Research on subjective well-being (SWB), which refers to how people experience and evaluate their lives and specific domains and activities in their lives, has been ongoing for decades, providing new information about the human condition. During the past decade, interest in the topic among policy makers, national statistical offices, academic researchers, the media, and the public has increased markedly because of its potential for shedding light on the economic, social, and health conditions of populations and for informing policy decisions across these domains.
An impetus to this movement came from the 2009 report of the Commission on the Measurement of Economic Performance and Social Progress (Stiglitz et al., 2009), which concluded that government population surveys should be oriented toward measuring people’s well-being, including the subjective dimension, as a way of assessing societal progress. The report emphasized that traditional market-based measures alone do not provide an adequate portrayal of quality of life, and recommended shifting the focus of economic measurement from production toward people’s well-being. The underlying argument is that national policies should better balance growth in market production with considerations of equality, sustainability, and nonmarket dimensions of well-being that cannot be captured well by conventional “objective” measures.
Reflecting this interest in broadening and deepening the research base on SWB, the U.S. National Institute on Aging and the UK Economic and Social Research Council asked the National Research Council’s Committee on National Statistics to convene an expert panel to (as described in the panel charge):
review the current state of research and evaluate methods for the measurement of subjective well-being (SWB) in population surveys … offer guidance about adopting SWB measures in official government surveys to inform social and economic policies … [and] consider whether research has advanced to a point which warrants the federal government collecting data that allow aspects of the population’s SWB to be tracked and associated with changing conditions…. The study will focus on experienced well-being (e.g., reports of momentary positive and rewarding, or negative and distressing, states) and time-based approaches…. The connections between experienced well-being and life-evaluative measures will also be considered.
It should be made explicit that the panel’s interpretation of its charge was to provide guidance primarily for measurement and data collection in the area of experienced (hedonic) well-being (ExWB). While acknowledging that measurement of the multiple dimensions of SWB is essential to a full understanding of it, this focus reflects the status of research on ExWB, which is less developed than it is for evaluative well-being, another dimension of SWB. Crucially, ExWB taps somewhat different domains of psychological functioning than do measures of evaluative well-being such as those dealing with life satisfaction. Indeed, many policy concerns—for example, those related to an aging population—center around quality of life and reduction of suffering on a day-to-day basis.
SWB data have already proven valuable to researchers, who have produced insights about the emotional states and experiences of people belonging to different groups, engaged in different activities, at different points in the life course, and living day to day in different family and community structures. Research has also revealed relationships between people’s self-reported, subjectively assessed states and their behavior and decisions. In the broadest sense, the promise of information about people’s ExWB rests in its capacity to enhance measures of (1) negative experiences, particularly where they can be shown to relate to longer-term suffering of specific populations in a way that provides insights into ways to reduce them, and (2) positive experiences, in a way that informs efforts to increase or enhance them. A reasonable analogy can be drawn with poverty. The policy goal in the 1960s in the United States to reduce poverty created the need to define and measure it. These tasks have been challenging, but knowledge generated by the process—although still incomplete, even after 50 years of effort—has proven essential for policy development and assessment. In the case of SWB, if, for example, long-term unemployment, depression, or lack of income are shown to be drivers of long-term suffering, then a strong case can be made for the inclusions in one or more datasets of such measures alongside information on employment status, mental health, and income.
In this report, a range of potential ExWB data applications are cited, from cost-benefit studies of health care delivery to commuting and transportation planning, environmental valuation, outdoor recreation resource monitoring, and assessment of end-of-life treatment options. Whether used to assess the consequence of people’s situations and policies that might affect them or to explore determinants of outcomes (the impact of positive emotional states on resistance to or ability to recover from illness is now an actively researched example of the latter), contextual and covariate data are needed alongside the SWB measures.
