Paternalistic nudges and the inner rational agent

Richard Thaler and Cass Sunstein (Reference Thaler and Sunstein2008) maintain that psychological factors that ‘distort’ people's preferences can be used to implement paternalist policies without limiting freedom of choice. For example, one can improve the nutritional choices of customers in a cafeteria by displaying healthier desserts more prominently than unhealthy desserts. But does the proper arrangement of desserts improve people's choices? After all, individuals would make different choices with a different arrangement of desserts. By what evaluative criterion can Thaler and Sunstein describe ‘libertarian paternalist’ policies as benefiting those who are ‘nudged’ by structuring choice situations?

The answer they give is that successful paternalist policies ‘make choosers better off, as judged by themselves’ (Reference Thaler and Sunstein2008, p. 5). The behavioral economist can (with a sigh of relief) continue to defer to the individual's preferences. This claim is paradoxical, however, because claims about the benefits of nudges conflict with the preferences that are implicit in people's choices. The value judgments in terms of which paternalist policies make people better off are supposed to be the individual's own judgments, yet not the judgments that can be inferred from their unguided choice behavior. Sugden argues that behavioral economists square this circle by attributing to individuals a ‘corrected’ or ‘true’ preference ordering in addition to the preference ordering that is implicit in people's choices. What the individual truly prefers is the healthy apple, but his or her willpower is no match for the vivid perception of the gooey chocolate cake that beckons at eye level. Welfare is still preference satisfaction – individuals are still the judges of value – but true preferences rather than manifest preferences are the standards of value. On this view, normative economists can continue to avoid making any value judgments. They can still defer to the values of individuals. But behavioral economists must look beneath the surface of choice behavior to find the true preferences.

Sugden is critical of this move, in part because it apparently justifies paternalistic policies, to which he is vehemently opposed. He maintains that there are no grounds upon which to hypothesize some underlying set of true preferences that are free from the context dependence of manifest preferences. There is, he maintains, no justification for the behavioral economist's postulation of an inner rational agent whose welfare is constituted by the satisfaction of that agent's true preferences. Where, he asks, is the psychological evidence for the existence of true preferences? How are they formed or modified and protected from the sensitivity of actual preferences to psychological mechanisms triggered by framing and defaults? Is it remotely plausible to suppose that the features of human psychology that give rise to people's actual preferences have no influence on true preferences? Moreover, how are economists supposed to determine what the individual's true preferences are? Fruit is, after all, not always better for everyone than gooey chocolate cake.

In Sugden's view, behavioral economists have conjured up fictitious inner rational agents, whose preferences paternalistic policies satisfy. The case for paternalism collapses without these ghosts, ready to approve of the wise and benevolent policies behavioral economists endorse. If all policy-makers can go on are people's actual preferences, as manifested in their choices, then there is no way to tell whether influencing their choices might benefit them and no justification for paternalism, whether coercive or ‘libertarian’.

Sugden goes on to argue that the dubious fiction of an inner rational agent has confused the thinking of philosophers as well as economists. Among those who have allegedly succumbed are Brynn Welch and myself in our essay ‘To Nudge or Not to Nudge’ (Hausman & Welch, Reference Hausman and Welch2010).

Welch and I regard nudges directed towards benefitting those who are nudged as paternalistic because they aim to ‘shape’ the choices of an agent, A, so as to bring about what the nudger, N, judges to be best for A, even though A's choices suggest that A has a different view of what would be best. In shaping A's choices, N's nudging threatens the control that A has over her own deliberation and evaluation of the alternatives (which Welch and I call A's ‘autonomy’). We assume that, in the absence of the deliberative foibles that lead A to choose badly and that provide N with the means to influence A's choices, A is able to evaluate alternatives and (if she has the freedom and means) to choose in accordance with her evaluation. As Sugden (Reference Sugden2018, p. 64) documents, we maintain that ‘shaping choices risks circumventing the individual's will’. Sugden (Reference Sugden2018, p. 65) goes on to argue that this view commits us to the existence of an inner rational agent ‘encased in an error-prone psychological shell’:

Welch's and my talk of autonomy suggests to Sugden that we hold a notion of agents as somehow escaping the psychological processes that determine people's actual preferences and their choices, and Sugden's criticism would be justified if that were the only way to interpret what we are saying. But there is a more humdrum interpretation: just as I can circumvent your will by giving you false information or a mind-altering drug, so I can redirect your deliberation by the way in which I frame some issue. The actual empirical mechanisms that govern the development and expression of values are, as psychologists and behavioral economists have shown, influenced by all sorts of contextual features, and preferences are far from stable and consistent. But in most people repeated experiences of similar evaluative problems in a variety of contexts give rise to a reasonably stable set of socially recognized values and to projects individuals are committed to owing to those values. Although far from complete or free from inconsistencies, these values distinguish which of the manifest preferences of individuals reflect what matters to them and which conflict with an individual's values, owing to confusion, emotional upset, false beliefs and foibles in deliberation and choice. Many of people's choices fail to reflect their semi-settled values. This truism does not depend on any specific view of deliberation or autonomy, and it does not imply that there is some perfectly rational homunculus within our breasts warring with our manifest wants and emotions.

Moreover, crucially, Welch and I do not suppose that what promotes an individual's well-being can be determined by identifying the individual's inner rational self, without any evaluative commitments on the part of economists. As the final section will elaborate, the findings of behavioral economists conclusively undermine the prospect of delegating the determination of what promotes one's welfare to the individual himself or herself.

Opportunity and preference satisfaction

If people choose alternatives that are bad for them (as they sometimes do), and it is, as Sugden argues, impossible to measure welfare by the satisfaction of the true preferences of an inner rational agent, what is left for normative economics? Sugden gives a radical answer: rather than ranking policies by the extent to which they contribute to well-being, as indicated by proxies for well-being such as willingness to pay, Sugden argues that normative economists should assess institutions and policies by the opportunity they provide, as measured by the range of alternatives open to individuals. He goes on to state and prove theorems that are analogous to the fundamental theorems of welfare economics. Rather than demonstrating that idealized markets are Pareto optimal, Sugden argues that they are ‘opportunity optimal’ in the following sense: in equilibrium, potentially interacting individuals have the opportunity to engage in any feasible set of transactions.

In Sugden's view, one outcome X provides more opportunity than another, Y, if and only if the opportunities Y provides are a proper subset of the opportunities provided by X. Because this is a controversial view of opportunity, I shall attach an ‘S’ to the word ‘opportunity’ whenever I am referring specifically to Sugden's notion.

Since Sugden regards the findings of behavioral economics as breaking both the connection between preference satisfaction and well-being and also the connection between well-being and the assessment of outcomes, institutions and policies, one might have thought that preferences would drop out of Sugden's account of normative assessment. But this is not the case. Sugden instead argues that the fact that agent A prefers F to G provides A and others with a pro tanto reason to bring about F rather than G. Sugden maintains that satisfying preferences is of non-instrumental value.

Since Sugden believes that economic policies should be appraised by the opportunity they provide rather than by their success at satisfying preferences, one might wonder why he bothers to defend the non-instrumental value of preference satisfaction. The explanation lies in his view that S-opportunity is of value because it enables us better to serve our interests. If what serves our interests promotes our well-being (as I maintainFootnote ¹), then Sugden would not be challenging the standard view that normative economics aims to promote well-being. But Sugden maintains that we have an interest in satisfying preferences that is independent of any concern to promote well-being. Having recognized that welfare does not coincide with preference satisfaction and having also rejected the identification of welfare with the satisfaction of true preferences, Sugden sees no legitimate way for normative economics to invoke welfare in order to appraise outcomes, institutions or policies. In addition, he is opposed to invoking non-preference-based welfare considerations, because doing so could justify paternalism.

All the same, if S-opportunity is to be of value because it enables people to promote their own interests, Sugden needs some concept of interests. He proposes that individuals have an interest in the satisfaction of their preferences, regardless of any connection between preference satisfaction and welfare. A connection between S-opportunity and preference satisfaction would show that increasing S-opportunity is in the interests of individuals.

Thus, there are two central steps in Sugden's defense of S-opportunity as a criterion for the assessment of institutions and policies. He needs to show that preference satisfaction is of normative importance, even if it does not enhance welfare, and he needs to show that more S-opportunity never diminishes the extent to which individuals are able to satisfy their preferences.

Is preference satisfaction of non-instrumental value?

