Theoretical studies have shown that capital gains taxes in the housing market may create lock-in effects but so far no empirical evidence has been presented regarding the size of these effects. For a panel of Swedish house owners in 1984-1990, we show that lock-in effects only appear for households with income reductions; the size of these lock-in effects crucially depends on the magnitude of the income loss. The theoretical model and features of the Swedish tax system imply that lock-in effects depend on the degree of mismatch in the current residence and whether the households buy up or by down.