Would you choose to be happy? Tradeoffs between happiness and the other dimensions of life in a large population survey
Introduction
A large body of academic literature examines the sources of subjective well-being (SWB). This literature is often based upon survey questions of respondents’ life satisfaction, and sometimes their positive or negative affect, or the sense of meaning or purpose in their lives. Respondents are also asked questions about other attributes, such as income, marital status, and employment status, so that the correlates and, ideally, causal determinants of the various components of SWB can be estimated (see Frey and Stutzer (2002), Di Tella and MacCulloch (2006), and Dolan et al. (2008) for overviews).
The research has advanced a great deal over recent years––not least because of the credibility SWB measures gained through a number of validation studies (see Diener et al. (2013) for an overview)––leading to an increasing interest amongst policymakers in using SWB measures to monitor progress and evaluate policy (Stiglitz et al., 2009, Fujiwara and Campbell, 2011, Helliwell et al., 2013). For example, the Office for National Statistics (ONS) in the UK has been measuring SWB in large general population samples since 2011. More recently, the OECD (2013) and a panel convened by the National Academy of Science in the US (National Research Council, 2013) released guidelines for the measurement of SWB.
Surprisingly, however, very little research has been undertaken examining the strength of individuals’ preference for SWB. When individuals consider tradeoffs between SWB and non-SWB aspects of their lives, how strongly do they care about SWB? Note that a regression-based estimate of the causal determinants of SWB is not equivalent to an estimate of the strength of preference for SWB (Adler, 2013). For example, an individual who is financially secure yet unhappy might experience little increased happiness as a result of additional income, but she nonetheless might still care more about getting richer than being happier. Whilst many psychologists would argue that this is a mistaken preference, economists and policymakers are greatly interested in people’s preferences and, whatever our views about the role of preferences versus experiences, we need to know much more than we do at present about how much SWB matters to people alongside (and arguably even before) establishing what causes SWB.
In thinking about the relation between preferences and SWB, it is helpful to distinguish between decision utility and experience utility (Kahneman et al., 1997). If an individual’s preferences are well-behaved, they can be represented by a utility function. An individual’s decision utility for a given situation (also referred to as preference utility (Adler, 2013)) is the numerical value assigned to that situation by her utility function. By contrast, an individual’s experience utility is a measure of the quality of her experiences.
A SWB survey provides information about experience utility. It also can be used as a strong proxy for decision (preference) utility if an individual’s preferences are focused solely upon her experiences. That is, if an individual’s overriding preference is to have a certain type of experience (such as feeling happy, experiencing a feeling of satisfaction, or experiencing a sense of purpose), then a SWB survey measuring that type of experience is a sufficient indicator of her decision utility.
But do individuals have an overriding preference for various experiences as opposed to the non-experiential components of their lives? This is an empirical question, and it is one that we directly address in this paper. Although SWB scholars sometimes assume without empirical investigation that individuals must prefer SWB over other aspects of their lives, there is nothing incoherent or irrational in having a preference for health, good relationships, liberty, accomplishment, knowledge, income, etc., in addition to a preference for various types of experiences (Nozick, 1974). Put differently, it is an empirical question whether SWB is a sufficient statistic for decision utility (Adler, 2013). Only a few studies have investigated the strength of individuals’ preferences for experiences as opposed to other aspect of their lives. This paper builds from those studies, and makes some important innovations.
In an exploratory study, Adler and Dolan (2008) ask a small group of respondents in the UK and the US to rank possible lives described in terms of income, life expectancy, health, and SWB (with SWB described specifically as the percentage of time spent in a good mood). All four components have statistically significant coefficients in a rank-ordered logistic estimation, suggesting that aspects beyond SWB matter. Health has the largest coefficient, followed by SWB.1
Benjamin et al. (2012) present respondents with pairs of “options” (possible lives), distinguished by two non-SWB dimensions. One option is higher on one of the dimensions and lower on a second, while in the second option the levels are reversed. For example, in one pairing, the two non-SWB dimensions are sleep and income, described as:
Option 1: A job paying $80,000 per year. The hours for this job are reasonable, and you would be able to get about 7.5 h of sleep on the average work night.
Option 2: A job paying $140,000 per year. However, this job requires you to go to work at unusual hours, and you would only be able to sleep around 6 h on the average work night.
