A Behaviorist Goes to Washington
Photo by Marcello Duarte
Simpler: The Future of Government
Cass R. Sunstein
Simon & Schuster, $26 (cloth)
Cass Sunstein went to Washington with the aim of putting some theory into practice. As administrator of the Office of Information and Regulatory Affairs (OIRA) during President Obama’s first term, he drew on the behavioral economics he helped develop as an academic. In his new book, Simpler, he reports on these efforts and elaborates a larger vision in which they exemplify “the future of government.”
Sunstein’s approach is inspired by a famous body of survey and laboratory research suggesting that individual decision-making is often irrational. For example, in surveys people express a willingness to pay more for insurance that covers injuries from terrorism than for insurance that covers injuries full stop, although logically the latter must be more valuable. This tendency appears to be a manifestation of “the availability heuristic”: mention of terrorism summons up more vivid imagery than mention of injury in the abstract. In the laboratory, when experimenters give subjects a mug and ask them how much they would sell it for, the subjects tend to demand considerably more than they offer to pay for the same mug when, instead of giving them the mug, the experimenters give them cash and an opportunity to buy it—an “endowment effect.” And many people who choose the one-ounce piece of chocolate when offered a choice between one- and two-ounce pieces choose the two-ounce piece when they also have a three-ounce option. It’s easier to choose the two-ounce piece when it is the middle choice—a “framing effect.” In general, the behaviorists find that people have strong tendencies to drift with the status quo rather than opt for change, to succumb unreflectively to rhetoric and imagery, and to excessively discount the future.
Such findings challenge libertarian and economistic accounts that portray individual choices as virtually always rational, or at least as only occasionally and unpredictably irrational. The behaviorists suggest that irrationality is both routine and predictable. Sunstein, a long-time defender of liberal politics, sees this research as supporting government intervention into private economic arrangements. Yet he has also absorbed the critiques of New-Deal-style bureaucracy, which insist that it does not follow from the fact that individuals choose sub-optimally that government can do better. Although the government may know that choices are biased in a particular direction, it may not be able to determine reliably whether that direction is contrary to the interests of the choosers. Some of the choices in the behaviorist studies, such as paying more for insurance with less coverage, are clearly against the choosers’ interests. But often, as with the chocolates, we cannot be sure, even after identifying cognitive bias, what the best choice is. Sunstein’s contribution is to elaborate intermediate interventions that protect against unconscious bias without categorically preempting choice.
Simpler reports some notable achievements, but it exaggerates the practical value of the behaviorist toolkit. The Obama administration’s most important policy initiatives make only minor use of it. Despite its upbeat tone, the book implies an oddly constrained conception of the means and ends of government. It sometimes calls to mind a doctor putting on a cheerful face to say that, while there is little he can do to arrest the disease, he will try to make the patient as comfortable as possible.
• • •
The book recounts a variety of Obama administration efforts to make government more transparent and user-friendly. These efforts have improved access to aggregate government data, enabled citizens to locate more of the specific data government has collected about them, provided clearer accounts of government activity, enhanced the intelligibility of government regulations and applications, and distributed helpful guidance on matters ranging from nutrition to energy efficiency.
These reforms are admirable, but there is nothing novel about them in principle, and they have been influenced only peripherally by behavioral research. The initiatives most identified with behavioral research involve what Sunstein calls “nudges.” A nudge seeks to influence choices by calibrating the way they are presented.
There are two types of nudges. The first concerns disclosure. The government can require sellers to more saliently and clearly communicate key factors the consumer might otherwise overlook or misunderstand. So, as a condition of government-backed loans to their students, vocational schools must prominently disclose to applicants their graduation rates and the employment rates of their graduates. Airlines must quote prices inclusive of airport fees and taxes. Auto dealers must report “annual fuel cost” as prominently as the less informative “miles per gallon.”
The second type of nudge involves changing the default rule—the rule that dictates what happens if the individual does not make a relevant choice. The most famous example concerns employer-sponsored retirement savings plans. The traditional default rule is that if the employee fails to enroll, she does not participate; the employer does not divert part of her pay to a tax-advantaged savings account. But behavioral science teaches—if we didn’t know it already—that people have biases in favor of the status quo and that they tend to over-discount their future needs. People who don’t save run a high risk of financial distress in retirement and are likely to regret their prior choices. So it makes more sense for the default rule to be automatic enrollment. If the employee does not want to save, she can take initiative to opt out, but if she drifts instead of choosing, the default rule leads her to save. Sunstein reports proudly that recent regulations encourage employers to shift from the traditional opt-in design to an opt-out one.
