Happiness surveys and public policy: what's the use?
The PR defense sees SWB surveys as evidencing preference utility, a measure of the extent to which individuals' preferences are realized (with individuals permitted to have an intrinsic preference for items that are at least partly nonexperiential, such as consumption, health, liberty, accomplishment, knowledge, and so forth). By contrast, the EQ defense argues that SWB surveys provide useful information about the quality of individuals' mental states. This defense is, in a way, much more straightforward. Although individuals are not infallible about the content of their mental states, surely each individual is generally more epistemically reliable about what she thinks and feels than about the occurrence of what she wants. Moreover, it seems straightforward to design survey questions focusing on experiential quality: for example, "How happy are you?"
Indeed, many scholars in the SWB literature offer (what appears to be) some version of the EQ defense. The most prominent example is Daniel Kahneman, who argues that information about experience utility should play a substantial role in structuring governmental choices. Kahneman and his collaborators (for short, "Kahneman") have pioneered a novel framework ("objective happiness") for measuring experience utility, and have empirically implemented this framework in several large-scale studies. (132) But Kahneman is hardly alone in seeing SWB surveys as evidencing experience utility. There are numerous other SWB researchers who present--or at least seem to present--the EQ defense.
That defense, once more, can take a strong or weak form. In the strong form, the EQ defense of SWB surveys endorses experientialism about well-being. A leading example of this approach is the best-selling book, Happiness: Lessons from a New Science, (133) written by Richard Layard, a prominent SWB researcher. Section A responds to Layard's arguments and, more generally, criticizes the strong EQ defense.
The weak EQ defense is more promising. It refrains from endorsing experientialism about well-being. Experientialism about well-being is, at a minimum, normatively controversial. The weak EQ defense of SWB surveys declines to take sides in that thorny debate. It claims only that good experiences are one aspect of well-being--a claim which seems very hard to deny.
More problematic is the assertion that policymakers should take account of the experiential impact of governmental policies via SWB surveys. In Section B, I illustrate the difficulties with the weak EQ defense of SWB surveys via a close analysis of Kahneman's "objective happiness" framework. Kahneman's work is by far the most systematic attempt, to date, to develop a policy-relevant measure of mental states. And Kahneman now acknowledges (or at least is willing to entertain) that well-being has nonmental aspects. The "objective happiness" framework should thus be understood as a concrete elaboration of the weak EQ defense: one that sees SWB surveys as evidence of experiential quality, and experiential quality as one important determinant (among others) of well-being.
Key objections to the "objective happiness" framework are its implausible presuppositions regarding temporal separability and, even more fundamentally, the separability of the hedonic from the nonhedonic. Relatedly, the framework offers no guidance in how policymakers should integrate information about hedonic utility with nonhedonic information. Finally, although the theoretical elaboration of "objective happiness" uses an "observer" to make time-tradeoff judgments so as to cardinalize hedonic utilities, the "observer" is not to be found in Kahneman's empirical studies, and even the theoretical elaboration fails to allow for heterogeneity in observer judgments/preferences.
Section C sketches a different and arguably more promising approach to integrating information about experiential quality into policy choice: an approach that includes happiness as one of the entries in individuals' preference utility functions, and that employs revealed or stated-preference studies to estimate the extent to which preference utility depends upon happiness.
A. Experientialism about Well-Being: The "Strong" Experience-Quality Defense of Subjective Well-Being Surveys
Experientialism about well-being has a distinguished intellectual history. Bentham, of course, was an experientialist, and so (in a different way) were John Stuart Mill and Henry Sidgwick. (134) But, more recently, the view fell into philosophical disfavor. The most influential twentieth-century work of moral philosophy, John Rawls's A Theory of Justice, (135) adopts a preference-based view of welfare, rather than equating well-being with pain and pleasure or, more generally, good experiences. (136) The other leading contemporary philosophical works on well-being (here, I am thinking of the work of Richard Brandt, (137) James Griffin, (138) John Finnis, (139) Martha Nussbaum, (140) and Wayne Sumner (141)) are also nonexperientialist.
Indeed, for a time, Robert Nozick's discussion of the "experience machine" was generally seen by philosophers as a decisive refutation of experientialism. (142) Over the last decade or so, some scholars have pushed back against this conventional wisdom. (143) Still, it remains the case that experientialism about welfare is a philosophically controversial position. As already explained in Part I, two widely accepted classes of well-being accounts--preferentialism (with the special exception of a preferentialist view that embraces an experientialist conception of "self-interested" (144)) and the objective-good approach--reject the experientialism requirement. Accounts within these classes reject the proposition that the only way to directly change someone's well-being is by changing the content of her mental states (what she thinks, feels, remembers, and so forth).
In Happiness, however, Layard adopts a Benthamite view of well-being and of the proper goals for social policy. He writes, "I believe that Bentham's idea"--that "all laws and all actions should aim at producing the greatest possible happiness"--"was right and that we should fearlessly adopt it." (145)
Happiness almost completely ignores the contemporary philosophical debates about well-being. Perhaps that is not surprising, given academic specialization (Layard is an economist) and the book's aim to reach a popular audience. Still, one can ask: has Layard advanced or at least sketched plausible arguments for why governmental policy should be solely focused on promoting individual happiness, rather than nonexperiential features of individual lives (except as an instrumental means to more happiness)? Layard argues as follows:
Why should we take the greatest happiness as the goal for our society? Why not some other goal--or indeed many? What about health, autonomy, accomplishment or freedom? The problem with many goals is that they often conflict, and then we have to balance one against the other. So we naturally look for one ultimate goal that enables us to judge other goals by how they contribute to it. Happiness is that ultimate goal because, unlike all other goals, it is self-evidently good. If we are asked why happiness matters, we can give no further, external reason. It just obviously does matter. As the American Declaration of Independence says, it is a "self-evident" objective. By contrast, if I ask you why you want people to be healthier, you can probably think of reasons why--people should not be in pain, they should be able to enjoy life and so on. Similarly, if I ask you about autonomy you will point out that people feel better if they can control their own lives. Likewise, freedom is good because slavery, prison and the secret police lead to nothing but misery. So goods like health, autonomy and freedom are "instrumental" goods--we can give further, more ultimate reasons for valuing them.... To help us promote the greatest happiness, we obviously need to understand what conditions affect people's happiness, and by how much. This is now becoming possible on an empirical basis.... But unless we can justify, our goals by how people feel, there is a real danger of paternalism. We ought never to say: this is good for you, even though it will never make you or others feel better. On the contrary, if we want to measure the quality of life, it must be based on how people feel. (146)
After addressing objections to happiness-maximization, Layard rearticulates his position in this way: "So we come back to the central idea of a humane ethic that values what people want for themselves, for their children and for their fellow citizens. And that is their happiness." (147)
In these passages, one can tease out three separate lines of defense of experientialism about well-being. None are particularly persuasive.
(1) Monism: "The problem with many goals is that they often conflict." (148) A view that equates well-being and happiness is unidimensional or "monistic." By contrast, if we allow well-being to depend upon both happiness and nonexperiential items such as health, autonomy, accomplishment or freedom, it becomes multidimensional. On such an account, how is government supposed to choose a course of action when the dimensions conflict?
What this argument overlooks is that there are very well-developed techniques in economics for assigning a given individual a single utility number as a function of her attainment on multiple dimensions. (149) Conversely, experientialism is hardly a sure recipe for avoiding the complexities of multidimensionality. On the most plausible accounts, experiential quality is itself multidimensional: a matter of good perceptions, cognitions, and memories, not just pains and pleasures. Many SWB researchers see subjective well-being (good mental states) as a composite of positive and negative affect and feelings of satisfaction. (150) Even the narrower, Benthamite view may not really be monistic. As discussed at greater length below, painfulness and pleasantness may turn out to be separate dimensions. (151)
(2) Only Happiness Has Self-Evident Intrinsic Value: "Happiness is that ultimate goal because, unlike all other goals, it is self-evidently good.... [G]oods like health, autonomy and freedom are 'instrumental' goods--we can give further, more ultimate reasons for valuing them." (152) Layard's idea of "self-evidence" invokes a standard feature of normative reasoning, namely that some normative propositions might be supported by direct appeal to intuitions. (153) Layard is surely correct that many of his (and this Article's) readers will strongly intuit that happiness has intrinsic welfare value: happiness, as such, increases well-being. But many readers will also intuit that physical health, autonomy, knowledge, relationships, or accomplishment have intrinsic welfare value. This relates to pluralism/monism: there is nothing at all confused in having the intuition that multiple types of individual attributes, both happiness and other goods, are intrinsic welfare components.
In characterizing happiness as a "self-evident" good, Layard not only appeals to intuitions in favor of happiness, but also emphasizes the strength of such intuitions. Assume, arguendo, that intuitions concerning the welfare relevance of health (for example) are generally weaker than intuitions concerning the welfare relevance of happiness. It would be a fallacy to conclude that only happiness, not health, has intrinsic welfare value. If I intuit that proposition P* is true, then (ceteris paribus) my normative views should be consistent with P*, even if I hold the intuition with less than complete certainty, and even if there is some other proposition P (logically consistent with P*) that I intuit more strongly.
In principle, Layard might try to undercut intuitions concerning the welfare relevance of nonexperiential items in a different way, by arguing that (a) such intuitions do not survive idealization (with full information, and thinking rationally, no one intuits that health, etc., has intrinsic welfare value), or that (b) such intuitions are much less widely shared than the intuition regarding the value of happiness. The first tack seems a nonstarter--undercut by the many serious thinkers who have embraced nonexperientialist accounts. The second tack is empirically speculative. Layard doesn't present (or attempt to present) evidence to support it. (154) And evolutionary considerations actually suggest that intuitions supporting the intrinsic significance of at least some nonexperiential items are likely to be quite widely shared. (155)
Layard's key argument against the intrinsic value of health, autonomy, and freedom is yet another one: namely, that we can give instrumental reasons for promoting these items. This argument is fallacious. Something can be both intrinsically and instrumentally valuable. In particular, imagine that there are two intrinsic components of welfare: being happy and being healthy. Then we can reason instrumentally about health (asking about the effect of health-promotion policies on happiness), but we can also point to the health-promotion effect of a policy as an intrinsic reason to support it. Symmetrically, we can reason instrumentally about happiness by asking about the effect of happiness-promotion policies on health. The health benefits of affects and cognitions are in fact a large topic of research among public health researchers! (156)
Thus, the premise that nonexperiential items possess instrumental value, as causal precursors to happiness, does not imply the conclusion that they lack intrinsic value.
(3) Paternalism: "[U]nless we can justify our goals by how people feel, there is a real danger of paternalism." (157) Government acts paternalistically toward some individual when it restricts her choices for her own good--when it limits what she can do, not because of third-party effects, but because it believes that she will fail to promote her own well-being.
Far from being antipaternalistic, experientialism has the real potential for paternalism. (158) If citizens self-interestedly prefer nonexperiential items, a government with a mandate to promote good experiences has a justification for restricting or altering their choices. For example, if mental health interventions are especially conducive to happiness, but some individuals have a substantial intrinsic desire to be physically healthy, an experientialist government might try to tilt those individuals' health expenditures away from their preferred mix as between physical and mental health. It is preference views--drawing an equivalence between what someone wants (under stipulated conditions) and what is good for that person--that have the conceptual resources to resist paternalism. (159) Conversely, the possible disjunction between the course of action that maximizes the realization of someone's self-interested preferences, and the course of action that maximizes the quality of her feelings or experiences, renders an experiential account potentially paternalist.
Layard obscures this potential via his assertion that individuals self-interestedly prefer their own happiness. "So we come back to the central idea of a humane ethic that values what people want for themselves, for their children and for their fellow citizens. And that is their happiness." (160) But this is an empirical assertion--one that Layard does not back up with evidence, and one that the studies reviewed in Part I.B. do not clearly support.
Of course, the sheer fact that a well-being account is paternalistic does not provide a decisive reason to reject it. Intuitively, some individuals do fail to promote their own well-being. But SWB scholars such as Layard should not inflate the virtues of experientialism by obscuring where it stands on paternalism.
I have parsed Layard's brief for experientialism and found it wanting. But better arguments for his position might be available. John Bronsteen, Christopher Buccafusco, and Jonathan Masur have recently offered a lengthy defense of experientially focused policy analysis (161). And, as mentioned, some contemporary academic philosophers have tried to mount a case for (or at least deflate some of the standard objections to) experientialism. (162) To my mind, the strongest reason to endorse an experientialism requirement is a reason that Layard does not mention: the "good-for" aspect of welfare. (163) Well-being is subject-relative goodness. Some occurrence enhances Sonya's well-being only if it is good for Sonya, rather than merely being good in an impersonal sense, or good for someone else. Call this the "non-remoteness" constraint on well-being:
an occurrence has a direct (164) impact on someone's welfare only if it is not too "remote" from her.
At first blush, the non-remoteness constraint presents a plausible case for experientialism. As Shelly Kagan explains:
Suppose I meet a stranger on a train. He tells me his story, and I form the desire that he succeed in his projects. We then part, and I never hear of him or even think of him again. If he does in fact succeed, then my desire has been satisfied. According to the desire theory, then, this makes me better off. But this is intuitively an absurd claim. Obviously my level of well-being is not affected at all by the success of the stranger.... This suggests that the unrestricted desire theory [of well-being] is hopelessly broad. A theory of well-being must explain which facts constitute my being better off. So they must be facts about me. Since my desires can range over facts that have nothing whatsoever to do with me, the satisfaction of such desires cannot constitute my well-being.... From this perspective, the position of mental statism no longer seems so arbitrary. At least it seems to keep the content of well-being within reasonable bounds, for facts about my mental states are certainly facts about me. In contrast, it is far from clear whether anything external to my mind ... can count as--in the relevant sense--facts about me. If not, then the limits of well-being must be drawn at the limits of our minds. (165)
The "stranger" case described by Kagan derives originally from philosopher Derek Parfit. (166) Eric Posner and I provided a similar case in our book on cost-benefit analysis, involving a person named Sheila and an endangered species, the Sri Lankan squirrel.
One outcome involves ... the continued existence of the Sri Lankan squirrel ...; the other is the extinction of this species. Sheila has never traveled to Sri Lanka, and never intends to, nor is she an environmentalist who's made species preservation her life's work, but she still (slightly, say) prefers the first outcome because she believes that morality includes environmental values disfavoring the disappearance of species. It seems odd to say that the nonextinction of the squirrel makes Sheila better off. (167)
The "stranger" and Sri Lankan squirrel cases underscore the non-remoteness constraint on well-being. More specifically, within the confines of a preference-based account of welfare, they show the need to restrict the category of welfare-relevant preferences. The cases illustrate that someone might have a fully laundered preference (a preference that is well-informed, rational, intelligible, and so forth) for outcome x over y, and still not be better off in x than y. In order to survive the remoteness objection, the preference-based account must say that person i is better off in x than y if and only if i has a self-interested, laundered preference for x over y. The reasons thus to restrict the category of welfare-relevant preferences include not just our intuitive reactions to particular hypothetical cases, but also deeper considerations regarding the possibility of self-sacrifice and of moral reasoning. (168)
Kagan concludes, however, that the non-remoteness constraint does not, on balance, argue for experientialism. I agree with Kagan. In particular, the literature on preferentialism has offered at least four possible general solutions to the problem of differentiating self-interested from non-self-interested preferences. (169) (1) Experientialism: An individual's preference is self-interested if and only if the fundamental arguments (170) for the preference are the individual's mental states. (2) Experientialism plus physicalism: An individual's preference is self-interested if and only if the fundamental arguments for the preference are the individual's mental states or physical states. (3) Existence-entailment: An individual's preference is self-interested if and only if the realization of the preference entails the individual's existence. (4) Sympathy: An individual's preference is self-interested if and only if based in self-sympathy, an attitude of care and concern for herself.
