How Change Happens

by Cass R Sunstein

  In support of that argument, it would be useful to focus directly on two kinds of consumer savings from fuel-economy standards, not involving externalities at all: money and time. In fact, the vast majority of the quantified benefits from recent fuel-economy standards come not from environmental improvements but from money saved at the pump; turned into monetary equivalents, the time savings are also significant. For the most recent and ambitious of those standards, the Department of Transportation found consumer economic savings of about $529 billion, time savings of $15 billion, energy security benefits of $25 billion, carbon dioxide emissions reductions benefits of $49 billion, other air pollution benefits of about $14 billion, and less than $1 billion in benefits from reduced fatalities.3 The total projected benefits are $633 billion over fifteen years, of which a remarkable 84 percent comes from savings at the pump and no less than 86 percent from those savings along with time savings.

  I have noted that on standard economic grounds, it is not at all clear that these consumer benefits are entitled to count in the analysis, because they are purely private savings and do not involve externalities in any way. In deciding which cars to buy, consumers can certainly take account of the private savings from fuel-efficient cars; if they choose not to buy such cars, it might be because they do not value fuel efficiency as compared to other vehicle attributes (such as safety, aesthetics, and performance). Where is the market failure? If the problem lies in a lack of information, the standard economic prescription overlaps with the behaviorally informed one: provide that information so that consumers can easily understand it.

  In this context, however, there is a risk that any kind of choice-preserving approach will be inadequate. Even with the best fuel-economy label in the world, consumers might well be insufficiently attentive to those benefits at the time of purchase—not because consumers have made a rational judgment that such benefits are outweighed by other factors, but simply because consumers focus on other variables.4 How many consumers think about time savings when they are deciding whether to buy a fuel-efficient vehicle?

  This question raises a host of empirical questions to which we lack full answers. But if consumers are not paying enough attention to savings in terms of money and time, a suitably designed fuel-economy mandate—hard paternalism, and no mere default—might well be justified because it would produce an outcome akin to what would be produced by consumers who are at once informed and attentive. If the benefits of the mandate greatly exceed the costs, and if there is no significant consumer welfare loss (e.g., in the form of reductions in safety, performance, or aesthetics), then the mandate does serve to correct a behavioral market failure. And indeed, the US government has so argued:

  The central conundrum has been referred to as the Energy Paradox in this setting (and in several others). In short, the problem is that consumers appear not to purchase products that are in their economic self-interest. There are strong theoretical reasons why this might be so:

  Consumers might be myopic and hence undervalue the long term.

  Consumers might lack information or a full appreciation of information even when it is presented.

  Consumers might be especially averse to the short-term losses associated with the higher prices of energy-efficient products relative to the uncertain future fuel savings, even if the expected present value of those fuel savings exceeds the cost (the behavioral phenomenon of “loss aversion”).

  Even if consumers have relevant knowledge, the benefits of energy-efficient vehicles might not be sufficiently salient to them at the time of purchase, and the lack of salience might lead consumers to neglect an attribute that it would be in their economic interest to consider.

  In the case of vehicle fuel efficiency, and perhaps as a result of one or more of the foregoing factors, consumers may have relatively few choices to purchase vehicles with greater fuel economy once other characteristics, such as vehicle class, are chosen.5

  To be sure, we should be cautious about accepting a behavioral argument on behalf of mandates or bans. Behavioral biases have to be demonstrated, not simply asserted; perhaps most consumers do pay a lot of attention to the benefits of fuel-efficient vehicles.6 The government’s numbers, projecting costs and benefits, might be wrong; recall the knowledge problem. Consumers have highly diverse preferences with respect to vehicles, and even though they are not mere defaults, fuel-economy standards should be designed to preserve a wide space for freedom of choice. The use of fleet-wide averages helps to ensure that such space is maintained.

  With these qualifications, the argument for fuel-economy standards, made by reference to behavioral market failures, is at least plausible. In this context, nudges (in the form of improved fuel-economy labels) and mandates (in the form of standards) might march hand in hand. With an understanding of behavioral findings, a command-and-control approach, promoting consumer welfare, might turn out to be far better than the standard economic remedy of corrective taxes.

  Less Risky

  The fuel-economy example is important, but it should not be read for more than it is worth. It certainly does not establish that in the face of human error, mandates are generally preferable to choice-preserving alternatives. As we have seen, such alternatives, above all defaults, reduce the costs of imposing solutions on heterogeneous populations, reduce the risks associated with government error, and avoid the many costs associated with eliminating freedom of choice. In light of the frequently unanticipated and sometimes harmful effects of mandates, default rules are generally less risky.

  No one should deny that in the end, mandates might turn out to be justified. But in a free society, it makes sense to give careful consideration to less intrusive, choice-preserving alternatives and, at least when standard market failures are not involved, to adopt a rebuttable presumption in their favor—with the rebuttal available only if it can be shown that a mandate will clearly improve social welfare.