For each pairing,2 respondents are asked both a choice question (“If you were limited to these two options, which do you think you would choose?”) and a predicted-SWB question, defined in terms of life satisfaction, overall happiness, and felt happiness. The authors then examine whether respondents tend to identify the same option as both preferred and predicted to produce greater SWB. They find that SWB and choice coincide 83% of the time, although the degree of convergence varies from below 50% to above 95% depending upon the non-SWB dimensions involved, as well as the study populations and question wording.3
Benjamin et al. (2014a) compile an extensive list of 136 aspects as inputs into a single index of well-being, including measures of SWB, goals and achievements, freedoms, morality, self-expression, relationships, and the well-being of others in society. Respondents are asked to choose between two options differentiated by two to six of these well-being aspects, with one option described as “much higher”, “somewhat higher” or “slightly higher” on each of the specified aspects. These responses are used to calculate marginal utilities for each of the 136 aspects. The authors find that measures of SWB and health have relatively large marginal utilities, as do family-related aspects, security, values of morality and meaning, freedom of choice and resources.
Fleurbaey and Schwandt (2015) ask a sample of US respondents to mainly think of feasible changes they could presently implement in their lives that could improve their SWB score reported in the beginning of the survey. Respondents are then additionally asked to think of feasible changes that could be implemented to again increase their SWB when presented with a specific list of life domains, including health, financial situation, family, education, etc.; and whether it is relatively easy for them to actually implement these changes. Their main result suggests that about 90% of respondents do seek to maximise their SWB.
It is critical to note at this point that the studies mentioned above focus on respondents’ anticipated change of happiness. In a recent study, Perez-Truglia (2015) directly addresses the validity of ‘currently’ stated happiness reports. Using food consumption and happiness data, the study infers preferences from happiness data via regression analysis, predicts how a utility-maximizing individual should act based on these estimates, and draws comparisons to actual behavior to find overall that, overall, actual happiness does predict choice.
Benjamin et al. (2014b) examine actual preferences of US medical graduate students’ rankings of residency programs for purposes of a matching algorithm that assigns students to programs. This is an incentive-compatible choice with substantial future career implications. For purposes of the algorithm, graduates list their four most preferred programs; in addition, they are asked by the researchers to report their anticipated SWB at each of these both during the residency period and beyond, and to rate each program based on a variety of attributes (e.g., residency prestige/status, stress, career prospects). They find large differences in the coefficients of these attributes between choice-based and anticipated SWB regressions.
This study builds on this small literature and makes two significant and related contributions: one addresses methodological concerns and the other deals with issues of preference heterogeneity. Regarding the first contribution of this study, we introduce a novel question format to test the strength of preferences for SWB. Individuals are asked for a pairwise ranking of two possible lives: one life is described as higher in some aspect of SWB, but lower in some non-SWB dimension; and vice-versa for the second life. In short, the level of SWB is directly incorporated in the life described.
Specifically, a series of pairs of lives are randomly presented to each respondent. Each pair is such that one life is higher on one of three SWB dimensions and lower on one of five non-SWB dimensions, while the second life is lower on that same SWB dimension and higher on that non-SWB dimension. The three SWB dimensions are life satisfaction, happiness, and a sense of purpose. The five non-SWB dimensions are income, physical health, family, career success, and education. The two lives are described either as brief scenarios or as vignettes. The latter offer a more extensive description of the possible lives, thus making the difference between the relatively high and low combinations of the SWB and non-SWB dimensions more salient. We ask respondents a choice (“Which life would you choose to lead?”) and/or a judgment question (“Which life is better?”).
