Normative economics makes many economists uncomfortable, because it seems to require that economists don the togas of moral philosophers. To say whether particular economic institutions, processes, outcomes, and policies are good or bad apparently requires commitments concerning what is good and what is bad. Some psychologists and behavioural economists, especially Daniel Kahneman and his many co-authors,Footnote 1 whom I shall call the ‘new hedonists’, maintain that normative economists must choose between (i) identifying welfare with preference (or choice) and (ii) identifying welfare with happiness or pleasure. They then go on to argue that the latter is the better criterion of economic evaluation. This essay argues in contrast that neither pleasure nor the satisfaction of actual preferences are plausible notions of welfareFootnote 2 and that economists should distance themselves from any philosophical theory of welfare.
This paper is divided into five sections. The first sketches the standard view of welfare endorsed by normative economics and its well-known difficulties. Section 2 criticizes the purported dichotomy between identifying welfare with the satisfaction of preferences and identifying welfare with a subjective state such as pleasure or happiness. Section 3 rejects the arguments in favour of transforming welfare economics into an inquiry into the means of increasing happiness. Section 4 develops a general critique of hedonistic welfare economics. Section 5 concludes with a qualified defence of standard welfare economics with respect to a limited domain and with the heterodox suggestion that normative economists should not be concerned only with welfare.
1. WELFARE AS CHOICE OR THE SATISFACTION OF PREFERENCE
Does normative economics in fact require any moral commitments? Particularly in the mid twentieth century welfare economists flirted with the possibility of maintaining that so-called ‘normative economics’ is in fact positive economics employed to explore the consequences of institutions, processes and policies for the objectives that people pursue. This view is not a relic of a superseded positivist past. For example, in a recent essay Gul and Pesandorfer write,
Economists use welfare analysis to examine how institutions mediate the interests of the participating individuals. Welfare-improving changes to an economic institution are defined to be changes to which the individual(s) would agree. Policy x is deemed better than policy y for an individual if and only if, given the opportunity, the individual would choose x over y (Reference Gul, Pesandorfer, Caplin and Schotter2008: 24 [italics added]).
The first sentence claims that ‘welfare analysis’ consists of positive economics devoted to the examination of how institutions mediate the interests of participating individuals. Gul and Pesandorfer intend the second and third sentences as definitions of apparently evaluative phrases in terms of individual assent or choice. The second sentence defines a change to be ‘welfare improving’ if and only if it is a change to which individuals would agree. The third sentence defines a policy x as ‘better than’ a policy y for an individual if and only if the individual would choose x over y if the individual had the choice. According to these definitions, what is welfare improving or better for individuals is entirely a factual matter. On Gul and Pesandorfer's view, to say that a change is welfare-improving or that a policy is better for individuals has by itself no implications, not even other things being equal, concerning whether the changes or the policies are good or bad. To draw normative conclusions requires a separate value judgement to the effect that changes to which individuals would agree are, other things being equal, good changes and that policies that individuals would choose are, other things being equal, good policies.
Since it is precisely these value judgements that motivate the inquiries, there seems little point to denying that normative economics rests on evaluative premises, and it is confusing to maintain that claims concerning what improves welfare and makes individuals better off are purely factual.
Rather than pretending to avoid moral commitments, orthodox normative economists have more typically attempted to keep them mild and plausible. One way to do this is to take the claim that x is better for an agent A than y if and only if A would choose x over y as a contentful thesis concerning individual welfare rather than merely a definition and to maintain that institutions, processes and policies should, other things being equal, promote individual welfare. The ‘other things being equal’ clause in this last claim enables economists to concede, sensibly, that other things may matter in addition to welfare.
There is, however, a well-known problem with identifying welfare with choice or with the satisfaction of people's actual preferences: People make mistakes. What they choose and prefer may make them worse off. For example, consider Bernheim and Rangel's case of an American tourist in London (Reference Bernheim and Rangel2004: 1561). Suppose the unfortunate tourist fails to look to the right and is run down by a taxi. Choosing to step into the street just then, as the tourist preferred to do, instead of remaining on the curb, does not make the tourist better off. It seems obvious that people do not always choose or prefer what is best for them.
Rather than admitting that people sometimes choose an action that is worse than some feasible alternative, Gul and Pesandorfer maintain that the object of the tourist's choice in this case is a strategy rather than a specific action, and they assert that no better alternative strategy was feasible. They argue that once one recognizes that there are subjective constraints on the strategies people can pursue, one can argue that apparent mistakes, such as stepping in front of a taxi are not mistakes at all, but the consequences of optimal feasible strategies (Reference Gul, Pesandorfer, Caplin and Schotter2008: 23). But this tactic fails to solve the problem. First, why believe that the tourist in fact made the optimal strategy choice? Second, and more importantly, whether the strategy choice was optimal or not, this is a case where interfering with the agent's choice would have enhanced the outcome. A bystander who forcibly prevented the tourist from stepping into the road would have improved the outcome.Footnote 3
Findings by psychologists and behavioural economists concerning the systematic biases and blunders in people's choices make the understanding of mistakes a pressing practical problem (Camerer et al. Reference Camerer, Issacharoff, Loewenstein, O'Donoghue and Rabin2003; Thaler and Sunstein Reference Thaler and Sunstein2003, Reference Thaler and Sunstein2008). For example, as Thaler and Sunstein document, people show a strong status-quo bias. Whether the default for a retirement savings plans calls for employees to opt in or to opt out has a huge effect on the rate of participation. Many people choose to save for their retirement when that is the default who would have chosen not to save for retirement if the default had been otherwise. Their behaviour thus reflects no settled preference between saving and not saving, and one cannot reasonably infer from their choices what is better for them. Which alternative then is better, and in what sense of ‘better’?
