Cass Sunstein: For decades, Hayek has been a hero of mine. Behavioral economics and its findings have been a focus of mine for not quite as long as I’ve had my admiration for Hayek, but still for some decades.

The question is whether there is such a thing as Hayekian behavioral economics. I’m going to suggest that there is in the sense that we can take with enthusiasm Hayek’s notions about the fallibility of planners and the epistemic problems planners of all sorts face, while also recognizing that human beings are sometimes insufficiently informed, something that I think would not surprise Hayek in the least, nor that this lack of information is sometimes behaviorally biased.

The ideal we’re getting at here is thinking of what choosers choose under epistemically favorable conditions. We can consider epistemically favorable conditions to be those where choosers are free, or at least free enough, from information gaps and from behavioral biases. Hayekian behavioral economics at its core is not going to ask, “What do planners know and believe and prefer?” Rather, it asks, “What do choosers know and believe and prefer under circumstances that are epistemically favorable?” That is how we can get at people’s true preferences to the best of our ability.

Hayek’s distinctive account of the reason to respect individual liberty is rooted in his critique of socialism and centralized planning. His emphasis was on the lack of knowledge on the part of planners compared with the knowledge that participants in markets have. The basic objection sketched in his great 1945 essay “The Use of Knowledge in Society” is that the price system is a marvel because it aggregates the information and tastes of lots of people, incorporating a lot more information than could possibly be assembled by central planners or groups or boards.

Hayek emphasized the unshared nature of information, the dispersed bits of incomplete and frequently contradictory knowledge that all the separate individuals possess. He stressed the very important and unorganized knowledge, the knowledge of particular circumstances of time and place. Hayek was not an emotional writer, but there’s a sense of soaring, at least, in his suggestion that it is more than a metaphor to describe the price system as a kind of machinery for registering changes, or a system of telecommunications that enables individual producers to watch merely the movement of a few pointers. On Hayek’s account, the price system is an extraordinary device for capturing collective intelligence in part because it collects in everyone what everyone knows and in part because it imposes the right incentives.

That’s the background. In light of modern behavioral findings about human error, it would be possible to object that the price system is not always so marvelous and that other institutions might do better. If consumers show limited attention, if they don’t pay attention to certain characteristics of products, let’s say, or if they show unrealistic optimism, or if they are more indifferent than they ought to be to risks that they face because they wrongly think they have a personal immunity against those risks, or if they suffer from present bias in the sense that they have implausibly high discount rates, then under those circumstances the price system might miss something important. Hayek’s “system of telecommunications” might give the wrong messages.

It is possible to agree with Hayek’s arguments about planning and prices while also thinking that certain forms of regulation and being alert to behavioral biases are not out of bounds. Hayek himself did not engage with behavioral findings for reasons that we can discuss, in some cases because they came after his time, but he did engage with the limitations of private markets.

Hayek wrote, “Probably nothing has done as much harm to the liberal cause as the wooden insistence of some liberals on some rough rules of thumb, above all the principle of laissez-faire.” Hayek didn’t choose his words carelessly, so we might pause over that provocative sentence. Or consider this, from The Road to Serfdom: “To prohibit the use of certain poisonous substances or to require certain precautions in their use, to limit working hours or to require certain sanitary arrangements, is fully compatible with the preservation of competition.”

The only question here is whether in a particular instance the advantages gained from an intervention are greater than the social costs that they impose. Hayek was, at least sometimes as we see from these passages, on board the costbenefit train and willing to concede some regulations even in a case where it’s not clearly a matter of externalities. In fact, limitation of working hours or requirement of certain sanitary arrangements doesn’t seem offhand to be a problem of externalities, but Hayek cited those as justifiable examples.

