While media reports have been rife with allegations that the nation would have responded more strongly to victims of Hurricane Katrina if these victims had not come to a large extent from minority groups and economically disadvantaged backgrounds, so far there is little evidence to rebut or
support these allegations. To examine this issue, we conduct a randomized experiment on about 2000 subjects with nationally representative demographic characteristics. The subjects take an internet-based survey instrument that measures their generosity towards Katrina victims in a small town by letting them divide $100 (implemented with real money in 10% of cases) between themselves and a charity helping Katrina victims in that town. Prior to measuring their generosity towards Katrina victims, we provide information about Katrina victims in that town in the form of a photo presentation with accompanying audio. After the giving experiment, we ask about their perceptions of Katrina victims in that town. We experimentally vary the race of the victims shown in the pictures as well as some elements of the audio text. This allows us to measure the causal effect of perceived recipient characteristics on giving. Finally, we ask a number of follow-up survey questions that provide us with suggestive evidence on the mechanisms at work.