Measuring consumption smoothing in CEX data

A new method of measuring the degree of consumption smoothing is proposed and implemented using data from the Consumer Expenditure Survey. The structure of this Survey is such that estimators previously used in the literature are inconsistent, simply because income is measured annually and consumption is measured quarterly. An AR(1) structure is imposed on the income process to obtain a proxy for quarterly income through a projection on annual income. By construction, this proxy gives rise to a measurement error which is orthogonal to the proxy itself—as opposed to the unobserved regressor—leading to a consistent estimator. Our estimates are contrasted with the output of two estimators used in the literature. This comparison shows that while the first (OLS) estimator tends to overstate the degree of risk sharing, the second (IV) estimator grossly understates it.

Research Highlights► Standard estimators (OLS and IV) of the degree of consumption smoothing are biased in the CEX. ► The bias results from income being measured annually while consumption is measured quarterly. ► A proxy for quarterly income can be constructed by specifying an income process. ► By construction, that proxy gives rise to a consistent estimator.