Imputing consumption in the PSID using food demand estimates from the CEX

This paper assesses the accuracy of decomposing income risk into permanent and transitory components using income and consumption data. We develop a specific approximation to the optimal consumption growth rule and use Monte Carlo evidence to show that this approximation can provide a robust method for decomposing income risk. The availability of asset data enables the use of a more accurate approximation allowing for partial sef-insurance against permanent shocks. We show that the use of data on median asset holdings corrects much of the error in the simple approximation which assumes no self-insurance against permanent shocks.