The experientialist conception of "self-interested" strikes me as too narrow. Surely occurrences within a person's physical body are not "remote" from her. As Kagan explains, even if we insist that a preference is self-interested only if its fundamental arguments are the person's nonrelational attributes, that would lead us to conception (2) in the previous paragraph, not (1). (171) We get from a nonrelationality requirement to experientialism only by adopting a controversial view of personal identity--namely, that a particular human person is just a bundle of psychological properties, rather than a particular human being who has a brain and a body. (172) Moreover, I would go wider than experientialism plus physicalism. Given that practical rationality, theoretical rationality, and affiliation are three central capacities of human persons--the capacities to make choices in the pursuit of goals, to acquire knowledge, and to form relationships with others--it seems to me that preferences for success in the exercise of those capacities can count as self-interested even if outside the scope of conceptions (1) and (2). Such preferences are not merely "intelligible" (as I discussed in Part I), (173) but intelligible as a matter of self-interest. A second case that Posner and I discussed in our book, the case of the deceived scholar, illustrates this point. (174)
Bronsteen, Buccafusco, and Masur argue that I have "directly contradict[ed]" myself in denying that the squirrel's survival benefits Sheila, and yet rejecting the experientialist conception of self-interest, conception (1). (175) Their characterization, if accurate, would also hold true of anyone else who adopts conception (2), (3), (4), or for that matter any view which analyzes self-interested preferences other than as preferences for experiences. But, in fact, there is no contradiction here: just a position intermediate between the narrow insistence that what is good for me must occur within my head, and an overbroad willingness to include the realization of any preference (including Sheila's, or a preference for the stranger's success in Parfit's case) as improving my welfare.
The brief discussion of the last several paragraphs will hardly satisfy the proponent of experientialism. The non-remoteness problem is very difficult, and the experientialist solution to that problem certainly does deserve close and serious consideration. Although I believe that this argument for experientialism fails--as do others advanced by contemporary philosophical experientialists, and by Bronsteen, Buccafusco, and Masur--I lack space to pursue the analysis here. Notwithstanding Nozick's "experience machine," experientialism should not be rejected out of hand.
Reciprocally, however, SWB scholars tempted to equate well-being and subjective well-being should recognize that they are entering a hornet's nest of disputation, and should be prepared to do serious normative battle in defense of experientialism. Lacking the appetite or ammunition for such battle, the SWB scholar might instead claim that good experiences are at least one aspect of well-being, whether or not the sole determinant.
That claim is hard to dispute. But does it justify the use of SWB surveys as a basis for policy choice? It is to that question that we now turn.
B. "Objective Happiness" and the "Weak" Experience-Quality Defense
This Section discusses Kahneman's "objective happiness" approach. (176) I remind the reader that "Kahneman" is shorthand for Kahneman and the distinguished collaborators with whom he has coauthored the theoretical and empirical articles developing this framework.
Kahneman disavows (or at least does not commit himself to) an experientialist conception of well-being. He writes:
Defining happiness by the temporal distribution of experienced affect appears very narrow, and so it is. The concept of objective happiness is not intended to stand on its own and is proposed only as a necessary element of a theory of human well-being. A comprehensive account of well-being inevitably brings in philosophical considerations and a moral conception of "the good life," which are not easily reduced to experienced utility. However, good mood and enjoyment of life are not incompatible with other psychological criteria of well-being that have been proposed, such as the maintenance of personal goals, social involvement, intense absorption in activities, and a sense that life is meaningful. Clearly, a life that is meaningful, satisfying, and cheerful should rank higher on the scale of well-being than a life that is equally meaningful and satisfying but sad or tense. Objective happiness is only one constituent of the quality of human life, but it is a significant one. (177)
Kahneman, here and elsewhere, tries to bolster the normative appeal of "objective happiness" by stressing that it is "only one constituent of the quality of human life." (178) The framework thus merits close critical attention, not only in its own right, but as a concrete elaboration of the weak EQ defense of SWB surveys.
The framework offers a methodology for measuring the hedonic or affective aspect of individual experience: whether a mental state is affectively positive or negative, that is, painful or pleasurable. (Kahneman often uses "hedonic," "affective," and "pain"/"pleasure" interchangeably, and I will follow his usage here.) An experience is hedonically/affectively positive, according to Kahneman, if it feels good and if the individual wants the experience to continue. An experience is hedonically/affectively negative if it feels bad and the individual wants it to stop.
The formal model underlying "objective happiness" presumes that there is a neutral hedonic level, and that a given individual (the subject), at a given moment, can meaningfully characterize her current, momentary, mental state as affectively positive, negative, or neutral--as hedonically better than, worse than, or equally good as the neutral level. Moreover, she can order possible momentary experiences within each of the two affective domains. Given two affectively positive experiences, the first can be ranked as more, less, or equally positive as the second--various momentary pleasures are not just pleasures simpliciter (better than neutral), but more or less pleasurable. Similarly, the subject can determine whether one hedonically negative momentary experience is more or less negative than another.
These intra-domain hedonic rankings of momentary experiences can be represented (Kahneman assumes) by numbers--by momentary hedonic utilities--which, thus far, are merely ordinal. If the subject assigns -3 to momentary experience A, and -6 to momentary experience B, those numbers mean that both are worse than neutral, and that B is more negative than A; the numbers -3 and -8 would equally well capture this hedonic information. In order to cardinalize momentary utilities, Kahneman introduces an "observer." (179) The observer ranks temporally extended episodes ("profiles"), each consisting of a series of affective states (positive, negative, or neutral) experienced for some length of time. The observer's ranking of the profiles is consistent with axioms of time neutrality, monotonicity, and separability. Time neutrality says that adding neutral time before or after a profile does not change where the profile is located in the observer's ranking. Monotonicity says that replacing one or more positive momentary experiences within a given profile with more intensively positive momentary experiences must improve how the profile is ranked by the observer, and symmetrically for negative experiences. And separability says that the ranking of two profiles does not depend on what is experienced during moments when experiential quality is the same. (180)
If the observer's ranking satisfies these axioms, then there is a cardinal measure of momentary hedonic value--call it the v(.) function--such that the observer ranks one profile over another if and only if the sum of the duration-weighted v(.) values for the first profile is greater than the sum of the duration-weighted v(.) values for the second. For example, let profile P consist of experiences A, B, and C, each for two units of time, and let P* consist of experience D for five units. Then if the observer prefers P to P*, it will be the case that 2v(A) +2v(B) + 2v(C) > 5v(D).
The v(.) function is cardinal in the sense that its values are added together to yield the overall value of a profile. The v(.) function, strictly speaking, is not unique, but unique up to a positive ratio transformation. Thus the summation or averaging of momentary v(.) values--and for that matter, the summation or averaging of the overall values of profiles--is a meaningful operation. (181) Because the v(.) function is cardinal, it can serve as the theoretical basis for the policy metrics that Kahneman recommends--gross national happiness and other such metrics that add or average measures of experiential quality across moments and persons.
To see how information about the observer's ranking allows us to arrive at a cardinal v(.) function, return to the case in which the subject experiences A and B as both negative, and A as less negative than B. Although this hedonic data--without more--is not sufficient to determine whether the numbers assigned to the two experiences are -3 and -6, or -3 and -8, imagine that we now learn that the observer is indifferent between a profile with T hours of A experience and a profile with T/2 hours of B experience. This information constrains the v(.) value of A to be a negative number which is half the v(.) value of B. Not only must it be the case that v(B) < v(A) < 0--which follows immediately from the fact that both feel painful to the subject, and B feels more painful. In order to represent the observer's indifference between the two profiles just mentioned, it must be the case that v(A)T = v(B)T/2, i.e., v(A)/v(B) = 1/2. This fact about the observer's ranking rules out the pair -3, -8.
To reiterate, the v(.) function is identified as that cardinal function which both respects the hedonic intensity of momentary affective experiences, as registered by the subject, and additively represents the observer's ranking of all possible temporally extended episodes. The observer's ranking of temporally extended episodes, as well as the subject's felt experience of momentary pains and pleasures, plays a critical role--in Kahneman's theoretical apparatus--in arriving at v(.).
The "objective happiness" framework, although an important advance, has substantial limitations as a method for integrating information about experiential quality into policy analysis--limitations either in the theoretical statement of the framework or in its empirical implementation by Kahneman. Those limitations, I believe, are the following.
1. Is the Hedonic Aspect of Experience Unidimensional or Bidimensional? Kahneman sees hedonic value as unidimensional. Pleasure and pain are twin poles of a single spectrum bifurcated at hedonic neutrality, just as positive and negative numbers correspond to two segments of a single line running through the number 0.
This assumption of the unidimensionality of affective experience is undermined by substantial psychological evidence for a two-dimensional (or, more generally, higher dimensional) affective space. (182) Such evidence includes (1) neurological evidence, showing that pain and pleasure arise in separate brain regions, which can be jointly activated; (2) survey research that looks to how individuals describe their current mental state, finding that words expressing negative affect (words such as "distressed," "upset," "hostile," "irritable," "scared," "afraid," "ashamed," "guilty." "nervous," and "jittery") are surprisingly uncorrelated with words expressing positive affect (words such as "attentive," "interested," "alert," "excited," "enthusiastic," "inspired." "proud," "determined," "strong" and "active"); (3) possible behavioral evidence of the dual activation of positive and negative affective systems; and (4) introspection, namely the possibility of states that are simultaneously painful and pleasurable. Nico Frijda eloquently summarizes the introspective evidence:
In actual fact, pleasure and pain can coexist at the same moment. and they may do so without annulling each other; masochistic pleasure is a case in point. Simultaneous arousal of both affects leads to complex interactions: the confusing feelings of ambivalence, the experience of conflict, a pungency and excitement added to the pleasure, a sweetness added to the pain, as in nostalgic remembrance. Ambivalence as well as bittersweet experience lead to the hypothesis that pleasure and pain are not the opposite outcomes of one single process, but the outcome of two somewhat independent processes. (183)
How, specifically, might the multidimensionality of affect jeopardize Kahneman's framework for measuring hedonic value? For simplicity, consider the two-dimensional case (the discussion generalizes). Assume that affect is "bivalent": positive and negative affect are not twin poles of a single dimension, but two, orthogonal aspects of experience that can be jointly realized: a moment of experience can possess positive affect, negative affect, or both. Kahneman's formal model does not appear to cover this last case. On that model, negative hedonic utilities are assigned to painful mental states, representing that their affective quality is worse than neutrality; and positive hedonic utilities are assigned to pleasurable mental states, representing that their affective quality is better than neutrality. But what sort of numbers should be assigned to bivalent experiences?
Kahneman, indeed, takes very seriously the bivalence objection, and at multiple junctures seeks to rebut it. First, Kahneman argues that bivalent experiences occur seldomly, because the brain's pain and pleasure systems are mutually inhibitory.
Cacioppo, Gardner, and Berntson (1999) point out that positive and negative affective states are processed by different neural systems and may be activated concurrently.... However, the systems are not functionally independent, and there is evidence that they inhibit each other. Lang (1995) has shown, for example, that watching pleasant pictures of food or smiling babies attenuates the startle response to a loud sound, whereas startle is actually enhanced in the presence of disgusting or horrible pictures. (184)
This empirical response argues that, even if the "objective happiness" measurement framework is only applicable to a subset of experiences--univalent ones--the framework remains a useful tool, because most experiences are univalent. However, the assertion of mutual inhibition and, thus, the rarity of bivalence, are empirically contestable.
But Kahneman also offers a second rebuttal to the bivalence objection--one that, if persuasive, would obviate the first. He argues (seemingly) that the brain assigns even bivalent experiences a net hedonic value, namely on balance positive or negative.
The bivalent nature of the Good/Bad system is not necessarily incompatible with the notion that most moments can be usefully characterized by a single value on a bipolar Good/Bad dimension. A bivalent system yields a bipolar dimension if the separate mechanisms that mediate Good and Bad are mutually inhibitory or reciprocally innervated or if the relevant output of the system is the difference between the levels of activity of the two mechanisms.... Davidson (1992) suggested that the brain may compute ... the difference of the levels of activity in the separate systems that mediate positive and negative affect. He proposed that the [Good/Bad] value corresponds to the difference.... (185)
If the claim, here, is that any experience feels on balance good or bad, that claim is undermined by the introspection data which Frijda mentions. But perhaps the claim is slightly different, namely that every bivalent experience (although perhaps ambivalent in how it feels) corresponds to some univalent experience for purposes of the observer's ranking of temporally extended experiential profiles. The claim, thus understood, is no longer about the psychology of momentary affect, but rather about how univalent, bivalent, and neutral experiences fit together to determine the well-being value of profiles. But is the claim true? For example, does a sequence of balanced bittersweet moments, wherein pain and pleasure have equal intensity, possess just the same well-being value as a sequence of neutral moments with neither affective system activated?
2. Do Observers Have the Same Ranking of Hedonic Profiles? Let us ignore, henceforth, the possibility of bivalent experiences. Let us also place to one side deep questions regarding the accessibility of hedonic (and more generally mental) life, namely whether a subject can ever communicate to an observer what she (the subject) is feeling like. Assume, instead, that subjects can classify experiences as hedonically good, bad or neutral; can rank experiences with the same valence; can describe experiences using labels or instantaneous utilities; and that these descriptors more or less succeed in communicating to a given observer what the experiences feel like. (186)
Kahneman's formal model assumes a single ranking of "profiles," that is, temporally extended hedonic episodes. In effect, he assumes that all observers will converge in ranking any given profile as better, worse, or equally good as any other. But why should this be the case? Kahneman is ambiguous as to whether observers are meant to rank profiles by consulting (a) their preferences (Which profile would I prefer to experience?) or (b) their judgments (Which profile do I think is better for the subject's well-being?). In either event, there is no good reason, conceptual or empirical, to suppose interobserver convergence.
Understand that interobserver convergence is not entailed by the requirement that each observer's v(.) function respect the subject's rating of momentary experiences, nor by the requirement that each observer's ranking of profiles satisfy axioms of time-neutrality, separability, and monotonicity. Two observers might each satisfy all these requirements and yet rank profiles differently.
For example, assume that the subject rates momentary experiences A through C as better than neutral and A better than B better than C. Observer One is indifferent between T hours of each experience and 2T hours of the subsequent experience (so between T hours of A and 2T hours of B, and between T hours of B and 2T hours of C). Observer Two is indifferent between T hours of each experience and 3T hours of the subsequent experience. Then Observer One's ranking is represented by the function [v.sub.1](.) such that [v.sub.1](A) > [v.sub.1](B) > [v.sub.1](C) > O, [v.sub.1](A)/[v.sub.1](B) = 2, [v.sub.1](B)/[v.sub.1](C) = 2. Observer Two's ranking is represented by the function [v.sub.2](.) such that [v.sub.2](A) > [v.sub.2](B) > [v.sub.2](C) > 0, [v.sub.2](A)/[v.sub.2](B) = 3, [v.sub.2](B)/[v.sub.2](C) = 3.
Each observer's ranking of profiles (using the sum of duration-weighted v(.) values) satisfies the axioms of time-neutrality, monotonicity, and separability, as well as respecting the subject's feelings. And yet the observers disagree in their ranking of certain profiles. For example, it is easy to see that Observer One will prefer the profile consisting of two hours of B followed by one hour of C, over the profile consisting of one hour of A, whereas Observer Two will have the opposite preference.
