The methods to select safety net beneficiaries are the subject of frequent policy debates. This paper presents the results from a randomized experiment analyzing how efficiency, legitimacy, and short-term program effectiveness vary across widely used targeting methods. The experiment was embedded in the roll-out of a national cash transfer program in Niger. Eligible villages were randomly assigned to have beneficiary households selected through community-based targeting, a proxy-means test, or a formula designed to identify the food-insecure. Proxy-means testing is found to outperform other methods in identifying households with lower consumption per capita. The methods perform similarly against other welfare benchmarks. Legitimacy is high across all methods, but local populations have a slight preference for formula-based approaches. Manipulation and information imperfections are found to affect community-based targeting, although triangulation across multiple selection committees mitigates the related risks. Finally, short-term program impacts on food security are largest among households selected by proxy-means testing. Overall, the differences in performance across targeting methods are small relative to the overall level of exclusion stemming from limited funding for social programs.