Accounts of wellbeing

In order for any account of wellbeing to be useful in informing policy-making, it must satisfy three general conditions. It must be: (i) theoretically rigorous; (ii) policy relevant; and (iii) empirically robust. By theoretically rigorous, we mean that the account of wellbeing is grounded in an accepted philosophical theory. This does not preclude it being subject to considerable controversy and criticism, of course. For more discussion on ‘what counts’ as theoretically rigorous, see Haybron (Reference Haybron, Eid and Larsen2008). By policy relevant, we mean that the account of wellbeing must be politically and socially acceptable, and also well understood in policy circles. Given different political, policy and stakeholder interests, no account is ever likely to receive wholesale endorsement but it must be generally acceptable to those using it and those affected by it, including the electorate. A fuller discussion of these issues is beyond the scope and focus of this paper, but the interested reader should consult Bok (Reference Bok2010).

By empirically rigorous, we mean that the account of wellbeing can be measured in a quantitative way that suggests that it is reliable and valid as an account of wellbeing. Ultimately, the measure should be sensitive to important changes in wellbeing and insensitive to spurious ones. In practice, distinguishing between the two is quite a challenge and often relies on judgement based on a priori expectations. The association between multiple measures can often provide useful guidance, where all measures moving generally in the same direction would be suggestive of important changes in wellbeing. These criteria are similar to those used by Griffin (Reference Griffin1986).

There are three accounts of wellbeing (Parfit, Reference Parfit1984; Sumner, Reference Sumner1996) that meet the three conditions (i) objective lists; (ii) preference satisfaction; and (iii) mental states (or SWB). We briefly describe the first two, which are well rehearsed, and focus more on SWB. Objective list accounts of wellbeing are based on assumptions about basic human needs and rights. In one of the best-known accounts of this approach, Sen (Reference Sen1999) argues that the fulfillment of these needs help provide people with the capabilities to ‘flourish’ as human beings. In simple terms, people can live well and flourish only if they first have enough food to eat, are free from persecution, have a security net to fall back on and so on. Thus, the aim of policy should be to provide the conditions whereby people are able to enhance their ‘capability sets’. The preference satisfaction account is closely associated with the economists’ account of wellbeing (Dolan and Peasgood, Reference Dolan and Peasgood2008). At the simplest level, ‘what is best for someone is what would best fulfil all of his desires’ (Parfit, Reference Parfit1984: 494).

SWB is a relative newcomer in terms of its relevance politically and its robustness empirically. Its theoretical rigour extends back to Bentham (Reference Bentham, Burns and Hart1996/1789) who provided an account of wellbeing that is based on pleasure and pain, and which provided the background for utilitarianism. Generally, SWB is measured by simply asking people about their happiness. In this sense, it shares the democratic aspect of preference satisfaction, in that it allows people to decide how good their life is going for them, without someone else deciding their wellbeing (Graham, Reference Graham2010).

There are some differences in the interpretation of agency in the SWB literature. Ryan and Sapp (Reference Ryan, Sapp, Gough and McGregor2007) believe that SWB concerns a person's capacity for optimal functioning, a confidence in being able to formulate and act to fulfil important goals and the motivation and energy to persist in the face of obstacles. Others believe that research in SWB reinforces particular social norms about what is desirable, which become a virtue that everyone can aspire to – but which are merely socially constructed (e.g., Edwards and Imrie, Reference Edwards and Imrie2008). It is important, however, for the different accounts of wellbeing not be conflated with one another, and that there is a clear separation between what is a determinant and what is an outcome of SWB. From the perspective of SWB, agency, social norms, virtues and individual behaviour are only important in so far as they have the ability to change the experiences of people's lives.

Of course, none of these issues would matter much in practical terms if the different accounts produced similar results – but they often do not. GDP, which is often used as a macro indicator of preference satisfaction since, all else equal, more (national) income allows us to satisfy more of our preferences, has been correlated with increases in life expectancy, an objective measure of wellbeing (Crafts, Reference Crafts2005). But there are some discrepancies. For example, GDP has also been correlated with objective measures suggestive of a lack of progress, such as increasing pollution and rising obesity (ONS, 2000, 2007). The evidence from the literature assessing the link between income and SWB is that this relationship is positive and modest (Stevenson and Wolfers, Reference Stevenson and Wolfers2008), although there are some differences depending on what measure of SWB is used (Kahneman and Deaton, Reference Kahneman and Deaton2010), and whether other people's income is included in the assessment (Luttmer, Reference Luttmer2005; Layard et al., Reference Layard, Mayraz, Nickell, Diener, Helliwell and Kahneman2010). We need to more carefully consider, even at a very general monitoring level, whether wellbeing has, in fact, gone up or down.

