Are Consumers Forward-Looking?

This paper establishes a stylized fact: in a cross-section of American consumers, consumption is an extremely poor predictor of future income. The PSID now contains the majority of the lifetime income stream for an early cross-section of consumers. Under the assumption that consumption is a function of the expected present value of income, I invert this function and compare the realized present value of income to consumption. Consumption proves to be a very poor predictor of future income, despite future income being predictable by past income. Under rational expectations, information known to the consumer should not enter the regression of present value on consumption. However, as an empirical matter lagged income is a better predictor than is consumption. This is true for both the present value of future income and for income at speci c future horizons. Indeed, the relation between consumption and income at increasingly longer horizons weakens quickly. One way to summarize results here is that if you want to forecast individual income and have to choose between consumption and a year-old income datum...go with the latter. The conclusion is that consumers appear \as if" they are not forward-looking and that theories of consumption need to be consistent with this stylized fact. Available data being imperfect much of the paper is devoted to robustness tests, none of which change the basic conclusion. keywords: consumption, forward-looking behavior, rational expectations, PSID JEL codes: E21