Figure 9. Alternative Definitions of Married and Single.

Note: Solid lines show within-groups estimated marriage health gaps β(a) from Equation (1) for different definitions of married and unmarried populations: excluding divorced/separated or widowed from the sample, and including cohabitants in the married group, as indicated. The regression is fitted to the PSID. Control variables include female, black, and college dummies, income, dummies for 0–3, 4–12, and 13–18 year-old children at home, and year of birth dummies; regressions also estimate α(a). The horizontal axis indicates age. In estimation, five-year age bins (20–24, through 60–64) are considered. The center point of the bin is represented in the figure. Weights are used in estimation. Confidence intervals of ± two standard errors are computed clustering at the household level.