How does Household Consumption Respond to Income Shocks? Evidence, Theory and Implications SBE/SES 0820519, Perri and 0820494, Krueger: Collaborative
The broad goal of our project is to understand the impact of income changes, deterministic or stochastic, on the welfare of households in modern societies. For example, the project allows us to give better answers to questions such as: What is the impact of a macroeconomic recession on household consumption? Should the government provide transfers to households facing a decline in their income?
The project uses the only micro-level data set (the Italian Survey of Household Income and Wealth) that follows individual households over time and records income, comprehensive consumption, labor supply and wealth data in order to document how households economically respond to a change in their income. Our study is the first to systematically study the joint dynamics of income, consumption and wealth in household micro data. The inclusion of wealth in the empirical analysis is novel and allows us to effectively discriminate between different economic theories that all predict that consumption responds to income shocks less than one-for-one. It also allows us to explicitly study the impact on consumption-savings decisions of changes in wealth that are independent of labor income dynamics (for example the appreciation in value of owner-occupied houses).
Our empirical analysis uncovers evidence of significant asymmetries (between income increases and income declines) in how consumption and wealth respond to a given income change. Furthermore we show that the magnitudes of consumption and wealth responses to income shocks have changed significantly over time. These empirical findings are not only the motivating stylized facts for the ensuing theoretical and quantitative analysis of our study, but they also provide the broader research community with readily available data against which it can test its micro-founded consumption-savings models.
We then determine which broad classes of consumption-savings models are qualitatively and quantitatively consistent with the facts established in the first part of our study. Important model inputs are the risky wage or labor earnings process and the risky asset prices (including property prices) that households face over their life cycle. Since the facts in the first part of our project reject both the notion that consumption responds to income changes one-for-one, and wealth not at all (as the hands-to-mouth theory of consumption would predict) as well as the notion that consumption does not respond to income shocks at all, with wealth taking up all the income variation (as the full insurance-complete markets model predicts), we focus, in our theoretical and quantitative analysis, on models that imply partial consumption insurance.
Our results indicate that, quantitatively, the most suitable consumption-savings model is one that implies more consumption insurance than the standard incomplete markets model, but less than models with limited commitment or private information. We use these findings to determine the degree of financial market incompleteness that is consistent with our stylized facts. The characterization of the appropriate (relative to the data) financial market structure is of independent interest to other researchers that want to use these models for applied policy analysis of social insurance programs (such as social security, disability insurance or progressive income taxation).
In the last part of our project we evaluate the theoretical and quantitative implications for asset prices and asset returns of the model favored in the analysis in the second part. As in any model with incomplete consumption insurance, the asset prices and returns in the model depend on higher moments of the consumption growth distribution. We first theoretically identify these crucial moments, then use the Italian household data to construct the corresponding empirical moments and finally evaluate whether the model-implied asset prices are consistent with the actually observed Italian asset pricing facts. We find that our preferred imperfect consumption insurance model predicts a significantly higher equity premium (the expected excess return of stocks over risk-free bonds) than the standard asset pricing model with full insurance.