Welfare economics is of course not merely a characterization of what welfare is and how to measure it. In addition, normative economists need to explain how welfare considerations should influence policy. The most common view denies that policies should be based on interpersonal comparisons of welfare, but there are some prominent economists such as Harsanyi (Reference Harsanyi, Sen and Williams1982) or Ng (Reference Ng1997) who are willing to make interpersonal comparisons and to defend utilitarianism. This essay takes no stand on the question of whether the welfare of individuals can or should be added up. It is concerned exclusively with the question of what normative economists should take welfare to be.
2. KAHNEMAN'S DICHOTOMY
If interfering with choice can make things better, then welfare cannot be defined by choice. How then should welfare be defined? What does it mean to maintain that it would have been better for the tourist not to have stepped into the road? In what sense is it better for people to save more for their retirement? The answer asserted by a growing number of economists is that the tourist would have been happier if he or she had not been run over by the taxi and that people would be happier if they saved more.Footnote 4 In their view, welfare is happiness. For example, Köszegi and Rabin simply identify welfare and happiness (Reference Köszegi and Rabin2008). Though not behavioural economists, so do Di Tella et al. (Reference Di Tella, MacCulloch and Oswald2003). Writing more than a half-century earlier in a very different context, Ian Little makes the same identification (Reference Little1957: ch. 1, 5).
In recent works Kahneman and several distinguished co-authors maintain that economists interpret the word ‘utility’ in two ways. Here is the way Kahneman expresses the claim in an essay co-authored with Robert Sugden:
Two different interpretations of the term ‘utility’ have been used in the literature of economics. In its original interpretation, which derives from Bentham, utility is interpreted in hedonistic terms, as a measure of pleasure and pain . . .
Of course, this is not the way that most economists have used the term ‘utility’ over the last 100 years. During this period, utility has usually been understood as decision utility. Decision utility is a representation of preferences, and the concept of preference is understood in terms of choice: a person's preferences are the mental entities that explain his choices, and are revealed in those choices (Kahneman and Sugden Reference Kahneman and Sugden2005: 162).Footnote 5
By ‘utility,’ here Kahneman need not mean welfare or anything evaluative. He could be interpreted as maintaining that economists have held two competing theories of choice in terms of pleasure or in terms of preferences, or perhaps that they simply have two unrelated uses for the same word. However, he and Sugden go on:
In 20th-century neoclassical welfare economics, the measure of welfare for any given individual is her utility. If utility is interpreted as decision utility and if economic agents are assumed to be rational utility-maximisers, there is no need to worry about whether they are choosing things that they will enjoy consuming. Merely by being rational in the formal sense of acting on consistent preferences, they are maximising decision utility. This allows economics to prove theorems about the welfare consequences of different economic institutions without addressing the empirical question of how those institutions impact on human happiness. (Reference Kahneman and Sugden2005: 162)
So the two concepts of utility give rise to two concepts of welfare: pleasure and preference satisfaction. In other words, Kahneman and Sugden maintain that economists face a choice between identifying welfare with pleasure or with the satisfaction of preferences. In many cases these two go hand-in-hand,Footnote 6 but not in all. When preferences are not self-interested or are based on false beliefs, or when individuals do not find out whether things turn out as they prefer, the satisfaction of the agent's preferences may easily fail to increase the agent's pleasure.
Accepting Kahneman's dichotomy between preference and pleasure leaves welfare economists with an unappetizing choice. Either they can maintain (absurdly) that people never make mistakes – that whatever people choose or prefer is best for them – or they can climb out on a shaky philosophical limb and espouse a hedonistic view of human welfare. With no obvious alternative other than the implausible denial that people make mistakes, some prominent contemporary economists have recently opted for hedonism.Footnote 7 For example, Richard Layard maintains that ‘Happiness should become the goal of policy, and the progress of national happiness should be measured and analysed as closely as the growth of GNP’ (Reference Layard2005: 147). David Blanchflower and Andrew Oswald write that economists are increasingly asking ‘whether we should, in one way or another, substitute the goal of gross national happiness for the more traditional economist's objective of gross national product’ (Reference Blanchflower and Oswald2005: 307). In a series of papers Oswald, Blanchflower, and others have explored the extent to which happiness depends on factors such as income, employment, sex, and death of a spouse (Clark and Oswald Reference Clark and Oswald2002; Di Tella et al. Reference Di Tella, MacCulloch and Oswald2003; Blanchflower and Oswald Reference Blanchflower and Oswald2004a, Reference Blanchflower and Oswald2004b; Oswald and Powdthavee Reference Oswald and Powdthavee2007). Kahneman and Krueger are more cautious: ‘A measure of Gross National Happiness would seem to us to be an overly ambitious goal in view of the present state of knowledge and limitations of subjective measurement’ (Reference Kahneman, Slovic and Lichtenstein2006: 22), but they conclude that ‘The U-index, or proportion of time people spend in an unpleasant emotional state, however, strikes us a promising measure of an important feature of society's well-being’ (Reference Kahneman, Slovic and Lichtenstein2006: 22).Footnote 8
3. A CRITIQUE OF FOUR ARGUMENTS IN DEFENCE OF HEDONISTIC WELFARE ECONOMICS
What grounds are there to favour this proposed shift from evaluating policies by preference satisfaction, which can often be imputed from market choices, to evaluating policies by their consequences for happiness or pleasure? There are four main arguments:
1. (The hedonism argument) Welfare is happiness, and so policies should be evaluated by their effects on happiness.
