns, and explores geographic segregation and migration as mechanisms.

Data: The authors use IPUMS-International harmonized census microdata, drawing representative samples (typically 10%) from 21 African countries, covering 2,286 districts. Religious affiliation is harmonized into Muslims, Christians, Traditionalists (Animists), and auxiliary categories (Other, No religion; the latter two are excluded from analysis across countries due to comparability issues). The main sample includes 7,188,717 individuals aged 14–18 cohabiting with an older generation member, with an extended sample of 13,018,904 aged 14–25 (cohabitation ~75%).

Measures of intergenerational mobility (IM): Primary school completion is the benchmark. Upward IM is the probability that children aged 14–18 whose cohabiting elder(s) have not completed primary education manage to complete primary school. Downward IM is the probability that children aged 14–18 whose elder(s) completed primary fail to do so.

Country- and district-level patterns: The authors compute faith-specific IM rates by country and district (administrative levels 2–3). They map Christian–Muslim and Christian–Traditionalist gaps and summarize within-country heterogeneity.

Explaining gaps—individual-level regressions: Linear probability models (OLS) relate children’s IM to indicators for Muslim and Traditionalist (Christians omitted), conditioning sequentially on: (1) country-by-birth cohort fixed effects and age; (2) detailed household structure and size (relationship to head; presence of parents; age of parents at birth; generational counts); (3) economic features (urban/rural; industry—six categories; occupation—ten categories—of household head); (4) district-by-religion older-generation literacy shares (completed primary), capturing initial group-specific local conditions; (5) district fixed effects interacted with urban status. Estimates are population-weighted to account for IPUMS coverage differences. Country-specific estimates and robustness checks are reported in extended/supplementary materials.

Childhood exposure design (movers): To disentangle regional childhood exposure from spatial sorting, the authors follow Chetty–Hendren-style age-at-move designs using 276,686 individuals aged 14–25 from 13 countries with age-of-move information. They relate the child’s primary completion (for those with uneducated elders) to the difference in upward IM between destination and origin districts among non-movers of the same cohort, interacting with the child’s age at move. The age-specific slope on destination–origin IM differences identifies exposure effects: earlier moves to higher-mobility districts should yield larger gains. They estimate versions using IM differences among all non-movers and among same-religion non-movers, and also within-household (sibling) variants.

District correlates: Within-country regressions relate district-level Christian–Muslim and Christian–Traditionalist gaps in upward IM to regional characteristics: (1) development at independence (population density, urban share, sectoral composition); (2) geography and location (distance to capital, coast, border, malaria ecology, terrain, resources, railroads); (3) historical factors (mission locations, precolonial and colonial state presence); (4) religious composition and fragmentation (own-religion shares; Herfindahl). Standardized coefficients are reported, with and without conditioning on older-generation literacy by religion.

Internal migration analysis: The authors compute internal migration rates (residing outside birthplace district at census) by religion across 20 countries (Nigeria excluded due to data), both in levels and de-meaned by birthplace to net out origin composition. They relate migration propensities to own-religion population shares in the birthplace, within countries, to examine segregation–migration links.

Interpretation: The analysis is descriptive and correlational. The authors do not claim causal identification for most relationships, except for exposure patterns inferred from age-at-move designs under the assumption that timing of moves is unrelated to children’s unobserved ability.