Sugden's argument that preference satisfaction is itself of value rests largely on three examples. Here is his first:

In accepting his friend's trust, Sugden has made a commitment to buy what his friend has asked for. That promise gives him reason to buy those items. Nothing in the example shows that his friend's preferences give him any additional reason to buy what his friend asked for. Indeed, the commitment does not depend on whether what his friend has asked for has any connection to his friend's preferences. She might, for example, have passed along some list of things given her by a family member without even looking at it. Moreover, we expect trustworthy friends to diverge from the shopping assignments we give them if they can benefit us by taking advantage of additional information concerning, for example, sales or food safety warnings. If his friend wants him to buy her a carton of cigarettes that Sugden thinks would be harmful to her, he would need to balance that harm against his duty to keep his promise, the displeasure his friend might feel at his not doing what he was asked to do and the effects of his presumption on his relationship with his friend. The example provides no argument that the preferences that may be implicit in his friend's request constitute reasons for Sugden to do as he was asked.

Sugden's second example is the following:

I disagree. The child's preference is not itself a reason to buy the toy. Suppose, unlike what the quotation assumes, the parent believes that the toy will give the child no pleasure. The parent predicts that, seeing the toy itself, rather than merely its picture, the child will be disgusted or frightened. Would there still be a reason to satisfy the child's preference? What gives the parent reason to purchase the desired toy is the parent's anticipation of the child's pleasure at receiving the gift (or the parent's fear of a tantrum otherwise), or perhaps the parent's concern for the child's developing a sense of control over her environment.

In a third example (Sugden, Reference Sugden2018, pp. 93–94), Sugden argues that if one shop contains more products than another and is no less convenient, then going to the larger shop will satisfy my preferences at least as well – whatever they are and regardless of their connection to my well-being or other worthy goals. In concluding that I have reason to go to the larger shop, I am, Sugden argues, taking the satisfaction of preference itself as of value. I value greater S-opportunity, because I value the satisfaction of preferences.

To simplify the discussion, suppose I am self-interested. The larger range of choices at the larger store is a reason to go there, because the additional goods may be better for me and, as stipulated, there are no disadvantages. What gives me reason to go to the larger shop is the costless possibility of benefit. Of course, if preferences indicate well-being, the larger array of goods at the larger shop will be at least as high up in my preference ranking as the goods in the smaller shop. But is this fact about my preferences an additional reason to go to the larger shop, beyond the advantages that the larger shop may offer? It must be, if Sugden is correct to claim that the satisfaction of preferences is itself of value. However, nothing about the case supports the claim that the satisfaction of preferences has normative significance in addition to or apart from the normative significance of what is preferred.

In all three examples, Sugden argues that preferences (of the friend, the child or one's own) are reasons for action, regardless of any connection they may have to well-being or other substantive values. In his view, one need look no further than preferences themselves to find normative reasons for action. Whether there are reasons to have those preferences is, Sugden maintains, a separate question. He concludes that providing people with the opportunity to act as they prefer should govern economic policy no matter what connection preferences have to welfare or any other value.

Sugden is apparently maintaining that economists should take the preferences of members of the population as determining the value of outcomes: the task of the policy economist is to figure out how best to follow the instructions of the population, no questions asked. Traditional welfare economists often came to much the same conclusion, because they took the preferences of individuals (as manifested in their choices) to determine what promotes their welfare. While the traditional view cannot survive the recognition that people may choose badly, Sugden's view purports not to depend on any connection between preference satisfaction and welfare or other values. If preferences themselves have normative significance, then in knowing them, the economist has a reason to favor some alternatives over others without regard to the considerations that gave rise to those preferences.

But once preferences are disjoined from welfare and other values, it is no longer plausible to maintain that normative economists should invariably defer to the preferences people reveal in their choices. Knowing what people prefer does not render the considerations that gave rise to the preferences irrelevant to policy assessment. Why might one think otherwise?

Sugden might think that this view follows from his contractualist vision of policy advice as addressed to individuals who seek terms of cooperation, rather than to idealized rulers. However, advisors may reasonably question the preferences of those they advise. Alternatively, one can argue – plausibly – that economists should, as a general rule, defer to the preferences of the populace, on the grounds that individuals are usually better informed about the details of their lives and more strongly motivated to promote their own well-being than are economists or government bureaucrats. But this argument is not available to Sugden, who places no weight on any connection between preference satisfaction and welfare. Moreover, counseling deference to the preferences of the population as a default is much weaker than asserting ‘that it is not the economist's business to enquire into his [an individual's] reasons for wanting what he wants’.

