Douglas Ginsburg is Circuit Judge, U.S. Court of Appeals for the District of Columbia.
Joshua Wright is Associate Professor, George Mason University School of Law.
The behavioral economics research agenda is an ambitious one for several reasons. The first reason is that behavioral economics requires a theory “true” preferences aside from – and in opposition to — the “revealed” preferences of the decision maker. A second reason is that while collecting and documenting individual biases in an ad hoc fashion can generate interesting results, policy relevance requires an integrative theory of errors that can predict the sufficient and necessary conditions under which cognitive biases will hamper the decision-making of economic agents. A third is not unique to behavioral economics but is nonetheless significant: demonstrating that behavioral economics improves predictive power. The core methodological commitment of the behavioral economics enterprise — as with economics generally at least since Friedman (1953) — is an empirical one: predictive power. Indeed, no less than Christine Jolls, Cass Sunstein and Richard Thaler have described the behavioralist research program as the economic analysis of law “with a higher R-squared,” that is, “a greater power to explain the observed data.”
As I’ve observed previously, there are some good reasons to believe that behavioral law and economics (BLE) scholars do not share these methodological commitments. I’ve discussed previously the example of failure of BLE scholars to even cite, much less grapple with, the work of Zeiler & Plott (or here) regarding the endowment effect. Zeiler & Plott present and support the provocative claim that current evidence supporting the endowment effect is better explained by experimental procedures than cognitive biases. Proponents of regulation based on the endowment effect, in my view, need not agree with this interpretation of these findings but they ought to respond to them if they want to be taken seriously. Unfortunately, out of the 342 articles in JLR discussing the “endowment effect” from 2006 to present, only 35 cite either Zeiler and Plott article. I find that ratio discouraging for the discipline of behavioral law and economics generally and the prevailing level of discourse.
Indeed, while David Levine is not referring to the BLE literature, he might as well have been when he writes:
Behavioral economics: love it or hate it – there seems to be no middle ground. Lovers take the obvious fact people are not frictionless maximizing machines together with the false premise that economists assume that they are to conclude that all of economics must be wrong. The haters take the equally obvious fact that laboratories are not the real world to dismiss all laboratory evidence that conflicts with their pet theories as irrelevant. In the end they seem primarily to talk past each other.
How can we improve the discourse and get discussion focused on predictive power and consequences of actual behavioral policies proposed or implemented? The burden here lies with the skeptics. As Richard Epstein points out, the behavioralists’ message has been clear and effective; indeed, Bar-Gill and Warren’s article generated the Consumer Financial Protection Bureau. Behavioral skepticism has proven less effective.
Skeptics, including myself, have been decidedly less effective in convincing their respective audiences that specific behavioral proposals should be rejected and conventional economic approaches should (at least for now) prevail in the market for ideas in the academy and in the policy world. It is true that the skeptics have a number of forces working against them. One is that BLE is new and exciting. Arguing that the “conventional” approach outperforms the newest tool in the toolkit is always an uphill battle. I’ve alluded to a second reason, failure of at least some of the BLE literature to engage with opposing ideas. But perhaps most important is the failure of the skeptics to present a comprehensive and convincing case that the conventional economic approach systematically can be expected to outperform BLE when the full social benefits and costs of the various approaches and institutions are accounted for. I’ve long been of the opinion that two primary reasons for this failure are that different strands of the skeptical literature have talked past one another, and that this has led to a failure to present the “full” case against BLE on the record to be evaluated.
Consistent with this view, the goal of this post is not to present any new ideas about behavioral economics or behavioral law and economics, but catalog the various objections that have been raised in the literature, discussing their interactions, and linking to some of the leading scholarship in the area calling into question the assumed superiority of the BLE approach on a variety of grounds.
The taxonomy of BLE-objections involves four general categories: (1) theoretical opposition on the methodological approach adopted by the BE literature in defining “true” preferences; (2) a variety of empirical objections to experimental and field evidence offered in support of BLE; (3) biases in behavioral cost-benefit analysis; and (4) slippery slope objections.
1. The Allure of “True” Preferences. The BLE claim to welfare-increasing intervention is built upon the proposition that actual choice behavior might deviate in systematic and predictable ways from that predicted by an economic agent’s true preferences. This proposition alone is not objectionable. The behavioralists, however, go on to assume that the possibility of erroneous choice behavior requires one to disregard the no-error possibility, that is, that actual choice behavior is evidence of welfare. This necessitates the behavioralist search for “true preferences,” defined to exclude revealed preferences. Indeed, the premise of BLE is that interventions can be designed to bring actual choice behavior closer to these so-called true preferences. Because true preferences are not revealed by choice, the social planner must find and define them. Sunstein and Thaler’s approach is to define true preferences as those preferences that would be expressed by the economic agent in the absence of any behavioral bias. Sunstein & Thaler then proceed to claim these preferences are the proper measure of welfare consequences from individual decision-making. Some economists have objected that the behavioralists’ basis for identifying true preferences amounts to no more than the adoption of a set of arbitrary assumptions in place of economic theory, and facilitates an unwarranted presumption that errors in decision making are prima facie evidence of welfare-reducing choice. As Armen Alchian has observed along these lines:
It is not even necessary to suppose that each firm acts as if it possessed the conventional diagrams and knew the analytical principles employed by economists in deriving optimum and equilibrium conditions. The atoms and electrons do not know the laws of nature; the physicist does not impart to each atom a willful scheme of action based on laws of conservation of energy, etc. The fact that an economist deals with human beings who have sense and ambitions does not automatically warrant imparting to these humans the great degree of foresight and motivations which the economist may require for his customary analysis as an outside observer or ‘oracle.’.
