Table 7.2. Summary of Methodological Issues with Subjective Well-Being Measures
StatLink 2 http://dx.doi.org/10.1787/888933839772.
Complementary approaches to survey questions have been, or are being, developed. Several approaches exist, or are being developed, that can complement survey questions. For example, big data is being used to “nowcast” subjective well-being (see sidebar, “Use of Big Data to ‘Nowcast’ Subjective Well-Being”), and new approaches are being developed to elicit people’s preferences, including information on how they value trade-offs between competing goals. “Automatic data capture,” which combines data such as GPS measurement or continuous health measures with survey questions on subjective well-being, is another example of these innovative approaches.
AGING AND SUBJECTIVE WELL-BEING
An intriguing line of research pursued by psychologists, sociologists, and economists is whether and how subjective well-being shifts with age. There are now dozens of articles examining this question. In short, the pattern is that evaluative well-being (as measured by instruments gauging life satisfaction, such as the Cantril ladder) is relatively high in an individual’s 20s, falls to its lowest point in their late 40s and early 50s, and then improves to the highest levels in their 70s (although there may be declines in older age; see Figure 7.1). This pattern holds for English-speaking, wealthy countries, but not for poorer, non-Western countries (Steptoe, Deaton, and Stone, 2015). There is also evidence that the pattern is not attributable to cohort effects, which could have explained the pattern by different cohorts of individuals experiencing various historical events (Blanchflower and Oswald, 2008). Less is known about how experiential subjective well-being changes with people’s age, but at least in the United States patterns of different affects are not U-shaped. For example, Stone et al. (2010) found that stress is high from age 20 through to about age 50, followed by a rapid decline through the 70s (the right side of this pattern is consistent with the evaluative well-being pattern of improving outcomes in older age). What is surprising about these patterns, at least for Western countries, is that the improvement in subjective well-being occurs at an age when the prevalence of chronic disease is on the increase, and that the presence of illness is associated with lower subjective well-being. Theoretical explanations, with some empirical support, focus on a shift in priorities and in social engagements and time use in older people, resulting in higher well-being.
Figure 7.1. Life Evaluations and People’s Age Across the World
Note: Mean life evaluations, plotted by 4-year age groups. The continuous line represents unadjusted data, while the dotted line represents data adjusted for 4 covariates (the share of the population who are women living with a partner, with a child at home, and unemployed).
Source: Stone, A.A. et al. (2010), “A snapshot of the age distribution of psychological well-being in the United States,” Proceedings of the National Academy of Sciences of the United States of America, Vol. 107(22), pp. 9985–9990. StatLink 2 http://dx.doi.org/10.1787/888933839791.
USE OF BIG DATA TO “NOWCAST” SUBJECTIVE WELL-BEING
Big data offer opportunities to complement NSO-generated measures of subjective well-being, though there are significant cautions that must also be considered. Subjective well-being measures derived from Big Data can potentially provide more timely estimates, high-frequency information, local level data, and early warning signals. Big Data are also multi-dimensional—Google search queries (a form of revealed behavior), for example, can be used to cover a wide range of states, such as pain. The wealth of big data means that the impact of various shocks (e.g., the impact of the financial crisis across and within US cities) can be investigated in a timely fashion.
There are, however, challenges. Much big data suffers from selection bias and noise, and disaggregating results by different population groups can be problematic. A particularly important problem is that it is difficult to credibly infer the intent of people who contribute data points: researchers have no way of validating their assumptions about why and when people tweet or search for particular terms on Google (which may also be considered selection factors). For this reason, interpretive exercises carried out using big data sources should be approached with caution; research focusing on subjective well-being constructs measured with big data as compared with more traditional data collection methods could illuminate our understanding of these selection factors.
Substantive Progress: New Knowledge About Subjective Well-Being
The Global Picture
Setting aside the measurement issues outlined above, which may impact the substance of what follows, Gallup World Poll data show that evaluative subjective well-being is highest, on average, in the Nordic countries, Switzerland, the Netherlands, Canada, New Zealand, and Australia; and lowest in the poorer countries of sub-Saharan Africa, and in countries experiencing war, such as Syria and Afghanistan (see sidebar, “Subjective Well-Being Across the World”). Changes in average reported life evaluation across countries since the SSF 2009 report also provide evidence that subjective well-being indicators are useful measures of social progress. The biggest drop in average life evaluations between 2005–07 and 2012–14 was in Greece, followed by Egypt and Italy (Helliwell, Layard, and Sachs, 2015)—although this analysis excludes 5 of the world’s current 10 lowest-ranking countries, including Syria and Afghanistan, due to lack of data in earlier waves of the Gallup World Poll.