Sugden's stronger claim is not defensible. Although economists have no authority to overrule popular sentiments, they should criticize those sentiments when they are based on mistake or malevolence. In such cases, they may justifiably advise individuals to change their preferences and urge legislators to vote against the wishes of their constituents. For example, encouraged by President Trump, many Americans would like to limit immigration drastically, because they mistakenly believe that Americans would be more prosperous and less subject to violent crime if immigrants were kept out. Similar mistaken beliefs lie behind much of the support for Brexit in the UK. Economists should not ‘bracket out’ the question of whether there are good reasons for these policy preferences. On the contrary, when preferences depend on erroneous grounds, economists should attempt to convince members of the populace of their error, and, unless there are other better reasons in defense of the policy, they should oppose it.

This critique brings us back to the central question: If choice and preference are unreliable indicators of well-being, on what basis can economists argue that largely eliminating immigration will be harmful for most Americans or that Brexit will not benefit most of the residents of the UK? Without any reference to well-being, economists can, of course, make arguments about the effects of policies on the level and distribution of national income. But gross domestic product is important with respect to some values and not with respect to others. What gives point to the economic analysis are values such as welfare, freedom, fairness, compassion, equality and rights. Normative economics without evaluative commitments is purposeless. Some evaluative criterion must be specified in order to rank outcomes and policies and to design and refine institutions. Preference satisfaction does not provide that criterion.

Appraisal by opportunity

What about opportunity as a basis for economic evaluation? The fact that preference satisfaction has no non-instrumental value undermines Sugden's defense of his opportunity criterion, but there may be other ways to defend it. One might argue that some notion of freedom, closely associated with opportunity, is of non-instrumental value, or one might argue that opportunity is instrumental for welfare rather than for preference satisfaction.

Moreover, many contemporary liberal political theorists argue that state policy should be neutral in its intent with respect to competing views of what makes life worthwhile. Rather than being a nanny or an evangelist, the state should be an impartial facilitator, providing opportunities for residents to pursue what is of value to them. In my view, like Rawls’ (Reference Rawls1971), how well people live in large part depends and should depend on their own choices, and the state should not aim to take individuals by the hand and lead them to live well, by the lights of some benevolent state nanny. The objective of the state should be instead to provide individuals with the ingredients out of which they can construct good lives for themselves. So the state should make sure that political liberties are secured, that basic needs for security, food, shelter and health care are met and that vital inputs into a good life such as education, the social bases of self-respect and access to a wide range of careers are available to all.

I am accordingly broadly in Sugden's camp, emphasizing the importance of opportunities or capabilities rather than well-being as the criterion with which to assess social policy. Sugden may be right to argue for a transformation of normative economics, even if not for the reasons he gives.

There is, however, a serious problem with taking S-opportunity to be a criterion with which to evaluate outcomes and the policies and institutions that give rise to them: if economists are going to evaluate alternatives with respect to the opportunities that they afford to individuals, then economists need some way to measure opportunities. And, as a major debate a generation ago established, a relevant measure of opportunity cannot be independent of an appraisal of the values of alternatives.Footnote ² In The Community of Advantage, Sugden avoids these issues, concerning which he was himself an early and influential contributor (Jones & Sugden, Reference Jones and Sugden1982). Instead, he stipulates that, regardless of its value, adding an additional alternative to an individual's set of opportunities, O, without taking any alternatives out of O increases S-opportunity. He argues that policies and institutions should be assessed by the S-opportunity they provide, because greater S-opportunity is in the interests of individuals in virtue of sometimes increasing and never diminishing the extent to which people's preferences are satisfied.

Sugden's defense of competitive markets rests on his argument that they satisfy what he calls ‘the Generalized Strong Interactive Opportunity Criterion’. This says, ‘[O]f every transaction that is feasible for all consumers collectively and is non-dominated for each consumer separately, that each consumer's component of that transaction is an element of his overall opportunity set’ (Sugden, Reference Sugden2018, p. 131). His argument shows that the market's contribution to opportunity has a wider significance than maximizing S-opportunity. But S-opportunity remains his criterion for assessing public policies.