If observers can have different rankings of profiles, Kahneman's framework for arriving at a cardinal measure of momentary hedonic experience collapses. If one observer prefers profile P to profile P*, and another observer has the opposite preference, then clearly there is no v(.) function such that the sum of duration-weighted v(.) values will both assign profile P a greater overall value than profile P*, and assign profile P* a greater overall value than profile P. Consider the case just mentioned in which Observer One prefers two hours of B followed by one hour of C to one hour of A, whereas Observer Two has the opposite preference. It is mathematically impossible to find a v(.) such that: 2v(B) + v(C) > v(A) and 2v(B) + v(C) < v(A).
In short, when observers have different rankings of profiles, there will be no v(.) function that represents all of their rankings. Rathen the cardinal measurernent of hedonic experience will be observer-relative. Observer One will have a cardinal function [v.sub.1](.), the sum of whose duration-weighted values will represent his ranking of profiles: Observer Two will have a different cardinal function [v.sub.2](.), the sure of whose duration-weighted values will represent his ranking of profiles: and so on. Policy aggregates such as gross national happiness, requiring a cardinal measure of hedonic quality, will also become observer-relative. For example, gross national happiness might be greater in outcome x than y, as calculated by adding up the momentary utilities that track Observer One's ranking of profiles; but greater in outcome y than x, as calculated by adding up the momentary utilities that track Observer Two's ranking of profiles. (187)
3. Circumventing the Observer in Empirical Implementation. In his empirical implementation of the "objective happiness" approach, Kahneman has not actually surveyed an observer or observers to rank hedonic profiles and thereby cardinalize hedonic utilities. (Consider, by way of contrast, empirical implementation of the QALY approach to policy analysis, in which observer preferences over health states are actually used to cardinalize health values.) (188) Rather, Kahneman has gone directly from subjects' descriptions or ratings of their health states, to conclusions about their "objective happiness."
Circumventing the observers renders the empirical exercise less onerous. It also suppresses difficulties that arise when observers have divergent rankings. But one can ask whether the results of such an exercise have much to do with the account of "objective happiness" that Kahneman defends in his theoretical work.
Consider the study of working women in Texas, which I mentioned in the Introduction. (189) Recall that each respondent was asked to recollect the episodes of her previous day, and to rate each episode on various scales of positive or negative affect, all ranging from values of 0 to 6: for example, how "happy" she felt on a scale of 0 to 6 during the episode, how "warm/friendly" she felt, how "frustrated/annoyed" she felt, or how "depressed/blue" she felt. A measure of the respondent's "positive affect" during the episode was calculated by averaging how she rated the episode on three of the scales (happy, warm/friendly, enjoyment), and her "negative affect" during the episode was similarly calculated by averaging her ratings on six other scales (frustrated/annoyed, depressed/blue, hassled/ pushed around, angry/hostile, worried/anxious, criticized/put down).
These positive and negative values were averaged, across subjects, to yield mean affect ratings of different activities. For example, Kahneman round that eating has an average positive-affect rating of 4.34, exercise of 4.31, and watching TV of 4.19. He used such averages not only to compare the affective value of different activities, but also to analyze the diurnal pattern of affect, and to assess whether affect is more influenced by an individual's income or her temperament.
However, the mean affect ratings employed in the Texas study seem quite arbitrary--arbitrary from the perspective of Kahneman's own account of objective happiness. Moreover, this is true even if one brackets the problem of interobserver divergence. If observers converge--a heroic assumption--then Kahneman's model allows for the assignment of observer-independent cardinal values to momentary experiences: cardinal values, the duration-weighted sums of which correspond to the single ranking of profiles that each and every observer shares. As already explained, these cardinal values express time-tradeoff judgments. If v(A) = 3v(B), then every observer is indifferent between one hour of A and three hours of B.
But there is no particular reason to think that the values elicited from subjects via the rating question in the Texas study correspond to these v(.) values. When a subject says that her happiness during commuting was "3" on a 0-to-6 scale, and that her happiness during socializing was "5" on a 0-to-6 scale, is she using these numbers to express her time-tradeoff judgments? Is she saying that she is indifferent between T hours commuting, and 3/5T hours socializing? Why assume that this is what she means? Perhaps she means the numbers only to have ordinal significance: "5" for socializing and "3" for commuting just means that socializing is more pleasurable than commuting. Or perhaps she intends the numbers to do more than represent the hedonic ordering of experiences, but to communicate some feature of those experiences, or of her ranking of temporal or probabilistic bundles of experiences, other than her time-tradeoff judgments. (190)
In his most recent empirical work, Kahneman employs a different approach (the "U-index") to measure the hedonic value of moments. (191) Subjects are still asked to assign numerical ratings to a given episode on different scales of positive and negative affect. But the episode is then assigned a value of 1 (if the highest rating was for a scale of negative affect), otherwise 0. Kahneman claires that this approach is purely ordinal. (192) That claim is somewhat suspect. (193) A truly ordinal, and more direct, approach would be to ask subjects a binary question, namely to characterize each episode as, on balance, worse or better than neutral experience--and to assign it the number 1 in the first case, 0 in the second. But perhaps this direct, binary, survey would be unreliable, and the actual U-index methodology can be seen as a reliable proxy for it.
In either event, assigning experiences a number of 1 (for unpleasant) or 0, and then adding up or averaging these binary numbers, is a very crude way to assess overall or average hedonic value of a person or a group. We lose all information about the relative intensity of hedonic experiences. Jim's profile of experiences on Tuesday might be assigned a larger average U-index value than Jim's profile today, even though all observers would converge in preferring the Tuesday profile or in judging it better. (194) The U-index study, like the Texas study, avoids the empirical encumbrance of actually introducing observers, but it does so at the cost of results that seem arbitrary by the lights of Kahneman's own theory of how to measure affects.
4. Temporal Separability. Kahneman's measurement scheme assumes the temporal separability of hedonic value. Roughly, this means that the sequencing of momentary utility does not affect the ranking of temporally extended episodes. More precisely, it says that if two episodes have the same hedonic value during some moment or moments, the ranking of the episodes does not depend on what particular value that is.
Kahneman's argument for temporal separability runs as follows:
The ordering of experiences can affect the utility they confer. For example, a strenuous tennis game and a large lunch yield a better experience in one order than in the other because the enjoyment of the tennis game is sharply reduced when it follows lunch. The condition of separability, which states that the contribution of an element to the global utility of the sequence is independent of the elements that preceded and followed it, is often violated when the sequences are described in terms of physical events such as lunch and a tennis game. In a moment-based treatment of total utility, however, the elements of the sequence that is to be evaluated are not events but rather moment utilities associated with events. Because all the effects of the order of events are already incorporated into moment utilities, the order of these moment utilities no longer matters. (195)
However, Kahneman's insistence that "the order of ... moment utilities no longer matters" is sheer ipse dixit. We are back to the issue of observer convergence and divergence. Might not some observers judge/prefer lives that trend upwards in terms of momentary hedonic value, just as some judge/prefer lives that trend upwards in terms of income or health? (196)
5. Integrating Hedonic Value with the Nonhedonic Components of a Good Life. Hedonic value is one determinant of well-being, but not the only determinant. It is implausible to claim that an individual's welfare is solely constituted by the affective quality of her experiences. First, such a claim would amount to a kind of experientialism about well-being; and, as I have already discussed, experientialism is quite controversial.
Second--and independent of the debate about experientialism--we should reject hedonism about mental well-being, that is, the position that hedonic value is the only welfare-relevant aspect of an individual's mental life. (197) Various examples can be constructed to drive home this point, but consider just one. Cheery is upbeat, but forgets much of what happens to him, and his stock of propositional knowledge is pretty mediocre. Grumpy goes through daily life in an affective state that is at or slightly below neutral, but he is keenly focused on the sights and sounds around him, can recollect his past in rich detail, and has educated himself in various fields. Assume, further, that Cheery and Grumpy are more or less the same in their objective characteristics (income, health, job status, social life). Is it clear that Cheery lives a better life than Grumpy? Hardly.
Indeed, hedonism about mental well-being is regularly rejected by SWB researchers themselves. Ed Diener, a leading figure, has consistently defined subjective well-being as a hybrid of affect and life satisfaction. (198) Paul Dolan, Richard Layard, and Robert Metcalfe, in a report laying the groundwork for large-scale gathering of SWB statistics by the government of the United Kingdom, recommend that an individual's SWB be assessed using three different kinds of measures: measures of affect, measures of life satisfaction, and "eudaimonic" measures, the last assessing how "rewarding" or "worthwhile" individuals perceive their activities to be. (199) Kahneman himself has reversed course and now concedes:
[E]xperienced and evaluated well-being should both be measured, and ... the measures should be explicitly separated. Contrary to a position that one of us espoused earlier (Kahneman 1999), measures of evaluated well-being are not simply flawed indicators of objective happiness (experienced well-being). Evaluation and memory are important on their own ... because people care deeply about the narrative of their life. (200)
However, Kahneman continues to argue that his "objective happiness" framework is a good tool for incorporating hedonic value into policy assessment. The hedonic utility of the positive and negative affective states experienced by each individual should be measured separately from whatever nonhedonic attributes are welfare-relevant; (201) and these hedonic utilities should be aggregated or averaged across persons and moments, as in the Texas and U-index studies.
But there are two difficulties--one obvious, the second more subtle--with this position. First, the "objective happiness" framework, itself, offers no guidance in balancing the hedonic and nonhedonic aspects of well-being. Imagine that we predict that a policy will increase the average "objective happiness" of some population, but reduce some measure of their average nonhedonic well-being. Is the policy worthwhile, all things considered?
The more subtle point is that the very assignment of hedonic utilities to individual moments--at least as Kahneman conceptualizes it--assumes the separability of hedonic and nonhedonic attributes. Remember his formal model. The observer considers temporally extended episodes ("profiles"), characterized in terms of the affects experienced by the subject during each moment, and ranks these--yielding a cardinal measure of momentary hedonic experience, what I referred to above as the v(.) value of that experience. But--now that Kahneman has, properly, conceded the welfare-relevance of nonhedonic attributes--the insistence that any given observer has such a ranking becomes problematic.
To be sure, the observer will have a conditional ranking of hedonic profiles. Assume that N is a possible profile of nonhedonic, welfare-relevant, attributes that a subject might possess. N includes both nonhedonic mental attributes (an individual's memory, perceptions, sense of satisfaction, and so forth) and whatever nonmental attributes are seen to be welfare-relevant (for example, health). H is one possible hedonic profile, that is, a sequence of hedonic attributes, [H.sup.*] a second possible hedonic profile, [H.sup.**] a third, and so forth. Then the observer will be able to rank NH versus N[H.sup.*] versus N[H.sup.**]. We can call this the ranking of hedonic episodes conditional on nonhedonic profile N. Similarly, if [N.sup.+] is a different specification of nonhedonic attributes, the observer will be able to rank [N.sup.+]H versus [N.sup.+][H.sup.*] versus [N.sup.+][H.sup.**]. We can call this second ranking the ranking of hedonic episodes conditional on nonhedonic profile [N.sup.+].
But it is a further question whether the observer's ranking of hedonic episodes is invariant to nonhedonic attributes, namely, that if the observer has a particular ranking of hedonic episodes conditional on N, then she has the very same ranking conditional on [N.sup.+], on [N.sup.++], and every other nonhedonic profile. Why believe that invariance holds true? Might it not be possible that the observer prefers NH to N[H.sup.*] to N[H.sup.**], but [N.sup.+][H.sup.**] to [N.sup.+]H to [N.sup.+][H.sup.*]? Invariance precludes interaction effects between affective content, on the one hand, and anything else occurring in someone's life on the other. The preclusion of such interaction is highly problematic. For example, who would want to experience a mental life in which the pairing of affects and memories is random, as opposed to being appropriate to what is remembered? Similarly, insofar as health is an intrinsic determinant of well-being, why think that the ranking of temporally extended sequences of pains and pleasures is invariant to the sequence of health states?
It is tempting to respond that the possible interaction between hedonic and nonhedonic attributes is a nuance which policy tools, necessarily crude and approximate, may need to ignore. But this response misunderstands the objection. Once we have assigned hedonic utilities to moments, we may (for practical reasons) need to ignore their correlation with nonhedonic attributes, in evaluating policies. However, we still need some rationale for this assignment. We still need some theory of what is being measured. Kahneman's theory is that hedonic utilities represent the preferences/judgments of an observer, ranking temporally extended hedonic episodes. But--once interaction effects are allowed into the picture--the observer may have no such ranking. If you ask her, "Do you prefer H to H* or [H.sup.*] to H?" she may respond: "The question is meaningless. I have various conditional preferences/judgments regarding the episodes, but no unconditional preference/judgment. To ask me for such a preference/judgment is like asking whether I prefer a long-sleeved or a short-sleeved shirt, to which I'll answer, "The first in cold weather, the second in warm."
C. A Better Approach?
In this Section, I very briefly sketch a different methodology than the "objective happiness" framework for incorporating information about experiential quality into policy analysis. I lack space to elaborate this methodology at length, but at least I can suggest how the limitations in Kahneman's framework might be remedied.
A preference-based view of well-being allows for an individual's happiness, or some other aspect of her experiential quality, to be one of the fundamental arguments for her preferences. Indeed, it would be absurd for the preferentialist to insist that only nonmental properties can be intrinsic determinants of well-being! However, the simplified utility functions adopted in classical economic theory and in much empirical work have ignored experiential arguments. Instead, preference utility is often modeled as a function of one or more nonmental attributes, quintessentially income, bundles of commodities, health, leisure, or the physical state of the environment.
This sort of simplified model of preferences, while more tractable, has a downside when coupled with policy tools such as cost-benefit analysis. Policy impacts on happiness, and thus on well-being, cannot be directly measured. Instead, it must be assumed (plausibly) that policies causally affect happiness via changes to individuals' nonmental attributes and (less plausibly) that individuals can accurately predict how changes in nonmental attributes affect their happiness and thus that their intrinsic preferences for happiness are fully captured by the instrumental component of their preferences for nonmental attributes. (202)
A richer model of preferences explicitly includes experiential properties. A given attribute bundle A is characterized as A = (M, N), where M describes one or more types of mental attributes, and N nonmental attributes. For short, call this a "hvbrid bundle." Individual i's preferences over hybrid attribute bundles are inferred from stated-preference surveys or, perhaps, behavioral evidence. Individual i's ordinal preference-utility function for hybrid bundles captures his ranking of these bundles. That ordinal preference-utility function takes the form [u.sub.i](M, N). If income is included among the nonmental attributes, this ordinal preference-utility function is sufficient to determine individual i's willingness to pay/accept for changes in the mental attributes.
Information about individual WTP/WTA for experiential changes allows for the direct incorporation of such changes in policy modeling. For example, imagine that a costly proposed regulation, lowering the ambient level of some feared environmental toxin, will have the benefit of reducing deaths and injuries and reducing anxiety about the toxin. Information about individual WTP/WTA for fear and anxiety allow us to undertake a cost-benefit test of the regulation that explicitly identifies fear-reduction as a separate "good" additional to physical harm-avoidance.
An individual's cardinal utility function for hybrid bundles would be determined by her ranking of bundles plus further facts about her preferences (for example, her ranking of bundle lotteries or her time-tradeoff preferences). Such cardinal utility functions are not required by cost-benefit analysis but might be required by other policy tools (for example, a GDP-like measure that aggregates/averages preference-utility across persons, with preference-utility in turn dependent on both experiential and nonexperiential attributes).