There are differences at the micro level too. For example, Peasgood (Reference Peasgood2008) used the British Household Panel Survey (BHPS) to examine objective wellbeing, preference satisfaction and SWB scores for the same individuals. She shows that there is a dramatic difference between the accounts for those with children, people who commute long distances, those with a degree and between men and women. Our choice of account of wellbeing could clearly have important implications for who we think of as doing well or badly in life.

Uses of the accounts

Ultimately, any account of wellbeing will be used for a specific policy purpose. We consider each of the three main policy purposes: (i) monitoring progress; (ii) informing policy design; and (iii) policy appraisal. Monitoring progress requires a frequent measure of wellbeing to determine fluctuations over time. Informing policy design requires us to measure wellbeing in different populations that may be affected by policy. Policy appraisal requires detailed measurement of wellbeing to show the costs and benefits of different allocation decisions.

Despite many unresolved questions about what should be on an objective list and how to weight the items on it, many governments and organisations have specific policies to target many of these needs (such as access to education and healthcare), suggesting that objective list accounts are an integral part of monitoring wellbeing. An example of this is the Human Development Index (HDI) (UNDP, 1990). By incorporating information on literacy rates and life expectancy (Hastings, Reference Hastings2009), the primary purpose of the HDI is to provide an alternative ranking of international development to that provided by GDP. It has been less useful in policy appraisal, where single measures of wellbeing, which weight the various components of wellbeing according to people's preferences or the impact they have on their life overall, are usually preferred (Srinivasan, Reference Srinivasan1994).

Perhaps the ‘exemplar’ monitoring account is preference satisfaction, through GDP, where rising incomes allow us to satisfy more of our preferences. According to standard theory, more choice allows us to satisfy more of our preferences and this idea has informed the design of policies in health and education. Preference satisfaction has also been used widely in policy appraisal. Cost–benefit analysis (CBA) values benefits according to people's willingness to pay (HM Treasury, Reference Treasury2003). Valuing benefits in monetary terms not only captures value in a single metric, it does so in a way that allows direct comparison to costs.

Most of the recent attempts at measuring SWB have focussed on providing information that can be used as inputs into monitoring progress. These attempts have been to generally move us away from focussing on the ‘ill-being’ reflected in rates of depression towards more positive notions of life satisfaction and positive affect (Lyubomirsky et al., Reference Lyubomirsky, King and Diener2005). Monitoring SWB could be important in ensuring that other changes that affect society do not reduce overall wellbeing. Similarities can be seen here between the current use of GDP, which is not used directly to inform policy but is monitored carefully, and sudden drops in SWB would have to be examined carefully and specific policies may then be developed to ensure it rises again.

Informing policy design requires the measurement of wellbeing in those populations affected by policy. For example, Friedli and Parsonage (Reference Friedli and Parsonage2007) cite SWB research as a primary reason for building a case for mental health promotion. More specifically, SWB could be used to make a strong case for unemployment programmes given the significant reduction in SWB associated with any periods of unemployment (Clark et al., Reference Clark, Diener, Georgellis and Lucas2004; Clark, Reference Clark, Diener, Helliwell and Kahneman2010). SWB can also be used for assessing the consequences of cigarette taxation (Gruber and Mullainatha, Reference Gruber and Mullainathan2005), air pollution (Luechinger, Reference Luechinger2009), flooding (Luechinger and Rasacky, Reference Luechinger and Raschky2009) and the risk of terrorism (Metcalfe et al., Reference Metcalfe, Powdthavee and Dolan2011). These are domains of policy that largely have a non-market consequence, and where SWB can potentially have a real impact on policy, especially as compared to preference satisfaction that focusses principally on changes in income.

Policy appraisal requires us to express the benefits of intervention in a single metric that can be compared to the costs of intervention. Using SWB data as a ‘yardstick’ could allow for the ranking of options across very different policy domains (Donovan and Halpern, Reference Donovan and Halpern2002; Dolan and White, Reference Dolan and White2007). Expected gains in SWB could be computed for different policy areas and this information could be used to decide which forms of spending will lead to the largest increases in SWB relative to their costs (Dolan and Metcalfe, Reference Dolan and Metcalfe2008). This is acknowledged in the current approaches to resource allocation from the UK government (HM Treasury, Reference Treasury2011).

