Recent research on household `vulnerability' has led to an increased appreciation of the welfare costs of risk. Measuring the risk borne by a particular household has generally involved the use of panel data, and in particular the use of time series variation in household expenditures to estimate the risk borne by the household in any given period. This has led researchers to focus on static measures of vulnerability, since once used to identify the distribution of consumption expenditures in a single period the time series variation can no longer be used to describe the intertemporal profile of the distribution of consumption expenditures--simultaneous estimation of inequality, risk, and time series variation in household vulnerability requires the additional structure of a dynamic model. Unfortunately, our present understanding of the economic circumstances in which most households are situated seems too limited to permit general agreement on what the right dynamic model is. We show that simple restrictions on households' intertemporal smoothing can be used to simultaneously estimate household risk preferences in a manner which is robust to a variety of different assumptions about the economic environment. Further, these simple restrictions and estimated preferences can then be used to robustly characterize the welfare costs of different sorts of variation in consumption expenditures.