The approach now under discussion might seem utopian (or dystopian, for skeptics about policy analysis) but in fact has some precedent in existing empirical work. A small literature does try to estimate individual WTP/WTA for fear and anxiety (with application both to environmental regulation and to policing strategies that mitigate the fear of crime). (203) The QALY literature looks at individual preferences over health states. The QALY number assigned to a health state (between 0 and 1) captures such preferences, for example, time-tradeoff preferences as between living a period of time in perfect health and living longer in a poorer health state. The health states thus valued sometimes encompass mental, not just physical health. It is not unusual to find QALY values for conditions such as depression, pain, or chronic anxiety. (204)
Estimating individuals' preference-utility for hybrid bundles via standard preference-elicitation techniques (stated-preference surveys or inference from behavioral evidence) is to be sharply distinguished from the methodology criticized in Part II: using SWB surveys as evidence of preference utility, that is, taking an individual's stated life satisfaction when in possession of bundle A as his preference utility for A. What is being recommended is not that an SWB number be employed as a proxy for preference utility, but rather that an individual's experiences be included among the determinants of preference utility.
Indeed, this approach does not essentially rely upon the assignment of numbers to mental states. In a stated-preference survey, the experiential component of a hybrid bundle could be described via a number ("imagine experiencing pain that you might rate at 4 on a scale of 1 to 5"), but it could also be described qualitatively--and indeed a qualitative description might be more successful in accurately communicating to respondents what the state feels like.
The approach also has key advantages over Kahneman's "objective happiness" framework. First, although that framework seeks only to quantify the well-being impact of affective states, in a hedonic utility number, the "hybrid bundle" approach is more generic. If A = (M, N), with M mental attributes, M itself might be unidimensional or multidimensional and, in either event, might include affective states, memory, cognition, sense of satisfaction or purpose, and so forth. Second, the "objective happiness" framework may be undermined by the bivalence of pain and pleasure--the fact that positive and negative affect may really be two, separate, dimensions of experience--but the hybrid-bundle approach is perfectly compatible with affective bivalence. Third, the hybrid-bundle approach is not committed to any kind of separability with respect to an individual's ranking of bundles: temporal separability, the separability of hedonic value from other aspects of experiential quality, or the separability of hedonic and nonmental attributes. Fourth, the approach provides a basis for making all-things-considered policy judgments, integrating information about policy impacts on experiential quality with other sorts of impacts. If we know someone's preference utility for different hybrid bundles, then we know how she makes tradeoffs between the experiential and nonexperiential components of these bundles. The governmental choice between multiple policies, characterized by different kinds of experiential and nonexperiential effects, can then be a function of the totality of individual tradeoffs--as operationalized via cost-benefit analysis (the sure of individual WTP/WTA amounts), or perhaps in some other way.
Fifth, at least insofar as this approach seeks only to infer individuals' ordinal preference-utility functions--which is all that is required for purposes of CBA and the determination of WTP/WTA amounts--the approach does not presuppose that individuals have the same preferences. (205) The ranking of hybrid bundles can vary from individual to individual. By contrast, as we have seen, Kahneman's "objective happiness" framework makes the implausible presupposition that "observers" will have the very same ranking of temporally extended hedonic episodes ("profiles"). (206)
The elicitation of preference utility for hybrid bundles presupposes that individuals have well-behaved preferences over such bundles--if not initially, then at least after debiasing and information-provision. Preference skeptics will deny this, pointing to violations of rationality conditions-(207)--and this challenge is a serious challenge to the "hybrid bundle" approach. But, as I have just reminded the reader, Kahneman's theory for assigning hedonic utilities to moments also depends upon an observer having a preference/judgment ranking of hedonic profiles. This observer is suppressed in Kahneman's empirical work--with consequent difficulties that I elaborated upon in discussing the Texas and U-index studies. It is ironic that Kahneman, in some scholarship, comes across as an ardent preference skeptic. (208) If actual individuals (even after debiasing) are incapable of satisfying rationality conditions requisite for holding a preference, then who is the observer whose well-behaved ordering of profiles is supposed to satisfy such conditions? In short, Kahneman's "objective happiness" framework is no more immune from preference skepticism than the hybrid-bundle approach.
Enthusiasm about the policy role of SWB surveys is premature. Why think that the number which someone assigns to her momentary or overall happiness, life satisfaction, positive or negative affect, or some other aspect of her experiential state offers real help in evaluating governmental policies? Two different answers to this question need to be teased apart. One says that a higher self-rated degree of life satisfaction shows that the respondent's preferences are more fully realized. In short, SWB surveys evidence preference utility. But the evidence would seem to be pretty poor. Preference and scale heterogeneity hamper the use of self-rated life satisfaction to make inferences about preference utility. Even if all respondents share the same underlying preferences and utility function, someone's answer to an SWB survey may well be skewed by evaluation error or miscommunication. This number may well be an inaccurate and, indeed, statistically biased indicator of the degree to which her life-circumstances realize her preferences.
Stated-preference surveys dominate SWB surveys as evidence of preference-realization. Anomalies using the stated-preference format suggest the importance of debiasing preferences--rendering them rational and well-informed. Perhaps debiasing is fruitless. But that would show that government policy choice must (somehow) find a normative foundation other than individuals' preferences, and not that preferences should be inferred via the SWB technique. Skepticism about the rationality of preferences hardly advances the PR (preference-realization) defense of SWB surveys.
The second answer to the "why" question takes a different tack, suggesting that a happiness, affect, or life-satisfaction rating is a measure of experiential quality. Thus goes the EQ defense of SWB surveys. Kahneman's "objective happiness" framework--using SWB surveys focused on momentary hedonic quality--is an important first step in developing a policy-relevant measure of experiential quality. Kahneman does not argue that well-being and good experiences are equivalent--but rather, much more plausibly, that good experiences are one important aspect of well-being.
However, a close examination of the "objective happiness" framework suggests significant limitations. The framework purports to cardinalize momentary hedonic utilities by appealing to an "observer's" ranking of temporally extended hedonic episodes, but presupposes--without justification--that observers have the same ranking, and that these rankings are separable from nonhedonic attributes. In empirical implementation, Kahneman has suppressed the observer and, most recently, abandoned any attempt at cardinalization--via a "U-index" that merely reports the fraction of time that individuals spend in an affectively unpleasant state. This is a crude measure of hedonic quality (let alone the nonhedonic aspects of experiential life, such as memory or a sense of meaning), because it does not tell us about the intensity of individuals' affective states.
It remains unclear whether SWB surveys--asking for a numerical rating of experiences--should be the central tool for incorporating information about experiential quality into policy analysis. At least in principle, a different approach, more closely continuous with traditional cost-benefit analysis, is available: namely, to use revealed or stated-preference evidence to infer individuals' preferences over "hybrid bundles," comprising both experiential and nonexperiential attributes. SWB surveys are at most an ancillary component of this approach. Its central focus is inferring preference utility, with experiential attributes merely one entry in the utility function.
Much more work remains to elaborate both this approach and frameworks (such as "objective happiness") that revolve around SWB surveys. In undertaking this effort, scholars should exercise caution, taking care not to muddy their concepts--taking care to understand that well-being need not reduce to good experiences, that individuals can have intrinsic preferences for aspects of their lives other than their mental states, and that someone's perceived degree of happiness or life satisfaction can diverge from her true preference utility.
(1.) For overviews, see generally LUIGINO BRUNI & PIER LUIGI PORTA, ECONOMICS AND HAPPINESS: FRAMING THE ANALYSIS (2005); BRUNO S. FREY, HAPPINESS: A REVOLUTION IN ECONOMICS (2008): BRUNO S. FREY & ALOIS STUTZER, HAPPINESS AND ECONOMICS: HOW THE ECONOMY AND INSTITUTIONS AFFECT WELL-BEING (2002): HANDBOOK ON THE ECONOMICS OF HAPPINESS (Luigino Bruni & Pier Luigi Porta eds., 2007): THE SCIENCE OF WELL-BEING (Felicia A. Huppert, Nick Baylis & Barry Keverne eds., 2005): WELL-BEING: THE FOUNDATIONS OF HEDONIC PSYCHOLOGY (Daniel Kahneman, Ed Diener & Norbert Schwarz eds., 1999); Ed Diener, Eunkook M. Sub, Richard E. Lucas & Heidi L. Smith, Subjective Well-Being: Three Decades of Progress, 125 PSYCHOL. BULL. 276 (1999); Paul Dolan, Tessa Peasgood & Malhew White, Do We Really Know What Makes Us Happy? A Review of the Economic Literature on the Factotw Associated with Subjective Well-Being, 29 J. ECON. PSYCHOL. 94 (2008).
(2.) The table is taken from Richard R. Layard, G. Mayraz & S. Nickell, The Marginal Utility of Income, 92 J. PUB. ECON. 1846, 1848 tbl.1 (2008). Many thanks to Marc Fleurbaey for the reference.
(3.) See infra note 176: see also infra Part III.B.
(4.) Alan B. Krueger, Daniel Kahneman, David Schkade, Norbert Schwarz & Arthur A. Stone, National Time Accounting: The Currency of Life, in MEASURING THE SUBJECTIVE WELL-BEING OF NATIONS: NATIONAL ACCOUNTS OF TIME USE AND WELL-BEING 9, 30-31 (Alan B. Krueger ed., 2009).
(7.) Daniel Kahneman, Alan B. Krueger, David A. Schkade, Norbert Schwarz & Arthur A. Stone, A Survey Method for Characterizing Daily Life Experience: The Day Reconstruction Method, 306 SCIENCE 1776, 1776-77 (2004).
(8.) Id. at 1777 tbl.1.
(9.) See Ed Diener, Weiting Ng, James Hatter & Raksha Arora, Wealth and Happiness Across the World: Material Prosperity Predicts Life Evaluation, Whereas Psychosocial Prosperity Predicts Positive Feeling, 99 J. PERSONALITY & SOC. PSYCHOL. 52, 58-60 (2010): Daniel Kahneman & Angus Deaton, High Income Improves Evaluation of Life but Not Emotional Well-Being, 107 PROC. NAT'L ACAD. SCI. 16,489, 16,489 (2010).
(10.) Scholarship endorsing the use of SWB surveys for law or policy purposes includes DEREK BOK. THE POLITICS OF HAPPINESS: WHAT GOVERNMENT CAN LEARN FROM THE NEW RESEARCH ON WELL-BEING (2010): ED DIENER, RICHARD E. LUCAS, ULRICH SCHIMMACK & JOHN F. HELLIWELL, WELL-BEING FOR PUBLIC POLICY (2009): FREY, supra note 1: FREY & STUTZER, supra note 1: CAROL GRAHAM, THE PURSUIT OF HAPPINESS: AN ECONOMY OF WELL-BEING (20111: RICHARD LAYARD, HAPPINESS: LESSONS FROM A NEW SCIENCE (rev. & updated ed. 2011): MEASURING THE SUBJECTIVE WELL-BEING OF NATIONS, supra note 4: Samuel R. Bagenstos & Margo Schlanger. Hedonic Damages, Hedonic Adaptation. and Disability, 60 VAND. L. REV. 745 (2007): John Bronsteen, Christopher Buccafusco & Jonathan S. Masur, Welfare as Happiness, 98 GEO. L.J. 1583 (2010) [hereinafter Bronsteen, Buccafusco and Masur, Welfare as Happiness]: John Bronsteen. Christopher Buccafusco & Jonathan S. Masur, Well-Being Analysis vs. Cost-Benefit Analysis, 62 DUKE L.J. 1603 (2013) [hereinafter Bronsteen, Buccafusco and Masur, Well-Being Analysis]: Ed Diener & Martin E.P. Seligman, Beyond Money: Toward an Economy of Well-Being, 5 PSYCHOL. SCI. PUB. INT. 1 (2004): Ed Diener, Subjective Well-Being: The Science of Happiness and a Proposal for a National Index, 55 AM. PSYCHOLOGIST 34 (2000): Paul Dolan & Mathew P. White, How Can Measures of Subjective Well-Being Be Used To Inform Public Policy?, 2 PERSP. PSYCHOL. SCI. 71 (2007): Peter Henry Huang, Happiness Studies and Legal Policy. 6 ANN. REV. L. & SOC. SCI. 405 (2010): and Daniel Kahneman, Alan B. Krueger, David Schkade. Norbert Schwarz & Arthur Stone. Toward National Well-Being Accounts, 94 AM. ECON. REV. (PAPERS & PROC.) 429 (2004). For further such scholarship, see infra notes 11.13.
(11.) For overviews of this approach, see DANIEL FUJIWARA & ROSS CAMPBELL, DEP'T FOR WORK & PENSIONS (UK). VALUATION TECHNIQUES FOR SOCIAL COST-BENEFIT ANALYSIS: STATED PREFERENCE, REVEALED PREFERENCE AND SUBJECTIVE WELL-BEING APPROACHES (2011); Andrew E. Clark & Andrew J. Oswald, A Simple Statistical Method for Measuring How Life Events Affect Happiness 31 INT'L J. EPIDEMIOLOGY 1139 (2002); Bruno S. Frey, Simon Luechinger & Alois Stutzer, The Life Satisfaction Approach to Environmental Valuation, 2 ANN. REV. RESOURCE ECON. 139 (2010): Heinz Welsch & Jan Kuhling, Using Happiness Data for Environmental Valuation: Issues and Applications, 23 J. ECON. SURVS. 385 (2009): and Paul Dolan, Daniel Fujiwara & Robert Metcalfe, A Step Towards Valuing Utility the Marginal and Cardinal Way (Ctr. for Econ. Performance, Discussion Paper No. 1062, 2011). Recent examples include Christopher L. Ambrey & Christopher M. Fleming, Valuing Scenic Amenity Using Life Satisfaction Data, 72 ECOLOGICAL ECON. 106 (2011): Arik Levinson, Valuing Public Goods Using Happiness Data: The Case of Air Quality, 96 J. PUB. ECON. 869 (2012); Tobias Menz & Heinz Welsch, Life-Cycle and Cohort Effects in the Valuation of Air Quality: Evidence from Subjective Well-Being Data, 88 LAND ECON. 300 (2012): and Nattavudh Powdthavee & Bernard van den Berg. Putting Different Price Tags on the Same Health Condition: Re-Evaluating the Well-Being Valuation Approach, 30 J. HEALTH ECON. 1032 (2011).
(12.) An increase in income by [c.sub.U]/[c.sub.y] dollars produces [c.sub.v]([c.sub.G]/[c.sub.v]) = [c.sub.G] increase in happiness, and thus is the equivalent, in happiness terms, of an increase in exposure to the good by 1 unit.