2. (The dichotomy argument) Economists can evaluate policies by their effects on decision utility or by their effects on experienced utility. The inadequacies of the first are arguments for the second.
3. (The expectation argument) Decision utility is a flawed expectation of experienced utility. Economists should assess policies in terms of what people ultimately care about rather than in terms of their imperfect expectations.
4. (The component argument) Because happiness contributes so significantly to welfare, economists should measure the effects of policies on happiness.
The hedonism argument
As already noted, some economists, such as Köszegi and Rabin, regard it as obvious that welfare or well-being is happiness. If, in addition to this premise, one holds – as most economists do – that ‘The goal of policies ought to be to maximize people's well-being . . .’ (Loewenstein and Ubel Reference Loewenstein and Ubel2008: 1804), then policies ought ideally to be evaluated in terms of their effects on happiness. The only possible reason to evaluate policies in terms of preference satisfaction rather than in terms of happiness would lie in the practical convenience of measuring preference satisfaction as a proxy for happiness.
Both the premises in this first argument for a hedonistic welfare economics are dubious, but apart from some comments near the end of this essay, I am going to grant the second premise, that (other things being equal) policies should promote welfare. Although versions of hedonism – the first premise, the view that welfare is happiness or pleasure – have had distinguished defenders, including especially Henry Sidgwick (Reference Sidgwick1907), the view faces serious objections and has few contemporary philosophical defenders. For recent sustained defences, see Feldman (Reference Feldman2004) and Crisp (Reference Crisp2006).
There are four main philosophical objections to hedonism and hence to this first argument in defence of a hedonistic welfare economics:
A. Hedonism is an implausible view of what people seek.
B. Happiness cannot be measured.
C. Welfare does not depend only on subjective experience.
D. ‘Happiness’ is ambiguous.
A. Hedonism is implausible as a view of people's aspirations. Agents have many goals other than happiness, and they do not usually form or revise their goals by calculating how their pursuit is likely to bear on subjective experience. Is Osama Bin Laden in pursuit of the happiest or most pleasurable subjective experiences? But if hedonism is properly understood more narrowly as a theory of welfare – that is, of what makes people better or worse off – rather than as a theory of what people care about, these objections are no longer relevant. It is possible to argue that what made Mother Teresa's life a good life was her subjective experience without maintaining (absurdly) that all she cared about was her own experience.
B. A second traditional objection to hedonism is that it makes welfare impossible to measure. If people can only observe the quality of their own subjective experience, then the systematic study of welfare is impossible. Kahneman and others have tackled this criticism head on and have shown that some subjective states can be reliably measured (Kahneman et al. Reference Kahneman, Wakker and Sarin1997, Reference Kahneman, Krueger, Schkade, Schwarz and Stone2004a, Reference Kahneman, Krueger, Schkade, Schwarz and Stone2004b; Kahneman Reference Kahneman, Kahneman and Tversky2000a, Reference Kahneman, Kahneman and Tversky2000b; Kahneman and Krueger Reference Kahneman and Krueger2006). In addition, large-scale empirical measures of life satisfaction are stable and consistent and appear to latch on to something real (Easterlin Reference Easterlin, David and Reder1974, Reference Easterlin1995; Clark and Oswald Reference Clark and Oswald2002; Di Tella et al. Reference Di Tella, MacCulloch and Oswald2003; Blanchflower and Oswald Reference Blanchflower and Oswald2004a, Reference Blanchflower and Oswald2004b; Oswald and Powdthavee Reference Oswald and Powdthavee2007, Reference Oswald and Powdthavee2008).
C. A third objection to hedonism, which unlike the first two is compelling, rests on the intuitive judgement that individuals with the same subjective experiences may not be equally well off. Consider the hypothetical case of two people, both of whom believe that they are loved by their spouses and children, successful in their projects, and honoured by their community. But the beliefs of only one of the two are true. The other is living in a fool's paradise, a victim of systematic deception. Since the subjective experience of the two individuals is the same, the hedonist (who takes well-being to consist in subjective experience) must say that the two are equally well off. The conclusion most philosophers have drawn is that how good lives are for the people who live them depends on more than just the quality of experience. Although happiness is an important ingredient in well-being, well-being is not happiness. Robert Nozick's well-known hypothetical example of an experience machine highlights this point (Reference Nozick1974). All hedonists can do in response to this objection is to contest the intuition that drives it. As discussed in section 4, thoughtful defenders of a role for ‘experience utility’ in welfare economics have noted that there are dimensions of human well-being that escape a hedonistic view of well-being, and they have distanced themselves from the view that welfare consists entirely of subjective experience.
D. A fourth difficulty stems from the ambiguities in the claim that welfare consists in happiness or pleasure. There are many different views of pleasure or happiness. Among the many views in the literature in philosophy and economics, one can find at least the following four:Footnote 9
1. Happiness consists of virtuous activity (Aristotle Reference Irwin2000).Footnote 10
2. Happiness consists of the exercise of the higher human faculties coupled with the experience of success in worthwhile projects (Mill Reference Mill and Sher1863).
