On the Covariance Structure of Changes in Consumption in the Health and Retirement Study

Using generalized method of moments on the covariance matrix, I test three models of consumption change on constructed consumption data from the Health and Retirement Study. Meant as a first step towards estimating life-cycle effects of subjective survival probabilities on consumption profiles, this study finds that the model that best describes the consumption data is a pure measurement error model. This is likely due to the large amount of error introduced in the process of inferring consumption from other financial data. This result casts significant doubt on the use of this data in estimating a life-cycle model. (Note: this paper duplicates some expository material from Perry (2005)