Bronsteen, Buccafusco and Masur have recently proposed that policies be evaluated via "well-being analysis" (WBA). whereby policy impacts, including income changes, would be converted into happiness units. See Bronsteen, Buccafusco and Masur. Welfare as Happiness, supra note 10, at 1627-40: Bronsteen, Buccafusco and Masur, Well-Being Analysis, supra note 10, at Part II. Although this proposal has some similarities to the use of SWB data to calculate monetary equivalents for nonmarket goods, it differs from the latter in important respects. First, Bronsteen, Buccafusco. and Masur adopt the "experience-quality" (EQ) defense of SWB surveys, indeed the strong EQ defense--to use a distinction I will develop later in the Article, see infra Part I.C. They see WBA as a way to take account of the information that SWB surveys provide about individuals' experience utility. By contrast, scholars using the monetary-equivalent approach do not universally adopt the EQ defense, at least not explicitly so. One possible reading of some of this scholarship is that it adopts the "preference-realization" (PR) defense: that the monetary-equivalent approach is meant to be sensitive to the information SWB surveys provide concerning preference utility. See infra Part 1.C. Indeed, Part II discusses the monetary-equivalent approach, at length, as a test case for the PR defense of SWB surveys. Second, WBA requires cardinal happiness data; the monetary-equivalent approach requires only ordinal data. See Bronsteen, Buccafusco and Masur, Well-Being Analysis, supra note 10. at Part II.B.4: infra Part II.A.3. Finally, WBA neutralizes wealth effects, whereas (in principle) the monetary-equivalent approach does not. To see this point, imagine that some nonmarket good has the very same effect on SWB for both rich and poor, and that money has a diminishing incremental effect. Then WBA will be neutral between a policy that provides the good for free to some number of rich individuals and one that does so for the same number of poor individuals, taxing some third population, whereas the monetary-equivalent approach will (in principle) favor the first policy.
(13.) See generally MEASURING THE SUBJECTIVE WELL-BEING OF NATIONS, supra note 4: Diener & Seligman, supra note 10: Kahneman et al., supra note 10: Ruut Veenhoven, Happy Life-Expectancy: A Comprehensive Measure of Quality-of-Life in Nations, 39 SOC. INDICATORS RES. 1 (1996): Ruut Veenhoven & Wim Kalmijn, Inequality-Adjusted Happiness in Nations: Egalitarianism and Utilitarianism Married in a New Index of Societal Performance. 6 J. HAPPINESS STUD. 421 (2005).
(14.) For discussion of cardinality and ordinality, see infra Part II.A.3.
(15.) See Winton Bates, Gross National Happiness, 23 ASIAN-PAC. ECON. LITERATURE 1, 12 (2009).
(16.) See JOSEPH E. STIGLITZ, AMARTYA SEN & JEAN-PAUL FITOUSSI, REPORT BY THE COMMISSION ON THE MEASUREMENT OF ECONOMIC PERFORMANCE AND SOCIAL PROGRESS 16 (2009), available at http://www.stiglitz-sen-fitoussi.fr/documents/rapport_anglais.pdf.
(17.) PAUL DOLAN, RICHARD LAYARD & ROBERT METCALFE, OFFICE FOR NAT'L STAT., MEASURING SUBJECTIVE WELL-BEING FOR PUBLIC POLICY 3 (2011).
(18.) See PAUL DOLAN & ROBERT METCALFE, OFFICE FOR NAT'L STAT., COMPARING MEASURES OF SUBJECTIVE WELL-BEING AND VIEWS ABOUT THE ROLE THEY SHOULD PLAY IN POLICY (2011); DOLAN ET AL., supra note 17 at 3: OFFICE FOR NAT'L STAT., FIRST ANNUAL ONS EXPERIMENTAL SUBJECTIVE WELL-BEING RESULTS (2012), available at http://www.ons.gov.uk/ons/dcp171766_272294.pdf; SAM WALDRON, OFFICE FOR NAT'L STAT., MEASURING SUBJECTIVE WELLBEING IN THE UK (2010).
(19.) For other, important, critical work on SWB surveys, see the chapters by George Loewenstein, William Nordhaus, and Erik Hurst in MEASLIRING THE SUBJECTIVE WELLBEING OF NATIONS, supra note 4: MARC FLEURBAEY & DIDIER BLANCHET, BEYOND GDP: MEASURING WELFARE AND ASSESSING SUSTAINABILITY (forthcoming 2013) (manuscript at ch. 5) (on file with the Duke Law Journal); Daniel M. Hausman, Hedonism and Welfare Economics, 26 ECON. & PHIL. 321 (2010): Mark Kelman. Hedonic Psychology and the Ambiguities of "Welfare," 33 PHIL. & PUB. AFF. 391 (2005): George Loewenstein & Peter A. Ubel, Hedonic Adaptation and the Role of Decision and Experience Utility in Public Policy, 92 J. PUB. ECON. 1795 (2008): Rick Swedloff & Peter H. Huang, Tort Damages and the New Science of Happiness, 85 IND. L.J. 553 (2010); and Marc Fleurbaey, Erik Schockkaert & Koen Decancq. What Good Is Happiness? (Ctr. for Operations Res. & Econometrics. Discussion Paper No. 17, 2009).
(20.) LAYARD, supra note 10. For a description and critique of Layard's views, see infra Part III.A.
(21.) See infra notes 177-178 and accompanying text.
(22.) I am grateful to Carol Graham for writing a commentary on this Article. Carol Graham, An Economist's Perspective on Well-Being Analysis and Cost-Benefit Analysis, 62 DUKE L.J. 1691 (2013). Graham distinguishes between two types of SWB surveys: hedonic well-being (HWB) surveys (which seek to measure individuals' affects) and evaluative well-being (EWB) surveys (which seek to measure how individuals evaluate their lives). Id. at 1692-93. To be clear, my intention in this Article is not to defend HWB surveys as contrasted with EWB surveys. Part II focuses on life-satisfaction surveys (a kind of EWB survey) as a comparatively more plausible indicator of preference utility than happiness surveys--but my aim in undertaking the analysis of Part II is to demonstrate that the PR defense with respect to SWB surveys of any kind is problematic. (Because EWB surveys fail in that role, as Part II shows, then a fortiori HWB surveys do.) Part III, in discussing the weak EQ defense, focuses on Kahneman's "objective happiness" framework. It does so because Kahneman's work is the most fully developed version of the weak EQ defense. As it happens, "objective happiness" relies on HWB surveys (asking about momentary affects) rather than EWB surveys. I do not mean to suggest that the weak EQ approach is best fleshed out by focusing wholly on affects, and indeed I criticize Kahneman for doing just that. See infra Part III.B.5.
(23.) Richard A. Easterlin, Income and Happiness: Towards a Unified Theory, 111 ECON. J. 465. 465 (2001).
(24.) See, e.g., DIENER ET AL., supra note 10, at 9-12 (suggesting that "economic theories of well-being equate well-being with utility," that "economists define utility as the satisfaction that a person experiences from the consumption of goods," and that the authors" proposed subjective definition of well-being, in terms of individuals' favorable self-evaluations, is "essentially identical to economists' concept of utility"); FREY, supra note 1, at 3 ("In general.... as in the literature, the terms 'happiness', 'well-being', and 'life satisfaction' are used interchangeably."); David G. Blanchflower & Andrew J. Oswald, Well-Being over Time in Britain and the USA, 88 J. PUB. ECON. 1359, 1360-62 (2004) (using well-being, happiness, and utility as equivalent terms); Diener & Seligman, supra note 10, at 1 (defining "[w]ell-being" as "peoples' positive evaluations of their lives"); Andrew J. Oswald, Happiness and Economic Performance, 107 ECON. J. 1815, 1815 (1997) (leaping from the premise that economic performance has no intrinsic normative significance, to the conclusion that "[e]conomic things matter only in so far as they make people happier"): Carol D. Ryff & Burton H. Singer, Know Thyself and Become What You Are: A Eudaimonic Approach to Psychological Well-Being, 9 J. HAPPINESS STUD. 13, 14-15 (2008) (viewing Aristotelian "eudaimonia" as a kind of psychological state, namely "psychological well-being"); Bernard M.S. van Praag, Perspectives from the Happiness Literature and the Role of New Instruments for Policy Analysis, 53 CESIFO ECON. STUD. 42, 42 & n.1 (2007) (assuming that "economic behaviour ... is motivated by maximization of utility, satisfaction, well-being or happiness" and noting that "[w]e will make no difference between these notions"). For further examples of SWB scholars using the concepts of well-being, subjective well-being, and/or happiness as equivalent, see Erik Angner. Are Subjective Measures of Well-Being 'Direct'?, 89 AUSTRALASIAN J. PHIL. 115, 119-20 (2011). 25. See supra note 23 and accompanying text.
(26.) See. e.g., DANIEL M. HAUSMAN, PREFERENCE. VALUE, CHOICE, AND WELFARE 77-81 (2012): DANIEL M. HAYBRON, THE PURSUIT OF UNHAPPINESS: THE ELUSIVE PSYCHOLOGY OF WELL-BEING 29-42 (2008): L.W. SUMNER, WELFARE, HAPPINESS, AND ETHICS 81-98 (1996): Erik Angner, Subjective Well-Being, 39 J. SOCIO-ECON. 361. 362-63 (2010).
(27.) For an overview of the philosophical literature on well-being, see MATTHEW D. ADLER, WELL-BEING AND FAIR DISTRIBUTION: BEYOND COST-BENEFIT ANALYSIS 158-85 (2012).
(28.) An "outcome" is a possible state of affairs: a description of some aspect of the world, which might possibly occur or have occurred in the past, present, and/or future. A maximally specified outcome is what philosophers term a "possible world," namely an outcome F such that for every other F* either F entails F* or F entails not-F*. In effect, a maximally specified outcome is a complete history of the universe, leaving no possible occurrence undetermined.
Of course, maximally specified outcomes are not items that humans can hold in consciousness. In particular, human policy analysts cannot use maximally specified outcomes to think about policy impacts. Rather, maximally specified outcomes are theoretical constructs that serve various theoretical purposes. For example, they help to categorize different accounts of well-being. An experientialist account says that the quality of an individual's experiences is the only intrinsic determinant of her well-being; changes in nonexperiential items can cause changes in well-being, but do not change it directly. This feature of experientialism is made precise by the definition in the text only if x and y are maximally specified. Even an experientialist account might allow the welfare ranking of partially described outcomes to be a partial function of nonmental attributes (given their causal role).
For example, assume that outcomes are partially described to specify an individual's income and how much pain or pleasure she feels, but not the nonhedonic aspects of her mental life (for example, her sense of life satisfaction or the quality of her memories). Assume that individual i's pains and pleasures are the same in x and y, but that she has more income in x. Then someone who reduces well-being to pains/pleasures and a sense of life satisfaction--a kind of sophisticated experientialist account--might say that individual i is better off in x than y despite the fact that all of i's specified mental attributes are identical in x and y. Having greater income might cause a change in unspecified, well-being-relevant mental attributes, namely, a sense of life satisfaction.
By contrast, if x and y are fully specified, the causal upshots of any nonmental attribute that an individual might possess are already "built into" the description of the outcomes, and so the only reason for a difference in some nonmental attribute to yield a difference in well-being is because of the attribute's intrinsic welfare significance.
(29.) See infra notes 197-200 and accompanying text.
(30.) See HAYBRON, supra note 26, at 30.
(31.) MARTHA C. NUSSBAUM, WOMEN AND HUMAN DEVELOPMENT: THE CAPABILITIES APPROACH 78-80 (2000); see also Martha C. Nussbaum, Who Is the Happy Warrior? Philosophy Poses Questions to Psychology, 37 J. LEGAL STUD. S81 (2008) (critically evaluating proposals to orient policy around SWB).
(32.) I have elsewhere endorsed a preference-based view of well-being. See MATTHEW D. ADLER & ERIC. A. POSNER. NEW FOUNDATIONS OF COST-BENEFIT ANALYSIS 25-61 (2006): ADLER, supra note 27, at 155-231. And the discussion in Part III.C--suggesting a methodology for incorporating information about experiential quality into policy--does presuppose that a preference-based view is more attractive than competing views. However, the remainder of the Article does not commit itself to this view. The nature of well-being is contested, and this Article hopes to address a wider readership than persuaded preferentialists. The aim of this Part is clarificatory: to bring to light distinctions between the various well-being accounts that are often obscured in the SWB literature. Part II argues that a potential defense of SWB surveys, in light of a preference-based view (the PR defense), is unpersuasive. This is an argument about the relation between the preference-based view and SWB surveys--not a defense of that view. Part III.A criticizes experientialism about well-being--but note that "objective good" as well as preference-based views reject experientialism.
(33.) See HAUSMAN, supra note 26, at 13-14: ANDREU MAS-COLLEL, MICHAEL D. WHINSTON & JERRY R. GREEN, MICROECONOMIC THEORY 6-9, 46-50 (7th prtg. 2009).
(34.) Amartya Sen famously argues for a closely related proposition, namely that social assessment should focus on the achieved "functionings" that are caused by income together with individuals' physical and social attributes, or opportunities to function, and not income itself. E.g., AMARTYA SEN, INEQUALITY REEXAMINED 28-30 (1992): see also Amartya Sen, Capability and Well-Being, in THE QUALITY OF LIFE 30, 41 (Martha Nussbaum & Amartya Sen eds., 1993) ("Since income is not desired for its own sake, any income-based notion of poverty must refer--directly or indirectly--to those basic ends which are produced by income as means.").
(35.) See ADLER, supra note 27, at 170-72.
(36.) See MAS-COLLEL ET AL.. supra note 33, at 41-50.
(37.) See Dan Moller, Wealth, Disability, and Happiness, 39 PHIL. & PUB. AFF. 177, 186-89 (2011).
(38.) See Beatrice Rey & Jean-Charles Rochet, Health and Wealth: How Do They Affect Individual Preferences?, 29 GENEVA PAPERS ON RISK & INS. THEORY 43, 44-46 (2004).
(39.) For a review of evidence that individuals have an intrinsic preference for health, see infra notes 76-84 and accompanying text: see also Carol Graham, Happiness and Health: Lessons--and Questions--for Public Policy, 27 HEALTH AFF. 72, 73-74 (2008) (reviewing evidence showing imperfect correlation between health and happiness); Carol Graham, Lucas Higuera & Eduardo Lora, Which Health Conditions Cause the Most Unhappiness?, 20 HEALTH ECON. 1431, 1432-33 (2011) (same): Peter A. Ubel & George Loewenstein, Pain and Suffering Awards: They Shouldn't Be (Just) About Pain and Suffering, 37 J. LEGAL STUD. S195 (2008) (arguing for the intelligibility of an intrinsic preference not to be physically injured).
(40.) See Betsey Stevenson & Justin Wolfers, The Paradox of Declining Female Happiness, 1 AM. ECON. J.: ECON. POL'Y 190, 191 (2009) (finding declining female happiness during a period of increased labor market opportunities and other options for women).
(41.) See Carol Graham & Stefano Pettinato. Frustrated Achievers: Winners, Losers and Subjective Well-Being in New Market Economies, J. DEV. STUD., Apr. 2002. at 100, 117-20 (providing evidence of "frustrated achievers": upward income mobility shifts income expectations and reduces happiness). See generally BARRY SCHWARTZ, THE PARADOX OF CHOICE: WHY MORE IS LESS (2004).
(42.) THOMAS HURKA. PERFECTIONISM 38-41 (1993): see also JOHN FINNIS, NATURAL LAW AND NATURAL RIGHTS 59-80 (1980) (arguing for the intrinsic value of knowledge): George Loewenstein, That Which Makes Life Worthwhile. in MEASURING THE SUBJECTIVE WELL-BEING OF NATIONS, supra note 4. at 87, 96-97 (same).
(43.) Richard Kraut, Aristotle's Ethics, STANFORD ENCYCLOPEDIA OF PHILOSOPHY (Mar. 29, 2010), http://plato.stanford.edu/entries/aristotle-ethics.
(44.) Moreover, knowledge may not produce happiness or life satisfaction. For evidence that SWB need not increase with education, see Dolan et al., supra note 1, at 99-100: and Joop Hartog & Hessel Oosterbeek. Health, Wealth and Happiness: Why Pursue a Higher Education?, 17 ECON. EDUC. REV. 245,251-54 (1998).
