Consumption or income, valued at prevailing market prices, is the workhorse metric of human welfare in economic analysis; poverty is almost universally defined in these terms, and the growth of national economies measured as such. Yet for almost as long as economic analysis has utilized these measures, various shortcomings have been noted in the ability of these constructs to comprehensively capture welfare. One example – these measures can’t fully account for access to non-market goods. More famously, with Amartya Sen’s emphasis on human functionings and capabilities, these measures may not fully capture an individual’s ability to achieve and exhibit agency.
In part inspired by this view that people intrinsically value capabilities and functionings as opposed to money-metric measures per se, a burgeoning sub-field of poverty research has proposed various measures of subjective, or self-reported, well-being (SWB). SWB is widely seen as multi-dimensional and unable to be captured in only one question. Hence there are numerous approaches to the measure of SWB, most notably combinations of evaluative/cognitive approaches, such those that inquire about life satisfaction, and hedonic/affective approaches such as those asking about happiness.
I think it’s uncontroversial if I claim that the field of economics is of mixed minds about the usefulness of SWB: these measures hold some promise for comprehensive welfare assessment yet there are various interpretive challenges. I’ve blogged about some of these challenges in the past. Most concerning is the worry that salient characteristics such as gender and education, which naturally vary in any population, influence how SWB questions are understood and reported, thus complicating cross-group comparisons. Now two recent papers have made advances in the field and, taken together, highlight both the pitfalls and the promise of SWB.
This first paper, which focuses on interpretive challenges, is a working paper by Bond & Lang. This study highlights the extreme gaps in our knowledge of how study subjects interpret subjective measures and how individual respondents discretize responses in relation to a presumed continuum of underlying subjective perception. The authors show that if, as is commonly assumed, subjective welfare such as happiness exists in a continuum to be experienced, yet question structure forces respondents to discretize their happiness to fit a scale (such as very happy or somewhat happy, etc.) then much interpretive difficulty follows. It turns out that if we don’t know the underlying distribution of SWB it’s very difficult to infer whether one group is happier than another group. Bond and Lang go through cases such as the Easterlin paradox and show that cross-group conclusions based on self-reported happiness (such as poor are “happier” than the rich) may not only be nullified but actually reversed with some monotonic transformation of the underlying happiness distribution.
This challenge is summarized succinctly by Bond & Lang:
and this challenge has to be taken seriously by the literature.
One key assumption in their approach is that utility is unbounded – but when it comes to SWB, I wonder about the realism of this assumption. Utility may diminish rapidly at higher levels of wealth or income, and may be bounded from below by a threshold level. These unknowns affect the applicability of the Bond & Lang critique and I think point to a key way forward: there is still much work to do on what I previously termed the hermeneutics of subjective welfare.
We need to understand variations in frames of reference and meaning interpretation underlying responses to seemingly well-developed scales. We need to understand the degree of differential interpretation of standard questions across key characteristics such as gender and education. And we need to understand how sensible it is to model SWB as a continuum and if so whether that continuum is bounded or unbounded.
Another recent paper in the SWB literature offers a new-ish way forward on another challenge with SWB – understanding exactly what dimensions of welfare matter most to the population of interest. The authors, Benjamin, Heffetz, Kimball, and Szembrot, implement something that I have often heard previously in conversation.
To explain this, let me back up a bit – a common complaint in standard multi-dimensional welfare measures is that the choice of weights used to aggregate different dimensions can be very ad-hoc – what Martin Ravallion has termed a “mash-up”. A fundamental problem in SWB studies is the lack of observable prices that are used to proxy for marginal utilities when constructing a consumption-based measure. Thus any attempt to aggregate SWB dimensions into a broader measure has traditionally relied on a “mash-up” of dimensions with little theory to inform a sensible weighting scheme.
In response to this problem, several discerning colleagues in conversation have asked: “well, why don’t we just ask a representative sample what they actually value and what dimensions of SWB they are willing to trade-off? By doing so we can elicit shadow prices for each dimension so that we can aggregate in a sensible manner.” This is exactly what Benjamin and co-authors do. They introduce personal choice scenarios to a representative U.S. sample in order to elicit responses in a form that can be aggregated into a unified welfare metric. This metric is based on the professed marginal trade-offs of different underlying SWB dimensions.
Benjamin and co-authors ask their sample not only about trade-offs between happiness and life satisfaction, but other aspects related to achievements, freedoms, relationships, and the well-being of others. In total they assess marginal tradeoffs across 113 personal-scenario aspects of well-being. When they pool across all respondents they find:
It turns out that the top four SWB aspects in their sample, in terms of relative marginal utility, are the following:
- The overall well-being of you and your family
- The happiness of your family
- Your health
- You being a good, moral person and living according to your personal values
And the bottom four are:
- Your enjoyment of winning, competing, and facing challenges
- How high your income is compared to the income of other people around you
- Your social status
- Your power over other people
Once estimated, these marginal utilities can then be used to aggregate different SWB aspects into one index applicable to the population studied. Of course a key, and strong, assumption here is that someone’s stated preference is an unbiased measure of true preference. But right now we have to face this assumption with any analysis of elicited preference data whether with respect to risk preference, willingness to pay, or SWB dimensions. Until we have ways to validate stated preference data (perhaps with respect to biological markers of well-being such as stress indicators?) it’s hard to know how to get around this assumption.
The elicited preferences in the Benjamin paper are both for a particular population and are only valid “locally” around the status quo (i.e. only valid at one point in time). We can imagine that if a war broke out or a new disease threat arose, then the elicited preferences may be quite different. Thus it seems critically important to replicate and measure the self-assessed marginal trade-offs of different SWB dimensions in a variety of settings to better understand the degree of generalizability across populations and across key characteristics such as gender. Interestingly, Benjamin and co-authors find that gender, income, age, and political affiliation are all correlated with distinctly different rankings of certain aspects. For example women rank significantly higher “people being good moral people…” while men rank higher “people getting the rewards and punishments they deserve…”
So there is a great deal left to be done before we can put some future SWB measure on the same pedestal as consumption or income valued at market prices, yet it is that very gap that makes research exciting.