While it is common to estimate the age effect on happiness with parametric terms (such as age and age- squared), this paper applies a semiparametric regression model that imposes much milder restrictions on the shape of age-happiness profile. Using data from the British Household Panel Study, the paper confirmed that the age-happiness profile remained U-shaped in the semiparametric specification as well. To trace sources of the U-shape, I applied the varying coefficient model of Hastie and Tibshirani (1993), and estimated differentiated age-profiles for major life circumstances, such as differences in marital status, the presence of children at home, and whether children were dependent or not. After introducing these differentiated age effects, the shape of the overall age-happiness profile changed in a notable way: it was no longer U-shaped, but flat. In contrast, the differentiated age profiles continued to have significant effects on happiness for most life circumstances. Evidently, the common finding that happiness is U-shaped in age may reflect a composite effect from differences in people’s adjustment to major life circumstances over the life cycle.