3. Happiness consists of the life-time sum of feelings of pleasure minus feelings of pain (Bentham Reference Bentham1789, KahnemanFootnote 11).
4. Happiness consists of the life-time sum of pleasurable attitudes minus painful attitudes (Sidgwick Reference Sidgwick1907, Feldman Reference Feldman2004).
Agreeing that welfare is happiness but disagreeing about what happiness is, defenders of these four views of happiness necessarily disagree about what welfare is. The plausibility of each of these views is an argument against the others.
The dichotomy argument
A second argument in defence of hedonistic welfare economics rests on Kahneman's dichotomy criticized above in section 2. If there are only two alternatives – welfare is either pleasure or preference satisfaction, then every objection to measuring welfare by preferences or choices becomes an argument in favour of measuring welfare by happiness or pleasure.
Yet it is a mistake to suppose that economists have to choose between just two views of welfare, which correspond to the two meanings of utility Kahneman distinguishes. There are other alternatives. Indeed, most contemporary philosophers reject both these views. Perhaps the most common contemporary view is that well-being consists in the satisfaction of self-interested, well-informed and rational preferences.Footnote 12 But this view, along with the simpler view of welfare as the satisfaction of actual preferences, has been subjected to trenchant criticisms in recent works by authors such as Thomas Scanlon (Reference Scanlon1998: ch. 3) and Richard Kraut (Reference Kraut2007: ch. 2).Footnote 13
The expectation argument
A third argument in defence of hedonism that runs through some of Kahneman's essays is that decision utility is a flawed expectation of experienced utility. These essays suggest that what matters to people and determines their preferences is experienced utility. Given the biases in the expectations people have concerning how they will feel, it is accordingly more sensible to evaluate institutions, outcomes and policies in terms of experienced utility. Thus Kahneman and Thaler write,
Utility maximization is usefully thought of as a goal. People are trying to make choices that will, on average, make them as well off as possible, as judged by themselves, not others. But to maximize utility successfully, one must start by making a forecast about how the various possible outcomes will be experienced.Footnote 14
To maintain that ‘to maximize utility successfully, one must start by making a forecast about how the various possible outcomes will be experienced’ assumes that what determines preferences are expectations of experiences. Similarly, earlier in the essay, Kahneman and Thaler maintain that the ‘ability of economic agents to make accurate, or at least unbiased, forecasts of the hedonic outcomes of potential choices’ is ‘a necessary condition for utility-maximizing choices’ (Reference Kahneman, Slovic and Lichtenstein2006: 222; see also Kahneman and Sugden Reference Kahneman and Sugden2005: 167). These remarks make sense only if one assumes, as Kahneman and Thaler apparently do, that maximizing utility is maximizing the quality of subjective experience, and that people construct their preference rankings on the basis of forecasts of ‘hedonic outcomes’.
On this view, welfare economics has been directed toward maximizing experienced utility all along. In Köszegi and Rabin's words, ‘all existing welfare conclusions in economics rely on choice-unobservable assumptions about what makes people happy’ (Reference Köszegi and Rabin2008: 1823). Similarly, Kahneman maintains that the standard approach in welfare economics equates what people choose with what they will most enjoy (Reference Kahneman, Slovic and Lichtenstein2006: 489). Kahneman suggests that in focusing on preferences, standard welfare economics has relied on biased and often mistaken expectations of experience utility rather than on the real thing. Instead of relying on preferences, which allegedly depend on the faulty judgements of individuals concerning their future experiences, those who seek to promote welfare should rely on the best available information concerning how outcomes will in fact be experienced. Provided that there are good ways to measure experience utility, welfare economists accordingly have reason to shift their interest from preferences to pleasures.
This argument in defence of hedonistic welfare economics relies on a false premise. In contrast to what Kahneman and Thaler maintain, whether an agent prefers x to y often does not depend mainly on the agent's beliefs about what his or her mental state would be if x is the case compared to what his or her mental state would be if y is the case. Indeed, Kahneman himself elsewhere disavows the view that preferences are determined by expectations of experienced utility. As Kahneman et al. very reasonably assert, ‘Our analysis applies to situations in which a separate value judgement designates experienced utility as a relevant criterion for evaluating outcomes. This set does not include all human outcomes’ (Reference Kahneman, Wakker and Sarin1997: 216). Kahneman maintains that one of his central goals is merely ‘to develop a moment-based conception of an aspect of human well-being’ (Reference Kahneman, Kahneman and Tversky2000b: 673; italics added).
People are not in fact always ‘trying to make choices that will, on average, make them as well-off as possible’. Just consider acts of pure hatred or humdrum habit. Nor do preferences mainly depend on forecasts of experiences. I strongly prefer the state of affairs in which after I die my children live long and happy lives to another in which they are incinerated in a nuclear holocaust, even though I do not expect to be feeling anything one way or another after my death. It is not just in odd cases around the edges that preferences do not depend on expectations of mental states. Even when there is a correlation between my preferences and my expected mental states, as there was when I last voted for President of the United States, it is at least as likely that my preferences (and the reasons for those preferences) explain my subsequent mental states as it is that my anticipation of my mental states explain my preferences. I was happy (as I expected to be) when Barack Obama won the 2008 election, because I preferred the policies he espoused. Those preferences rather than my anticipation of greater happiness if Obama won explained my voting.
