Read an Excerpt
Law and Happiness
The University of Chicago PressCopyright © 2010 The University of Chicago
All right reserved.
Chapter OneMeasuring Well-Being for Public Policy: Preferences or Experiences?
Paul Dolan and Tessa Peasgood
Policy makers seeking to enhance well-being are faced with a choice of possible measures that may offer contrasting views about how well an individual's life is going. We suggest that choice of well-being measure should be based on three general criteria: (1) the measure must be conceptually appropriate (that is, are we measuring the right sort of concept for public policy?), (2) it must be valid (that is, is it a good measure of that concept?), and (3) it must be empirically useful (that is, does it provide information in a format that can be readily used by policy makers?). Preference-based measures (as represented by income) are compared to experience-based measures (as represented by subjective evaluations of life) according to these criteria. Neither set of measures meets ideal standards, but experiences do fare at least as well as preferences, and subjective evaluations perform much better than income alone as a measure of well-being.
In various ways, policy makers seek to improve of the well-being of the populations they serve. The question is, do they have a clear idea about what constitutes well-being? Economists have for some time framed this in terms of utility, as represented by preferences (Fisher 1918). The degree to which preferences are fulfilled is determined primarily by an individual's budget constraint, which is determined by her income and prevailing prices. It is not surprising, then, that economists have paid so much attention to national income as a proxy for well-being. More recently, some economists have sought to improve the use of income as a proxy for well-being by focusing only on those aspects of income that are deemed to bring genuine improvements in well-being. Examples of this approach include adjusted income accounts, such as the Index of Economic Well-Being (Osberg 1985). A preference satisfaction (or desire fulfillment) account of well-being has also been at the heart of philosophical discourse.
Many alternative approaches in philosophy adopt mental-state accounts of well-being. These accounts view well-being as a psychological phenomenon characterized by feelings of pleasure and displeasure, happiness and sadness, and satisfaction and dissatisfaction. Such accounts of well-being are generally grounded in hedonistic philosophies (Kahneman 2000), but we use the term "subjective evaluation" to refer to a more general account that considers how people evaluate their lives, as well as how they feel in a strict hedonic sense. Direct measures of subjective evaluation—usually asking people how satisfied they are with their lives overall—have been used by psychologists for 50 years, and they are now becoming popular among economists (Dolan, Peasgood, and White 2006). Subjective evaluation is also making its way up the policy agenda, particularly in the United Kingdom (see, for example, Defra 2005).
Of course, the debate about what constitutes well-being is longstanding. Despite a lack of theoretical agreement, policy decisions reflect judgments, at some level, about the well-being of those who may be affected by the decisions. Moreover, the choice of well-being measure may have very different implications for which people we judge to have high and low levels of well-being. Consider the following data from the British Household Panel Survey (BHPS) in the United Kingdom. The BHPS is an annual survey of about 5,000 households (about 10,000 individuals), has been running since 1991, and is broadly representative of the British population (Nathan 1999).
For subgroups of respondents in 2004/2005 (wave 14), Figures 1–4 show average standardized scores of income (net current household income controlling for household size), consumption (for a limited range of consumption items and controlling for household size), a General Health Questionnaire (GHQ) measure of anxiety and depression (12 questions with four possible responses, giving a score out of 36), life satisfaction (scored from 1 to 7), and weighted responses to satisfaction in eight domains. The domains are health, household income, house or flat, spouse or partner, job, social life, amount of leisure time, and use of leisure time. Weights are derived from a regression model predicting overall life satisfaction.
It is clear that different groups have different levels of well-being based whether preference satisfaction (income or consumption) or subjective evaluation (GHQ, life satisfaction, or domain satisfaction scores) is used. In the case of people over age 70, the difference between consumption and life satisfaction is close to 1 standard deviation. Given that the choice of well-being measure is important, it is necessary to decide which measure is most appropriate to use in resource allocation decisions.
It is important that the choice of well-being measure be based on clear criteria that are relevant to the policy context. It is surprising that there has been very little consideration given to precisely what conditions a well-being measure for policy purposes should satisfy. Sumner (1996) does set out criteria for a theory of welfare of "descriptive adequacy," which, among other things, requires compatibility with widely held intuitions about well-being and shares many of the features we consider here. However, our emphasis is on the need for a measure of well-being that can actively be used for policy rather than to establish criteria for an account of well-being for philosophical dialogue.
