A Political Justification of Nudging

Abstract

Thaler and Sunstein justify nudge policies from welfaristic premises: nudges are acceptable because they benefit the individuals who are nudged. A tacit assumption behind this strategy is that we can identify the true preferences of decision-makers. We argue that this assumption is often unwarranted, and that as a consequence nudge policies must be justified in a different way. A possible strategy is to abandon welfarism and endorse genuine paternalism. Another one is to argue that the biases of decision that choice architects attempt to eliminate create externalities. For example, in the case of intertemporal discounting, the costs of preference reversals are not always paid by the discounters, because they are transferred onto other individuals. But if this is the case, then nudges are best justified from a political rather than welfaristic standpoint.

1 Introduction

A “nudge” is a policy intervention that targets the environment in which citizens make decisions. The environment ideally is manipulated in such a way as to preserve the full range of options that citizens already have, that is, without reducing their freedom of choice. Subtle features of the “choice architecture”, however, are arranged so as to facilitate the choice of options that are beneficial to the decision-makers. Nudge policies typically remove psychological biases that prevent people from making the right decisions, or use the biases to direct behaviour towards better options. Because choice architects aim at improving people’s lives without reducing their freedom, the policy of nudging has been advertised using the seemingly paradoxical label of “Libertarian Paternalism”.¹

Philosophical discussions of Libertarian Paternalism so far have mostly focused on this oxymoron. Critics have argued for example that nudges may preserve so-called option-freedom, but surely infringe the autonomy-freedom of individuals. Since option freedom without autonomy is hardly valuable, a genuine libertarian should be unhappy with the policy of nudging (Hausman and Welch 2010, Grune-Yanoff 2012). Another line of criticism has focused on paternalism: the problem is that the policies advocated by nudgers are aimed at improving the well-being of nudged citizens, but well-being is defined with respect to the preferences of the citizens themselves.

In our understanding, a policy is ‘paternalistic’ if it tries to influence choices in a way that will make choosers better off, “as judged by themselves” (Thaler and Sunstein 2008: 5).

This understanding is idiosyncratic: genuine paternalism imposes a view of well-being on recalcitrant subjects, while nudges simply help people achieve what they want (e.g. Hausman and Welch 2010). In fact Thaler and Sunstein’s defense of nudging is welfaristic, rather than paternalistic. Policies should be assessed on the basis of their consequences for people’s well-being, and well-being depends on the satisfaction of people’s preferences.²

The problem is not merely semantic, as we shall see shortly. One consequence of Sunstein and Thaler’s strategy is that nudges must help nudged individuals to attain what they want. But some of the policies advocated by libertarian paternalists are not welfare-enhancing in this sense: by inducing people to become organ donors, for example, policy makers intend to benefit the potential recipients rather than the nudged donors (e.g. Thaler and Sunstein 2008, ch. 11). So a welfarist argument seems unable to justify the whole range of policies that choice architects propose. But there is more: we shall argue that welfarism fails to justify even those policies that seem to be designed to help nudged individuals, for example by correcting intertemporal inconsistencies caused by choice myopia. Part of the problem is that the identification of people’s preferences is extremely difficult, if not impossible, when choice myopia is involved. So the decision to nudge people in one direction rather than another cannot be justified in welfaristic terms. Two options seem to be available at this point: on the one hand, one may take a genuinely paternalistic stance and argue that nudged individuals are better off independently of their preferences. On the other, it is possible to argue that choice myopia generates costs that are not paid entirely by the decision-makers. In the real world the costs of biases are often transferred onto other individuals. But if this is the case, then the biases create externalities and nudges should be justified from a political rather than welfaristic standpoint.

The paper is organized as follows: section two provides a quick overview of standard justifications of government intervention in the economic realm. This section is largely descriptive and paves the way for the behavioural economics account of market failures, in section three. Section four illustrates a specific example, focusing on temporal discounting. The main thesis of this paper is formulated and defended in sections five and six, where we argue that biases like temporal discounting can create externalities and therefore nudge policies are better justified from non-welfaristic premises.