The component argument
Although welfare is not the same thing as happiness or pleasure, happiness contributes substantially to welfare. If happiness can be measured reliably, then its measurement will provide useful information to policy makers. For example Blanchflower and Oswald (Reference Blanchflower and Oswald2005) point out that Australia ranks third in the world in terms of the Human Development Index but 12th in a sample of 35 nations in terms of happiness. Without accepting the authors’ identification of welfare with happiness, economists can still recognize that happiness is a crucial component of welfare, and they can ask what is responsible for the discrepancy between measures of happiness and measures of human development. Similarly, although Loewenstein and Ubel harshly criticize the proposal to measure welfare by experience utility, they think that measures of experience utility are still useful. They suggest, for example, that ‘the best approach may involve decision utility measures among people who are thoroughly and convincingly informed about the relevant research on experience utility’ (Reference Loewenstein and Ubel2008: 1807). Despite championing measures of experience utility, Kahneman sometimes makes only modest claims for such measures. In the conclusion to ‘Experienced Utility and Objective Happiness: A Moment-Based Approach’, he writes
Objective happinessFootnote 15 is not proposed as a comprehensive concept of human wellbeing, only as a significant constituent of it. Maximizing the time spent on the right side of the affect grid is not the most significant value in life, and adopting this criterion as a guide to life may be morally wrong, and perhaps also self-defeating. (Reference Kahneman, Kahneman and Tversky2000b: 688)
Kahneman maintains only that ‘the experiencing subject deserves a voice’ in the assessment of well-being (Reference Kahneman, Kahneman and Tversky2000b: 687).
Yet it is not clear how loud this voice should be or what tune it should sing. ‘Happiness’ remains ambiguous, and it is not clear how to integrate concerns about happiness with other aspects of welfare. The ambiguities concerning what counts as happiness haunt even the modest proposal of supplementing standard welfare economics with measurements of happiness. On the one hand, there is an empirical literature that goes back to Easterlin (Reference Easterlin, David and Reder1974, Reference Easterlin1995) that measures happiness by responses to questions such as ‘Taking all things together, how would you say things are these days – would you say you're very happy, fairly happy, or not too happy these days?’ or ‘On the whole, are you very satisfied, fairly satisfied, not very satisfied, or not at all satisfied with the life you lead?’ (Di Tella et al. Reference Di Tella, MacCulloch and Oswald2003: 810–811). These questions are soliciting retrospective judgements concerning either how happy individuals believe themselves to be when they are surveyed or how satisfied they are with their lives. Rather than measuring the quality of the individual's mood or how the individual is feeling, the surveys elicit attitudes. Such surveys measure happiness as a cumulative attitude, which is the fourth notion of happiness discussed above. On the other hand, Kahneman and his co-authors (whom I have called ‘the new hedonists’) have argued for measuring momentary feelings or mood, roughly corresponding to the third notion of happiness as a sum of feelings.
Even if economists knew what happiness is, they would face serious problems weighing happiness against other considerations influencing welfare. For example, in conjunction with others (Smith et al. Reference Smith, Sherriff, Damschroder, Loewenstein and Ubel2006), Loewenstein and Ubel found that ‘despite experiencing moods similar to people without colostomies, they [colostomy patients in their sample] were willing to give up 15% of their projected life span to rid themselves of a colostomy’ (Reference Loewenstein and Ubel2008: 1799). If one takes measures of subjective experience as indicating well-being, colostomies do not diminish welfare. If one takes preferences (as indicated here by a time trade-off) as measuring well-being, colostomies significantly diminish welfare. Just what sort of ‘voice’ should experienced utility have?
Furthermore, if hedonistic welfare economics is proposed merely as a supplement to standard preference-based welfare economics, which is designed to pick up additional data concerning happiness, then it offers no alternative account of welfare that will resolve the conundrum that standard welfare economics appears to rule out the possibility of mistakes.
4. A CRITIQUE OF HEDONISTIC WELFARE ECONOMICS
Whether proposed as an alternative to the standard view of welfare employed by normative economists or merely as a supplement to it, defenders of the view that welfare is happiness or that it depends mainly on happiness owe us an account of what happiness is. The two notions of happiness economists have been mainly concerned with are on the one hand happiness as a judgement or an attitude – as an appraisal of how one's life is going – and happiness as a feeling, as a matter of mood. These two notions, which I shall call ‘attitudinal’ and ‘experienced’ happiness are clearly different, and, as Kahneman and various co-authors have demonstrated, their differences have empirical consequences. In fact, measures of what Kahneman calls ‘objective happiness’, which result from summing momentary pleasures, may conflict with retrospective reports and attitudes. One important discovery that contemporary psychologists have made concerning subjective experiences is that people are not good at anticipating them or remembering them. In one memorable experiment (see Redelmeier et al. Reference Redelmeier, Katz and Kahneman2003), in some patients undergoing a sigmoidoscopy, the doctor left part of the probe inserted for an additional period. Since this is uncomfortable, those in whom the probe remained inserted for an additional period experienced a larger total quantity of discomfort. Yet those subjects reported on average that the experience was less painful than those in whom the probe was removed as usual. In recalling a painful episode, people average between their memories of the discomfort at the end and at the peak, paying insufficient attention to duration. Thus the inconsistency between the discomfort people remember and the ‘objective’ discomfort psychologists can measure by integrating momentary discomfort over time.
