This paper evaluates the role of rural infrastructure on the performance of some agricultural crops in Colombia. The study utilizes geo-referenced cross-sectional data of four crops, coffee, rice, beans and plantains, collected for the majority of municipalities. Using genetic matching models, we find that both having access to irrigation and drainage systems and better infrastructure for marketing –rural roads and nearby retail and wholesale centers– significantly increase crop yield as well as planted and harvested areas. Results are robust to a suitable set of matching algorithms. The positive and significant impact on agricultural development provides support to reorient agricultural policy towards the supply of public goods that pushes up productivity.