Does happiness vary with age? The evidence is inconclusive. Some studies show happiness to increase with age (Diener et al. 1999; Argyle 2001). Others hold that the association is U-shaped with either highest depression rates (Mroczek and Christian 1998; Blanchflower and Oswald 2008) or highest happiness levels occurring during middle age (Easterlin 2006). Current studies suffer from two shortcomings. Firstly, they do not control for three confounding time variables: age, period and cohort effects. Secondly, all empirical research lacks a theoretical explanation as to why age affects happiness. The purpose of our analysis is to contribute to closing both of these research gaps. A social investment model frames the dynamics of happiness across the life-span. The empirical test draws on West German panel data that followed individuals from 1984 to 2005. Descriptive analysis shows a cubic age function with the lowest level at middle age. However, hierarchical three-level variance component models (Rabe-Hesketh and Skrondal 2005), find significant differences across pre-war and post-war cohorts, baby boomers and offspring of the baby bust as well as deviations during reunification. Yet, cohort and period effects account for less than 10% of the variance. (Un)happiness in midlife is more strongly determined by gender-specific occasional influences and individual characteristics. Both define objective and subjective returns of professional and personal life investments. These social investment decisions date back to early adulthood and bear a high risk of failure during midlife. Unforeseen consequences and long-term private and professional commitments make it costly to adjust, but at the same time new investments may pay off in a pro-longed future. This dilemma turns many middle-aged people into "frustrated achievers".