DEFINING AND CHARACTERIZING ExWB
SWB is multifaceted and, for it to be a useful analytic construct, its components must be disentangled and understood. Evaluative well-being refers to judgments of how satisfying one’s life is; these judgments are sometimes applied to specific aspects of life, such as relationships, community, health, and work. Experienced well-being—the focus of this report—is concerned with people’s emotional states and may also include effects associated with sensations (e.g., pain, arousal) and other factors such as feelings of purpose or pointlessness that may be closely associated with emotional states and assessments of those states. The term “hedonic” is typically used to denote the narrower, emotional component of ExWB, which can be measured as a series of momentary states that take place through time. ExWB is often further divided into positive experiences, which may be characterized by terms such as joy, contentment, and happiness, and negative experiences, which may be characterized by sadness, stress, worry, pain, or suffering.
In some ways separate but also intertwined with the evaluative and experienced dimensions is eudaimonic well-being, which refers to a person’s perceptions of meaningfulness, sense of purpose, and the value of his or her life. For thinking about some questions—such as the worthwhileness of specific activities, or the role of purpose in assessments of overall satisfaction with life—eudaimonic sentiments may be relevant to both experienced and evaluative measures of well-being. The most common assessment of eudaimonia refers to individuals’ overall assessments of meaning and purpose.
In practice, a number of ExWB measurement approaches and objectives coexist, ranging from the moment-to-moment assessments of emotional states to questionnaires and interviews that require reflection by respondents about somewhat longer time periods, such as a whole day. ExWB measures can, in a sense, be viewed as a subspectrum of the overall SWB continuum that at one end involves a point-in-time reference period and is purely hedonic (“How do you feel at this moment?”) and, at the other, involves an unstated but presumably much longer temporal reference period
that is evaluative (“Taking all things together, how happy are you?”). As used in this report, ExWB includes the portion of the spectrum ranging from reports about feelings at a given moment to global-day assessments or reconstructions. Specification of the reference period has a strong impact on what will ultimately be measured. As the reference and recall periods lengthen, SWB measures take on more characteristics of life evaluation.
Unfortunately, in the literature, these temporal distinctions have often been blurred, which has led to ambiguous terminology and other confusions. “Happiness” has been used in reference to momentary emotional states and also as a way of describing overall life evaluations; such lack of specificity has at times muddled the discourse. Moreover, the multiple dimensions of well-being, such as suffering, pain, stress, contentment, excitement, purpose, and many others, cannot be ignored if investigators are to have any hope of understanding the complexities known to coexist. For example, a person who is engaged in stressful or difficult activities, such as working toward an education or a job promotion, may at the same time more broadly find meaning or satisfaction with life overall. Or a person who is chronically suffering or lacking hope may experience temporary reprieve in an enjoyable moment.
CONCLUSION 2.1: Although life evaluation, positive experience, and negative experience are not completely separable—they correlate to some extent—there is strong evidence that multiple dimensions of SWB coexist. ExWB is distinctive enough from overall life evaluation to warrant pursuing it as a separate element in surveys; their level of independence demands that they be assessed as distinct dimensions.
The ExWB dimension of SWB itself can and often needs to be parsed more finely.
CONCLUSION 2.3: Both positive and negative emotions must be accounted for in ExWB measurement, as research shows that they do not simply move in an inverse way. For example, an activity may produce both negative and positive feelings in a person, or certain individuals may be predisposed to experience both positives and negatives more strongly. Therefore, assessments of ExWB should include both positive and negative dimensions in order for meaningful inferences to be drawn.
Additionally, the observation that the aspects of negative experience—sadness, worry, stress, anger, etc.—tend to be more differentiated than those on the positive side, which are more unidimensional, carries implications for data collection.
RECOMMENDATION 2.1: When more than two ExWB questions can be accommodated on a survey, it is important to include additional ones that differentiate among negative emotions because—relative to the positive side—they are more complex and do not track in parallel (as the positive emotion questions tend to do).
At this point, empirical evidence does not indicate whether either the positive or negative ExWB dimension is more relevant to policy. But, as described in Chapter 5 and elsewhere in the report, reducing negative experiences, particularly those linked to prolonged suffering, is often a central policy objective, even if the exact levers have not been identified. To this end, development of a scale of “suffering” that has a duration dimension should be a pressing research concern. Such a measure might capture and distinguish between things like minutes of pain or stress versus ongoing poverty, hunger, and so on. Suffering is not the complete absence of happiness or the presence of exclusively negative experiences and emotions, and the scale should reflect this in a way that suggests relevant classes of policies.
