Welfare persistence is estimated and compared between Swedish-born and foreign-born households during the 1990s. This is done within the framework of a dynamic discrete choice model controlling for the initial condition and permanent unobserved heterogeneity. We control for three types of persistence in terms of observed and unobserved heterogeneity, serial correlation, and structural state dependence, the focus being on the latter measure. The results show that state dependence in Swedish welfare participation was strong. This effect was three times as large for the foreign-born compared to Swedish-born, but when this effect is distributed over time, it disappears after three years for both groups. Contrary to previous studies, our results for foreign-born are that both country of origin and time in the country of destination have only small impacts on welfare participation.