(45.) See Dolan et al., supra note 1, at 107 (finding mixed evidence of SWB impact of having children): Thomas Hansen, Parenthood and Happiness: A Review of Folk Theories Versus Empirical Evidence, 108 SOC. INDICATORS RES. 29. 44 (2012) (finding that resident children tend to reduce SWB).
(46.) ROBERT NOZICK, ANARCHY, STATE. AND UTOPIA 42-43 (1974).
(47.) Cf. SUMNER, supra note 26, at 139 (analyzing well-being as "authentic happiness," where authenticity requires inter alia that the individual be factually well-informed): Shelly Kagan, Well-Being as Enjoying the Good, 23 PHIL. PERSP. 253, 255 (2009) (suggesting that well-being may consist of taking pleasure in the attainment of objective goods).
(48.) See infra Part III.A.
(49.) A different objection is that any person j whose preferences have non-mental-state fundamental arguments can be mimicked by a "doppelganger" [j.sup.*] who cares only about her mental states. In particular, whenever j intrinsically prefers nonmental fact F, j* has an intrinsic preference concerning her beliefs; namely, [j.sup.*] prefers that she believe F. Thus, policy modelers can ignore the possibility of preferences with non-mental-state fundamental arguments--instead "translating" apparent intrinsic preferences for nonexperiential items into intrinsic preferences for the corresponding beliefs.
But one implication of Nozick's "experience machine" example is that j and [j.sup.*] will not necessarily make identical choices. If the machine will cause the false belief F, then [j.sup.*] might enter the machine while j refuses. Moreover, third parties might act very differently toward j and [j.sup.*]. If j prefers to have a faithful spouse, and a friend observes that j's spouse is unfaithful, then the friend might tell j but not do the same in [j.sup.*]'s case. If j wants to be in good physical health, the government might fund certain health interventions that it would not fund in the case of [j.sup.*]. who simply wants to believe she is in good health.
(50.) See supra note 24 and accompanying text.
(51.) See, e.g., FREY, supra note 1, at 5 ("[H]appiness is undoubtedly an overriding goal in most people's lives."); LAYARD, supra note 10, at 124 (arguing that happiness-maximization "values what people want for themselves, for their children and for their fellow citizens," namely "their happiness"): Daniel J. Benjamin, Ori Heffetz, Miles S. Kimball & Alex Rees-Jones. What Do You Think Would Make You Happier? What Do You Think You Would Choose?, 102 AM. ECON. REV. 2083, 2107 (2012) (noting that the implicit view in much of the economics of happiness literature is that SWB is the sole argument for idealized preferences).
(52.) This can occur insofar as individuals incorrectly predict the causal impact of nonmental attributes. Imagine that individual i cares only about his pains and pleasures. If x and y are (incompletely specified) outcomes that describe individuals' income, health, and various other nonhedonic attributes, then it could be the case that [u.sub.i](x) > [u.sub.i](y), because i will end up with more pleasure in x, but that i incorrectly believes his bundle of nonhedonic attributes in y will cause him more pleasure. (This is just a kind of hedonic forecasting error.) In particular, an individual who cares only about pains and pleasures might give an incorrect answer to a question (such as the following) asking him to compare his preference-utility in the actual world x to what his utility would be in y, given the counterfactual bundle of nonhedonic attributes specified in y. "Consider your actual income and health. Is your preference-utility higher than it would be if, instead, you had this level of health and this much income?"
(53.) On what it means to "satisfy" preferences, see generally HAUSMAN, supra note 26.
(54.) DIENER ET AL., supra note 10, at 11 (citation omitted).
(55.) Id. at 11 (citation omitted).
(56.) Paul Dolan & Daniel Kahneman, Interpretations" of Utility and Their Implications .for the Valuation of Health, 118 ECON. J. 215,215 (2008) (citations omitted).
(57.) For a discussion of both the stated-preference methodology and the traditional view favoring behavioral evidence of preferences, see sources cited infra notes 119, 121.
(58.) For example, Kahneman assigns experience utility to moments. See infra Part III.B.
(59.) See supra note 45 (describing evidence of SWB impact of parenting).
(60.) See supra note 44 (describing evidence of SWB impact of education).
(61.) Laura A. King & Christie K. Napa, What Makes a Life Good?, 75 J. PERSONALITY & SOC. PSYCHOL. 156, 158 (1998).
(62.) Id. at 158-61.
(63.) Matthew Adler & Paul Dolan, Introducing a "Different Lives" Approach to the Valuation of Health and Well-Being 12 (Kenan Inst. for Ethics at Duke Univ., Working Paper No. 5, 2012).
(64.) Benjamin et al., supra note 51, at 2087.
(66.) Id. at 2087-88.
(67.) Id. at 2097.
(68.) Id. at 2085.
(70.) Amos Tversky & Dale Griffin, Endowments and Contrast in Judgments of Well-Being, in CHOICES, VALUES, AND FRAMES 709, 722-23 (Daniel Kahneman & Amos Tversky eds., 2000).
(71.) Daniel J. Benjamin, Ori Heffetz, Miles S. Kimball & Alex Rees-Jones, Can Marginal Rates of Substitution Be Inferred from Happiness Data? Evidence from Residency Choices 9-15, 31 (Johnson Sch. Research Paper Series, No. 8-2013. 2013). available at http://papers.ssrn.com/ sol3/papers.cfm?abstract_id=2221538.
(72.) DOLAN & METCALFE. supra note 18, at 11, 20; see also Ed Diener & Christie Scollon, Subjective Well-Being Is Desirable, but Not the Summum Bonum 8-9, 20 (July 2. 2003) (unpublished manuscript) (on file with the Duke Law Journal) (asking college students in various countries to rate happiness compared to other values such as wealth, love, health, and so forth).
(73.) DOLAN & METCALFE, supra note 18. at 5.20 tbl.10.
(74.) Ann Bowling & Joy Windsor, Towards the Good Life: A Population Survey of Dimensions of Quality of Life, 2 J. HAPPINESS STUD. 55, 61 (2001).
(75.) Id. at 59, 61, 64.
(76.) See generally Xavier Badia, Michael Herdman & Paul Kind, The Influence of Ill-Health Experience on the Valuation of Health, 13 PHARMACOECONOMICS 687 (1998); Laura J. Damschroder, Brian J. Zikmund-Fisher & Peter A. Ubel, Considering Adaptation in Preference Elicitations, 27 HEALTH PSYCHOL. 394 (2008) [hereinafter Damschroder et al., Considering Adaptation]; Laura J. Damschroder, Brian J. Zikmund-Fisher & Peter A. Ubel, The Impact of Considering Adaptation in Health State Valuation 61 SOC. SCI. & MED. 267 (2005); Jason Riis et al.. Ignorance of Hedonic Adaptation to Hemodialysis: A Study Using Ecological Momentary Assessment, 134 J. EXPERIMENTAL PSYCHOL.: GEN. 3 (2005): Dylan Smith et al., Mispredicting and Misremembering: Patients with Renal Failure Overestimate Improvements in Quality of Life After a Kidney Transplant, 27 HEALTH PSYCHOL. 653 (2008); Dylan M. Smith, Stephanie L. Brown & Peter A. Ubel, Mispredictions and Misrecollections: Challenges for Subjective Outcome Measurement, 30 DISABILITY & REHABILITATION 418 (2008); Dylan M. Smith, Ryan L. Sherriff, Laura Damschroder, George Loewenstein & Peter A. Ubel, Misremembering Colostomies? Former Patients Give Lower Utility Ratings Than Do Current Patients, 25 HEALTH PSYCHOL. 688 (2006) [hereinafter Smith et al., Misremembering Colostomies]; Peter A. Ubel, George Loewenstein & Christopher Jepson, Disability and Sunshine: Can Hedonic Predictions Be Improved by Drawing Attention to Focusing Illusions or Emotional Adaptation?, 11 J. EXPERIMENTAL PSYCHOL: APPLIED 111 (2005); Peter A. Ubel et al., Do Nonpatients Underestimate the Quality of Life Associated with Chronic Health Conditions Because of a Focusing Illusion?, 21 MED. DECISION MAKING 190 (2001): Peter A. Ubel, George Loewenstein, Norbert Schwarz & Dylan Smith, Misimagining the Unimaginable: The Disability Paradox and Health Care Decision Making, 24 HEALTH PSYCHOL. $57 (2005) [hereinafter Ubel et al., Misimagining the Unimaginable]; Peter A. Ubel, George Loewenstein & Christopher Jepson, Whose Quality of Life? A Commentary Exploring Discrepancies Between Health State Evaluations of Patients and the General Public, 12 QUALITY LIFE RES. 599 (2003) [hereinafter Ubel et al., Whose Quality of Life?].
(77.) See Matthew D. Adler, QALYs and Policy Evaluation: A New Perspective, 6 YALE J. HEALTH POL'Y L. & ETHICS 1, 1 & n.1, 8-9 nn. 25-26 (2006) (collecting sources).
(78.) Loewenstein & Ubel, supra note 19, at 1799.
(79.) This finding has been termed the "disability paradox." Ubel et al., Misimagining the Unimaginable, supra note 76, at $57.
(80.) Damschroder et al., Considering Adaptation, supra note 76, at 394.
(81.) See Loewenstein & Ubel, supra note 19, at 1799-1800, 1803.
(82.) Smith et al., Misremembering Colostomies, supra note 76, at 689-90.
(83.) See id. at 192 tbl.2. The life-satisfaction and mood differences between current and former patients were not statistically significant, Id.
(84.) It is not conclusive because colostomy patients may be mispredicting the hedonic benefit of a return to perfect health.
(85.) See supra note 56 and accompanying text.
(86.) Andrew E. Clark, Paul Frijters & Michael A. Shields, Relative Income, Happiness, and Utility: An Explanation for the Easterlin Paradox and Other Puzzles, 46 J. ECON. LITERATURE 95, 99-106, 115 (2008). The "paradox," first observed by Richard Easterlin, is the substantial growth of income in many countries without a corresponding rise in happiness levels. Id. at 95. But see Daniel W. Sacks, Betsey Stevenson & Justin Wolfers, Subjective Wellbeing, Income, Economic Development and Growth, in ... AND THE PURSUIT OF HAPPINESS: WELL-BEING AND THE ROLE OF GOVERNMENT 59, 59 (Philip Booth ed.. 2012) (challenging the existence of the paradox).
(87.) Clark et al., supra note 86, at 99 (emphasis added).
(88.) Id. at 115 (emphasis added). At the conclusion of Section 4 of the article, Clark et al. note reason for caution about "the link between happiness and utility" and here, again, make explicit that by "utility" they mean "decision utility." Id. at 121: see also Erik Angner, Subjective Well-Being: When. and Why, It Matters 18 (Aug. 31, 2012) (unpublished manuscript), available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2157140 (suggesting that a happiness measure can provide useful information about preference-realization insofar as individuals want happiness or happiness is sufficiently correlated with preference-realization); Peter Railton, Subjective Well-Being as Information and Guidance 33 (Mar. 2012) (unpublished manuscript), available at http://profron.net/happiness/files/readings/Railton_SubjectiveWellBeing.pdf (suggesting that the affective and life-satisfaction components of an individual's SWB provide information about her success in attaining, respectively, short-term and long-term goals).
(89.) See, e.g., Blanchflower & Oswald, supra note 24, at 1361; Rafael
Di Tella & Robert MacCulloch. Some Uses of Happiness Data in Economics, J. ECON. PERSP., Winter 2006. at 25, 28.
(90.) See infra Part II.A.3.
(91.) To say that a given individual i prefers bundle A to bundle [A.sup.*] is a shorthand for saying: i prefers it to be the case that she has attributes A, as opposed to it being the case that she has attributes [A.sup.*]. Thus, when we compare how various individuals rank a given set of attribute bundles, we are not comparing how those individuals rank the very same states of affairs. Rather, we are comparing how one individual ranks states of affairs specified as those in which she has various possible attributes, to how a second individual ranks states of affairs specified as those in which he has various possible attributes.
(92.) Let u(.) be one function of attribute bundles, [u.sup.*](.) a second. To say that [u.sup.*](.) is an increasing transformation of u(.) means that whenever u(A) = u(B). [u.sup.*](A) = [u.sup.*](B) and whenever .(A) > u(B), [u.sup.*](A) > [u.sup.*](B).
(93.) For a formal model of adaptive preferences, see Clark et al., supra note 86, at 104-06.
(94.) See ADLER. supra note 27, at 279 n.57 (collecting sources).
(95.) E.g., Wim Groot, Adaptation and Scale of Reference Bias in Self-Assessments of Quality of Life, 19 J. HEALTH ECON. 403 (2000): Heather P. Lacey et al., Are They Really That Happy? Exploring Scale Recalibration in Estimates of Well-Being, 27 HEALTH PSYCHOL. 669 (2008): Maarten Lindeboom & Eddy van Doorslaer, Cut-Point Shift and Index Shift in Self-Reported Health, 23 J. HEALTH ECON. 1083 (2004): Erik Meijer, Arie Kapteyn & Tatiana Andreyeva, Internationally Comparable Health Indices, 20 HEALTH ECON. 600 (2011): Debby Postulart & Eddy M.M. Adang, Response Shift and Adaptation in Chronically Ill Patients, 20 MED. DECISION MAKING 186 (2000): Joshua A. Salomon, Ajay Tandon & Christopher J.L. Murray, Comparability of Self Rated Health: Cross Sectional Multi-Country Survey Using Anchoring Vignettes, 328 BRIT. MED. J. 258 (2004), doi:10.1136/bmj.37963.691632.44: Amir Shmueli, Reporting Heterogeneity in the Measurement of Health and Health-Related Quality of Life, 20 PHARMACOECONOMICS 405 (2002): Amir Shmueli, Socio-Economic and Demographic Variation in Health and in Its Measures: The Issue of Reporting Heterogeneity, 57 SOC. SCI. & MED. 125 (2003): Miriam A.G. Sprangers & Carolyn E. Schwartz, Integrating Response Shift into Health-Related Quality of Life Research: A Theoretical Model, 48 SOC. SCI. & MED. 1507 (1999): Peter A. Ubel. Aleksandra Jankovic, Dylan Smith, Kenneth M. Langa & Angela Fagerlin, What Is Perfect Health to an 85-Year-Old?: Evidence for Scale Recalibration in Subjective Health Ratings, 43 MED. CARE 1054 (2005).
(96.) Ubel et al., supra note 95, at 1055 tbl.1.
(97.) Id. at 1056.
(98.) Salomon et al., supra note 95, at 2.
(99.) Id. at 3-4.
(100.) See supra notes 76-84 and accompanying text.
(101.) To be sure, scale recalibration would only be relevant for QALY values elicited via ratings, rather than via time-tradeoff or standard-gamble questions.
(102.) See Ubel et al., Whose Quality of Life?, supra note 76, at 604-05.
(103.) See Andrew J. Oswald, On the Curvature of the Reporting Function from Objective Reality to Subjective Feelings, 100 ECON. LETTERS 369, 370-71 (2008).
(104.) See Gary King, Christopher J.L. Murray, Joshua A. Salomon & Ajay Tandon, Enhancing the Validity and Cross-Cultural Comparability of Measurement in Survey Research, 98 AM. POL. SCI. REv. 191 (2004) (discussing the use of "vignettes'" to correct for scale heterogeneity, and illustrating this technique with respect to measures of political efficacy and visual acuity): see also Mary Steffel & Daniel M. Oppenheimer, Happy by What Standard? The Role of Interpersonal and Intrapersonal Comparisons in Ratings of Happiness, 92 SOC. INDICATORS RES. 69 (2009) (finding scale heterogeneity with respect to happiness scales).