To Sunstein, nudges have two sorts of advantages over mandatory regulation. First, they constrain liberty less. A mandatory rule might deny government-backed loans for tuition at schools with low graduation rates or force participation in employer-sponsored retirement programs. A nudge leaves choice open. To be sure, the nudge represents an exercise of government power on the basis of official judgments about people’s interests, but that power is exercised weakly and the judgments are more tentative than a mandatory rule would require. If nudges are paternalistic, they are “soft” paternalism rather than the hard paternalism of mandatory rules.
Second, effective nudges demand less government information than effective mandatory rules. A nudging government still needs some basis for deciding what direction to nudge in, but as long as individuals are free to push back, they can escape the effect of the government’s mistakes if they care enough to make an effort. Moreover, nudges permit individuals to act on particular tastes or take account of unusual circumstances in a way that a uniform rule would not. Even if it is in the interests of employees generally to save more, certain employees may have good reasons for not doing so at a given moment. Maybe they have extraordinary medical bills or need a down payment for a house. A regulator writing a mandatory rule could not factor in such varying circumstances, but a nudge does the trick.
• • •
Yet, while Sunstein gives many examples of plausible nudges, most of them, like the retirement-plan default, seem to have fairly modest impacts. Only 20 percent of the income of elderly retirees comes from pension funds; most of the rest comes from the traditional hard-paternalist Social Security program. You can get an idea of how limited the domain of the behaviorist tools is by surveying the Obama administration’s most ambitious domestic initiatives: the Food Safety Modernization Act, the Dodd-Frank Wall Street Reform and Consumer Protection Act, the Patient Protection and Affordable Care Act (a.k.a. Obamacare), and Race to the Top and related education initiatives. Nudges do not play central roles in any of them.
Choice architecture plays no role in the food safety statute. The law could have created a rating system that advised consumers of the probability that food from a given processor would make them sick, while permitting processors to market and consumers to purchase low-rated products. (Some cities do something like that with restaurant sanitation regulation.) Instead, the law demands a high degree of safety from all processors and prohibits those who do not meet it from operating. On Sunstein’s criteria, there is no need for halfway measures because the government understanding of consumer interests is strong and there is little variation in relevant tastes.
Sunstein mentions in passing that choice-architecture tools will be useful in implementing the Affordable Care Act, and indeed the statute is chock full of nudges, including a requirement of automatic enrollment in employer health plans and a directive to the Food and Drug Administration to study standardized disclosure of risks and benefits of prescription drugs.
In any event, the choice-enhancing features of the statute seem considerably less important than the features of the bill that are designed to narrow choice. Consumers no longer have the choice to remain uninsured. This mandate reflects in part a hard paternalist belief that few people could rationally choose not to insure, as well as concerns about adverse selection—the tendency of relatively healthy people to undermine risk pooling by opting out. Providers have less choice about which applicants they can reject. They can’t reject any on the basis of pre-existing conditions, otherwise they would “cream”—compete on their ability to identify relatively healthy people rather than on the quality of their service. If the Medicare cost controls work, they will necessarily narrow choices of drugs and services. And the statute encourages states to restrict the rates insurers can charge. Rate regulation arises from fear that insurance markets will become oligopolistic. Disclosure is no antidote for monopoly power.
Sunstein’s vision calls to mind a doctor with a cheerful face saying there’s little he can do.
Dodd-Frank has a bigger role for choice architecture than do the other initiatives. It creates a Consumer Financial Protection Bureau to regulate consumer credit contracts, and no doubt much of the regulation will address disclosure and default matters. But the law is also likely to lead to choice-limiting prohibitions of abusive contract terms. More importantly, Dodd-Frank is primarily an effort to avoid further financial system crises, and irrational consumer choices are not the major source of risk here. The most important causes of the recent crisis involved the externalization of risk by borrowers and lenders, ultimately to the federal government through its deposit insurance, lender-of-last-resort, and too-big-to-fail bailout practices. Many risk-creating transactions, including consumer “liar loans” and no-down-payment deals, were rational for the immediate parties. No amount of choice structuring in connection with a decision can reliably prevent harm that will occur to someone other than the decision-maker. Thus, the central thrust of Dodd-Frank is to constrain risk-taking by restricting financial institutions’ choices about such matters as proprietary trading and capital levels.
Finally, Race to the Top also has a choice feature. As a condition of eligibility for the relevant grants, states must make some provision for charter schools. Yet the charter school requirements are very weak. The main thrust of the administration’s education initiatives is to induce states to adopt standards, data systems, diagnostic tests, and remedial interventions that permit schools to engage in continuous re-assessment and adaptation of teaching practices—what the regulations call “instructional improvement systems.”