There are three problems with this criterion. First, opportunity sets that are not nested are not comparable with respect to opportunity. That means that an increase in S-opportunity is not necessary for an increase in preference satisfaction (or, intuitively, an increase in well-being). An extremely limited opportunity set, tiny-O, would be no worse than some enormous set of extremely valuable alternatives if tiny-O contained even just one alternative X that is unavailable in big-O. Assessing policies with respect to S-opportunity will often fail to distinguish among policies and institutions whose consequences for preference satisfaction and welfare differ enormously.

Second, the greater S-opportunity that results from adding new options that are nowhere near the top of anybody's preference ranking does not enable individuals better to satisfy their preferences. Sugden can respond that an increase in S-opportunity is at least never worse with respect to preference satisfaction. Against this claim, one can point to the losses that very often accompany the expansion of opportunity sets. In response, Sugden can reasonably reply that if there are such losses, then there has not been any increase in S-opportunity. For example, if the opening of a Walmart on the outskirts of town, with its wider variety of goods, eliminates the option of walking to the corner clothing shop and schmoozing, then it provides no increase in S-opportunity.

Although this response successfully defends the claim that greater S-opportunity never diminishes preference satisfaction, it is a pyrrhic victory, because it highlights the fact that additions to opportunity sets very often at the same time eliminate opportunities. Opportunity sets are rarely nested, and consequently, they are rarely comparable with respect to S-opportunity.

Third, the thesis that more S-opportunity never diminishes preference satisfaction runs into problems when one considers the sequencing of policies. Suppose that at t₀ there is a set of joint opportunities O. Policy P’ is adopted, which results at t₁ in a joint opportunity set O’ that is an increase in S-opportunity over O. O’ does not diminish the extent to which preferences are satisfied – so far so good. At t₂, policy P* is proposed that, if adopted, would result in the opportunity set O*. If P* had been introduced before P’ had been implemented, it would have been adopted, because O* increases S-opportunity over O. But O* does not include some opportunities in O’, which results from P’, and O* is hence not an increase in S-opportunity over O’. With respect to both preference satisfaction and an intuitive appraisal of the worth of the opportunities, it would have been much better if P’ had not been adopted at t₁, so that P* would have been available as an improvement at t₂. For example, owing to their implications for the Earth's climate, the enormous expansions of opportunities in the twentieth century provided by petroleum-based technologies may drastically limit opportunities in later centuries. Although Sugden's opportunity criterion permits an important defense of the market, it is inadequate as a guide to policy choice.

A modest proposal for a welfare economics informed by behavioral economics

Behavioral economics has shown that preferences are an unreliable guide to well-being, but an unreliable guide may still be of value. In my view, it is premature to conclude that welfare should have no role in guiding social policy or that preferences provide no guidance concerning welfare. An understanding of the circumstances in which preferences tend reliably to indicate well-being (which presupposes a substantive conception of well-being) makes it possible to make well-founded judgments concerning whether policies enhance well-being. Without supposing that there is a value-free way to determine the ‘true’ preferences of an inner rational agent, economists can use the findings of behavioral economics coupled with some weak premises concerning what conduces to well-being to identify factors that sometimes make preferences poor indicators of well-being. When these factors are not present, preferences are useful (albeit fallible) guides to well-being. Building on Hausman (Reference Hausman2012), one can identify four such factors:

1. (1) False beliefs. Suppose (to use one of Sugden's own examples; Reference Sugden2018, p. 73) Ann aims to eat the least fattening dessert and chooses the cake at 200 kcal over the fruit at 250 kJ, not realizing that 1 kcal = 4.184 kJ. In that case, Ann's preference for the cake provides no grounds to conclude that the cake would be better for her.

2. (2) Poor evaluative judgment. If Ann is foolish and irresponsible, confused by probabilities, a young child or cognitively disabled, then her preferences will not be a reliable guide to her well-being.

3. (3) Temporary incompetence. Ann might be subject to some condition that temporarily damages her judgment. Perhaps she has taken drugs or she is emotionally distressed or sleep deprived. Framing and contextual features might also elicit some serious cognitive flaw.

4. (4) Non-self-interested preferences. If Ann chooses the last piece of cake maliciously to spite her husband, who loves cake (even though she knows there will be a nasty row as a result), there is no good reason (apart from possible schadenfreude) to infer that the cake is better for her.

