Escaping Paternalism: Rationality, Behavioral Economics, and Public Policy (Cambridge Studies in Economics, Choice, and Society)

by Mario J. Rizzo

The burgeoning field of behavioral economics has produced a new set of justifications for paternalism. This book challenges behavioral paternalism on multiple levels, from the abstract and conceptual to the pragmatic and applied. Behavioral paternalism relies on a needlessly restrictive definition of rational behavior. It neglects nonstandard preferences, experimentation, and self-discovery. It relies on behavioral research that is often incomplete and unreliable. It demands a level of knowledge from policymakers that they cannot reasonably obtain. It assumes a political process largely immune to the effects of ignorance, irrationality, and the influence of special interests and moralists. Overall, behavioral paternalism underestimates the capacity of people to solve their own problems, while overestimating the ability of experts and policymakers to design beneficial interventions.

Much has been written about whether real people are “rational” and the kinds of errors they may be subject to. Unfortunately, economists, and perhaps other social scientists as well, have been boxed into an excessively narrow and technical concept of rationality. This has left its imprint on the influential heuristics-and-biases literature, which claims people are not fully rational relative to that standard. It worries us that much of the core behavioral literature is often presented as essentially settled, and many alleged biases have now attained the status of truisms. One of our tasks in this book is to show that the evidence is far weaker than has been often assumed. Similarly, the behavioral policy goal of moving people in the direction of their “true preferences” – that is, their well-considered preferences purged of all cognitive and behavioral contaminants – is often a mirage. The behavioral policy discussion frequently suffers from significant empirical difficulties and conceptual confusions.

Experimental evidence suggests that individuals may systematically deviate from the behavior that would best satisfy their own preferences. The list of alleged deviations from strict rationality includes – but is not limited to – status quo bias, optimism bias, susceptibility to framing effects, poor processing of information, and lack of willpower or self-control. To the extent that these phenomena cause people to make errors, paternalist policies can in principle help them to do better – not by some exogenous standard, but by their own standards. This is the defining feature of new paternalism that distinguishes it from the old: the new paternalists claim to help people better achieve their own preferences, not someone else’s.

In this book, we will defend what we call inclusive rationality. Inclusive rationality means purposeful behavior based on subjective preferences and beliefs, in the presence of both environmental and cognitive constraints. This notion of rationality preserves the core notion of purposefulness, and in that sense it should seem familiar. But unlike other notions of rationality – many of which were invented for modeling purposes but have since taken on a life of their own – inclusive rationality does not dictate the normative structure of preferences and beliefs a priori. Instead, it allows a wide range of possibilities in terms of how real people select their goals, form and revise their beliefs, structure their decisions, and conceptualize the world. Their preferences and beliefs may be inchoate, incomplete, inconsistent, mutable, and dependent on context. Inclusive rationality can thus encompass choices and strategies that would not make sense under more restrictive notions of rationality.

Modeled rationality is always incomplete and bounded relative to what it could be in another model. Recognizing this fact does not automatically open the door to normative criticism. There will always be some standpoint from which agents are deficient. The key issue is whether that standpoint is important and relevant to the agents themselves.

Rationality does not imply that human beings have their minds fully made up on the first day about the relative desirability of all alternatives in all states of the world at all points in time. To assert this is an artificial construct of modeling. Whatever the usefulness of the construct in explaining or predicting aggregate behavior, we can see no plausible argument for declaring individuals “irrational” who may be simply undecided. They are in a process of self-discovery. They do not need fixing.

How does embracing an ongoing process of preference formation affect the behavioral paternalists’ normative project? To be blunt, it robs them of the Archimedean point that they would use to judge outcomes. If preferences do not exist independently of the act of choice, then there is no preference set against which to judge the individual’s choices as deficient.

What happens when one belief implies a further statement? Must the agent submit to logic and believe the implied statement? Not necessarily and certainly not generally. The individual may find that the implied statement is completely implausible and therefore decide to reject the initial premise. This occurs often and is good reasoning. The rules of logic are constraints on the relation between propositions. They do not tell people what to believe, pure and simple. Thus, logic is not the whole of reason.

We have found that the literature on cognitive biases, vast though it is, tends to fail in one fundamental respect: recognizing the pragmatic and contextual nature of rational decision-making. The mistake that is constantly and consistently made is to equate rationality with an abstract system of thought unrelated to the purposes and plans of individuals in the environments in which they find themselves. In a related manner, the literature also fails to take into account the socially legitimate expectations of the participants in experiments that the researchers should not provide extraneous or misleading information. These are problems that go to the very heart of the “heuristics and biases” research program. Our perspective, by contrast, recognizes that beliefs serve a purpose – and that purpose is not always truth-tracking. Beliefs can direct attention and provide motivation. Beliefs can be a source of direct satisfaction. Even when beliefs perform a primarily truth-tracking function, there is no uniquely correct way to form and revise beliefs in real-world environments characterized by uncertainty and change.