SUBJECTIVE WELL-BEING ACROSS THE WORLD
New evidence from the Gallup World Poll shows that the positive relationship between evaluative well-being and income runs throughout the range of GDP per capita, from the poorest to the richest countries. This pattern was not easily seen from the World Values Survey, which does not include the really poor countries in Africa, and had to wait until the Gallup World Poll came along to become firmly established. Prior to that, many researchers had observed that, when plotted against per capita GDP, countries’ evaluative well-being flattened out at some point so that, at income levels above those of a country like Morocco, higher average income did not lead to higher evaluative well-being. This led many researchers in the field to conclude that income did not matter once countries were no longer poor. In his early papers, Easterlin used this as evidence that income did not improve the human lot, at least once we were not dealing with real poverty anymore (Easterlin, 1974).
Using the Gallup World Poll, Deaton (2008) showed that this conclusion was wrong. When life evaluation is plotted against the logarithm of income, the result is very close to a straight line. There is certainly diminishing marginal utility with respect to GDP per capita, but doubling income has the same effect at the bottom as at the top of the income scale, though the absolute changes in income are much smaller at the bottom than at the top (Figure 7.2). Of course, there is a lot of scatter, so that if you do not take the whole range, the relationship is much less obvious. For example, using plots limited to rich countries, an observer might conclude that there is not much of a relation there.
The variability in the association is also important, because it shows that countries are not trapped by their level of GDP per capita. Some countries do much worse than others, some much better. One interpretation is that these deviations indicate the policy space: governments can promote high subjective well-being even with limited resources, and countries like the United States can have relatively poor outcomes in terms of subjective well-being, even when they are very rich in terms of GDP per capita. Many factors besides GDP per capita determine people’s subjective well-being levels, from employment status to health, and from environmental quality to social relationships. Nevertheless, some of the international variation may also be due to reporting styles. It is certainly plausible that people with different cultures use the scale differently. For example, countries of the former Soviet Union are way below the line, but first, there are good reasons for this, and second, income measures for these countries are likely to be affected by large measurement
errors, so perhaps there is no need to invoke “Slavic dourness.” In general, we should also remember that in these global comparisons involving per capita GDP measured in purchasing power parity (PPP) terms, there is huge uncertainty about GDP measures; so not everything has to be attributed to oddities in life evaluation measures.
Figure 7.2. Log GDP per Capita Is Associated with Life Evaluations Worldwide
Note: N = 107 countries and territories. Pooled observations, 2009–13.
Source: Gallup World Poll and World Bank World Development data; Exton, C., C. Smith, and D. Vandendriessche (2015), “Comparing happiness across the world: Does culture matter?,” OECD Statistics Working Papers, 2015/04, OECD Publishing, Paris. http://dx.doi.org/10.1787/jrqppzd9bs2-en.
Another remarkable finding from the Gallup World Poll data is that experiential well-being measures are much less associated with per capita GDP than evaluative well-being, such as that measured using the Cantril ladder. At the country average level, there is only a small positive correlation between positive emotions (the sum of smiling/laughing, enjoyment, and feeling well rested a lot “yesterday”) and log GDP per capita, while negative emotions (feeling anger, worry, and sadness a lot “yesterday”) show essentially no relationship (Exton, Smith, and Vandendriessche, 2015). People in some African countries report as many instances of positive emotions yesterday as do people in much richer countries. If, as argued by Benthamite hedonistic utilitarians, the purpose of policy should be to maximize experiential happiness, then Gallup World Poll data would imply that Kuwait, Trinidad and Tobago, and Paraguay should be giving aid to Syria, Iraq, and Armenia, at least if aid improves happiness. Stress (for which Philippines is champion, and the United States is near the top), worry (Iraq), and anger (Algeria, Iran, Iraq, Turkey) are also not strongly related to income, while pain is highest in the Middle East. So higher national income tends to come with higher life evaluation, but does little to improve the emotional lives of citizens.
Despite these findings, the association between money income and subjective well-being is not yet settled. Stevenson and Wolfers (2012), for example, claim that the derivative of the Cantril ladder–based evaluative measure with respect to the log of income is around 0.30, pretty much no matter where you look over time.
Other findings from Gallup data show that the U-shape between life evaluation and age is not universal across countries, or at least across regions of the world (Figure 7.3). It is no puzzle that the ladder falls with age in the former Soviet Union countries, given that the elderly experienced the greatest losses from the transition away from a planned economy. But there are also areas of the world, like sub-Saharan Africa, where the relation between life evaluation and age is flat; others, like Latin America and southern Europe, where life evaluations fall with age; and some, like China, that share the English-speaking U-pattern. Of course, these are cross-sectional results, but it is not clear how to reconcile them with a universal U that is biological, based on evidence from primates, as has been argued (Weiss et al., 2012). The life-cycle patterns of experiential well-being are more uniform across the world, with a lot of what originally reported in Stone et al. (2010) showing up in many places. Negative emotions really do seem to become less prevalent with age around the world. This should perhaps replace the U-shape in evaluative well-being as the new stylized fact.
Figure 7.3. Life Evaluations and Age in Four World Regions
Source: Steptoe, A., A. Deaton, and A.A. Stone (2015), “Subjective wellbeing, health, and ageing,” The Lancet, Vol. 385(9968), pp. 640–648. StatLink 2 http://dx.doi.org/10.1787/888933839810.