(105.) See BERNARD M.S. VAN PRAAG & ADA FERRER-I-CARBONELL, HAPPINESS QUANTIFIED: A SATISFACTION CALCULUS APPROACH 5-6, 34-42 (2004).
(106.) Norbert Schwarz & Fritz Strack, Reports of Subjective Well-Being: Judgmental Processes and Their Methodological Implications, in WELL-BEING: THE FOUNDATIONS OF HEDONIC PSYCHOLOGY. supra note 1. at 61. The terms "'evaluation error" and "miscommunication'" are my own; Schwarz and Strack do not use these terms.
(107.) Cf Hendrik Jurges. Unemployment, Life Satisfaction and Retrospective Error, 170 J. ROYAL STAT. SOC'Y 43, 44 (2007) (discussing the effect of inaccurate memory on survey responses).
(108.) Schwarz & Strack. supra note 106, at 63 (citations omitted). A recent review concludes: "What information people attend to when responding to surveys can strongly affect life satisfaction judgments." Ed Diener. Ronald Inglehart & Louis Tay, Theory and Validity of Life Satisfaction Scales, SOC. INDICATORS RES. 11 (2012), link.springer.corn/article/10.1007 %2Fs11205-012-0076-y?LI=true.
(109.) Schwarz & Strack, supra note 106, at 63.
(110.) Id. at 74 (citations omitted). Subsequent findings with respect to weather effects have been mixed. See Diener et al., supra note 108, at 18-19: Sylvia Kampfer & Michael Mutz, On the Sunny Side of Life: Sunshine Effects on Life Satisfaction, 110 SOC. INDICATORS RES. 579 (2013).
(111.) On cultural differences in responses to SWB surveys, see generally Shigehiro Oishi & Ed Diener, Culture and Well-Being: The Cycle of Action. Evaluation, and Decision, 29 PERSONALITY & SOC. PSYCHOL. 939 (2003).
(112.) See Schwarz & Strack, supra note 106, at 64.
(113.) Defining cardinality is a subtle matter, and the definition offered here is rough, but will suffice for the purpose of this Article. Assume that i has a preference structure represented by v(.) and by any other function [v.sup.*](.) just in case [v.sup.*] (.) is an eligible transformation of v(.). For example, it is well known that, in the case of preferences regarding lotteries, [v.sup.*] (.) needs to be a positive linear transformation of v(.), that is, [v.sup.*] (.) = av(.) + b, with a positive. Similarly, in the case of time-tradeoff preferences, [v.sup.*] (.) may need to be a positive ratio transformation of v(.). that is, [v.sup.*] (.) = av(.), with a positive.
Now consider two possible series of attribute packages: A, B, ... versus A', B',.... Utility function v(.) is cardinal if it represents a feature of preferences such that. for every eligible [v.sup.*] (.), v(A) + v(B) + ... [greater than or equal to] v(A') + v(B') + ... if and only if [v.sup.*] (A) + [v.sup.*] (B) + ... [greater than or equal to] [v.sup.*] (A') + [v.sup.*] (B') +.... Utility function v(.) and all eligible transformations thereof assign overall sums to series of packages so as to rank these series the same way. It is therefore "meaningful" to engage in an operation such as v(A) + v(B) + ..., and so v(.) can be termed "cardinal."
Note that. if [v.sup.*] (.) is a positive ratio transformation, v(.) is cardinal. Moreover, if all series being compared have the same number of terms, and [v.sup.*] (.) is a positive linear transformation, v(.) is cardinal. Thus "cardinality." in the scholarly literature, is often associated with being a ratio and/or linear transformation.
(114.) For helpful discussions of econometric issues in SWB surveys, see Andrew Clark, Fabrice Etile, Fabien Postel-Vinay, Claudia Senik & Karine Van der Straeten, Heterogeneity in Reported Well-Being, 115 ECON. J. C118 (2005): Ada Ferrer-i-Carbonell & Paul Frijters, How Important Is Methodology for the Estimates of the Determinants of Happiness?, 114 ECON. J. 641 (2004); Simon Luechinger, Valuing Air Quality Using the Life Satisfaction Approach, 119 ECON. J. 482 (2009): Erzo F.P. Luttmer, Neighbors as Negatives. Relative Earnings and Well-Being, 120 Q.J. ECON. 963 (2005): and Nattavudh Powdthavee, How Much Does Money Really Matter? Estimating the Causal Effects of Income on Happiness, 39 EMPIRICAL ECON. 77 (2010): Dolan et al., supra note 11.
(115.) Blanchflower & Oswald, supra note 24, at 1361-62.
(116.) See supra note 114: see also Marianne Bertrand & Sendhil Mullainathan, Do People Mean What They Say? Implications for Subjective Survey Data, 91 AM. ECON. REV. (PAPERS & PROC.) 67, 67, 71-72 (2001) (arguing against the use of SWB and similar surveys to predict the effect of observable individual attributes on individuals' attitudes because of the correlation of measurement error with those attributes).
(117.) See Clark et al., supra note 114. at C119.
(118.) See Powdthavee, supra note 114, at 79: Paul Dolan & Robert Metcalfe, Comparing Willingness-to-Pay and Subjective Well-Being in the Context of Non-Market Goods 20 (Ctr. for Econ. Performance, Discussion Paper No. 890, 2008).
(119.) For overviews of this approach, see generally IAN J. BATEMAN ET AL., ECONOMIC VALUATION WITH STATED PREFERENCE TECHNIQUES: A MANUAL (2002): RICHARD T. CARSON, CONTINGENT VALUATION: A COMPREHENSIVE BIBLIOGRAPHY AND HISTORY (2011); HANDBOOK ON CONTINGENT VALUATION (Anna Alberini & James R. Kahn eds., 2006); A PRIMER ON NONMARKET VALUATION (Patricia A. Champ, Kevin J. Boyle & Thomas C. Brown eds., 2003): VALUING ENVIRONMENTAL AMENITIES USING STATED CHOICE STUDIES (Barbara J. Kanninen ed., 2007); VALUING ENVIRONMENTAL PREFERENCES: THEORY AND PRACTICE OF THE CONTINGENT VALUATION METHOD IN THE US, EU, AND DEVELOPING COUNTRIES (Ian J. Bateman & Kenneth G. Willis eds., 2001); Richard T. Carson & W. Michael Hanemann, Contingent Valuation, in 2 HANDBOOK OF ENVIRONMENTAL ECONOMICS 821 (Karl-Goran Maler & Jeffrey R. Vincent eds., 2005): L. Venkatachalam, The Contingent Valuation Method: A Review, 24 ENVTL. IMPACT ASSESSMENT REV. 89 (2004).
(120.) A different format, less utilized at present, is an "open-ended" question that asks for the maximum the respondent is willing to pay for some good.
(121.) See supra note 119: see also FUJIWARA & CAMPBELL, supra note 11: Matthew D. Adler, Welfare Polls: A Synthesis, 81 N.Y.U.L. REV. 1875 (2006): Richard T. Carson, Nicholas E. Flores & Norman F. Meade, Contingent Valuation: Controversies and Evidence, 19 ENVTL. & RESOURCE ECON. 173 (2001); Daniel Kahneman, Ilana Ritov & David Schkade, Economic Preferences or Attitude Expressions?: An Analysis of Dollar Responses to Public Issues, 19 J. RISK & UNCERTAINTY 203 (1999); Daniel Kahneman & Robert Sugden, Experienced Utility as a Standard of Policy Evaluation, 32 ENVTL. & RESOURCE ECON. 161 (2005); Robert Sugden, Anomalies and Stated Preference Techniques: A Framework ,for a Discussion of Coping Strategies, 32 ENVTL. & RESOURCE ECON. 1 (2005).
(122.) See supra text accompanying notes 35-48 (discussing the laundering of preferences).
(123.) See, e.g., Jerry Hausman, Contingent Valuation: From Dubious to Hopeless, J. ECON. PERSP., Fall 2012, at 43, 46-49.
(124.) This is a violation of expected utility theory. See James K. Hammitt & John D. Graham, Willingness To Pay for Health Protection: Inadequate Sensitivity to Probability?, 18 J. RISK & UNCERTAINTY 33, 34-35 (1999).
(125.) John W. Payne, James R. Bettman & David A. Schkade, Measuring Constructed Preferences: Towards a Building Code, 19 J. RISK & UNCERTAINTY 243 (1999).
(126.) Cf Dan Ariely, George Loewenstein & Drazen Prelec, "Coherent Arbitrariness": Stable Demand Curves Without Stable Preferences, 118 Q.J. ECON. 73, 73 (2003) (finding that market behavior may not reveal stable underlying preferences).
(127.) See CARSON, supra note 119, at 13 (noting that hypothetical bias and strategic misstatement are two enduring concerns about stated-preference surveys).
(128.) See Bruno S. Frey & Alois Stutzer, The Use of Happiness Research for Public Policy, 38 SOC. CHOICE & WELFARE 659,670 (2012) ("When happiness indicators influence the behaviour of political actors and their policy choices, individuals have an incentive to misstate their well-being.").
(129.) For one particularly intensive effort to encourage deliberation about preferences as part of a stated-preference survey, see Douglas MacMillan, Nick Hanley & Nele Lienhoop, Contingent Valuation: Environmental Polling or Preference Engine?, 60 ECOLOGICAL ECON. 299 (2006).
(130.) In practice, stated-preference practitioners often use a so-called "dichotomous choice" framework which, in effect, asks each respondent to rank only two attribute bundles (bundles consisting of some level of a good and some amount of income) out of a larger set. Dichotomous-choice questions, besides having desirable properties as regards strategic bias, are less cognitively demanding than questions asking the respondent for fuller ranking information. Answers to a survey posing each respondent a dichotomous-choice question over some pair of bundles, together with assumptions about the homogeneity of respondents' preferences, can be used to infer the common preferences over the entire set. But it is also quite possible to ask each respondent to rank three or more bundles, or to pick the most preferred of three or more bundles: and stated-preference surveys sometimes do pose such questions, as in the so-called "choice experiment" or "contingent ranking" format.
(131.) Some critics of using stated-preference surveys to estimate WTP/WTA values have argued that these surveys impose a cognitive burden by requiring respondents to rank multiple bundles--and have suggested that such burden is reduced by the SWB format. See, e.g., Frey et al., supra note 11, at 148. But this cognitive-burden defense of the SWB format is persuasive only if SWB surveys are taken as evidence of something about respondents other than their preferences. The "'burden" of ranking multiple alternatives is an inextricable part of having a preference.
(132.) See Infra Part III.B.
(133.) LAYARD. supra note 10.
(134.) SUMNER, supra note 26, at 83-92.
(135.) JOHN RAWLS, A THEORY OF JUSTICE (rev. ed. 1999).
(136.) See id. at 358-80.
(137.) See generally RICHARD B. BRANDT, A THEORY OF THE GOOD AND THE RIGHT (1979).
(138.) See generally JAMES GRIFFIN, WELL-BEING: ITS MEANING, MEASUREMENT, AND MORAL IMPORTANCE (1986).
(139.) See generally FINNIS, supra note 42.
(140.) See generally NUSSBAUM, supra note 31.
(141.) See generally SUMNER. supra note 26.
(142.) See, e.g.. Sharon Hewitt, What Do Our Intuitions About the Experience Machine Really Tell Us About Hedonism?, 151 PHIL. STUD. 331,332 (2010): Matthew Silverstein, hi Defense of Happiness: A Response to the Experience Machine, 26 SOC. THEORY & PRAC. 279, 281-82 (2000).
(143.) See, e.g., ROGER CRISP. REASONS AND THE GOOD (2006): FRED FELDMAN. PLEASURE AND THE GOOD LIFE: CONCERNING THE NATURE, VARIETIES, AND PLAUSIBILITY OF HEDONISM (2004): FRED FELDMAN, WHAT IS THIS THING CALLED HAPPINESS? (2010): Mark Bernstein. Well-Being, 35 AM. PHIL. Q. 39 (1998): Torbjorn Tannsjo, Narrow Hedonism, 8 J. HAPPINESS STUD. 79 (2007).
(144.) See supra text accompanying note 48: infra notes 163-175 and accompanying text.
(145.) LAYARD. supra note 10. at 111-12. Layard's book was originally published in 2005. Citations are to the revised and updated edition, published in 2011, which added a new part but did not change the main text. See id. at xiii. The new part adds little to the original argument for happiness maximization, except a slight addendum to the argument about the uniquely self-evident intrinsic value of happiness. See infra note 154.
(146.) Id. at 112-13 (footnote omitted).
(147.) Id. at 124.
(148.) Id. at 112.
(149.) See generally RALPH L. KEENEY & HOWARD RAIFFA, DECISIONS WITH MULTIPLE OBJECTIVES: PREFERENCES AND VALUE TRADEOFFS (1993).
(150.) For a fuller discussion of the multidimensionality of good experiences, see infra notes 197-200 and accompanying text.
(151.) See infra Part III.B.1.
(152.) LAYARD, supra note 10, at 113.
(153.) See NORMAN DANIELS, JUSTICE AND JUSTIFICATION: REFLECTIVE EQUILIBRIUM IN THEORY AND PRACTICE (1996).
(154.) In the new part to his book, added in the revised and updated edition, Layard claims--without empirical support--that happiness "is the only good which would be generally accepted as an end in itself." LAYARD. supra note 10, at 240.
(155.) See ELLIOTT SOBER & DAVID SLOAN WILSON. UNTO OTHERS: THE EVOLUTION AND PSYCHOLOGY OF UNSELFISH BEHAVIOR 296-327 (1998).
(156.) See Diener & Seligman, supra note 10, at 13-15.
(157.) LAYARD. supra note 10. at 113.
(158.) Robert Sugden, Capability, Happiness, and Opportunity, in CAPABILITIES AND HAPPINESS 299. 300 (Luigino Bruni, Flavio Comim & Maurizio Pugno eds., 2008).
(159.) To be sure, preference views become increasingly paternalistic as they impose increasingly stringent rationality and informational conditions on preferences. The preference component of any such view pushes against paternalism, whereas the idealization component pushes toward it.
(160.) LAYARD. supra note 10, at 124.
(161.) See. e.g., Bronsteen, Buccafusco and Masur, Welfare as Happiness, supra note 10: Bronsteen. Buccafusco and Masur, Well-Being Analysis, supra note 10.
(162.) See supra note 143.
(163.) See SUMNER, supra note 26: Bernstein, supra note 143, at 45-46: Shelly Kagan, The Limits of Well-Being, SOC. PHIL. & POL'Y, June 1992, at 169, 171-72.
(164.) The word "direct" is needed because remote occurrences might have an instrumental, causal, impact on someone's well-being.
(165.) Kagan, supra note 163, at 171-72.
(166.) See DEREK PARFIT, REASONS AND PERSONS 494 (1987).
(167.) ADLER & POSNER, supra note 32, at 34.
(168.) See id. at 34-35; ADLER, supra note 27, at 174-78.
(169.) See ADLER, supra note 27, at 178-81.
(170.) See supra notes 31-34 and accompanying text.
(171.) Kagan, supra note 163, at 180-89.
(172.) See generally DAVID DEGRAZIA, HUMAN IDENTITY AND BIOETHICS (2005): ERIC T. OLSON. THE HUMAN ANIMAL: PERSONAL IDENTITY WITHOUT PSYCHOLOGY (1997).
(173.) See supra notes 34-47 and accompanying text.
