Program evaluation of agricultural input subsidies in Malawi using treatment effects: Methods and practicability based on propensity scores

Several evaluations have been conducted to assess the impact of agricultural input subsidies in Malawi but have been mostly either descriptive or qualitatively inferred of the intervention impacts on the overall goal of the subsidy program. In most studies cited in this paper their approaches do not control for misspecification errors that might arise due to selection bias. One common erroneous approach is the lack of controlling for treatment effects. In this study we employ quasi-experimental econometric techniques using propensity scores to control for selection bias by creating control groups for those individuals that benefit from agricultural input subsidies. The study utilizes raw household data from two surveys conducted through the Malawi National Statistical Office in 2004/05 and 2006/07 production seasons. A household model for each dataset is estimated together with Average Treatment Effects on the Treated to assess the impact of targeted fertilizer input subsidies in 2004/05 and a refined program adopted in 2006/07 production periods. The evidence suggest that the starter pack or targeted input program implemented before 2004/05 focusing on one tenth of a hectare had a significant negative impact on household food expenditures compared to the refined program in 2006/07 that targeted about half a hectare for marginalized smallholder farmers. The latter, though portraying mostly insignificant results, showed positive impacts on household food expenditures. The approach adopted also proposes ways in which policy makers can effectively and independently evaluate the impact of public programs on social and economic welfare.