2 Market Failures and Policy Intervention

The justification of government intervention is a central topic in contemporary political economy. When, where, and how much interference with individual choice is legitimate? On the far right of the political spectrum, libertarians support the idea of a legal system exclusively concerned with non-economic matters, and a minimal apparatus of state regulation in the economic realm. Libertarians rely on theorems proving that, in the appropriate circumstances, free markets deliver allocations of goods that are efficient in Pareto’s sense. The proviso is important, because neoclassical economics has devoted significant resources to the analysis of “market failures”. Classic cases include the existence of natural monopolies; the transaction costs caused by asymmetric information; externalities generated by the impossibility to attribute property rights; and the existence of non-excludable, non-rival public goods.³

The neoclassical theory – through the fundamental theorems of welfare economics – suggests that state intervention can be extended to correct these imperfections, with the exclusive aim to restore efficiency. According to another, more radical approach, markets may be inefficient even when they operate in ideal conditions, unencumbered by information asymmetries, externalities, or monopolies (e.g. Keynes 1936). And finally, it is possible to argue that market outcomes must be corrected on the basis of purely normative considerations (to reduce poverty or inequality, for example).

In practice, however, the “political economic” justification of the welfare state system of contemporary Western countries is a blend of arguments taken from neoclassical economics and from Keynesian theory, with a pinch of normative advice. In this context, Behavioural Economics and Libertarian Paternalism have introduced a new set of considerations, in particular the idea that market failures may be caused by biases of individual decision-making. The source of the problem here is not an architectural flaw of the market but a widespread failure of rationality. Individual agents may fail to behave according to their “true” preferences. The interventions prescribed by behavioural economists therefore are aimed at improving subjective well-being by replacing irrational agents with well-behaved ones. Nudging, in this specific sense, may be considered an ally of the neoclassical (welfaristic) approach to the analysis of market behaviour.

3 Bounded Rationality and “Genuine” Preferences

An irrational choice, according to behavioural economists, is a choice that does not reflect a consistent structure of preferences. The work of contemporary behavioural economists is in many ways a descendent of the theory of bounded rationality developed by Herbert Simon in the 1950s. But according to Simon (1955), the bounded decision maker is not incoherent: rather, she is forced by her cognitive limitations to consider only a subset of the available alternatives. Because she cannot compare more than a certain number of alternatives, the agent must define a boundary for the comparative process and then choose within this boundary. So Simon’s decision maker does not suffer from internal inconsistency of preference: any apparent behavioural incoherence that we observe is due to a change in the alternatives that are included in her opportunity (sub)set.

Contemporary behavioural economists, in contrast, make stronger claims:

Our major emphasis is on the fact that in many domains, people lack clear, stable, or well-ordered preferences. What they choose is strongly influenced by details of the context in which they make their choice, for example default rules, framing effects (that is, the wording of possible options), and starting points. These contextual influences render the very meaning of the term ‘preferences’ unclear. If social planners are asked to respect preferences, or if they are told that respect for preferences promotes well-being, they often will be unable to know what they should do. (Sunstein and Thaler 2006: 233)

While Simon’s agent has a clear preference ranking but is bound to consider only part of the option space, Sunstein and Thaler’s agent suffers from a deeper failure of rationality that puts the very notion of preference in doubt.

What is the link between choice consistency and the notion of preference? Intuitively a preference is a judgment that, all things considered, x is better than y, or, following Hausman (2011), total comparative evaluations of the consequences of actions. There are three important elements in this definition: first, a preference is an evaluation. Second, it is comparative, not absolute. It must involve at least two items and is expressed via a ranking. Third, it is a total evaluation: nothing that matters must be excluded. The “total evaluation” requirement is a consequence of the principle of independence, which is crucial for the stability of the preference relation.⁴ But the independence principle has been shown to be empirically violated in a number of experiments, which have fuelled behavioural economists’ scepticism concerning preferences.

Violations of independence may be caused by failures to consider attributes that matter to the decision-maker. Intrinsic inconsistencies at the level of preferences therefore may result from a mechanism that is in many ways similar to the one discussed by Simon. A boundedly rational agent in fact may not only be unable to search the full space of options; she may also be unable to search the full space of attributes of each option. The latter is especially relevant if the options are complex, or if choices have ramifications with uncertain future consequences. A limited human being in such circumstances may only be able to consider the most salient characteristics of each option, overlooking others. If salience varies across time or across circumstances, choice inconsistencies may occur.⁵

4 An Example: Temporal Discounting

In the course of the paper we shall often refer to a bias that plays a prominent role in the libertarian-paternalistic literature, namely temporal discounting. Thaler and Sunstein devote many pages to describe programmes – like “Save More Tomorrow” – designed to nudge people toward putting more money in their pension schemes. These programmes rely on techniques, such as the manipulation of default options, that influence choice without reducing substantially the freedom of decision-makers. In general, nudges try to curtail our tendency to sacrifice future well-being in order to satisfy immediate wants. This tendency of course is not necessarily problematic, unless people come to regret their past decision to consume rather than save. But when this happens, temporal discounting leads to temporal inconsistency of preference and inconsistent choice.