This critique has led some economists, such as Pesendorfer & Gul, to entirely reject welfare analysis based upon the “multiple selves” models prevalent in the hyperbolic discounting literature, observing that “standard economics has neither need nor use for a welfare criterion that trades off utility among the various selves of a single individual.” They argue that the multiple selves model, in the hands of the behavioralists, provides “both an opportunity and a rationale for activism.” Similarly, Greg Mitchell makes the related point that evidence of irrational choice behavior cannot support conclusions about individual economic welfare, but “only that the irrational individual has failed to do what he or she most prefers, for rational choice theory employs an ordinal definition of utility that does not permit the kinds of external normative evaluations or interpersonal welfare comparisons that an objective measure of welfare, such as monetary wealth or healthiness, would permit.”
2. Empirical Objections. A second class of objections to the BLE enterprise is empirical rather than theoretical. Here, the objections fall into three sub-categories: (1) the utility of laboratory evidence in markets, (2) robustness of experimental evidence on its own terms, (3) and data interpretation and policy errors. The first category of objections is that while experimental evidence contains some value even if robust and reliable, documentation of these biases are of little benefit in informing policy analysis. Many — but not all — biases documented in laboratory settings disappear when exposed to market mechanisms, institutions, and the profit motive. These institutions create incentives for specialization and learning that reduces errors; these features do not exist in the laboratory. Indeed, one need not reject the existence of behavioral biases in order to raise doubts about the policy relevance of purely experimental results that have not been verified in field or quasi-market setting. As Edward L. Glaeser notes in Paternalism and Psychology, 73 U. Chi. L. Rev. 133, 140 (2006), “in experiments, individuals have few tools with which to improve their reasoning, and their only real method of responding to incentives is to think harder.” Unfortunately, BLE scholars have advanced policy proposals in many cases that ignore these critical distinctions and thereby stretch the existing evidence beyond what it can possibly justify.
A second sub-category of empirical objections involves challenging the robustness of experimental evidence on its own terms. I discussed the Zeiler & Plott paper above, and so I will not repeat that point here. But similar problems of robustness plague the literature on framing effects. As Gregory Mitchell has observed, while the existence of framing effects is not disputed, the effects are not robust to even small changes in experimental settings. For example, small manipulations in the decision-making context, such as asking choosers to think about the possible success or failure of options, to give reasons for their choices, or to deliberate more analytically, can reduce or eliminate the influence of framing effects. Mitchell also highlights evidence that stable preferences prevail in choice settings where choices are made frequently, involve less emotion, involve deliberation or reflection, involve a small number of options, or where the chooser is well informed. Further, framing effects might be reduced or eliminated at low cost without the sort of interventions proposed by the libertarian paternalists. Similarly, I’ve argued elsewhere that conventional economic theories of consumer contracts offer superior predictive power to their behavioral counterparts.
A third sub-category of empirical objections to BLE involves data interpretation and policy error. Consider, for example, the recent discussion of the work of Elizabeth Warren involving consumer credit. As Richard Epstein pointed out in his earlier post, and has frequently been a topic of conversation here at TOTM, Professor Warren has misinterpreted data as supporting the behavioralist approach to consumer credit.
3. Omissions in Behavioral Cost-Benefit Analysis. Some scholars have expressed skepticism about the behavioralist policy agenda on the ground that a more complete analysis of the long-run costs and benefits of paternalistic regulations would support a more limited role for government intervention. These scholars have pointed to specific costs of behavioral policy interventions — including the diminished incentive to learn and invest in cognitive capacity and human capital that would reduce errors and improve decision-making. Prominent critics in this area include Jonathan Klick & (again) Greg Mitchell, who have focused on the potential for behavioral intervention to create a particular form of dynamic inefficiency. Individuals have a greater incentive to invest time and money into human capital and assets that improve decision making when they bear the costs of their errors. This idea is not new, and has deep roots not only in economic theory, but also in libertarian philosophy observing that restraints upon human behavior that does not harm others impedes the individual development. Klick & Mitchell describe this cost of libertarian paternalism as a type of moral hazard, which in the long run raises error rates as people invest less in error-correction. Behavioralist cost-benefit analyses generally omit the potential for these dynamic costs when assessing the potential economic effects of proposed policy interventions.