It is possible to estimate monetary values for non-market goods from SWB data by estimating the amount of income that has exactly the same effect on SWB as the non-market good (Dolan et al., Reference Dolan, Fujiwara and Metcalfe2011). Alternatively, benefits can remain expressed in SWB units, as health benefits are often expressed in quality-adjusted life years (QALYs) (NICE, 2008). Indeed, the health economics literature is now showing that different quality of life weights would be derived from SWB data than from using more standard preference-based methods (Dolan and Kahneman, Reference Dolan and Kahneman2008; Dolan et al., forthcoming).

Measuring SWB

There have been many attempts to classify the different ways in which in SWB can be measured for policy purposes (Kahneman and Riis, Reference Kahneman, Riis, Huppert, Baylis and Kaverne2005; Dolan et al., Reference Dolan, Peasgood and White2006, Waldron, Reference Waldron2010). Here we distinguish between three broad categories of measure: (i) evaluation; (ii) experience; and (iii) ‘eudemonic’.

Some challenges

Before recommending any specific measures, we need to consider some key conceptual and methodological issues that apply to all three ways of measuring SWB. These are not fundamental flaws but rather issues to address when advancing any measure of wellbeing. Two distinct yet related conceptual concerns are expectations and adaptation. These lead nicely into the first of three methodological concerns: scaling. The other two methodological issues are salience and selection.

Expectations may change over time as individuals experience different life events. For instance, if governments propose policies to improve social conditions, this then becomes the new standard and effectively could become a social norm that people's SWB is based around (Duncan, Reference Duncan2005). So an improvement in social and economic conditions may not be correlated with improvements in SWB. A classic example of this is the frustrated achievers identified by Graham and Pettinato (Reference Graham and Pettinato2002). They find that upwardly mobile individuals are most likely to look beyond their original cohort for reference groups, and raise their expectations accordingly to fit their new reference group. This suggests that expectations and reference points are not static, but dynamic (Diener et al., Reference Diener, Lucas and Scollon2006).

Issues of adaptation are distinct from expectations but closely related. For instance, if a person has been endowed with a life-changing event (e.g. disability and divorce), their psychological immune system will ‘kick in’ (Gilbert et al., Reference Gilbert, Pinel, Wilson, Blumberg and Wheatley1998), and, over time, will offset some or even all of the loss in SWB. According to Kahneman and Thaler (Reference Kahneman and Thaler2005), a main reason for adaptation is attention, whereby the novelty and attention-seeking nature of many circumstances and events wear off over time. Some conditions, like unemployment it seems, continue to draw attention to them and continue to affect SWB long after the event (Lucas et al., Reference Lucas, Clark, Georgellis and Diener2003; Lucas, Reference Lucas2005).

Expectations and adaptation are also important for preference satisfaction too. For instance, once individuals expect that they will be endowed with a certain good, their value of the good changes (Shogren et al. Reference Shogren, Shin, Hayes and Kliebenstein1994). It has also been previously recognised that projects or policies are capable of changing people's preferences (Elster, Reference Elster1983; Bowles, Reference Bowles1998), and this is largely due to adaptive processes.

Adaptive processes underpin Amartya Sen's famous ‘happy slave’ example – both in terms of mental states: ‘if a starving wreck, ravished by famine, buffeted by disease, is made happy through some mental conditioning . . . the person will be seen as doing well on this mental states perspective’ (Sen, Reference Sen1985: 188) and preferences: ‘The defeated and the downtrodden come to lack the courage to desire things that others more favourably treated by society desire with easy confidence’ (Sen, Reference Sen1985: 15).

In principle, objective list accounts can ‘statically’ determine the value of different components of wellbeing. In practice, what counts as a good life is going to be dynamic too. As argued by Clark (Reference Clark2009: 34), ‘adaptation may pose serious problems for the [objective list account of wellbeing] if the relevant capabilities are to be identified through democratic or participatory techniques’. The degree to which we should allow wellbeing to vary according to expectations and adaptation are vexing moral problems and cannot be resolved here. But information on all three accounts of wellbeing can illuminate what difference expectations and adaptation make in practice and such information can be fed into the normative debate.