• • •
The obverse of Sunstein’s preoccupation with choice architecture is his relative indifference to other approaches to making administration less rigid. Recall that among the problems Sunstein sees with conventional regulation are, first, that it mandates conduct in situations where the regulator doesn’t know with confidence what is the right thing to do, and second, that it is insufficiently sensitive to relevant local variations in taste or circumstances.
The most common way to deal with the first problem—insufficient information—is to build learning into the process of intervention: the regulator intervenes provisionally, studies the effects of her intervention, and adapts as she learns. It is commonplace for statutes to mandate or fund demonstration or pilot projects. More importantly, statutes often demand that both top administrators and frontline workers reassess and adjust their practices continuously. This approach is the central and explicit thrust of Race to the Top’s “instructional improvement systems,” and it recurs prominently in all the statutes mentioned so far.
These statutes also strive to accommodate local variation. Increasingly, regulation leaves regulated actors discretion over practices as long as they achieve minimally adequate measurable results. Thus, for example, the Food Safety Modernization Act does not mandate that the Food and Drug Administration prescribe how processing plants should operate. Rather, it provides that each processor formulate its own safety plan and that the FDA then measure the efficacy of the facility’s efforts against a set of indicators. Moreover, the U.S. federal structure is significantly responsive to the demands of learning and context. All four initiatives pay homage to the idea of states as “laboratories of democracy” by giving them significant roles, according them broad discretion as to how to perform them, and encouraging peer exchanges or central aggregation of information in ways that promote learning.
Sunstein knows all this and occasionally refers in the book to non-nudge forms of regulatory loosening, but they are not central for him. Why not? I can think of three reasons.
First, Sunstein seems overly impressed by methodological rigor. The studies that have influenced him have the glamor of scientific form and the force of controlled empiricism. But rigorous control comes at a cost. It is impossible in some circumstances and expensive in most others, and results are often ambiguous. Small changes in the intervention studied might have produced very different results, and the experimental measures may not capture all the important effects. Drug regulation is the field in which rigorous empiricism has played the most central role, but even here recent legislation has loosened requirements because of concerns that they excessively delay or increase the cost of new therapies. Elsewhere, it is often more productive to have people with diverse perspectives and knowledge pool information informally. The most common form of empiricism in most policy contexts is peer review, in which knowledgeable people address problems or assess data deliberatively. In the business world, it is routine to re-assess strategy continuously in the light of whatever data is available without the benefit of formal empirical study. (Steve Jobs, whose commitment to “user-friendly” products Sunstein invokes, famously disdained market research.)
Second, Sunstein, both as a teacher of administrative law and an OIRA official, has been immersed throughout his career in an anachronistic model of administrative government. The model was developed in the aftermath of the New Deal. It combines the New Deal understanding of administration as a matter of rigid and stable regulations with the conservative suspicion of government initiative. The model was thus designed to make it difficult to promulgate regulations, but it also can make it difficult to change them or at least easy to leave them in place. OIRA review is just one stage in the promulgation process, which often takes many years. Recent administrations have thickened OIRA’s role by requiring that agencies show that a proposed regulation’s benefits exceed its costs. In Republican administrations, OIRA has often stalled or strangled regulations by making implausible demands for methodological rigor or biased estimates of costs and benefits.
Sunstein argues convincingly that cost-benefit analysis has no anti-regulatory bias in principle and is a helpful inducement to clarity and disciplined analysis. Nevertheless, the process remains heavily front-loaded. Efficacy must be rigorously demonstrated when a regulation is proposed, but there is little pressure to reassess once it is in place. This process is poorly adapted to situations, such as those that prompted the four initiatives mentioned above, where there is strong uncertainty about the nature of a problem or its solution. For example, because bankers constantly find new ways to transact around regulatory constraints, no rule can anticipate all the risky moves that should be banned. What’s needed is a rule that mandates ongoing disclosure, review, and adaptation. Assessments of the efficacy of such a rule will be most valuable after it has been implemented.
To its credit, the Obama administration, under Sunstein’s influence, showed some awareness of the front-loading problem when it encouraged agencies to review already-promulgated rules. But the review duty is vague and discretionary. The duty of upfront cost-benefit analysis is detailed and categorical. Sunstein’s preoccupation with choice as a response to uncertainty seems to reflect an assumption that administrative practice has limited capacity to learn and adapt. In fact, however, the process of regulatory enactment he oversaw is partly responsible for this limitation.
Third, the modes of adaptation that Sunstein slights tend to employ collaborative decision-making. In one example that can be found in contexts as disparate as nuclear power regulation and child protective services, frontline workers are expected to depart from the rules when following them would not serve their purposes. But when they depart, they must trigger a review in which their actions are assessed by peers and supervisors. When the departures are sustained, the rules are rewritten to reflect the enhanced understanding. Decisions at all levels of the process involve deliberation by groups, often interdisciplinary ones.
