Is the apparent U-shape of well-being over the life course a result of inappropriate use of control variables? A commentary on Blanchflower and Oswald (66: 8, 2008, 1733–1749)

Abstract

In their article in this journal "Is well-being U-shaped over the life cycle?" Blanchflower and Oswald (Blanchflower, D.G., & Oswald, A.J. (2008). Is well-being U-shaped over the life cycle? Social Science & Medicine, 66, 1733-1749) report the results of an ambitious cross-national study of age and well-being and conclude that in most societies studied the well-being of adults tends to be high in young adulthood and old age and lowest around age 40. I argue that the appearance of this U-shaped curve of well-being is the result of the use of inappropriate and questionable control variables. The most clearly inappropriate control variable is marital status, the control of which to a large extent accounts for the U-shaped curve. Most researchers who have studied the relationship between being married and being happy believe that happiness affects marital status (happier people are more likely to marry and stay married), and of course a variable that is affected by the dependent variable should not be included as a control variable in a simple recursive model. Such control variables as income and education are suspect because the relationship between them and well-being is likely to be partially spurious, and such variables as race and whether or not persons lived with both parents at age 16 should not be controlled, because they cannot affect or be affected by age. Finally, year of survey should not be controlled because of the age-period-cohort conundrum, which makes including age, period, and cohort all as predictor variables in a regression inappropriate (and impossible if the three variables are measured precisely and comparably). The only clearly appropriate control variable is birth cohort, and when only it is controlled, the regression data become estimates of how the well-being of persons has actually changed as they have gone through the life course. I argue that such estimates are much more useful than the counterfactual abstractions provided by Blanchflower and Oswald (Blanchflower, D.G., & Oswald, A.J. (2008). Is well-being U-shaped over the life cycle? Social Science & Medicine, 66, 1733-1749), and I conclude that those authors (or someone else) could make a very important contribution by redoing their analyses with birth cohort as the only control variable. I do that with the American happiness data and find that the results do not come close to the U-shaped pattern.