In this paper, we consider three general criteria that any measure for policy should be evaluated against: (1) the measure must be conceptually appropriate, (2) it must be valid, and (3) it must be empirically useful. We suggest that a conceptually appropriate measure is one that is a complete measure of prudential value (that is, what is good for the individual rather than what might be considered to be the good life). An appropriate measure will measure what it purports to, but validity is problematic in the absence of a gold standard for well-being. Nonetheless, the measure should allow for comparisons across time and people and should converge with and predict things (such as health) commonly thought to be associated with well-being. A measure of well-being will be empirically useful if it is cardinal, unbiased, sensitive to changes in well-being, and practical to collect.
We evaluate the preference satisfaction account (as largely proxied by income) and the subjective-evaluation account against these criteria as measures of individual (rather than social) well-being.
Subjective evaluation is usually identified through survey questions, which vary from single questions (for example, "All things considered, how satisfied are you with your life as a whole these days?" [World Values Survey 2000]) to multiple questions (for example, Satisfaction with Life Scale [Diener et al. 1985]). Questions may also use the terminology of "happiness" (for example, "Taking all things together, would you say you are ... Very happy, Quite happy, Not very happy, Not at all happy" [World Values Survey 2000]). Responses to questions using the language of satisfaction generally correlate highly with those asking about happiness (van den Berg and Ferrer-i-Carbonell 2007; Di Tella et al. 2003; Helliwell and Putnam 2004). Therefore, we use the term "subjective evaluation" to refer all of these general measures. We do not, however, explore the advantages and disadvantages of more hedonistic measures of well-being, such as measures of affect balance or measures that aggregate daily affect (for example, Daily Reconstruction Method [Kahneman et al. 2004]).
We recognize that governments do not rely on income alone as a measure of well-being, and many social indicators are used to judge the effectiveness and the distributional consequences of government policy. However, willingness to pay (WTP) is widely used as a measure of benefit in cost-benefit analysis, and income is sometimes treated as a complete measure of value. Therefore, it is useful to compare income with life satisfaction, which is increasingly being advocated as an alternative measure of well-being for policy.
We also recognize that a measure of well-being could be used in many policy contexts: in economic evaluations such as cost-benefit or cost-effectiveness studies in (for example) health care, in research on understanding issues such as the causes of well-being throughout the life course, in measuring individual well-being and aggregate well-being for macro policy design, in monitoring distributional and equity concerns and evaluating policy initiatives aimed at distributional issues, and in international comparisons aimed at judging relative country performance and contributing to national-level policy agendas. Precisely what is required of the measure may vary according to context, but the criteria should apply in all circumstances, even if the relative importance of the criteria may vary depending upon the application of the measure.
Notwithstanding some variation according to specific contexts, our general conclusion is that income as a proxy for preference satisfaction performs no better, and sometimes worse, than life satisfaction ratings as a measure of individual well-being for public policy. In Sections 2 and 3, we set out the criteria and assess the accounts according to them before considering some of the implications of our conclusion for research and policy in Section 4.
2. THE CRITERIA
2.1. The Measure Is Conceptually Appropriate
While it may be philosophically interesting to consider the best and most complete account of well-being, policy makers require knowledge only about that part of well-being that is relevant for public policy. This means the account and the measure must include only those things that are relevant to policy—and it must exclude those things that are not relevant. In relation to the former consideration, a distinction can be made between the good life and a life that is good for the individual concerned. The concept of a good life includes values beyond what is good for the individual, such as moral, spiritual, and aesthetic concerns. The concept of well-being most suitable for public policy is that which is good for the individual, that is, prudential value. This is not to say that nonprudential values should not be of concern for government. However, for transparency and clarity, this can be in addition to well-being rather than combined in one well-being measure.
The measure should incorporate all those attributes that are seen to make someone's life better for her; hence, it is an exhaustive measure of prudential value. For example, if the measure excludes something that the individual could have, and society cares about whether she has it for her sake, then the measure may not be sufficiently complete. If an attribute that we care about people having is not picked up by the measure, and this attribute differs between members of the society, then we need to cast doubt either on our intuitions that the attribute is something we should care about or on the measure of well-being.
