The earliest studies had data only for a few mainly developed countries. They found that richer individuals within a country were generally happier than poorer individuals, but that average degree of happiness was not greater in richer countries than in poorer ones. This led to various attempted explanations, but the most common was the claim that happiness depends on how rich a person is relative to others in the same country or region. This was supposed to be the reason why average happiness did not appear to be higher in richer than poorer countries since one’s peers also have higher incomes in richer countries. Such a relative income hypothesis was not implausible, but like many other seemingly plausible theories based on limited data, it turned out to be wrong when happiness data became available for many countries at very different stages of economic development.
Stevenson and Wolfers’ “Economic Growth and Subjective Well-Being”, published in Brookings Papers, Spring 2008, is the best discussion I have seen of happiness data for a large number of countries (I comment on their paper in the same volume). They reproduce the result found with the early data that high-income persons within a country are much happier on average than poor persons. They also find, however, contrary to earlier findings, that average degree of happiness is higher in countries with higher average per capita incomes, and that the relation between income and happiness among countries appears to be about as strong as the relation within countries.
Happiness in the United States does not appear to be exceptional in these cross-country comparisons since Americans are much happier on average than are individuals in much poorer nations. The US data show the high average degree of happiness that is appropriate to its high average incomes. Technically, what I mean is that the US falls about on the curve that bests fits the relation between happiness measures and incomes –see Stevenson and Wolfers, Figures 4 and 5. They argue, correctly I believe, that it is best, as in these figures, to compare the relation between changes in degree of happiness with percentage, not absolute, changes in levels of average incomes across countries. A $1000 increase in per capita income means a lot more to persons in poor countries than to those in rich countries.
On the whole, the data also indicate that reported average happiness tends to rise over time within a country as per capita incomes rise, or falls when per capita incomes fall, as in countries after the fall of communism. One apparent exception is the United States, where reported average happiness did not increase during the past three decades even though average incomes did. Stevenson and Wolfers provide a number of possible reasons why this occurred, including non-comparable data over time, and increasing income inequality in the US during this period. Still, this result remains something of a puzzle, given all the other evidence on the positive relation between reported happiness and income.
The big question that is not answered by their results, or by those of other economists who have used happiness data, is what happiness data tell us about wellbeing. That is, about lifetime “utility” of persons who differ by income and other characteristics? Virtually all economists who have written on happiness automatically assume that it is a quantitative measure of utility, and that this provides a way to make interpersonal comparisons of utility and wellbeing. But I argue in my comment on the Stevenson-Wolfers paper that “happiness”, even if accurately measured by these surveys, is not the same as utility or wellbeing. Rather, happiness may be an important component of utility that often, but not always, moves in the same direction as utility.
I use health as an analogy to happiness. Individuals generally get more utility when they expect to live longer and are in better health. However, they do not try to simply maximize how long they live and the quality of their health since they may trade off lower life expectancy for higher income by taking jobs with greater risks to their lives, or for pleasures of food, drink, driving fast, and other activities. The same is true for “happiness. Yes, individuals generally prefer to be happier, but sometimes they are willing to trade off happiness for other behavior that gives them greater utility.
This distinction helps explain why so many persons want to immigrate, and many have immigrated, to the United States (and other richer countries). Their lives are often very difficult for a number of years since they usually come with little money, do not have jobs, may not know English, encounter discrimination, and experience other obstacles and difficulties. They may be quite unhappy for a number of years-I have not found happiness data for immigrants- but many of them stick it out, while the unhappiest immigrants may return to the countries they came from. The immigrants who remain do not believe they made the wrong decision to immigrate, but anticipate that their lives will get better, and especially that their children will have much better opportunities in the United States than they could have had in their home countries.
I am also doubtful that the decline since 1970 in reported happiness of women compared to men in many countries is an indicator that the utility of women has declined, either absolutely or relatively. Many women who work as well as do most of the housework and childcare do not report themselves as happy, but they reveal a preference for the higher income and status that comes from working compared to staying home full time.
My conclusion is that happiness data have been useful, and the relation with income is plausible. Yet happiness data do not enable us to directly measure utility and wellbeing. I admit I do not know why average degree of happiness has not risen in recent decades in the US as incomes rose. Perhaps the considerations advanced by Stephenson and Wolfers explain why, or perhaps utility has in fact not improved over time, or perhaps more likely happiness statistics are deviating from unmeasured increases in utility.