For pressing initiatives, the behaviorist toolkit is of limited use.
It is not clear what Sunstein thinks about such processes. Simpler is focused on individual decision-making. Sunstein sees individual decision-making as prone to error but also as corrigible with the help of behaviorist learning. He says nothing here about the psychology of group decision-making. Elsewhere, however, he has written of it with skepticism. In particular, he has explored the pathologies of “group polarization”—the tendency of deliberators to become more extreme in their views when they deliberate with like-minded others—and “groupthink”—the tendency of deliberators to repress or shape their views to conform to what they sense, sometimes inaccurately, is the dominant tendency in the group.
Sunstein’s discussion of the challenges of group decision-making does not have the meliorist tone he adopts with respect to individual decision-making. Yet there is a social science literature suggesting that the problems of group decision-making are also treatable. For example, disinterested conveners or facilitators can improve deliberation by inducing presentation of balanced information and eliciting a full range of views. This literature also suggests that group decision-making can mitigate some of the pathologies the behaviorists find in individual decision-making. For example, the need of each group member to articulate and explain her position to the others forces self-consciousness that disrupts reliance on the unreflective thought patterns the behaviorists document.
• • •
The nudge approach is influenced by ideological strategy as well as by social science. Sunstein seems to think that for liberalism to reclaim the support it has lost in recent decades the key task is to find common ground with the libertarian right. Hence he emphasizes the liberty-respecting dimension of choice-architecture regulation. A good part of the book engages libertarian critiques of government respectfully, indeed timidly. Sunstein also shares the libertarian focus on the danger of excessive, as opposed to insufficient, regulation. At OIRA, he enthusiastically implemented President Obama’s directive that agencies seek to identify “unnecessary” regulations for repeal or cutback, acknowledging only as an afterthought that there might be some regulations that should be strengthened.
Although Sunstein seems unchast-ened by it, there is evidence in his book that the ideological strategy is a failure. Libertarians have not been placated. Thirty-four Republican Senators voted against Sunstein’s confirmation. In a series of television rants, Glenn Beck portrayed Sunstein as “the most dangerous man in America,” attacking nudges as an insidious form of covert control. The economist Edward Glaeser made the same argument with less hyperbole: fiddling with choice architecture is more dangerous than enacting mandatory rules because the interventions are less noticeable and hence less likely to trigger political opposition.
Sunstein’s strategy misconceives where liberals should be looking for allies and what they need to do to win them. They should be looking in the center, not on the fringe, and the key to winning centrist support for liberal economic programs is to demonstrate their capacity to deal effectively with public problems, not to increase their accommodation of individual choice. Most Americans are not strong libertarians in economic matters. They do not see the capacity to choose among health insurance plans or to buy tax-free cigarettes as matters of liberty in any sense akin to rights of free speech, due process, political participation, or (for some) gun possession. They see choice in the economic domain largely in utilitarian terms. If regimes that allow choice leave most people better off, they are good. But choice should be readily sacrificed when doing so leads to more efficient provision of services.
Consider that the situation in current public discourse is virtually the opposite of that portrayed by Glaeser. Minor, indirect efforts to influence choices, such as Mayor Michael Bloomberg’s restriction on soda servings, often generate noisy debate about whether their trivial restraints on liberty can be justified. Because libertarian rhetoric is a kind of background music in American culture, debates about paternalism have a certain entertainment value. Yet massive and directly coercive programs are rarely attacked as infringements of liberty and are often taken for granted. Social Security is the standout example, but there are many others, including Medicare, unemployment insurance, workplace safety, securities regulation, and defective-product regulation. All these programs rest in substantial part on hard paternalist rationales. Yet when they are criticized, they are usually charged with ineffectiveness, not with infringing liberty. Even the right rarely attacks Social Security as paternalist anymore. Its complaints mainly assert that the program is inadequately funded and unlikely to deliver promised benefits. Social Security’s defenders spend most of their time showing that the program is sound or can be made so with modest reforms, not trying to make the program more palatable to libertarians.
The biggest current liability for liberals is that many people have lost faith in the capacity of government to solve the problems they care about. Perhaps the most prominent of these problems are unemployment, economic inequality, the deterioration of the natural environment, and national security. The behaviorist toolkit is not much help here. Sunstein’s account of the future of government has nothing to say about unemployment, inequality, or national security, and its contribution to environmental protection is limited to consumer labeling of cars and appliances. Sunstein is right that government needs to be sensitive to the limits of its knowledge and understanding and that intervention needs to be more flexible and adaptive. But it seems unlikely that many major problems can be solved without more direct intervention and more collective decision-making than his strategy contemplates.
Editors' Note: This article appeared in the July/August 2013 print edition.