Expectations and adaptation feed directly into scaling effects. In order to make meaningful comparisons over time and across people, we need to understand how interpretations of the scales may change over time. Frick et al. (Reference Frick, Goebel, Schechtman, Wagner and Yitzhaki2006) show that respondents in the German Socio-Economic Panel have a tendency to move away from the endpoints of response scales over time. The relationship between earlier and later responses can be seen as an issue of scaling and salience: if later responses are influenced by earlier ones, then the earlier ones are salient at the time of the later assessment (as shown in the study by Dolan and Metcalfe, Reference Dolan and Metcalfe2010). It is possible that the interpretation of endpoints on a scale change when circumstances change and when key life events happen and it is important that we conduct more focussed empirical research into this issue. It is not at all clear, though, whether this actually matters for policy purposes since a seven out of ten before and after having children, for example, is still, in fact, seven out of ten.

On salience, any question focusses attention on something and we must be clear about where we want respondents’ attention to be directed, and where it might in fact be directed. We should like to have attention focussed on those things that will matter to the respondent when they are experiencing their lives possibly including any thoughts they may have about their health, which might matter more than the state itself (Dolan, Reference Dolan2010). It must be recognised for example, that the mere act of asking a happiness question might affect experiences (Wilson et al., Reference Wilson, Lisle, Schooler, Hodges, Klaaren and LaFleur1993). Responses will be influenced by salient cues, such as the previous question (Schwarz et al., Reference Schwarz, Strack, Kommer and Wagner1987), and perhaps also by the organisation carrying out the survey. The general consensus, however, is that there are stable and reliable patterns in happiness, even over the course of many years (Fujita and Diener, Reference Fujita and Diener2005).

On selection, who chooses to be part of a survey that contains SWB measures is important to establishing whether the effects of any factor associated with happiness are generalisable or specific to the sample population. Attrition of certain types of people in different types of happiness surveys is also important in generalising treatment effects. Watson and Wooden (Reference Watson and Wooden2004) show that people with lower life satisfaction are less likely to be involved in longitudinal surveys. Moreover, people self-select into particular circumstances that make it difficult for us to say anything meaningful about how those circumstances would affect other people. Take the effects of volunteering as an example. There is generally a positive association between volunteering and SWB but it is possible that those choosing to volunteer are those most likely to benefit from it and those with greater SWB may be those most likely to volunteer in the first place. Part of any correlation will then be picking up the causality from SWB to volunteering. Telling the chicken from the egg in SWB, as it is elsewhere, is crucial for effective policymaking.

Recommendations and conclusion

There are three main accounts of wellbeing (objective lists, preference satisfaction and subjective wellbeing), and all three accounts are important for policy purposes. There are on-going discussions about how to measure objective lists (Dean, Reference Dean2009) and how to capture preferences more effectively (Taylor-Gooby, Reference Taylor-Gooby2008; Greener and Powell, Reference Greener and Powell2009). Our paper focusses on how to measure SWB but we recognise the importance of the other accounts. Indeed, there may be significant differences between the accounts (Burchardt, Reference Burchardt2005; Anand and van Hees, Reference Anand and Van Hees2006), but better data on SWB will allow us to better highlight the synergies and tensions between the accounts of wellbeing.

In the spirit of the Stiglitz et al. (Reference Stiglitz, Sen and Fitoussi2009), who suggest measuring the different components of SWB separately, we suggest measuring each of evaluative, experience and ‘eudemonia’ separately. Table 1 provides our specific recommendations for each policy purpose. We strongly recommend: (1) routine collection of column 1; (2) collection of column 2 where possible; and (3) policy appraisal should include more detailed (e.g. time use) measures. We recommend that governments and statistical agencies ask the following questions:

1. (i) Overall, how satisfied are you with your life nowadays?

2. (ii) Overall, how happy did you feel yesterday?

3. (iii) Overall, how anxious did you feel yesterday?

4. (iv) Overall, how worthwhile are the things that you do in your life?

TABLE 1.

Recommended measures of SWB¹

Notes:

1. Reviews of the different measures can be found in Dolan et al. (Reference Dolan, Peasgood and White2006) and Waldron (Reference Waldron2010).