A second error of omission in behavioralist analyses of policy interventions is underestimating or ignoring the cost of opting-out. The claimed “libertarian” aspect of behavioral interventions is that the manipulation of choice frames respects freedom of choice, and so the individual can always reject the regulator’s preferred choice in favor of expressing his own preference, even if irrational. As we discuss above, many of the proposed behavioral interventions simply do not live up to claim of “choice-neutrality” and, upon close inspection, can be seen to reduce the set of available choices. Many of the proposed behavioral policy interventions, including sin taxes and product bans, contemplate eliminating the ability to opt-out entirely, which overtly reduces available choices. Behavioralist policy analysis that simply assumes the cost of opting out is zero, or at least sufficiently low that it can be assumed to be negligible, skews in favor of intervention that may reduce welfare.
A third distinct sub-category of skewed cost-benefit analysis of behavioral intervention involves a tendency to underestimate the costs of government actors implementing the proposed policies. This particular objection to BLE proposals has been fairly broadly discussed in the legal literature, emphasized by Judge Posner (among others) and in Judd Stone’s most recent post in this symposium. The question is not whether governments or private individuals are irrational, or whether we can expect a government agency led by a group of behaviorally-biased individuals to implement error-free policy. The question is, as it always is when analyzing the relative economic merits of different institutional arrangements, “compared to what?” How costly will government policy errors be if government actors suffer from hyperbolic discounting, or status quo bias, or are sensitive to framing effects? What will be the frequency and magnitude of those errors relative to relying upon private decision makers to correct their own errors? Can we trust behavioral regulators suffering from confirmation bias reliably to identify the true preferences of individuals in order to implement behavioral polices? The correct focus is on the consequences of the choice to rely upon the government rather than private decision-makers to correct errors.
4. Slippery Slope Concerns. Perhaps the most often discussed objection to BLE proposals has been the issue of slippery slopes. These arguments, from the skeptics’ side, have been pressed by Mario Rizzo & Glen Whitman in a series of papers and are the subject of a recent CATO exchange on the topic. Paternalist policies are prone to slopes. And regulatory missions have a tendency to expand; mission creep assures that the government agency will need more “resources,” meaning money and staff, and forestalls the danger of its actually accomplishing its mission and becoming redundant. Mission creep is a concern regardless whether an agency’s purpose is paternalistic, but with a mandate to regulate conduct for the benefit of the regulated individuals, there is no end to the good an agency may attempt to do.
Smoking bans are a case in point. The government first determined that cigarette smoking is bad for one’s health and so advised the public. When the public did not respond adequately, i.e. , not everyone quit smoking, warnings were required on every pack of cigarettes, “sin” taxes were put on cigarettes so as to raise the price and thereby reduce the quantity consumed, and tobacco advertisements were banned from television. Notwithstanding the government’s paternal concern for their health, millions of people continued to smoke cigarettes. The government then publicized the hazard smoking posed to non-smokers, which provided a new rationale for banning smoking, namely, the externality it imposed upon others. The rationale was flawed, of course, because there was no gap in the relevant property rights: Patrons of restaurants and bars who did not want to be exposed to second-hand smoke could take their custom elsewhere; the expressed concern for employees of those establishments was similarly flawed in that, unlike the unfortunate subjects of the Soviet system, they were free to change their place of employment. Eventually, the ban on smoking in bars and restaurants was generalized to all indoor spaces and recently has spawned proposals to ban smoking even out of doors in some localities, notwithstanding the lack of any reason to be concerned with third parties.
Richard Thaler insists that “slippery slope” objections are themselves an irrational fear:
There are a lot of things out there to be afraid of, and there seem to be phobias named for each one. For example, you may not be familiar with bathmophobia, which is an abnormal and persistent fear of stairs or steep slopes, or a fear of falling. Less well known is “nudgephobia,” also known as the Whitman-Rizzo syndrome, which is the fear of being gently nudged down a slope while standing on a completely flat surface. This phobia is sometimes associated with other disorders such as the fear of being given helpful directions when lost; the fear of obtaining reliable medical advice when sick; and, in rare cases, some have even suffered from a fear of having someone recommend a book or movie that you will really like.
But the glib dismissal of these arguments as trivial for assessing the costs and benefits of a proposed behavioralist proposal appears unmoored from theoretical or empirical grounds. And while Thaler characterizes behavioral economics as about recommending movies, helping the sick, increasing savings, or putting healthy foods closer to the kids in the lunchline, he does not address head-on more serious concerns about more slope-prone behavioral policies including eliminating at-will employment, sin taxes, and prohibiting certain products outright.
With the likelihood of proposed behavioral interventions high and undoubtedly increasing over time, and with the development and improvement of the insights from behavioral economics continuing apace, vast improvements are called for with respect to the level of discourse in the legal academy concerning the evaluation of the full social costs and benefits of these interventions. One part of improving discourse in this area is to simply put on the table all of the relevant social costs and benefits in order to begin assessing their relative magnitudes. The BLE skeptics, as I’ve called them, have not carried their burden in this regard. Some discuss theoretical objections, others whether BLE can be applied without loss to liberty, others focus upon the details of slippery slope mechanisms, a few on the experimental and field evidence, and still others emphasize the costs of behavioral interventions on economic welfare.
The modest goal of this post is to provide a quick and dirty taxonomy of these various objections in order to spur discussion on the policy-relevant questions regarding a full social accounting of BLE interventions compared to feasible alternatives.