Preferences for redistribution and social spending are correlated with income and unemployment risk, but it is unclear how these relationships come about. I build a theory emphasizing that only large changes in economic circumstances provide the information and motivation needed for people to change their preferences. Stable long-run preferences are shaped mainly by early socialization, which includes economic and ideological influences from the family, and early labor market experiences. Enduring shocks, low intergenerational mobility and the tendency of left-wing parents to be poorer generate correlations between circumstances and preferences. Because preferences are stable, greater inequality may not increase aggregate support for redistribution. Support is found for the theory with panel data from Switzerland, using a range of empirical tests.