  Notes

  1. See Sarah Conly, Against Autonomy (2012); Ryan Bubb & Richard Pildes, How Behavioral Economics Trims Its Sails and Why, 127 Harv. L. Rev. 1593 (2014).

  2. I am bracketing the question of definition, but note that freedom of choice is, by any reasonable account, an important ingredient in social welfare. See Daniel Benjamin et al., Beyond Happiness and Satisfaction: Toward Well-Being Indices Based on Stated Preference, 104 Am. Econ. Rev. 2698 (2014); Björn Bartling et al., The Intrinsic Value of Decision Rights (U. of Zurich, Dept. of Econ. Working Paper No. 120, 2013), http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2255992. For valuable discussion of foundational issues, see Matthew Adler, Well-Being and Fair Distribution: Beyond Cost–Benefit Analysis (2011).

  3. Nat’l High. Traf. Safety Administration, Final Regulatory Impact Analysis: Corporate Average Fuel Economy for MY 2017–MY 2025, August 2012, table 13.

  4. See Xavier Gabaix & David Laibson, Shrouded Attributes, Consumer Myopia, and Information Suppression in Competitive Markets, 121 Q. J. Econ. 505, 511 (2006).

  5. See Light-Duty Vehicle Greenhouse Gas Emission Standards and Corporate Average Fuel Economy Standards; Final Rule, Part II, 75 Fed. Reg. 25,324, 25,510–25,511 (May 7, 2010), https://www.gpo.gov/fdsys/pkg/FR-2010-05-07/pdf/2010-8159.pdf.

  6. Hunt Allcott & Michael Greenstone, Is There an Energy Efficiency Gap?, 26 J. Econ. Persp. 3 (2012).

  11

  On Preferring A to B, While Also Preferring B to A

  In the last quarter century, one of the most intriguing findings in behavioral science goes under the unlovely name of “preference reversals between joint and separate evaluations of options.”1 The basic idea is that when people evaluate options A and B separately, they prefer A to B; but when they evaluate the two jointly, they prefer B to A. Preference reversals of this kind have received far too little attention; they bear on many questions in law and policy. They also raise fundamental questions about choice, rationality, choice architecture, and the relationship between choice and welfare. In subtler ways, they tell us som
ething important about when change will occur.

  In many circumstances, people are making choices in separate evaluation. They are assessing an option: an appliance, a book, a movie, a policy option, a political candidate, or a potential romance. They focus intently on the option in isolation. In other circumstances, people are making choices in joint evaluation. They are assessing two options, and they compare them along one or another dimension. They focus intently on the two options and the particular dimensions along which they differ.

  Of course they might also assess three options or four or five hundred. The distinction between separate and joint evaluation might be unrealistic in situations in which the alternative to separate is better described as multiple. I will return to that point, but let us bracket it for now. The central idea is that in assessing an option, people may not be comparing it with other options in the same general vicinity.

  My principal goal here is to show that both separate and joint evaluation often lead to bad outcomes, though for intriguingly different reasons. To simplify the story, a characteristic problem with separate evaluation is a lack of relevant information. Some features of an option are hard to assess in isolation; their meaning, for life or actual experience, is unclear. For that reason, those characteristics may be ignored. By contrast, a characteristic problem with joint evaluation is the undue salience of a single factor (or a subset of factors). People focus on what distinguishes the two options, whether or not they mean much for either life or experience.

  I devote considerable space to elaborating these problems, which show why individual choices often go wrong in both markets and politics. Contrary to a widespread view, I argue that there is no justification for preferring separate evaluation to joint evaluation or vice versa. For purposes of producing good decisions and good outcomes, both modes of evaluation run into serious problems. We should seek structures that avoid the problems and pathologies associated with each. At first glance, global evaluation seems better than separate or joint evaluation, but that conclusion is too simple. Such structures must be based on an assessment of the goal of the relevant task: Is it to increase consumer welfare? To promote social welfare more broadly? To ensure optimal deterrence? An understanding of the relevant problems and pathologies paves the way toward identification of the appropriate structures.

  Consumer Goods

  For consumers, here is an example of the kind of preference reversal on which I will be focusing:2

  Dictionary A: twenty thousand entries, torn cover, but otherwise like new

  Dictionary B: ten thousand entries, like new

  When the two options are assessed separately, people are willing to pay more for B; when they are assessed jointly, they are willing to pay more for A. A prominent explanation for such preference reversals points to evaluability.3 In separate evaluation, it is difficult for most people to know how many words a dictionary should have or whether ten thousand is a lot or a little.4 For that reason, ten thousand and twenty thousand might be taken to be indistinguishable in separate evaluation and hence produce essentially the same willingness to pay. The numbers do not and indeed cannot matter, to the extent that it is difficult to know what they mean. By contrast, torn cover is clearly a negative, and like new is clearly a positive, even in separate evaluation. That characteristic looms large. Who wants a dictionary with a torn cover?