We believe that this format provides new information into the strength of individuals’ preferences for SWB. One potential limitation of the Benjamin et al. (2012) format is that the question there posed may perhaps yield an overestimate of the degree to which individuals prefer SWB. Imagine that a survey respondent is told about two hypothetical lives, A and B, and says both (1) that she prefers life A over life B and (2) that she predicts SWB to be higher with life A than with life B. One explanation of this convergence between the choice and predicted-SWB question is that the respondent has an overriding preference for SWB, as opposed to health, relationships, income, knowledge, etc. A different explanation is that the respondent does care about the non-SWB features of her life, but her “hedonic forecasts” are such that she believes a higher level of the non-SWB features will increase her level of SWB. By incorporating SWB itself into the hypothetical lives, and presenting respondents with a clear pairwise choice between a higher level of SWB and a higher level of a non-SWB dimension, we aim to create a format that reveals the respondent’s fundamental preferences for the SWB and non-SWB aspects of life, and that minimizes the confounding effect of hedonic forecasting.4
The second contribution of the current article is to delve deeply into preference hetereogeneity. Our methodology explores systematic differences between different types of SWB. Life satisfaction is the evaluative component of SWB; happiness is the affective component; a sense of purpose is the eudaimonic component. Recent recommendations regarding the measurement of SWB suggest that SWB surveys distinguish between the three components (Dolan and Metcalfe, 2012; National Research Council, 2013, OECD, 2013). They have also been shown to have different determinants (Keyes et al., 2002, White and Dolan, 2009). Our innovation is to separate the evaluative, affective, and eudaimonic components of SWB in the context of measuring the strength of preference for SWB. If we are to say how much SWB matters relative to other concerns, we need to be clearer about the type of ‘happiness’ we are referring to.
We gather data from a large sample of nearly 13,000 UK and US respondents, which also means that we are able to consider whether the preferences in a country where SWB data are being gathered for monitoring and policy appraisal purposes (the UK) differ from those in a country where SWB arguably has less policy resonance (the US).
Our main findings are as follows. For the brief scenarios, averaging across all possible combinations of SWB and non-SWB dimensions, roughly three-fifths of both UK and US respondents prefer the life higher in SWB. The probability of the respondent preferring the low SWB/high non-SWB life increases dramatically when the non-SWB dimension is health. Overall, individuals seem to have a stronger preference for the affective component of SWB (i.e., happiness) as compared to the evaluative and eudaimonic components. We find that respondents’ own attributes have the expected relationship with their preferences: ceteris paribus, a higher attainment with respect to a non-SWB dimension is associated with an increased probability she will prefer the low-SWB life.
Section 2 offers an overview of the survey and describes the econometric approach of this study. Section 3 presents descriptive statistics and estimated results. Section 4 discusses and concludes.
Section snippets
The data
We collect data from a representative sample––in terms of race and ethnicity––of 6437 UK and 6555 US individual (Table A1 in Appendix A). Table 1 summarises the survey’s design. To assess respondents’ own SWB, we first ask the following four questions, using an 11-point scale (0–10), used by the ONS in the UK (Dolan and Metcalfe, 2012): Overall, how satisfied are you with your life nowadays? (measuring the evaluative component of SWB); Overall, how happy did you feel yesterday? and Overall, how
Descriptive statistics
The two respondent populations are fairly similar − see Table A2 in Appendix A. An analysis of SWB reports by nation suggests that the US sample reports, on average, higher levels of SWB, compared to the UK one, despite also being more anxious. The histograms presented in Appendix B suggest that this average difference is mainly due to the higher-end concentration of responses in the US.
In the UK, the overall percentage of respondents preferring the high-SWB brief scenario are 60% (61%) with
Discussion
The evidence on what causes SWB is ever increasing, but there is scant information about how much SWB matters to people in the first place. Against this general background, this study uses a sample of more than 6000 respondents in the UK and the US, respectively, to make a number of important innovations. We use a novel question format to elicit preferences for SWB. The question asks for a pairwise ranking of two lives, one higher in SWB and lower in some non-SWB attribute, the other higher in
Acknowledgements
The authors wish to thank participants at a session on ‘subjective wellbeing and utility’ at the American Economic Association annual meeting 2017, the ‘Preferences, Well-being and Discrete Choice Modeling’ workshop at the University of Antwerp, and faculty workshops/seminars at Cornell Law School, Duke University Law School, LSE’s Centre for Economic Performance, the University of Chicago Law School, as well as Dan Benjamin, Ori Heffetz, Ayse Yemiscigil, and three anonymous reviewers for
References (52)
- et al.
Do we really know what makes us happy? A review of the economic literature on the factors associated with subjective well-being
J. Econ. Psychol.
(2008) - et al.
Does happiness pay? An exploration based on panel data from Russia
J. Econ. Behav. Org.
(2004) - et al.
Does affluence impoverish the experience of parenting?