To answer some kinds of questions, additional nuances beyond the positive and negative distinction are required. Thinking in terms of “experiences,” as opposed to only “emotions,” allows for consideration of an expanded set of factors—such as sense of purpose, hostility, pain, and others—which may also be important to developing a full picture of well-being.
CONCLUSION 2.4: An important part of people’s experiences may be overlooked if concepts associated with purpose and purposelessness are not included alongside hedonic ones like pleasure and pain in measures of ExWB. Crucially, central drivers of behavior may also go missing. People do many things because they are deemed purposeful or worthwhile, even if they are not especially pleasurable (e.g., reading the same story over and over again to a child, visiting a sick friend, or volunteering); they also do many things that are pleasant even if they are not viewed as having much long-term meaning in the imagined future.
When to include factors beyond the hedonic core depends on the research or policy question. For example, in studies of housing conditions or medical treatment effectiveness, sensations such as physical pain, numbness, heat, or cold, which enhance or degrade momentary experience, have an obvious relevance.
A range of techniques is available for measuring ExWB. At the short recall period end of the temporal spectrum are approaches that register emotional states in the moment.
CONCLUSION 3.1: Momentary assessment methods are often regarded as the gold standard for capturing experiential states. However, these methods have not typically been practical for general population surveys because they involve highly intensive methods that are difficult to scale up to the level of nationally representative surveys and involve considerable respondent burden, which can lead to low response rates.
For these reasons, while momentary assessment methods have proven important in research, they have not typically been in the purview of federal statistical agencies.
This conclusion reflects the current (and past) state of technology. The ways in which government agencies administer surveys is changing rapidly and, as monitoring technologies continue to evolve, new measurement opportunities will arise. For many, it may be less intrusive and burdensome to respond to a prompt from a programmed smartphone designed to sample real-time experience than to fill out a traditional survey. Use of such modes will become increasingly feasible, even for large-scale surveys, at reasonable cost.
The most frequently used alternatives, or compromises, to momentary assessment instruments are single-day measures, which involve questions asked at the end of the day or the day following the reference period (that is, about yesterday). Single-day measures have been shown to yield credible though somewhat different kinds of information about people’s daily experiences. End-of-day methods, typically used in smaller-scale studies, cannot work with surveys that rely on interviews administered throughout the day. Given these constraints for momentary assessment and end-of-day approaches, global-yesterday questions have most often been used in large surveys.
CONCLUSION 3.2: Global-yesterday measures represent a practical methodology for use in large population surveys. Data from such surveys have yielded important insights—for example, about the relationships between ExWB and income, age, health status, employment status, and other social and demographic characteristics. Research using these data has also revealed how these relationships differ from those associated with measures of evaluative well-being. Even so, there is much still to be learned about single-day measures, and it is pos-
sible that much of what has been concluded so far may end up being contested.
For some research and policy questions, contextual information about activities, specific behaviors, and proximate determinants is essential. For example, if the question is how people feel during job search activities, while undergoing medical procedures, or engaged in child care, more detailed information than can be typically ascertained from a global daily assessment is needed. Activity-based or time-use methods—such as the Day Reconstruction Method (DRM)—attempt to fill this measurement need. The DRM asks respondents to describe the day’s events by type of activity (e.g., commuting to work, having a meal, exercising) and provide a detailed rating of their emotional state during the activity. The DRM therefore goes beyond asking who is happy to asking when they are happy. This time-use dimension potentially establishes links to policy levers.
CONCLUSION 3.6: Capturing the time-use and activity details of survey respondents enhances the policy relevance of ExWB measures by embedding information about relationships between emotional states and specific activities of daily life.
The nature of the question under consideration dictates the appropriate measurement method and may suggest an appropriate data collection modality. For example, if the particular SWB dimension of interest is thought to be sensitive on a short time frame—to daily activities (e.g., going for a run) or events (e.g., a big win by one’s favorite team)—a large cross-sectional data collection conducted every 2 years is unlikely to be useful. In such cases a high-frequency approach (even if it involves much smaller samples) might be more informative, and less costly.
