Distinguishing between cognitive biases

Economists have recently taken increased interest in a number of cognitive biases and heuristics first documented by psychologists.1 In theoretical studies, economists typically introduce such biases and heuristics into stylized models with a goal of understanding how small, but psychologically relevant, deviations from the standard economic framework can influence decisions such as saving and consumption (Harris and Laibson 2001), investment (Barberis and Huang 2001), and labor supply (Fang and Silverman 2004a). In empirical studies, economists have followed two basic strategies. The first is to derive distinctive empirical implications from a model of a particular bias or heuristic and then check if the data are qualitatively consistent with the bias model's predictions but inconsistent with the standard model's (see, for example, Babcock et al. 1997; Genesove and Mayer 2001; Della Vigna and Malmandier 2004). The second involves estimating structural models that allow a particular bias and attempt to measure the degree of that bias and its implications (Fang and Silverman 2004b; Paserman 2004). It, unlike the first strategy, assumes an explicit model that permits simulations of the behavioral and welfare consequences of counterfactual policy experiments. An important motivation for incorporating cognitive biases and heuristics in economic analyses is public policy. When a public economist evaluates a policy, the typical first step is to consult the formalization of Adam Smith's invisible hand in the first fundamental theorem of welfare. According to this theorem, when competitive markets exist and their participants share information commonly, the allocations of those markets are efficient. That is, no other feasible allocation could make one person better off without making someone else worse off. It follows that if markets are missing, or imperfectly competitive, or if economic agents are acting with incomplete information, policy interventions may be justified on efficiency grounds. The implications of cognitive biases and heuristics for decision making form the basis of another rationale for policy interventions into economic activity. Biases and heuristics may drive a wedge between normative or long-term preferences and revealed preferences. That is, biases and heuristics can make what an individual actually chooses to do different from what it would be if bias did not color perception, if choices were made from a temporal distance, or if careful attention to decisions might be paid at zero cost. If biases or heuristics lead economic agents to decisions at odds with their normative or long-term preferences, then public policy interventions could in principle make some better off yet none worse off. Thus, even in cases where perfect markets exist, the influence of cognitive biases and heuristics may justify public policy interventions on efficiency grounds. Although investigations into the relevance of cognitive biases and heuristics for economic decision making may have profound policy implications, those with such a research agenda face a fundamental difficulty: given the large number of deviations from strict rationality that psychologists have documented, how should we determine which ones apply to which economic settings? So far, both the theoretical and the empirical studies in economics have tended to investigate the implications of cognitive biases and heuristics one bias at a time (Barberis and Huang 2001 is an exception). The strategy of limiting attention to one potential bias may, in some cases, be justified by a priori indications of the primacy of that bias in determining behavior. Even when there is no such logic, an incremental approach offers some clear advantages. For theory, isolating the influence of a minimal deviation from the standard framework has intrinsic interest and provides greater potential for clarity and tractability. An additional advantage derives from a concern for distinguishing among the effects of various biases. In some quite standard settings it is impossible to empirically distinguish between a model of even a single bias and a traditional model (see, for example, Barro 1999). Distinguishing among multiple models of bias at once may present substantial challenges. Nevertheless, in many cases several biases might plausibly explain behavior. The obvious question for empirical research is whether we can distinguish these various biases from each other, and from a traditional model, using readily available data. If such data are not sufficient, it then becomes important to investigate what additional data should be collected. Distinguishing among biases is important because, as we will demonstrate, different biases may lead to very different policy recommendations. Here our primary task is to determine whether certain biases can be distinguished using data on labor supply and welfare program participation. In the context of a simple model of work and welfare program participation, we investigate whether it is possible to distinguish between two psychological biases: Time inconsistent discounting in the form of present-biased (hyperbolic) time preferences and nonrational beliefs in the form of projection bias. We show formally that indeed these can be distinguished from each other, and from a conventional model using standard data. In addition, we highlight a novel distinction between the two biases and argue that individuals under the two biases may exhibit different attitudes and changes in attitude toward welfare eligibility restrictions such as time limits. To the extent that such data may be collected, these differences in attitudes provide an unexplored channel that researchers may exploit to distinguish biases in belief and time discounting.