Do people really want to be nudged towards healthy lifestyles?

In the passages I have quoted, Thaler and Sunstein are declaring that their normative criterion is based on individuals’ judgements about what makes them better off, and about what their goals are. There is a clear implication that these judgements are subjective: different individuals in similar circumstances may make different judgements about what is good for them, without those differences showing that anyone’s judgement is in error. The policy maker is expected to respect those judgements, rather than to implement her own judgements about what is good for people, or to implement the judgements of some consensus of experts. But how are individuals’ subjective judgements defined, and how is the policy maker (or the normative economist who advises her) to elicit them?

This approach of behavioural welfare economics is widely used by prominent behavioural economists (Bleichrodt et al. 2001; Camerer et al. 2003; Kőszegi and Rabin 2007; Bershears et al. 2008; Salant and Rubinstein 2008; Bernheim and Rangel 2009) and has been endorsed from the perspectives of economic philosophy (Hausman 2012) and the economics of social policy (Le Grand and New 2015). In a recent paper, Gerardo Infante, Guilhem Lecouteux and I have documented the extent to which this approach is used, reconstructed its logic, and subjected it to a methodological critique (Infante et al. 2016). We argue that its implicit model of a human decision maker is that of an inner rational agent interacting with the world through an error-prone psychological shell. The inner agent is imagined to have stable, context-independent preferences of a broadly neoclassical kind. These are the latent preferences that policy makers should try to satisfy, and which express the individual’s judgements about what is good for her. Psychological factors appear in the model only as obstacles that can prevent the inner agent from achieving its goals—obstacles such as inattention, cognitive overload and lack of self-control. Despite Thaler and Sunstein’s label, the decision maker described by this model is not a Human in the ordinary sense of the word. It is a faulty Econ.

Infante, Lecouteux and I argue that this model lacks adequate psychological foundations. The existence and rationality of latent preferences is assumed, not explained. Latent preferences are defined as the subjective judgements that a person’s choices would reveal in the absence of error, but we are given no independent criterion by which errors can be identified. Nor are we given any reason to suppose that, after the effects of specific errors have been filtered out, individuals’ choices will be free of the ‘anomalies’ that are the raison d’être of behavioural welfare economics. Take the familiar case of Thaler and Sunstein’s cafeteria, in which customers’ choices between food items vary according to the display chosen by the cafeteria director. Thaler and Sunstein’s recommendation is that the director displays the items so that customers tend to choose as they would have done, had their reasoning not been distorted by error. But if error-free choices are affected by the display too, we are no further forward.

The remark about people being willing to pay third parties to help them make better decisions may suggest a different reading of ‘as judged by themselves’. On this alternative reading, latent preferences are not a hypothetical construct, defined by what the individual would prefer if ideally rational. Instead, they are the preferences that the individual actually endorses in some independently definable circumstances (perhaps in a cool emotional state, reflecting about her welfare), and which in some sense she continues to acknowledge even when she fails to act on them. In other words, Thaler and Sunstein may be using a model of self-acknowledged akrasia (that is, failure of self-control).

Some of the cases they discuss certainly fit this interpretation. For example, their chapter on ‘Resisting Temptation’ begins with a story about a dinner party hosted by Thaler. He has put out a large bowl of cashew nuts for pre-dinner nibbles. Realising that the nuts are being eaten very quickly and that his guests’ appetites for dinner will be impaired, he removes the bowl to the kitchen, where it is out of sight. Crucially for the story, the guests immediately thank him for doing this (pp. 40–41). If we take it that the guests are free to walk into the kitchen, Thaler’s action is a paradigm case of a nudge (and is very similar to the case of the cafeteria director who manipulates the display of food items). But the guests’ thanks add an extra dimension to the story. On the most natural interpretation, the guests were aware that the sight of the nuts induced an urge to eat they found hard to resist but which, even at the moment of eating, they felt a desire not to act on. As soon as the visual stimulus was removed, they endorsed that desire as their true preference. It is reasonable to conclude, as Thaler and Sunstein do, that the guests wanted the outcome that they were being nudged towards.

