Using a large sample of individual credit information provided by a US credit bureau, this paper investigates the empirical relevance of stigma and information sharing on household bankruptcy and its trend. Many observers of bankruptcy patterns have conjectured that there exists an increased willingness to default that reflects a diminution of social stigma. In this paper, we use a new methodology to disentangle stigma and social learning-two acknowledgedly important social factors affecting default. Although our results indicate a large and important role for stigma, changes in information costs seem to be the more relevant factor in explaining the observed bankruptcy trends. Furthermore, we show that this aggregate trend disguises enormous heterogeneity. While social factors appear quite important among the very poor and less educated, stigma seems to have increased and information costs to have decreased among these very groups. On the contrary, we show that it is primarily among the relatively rich and well educated that stigma has declined. These compositional findings further suggest that the overall increase in the bankruptcy rates cannot be explained by a decrease in social stigma. We argue that the secular increase in bankruptcy is more likely attributable to decreased information costs rather than to changes in social stigma.