Having identified this inconsistency between cognitively complex anticipations, memories, and appraisals of mental states on the one hand and cognitively simpler immediate experiences of them, the new hedonists have argued for the importance of measuring ‘moment utility’ or ‘instant utility’ – that is, the quality of momentary experience – and then aggregating these quantities to measure ‘total utility’ or ‘objective happiness’ (Kahneman Reference Kahneman, Kahneman and Tversky2000b). In the discussion that follows, I shall first present some qualms concerning the techniques used to measure objective happiness. Then I shall argue that designing policies to maximize objective happiness demands far more information than designing policies to satisfy preferences and more information than economists are likely to have. I shall then question whether objective happiness is an ethically plausible objective. Finally, I shall argue that conflicts between objective happiness and retrospective appraisals diminish whatever plausibility there may be in designing policies that aim to augment happiness.
Measurement problems
There are several ways to measure moment utility. One that Kahneman has emphasized is the daily reconstruction method. Using this technique, Kahneman et al. (Reference Kahneman, Krueger, Schkade, Schwarz and Stone2004a) assembled the information in Table 1 concerning a sample of slightly over 900 Texan working women. The women filled out a diary of episodes from the previous day, noting when each episode began and ended, what they were doing, where they were and whom they were with. They then rated each episode along dimensions such as ‘happy’, ‘warm/friendly’, ‘angry/hostile’ on a 0–6 point scale ranging from ‘not at all’ to ‘very much’. Though these are retrospective appraisals, they are done soon after the experiences, and Kahneman et al. and Kahneman and Krueger (Reference Kahneman, Slovic and Lichtenstein2006: 10–11) argue that they are a good approximation to the results of more detailed studies in which experimental subjects carry around electronic diaries and periodically report their current mood.
Reproduced from Kahneman et al. (Reference Kahneman, Krueger, Schkade, Schwarz and Stone2004b: 432).
‘Net affect’ in Table 1 is the average of the ratings (on a 0–6 scale) of three positive adjectives (‘enjoyment’, ‘warm/friendly’ and ‘happy’) minus the average of five negative adjectives (frustrated, depressed, angry, hassled and criticized). Some of the activities can be pursued at the same time. So, for example, the time spent socializing after work might include time spent at dinner. The average time column in the table is potentially confusing, because the time reported is not conditional on having engaged in the activity; to determine how long on average those women who engaged in an activity spent on it, one needs to divide the average time for the whole sample, by the percentage of the sample who engaged in the activity. The results are surprising, at least to this ageing academic. The women who reported having intimate relations, claimed to have spent on average about two hours at it, and those who claim to have prayed claimed to have spent about two hours praying as well. Praying might be done while also doing housework or commuting, but presumably having sex excludes the other activities, except perhaps prayer and watching TV. These data do not inspire confidence.
Problems of implementation
A more serious problem lies in how to interpret and how to apply the net affect numbers. As Kahneman et al. note, the results reported here contrast dramatically with the results of surveys that purport to measure enjoyment of activities. In those surveys, interacting with one's children usually comes out at or close to the top. In this study, in contrast, spending time with one's children is near the bottom of the list, approximately tied with housework.
On the basis of studies of the determinants of objective happiness, Kahneman and Krueger suggest two general policies to increase welfare: to focus on mental health interventions for those who are miserable and to reduce the amount of time people spend doing things that they find unpleasant (Reference Kahneman, Slovic and Lichtenstein2006: 21). The second of these two policies apparently implies that it would increase welfare if adults spent less time taking care of children. Could social policy really improve the well-being of adults (ignoring for the moment the consequences for children) by shortening the amount of time that they spend taking care of their children? Kahneman and Krueger do not say so, and I suspect that they would reject this apparent implication of their policy ideas.
It seems sensible to suggest that people would be better off if they spent less time doing things they find unpleasant. Yet the example of childcare shows that new hedonist welfare economists need to tread cautiously. A crucial problem with the proposal to diminish the time people spend doing things they find unpleasant is that a myopic policy of maximizing current net pleasure is no more likely to maximize net pleasure over a lifetime than is a policy of maximizing weekly profits likely to maximize profits over a decade. For example, the pleasure men and women take in interacting with their children depends on the other demands on their time and the amount of time they have with their children (Hochschild Reference Hochschild2001). For many parents, the quality of the momentary experiences of childcare may be initially low but increasing with the amount of time they have with their children before eventually diminishing. Parents may enjoy their time with their children more if their jobs allow them to spend more rather than less time with their children or to be with their children when parents and children are not both exhausted at the end of long days at work and day care.
Childcare is obviously not only a consumption good. It is also an investment in both the child's development and the parent's own future satisfaction. As Kahneman and Krueger would no doubt agree, policy makers consequently cannot ignore the consequences of shortening the time parents spend with their children for the children's development or for the parent's later satisfaction. Though this is only an issue of implementing the new hedonism, it is a serious one: Realizing a policy of maximizing objective happiness is immensely more difficult than carrying out a more traditional hedonist policy or a policy of maximizing preference satisfaction. The reason is that people's preferences and their anticipations of happiness are influenced by their own calculations of tradeoffs between the present and the future. Despite the myopia and irrationality documented by behavioural economics, these calculations are not all stupid or uninformed. People do things that they expect to be unpleasant, such as going to the dentist, because they anticipate that doing so will better satisfy their preferences than allowing their teeth to decay. They anticipate that some pain now will lead to less pain later. There is such a thing as prudence, and both traditional hedonists who rely on the predictions individuals make concerning their future happiness and non-hedonists who take welfare to be preference satisfaction can piggyback on prudence when examining the welfare consequences of policies. The new hedonists cannot.
