In this paper we model the evolution of income risk and consumption growth. We decompose the time series innovation of the income process into its common and cohort-specific components. From these we compute conditional variances which are used as separate risk terms in a consumption growth equation. Using a long series of British household data we exploit the time-series variation to identify precautionary saving effects and find strong evidence of their importance. Specifically, after allowing for demographic and labour market status, there is an independent role for income risk in explaining consumption growth. Rather than the component that is common across cohorts, however, it is the cohort-specific element that is important in determining changes in consumption growth. This result points to a failure of between-cohort insurance mechanisms.