Temporal inconsistency can be explained along the lines illustrated in the previous section, using Construal Level Theory (CLT), a framework developed by Nira Libermann and Yaacov Trope.⁶ When a choice produces consequences in the future, the decision-maker must construct a mental representation of the future events that may be influenced by her decision. The level of abstraction of this mental representation, according to CLT, depends on temporal distance: the further in time the event is, the more abstract its representation. This cognitive process can be described by positing that future events are characterized by several attributes. Because of her cognitive limitations, the decision maker cannot evaluate all of them, but will rather select a subset that she will consider for her decision. If we accept the empirical evidence reported by CLT, the selection of the subset of attributes depends at least in part on temporal distance. As time goes by, the composition of the subset changes, and choices may change too. This process may lead to a classical situation of inter-temporal incoherence.

Suppose for example that Ann is facing a saving decision early in her life. One option is to spend immediately all the extra money she is earning, buying for example holiday trips around the world. Another option is to invest in a pension fund. Let us imagine that, if she decides to save, her youth will be fairly boring compared to the fun she could have if she decided to spend the money on holidays and other treats. But if she saves, she will be able to rest and enjoy a wealthy retirement later in life. If she spends it all, in contrast, she will face the alternative between working until a much later age or retiring with a low pension.

Ann’s decision has several different implications (attributes). To simplify, we limit the analysis to six attributes on three different levels or dimensions (fun vs. boring youth, wealthy vs. poor retirement, resting vs. toiling in old age). Let us suppose that Ann’s preferences along each dimension are organised as follows:

• She prefers to have Fun rather than to have a Boring youth (F > B);

• She prefers to be Wealthy rather than Poor later in life (W > P);

• She prefers to Rest rather than Toil when she is old (R > T).

A rational agent should be able to form a coherent preference ranking across all the combinations of these attributes, or possible life-styles. But like most people Ann has cognitive limitations. Let us assume that she can only manage to consider two attributes at a time. We shall also assume that she suffers from the myopia of CLT: when the consequences of a decision (attributes) are far away in time, they become less salient and tend to fade. As a consequence, her choices are going to be affected by the subset of attributes that are cognitively salient.

At the time of making her saving decision (t₁) only the first dimension (F or B) is salient. But the choice made at t₁ will determine which options are available later in life (at t₂):

• If Ann chooses B, then the set of options at t₂ is {BWR}, {BWT}, {BPR}, {BPT};

• If Ann chooses F, then her set of options at t₂ is {FWT}, {FPR}, {FPT}.

CLT predicts that at t₂ the attributes F and B will cease to be salient, and Ann will focus on the second and third attributes only. Let us suppose that her preference ranking over these pairs of attributes is {WR} > {WT} > {PR} > {PT}.

Her preferred option is to be wealthy without having to work; second-best is to toil in order to be wealthy; then to rest and be poor; worst of all is to be poor and toil. But notice that the option WR is not available, if she has chosen F when she was young. At t₂ nevertheless Ann can still compare the option BWR she had at t₁ with the best option she has now (FWT): since B and F are not salient anymore, and WR > WT, at t₂ she regrets the choice that she has made when she was young.

Notice that Ann’s choice at t₁ has narrowed her options at t₂. Giving up R is the cost that she has to pay at t₂ in order to reverse her previous decision. Initially she could have had both W and R (at the cost of F); at t₂ she cannot. If she wants W she must pay a cost in terms of R. Psychologists’ examples in the CLT literature tend to obscure this point, by focusing on seemingly reversible choices. But some choices are strictly speaking irreversible: you cannot go back in time. The option WR is unavailable to Ann after she has chosen F. What Ann can do is pay a cost to exchange PR for something she likes better (at t₂). But she cannot decide to have F and R simultaneously.

Ideally, a rational agent should be able to aggregate or arbitrate among the various dimensions of the attributes. She should be able to build a single utility function assigning specific weights to the various dimensions of the consequences. So in order to help Ann choose what she really wants, the choice architect should somehow identify her intertemporal utility function. But how? In practice, choice architects typically decide to privilege the preferences of old Ann, nudging young Ann to make decisions that will be more likeable to her future self. In a CLT perspective, for example, one may reduce the effect of temporal distance by enhancing the salience of some attributes that will become important in the future. This is equivalent to help people form a comprehensive utility function by “putting the child in the adult’s shoes”. But in principle the choice architect might as well put elderly people in the shoes of their younger selves; arguably one has a lot more fun traveling, partying, spending money when she is young than when she is old. Provided she keeps fond memories, old Ann might be helped or nudged to see her life in such a way that she feels no regret for earlier decisions. Which way should we go?