The cost of avoiding the mistakes individuals make in predicting future happiness is to deprive oneself of the knowledge that is encapsulated in their preferences and anticipations. To determine whether to encourage activities such as visits to the dentist, which typically cause a short-term dip in objective happiness and are desirable only as investments in future well-being, new hedonist policy analysts (who identify well-being with momentary experience) have to make their own judgements about whether such investment is worthwhile. While certainly not impossible (and, in the case of dentists, easy), this task is in general very difficult. How can the policy maker determine what amount of time spent taking care of one's children would maximize the sum of people's life-time momentary pleasures?
Is maximizing objective happiness an ethically plausible objective?
In addition to these problems of measurement and implementation, the proposal to identify welfare with experience utility or objective happiness faces ethical objections. Consider for example the clash between the great value that people claim to place on interactions with their children and the low net affect of childcare in the Texas study. How might it be explained? Kahneman et al. point out that surveys concerning generic attitudes are bound to be misleading: Few people will admit that they don't enjoy taking care of their kids, no matter how sick of changing diapers they may be. Fair enough. But would it be plausible to maintain that the value of interactions with one's children can be captured by the subjective experience of those interactions? Whether an activity makes someone's life better is not the same thing as the feelings the individual has while engaging in the activity. If psychological investigation showed that the lifetime sum of my momentary utilities would have been larger if I had not had children, it would not follow that my life would have been better without my sons.
The new hedonists maintain that the conflict between retrospective appraisals and objective happiness show that hedonists should focus on objective happiness. Since those subjects in the sigmoidoscopy experiment from whom the probe was promptly removed had subjective experiences that were at each moment no worse than those with the prolonged period with the probe and at some moments better, Kahneman and his collaborators maintain that those subjects are better off. Their view seems initially plausible. Knowing the conditions of the experiment, who would choose to be in the group in which the probe was left in for a longer period?
On the other hand, looking back and summing up their experiences, those subjects in whom the probe was left for the longer period judge that their own life in terms of their own (remembered) subjective experience is better. Recognizing this, should a sensible hedonist prefer to have the probe removed promptly? Why not instead regard the additional period of discomfort as an investment in achieving a life that seems to oneself to have had less discomfort in it? (Indeed if the subjective experience of recollecting differs sufficiently between the two groups, those who experience more momentary discomfort during the procedure may have a higher life-time sum of momentary pleasure.) Given a choice in a surgery between a more effective (but not completely effective) anaesthetic on the one hand and the combination of a less effective anaesthetic and an amnesiac that leaves one with no memory of the pain, is it obvious that one should choose the more effective anaesthetic with its greater objective happiness?
Should those attracted to hedonism because of their concern with subjective experience be interested in what Kahneman calls ‘objective happiness’ – the objective measure of the sum of momentary subjective experiences – or should those sympathetic to hedonism instead focus on subjective appraisals of lives or extended episodes within lives? If what matters is the subjective value of one's life, which is more important, the subjective appraisal of experience – that is, people's attitudes toward their subjective experience – or subjective experiences themselves? If the sum of some measure of pleasures minus pains over time does not match an individual's own appraisal of how good his or her life was over that period, why suppose that it is momentary experience rather than retrospective appraisal that matters?
Kahneman himself has eloquently expressed some of these concerns. He writes,
But memories have an attribute of permanence which lends them a weightiness that the fleeting present lacks: they endure and populate the mind. In the words of the novelist Penelope Lively (Reference Lively1993:15), ‘A narrative is a sequence of present moments but the present does not exist.’ Because memories and stories of the past are all we ultimately get to keep, memories and stories often appear to be all that matters. (Reference Kahneman, Kahneman and Tversky2000b: 687)
A good life is not a sum of the net goodness of its moments, the better the moments, the better the life. With some gyrations, old-fashioned hedonists can agree with this platitude. But the new hedonists’ concern with net momentary pleasures commits them to an implausible ‘toting up’ view (Griffin Reference Griffin1986: 34–35) of well-being, tempered only by the complicated interdependencies between momentary pleasures and pains.
I conclude that the conflicts between the quality of experiences and people's recollections of experiences underscores the limited and often indirect relevance of pleasures and pains to a good life. Psychologists are now better able to measure the quality of momentary experiences. But the cost of this ability is the divorce of ‘objective happiness’ from people's retrospective and prospective views of their own experience. And this separation makes the relevance of momentary experiences to welfare questionable.
Consider the fact that occasionally people with advanced Alzheimer's disease appear to be very happy. If mood is the measure of welfare, their lives are going well. Like most people, I believe instead that those who are happily demented are very badly off, although not as badly off as those who are both demented and unhappy.Footnote 16 A hedonist who identifies happiness with attitudes can accommodate the belief that, regardless of their mood, those who are senile are living badly. Although dementia can sometimes bring pleasure, the achievements and experiences that lead people to judge themselves happy consist in a very different sort of experience. New hedonists cannot make this response, because what diminishes the subjective quality of life of those who are demented but nevertheless in a good mood is the lack of connection of the momentary states to the past and future. Evaluating the momentary mental states and then adding them up, one may arrive at a large quantity of positive feelings. But these feelings don't add up in the way that the experiences in a good life cumulate.
In one of the few passages where Kahneman discusses these concerns, he writes as follows:
The goal of this discussion is not to reject the memory-based view, which is indeed irresistibly appealing, but to point out that intuition is strongly biased against a moment-based view. The approach proposed here is bound to be counter-intuitive even if it has merit – that was one of the reasons for proposing it. Although wholly devoid of permanence, the experiencing subject deserves a voice. (Reference Kahneman, Kahneman and Tversky2000b: 687)
Perhaps. But, as already noted, it is far from clear what authority or weight this voice deserves.