Thaler and Sunstein go on to argue that, because of what George Loewenstein (1996) has diagnosed as the ‘hot–cold empathy gap’, people often fail to anticipate the power of temptation, and so expose themselves to temptations that they mistakenly expect to be able to resist. In their story: ‘Tom is on a diet and agrees to go out to a business dinner, thinking he will be able to limit himself to one glass of wine and no dessert. But the host orders a second bottle of wine and the waiter brings by the dessert cart, and all bets are off’ (p. 42). It is easy to imagine a Tom who, even while accepting a share of the second bottle and choosing a crême brulée, acknowledges that he is acting against his better judgement. Such a Tom might reasonably be said to want to be nudged towards keeping his resolutions—at least if, when accepting the dinner invitation, he was not 100 % confident of his willpower.

If these arguments for nudging are to be consistent with the ‘as judged by themselves’ clause, we need to interpret that clause in terms of hypothetical latent preferences, not preferences that individuals actually endorse but are sometimes tempted to act against. The point of the story of the person with the high-fat diet (let us call him Bill) is surely that he does not think that this diet is bad for his heart. Over many years, he has a consistent sense that he is eating sensibly. It is Thaler and Sunstein, in the role of experts, who are questioning Bill’s judgement. They are appealing to the hypothetical choices of an inner rational agent whose reasoning takes account of information that the real Bill is unaware of or does not attend to, or whose significance he does not appreciate.

Thaler and Sunstein might reasonably argue that, over the years of his high-fat diet, the real Bill must have been exposed to information and experiences that provided good grounds for the belief that such a diet increases the risk of heart disease. His failure to make this inference might be interpreted by an outside observer as an error of reasoning. But recall Thaler’s claim that he and Sunstein ‘just want to reduce what people would themselves call errors’. The outside observer is not entitled to infer that the high-fat diet was bad for the real Bill, as judged by himself and before he experienced his own heart attack. Thus, she is not entitled to infer that the real Bill wanted to be nudged towards eating less fat.

As an aid to thinking about the difference between the two interpretations of ‘as judged by themselves’, I offer the following thought experiment. Consider Jane, one of the customers at Sunstein and Thaler’s cafeteria. Jane is currently in good health but seriously overweight. Professional dieticians would agree that her long-term health prospects would be better if she ate less high-fat, sugar-rich food (such as the cafeteria’s cream cake) and more fruit and vegetables (such as the cafeteria’s fresh fruit option). But, despite knowing what dieticians advise, she usually chooses the cake. Imagine that, in an attempt to uncover the reasoning that leads people like Jane to ignore dietary advice, we are designing a questionnaire which asks: ‘Which of the following statements best represents your reasons for choosing cake rather than fruit, contrary to the recommendations of health experts?’ The thought experiment is to produce statements that Jane might plausibly assent to, and that could be used to diagnose her mode of reasoning.

The point of this thought experiment is that there are many ways in which a person can explain why she knowingly acts against well-grounded but unwelcome recommendations from experts. Failure of self-control is only one of these ways. My conjecture, based both on casual social experience and on general scientific knowledge about the prevalence of empathy gaps, procrastination, conformity, cognitive dissonance, self-deception and the tendency to overweight evidence from personal experience, is that lack of self-control would not be particularly common as a self-ascribed explanation; statements like (b), (c), (d), (e), (f), (g) and (h) would be more common.

Notice that each of the eight statements in my questionnaire represents a mode of reasoning, heuristic or bias that is known to be a common property of human psychology. However, (a) differs from the others in admitting to a failure on the part of the respondent herself: she has failed to choose as she intended (and, by implication, has failed to choose according to her own best judgement). All the other statements maintain that the respondent knows what she is doing and has chosen what, on balance, she believes is best for her. They maintain that the error is in the experts’ recommendation, applied without recognition of the respondent’s circumstances, not in the respondent’s failure to follow it. My guess is that most people prefer not to explain their own behaviour as the result of error.

My claim is that self-acknowledged self-control problems are a lot less common than many behavioural economists seem to think. Even if behavioural economists or policy makers feel confident that people’s lifestyle choices are based on some kind of error, they should not jump to the conclusion that the error is a self-acknowledged failure of self-control, or that (as Thaler puts it) it is what those people would themselves call an error.

I have argued that, in justifying policies which nudge people towards healthy lifestyle choices, Thaler and Sunstein conflate two different versions of the claim that their proposals are designed to make choosers better off, as judged by themselves. Because of this ambiguity, they can give what appear to be non-paternalistic justifications for interventions that in fact are paternalistic in intent. They can give what appears to be support for the claim that the people at whom their nudges are directed want to be nudged. But these appearances are misleading. Advocates of nudging should come clean about the paternalism of their position and defend it directly.