  The problem of evaluability, used in the psychology literature, is best understood in more conventional economic terms. It points to a lack of adequate information, which can be costly to obtain, and which people might not seek out even if obtaining it is not costly. In separate evaluation, insufficient information is a pervasive problem. (Whether it is rational to try to obtain it is a separate question, which depends on the costs and benefits of the effort.) Many characteristics of options—dictionaries, appliances, cell phones, jobs, people, cities—are essentially meaningless in the abstract. Consistent with standard practice, I shall use the term evaluability throughout, understanding it to refer to a lack of adequate information.

  Some characteristics that are hard or impossible to evaluate are numbers, whose meanings depend on context and background. For Geekbench 3 SC 32, a laptop might show a number of 3680, but it might be hard or impossible for consumers to know the consequences of that number for what they care about. With respect to some numbers (such as battery life), many or most consumers might have a good understanding even in separate evaluation. But even for such numbers, separate evaluation might not make it easy for people to make appropriate distinctions between impressive numbers (nine hours, ten hours) or less impressive ones (six hours, five hours). To overcome a problem of evaluability (understood as lack of adequate information), consumers must do some work, and they will often decline to do it.

  The problem of evaluability, in this sense, belongs in the same family as opportunity cost neglect: people might be willing to pay $X for a certain good, but not if they are focused on other things for which they might pay $X.5 Drawing people’s attention to opportunity costs is analytically similar to joint evaluation; it broadens the viewscreen to focus attention on a comparison to which they would otherwise be oblivious.

  For consumer goods, and indeed for countless options, evaluability presents a serious challenge—and at the time of choice, characteristics that can be readily evaluated will dominate people’s decisions. We can understand the behavioral phenomenon of present bias in similar terms.6 The present is often easy to evaluate; the future is often surrounded in some kind of cloud. A challenge of evaluability, with respect to future states of affairs, may well contribute to present bias. For choices in politics and law, evaluability of certain characteristics of options may be especially difficult—a point to which I will return.

  In joint evaluation, by contrast, it is easy to see that ten thousand words is less good than twenty thousand words. To consumers, those numbers greatly matter, and they become more meaningful in a comparative setting. Because the point of a dictionary is to define words, a dictionary with twenty thousand words seems much better than one with half that amount. If one dictionary has twice as many words as the other, who cares whether its cover is torn?

  Or consider this example7 (involving an admittedly outdated technology):

  CD changer A: can hold five CDs; total harmonic distortion = .003 percent

  CD changer B: can hold twenty CDs; total harmonic distortion = .01 percent

  Subjects were informed that the smaller the total harmonic distortion, the better the sound quality. In separate evaluation, they were willing to pay more for CD changer B. In joint evaluation, they were willing to pay more for CD changer A.8 Here too evaluability is the most plausible explanation. Even if an individual knows that a lower figure is preferable for total harmonic distortion, 0.01 percent sounds very low in the abstract, and it is not particularly meaningful. In separate evaluation, 0.01 percent and 0.003 percent might seem the same (low). But in joint evaluation, 0.003 percent is obviously much better. Apparently, people think that for a CD changer, what most matters is the sound quality, and hence they would willingly sacrifice on a relatively unimportant dimension (the number of CDs held) in return for better sound quality.

  Here is one more example, involving data from an actual marketplace:9

  Baseball card package A: ten valuable baseball cards, three not-so-valuable baseball cards

  Baseball card package B: ten valuable baseball cards

  In separate evaluation, inexperienced baseball card traders would pay more for package B. In joint evaluation, they would pay more for package A. Intriguingly, experienced traders also show a reversal, though it is less stark. They too would pay more for package A in joint evaluation. In separate evaluation, they would pay more for package B (though the difference is not statistically significant). For experienced traders, it is fair to say that in a relatively small population, the preference is for package A in joint evaluation, but there is no preference in separate evaluation.

  The
explanation is similar to what we have seen before, and it is straightforward for inexperienced traders. In joint evaluation, it is easy to see that package A is better. People get something of value that they do not get with package B. In separate evaluation, it is hard to assess the two packages or to know which one is better. Some people might use a version of the representativeness heuristic to downgrade package A; because it contains cards that are not so valuable, the entire package looks worse. As John List explains, this particular reversal, “examining real choices in actual markets,” produces a nice test of “the stability of preferences in riskless decisionmaking.”10 The central finding is that preferences are not stable.

  Politics

  Here is one more example, from a different domain:11

  Congressional candidate A: would create five thousand jobs; has been convicted of a misdemeanor

  Congressional candidate B: would create one thousand jobs; has no criminal convictions

  In separate evaluation, people rated candidate B more favorably, but in joint evaluation, they preferred candidate A. The best explanation should now be familiar. In the abstract, it is hard to know whether one thousand or five thousand jobs is a large number, and a great deal of work would be necessary to make sense of the difference. But a misdemeanor conviction is obviously a strong negative—which explains the relatively greater appeal of candidate B in separate evaluation. But in joint evaluation, most people think that the benefit of four thousand additional jobs outweighs the cost of having a member of Congress with a misdemeanor conviction, and hence candidate A prevails.