Clearly we are not claiming that people should be encouraged to be irresponsible when they are young. The example is only meant to show that the choice between different nudges seems arbitrary. An old self may be nudged to empathise with a young self, or the other way around. The literature on choice architecture hides the problem under the carpet, by simply assuming that we ought to help youngsters to save more. But nothing in the behavioural economics or libertarian-paternalistic framework justifies this assumption.

This does not hold for all types of irrationality, to be sure: in some cases it is relatively easy to identify “true” preferences and to justify policies aimed at restoring consistency. In typical cases of akrasia and addiction, for example, the preferences of the decision-maker are stable for long stretches of time, both before and after the consumption opportunity has become available; only for a short period, when the consumption opportunity is approaching, there is a temporary reversal of preferences. In such cases, explaining inconsistencies as temporary deviations from rationality seems to make sense.⁷ The case of saving is different – and challenging for the libertarian paternalist – because there really seem to be two stable agents with different preference orderings, hence two equivalent ways of restoring coherent preferences; moreover, there is no reason to privilege one of them, based on purely welfaristic considerations. One could of course decide that old Ann is right and young Ann is wrong, but this would imply taking a paternalistic stance in the standard sense of the term – a stance that Thaler and Sunstein apparently are unwilling to take.

5 The Politics of Nudging

We suspect that this conundrum is the consequence of taking the wrong perspective, right from the start. Let us step back and reflect on what nudgers have been trying to do: the supporters of nudge policies seek a solution that is in some sense a-political. Libertarian Paternalism is welfaristic to the extent that its policies promote the well-being of the nudged individuals. “I – says the choice architect – intervene because I care about your well-being”. But unfortunately, as we have just seen, it is not entirely clear what is the “well-being” to be maximized. If nudgers, like most economists, endorse a subjective view of well-being (as equivalent to the satisfaction of people’s preferences), they end up in a dead-end alley.

Since we sympathize with the general spirit of nudge policies, we propose a simple change of perspective: we shall argue that seeking a political justification of nudging may work well for those who are unwilling to take a genuine paternalistic stance. The idea, roughly, is that choice architects are often justified to intervene to protect other people from the damage that may be caused by irresponsible individuals. Nudge policies are not (or not only) for the good of the nudged, but for the good of third parties that otherwise are going to be harmed.

To build this argument we have to reconsider the costs of preference reversals. The examples examined so far presuppose that all the costs are internalised. Once Ann has chosen Fun, for example, she can only have Wealth at the expense of Rest. She pays the full cost of her myopia. But there are reasons to believe that things may not always go this way. Real people are skilful at sharing the costs of their mistakes. If somebody else can be made to pay, then myopic Ann may be able to have fun and wealth and rest. The best of all worlds.

How do you make others pay? The standard way in modern democracies is to form a coalition and legislate. This is particularly easy when the “others” cannot vote because they are not even born: irresponsible spenders can make their children and grandchildren pay the debts accumulated by previous generations. The one who pays for Ann’s past mistakes must not always be Ann’s future self; it may also be somebody else’s future self.

This brings a new philosophical perspective on the issue of nudging, because it opens the door to the classic arguments for policy intervention that we have reviewed in section two. The claim that people ought to be nudged is easily defensible once we realize that saving for the future is not entirely a private matter. The problem with not saving enough is not merely that people will regret it when they will be old. It is also that we will have to deal with plenty of old people who cannot support themselves. Seen from this perspective, the focus on preferences and paternalism shifts the spotlight away from some important issues. The key political problem is that a society with a large number of elderly people who cannot support themselves is not viable. The elderlies would burden younger people with large costs, and the latter would try to resist, generating conflict and unrest.

This argument applies in principle to any policy intervention that tries to correct biases of intertemporal choice. Most cases discussed by libertarian paternalists, including smoking and dieting, are potential sources of externalities. If these behaviours were entirely private, the case for nudging would be much weaker. But although smoking, dieting, and saving undoubtedly affect the welfare of each individual smoker, eater or saver, they also have non-negligible effects on others. In our own country, to take a concrete example, a couple of generations raised with unrealistic expectations of public welfare support have burdened another couple of generations with a mountain of public debt, plus the prospect of low employment and low pensions. This situation is not easily redressed, because as younger people in Italy have learnt all too well, the votes of an ageing population can steer politicians toward policies that are detrimental for the new generations.