Let us say that when none of the first three disabling factors is present, then Ann is a ‘competent evaluator’. When the fourth factor is also absent, then Ann is both a competent and a self-interested evaluator. It is plausible to maintain that, other things being equal, Ann's preferences are a reasonable guide to her welfare when she is a self-interested, competent evaluator. If her preferences were bizarre or they were constantly fluctuating, one would retract this judgment, but not the general view that information about preferences can sometimes guide judgments about what is good or bad for people. Notice that the inferences concerning Ann's well-being rest on regarding her manifest preferences – albeit under the right conditions – as a guide to her welfare. Such a view is only justified if one has some conception of welfare.

In this view, unlike Sugden's, preferences matter to the appraisal of policies because preferences are often readily available, and, in some contexts, they are reasonably reliable indicators of well-being. In many circumstances, individuals have well-justified beliefs about the relevant facts, they are self-interested, they are competent judges of the relevant values and they are well placed to make good value judgments. When these conditions are satisfied, their preferences are reasonably reliable indicators of what is good for them, and economists who aim to promote welfare can find guidance in people's preferences.

As already mentioned, the assertion that preferences are in some circumstances a good indicator of well-being must rest on some conception of well-being. It cannot be the view that well-being consists in preference satisfaction, which implies that satisfying preferences always enhances well-being. It could be a hedonist view that well-being consists in pleasure. But there are serious philosophical objections to the view that pleasure and the absence of pain are the sole non-instrumental goods. There is, I assert (here, without argument), no well-established philosophical theory of well-being that economists can rely on.

Rather than committing themselves to any philosophical theory of well-being, it seems to me that economists are better advised to rely on nothing more than truisms concerning well-being that are widely accepted (and indeed form the ‘evidence’ against which philosophical theories of well-being are tested). These truisms, which I call a ‘folk theory’ of well-being, consist of claims about what things make life better, which are more important, how we can know about values and how values should influence actions. For example, when people are enormously upset, they are much more likely to make choices that have consequences such as giving offense to close friends or exposing themselves to risks of physical harm. On the basis of the folk theory of well-being, one can make the commonsensical value judgment that consequences such as these are generally harmful to individuals. The facts about the consequences of emotional upset coupled with these value judgments ground the conclusion that preferences are not good indicators of well-being when individuals are emotionally upset.

Truisms concerning well-being coupled with an appreciation of the circumstances in which preferences are reasonably reliable indicators of well-being permit nothing more than coarse-grained judgments concerning when alternatives are better for most individuals. There is little guidance to be found here for difficult individual decisions. Since the state is not, in my view, in the business of attempting to run the lives of individuals, the limitations in the abilities of its agents to determine which alternatives are better or worse for individuals are not as disabling as they might at first glance appear to be. There is little question that everyone, or almost everyone, has certain needs that must be met if individuals are to live well: nutrition, housing, clothing, health care, security, social and intimate relations, recreation and some control over one's environment. Controversy is inevitable concerning what it takes to satisfy these needs and how to make tradeoffs between more fully satisfying some and less fully satisfying others. As Sugden urges, there is a great deal to be said for relying largely on the market within an institutional framework that secures people's civil and political rights and provides a backstop for those whom the market does not serve well. Attention to civil and political rights also places procedural limits on the consideration and promulgation of policies and regulations.

Although direct democracy is not feasible in a large state and, owing to unavoidable ignorance and mistake, is prone to error, the preferences of the population must have a large role in determining the policies of a liberal representative democracy. Policy would be more sensible if the only preferences that count were those of competent evaluators, but it is not easy to distinguish the competent sheep from the confused goats, and there are good procedural reasons why the voices and votes of both should count.

Mainstream welfare economics can proceed cautiously, because in the right conditions, preferences are a good guide to well-being and choices are often a good guide to preferences. Moreover, without the need to conjure into existence inner rational agents, behavioral economics, conjoined with the folk theory of well-being and an everyday understanding of deliberation, provides guidance concerning when to regard preferences as reliable indicators of well-being, how to elicit less distorted preferences and how to correct for the distortions caused by false beliefs and deliberative foibles. Life is complicated for behavioral welfare economists – but then I conjecture that few ever really believed that there were no normative challenges to be faced in the conduct of welfare economics.