ADDITIONAL MEASUREMENT ISSUES
This report addresses a number of conceptual and survey methodology issues pertaining to SWB measurement; among the most crucial are
Sensitivity of measures to changing conditions, situations. A prerequisite to applying SWB data to policy is understanding what constitutes a meaningful change in a measure. In thinking about “sensitivity” and how measures are calibrated, it is instructive to consider standards applied to existing statistics. A change in the unemployment rate, for example, from 6 percent to 6.1 percent reflects a change in status of only 1 in 1,000 people in the workforce. Over the 50 years that the unemployment statistic has existed, analysts have had time to learn how to interpret what appears to be a small
change; key here is that a 0.1 percent change in the unemployment rate represents a much larger impact among the population defined as actively looking for work than for the total workforce. Similarly, it will take time to understand how to interpret SWB time series data.
Survey context, ordering, and mode effects. Although a survey methodology concern generally, question ordering and contextual factors appear to be especially serious for subjective well-being. An experimental split-sample randomized trial conducted by the UK Office for National Statistics (ONS) reported a significant question-order effect for multiple-item positive and negative affect questions: it mattered whether the positive questions or the negative questions were answered first. Deaton’s (2012) analysis of Gallup-Healthways data demonstrated the importance of the type of questions (and responses) that precede well-being assessments. Specifically, asking questions about political topics first had a substantial impact on a subsequent measure of evaluative well-being, though it had relatively little effect on the ExWB measure. Insertion of buffer questions has in some cases been shown to virtually eliminate item-order effects, suggesting that careful survey design has the potential to greatly minimize these problems. As noted in section 4.6, many of these design questions can be addressed using fairly straightforward experiments that will ultimately lead to better surveys.
Survey mode refers to how questions are posed to respondents—for example, by personal interview, telephone, or Internet instrument. Results from another split sample of the ONS survey found significantly higher life-satisfaction, happiness, and worthwhile scores (and lower anxiety scores) for telephone interviews compared to face-to face interviews, suggesting that survey mode can have a significant impact on respondent ratings.
RECOMMENDATION 4.3: Given the potential magnitude of survey-mode and contextual effects (as shown in findings related to work by the UK Office for National Statistics and elsewhere), research on the magnitude of these effects and methods for mitigating them should be a priority for statistical agencies during the process of experimentation and testing of new SWB modules.
Another important methodological issue that has arisen in the literature, discussed in sections 4.1 and 4.2, is whether respondents’ answers to SWB questions are subject to biases among groups—defined by culture, age, or other traits—that may invite misleading conclusions about actual experiences. Research has shown systematic variations in reported well-being that appear to be associated with cultural norms about ideal emotional states. Another potential threat to the validity of ExWB measures, discussed in section 4.4, is adaptation: the psychological process whereby people adjust
to and become accustomed to a positive or negative stimulus brought on by changed circumstances.
A major research challenge is to improve the knowledge base about causal pathways—both between SWB and its determinants and between SWB and various outcomes—in a way that would be suggestive of policy mechanisms. Understanding causal properties is, of course, a difficult problem in many areas of social science, not just for research on SWB. Heckman (2000, p. 91) has aptly described this general difficulty for his own discipline:
Some of the disagreement that arises in interpreting a given body of data is intrinsic to the field of economics because of the conditional nature of causal knowledge. The information in any body of data is usually too weak to eliminate competing causal explanations of the same phenomenon. There is no mechanical algorithm for producing a set of “assumption free” facts or causal estimates based on those facts.
This critique seems especially pertinent for analyses of SWB data. In many situations, it is not known whether positive and negative emotions are the predictor or outcome or if the association is reciprocal. For example, the observed association between positive emotional states and better health may be causally linked in that order, or better health may create conditions for happiness. Clearly, both can be taking place. Income and well-being could also embody this kind of circular interaction, in which distinguishing cause and effect is difficult.