Correlates and Determinants of Subjective Well-Being
At an individual level, there is a growing consensus around the factors that are correlated to higher life satisfaction: being employed and having higher income, better health, and stronger relationships are among the most important factors (see, for example, Eurostat’s “Analytical Report on Subjective Well-Being,” published in 2016). A large number of other correlates has been identified in some data sets, such as environmental conditions and pollution (Silva, de Keulenaer, and Johnstone, 2012). See the sidebar above for a partial list of the research findings on correlates of subjective well-being.
There is also new work on the importance of childhood as a critical period for later subjective well-being. Children’s emotional health is the largest predictor of adult life satisfaction, above cognitive skills (Layard et al., 2014; OECD, 2015b). This indicates that while children’s subjective well-being is important in and of itself, it matters also because it is likely to be a driver of adolescent and adult outcomes (such as adult life satisfaction, employment, or school achievement). Children’s subjective well-being and emotional health are, in turn, correlated with a variety of family characteristics such as financial difficulties, family structure, moving to a different residence, and the quality of the parent relationship. One way that this research might help policy-makers would be to better understand why some children are resilient to detrimental circumstances or events, but others are not, and what the implications for public service investments are. Eurostat is currently developing a module for EU-SILC on children’s health and material well-being that will likely be collected every 3 years in the future. When these data become available, they will provide many opportunities for deeper analysis on this topic.
Levels of subjective well-being are not only determined by the events that make people better or worse off, but also by the degree to which they “bounce back” after such events. Resilience is a concept closely related to that of (true) adaptation, and several papers in recent years have added to this literature. The picture that is painted by these studies is, however, mixed: there is evidence that life satisfaction adapts to some life events (such as marriage or childbirth), but less so to others (such as disability, entry into poverty, international migration, or unemployment; see, for example, Lucas, 2007; Clark et al., 2008; Oswald and Powdthavee, 2008; Frijters, Johnston, and Shields, 2011; Clark and Georgellis, 2013; Clark, D’Ambrosio, and Ghislandi, 2016; and Helliwell, Bonikowska, and Shiplett, 2016). One potential explanation is that people adapt more to positive life events than to negative life events—which may point to a relationship between loss aversion and adaptation. Again, we point out the importance of specifying which type of subjective well-being is being assessed: this is because experiential and evaluative subjective well-being are likely to have different patterns of adaptation.
Progress has also been made in analyzing subjective well-being not only as an outcome, but also as a predictor or, in the framework of a production function, as an input to other life outcomes. Steptoe, Deaton, and Stone (2015) have shown that, for example, there is evidence that the three components of subjective well-being predict individual morbidity and mortality even when controlling for a wide variety of individual characteristics. Similarly, a growing body of research has supported the idea that meaning and purpose in life (i.e., eudaemonic well-being) is associated with health and mortality. For example, recent findings from the Midlife in the United States study demonstrate that subjective well-being is linked to the metabolic syndrome, a group of factors that raises the risk for heart disease and other health problems such as diabetes and stroke (Boylan and Ryff, 2015). However, recent data from the UK Million Women study show that ratings of “happiness” (admittedly an ambiguous construct from the multi-dimensional subjective well-being perspective advanced here—see “What Are Subjective Well-Being Measures?”) are not linked to mortality when personal and health conditions at the first assessment point were considered in the regressions (Liu et al., 2016). Although the study omitted men, it could have been more refined in the subjective well-being measures employed, and it may have over-controlled co-varying health status, its conclusions challenge the prevailing sentiment about this issue.
What Do People Mean When They Say They Are “Satisfied with Their Life”?
To know what is behind these measures, and in parti
cular to help understand how different measures relate to each other and might be combined, it is very important to understand what people mean when they say that they are satisfied with their lives, and how they weigh different well-being outcomes—that is, what matters for people’s subjective well-being.
The correlates of the three types of subjective well-being present a generally coherent picture and provide predictive evidence that the measures are performing as expected. Nevertheless, a strong case can be made for more deeply understanding the origins of the ratings—that is, how people are generating them. In this regard, most work has been done on evaluative measures like life satisfaction, where there have been investigations into the aspects of life that bear the most on judgments of life satisfaction by simply regressing overall satisfaction measures on ratings of specific domains of life such as work satisfaction, partner satisfaction, social satisfaction, and so on (Helliwell, Layard, and Sachs, 2016, special section). However, simple, atheoretical analytic approaches may result in misleading conclusions and, importantly, lead to incorrect policy inferences. More recently, an econometric approach to decomposing global life satisfaction using stated preferences has been proposed and tested by Benjamin et al. (2014). This study has shown that people’s decisions about the future are based on a complex weighting of ratings of anticipated well-being in several imagined outcomes. In another approach to the question, new work is ongoing employing traditional qualitative methods to understand the thought processes associated with making life satisfaction ratings—for example, via the use of “think aloud” techniques (Broderick et al., 2016).
For Good Measure Page 29