Maybe a mandatory seatbelt law, a ban on trans fats, or regulation of exposure to certain carcinogens in the workplace would be unobjectionable if the only question to consider is costs versus benefits, according to Hayek. The question remains, do Hayek’s other arguments about the knowledge problem count against, for example, cigarette taxes or taxes on sugarsweetened beverages? Are they a large-scale objection to paternalism from public institutions? Hayek didn’t answer these questions. His high-level concerns about coercion and about the deficiencies of planners can’t answer concrete questions such as whether salient disclosure requirements are essential to overcome limited attention or whether energy efficiency mandates are an appropriate response to present bias and myopic loss aversion. Those just aren’t questions he ever directly answered.

I would contend that any form of Hayekian behavioral economics would firmly reject the view that public officials should be content to merely identify individual errors and to declare victory. We have to ask on Hayekian grounds how costly the errors are compared with the errors that would be induced by corrective measures. To engage in that analysis, we have to know something about relative institutions. What I’m suggesting is that if we do decide to proceed with a remedy, a Hayekian approach would try to reduce the knowledge problem by asking what individuals would do under epistemically favorable conditions.

The good news is a stream of research is asking exactly that question: What do choosers in fact do under epistemically favorable conditions? It’s not about what planners want given their own values and desires and commitments; it’s about getting at those subjective individual preferences. And behavioral economists have developed a range of conceptual tools and empirical methods to do that, ways to understand and analyze how people respond when they are freed from undesirable biases and insufficient information.

So, the best approach I’m suggesting is to ask what are active, informed choosers who are free of behavioral biases, who have broad view screens, and who are unaffected by clearly irrelevant factors and frames to do? Now, this might seem like an abstract question. It might even seem the sort of question for which one ought to, on Hayekian grounds, be most distrustful of those who even dare to ask it. But before you get there, let’s just notice there’s a stream of research that’s asking exactly these questions. At least to some degree, the answers are not unknowable, and they are not irretrievably hidden behind the knowledge problem.

It would be extravagant maybe to claim that those interventions defended by reference to people’s choices under epistemically favorable conditions are Hayekian in the sense that Hayek advocated them. But it may not be going too far to insist that they’re in Hayek’s general spirit and respectful of his most fundamental concerns. They might, if we’re lucky, provide an orientation for both theory and practice, now in its early stages, that promises to preserve and to cherish freedom while also improving human lives, not least by lengthening them.

Mario Rizzo: I don’t intend to spend much time on whether Hayek would have approved a behavioral public policy. However, as many of you know, May 8 was Hayek’s 122nd birthday. He’s quite old right now. And in my birthday seance with him, I asked whether he approved of behavioral public policy, and he said simply that he didn’t much care for it. And that is all he had to say. Well, that’s not quite enough. So what I will do here is answer the question, has behavioral public policy solved the Hayekian knowledge problem?

Let me review a few things to start. And on these points, I think Cass and I are in perfect agreement. Behavioral public policy is not just one thing. I view it as a combination of three things. First, a set of normative standards. And these range from subjective true preferences, which was a term previously used, to objective direct welfare measurement. Second, it’s a continuum of policies, from soft to hard, advice, nudges, nonwaivable defaults, sin taxes, and mandates. And third, it claims to be evidence based. That means no a priori assumptions about individual behavior.

So what is the Hayekian knowledge problem? I think Cass correctly identified it: we have to contrast general scientific knowledge, propositions, rules derived from science, with the concrete knowledge of the circumstances of time and place. For example, certain biases may be found in lab experiments. However, in the real world, these are heavily dependent on context, both as to existence and quantitative magnitude.

One of the leading psychologists of the 20th century, Jerome Kagan, in his book Psychology’s Ghost, emphasizes what he considers to be a fundamental problem of psychology: “Few psychological concepts intended to represent a person’s tendency to react in a certain way apply across diverse settings.”

This is a recognition, I think, of essentially the Hayekian knowledge problem, only in psychology. We can have generalizations about psychological propensities, but they are very, very dependent on context. So, is behavioral public policy evidence based? I’m going to talk about what I consider to be some of the deadly knowledge problems of behavioral public policy. The first relates to the question of true preferences. These are essentially counterfactual preferences, preferences that people would have if they had all relevant information, no deficit of cognitive abilities, and complete willpower. But this is essentially the picture of the perfect neoclassical agent, which behavioral economists say does not exist.