2. This is similar to the question used in the BHPS, GSOEP (German Socio-Economic Panel) and World Values Survey (WVS), the Latinobarometer and the recent Defra surveys. The GSOEP, WVS and Defra surveys use a 0–10 scale. Some of these surveys use a scale running from completely dissatisfied to completely satisfied, and they do not make clear where on the scale dissatisfied stops and satisfied starts. This makes it difficult to interpret the scores. Moreover, we seek consistency across the different measures of SWB, at least at the level of monitoring, and the experience measures generally calibrate the scales from ‘not at all’.

3. These are largely taken from the BHPS domains. The BHPS does not ask about satisfaction with mental wellbeing and with the wellbeing of your children, which are obvious omissions from the BHPS domains. Both of these domains are potentially important determinants of wellbeing (as distinct, in the case of children, from simply knowing whether someone has children or not). It is important to ask about general mental wellbeing and not mental health, since the latter is most likely to only pick up the negative side of the domain.

4. See Van Praag and Ferrer-i-Carbonell (Reference van Praag and Ferrer-i-Carbonell2004) for considering of the sub-domains that go into job satisfaction.

5. See for example the UK Local Authority Surveys, conducted by IpsosMORI (2004).

6. Happy and worried are the two main adjectives used in the original DRM by Kahneman et al. (Reference Kahneman, Krueger, Schkade, Schwarz and Stone2004). For the U-index developed by Kahneman et al. (Reference Kahneman, Krueger, Schkade, Schwarz and Stone2004) and Krueger et al. (Reference Krueger, Kahneman, Schkade, Schwarz and Stone2009), they measure the percent of moments spent in an unpleasant state during each activity, where an unpleasant state is defined as one where a negative emotion (sad, stress or pain) strictly dominates the positive emotions (happy). Using these two adjectives is consistent with the main headline indicators in the Gallup–Healthways data. Gallup measure the percentage of Americans who, reflecting on the day before they were surveyed, say they experienced a lot of happiness and enjoyment without a lot of stress and worry versus the percentage who say they experienced daily worry and stress far outweighing their happiness and enjoyment.

7. Some of these adjectives can be taken from the Gallup World Poll questions. We would also recommend that data using well-established measures of mental health (e.g. the PHQ9 and GAD7, which are being used to evaluate the impact of cognitive behavioural therapies) be collected periodically.

8. See Kahneman et al. (Reference Kahneman, Krueger, Schkade, Schwarz and Stone2004) and Krueger and Stone (2008).

9. See Smallword and Schooler (Reference Smallwood and Schooler2006) and Dolan (Reference Dolan2011).

10. The eudomonic measures are traditionally quite demanding in terms of the number of questions and time taken to complete (Dolan et al., Reference Dolan, Peasgood and White2006). There are no general questions about purpose and meaning in life and so we have based our recommendations on a suggestion by Felicia Huppert.

11. See White and Dolan (Reference White and Dolan2009).

Policymakers may wish to aggregate across the four questions above (and thus in the first column of Table 1) for the purposes of monitoring progress but important differences across the measures need to be made clear. For informing and appraising public policy, the questions need to be refined further, to include more domain level experiences and evaluations, and also more SWB data related to time-use (see Kahneman et al., Reference Kahneman, Krueger, Schkade, Schwarz and Stone2004).

By using the three SWB measures across countries, we will be able to provide more empirically robust data that will allow us to gain a better insight into the challenges raised in the previous section. The time has certainly come for regular measurement of SWB in the largest standard government surveys. Gathering such data will allow us to test a number of theories in economic and social sciences and policy. For example, Taylor (Reference Taylor2011) suggests that SWB is influenced by a range of choices that not all people in society are able to make. Having SWB on large surveys will allow us to test the opportunity of people to obtain higher SWB and the important objective circumstances that allow people to have higher SWB. Moreover, it will allow us to test the different measures of SWB with the different types of objective accounts of wellbeing that has not been previously conducted.

By aggregating over years, data should be available at local authority level and reliable quarterly data should be produced at the national level, especially if the survey involved overlapping panels. There are many potential surveys that could include the measures in Table 1, such as the Integrated Household Survey (IHS) in the UK (see Waldron, Reference Waldron2010, for details of the candidate surveys). Statistical agencies, academics, public policy officials, and the general public at large have a fantastic opportunity to measure SWB in ways that will enhance the monitoring of progress, and better inform the design and appraisal of social policy.