In relation to what might be excluded from prudential value, some sources of well-being may be illegal or socially illegitimate (Feldman 2002). A contentious issue is whether one individual's well-being is allowed (for policy purposes) to decline when another person's income or consumption rise. Harsanyi (1982, p. 56) proposed that the social welfare function should exclude antisocial preferences, such as "sadism, envy, resentment, and malice." However, one person's consumption may also be viewed as imposing an externality onto other people that they are unable to avoid (Frank 1997; Layard 2006). We may also find in12 appropriate claims on resources arising from the cultivation of expensive tastes (Cohen 1993; Rawls 1982).
Despite these concerns, we suggest that all sources of well-being should be included at the measurement stage. It is impractical to measure the extent to which an individual's well-being has been derived from illegitimate sources, even assuming those sources could be agreed upon and remain stable. Additional information on the sources of well-being or changes in well-being may still be relevant for the formulation of policy, and methods might be developed to determine these where they are considered particularly important.
Well-being may also arise from areas that society does not perceive to be appropriate areas of concern for government intervention, for example, those relating to religious beliefs. However, attributes of well-being should be excluded from a public policy measure only if they cannot be affected by any government policies (and this is unlikely to be known in advance). While it may be appropriate to exclude some attributes of well-being from being targets of government policy, the consequences of any action designed to improve well-being should be judged by its impact on well-being overall.
Therefore, an ideal measure would include all important consequences to the individual, and whether these should all be considered the remit of policy is then a separate question. Of course, a measure of well-being for public policy needs to be acceptable to policy makers and the public. A measure of prudential value is likely to be have more legitimacy when it is placed alongside (and sometimes traded off for) other objectives (for example, truth, justice, freedoms, knowledge, beauty, and so on), and this is preferable to trying to incorporate these other values within the well-being measure.
2.2. The Measure Is Valid
To be useful for policy, a measure should measure what it purports to. However, it is impossible to fully establish validity in the absence of a gold standard for measuring well-being. Nonetheless, the measure should show how an individual's well-being changes over time and how it compares to that of other individuals. Interpersonal comparison requires that measurement scales referring to different people can be meaningfully compared. Economists have been remarkably reluctant to make interpersonal comparisons (as a good example, see Robbins 1932). However, we cannot escape the fact that most policy decisions involve, either explicitly or implicitly, comparisons of the costs and benefits incurred by different people.
One way of considering whether a measure of well-being is valid is to consider the degree to which the measure is similar to (converges on) other measures to which theoretically it should be similar. If the measure is not correlated with factors that we take to be signs of an individual's well-being—such as health, material resources, facial expressions, her opinions about how her life is going, and opinions of those close to her—then this would be of concern for the measure unless a reasonable explanation could be offered. It should be noted that our intuition on sources of well-being may be incorrect, so correlations of a measure of well-being with an attribute that is commonly thought to be a source of well-being should be treated with caution. A measure has predictive validity if it can correctly predict something that we theoretically think it should be able to predict. For example, low levels of well-being may predict attempts at suicide.
2.3. The Measure Is Empirically Useful
For most policy evaluations, we should like to know something about how much well-being changes as well as whether it goes up or down; that is, the measure should be cardinal. Since Pareto ( 1971), many economists have been reluctant to think in terms of cardinal differences in well-being. However, Ng (1997) argues that people can make statements about how much more or less happy they are in one state over another. Moreover, people do not seem to find it difficult to make decisions between ways of spending their time, money, and energy on things that yield noncomparable benefits: it seems an individual can get by without having a clear idea of exactly how much additional well-being she gets from owning a cat, her relationship with her sister, eating chocolate ice cream, or having a detached garage. Griffin (1986) describes individual judgments of intrinsic reward as "roughly cardinal" and capable of distinguishing big from small differences finely enough to guide individual decisions on what is worth sacrificing or risking now for future gains.
Excerpted from Law and Happiness Copyright © 2010 by The University of Chicago. Excerpted by permission of The University of Chicago Press. All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.