The behavioral life-cycle hypothesis

Shefrin dan Thaler (1988) Abstract Self-control, mental accounting, and framing are incorporated in a behavioral enrichment of the life-cycle theory of saving called the behevioral life-cycle (BLC) hypothesis. The key assumption of the BLC theory is that households treat components of their wealth as nonfungible, even in the absence of credit rationing. Specifically, wealth is assumed to be divided into three mental accounts: current income, current assets, and future income. The temptation to spend is assumed to be greatest for current income and least for future income. Considerable empirical support for the BLC theory is presented, primarily drawn from published econometric studies. Introduction Modigliani and Brumberg's life-cycle theory of saving (1954) and Friedman's similar permanent income hypothesis (1957) are classic examples of economic theorizing. The life-cycle (LC) model makes some simplifying assumptions in order to characterize a well-defined optimization problem which is then solved. The solution to that optimization problem provides the core of the theory. Attempts to test the life-cycle hypothesis have met with mixed success. As summarized by Courant et al. (1986, 279-80), "But for all its elegance and rationality, the life-cycle model has not tested out very well... Nor have efforts to test the life-cycle model with cross-sectional microdata worked out very successfully." Various alterations to the theory have been proposed to help it accomodate the data: add a bequest motive, hypothesize capital market imperfections, assume that the utility function for consumption changes over time, or specify a particular form of expectations regarding future income. These modifications often appear to be ad hoc, since different assumptions are necessary to explain each anomalous empirical result. This paper suggests that the data can be explained in a parsimonious manner by making modifications to the life-cycle theory that are quite different in spirit from those cited above, namely modifications aimed at making the theory more behaviorally realistic. We call this enriched model the Behavioral Life Cycle (BLC) hypothesis. ... Conclusion The LC model is clearly in the mainstream tradition of microeconomic theory. It is typical of the general approach in microeconomics, which is to use a normative-based maximizing model for descriptive purposes. The recent papers by Hall and Mishkin (1982) and Courant et al. (1986) are really advances in the LC tradition. Our model is quite different in spirit. First of all, our agents have very human limitations, and they use simple rule of thumb that are, by nature, second-best. While the LC model is a special case of our model (when either a first-best rule exists or there is no self-control problem), our model was developed specifically to describe actual behavior, not to characterize rational behavior. It differs from a standard approach in three important ways. (1) It is consistent with behavior that cannot be reconciled with a single utility function. (2) It permits "irrelevant" factors (i.e. those other than age and wealth) to affect consumption. Even the form of payment can matter. (3) Actual choices can be strictly within the budget set (as a Christmas club). The relationship between the self-control model and the LC model is similar to the relationship between Daniel Kahneman and Arnos Tversky (1979) prospect theory and expected utility theory. Expected utility theory is a well-established standard for rational choice under uncertainty. Its failure to describe individual behavior has led to the development of other models (such as prospect theory) that appear to do a better job at the tasks od description and prediction. The superiority of prospect theory as a predictive model, of course, in no way weakens expected utility theory's value as a prescriptive norm. Similarly, since we view the LC model as capturing the preferences of our planner, we do not wish to question its value to prescriptive economic theory. The LC model has also served an enormously useful role in providing the theory against which empirical evidence can be judged. For example, the one-to-one pension offset was a result derived from the LC model (without bequests), and the numerous studies we cite were no doubt stimulated by the opportunity to test this prediction. Saving adequacy even more directly requires a life-cycle criterion of appropriate saving with which actual saving can be compared. At times we have argued that the use of ad hoc assumptions, added to the theory after anomalous empirical evidence has been brought forward, renders the LC model unstable. It is reasonable to ask whether our model is testable. We think that it is. Every one of the propositions we examined in this paper represents a test our model might have failed. For example, if the estimated pension offsets were mostly close to -1.0 instead of mostly close to zero, we would have taken that as evidence that self-control problems are empirically unimportant. Similarly, the effects of bonuses on saving could have been negligible, implying that mental accounting has little to add. Other tests are also possible. Our theory suggests the following additional propositions...