Abstract We study the role of social identity in determining the impact of social fragmentation on public good provision using laboratory experiments. We find that as long as there is some degree of social fragmentation, increasing it leads to lower public good provision by majority group members. This is mainly because the share of those in the majority group who contribute fully to the public good diminishes with social fragmentation, while the share of free-riders is unchanged. This suggests social identity preferences drive our result, as opposed to self-interest. Importantly, we find no difference in contribution between homogeneous and maximally-fragmented treatments, reinforcing our finding that majority groups contribute most in the presence of some diversity.