LAUSR

National agricultural investment plans in developing countries are expected to be evidence-based, reflect broad development processes, and measure contributions to high-level outcomes, such as economic growth and poverty reduction. We propose an economywide systems-approach that combines ex post household econometric analysis (using propensity score matching) of investment impacts with ex ante modeling of growth and poverty linkages (using a spatially-disaggregated dynamic computable general equilibrium model). We apply this approach retrospectively to Mozambique. Simulation results indicate that the country's investment plan from 2012 to 2017 would not achieve national growth targets, despite doubling public spending on agriculture. Rather than increasing spending, the government should have reallocated resources towards agricultural research and extension, instead of irrigation and fertilizer subsidies. Providing extension services to smallholders is most effective at raising growth and reducing poverty in all regions of the country. Investing in irrigation was more likely benefit growth in the country's southern region due to less favorable agroecological conditions. These conclusions are robust across assumptions about investment efficiency. As demonstrated in Mozambique, our approach provides a consistent framework for evaluating ex ante sector-wide agricultural investment plans based on growth, poverty and regional equity considerations. Our approach complements household-level evaluations by enhancing their relevance for national planning.