We analyze how relative wage movements among birth cohorts and educati
on groups affected the distribution of household consumption and econo
mic welfare. Our empirical work draws on the best available cross-sect
ional data sets to construct synthetic panel data on U.S. consumption,
labor supply, and wages during the 1980s. We find that low-frequency
movements in the cohort-education structure of pretax hourly wages amo
ng men drove large changes in the distribution of household consumptio
n. The results constitute a spectacular failure of between-group consu
mption insurance, a failure not explained by existing theories of info
rmationally constrained optimal consumption behavior. A welfare analys
is indicates that the cost of between-group consumption variability is
larger than the cost of aggregate consumption variability by two orde
rs of magnitude.