Why the conflict between experienced and attitudinal hedonism undermines both
In detailing the conflicts between aggregating measures of momentary experience on the one hand and anticipations and recollections of the quality of the experience of life episodes, the work of Kahneman and others diminishes whatever limited appeal hedonism of either sort might have. As Kahneman argues, his results raise serious questions about the reliability of traditional measures of (attitudinal) happiness based on questions such as ‘On the whole, are you very satisfied, fairly satisfied, not very satisfied, or not at all satisfied with the life you lead?’ At the same time, the conflicts between the sum of momentary pleasures and retrospective appraisals expose how tenuous is the connection between ‘objective happiness’ and well-being. If objective happiness and retrospective appraisal do not coincide, one has less reason to regard either as a plausible view of well-being. Rather than shoring up hedonism by demonstrating that objective happiness can be measured reliably, the work of the new hedonists provides an additional reason to reject hedonism.
5. CONCLUSIONS
If the choice economists faced were between the current preference-based conception of welfare and a hedonistic conception, economists would, I maintain, be better off sticking with what they have got. But there are better alternatives.
This essay began with a problem: If welfare is identified with choice, then mistakes are impossible. But mistakes are obviously not impossible. Non-suicidal American tourists who find themselves beneath London taxis have made serious mistakes. Those who fail to opt into a retirement plan, yet who would not have opted out if the default had been set differently may be making a mistake. Unless economists concede that people sometimes choose what is bad for them, they will be unable even to consider the possibility of mistakes.
One solution to this problem, which this essay has criticized, is to adopt a view of welfare as happiness or pleasure. This proposal is ambiguous, because there are widely differing interpretations of happiness or pleasure. This proposal is implausible, because people whose subjective states are just the same are not necessarily equally well off. ‘Objective happiness’, the new hedonist's concept of experience utility, is measurable but of limited relevance to overall well-being. It implies that those who are happily demented are living good lives. Measures of attitudinal happiness like those used in large-scale surveys have no such absurd consequences and provide some information; but it is not clear what economists should do with that information.
What should normative economists do? One possibility is to adopt a more sophisticated preference-based theory of welfare. Satisfying people's actual preferences does not always make them better off, because their preferences may be based on false beliefs or because their preferences are not self-interested. But satisfying actual preferences is nevertheless typically a good way to make people better off, since people's preferences often are self-interested and their beliefs often reliable. So economists can go on using preference-based techniques for evaluating institutions, outcomes and policies, though with a note of caution. (The findings of behavioural economics suggest that this note of caution should be marked fortissimo.) When defenders of ‘libertarian’ or ‘asymmetric’ paternalism maintain that the test of whether their proposals benefit people lies with people's own judgements or expressed preferences, they may have in mind that welfare is the satisfaction of self-interested, rational and well-informed preferences.Footnote 17
A second possibility is for economists to avoid committing themselves to any philosophical theory of welfare and to take preference satisfaction merely as evidence of well-being, whatever well-being may be (Hausman and McPherson Reference Hausman and McPherson2009). One might object that such a proposal is incoherent. Indeed Hausman and McPherson themselves previously argued that ‘there is no way to defend the claim that preference satisfaction is the best measure of welfare until one has said what welfare is’ (Reference Hausman and McPherson2006: 121). This claim and the incoherence objection assume, erroneously, that words get meanings only through definitions or explicit theories. But this had better not be the case, lest most of what people say be counted as meaningless. Knowing that good health, happiness, enjoyment, the respect of others, intimate friends and so forth generally contribute to welfare gives content to talk of welfare without defining the term. Possessing some idea of what things generally make people better or worse off, economists can reasonably conclude that a good deal of what people do is directed toward increasing their own well-being. Furthermore, it is plausible to maintain that in many circumstances individuals are better judges of what will benefit them than are others. When these conditions hold (that people aim at their own advantage and are good judges of what will benefit them), economists can rely on preferences as evidence concerning welfare. In doing so, they are not supposing that the satisfaction of preferences constitutes welfare, nor are they committed to any other theory of welfare. If people aim at their own advantage and are good judges of what will benefit them, then their preferences will be good evidence of what is good for them. Of course, people do not always aim at their own advantage, and people are sometimes very bad judges of what will benefit them. They are not always self-interested, and they make mistakes. So preference and welfare will sometimes diverge. But preferences can be evidence of welfare without being infallible evidence.
Let me conclude with the suggestion that normative economists should not be concerned exclusively with welfare. There are many other social values, and they are not all beyond the reach of economic modelling. Consider, for example, the costs of illiteracy to the person who cannot read. These can in part be captured by unhappiness or frustrated preferences, but illiteracy need not lead to unhappiness, and someone without an education may not want one. In that case, does illiteracy no longer impose costs on those who cannot read? In much the same way, some disabilities, such as deafness, need not lead to unhappiness, and some of the deaf prefer deafness to being able to hear. When deafness does not lead to unhappiness or frustrate preferences, is deafness no longer a loss? Perhaps the loss attached to an inability to read or to hear can be captured by a more sophisticated view of welfare than happiness or preference. But it might also be, as Amartya Sen in particular has argued (Reference Sen, Nussbaum and Sen1993), that welfare is not all that matters.
