When managing the economy, governments make decisions that influence not only overall growth but also its distribution. How do voters judge incumbents for this? I revisit the idea of group-based retrospective voting and argue that voters assess the economic performance of their social in-groups relative to the national economy. By sanctioning the incumbent for in-group performance, voters can incentivize policy-making that better aligns with their interests. I test the theory, first, by estimating the relationship between in-group performance and incumbent support in panel data. This relationship is comparable in magnitude to sociotropic voting. I further conduct three experiments in Denmark and the United States, randomizing information about the performance of groups defined by geography, age, education, ethnicity and class. The findings suggest there are limits to sociotropic voting, as voters want their groups to follow or beat the national trend. This has important implications for electoral accountability and party competition.