Abstract
Gender parity in public programme participation is often interpreted as evidence of equity, yet similar participation rates may conceal unequal access. This study asks whether gender disparities in Zambia’s agricultural subsidy system arise mainly at the point of programme entry or through differences in welfare returns after participation. Using microdata from an agricultural subsidy programme in Mungwi district, the study combines Fairlie nonlinear decomposition with inverse probability weighting to separate access mechanisms from treatment effects. Nonlinear decomposition indicates an estimated 8.4 percentage-point participation gap favouring male-headed households, of which roughly 74% cannot be explained by observable characteristics. In contrast, inverse probability weighting estimates further suggest that participation improves food security by approximately 18.6 percentage points, with no statistically significant gender difference in returns between male- and female-headed households. The results suggest that inequality operates primarily through access mechanisms rather than differences in programme benefits once access is secured. Evaluations that focus solely on treatment effects may therefore overlook disparities that occur at the programme entry stage. Policy efforts should therefore strengthen transparency in beneficiary selection, verification, and grievance mechanisms, rather than relying only on average treatment effects to assess programme equity.