Therefore, the satisfaction of these sorts of theoretical preferences does not satisfy the preferences of any real living person, even if they can be discovered. This is an interesting result that’s not often emphasized, and I’m going to ignore it for the rest of what I have to say, but I think it’s an important point. Second, there’s the question of the degree of bias. It’s not enough to know the presence of a bias; we must know its quantitative extent. For example, a properly calibrated sin tax, let’s take cognizance of the degree to which people prefer healthier food later rather than now, their present bias with respect to food. The weaker the bias, the lower the tax should be, and vice versa. But quantitative estimates of bias are unreliable and do not generalize into the real world.

The third one that I want to emphasize is the question of bias interactions. Normally behavioralists analyze the effect of only one bias at a time. Wikipedia lists, however, 175 cognitive biases. In all fairness, not all of these 175 biases are that distinct, but nevertheless, there are an awful lot of cognitive biases. A recent National Bureau of Economic Research study shows that biases are very often highly correlated with each other. Since biases do not all move in the same direction, or to the same degree, the net effects can cancel, or efforts to counterbalance a particular one can make matters worse.

For example, a person may be present biased and intend to save too little, but if he also exhibits projection bias—that is overestimating his future consumption needs and plans for too long a retirement—the overall effect of the biases unclear, as they move in different directions.

Fourth, there’s population heterogeneity. Behavioral policies are usually one-size-fits-all, one sin tax for all consumers of sugary drinks, but biases are not uniform. In one major study, heterogeneity was substantial. For example, 29 percent of the sample had present bias, while 33 percent had future bias with respect to money. With respect to food—that is healthier food now or later—15 percent had present bias, 7 percent had future bias, and the rest were unbiased.

Counteracting behaviors is my fifth problem. I give two examples. First, behavioral economists love to attack sugary soft drinks. Would it do any good even if consumption were reduced? The most likely substitute are soft drinks with noncaloric sweeteners, but there is no consistent evidence that the substitution does any good from a health perspective. Why? People respond by consuming more calories elsewhere. Example two: those firms that introduce automatic enrollment in retirement savings programs saw their enrollees offset 40 percent more of the retirement savings with loans and withdrawals after eight years, compared to those who actively opted in.

The sixth problem is one relating to self-regulation and small-group debiasing. We need to know not only the tendencies toward biases but also the operative amount in any given situation. This is affected by first, self-regulation, which is idiosyncratic and hard to identify. Mary eats junk food on weekends; is she breaking her diet, or is this the glue, the exception, that keeps the diet in place? Second, growing research shows that when people make decisions after discussion in small groups, biases are eroded and sometimes disappear entirely. In the real world, this would include family members, friends, colleagues, and other advice-givers.

The last problem that I want to mention is the dynamic impact on self-regulation. Considerable research shows that regulation by external sources and self-regulation are substitutes. Behavioral public policy is supposed to be in part a remedy for deficient willpower. But when internal or self-control is not exercised, it deteriorates in the long run, even in areas unrelated to the initial external regulation. The loss of self-control and self-regulatory capacity generalizes.

In conclusion, nothing I said should be construed as claiming that people do not make mistakes, that they are not sometimes foolish, or that they have perfect willpower; people are fallible. But it is one thing to recognize human foibles on a general level yet another thing to ascertain them in specific instances, given the myriad of local and personal factors that must be considered in decisionmaking. People need good information to make good decisions. The market can provide much of it. In some cases, the government may need to provide it. But providing information is not a nudge, it just boosts our decisionmaking capacity. We already knew that before behavioral economics. The more aggressive interventions Cass proposes, even when they still leave some degree of choice, are still both centralizing and coercive in ways that do not entirely escape Hayek’s objections to coercive central planning.