This paper documents the vulnerability and exposure to flood risk with fine-grain data for the universe of homeowners, landlords and tenants in France. Using data at the dwelling level, I distinguish the exposure of owner-occupied, rented, second, and vacant homes, and link these properties to the disposable income and housing wealth of their owners and tenants. Vulnerability to floods exhibits large heterogeneity. Tenants and homeowners owning only their primary residence are highly vulnerable due to limited housing wealth, whereas most landlords diversify their assets, exposing only a fraction of their housing wealth to flood risks. In addition, I find that high-income homeowners and tenants are underrepresented in flood-prone areas. As a result, the income of homeowners is 6% lower in risky areas, and the income of tenants is 9% lower. On the contrary, second homes owned by middle and high-income households are overrepresented in flood-exposed areas, with a twofold higher likelihood. These patterns of residential distribution appear to be driven by high-income homeowners and tenants actively avoiding flood risks and by an amenity effect influencing second home owners. I show that these findings matter for adaptation policies, studying two types of policies. (1) The cost of buying-back homes at risk could be divided by five if it focused on low-income homeowners. (2) A subsidized insurance policy such as the current French CatNat system redistributes money away from low-income households to owners of second homes through a regressive tax.