Working Paper: NBER ID: w2012
Authors: Joseph G. Altonji; Aloysius Siow
Abstract: This paper tests the rational expectations lifecycle model of consumption against (1) a simple Keynesian model and (ii) the rational expectations lifecycle model with imperfect capital markets. The tests are based upon the relative responsiveness of consumption to income changes which can be predicted from past information and income changes which cannot be predicted. Since there is strong evidence that panel data contains substantial measurement error, the tests are especially constructed to allow for measurement error in the income process. They also allow for more general income processes than have been considered to date in the literature. The results reject the Keynesian model and generally support the lifecycle model, although the tests are not sufficiently precise to rule out the possibility that some households are liquidity constrained. Measurement error does have a strong influence on the relationship between consumption and income. When it is ignored our tests do not reject the Keynesian model. We show that consideration of measurement error may also reconcile differences in the results of Hall and Mishkin (1982) and Bernanke(1984). Nevertheless, our most important conclusion is that Hall and Mishkin's, Bernanke's, and Hayashi's (198 ) qualitative finding that the vast majority of households obey the lifecycle model is not an artifact of failure to account for measurement error in the income data.
Keywords: Consumption; Income Changes; Measurement Error; Lifecycle Model
JEL Codes: D12; E21; C23
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Ignoring measurement error (C20) | Incorrect inferences regarding the validity of the lifecycle model (C52) |
| Determinants of income are exogenous (E25) | Clearer causal interpretation of the relationships examined (C21) |
| Measurement error (C20) | Underestimated true sensitivity of consumption to income changes (D12) |
| Results reject Keynesian model (E12) | Support lifecycle model (C41) |
| Most households follow lifecycle model (D15) | Robust to measurement error considerations (C20) |
| Measurement error (C20) | Downward bias in OLS estimates of consumption response to income changes (D12) |
| Using instrumental variables (C36) | Consumption response to income changes is approximately three times larger than OLS estimate (E21) |