(174.) ADLER & POSNER, supra note 32, at 30. Let me modify the case somewhat to highlight the connection with goals, knowledge, and relationships. Imagine that David, an academic scientist, works hard at his research. He hopes to make a scientific discovery and to be respected for doing so by his colleagues. In considering possible outcomes, David says that he prefers x (an outcome in which he actually discovers some new and significant scientific fact), as opposed to y (an outcome in which he falsely believes to have done so, abetted in this belief by colleagues who want to spare his feelings but pity David behind his back). Even though David's mental states are identical in these possible outcomes, there are various important respects in which his preference seems self-interested whereas Sheila's does not: in y he has failed at his career goal, he knows less (and has made no contribution to human knowledge), and a thread of deceit runs through his relationships with his colleagues.
Note that both conceptions (3) and (4) would count David's preference for x as self-interested. A preference that I contribute to finding a cure for cancer--by contrast with a preference that a cure for cancer be discovered--is existence-entailing, because all outcomes in which I do not exist are ranked equal by the first but not the second preference. David's preference is that he make a scientific breakthrough and that he be respected by his colleagues as a result. Moreover, someone caring about David (whether David himself, or someone else) would be motivated to pursue x. For example, if David has a sympathetic friend who knows of David's capacity for self-deception, the propensities of David's colleagues, and so forth, and believes that David's research is heading down a false path--and thus taking David in the direction of y not x--the friend might try to get David to redirect his research.
(175.) Bronsteen, Buccafusco and Masur, Welfare as Happiness, supra note 10, at 1621. They also point out that an account of self-interest which categorizes as "self-interested" the deceived scholar's preference for genuine academic success (as opposed to the mere belief thereof) also thus categorizes the preference for success of a "driven scholar" willing to sacrifice his happiness and family life. Id. at 1625-26. This is true. But it is a further question whether this latter preference survives full information and rational reflection on the driven scholar's part, and yet a further question whether it would be widely shared (a point relevant to interpersonal comparisons, see ADLER, supra note 27, at 185-225). Note, here, that the critic of the experientialism requirement has a pretty easy argumentative burden. What she needs to show is that there are some cases in which someone's well-being is directly improved by nonexperiential changes, not that such changes always override experiential losses.
(176.) For the theoretical elaboration of the objective-happiness framework, see Daniel Kahneman, Peter P. Wakker & Rakesh Sarin, Back to Bentham? Explorations of Experienced Utility. 112 Q.J. ECON. 375 (1997); Daniel Kahneman, Experienced Utility and Objective Happiness: A Moment-Based Approach, in CHOICES. VALUES, AND FRAMES, supra note 70, at 673 [hereinafter Kahneman, Experienced Utility]; Daniel Kahneman & Jacob Riis, Living, and Thinking About It. Two Perspectives on Life, in THE SCIENCE OF WELL-BEING, supra note 1, at 285: and Daniel Kahneman, Objective Happiness, in WELL-BEING: THE FOUNDATIONS OF HEDONIC PSYCHOLOGY, supra note 1, at 3, 3-12 [hereinafter Kahneman, Objective Happiness]. For empirical implementation, see generally MEASURING THE SUBJECTIVE WELL-BEING OF NATIONS, supra note 4; Kahneman et al., supra note 7: and Alan B. Krueger et al.. Time Use and Subjective Well-Being in France and the U.S., 93 SOC. INDICATORS RES. 7 (2009). For further discussions of the framework, see Dolan & Kahneman. supra note 56: Daniel Kahneman & Alan B. Krueger, Developments in the Measurement of Subjective Well-Being, J. ECON. PERSP., Winter 2006, at 3; Kahneman et al., supra note 10; Kahneman & Sugden, supra note 121.
(177.) Kahneman, Experienced Utility, supra note 176, at 683 (citations omitted). For other passages in which Kahneman declines to commit himself to an experientialist account of well-being, see Dolan & Kahneman, supra note 56, at 229-30; Kahneman et al., supra note 176, at 377 & n.3, Kahneman & Sugden, supra note 121, at 178 n.11; Kahneman & Riis, supra note 176, at 288-89.
(178.) Kahneman, Experienced Utility, supra note 176, at 683.
(179.) See Kahneman et al., supra note 176, at 388-89, 91: Kahneman, Experienced Utility, supra note 176, at 680-81: Kahneman, Objective Happiness, supra note 176, at 5-6.
(180.) It might seem absurd to compare profiles with short temporal duration. For example, if P consists of an hour of certain experiences, and P* an hour of different experiences, is the observer meant to contemplate a life that endures for only an hour? Or (almost as absurdly) a life that is entirely neutral except for P, as compared to a life entirely, neutral except for P*? Given separability, this problem can be avoided. The observer can rank P and P* by contemplating some arbitrary profile M followed by P, as compared to the very same profile M followed by P*.
(181.) Assume that v(.) is such that its duration-weighted values represent the observer's ranking of profiles. Let [v.sup.*] (.) be defined as follows: for every momentary experience A, [v.sup.*] (A) = av(A), where a is a positive constant. Note now that duration-weighted [v.sup.*] (.) values equally well represent the observer's ranking of profiles. Why? Let (A, B, ...) be a series of momentary experiences and (A', B', ....) a different series, with each experience A enduring for time t(A). Then whenever v(A)t(A) + v(B)t(B) + ... [greater than or equal to] v(A')t(A') + v(B')t(B') + ..., it will also be true that [v.sup.*] (A)t(A) + [v.sup.*] (B)t(B) + ... [greater than or equal to] [v.sup.*] (A')t(A') + [v.sup.*] (B')t(B') +.... Kahneman goes further, showing that [v.sup.*] (.) represents the observer's ranking not just if, but only if, it is a positive ratio transformation of v(.). Kahneman et al., supra note 176, at 398-402.
(182.) See, e.g.. John T. Cacioppo & Gary G. Berntson, Relationship Between Attitudes and Evaluative Space: A Critical Review, with Emphasis on the Separability off Positive and Negative Substrates, 115 PSYCHOL. BULL. 401 (1994); Richard J. Davidson. Anterior Cerebral Asymmetry and the Nature of Emotion, 20 BRAIN & COGNITION 125 (1992): Ed Diener & Robert A. Emmons. The Independence of Positive and Negative Affect, 47 J. PERSONALITY & SOC. PSYCHOL. 1105 (1985): Nico H. Frijda, Emotions and Hedonic Experience, in WELL-BEING: THE FOUNDATIONS OF HEDONIC PSYCHOLOGY, supra note 1, at 190: Peter J. Lang, The Emotion Probe: Studies off Motivation and Attention, 50 AM. PSYCHOLOGIST 372 (1995): Richard E. Lucas. Ed Diener & Eunkook Suh, Discriminant Validity of Well-Being Measures, 71 J. PERSONALITY & SOC. PSYCHOL. 616 (1996): David Watson, Lee Anna Clark & Auke Tellegen, Development and Validation of Brief Measures of Positive and Negative Affect: The PANAS Scales, 54 J. PERSONALITY & SOC. PSYCHOL. 1063 (1988): David Watson & Auke Tellegen, Toward a Consensual Structure of Mood, 98 PSYCHOL. BULL. 219 (1985).
(183.) Frijda, supra note 182, at 195.
(184.) Kahneman, Experienced Utility, supra note 176, at 682.
(185.) Kahneman. Objective Happiness, supra note 176, at 8 (emphasis added) (citation omitted) (citing Davidson. supra note 182).
(186.) Inaccessibility would jeopardize any approach that counts mental experience as one component of well-being (including the approach I argue for in the next Section), not merely the "objective happiness" framework.
(187.) Kahneman at one point suggests that ordinal rather than cardinal momentary utilities will typically be adequate for policy purposes. See Kahneman & Riis. supra note 176. at 290 ("Except for the rare cases in which cumulative distributions [of moment utility over time] cross, the mean (or the median) of the distribution of moment utility is an ordinal measure of total utility that can be compared across situations, people and populations."). The assertion that the crossing of cumulative distributions will be "rare" is pure speculation.
(188.) See supra notes 77-84 and accompanying text.
(189.) See supra notes 7-8 and accompanying text.
(190.) The numbers might be cardinal, but in the sense of (1) representing the respondent's ranking of lotteries over experiences, as per expected utility theory: (2) representing the respondent's judgments regarding the hedonic differences between experiences: (3) deriving from a multiattribute function that aggregates subutility functions corresponding to the multiple dimensions of hedonic experience: or (4) communicating a primitive, cardinal measure of the felt intensity of the experiences. On this point, note too that QALY values elicited via explicit time-tradeoff questions need not--and in practice, do not--correspond to QALY values elicited through a direct rating question or the standard-gamble question. See supra note 77.
(191.) Krueger et al.. supra note 4, at 9.
(192.) See id. at 19 (stating that "[t]he U-index is an ordinal measure at the level of feelings" because the classification of an episode as unpleasant or pleasant "relies purely on an ordinal ranking of the feelings within each episode").
(193.) In the theoretical presentation of the "objective happiness" framework, it is assumed that momentary experiences can be compared to the neutral level, and that experiences within each hedonic domain (positive or negative) can be compared to each other. So pains are ordered, pleasures are ordered, but pains and pleasures are not necessarily ordered vis-a-vis each other except in the sense of being categorized as worse or better than neutral. The U-index procedure assumes that the intensity of any positive or negative experience can be compared to the intensity of any other positive or negative experience. See id. at 19 n.13. It is much more contestable whether subjects can make such comparisons.
(194.) Imagine that Jim on Tuesday has more negative moments than Jim today, but that his positive moments on Tuesday--when they occur--are intensely positive, whereas nothing so joyful happens to him today.
(195.) Kahneman, Experienced Utility, supra note 176, at 678.
(196.) See ADLER. supra note 27, at 419-20; see also Loewenstein. supra note 42, at 100-02 (challenging an approach that measures experiential quality by summing moment utilities, and noting both that peak negative and positive experiences have an importance that is disproportionate to their duration and that individuals encode experiences in terms of "episodes" rather than moments).
(197.) See, e.g., Loewenstein & Ubel, supra note 19 at 1801-04: Loewenstein. supra note 42 at 94-96.
(198.) See, e.g., Diener et al., supra note 1, at 277.
(199.) DOLAN ET AL., supra note 17, at 6-9.
(200.) Kahneman & Riis, supra note 176, at 289 (citing Kahneman, Objective Happiness, supra note 176).
(201.) See id.
(202.) The literature on errors in individual affective forecasting makes this latter premise less plausible. See, e.g., Kahneman & Sugden, supra note 121. at 168-73.
(203.) E.g., Matthew D. Adler, Fear Assessment: Cost-Benefit Analysis and the Pricing of Fear and Anxiety, 79 CH1.-KENT L. REV. 977, 1024-30 (2004); Paul Dolan, Graham Loomes, Tessa Peasgood & Aki Tsuchiya, Estimating the lntangible Victim Costs of Violent Crime, 45 BRIT. J. CRIMINOLOGY 958,959-64 (2005).
(204.) See Adler, supra note 203, at 979-80, 1043-52: Paul Dolan & Tessa Peasgood, Estimating the Economic and Social Costs o.f the Fear of Crime, 47 BRIT. J. CRIMINOLOGY 121, 126 (2007).
(205.) Let x and y be two outcomes. In order to determine what a given individual is WTP/WTA for the move from x to y, all we need to know, in principle, is how he orders attribute bundles (including both income and whatever non-income attributes are specified in the outcomes). This ordering is captured by his ordinal utility function. Moreover, individuals i and j might have different rankings of attribute bundles, and suitable estimation techniques (such as stated-preference surveys) are robust to such heterogeneity. See supra Part II.C.
Admittedly. if we were to compare x and v by using a cardinal utility function for attribute bundles, heterogeneity in utility functions would become a problem. Comparing the sums or averages of bundle utilities using one method for assigning cardinal utilities could yield a different result than comparing those sums or averages using a different method.
(206.) See supra notes 186-187 and accompanying text.
(207.) Either the minimal conditions required to have a preference at all, or the conditions requisite for the preference to have normative "'bite." See supra notes 35-48.
(208.) See Dolan & Kahneman, supra note 56, at 215-16 (questioning the usefulness of measures of "'decision utility," that is, preference utility, as a basis for valuing health states, because individuals' preferences regarding health are inevitably biased to some extent); Kahneman et al., supra note 121. at 228-29 (claiming that "people are better described as having attitudes than preferences," with attitudes "lack[ing] some of the essential properties that economic theory requires of preferences").
MATTHEW D. ADLER ([dagger])
([dagger]) Richard A. Horvitz Professor of Law and Professor of Economics, Philosophy and Public Policy, Duke University. Many thanks for helpful comments to faculty workshop participants at the law schools of Duke University, Loyola University Chicago, the University of California, Berkeley, the University of Illinois, and the University of Pennsylvania; to faculty workshop participants at Duke's Sanford School of Public Policy; to participants in the Duke Law Journal's 43rd Annual Administrative Law Symposium: "A Happiness Approach to Cost-Benefit Analysis"; and to Dan Benjamin, John Bronsteen, Christopher Buccafusco, Maureen Cropper, David DePianto, Ed Diener, Paul Dolan, Marc Fleurbaey. Stephen Galoob, Carol Graham, Jennifer Hawkins, Ori Heffetz, Rob Kar, Jon Masur, Michael Moore, Doug Noonan, Galen Panger, Eric Posner, Arden Rowell. Rick Swedloff, Peter Ubel, and Jonathan Wiener. The usual disclaimer applies. Thanks to David Merritt and Melinda Patterson for excellent research assistance and to Bill Draper and Kelley Leong for amazing library support.
Table 1. Widely Used SWB Questions Survey Variable Question General Social Happiness Taken all together, how would you say Survey things are these days? Would you say that you are very happy, pretty happy. or not too happy? World Values Life sat. All things considered, how satisfied are Survey you with your life as a whole these days'? Please use this card to help with your answer. [range of 1-10 with 1 labelled "Very Dissatisfied" and 10 labelled "Very Satisfied"] European Social Happiness Taking all things together, how happy Survey would you say you are? Please use this card [range of 0-10 with 0 labelled "Extremely unhappy" and 10 labelled "Extremely happy"] European Social Life sat. All things considered, how satisfied are Survey you with your life as a whole nowadays? Please answer using this card, where 0 means extremely dissatisfied and 10 means extremely satisfied. [range of 0-10 with 0 labelled "Extremely dissatisfied" and 10 labelled "Extremely satisfied"] European Happiness Taking all things together on a scale of Quality of Life 1 of 10. how happy would you say you Survey are? Here 1 means you are very unhappy and 10 means you are very happy. European Life sat. All things considered, how satisfied Quality of Life would you say you are with your life Survey these days? Please tell me on a scale of 1 to 10, where 1 means very dissatisfied and 10 means very satisfied. German Socio- Life sat. In conclusion, we would like to ask you Economic Panel about your satisfaction with your life in general. Please answer according to the following scale: 0 means 'completely dissatisfied', 10 means 'completely satisfied'. How satisfied are you with your life, all things considered? British Life sat. How dissatisfied or satisfied are you Household with your life overall? [range of 1-7 Panel Survey with 1 labelled "Not satisfied at all" and 7 labelled "Completely satisfied".]
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||III. Subjective Well-Being Surveys as Evidence of Experience Utility through Conclusion, with footnotes, p. 1568-1601; A Happiness Approach to Cost-Benefit Analysis|
|Author:||Adler, Matthew D.|
|Publication:||Duke Law Journal|
|Date:||May 1, 2013|
|Previous Article:||Happiness surveys and public policy: what's the use?|
|Next Article:||Well-being analysis vs. cost-benefit analysis.|