--- - |2+ Objective Attribution of responsibility for government performance in a federalist system is challenging but necessary for democratic accountability. We identify competing hypotheses for how attributions for government performance might arise in social media and test our expectations using data drawn from Twitter following Hurricane Sandy. Method We use a novel linguistic approach to measure blame attributions in text and compare patterns of blame attributions toward multiple levels of government over time. Results Social media blame attributions emerge at the outset of the storm and are more likely to center on federal actors, followed by local actors. State actors received the least blame. We find similar patterns in retweets. Conclusion Our results suggest that social media privileges attributions that target broadly salient national political actors; however, social media accounts of disaster may make it easier for the public to assess performance of local and state government. - 'Social Science Quarterly, EarlyView. '