The unique policy value of ExWB measures may not be in new assessments of how income does or does not relate to SWB or in an aggregate-level tracking of experiential states. Rather, their value may come from the discovery of actionable relationships for specific policies—in such diverse areas as health, city planning and neighborhood amenities, divorce and child care practices and laws, commuting infrastructure, recreation and exercise, social connectedness, and corruption—that may otherwise escape attention.
CONCLUSION 5.1: ExWB data are most relevant and valuable for informing specific, targeted policy questions, as opposed to general monitoring purposes. At this time, the panel is skeptical about the usefulness of an aggregate measure intended to track some average of an entire population.
Perhaps the most compelling reason for pursuing ExWB data collection is its potential to identify subpopulations that are suffering and to inform
research into the sources of and solutions to that suffering. Again, the panel emphasizes the necessity of measuring both experienced and evaluative dimensions of self-reported well-being. Certain policies may aim to enhance one or the other of these dimensions but may end up affecting both. For instance, an action designed to enhance day-to-day living quality at the end of life may have an impact on life satisfaction as well. And policies that aim to enhance longer-term opportunities of the young may in turn have short-term negative effects on momentary emotional experience—as in the case of a student who must work hard in school, which may at times be unpleasant, but pays off later in terms of higher life satisfaction.
CONCLUSION 5.2: To make well-informed policy decisions, data are needed on both ExWB and evaluative well-being. Considering only one or the other could lead to a distorted conception of the relationship between SWB and the issues it is capable of informing, a truncated basis for predicting peoples’ behavior and choices, and ultimately compromised policy prescriptions.
DATA COLECTION STRATEGIES
Because self-reported well-being embodies multiple dimensions and sheds light on behavior and conditions at different levels of aggregation, an ideal measurement infrastructure requires a multipronged approach.
The Measurement Ideal
One prong of a comprehensive SWB measurement program involves inclusion of modules in large-scale population surveys such as those in the ONS Integrated Household Survey and the Gallup World Poll. The repeated cross-sectional structure of such surveys allows both evaluative well-being and ExWB to be tracked. These sources are capable of identifying suffering or thriving subgroups, facilitating qualitative research for special populations, and perhaps providing useful policy information at the macro level. The Gallup data have also been used for nation-to-nation comparisons.
The second prong of a comprehensive measurement program involves inclusion of SWB questions in specialized, focused data collections. Examples include health interview surveys, time-use surveys, and neighborhood environment surveys. Question modules may be constructed as experiments or pilots within existing large survey programs (the American Time Use Survey [ATUS] module, for example, uses outgoing samples of the Current Population Survey), or they may stand alone, in which case they may be designed to include covariates shown or thought to have the strongest associations with ExWB. The advantage of targeted studies is that they
can be tailored to address specific questions—whether about health care, city planning, or airport noise management—and can sometimes be attached to ongoing surveys for which the surrounding content is appropriate. Another example is the American Housing Survey’s new Neighborhood Social Capital module; adding ExWB questions would allow researchers to explore links to community characteristics, connectedness, and resilience—associations specifically cited by Stiglitz et al. (2009) as very important and potentially alterable by policy. Because research continues to reveal details about the links between healthy emotional states and healthy physical states, health surveys provide an increasingly secure foothold for ExWB measurement. An appealing feature of smaller-scale or special-purpose surveys is that they can often be supported by funding agencies in such a way that content matches well with their organizational missions.
The third prong to an ideal data infrastructure consists of panel data collection. Information about how individuals’ SWB changes over time and in reaction to events and life circumstances cannot be fully understood without longitudinal information; such data are also crucial for addressing questions of causality (e.g., does getting married make people happier, or are happier people more likely to get married?). Krueger and Mueller (2012), for example, were able to examine the emotional impact associated with job search and other daily activities for the unemployed, both during joblessness and upon reemployment, using longitudinal time-use data. Just as panel data have allowed researchers to learn more about the characteristics of poverty (revealing less chronic poverty and more movement in and out of poverty than was once thought), panel data on ExWB may be useful to researchers studying the duration of depression and suffering at the individual level and whether these conditions tend to be chronic or if there is movement in and out of suffering states and groups. It is difficult to study such phenomena without panel data that are collected on a regular and frequent basis.