When reformers are concerned that people do not pay enough attention to their future needs, then, they are not just making a normative moral statement (“we must help myopic people, promote their own well-being”). They are also making a normative political statement, assuming implicitly the point of view of society.⁸ And they do it because saving decisions have important externalities. But if externalities do play a role, then the nudges that behavioural economists are proposing should not be defended solely or mainly on welfaristic grounds.

6 Externalities and Norms

A plea for a political justification of nudging faces some potential counter-arguments. A libertarian sceptic, for example, may argue that in an ideal free society the youngsters should be entitled to refuse welfare support to their parents and grandparents, if so they wish. Constitutional architectures should be devised to protect born and unborn citizens from the threat of organized coalitions like the ones we have envisaged above. In this ideal society nudges would be superfluous because the externalities would not arise in the first place.

Another related argument challenges the political justification as follows: suppose that we could create an institutional arrangement where the costs of myopia cannot be externalized. Then, one may argue that we would still have a moral duty to help those who have made mistakes in the past. Suppose for example that, as a matter of fact, all smokers were to die at the age of sixty without burdening the public healthcare system. Shouldn’t we nudge young potential smokers towards a healthier lifestyle in any case, for moral reasons only?⁹

Our reply is twofold. First, let us consider the prospect of setting up constitutional barriers against coalitions that want to externalise costs. Although the idea seems fair in principle, it may be very difficult to implement it in practice. A large number of low savers can easily lobby politicians and bend the rules in their favour. Constitutional defensive barriers must be constantly guarded, so even the preservation of liberties entails costs. A policy that corrects for myopia at low cost then may look very attractive even to the libertarian constitutionalist.

Our second point is that we must be careful not to take a narrow view of the relevant costs. As the second argument correctly points out, there are social ties and moral norms that regulate insurance and cost-sharing quite independently of legislation. People may be under strong social pressure to support the members of older generations, even if the latter do not try to coerce them to do so. This is an area where cultural differences play a major role. Consider dieting – another example that features prominently in the libertarian paternalistic literature. Obesity is in many ways similar to irresponsible spending: although the consequences seem to be a private matter, in some situations they can constitute a social problem. However in North America the idea that medical insurance is an individual responsibility is very popular. In Europe, in contrast, public healthcare is such an established institution that the idea of letting people die if they cannot pay their medical bills would strike most people as strongly counter-normative.¹⁰

Since it would be morally unacceptable to transfer all the costs to the patients, an obesity epidemics could have significant negative externalities in Europe. In this context nudges are attractive for two reasons: (1) because they prevent a situation (the death of many people) that the majority of citizens would find morally wrong; and (2) because they reduce the probability that the majority will feel morally obliged to intervene. Notice that neither of these justifications is welfaristic: in the first case, we make young Ann pay a cost to make sure that old Ann does not get ill, but we do not require young Ann to agree that she is made better off by the nudge policy. The intervention would be justified in genuinely paternalistic, rather than in libertarian “paternalistic” (that is, welfaristic) terms. The second justification is based on the prevention of externalities: nudges are a cheap way of preventing later interventions that would be costly for the community. Moreover, they prevent institutions that we value highly, such as public welfare, from being subjected to excessive strain. In the worst-case scenario such institutions may even collapse, so avoiding future stress has clear advantages if it can be done at low cost.

Our final point in this paper, then, is that the concept of externality is context-specific, and may depend heavily on local norms. The reason why we feel that people should save more is that a society that does not save enough may face catastrophic consequences in the future, and the costs will be shared at least in part with people who have no responsibility. The degree of this sharing will depend in part on the existence of social or moral norms that prescribe assistance towards the weakest members of society. Where such norms are strong – as in the case of medical assistance in Europe – we feel even more justified to prevent unhealthy eating, smoking, and similar behaviours, but the justification is not based on welfaristic premises.

All these considerations suggest that welfarism is a red herring. Coercion may be licenced even if the preferences of individuals regarding their own lifestyles are stable and consistent. In the case of inconsistent preferences, the policy of nudging has the further attraction of being cheaper and of keeping the options open to the decision-maker. But we should not worry about preferences too much: we should stop looking for welfaristic justifications of nudge policies. An important reason why we think that people ought to take care of their future is that a failure to do so will probably affect their children and the children of others. In addition, we often feel that a failure to plan in advance for one’s future would create situations that are unacceptable quite independently of what irresponsible agents prefer. Nudge policies are attractive because they tackle these issues in a cheaper and less intrusive ways than traditional alternatives. For all these reasons they should be welcomed and put to use whenever they help create a more viable society. The libertarian “paternalistic” tag may be a good selling point in the current political climate, but it fails to capture the real reasons why we feel entitled to nudge saving, dieting, and the like.