How to target social program benefits to the poorest households in the developing world remains a key question for policymakers. This work examines community-based targeting (CBT), a targeting method in which local community members identify poor households to receive program benefits. The first chapter explores issues that arise in assessing the accuracy of community-based targeting exercises, especially for the purposes of comparison with more traditional proxy-based targeting methods. I propose the index of the marginal utility of expenditures (MUE) developed in Ligon (2020) as a potential welfare benchmark to use in such comparisons of accuracy. I show that unlike with accuracy comparisons using a standard consumption benchmark, community-based targeting may yield more accurate targeting outcomes than proxy-based methods under an MUE benchmark. The second chapter uses a set of lab-in-the-field experiments in Central Java, Indonesia to better understand the types of welfare information about others that community members have and use when making community-based targeting decisions. When asked to provide information on specific welfare benchmarks, community members appear to have little dynamic, current welfare information about others in their community. The results suggest that they instead rely on more static welfare attributes (such as asset ownership and demographic characteristics) when making targeting decisions. Finally, the third chapter investigates whether there are accuracy gains associated with joint community-based targeting procedures in which multiple community members make targeting decisions collaboratively, as opposed procedures where individuals make similar targeting decisions by themselves. Building on the lab-in-the field exercises from the previous chapter, I compare observations of the same individuals making joint and independent targeting decisions. There seem to be fairly negligible accuracy gains from convening a group targeting exercise as opposed to having the average community member make targeting decisions.