Female educational inequality and household cooking fuel choice in Ghana: evidence from the 2022 Demographic and Health Survey

Technical Summary: Female Educational Inequality and Household Cooking Fuel Choice in Ghana

Problem Statement
Despite global efforts to transition to clean cooking energy, access remains limited in Ghana, where over 80% of households rely on polluting solid fuels (wood and charcoal). This reliance imposes severe health, gender, and environmental costs, particularly on women who bear the brunt of indoor air pollution. While existing literature establishes that higher individual education levels correlate with clean fuel adoption, a critical gap remains regarding the role of educational inequality. Specifically, it is unknown whether the distribution of women's educational attainment within a community—distinct from the average level of attainment—acts as an independent barrier to the adoption of clean cooking technologies. This study addresses whether female educational inequality is associated with household cooking fuel choice in Ghana and whether this association can be disentangled from the effects of low average education and regional structural factors.

Methodology
The study utilizes data from the 2022 Ghana Demographic and Health Survey (GDHS), focusing on a nationally representative sample of 8,557 married or cohabiting women across 618 enumeration-area clusters.

• Dependent Variable: A binary indicator of whether a household's main cooking fuel is "clean" (electricity, solar, LPG, piped natural gas, biogas, or alcohol/ethanol) versus "traditional" (wood, charcoal, crop residue, dung, etc.).
• Key Explanatory Variables: Two community-level measures of female educational inequality are constructed using the single years of schooling (0–18) for all women aged 15–49 in each cluster:
  1. The Gini coefficient of education.
  2. The Theil index (GE(1)) of education, which is particularly sensitive to deprivation in the lower tail of the distribution.
• Empirical Strategy: The authors employ survey-weighted logistic regression with clustering at the primary sampling unit (PSU) level to account for the complex sampling design. The estimation follows a sequential specification strategy:
  1. Model 1: Unconditional association between inequality and fuel choice.
  2. Models 2–13: Progressive addition of control variables, including household wealth quintiles, headship characteristics, household size, rural/urban residence, spousal education and employment, the woman's employment, electricity access, financial inclusion, and age groups.
  3. Robustness Checks: The analysis is replicated using the Theil index. Heterogeneity is tested by splitting the sample into rural and urban subsamples. Crucially, the study tests whether the inequality association is distinct from the level of education by adding cluster mean schooling and regional fixed effects (16 regions) to the full specification.
• Diagnostics: Multicollinearity is assessed via Variance Inflation Factors (VIFs), and model fit is evaluated using McFadden's pseudo-$R^2$ and classification accuracy.

Key Results

1. Inequality and Fuel Choice: There is a strong, negative association between female educational inequality and the use of clean cooking fuels. In the fully adjusted model (Model 13), households in communities with maximal female educational inequality have approximately 78% lower odds of using clean fuels compared to those in communities with perfect equality, holding wealth and demographics constant. This result is robust across both the Gini and Theil indices and persists across thirteen progressively adjusted specifications.
2. Urban-Rural Heterogeneity: The association is significantly stronger in urban areas (OR = 0.055) than in rural areas (OR = 0.265). The authors suggest that in urban settings, where clean fuels are physically available, the community's educational environment becomes the primary differentiator of adoption. In rural areas, supply and income constraints bind for nearly all households, compressing the marginal impact of educational disparities.
3. The "Level vs. Distribution" Identification: When cluster mean schooling is introduced into the model, the statistical significance of both inequality indices disappears. Similarly, the inclusion of regional fixed effects absorbs the inequality association. This indicates that in the Ghanaian context, where average attainment is low, the distribution of education and the level of education are empirically inseparable. High inequality is mechanically tied to low average attainment; thus, the observed effect is best interpreted as a composite of community-level female educational disadvantage rather than a distinct effect of inequality alone.
4. Other Determinants: Household wealth remains the dominant predictor, with the richest quintile showing odds of clean fuel use orders of magnitude higher than the poorest. Household size and the age of the household head are negatively associated with clean fuel use, while the husband's higher education significantly increases the odds of adoption.

Key Contributions

• Measurement Innovation: The study is among the first to apply distribution-sensitive measures (Gini and Theil indices) of female education to household energy choices, moving beyond the standard focus on average attainment levels.
• Theoretical Refinement: By testing whether the inequality effect survives the introduction of mean schooling and regional controls, the paper provides a nuanced understanding of the education-energy nexus. It argues that in low-attainment settings, "inequality" and "low level" are facets of the same construct: community-level educational disadvantage.
• Policy-Relevant Evidence: The study demonstrates that the uneven progress in girls' education and the stalled transition away from biomass are not separate policy problems but are deeply interconnected.

Significance and Claims
The paper claims that expanding and equalizing women's education, particularly in educationally deprived communities, is complementary to clean-energy strategies for advancing the Sustainable Development Goals (SDGs). The authors caution against interpreting their findings as a causal effect of inequality per se due to the cross-sectional design and the identification issue between level and distribution. Instead, they posit that the composite construct of community-level female educational disadvantage is the policy-relevant object.

The study concludes that clean cooking policy and girls' education policy should be treated as complements. Targeting districts with high educational disadvantage with combined interventions (e.g., retention support for girls alongside clean fuel promotion) is likely to be more effective than isolated approaches. Furthermore, the findings suggest that awareness-based interventions may have the greatest marginal impact in urban and peri-urban communities where supply exists but adoption lags, whereas rural areas require supply-side expansion first. Finally, the paper highlights that widespread electricity access does not automatically translate to clean cooking, necessitating distinct policy tracks for cooking energy under SDG 7.