• Christians have the highest upward IM in 15 of 21 countries (cross-country mean ~41.2%). Traditionalists generally have the lowest upward IM (e.g., <10% in Burkina Faso, Sierra Leone, Rwanda, Malawi, Uganda, Ethiopia).
• In Nigeria, upward IM: Christians 78.6% vs Muslims 46.6% (gap 32 pp). In Ethiopia: Christians 13.8% vs Muslims 8.2%. In West Africa, mean Christian–Muslim upward IM gap ≈ 22.1 pp.
• Downward IM is higher for Muslims (mean 27.5%) and Traditionalists (mean 42.6%) than for Christians. Examples: Cameroon downward IM 4.1% (Christians) vs 19.6% (Muslims); Nigeria 7.8% (Christians) vs 16.2% (Muslims).
• Within districts, Christians still outperform: median district upward IM 0.44 (Christians), 0.33 (Muslims), 0.21 (Traditionalists); downward IM 0.25 (Christians), 0.29 (Muslims), 0.35 (Traditionalists).
• Baseline regressions (conditioning on country–cohort and age): Christian advantage over Muslims is ~16 pp in upward IM and Muslims are ~7 pp more likely to experience downward IM; Christian advantage over Traditionalists is ~20 pp upward IM (downward IM gap smaller than for Muslims).
• Household size/structure controls reduce gaps slightly (Muslim upward gap ~0.165 to ~0.15; Traditionalists ~0.20 to ~0.18). Economic features (urban/rural, industry, occupation) reduce Christian–Traditionalist gaps but do not explain Christian–Muslim upward gap.
• Controlling for district-by-religion older-generation literacy substantially attenuates gaps (Christian–Muslim upward gap ~0.15 to ~0.085; downward gap ~0.07 to ~0.04). Adding district fixed effects (with urban interaction) further reduces gaps to ~0.05 (upward) and ~0.03 (downward). Gaps remain even in the most comparable half of districts.
• Movers’ exposure effects: Early-life moves to higher-mobility districts increase primary completion probabilities by about 2 percentage points per additional year of exposure before age 12, for both Christians and Muslims (similar results using own-religion IM differences and within-household designs). Exposure slopes diminish with age and are positive after age 13–14 (interpreted as selection).
• Segregation: Christian–Muslim upward IM gaps are strongly positively associated with the Muslim share in a district; Muslims underperform where they are majorities, even conditioning on elders’ literacy. Christian upward IM is roughly flat across own-religion shares; Muslim upward IM declines with own-religion share.
• Residence patterns: Muslims and Traditionalists live in more remote, less urbanized districts, farther from capitals and coasts, with fewer colonial investments and missions, and lower initial development.
• Internal migration: Christians have higher internal migration rates than Muslims in 17 of 20 countries (pooled means: Christians 0.298; Muslims 0.222; Traditionalists 0.194). Within-birthplace comparisons still show Christians 3.6 pp more likely to migrate than Muslims. Muslim migration is especially low in predominantly Muslim districts; the negative relationship between migration propensity and own-religion share is three times stronger for Muslims (slope ≈ -0.33) than for Christians (≈ -0.12).

The findings show that religion-specific educational mobility disparities in Africa are large and persistent, even when comparing youth within the same district and similar household backgrounds. Regional characteristics—captured by district fixed effects and religion-specific literacy among elders—explain roughly two-thirds of the gaps, underscoring the centrality of place. The movers’ design demonstrates that both Muslim and Christian children benefit equally from early-life exposure to higher-mobility regions, which points away from inherent differences in returns to locality by faith and toward constraints on accessing opportunity. Lower internal migration among Muslims, especially in Muslim-majority (segregated) areas that are less urbanized and more remote with limited infrastructure, appears to amplify initial educational disadvantages. Furthermore, segregation is closely linked to larger Christian–Muslim mobility gaps: Christians’ upward mobility is relatively insensitive to own-religion shares, whereas Muslims’ upward mobility declines in Muslim-dominated districts, consistent with underinvestment equilibria in segregated settings. These results highlight the importance of spatial opportunity, integration, and mobility for reducing interfaith educational inequalities and suggest that policy effectiveness may be heterogeneous across religious contexts due to differential take-up and geographic constraints.

This paper compiles the first continent-wide, religion-specific measures of educational intergenerational mobility across 21 African countries and 2,286 districts and documents sizable and persistent gaps: Christians generally exhibit higher upward mobility than Muslims and Traditionalists. Regional initial conditions and current place effects explain much of the variation, yet residual gaps remain within regions and household strata. Both Christians and Muslims gain similarly from early exposure to high-mobility regions, but Muslims’ lower internal migration—particularly in segregated, Muslim-majority areas—exacerbates disadvantages. The study underscores the need for policies that account for religious segregation, improve access to quality schooling in remote areas, and reduce barriers to migration. Future research should: (1) examine within-faith heterogeneity (Islamic jurisprudence traditions; Christian denominations); (2) measure learning quality and faith-specific private and social returns to schooling; (3) analyze heterogeneous economic returns to migration by religion; and (4) evaluate the differential take-up of education policies in religiously segregated communities using georeferenced school and learning data.

• Identification: Most analyses are descriptive/correlational; OLS regressions with controls do not establish causal effects (except suggestive exposure patterns relying on timing-of-move exogeneity).
• Measurement: IM focuses on primary school completion and does not capture learning quality; some Traditionalist categories blend heterogeneous local religions; religion coding varies across censuses (Other/No religion excluded).
• Sample and coverage: Age-of-move data are available in only 13 countries; internal migration analysis excludes Nigeria; cohabitation-based matching may miss non-cohabiting parents/guardians.
• Controls and aggregation: Economic categories (industry/occupation) are coarse; unobserved cultural or institutional mechanisms are not directly measured; regional proxies for historical investments and geography may be imperfect.
• Generalizability: Results pertain to the countries and districts with available censuses and may not capture all African contexts or time periods.