In this study we conduct an impact evaluation of a complex rural development project in Central America with multiple treatments taking place simultaneously, purposive program placement and project participant freedom to opt in to project interventions. For this purpose we use propensity score matching difference-in-differences estimation, and compare results of this method with weighted propensity score regression and simple difference-in-differences estimation. We find short term project impacts in household savings, in participation in groups and associations, and in reduction of stored grain losses. However, we find no project impacts in long-term outcomes associated with increased agricultural income or household asset accumulation. These results are not surprising, since the project evaluation was conducted two years into a three-year project, before beneficiaries had realized its full benefits. Our study calls attention to the need of more research on linking short term to long-term impacts and on longer term strategies to evaluate impacts of agricultural technology.