In this paper we analyze mortality caused by 2,194 large flood events between 1985 and 2008 in 108 countries. Unlike previous studies that looked at natural-disaster mortality, we find that year-to-year changes in income and institutional determinants of vulnerability do not affect flood mortality directly. Income and institutions influence mortality only indirectly, through their impact on the intensity and frequency of floods. Population exposure affects the number of deaths both directly and indirectly. Higher population exposure results in more deaths once the flood has occurred, but it is associated with smaller floods. In developing countries it also reduces the count of floods.