A final component of an ideal ExWB data collection strategy is real-time data collection. As described above, momentary sampling methods have been central to ExWB research but are often impractical for national statistical offices. For the immediate future, the primary means for measuring and tracking ExWB, and SWB more broadly, will continue to be survey based. Neither the technical or economic challenges to “traditional” survey methods nor the promises of alternative ways for measuring the public’s behaviors and views have reached a point where it is sensible to transition away completely from the former.
However, although real-time, momentary monitoring may not now be practical for major surveys such as the American Community Survey or the Current Population Survey, it may be (or become) a reality for a number of other surveys, particularly in the health realm. Knowing how
people are feeling and what they are doing at the same moment can shed light on the relationships between ExWB and a long list of correlates from commuting, to air pollution, to child care, with clear ties to policy. As the ways in which government agencies administer surveys change—in reaction not just to rapidly evolving technology but also to declining response rates and escalating survey costs—new measurement opportunities will arise. For government data collection to stay relevant and feasible, statistical agencies will need to apportion some of their resources to understanding and adapting to emerging survey methods, new “big data” sources, and alternative computational science methods for measuring people’s behavior, attitudes, and states of well-being.
Assessment of Current Data Collection
Very few if any national statistical offices have the resources needed to pursue data collection on all the fronts identified above as parts of the ideal strategy. At this point, some data collection modes are better understood and better supported by evidence linking them to outcomes than others; phasing in SWB data collection should reflect this. While recent research has rapidly advanced our understanding of the properties of SWB measures and their determinants, ExWB metrics are not yet ready to be published and presented as “official statistics.”
RECOMMENDATION 6.1: ExWB measurement should, at this point, still be pursued in experimental survey modules. The panel encourages inclusion of ExWB questions in a wide range of surveys so that the properties of data generated by them can be studied further; at this time, ExWB questions should only be considered for inclusion in flagship surveys on a piloted basis. Numerous unresolved methodological issues such as mode and question-order effects, question wording, and interpretation of response biases need to be better understood before a module should be considered for implementation on a permanent basis.
The United Kingdom, because of its more centralized statistical system and the opportunity raised by the current government’s interest in well-being measurement, has been able to push further than has the United States on the first prong of the comprehensive measurement infrastructure laid out above. The cautions noted above notwithstanding, it is important to recognize and commend the opportunity that the ONS initiative has provided to begin analyzing data properties, interpreting the results, and generally using it as a test bed for further development of SWB measurement. However, for the United States, the panel recommends prioritizing development of SWB
modules for inclusion in targeted, specialized surveys above the development of instruments for the large general population surveys.
RECOMMENDATION 6.2: ExWB questions or modules should be included (or should continue to be included) in surveys where a strong case for subject-matter relevance can be made—those used to address targeted questions where SWB links have been well researched and where plausible associations to important outcomes can be tested. Good candidates include the Survey of Income and Program Participation (which offers income, program participation, and care-giver links); the Health and Retirement Study (health, aging, and work transition links); the American Housing Survey’s Neighborhood Social Capital module (community amenities and social connectedness links); the Panel Study of Income Dynamics (care-giving arrangements, connectedness, and health links); the National Longitudinal Survey of Youth (understanding patterns of obesity); and the National Health Interview Survey and the National Health and Nutrition Examination Survey (health and health care links).
The ATUS modified DRM module is the most important U.S. government ExWB data collection, and its continuation would enhance SWB research. It also provides an appropriate vehicle for experiments to improve the structure of abbreviated DRM-type surveys. The ATUS SWB module is the only federal government data source of its kind—linking self-reported information on individuals’ well-being to their activities and time use. Though there are no plans to field it in 2014 (or beyond) at this point, the SWB module is practical, inexpensive, and worth continuing as a component of ATUS. Not only does the ATUS SWB module support research, it also provides additional information to help refine SWB measures that may one day be added to the body of official statistics.