The Widow Who Said She Was Well: Two Problems with Happiness Surveys — Epoche B2
The problem: can we simply ask? A common belief guides much modern policy: to find out how well people's lives are going, ask them how happy they feel and add up the answers. The method looks democratic, because it lets people speak for themselves instead of letting an official decide what counts as a good life, and it is cheap, since a telephone question costs a small fraction of a clinical examination or a household audit. Yet it faces two serious problems, one empirical and one philosophical, and each demands a different solution. The empirical problem is that the single word "happiness" covers at least two distinct quantities which behave differently when measured. The philosophical problem is that what a person says about her satisfaction is shaped by what she has learnt to expect, so that the worst-treated may report the least dissatisfaction. The first problem can be repaired by better instruments. The second cannot. First problem: "happiness" is not one thing Kahneman and Deaton (2010) analysed over 450,000 American survey responses. The data came from the Gallup-Healthways Well-Being Index, which telephoned around a thousand United States residents every day through 2008 and 2009 — the size matters, because it allows the sample to be cut finely by income without the cells becoming too small to interpret. Each respondent answered two different kinds of question, and it is the pairing that makes the study decisive: the same people, at the same moment, supplied both. The first is a life evaluation : a reflective judgement about a life taken as a whole. Gallup used the Cantril ladder, which asks the respondent to picture a ladder numbered from 0 at the bottom to 10 at the top, where the top step is the best possible life for her and the bottom step the worst, and to say on which step she stands. Answering requires her to summon a standard of comparison — her ambitions, her neighbours, her earlier life — and to place herself against it. The second is emotional wellbeing : the affective quality of yesterday, obtained from yes-or-no questions about whether she experienced a great deal of enjoyment, happiness, worry, sadness, stress or anger during the previous day. This asks for a report close to experience, with no summing-up and no standard of comparison. The two came apart in a specific way. Life evaluation rose steadily with the logarithm of household income. The logarithm counts doublings rather than dollars: it advances by the same step from 25,000 to 50,000 as from 50,000 to 100,000. To say that the ladder score is a straight line in log income is to say that $$L = a + b\,\ln y,$$ where $y$ is income and $b$ is a fixed slope, so that each doubling of income adds the same amount, $b\ln 2$, to the ladder score. Two consequences follow from the form alone. First, there is no ceiling: within the range Gallup observed, the rich reported higher ladder scores than the merely comfortable by about as much as the comfortable exceeded the poor. Second, and more importantly for policy, the same absolute gain in evaluated life costs an ever-larger absolute sum of money: moving a household from 12,500 to 25,000 buys as much ladder as moving one from 200,000 to 400,000. A distributive argument is thus built into the functional form, not added to it as a value judgement. Emotional wellbeing behaved differently. It also rose with income, but the rise flattened around a household income near 75,000 US dollars a year — and since those are 2008-9 dollars, the figure is a snapshot of one country in one period and should never be quoted as a timeless threshold. Kahneman and Deaton's more robust finding is one the popular summaries drop: low income does not so much fail to buy joy as amplify pain. Among people suffering divorce, ill health or loneliness, the emotional cost of the misfortune was markedly greater for those on low incomes. Poverty, on this reading, is less an absence of pleasure than a loss of protection. Here the essay as originally written must be corrected. It stated flatly that emotional wellbeing "stops improving" beyond about 75,000 dollars. That claim has not survived. Killingsworth (2021) used a different instrument — an application that pinged 33,391 employed American adults at random moments and asked how they felt right now , gathering over 1.7 million such reports — and found experienced wellbeing continuing to rise with log income well above 75,000. Because remembering yesterday and reporting the present moment are not the same act, the two studies were not directly comparable, and the dispute could easily have hardened. Instead Killingsworth, Kahneman and Mellers (2023) conducted an adversarial collaboration: opposed researchers agreed in advance, with a neutral arbiter, on how the data would be reanalysed, so that neither side could select the cut that favoured it. The resolution was that both had captured something real. For most people, emotional wellbeing keeps rising with log income past 75,000; a flattening appears only among the least happy portion of the sample, roughly the unhappiest fifth, whose suffering has sources that money does not reach. The plateau is a feature of a minority, not of humanity. What survives intact is the essay's structural point, which the correction actually strengthens. "Adding up the answers" is impossible until we decide which answers count, and the two measures recommend different policies. If life evaluation is the target, income and rank matter, because the ladder question explicitly invites comparison. If emotional wellbeing is the target, the daily determinants of feeling matter — pain, caregiving burden, isolation, sleep — and the strongest case for transfers is that they buffer misfortune rather than that they purchase pleasure. The first solution, then, is precision: any survey must state which of these two things it measures. Second problem: the deprived may report contentment Amartya Sen's objection cuts deeper, and it is best