J. Exp. Soc. Psychol.
(2012) - et al.
Hedonic adaptation and the role of decision and experience utility in public policy
J. Publ. Econ.
(2008) A Samuelsonian validation test for happiness data
J. Econ. Psychol.
(2015)- et al.
Does marriage make people happy, or do happy people get married?
J. Soc. Econ.
(2006) - et al.
Choice and happiness in South Africa
Econ. Lett.
(2017) Human capital, higher education institutions, and quality of life
Reg. Sci. Urban Econ.
(2011)- Adler M.D., Dolan P. (2008) Introducing a different lives approach to the valuation of health and well-being. Institute...
Happiness surveys and public policy: what’s the use?
Duke Law J.
(2013)
Do Danes and Italians rate life satisfaction in the same way? Using vignettes to correct for individual-specific scale biases
Oxf. Bull. Econ. Stat.
Economic growth and subjective well-being: reassessing the easterlin paradox. comments and discussion
Brookings Pap. Econ. Activity
What do you think would make you happier? What do you think you would choose?
Am. Econ. Rev.
Beyond happiness and satisfaction: toward well-being indices based on stated preference
Am. Econ. Rev.
Can marginal rates of substitution be inferred from happiness data? Evidence from residency choices
Am. Econ. Rev.
The importance of impaired physical health and age in normal cognitive aging
Scand. J. Psychol.
Estimating the influence of life satisfaction and positive affect on later income using sibling fixed effects
Proc. Natl. Acad. Sci.
Some uses of happiness data in economics
J. Econ. Perspect.
Theory and validity of life satisfaction scales
Soc. Indicat. Res.
Interpretations of utility and their implications for the valuation of health
Econ. J.
Measuring subjective wellbeing: recommendations on measures for use by national governments
J. Soc. Policy
Happiness by Design
Cognitive dissonance
Sci. Am.
Do People Seek to Maximize Their Subjective Well-being? IZA Discussion Paper No. 9450
Positive emotions speed recovery from cardiovascular sequelae of negative emotions
Cognit. Emotion
What can economists learn from happiness research?
J. Econ. Lit.
Cited by (48)
Should bads be inflicted all at once, like Machiavelli said? Evidence from life-satisfaction data
2023, Journal of Economic Behavior and OrganizationIntrinsic preferences for unhappy news
2022, Journal of Economic Behavior and OrganizationOffshoring and well-being of workers
2022, Journal of Economic Behavior and OrganizationCitation Excerpt :The current study aims at contributing to this literature by studying offshoring and workers’ SWB - measured using the “life satisfaction” measure - to capture the effect of offshoring on the wider domains of well-being. Several studies report that SWB is considered to be a reliable and valid proxy of individuals’ (experienced) utility (e.g., Kahneman and Sugden, 2005; Benjamin et al., 2014; Adler et al., 2017). This measure is also found to be highly correlated with other objective and subjective measures of stress and mental health (e.g., Clark et al., 2008; Oswald and Wu, 2010).
People versus machines: The impact of being in an automatable job on Australian worker's mental health and life satisfaction
2022, Economics and Human BiologyCitation Excerpt :Diener et al., 2018). The rapidly expanding literature on life satisfaction has identified a range of factors which influences an individual’s life satisfaction, including income, education, health and unemployment (Adler et al., 2017; Das et al., 2020). These studies typically rely on survey responses regarding self-reported life satisfaction (as we do here), positive or negative affect, or the sense of purpose or meaning in one’s life (Dolan and Metcalfe, 2012).
“Better the devil you know”: Are stated preferences over health and happiness determined by how healthy and happy people are?
2022, Social Science and MedicineCitation Excerpt :Since we are asking for preferences over different possible lives for a friend, and not between a friend and someone else (e.g., a stranger), these preferences reflect a more detached view of the trade-off between PH and SWB and not nepotism or cronyism as might commonly be associated with choosing a friend over someone else. Next, following Adler et al. (2017), the survey presents one of the three “you” pair of lives at random and asks participants to rate the credibility of each hypothetical life in that pair. In order for a survey eliciting preferences between hypothetical lives to be of any value, the hypothetical lives must be considered by respondents to be meaningful.
Using memories to assess the intrapersonal comparability of wellbeing reports
2022, Journal of Economic Behavior and Organization