Using unique data collected from October to December 2012, we estimate the link between commuting to and from work and the level of household exposure to floods. The result suggests an empirical puzzle - individuals affected by only one flood are roughly 10% more likely to engage in the commuting activity, whereas households affected by two floods are 13% less likely to do so. We check the robustness of this result by operationalizing the past exposure to floods with variables that describe the geographical location of the house and its characteristics. We explain the puzzle by the fact that individuals commute to work in order to accumulate resources to decrease the household's vulnerability to flood risk, amongst other reasons. When the flood risk is high, some households out-migrate, and stayers commute less, probably, for similar reasons as why they stay. Further, we find evidence in support of the "network effect" hypothesis - an individual with an active commuter in the household is by 47% more likely to commence commuting. We also find that flood affected commuters travel shorter distance for work.