Prioritizing public investments requires information on relative returns that are difficult to derive from disparate evaluation studies. This paper presents a 'hybrid' approach that combines ex post evaluation data with an economy-wide model for experimenting ex ante with alternative investment portfolios within a consistent, structural framework. The approach is used to evaluate rural investments in Uganda. Agricultural research and extension services are found to be much more effective at promoting economic growth and poverty reduction than either rural feeder roads or irrigation infrastructure. This suggests that the government's recent shift in emphasis from extension services to irrigation is potentially misguided.