Who Are the Hand-to-Mouth?

The literature has singled out households with little net wealth and/or little liquid assets as having high marginal propensities to consume (MPC’s) or, in the extreme, behaving hand-to-mouth (e.g., Zeldes, 1989, Kaplan, Violante, and Weidner, 2014). We show that standard models predict that high MPC households will display higher expected growth in consumption and, conditional on income, less volatile spending growth. (For a household literally hand-to-mouth, does not deviate from income.) But data on household expenditures in the Panel Study of Income Dynamics (PSID) and Consumer Expenditure Surveys (CE) show that neither of these predictions hold. Low-wealth households do not display higher spending growth, despite faster income growth; and low-wealth households show larger fluctuations in spending conditional on income. Furthermore, we find that spending responses to the 2008 stimulus act payments, captured by the CE, are actually smaller for low-wealth households. Motivated by these facts, we develop a model with heterogeneity in preferences. We allow not only that some households are less patient, but also that some households are more amenable to fluctuations in their spending, displaying a higher intertemporal elasticity of substitution (IES). Our motivation for the latter heterogeneity is partly driven by observing that low-wealth households display fundamentally different spending choices, even conditional on how much they spend. More exactly, they allocate a given amount of spending across fewer categories of good, but fluctuations in their spending are much more driven by dropping or adding categories. We show in a simple setting that a more active extensive margin for spending across categories can rationalize a higher intertemporal elasticity. In our model, high IES households hold less wealth, especially less liquid wealth. These low-wealth households display lower spending growth, but that spending is more volatile, consistent with the patterns in the PSID and CE. We explore how to empirically identify households with a high MPC in the context of a model with preference heterogeneity. Asset dispersion now reflects two components: dispersion in “targeted” asset positions as well as deviations from those targets driven by shocks. We show that the MPC is much more related to deviations from target than dispersion in target assets. We show that average propensity to consume (APC) can be a better indicator of MPC than asset holdings. Conditional on income, a household with below-target assets will both (a) be accumulating assets, which means a low APC, and (b) exhibit a high MPC. So low APC is associated with high MPC. In fact, we find that a low APC does predict a high spending response to the 2008 stimulus payments in the CE data, even though holding low assets does not.