Heterogeneity and Asymmetry in the Marginal Propensity to Consume

© 2021 Alexander BallantyneBuilding on the 'partial insurance' framework of Blundell, Pistaferri and Preston (2008), I estimate households' marginal propensity to consume (MPC) out of transitory and permanent income shocks using data from the US Panel Study of Income Dynamics over 1999-2019. Following Chatterjee, Morley and Singh (2021), I use the state space representation of the model but estimate it with Bayesian methods, which provide a natural setting for estimating high dimensional models with unobserved components. I extend the model in different directions by allowing for: i) asymmetry in MPCs with respect to the sign of transitory income shocks; ii) observable and unobservable heterogeneity in MPCs; iii) robust aggregation of population statistics; and iv) selection of the main drivers of MPC variation from a rich set of covariates. The results challenge some conventional understanding of consumption smoothing behaviour. I find clear evidence of positive asymmetry - larger responses to positive transitory income shocks - for households with substantial home equity. This contrasts the existing survey evidence on 'reported preferences' and is suggestive of a behavioural interpretation. However, evidence of negative asymmetry for so called 'poor hand-to-mouth' households (who have low liquidity and low wealth) is consistent with the Kaplan-Violante two-asset model. Households with heads aged 55-64 are found to exhibit negative asymmetry, irrespective of their financial position, which is interpreted as a savings preference before retirement. Under the assumption of symmetry, I look for evidence of variation in MPCs that would accord with theoretical predictions. Distributions of MPCs along balance sheet positions show some evidence of financial frictions, but demographic characteristics account for most of the variation in MPCs. I estimate unobserved MPC heterogeneity and find that it is large for transitory MPCs, but that observed variation explains much of the heterogeneity for permanent shocks. Population-representative dollar-for-dollar MPCs for the US are carefully aggregated from microdata elasticities using an extended dataset that includes retired households. Households increase permanent nondurables consumption by 7 cents per dollar for transitory income shocks, and 38 cents for permanent shocks. Using a rich set of demographic and financial covariates, Bayesian shrinkage and sparsification techniques are applied to identify the key drivers of MPC heterogeneity. Households with heads who are black, female, or born in the 1980s have larger responses to permanent income shocks, while those with four or more years of college have lower responses. These four covariates drive the overwhelming majority of heterogeneity in MPCs out of permanent income shocks. The pattern of results is consistent with financial literacy being an underlying mechanism; however, incorporating a proxy for financial literacy does not remove the demographic effects. This thesis provides a nuanced view of how different households respond to income shocks, which can help guide better targeted policy interventions. The findings are consistent with some aspects of the life-cycle literature; however, they also challenge the conventional assumption of symmetry and focus on financial drivers of MPCs. Further research into behavioural factors that affect consumption smoothing and demographic differences across the life cycle is warranted