Standard economics is particularly ill-equipped to deal with the process of automatic counteractive self-control. Most choice problems are conceived as the agent maximizing an objective function. In the more recent extended versions, designed to accommodate behavioral insights, an extra constraint is added: self-control. This is costly impulse control – the deliberate application of cognitive effort to resist temptations. But the modeled agent’s perception of the problem, and thus the framework of decision, is itself static.58 The agent takes the parameters of the problem, including the costs and benefits assigned to outcomes, as given. By contrast, a central idea of counteractive self-control is that the individual spontaneously revalues the costs and benefits of resisting temptation. Rather than drawing on his limited supply of willpower, the agent reconceptualizes the problem he faces.

Theoretically, i.e., if everything else is held constant, the replacement of a sugary product with an artificially sweetened product that yields no energy should result in the loss of weight. But people do not live in an everything-else-held-constant world. The theoretical potential for weight loss will be realized only if the artificially sweetened beverages are actually used in the right way.5 But “evidence that use of NNS [nonnutritive sweeteners] in free-living individuals results in improved weight loss or maintenance is lacking” (Mattes and Popkin 2009, 9).

The new, lower thresholds increased the number of Americans classified as overweight or obese by more than 35 million overnight (Kuczmarski and Flegal 2000). What motivated the WHO and NIH to adopt the new recommendations? We cannot say for sure, but we can venture an educated guess. The WHO report, which heavily influenced the NIH’s later decision to change the BMI thresholds, “was drafted and written under the auspices of the International Obesity Task Force (IOTF),” an organization “primarily funded by Hoffman-LaRoche (the maker of the weight-loss drug Xenical) and Abbott Laboratories (the maker of the weight-loss drug Meridia)” (Oliver 2006, 28–29). These companies stood to benefit from more people being qualified for coverage of their products by private and governmental health insurance. Yet the involvement of the WHO and NIH, both organizations associated with unselfish advocacy of public health, provided cover for the private interests involved.

Behavioral economics indicates that certain types of argument will be more likely to succeed in the political sphere: those that emphasize the urgent need for taking action; those that downplay complexity and emphasize simple solutions; those that flatter people’s current beliefs and attitudes; those that rely on easily recalled and vivid illustrations of alleged problems; and those that emphasize the benevolent goals of the policies in question. Given these tendencies, we should expect the highly motivated parties mentioned earlier to exploit them to advance their agendas. Activists, academics, experts, and industry lobbyists have strong rational incentives to craft their policy proposals so as to maximize their appeal to irrational voters and legislators.

In the world of puppet rationality, there is little room for people to learn about themselves, to discover their own weaknesses and foibles, or to find ways to shape and correct their own behavior. Why not? Because the work has already been done, instantaneously and costlessly, by assumption. Their preferences are fully formed and comprehensive, evaluating and ranking all objects of choice in the present and future in all possible states of the world. Their beliefs about all matters, significant and trivial, are complete, unambiguous, and fully consistent with one another. Real people, on the other hand, have better things to do than ensuring the comprehensiveness and consistency of all their mental states.

At the conceptual and philosophical level, behavioral paternalists have yet to grapple adequately with the thicket of thorny issues that arise in trying to identify people’s true preferences. They have not yet managed to fill in the gaping chasm between identifying inconsistencies among individual preferences on the one hand and choosing which preferences to favor on the other. They have also not yet grappled with the evidence indicating that fast-and-frugal non-Bayesian heuristics can perform better in real-world environments. They have hardly begun to fill the gaps in knowledge required to craft policies that reliably improve human welfare by their chosen standard.

One possible conclusion is that we should reject behavioral economics in its entirety and revert to the neoclassical approach – one that largely rejected paternalistic thinking. But that would be far too simplistic. It would be a mistake to cling tightly to sterile models of human behavior that have no room for inchoate preferences, self-discovery, trial and error, and self-regulation. While the simplified models of neoclassical theory may still be useful for particular purposes and contexts, they cannot provide a panoptic picture of human decision-making. They cannot capture the full depth and complexity of human choice. This is true in general, but particularly with regard to normative analysis. Reliance on formalistic models carries the distinct danger of measuring real humans relative to the model – and judging them deficient when they depart from it. This is a failing of both neoclassical and behavioral economics.

What the discipline does need, however, is a strong dose of humility – particularly for those behavioral researchers looking for policy applications of their work. It is jarring, to say the least, to see social scientists pointing out the errors of private individuals – and then failing to consider that social scientists and policymakers are also subject to error. It is frustrating to see behavioral researchers demonstrating the complexity of real decision-making processes – and then ignoring that complexity when recommending regulatory corrections of those very processes. It is simply baffling to see behavioral economists showing how real behavior deviates from neoclassical norms – and then insisting that behavior must conform to those norms or else be judged deficient. Our hope is that, equipped with greater humility, greater respect for nonstandard preferences, and greater awareness of the surprising functionality of real-world behavior, behavioral researchers will be less inclined to approach humanity from a position of presumed superiority, like puppet masters correcting the behavior of errant puppets. Instead, they will approach them as fellow human beings doing the best they can, trying to improve their own choices, and offering friendly advice on how others might do the same.