Do they think that joy and misery are temporary? Comparing trajectories of current and predicted life satisfaction across life events

ABSTRACT

When people marry or have their first baby, their life satisfaction spikes but fades soon thereafter. Similarly, when people separate from a partner or become widowed, their satisfaction decreases but largely recovers within a few years. I analyse to what degree people who experience these life events expect that their current shift towards joy or misery is temporary vs. long-lasting. Data are from the German Socio-Economic Panel (SOEP) and are analysed with a novel estimation strategy that uses month-specific information on the timing of interviews and life events. I compare trajectories of people’s current life satisfaction and anticipated life satisfaction, i.e. the level of satisfaction they expect to have five years into the future. Results show that people expect some of these life events’ joy or misery to stay, but the majority to fade. These findings reveal that, to a substantial degree, people are able to see their situation from a distance and separate their current feelings from their long-term life prospects. This may contribute to people’s overall well-being and sustainable life strategies.

KEYWORDS:

• Life satisfaction
• Life events
• Set-point theory
• Panel data
• Smoothed monthly estimates
• Well-being
• Anticipated life satisfaction

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 This is how the ‘control’ groups are defined for each life event. Marriage: people who are not married, divorced, or widowed (n_individuals = 12,267; n_observations = 63,129). Transition to parenthood: people who have no children (n_individuals = 13,406; n_observations = 81,942). Separation: people who have a partner living in the household (n_individuals = 30,748; n_observations = 208,932). Death of partner: people who have a partner living in the household (n_individuals = 32,609; n_observations = 222,802): This ‘control’ sample also includes people-years that are at least five years before the event. The underlying assumption is that, for these observations, estimates of age- and period-coefficients are not subject to anticipatory effects of the life event.

2 The results figure shows this period (36 months before to 60 months after the transition to parenthood), but these values are calculated from a slightly larger analytical sample, which covers this period plus the bandwidth in both directions (from 42 [36 + 6] months before the event to 66 [60 + 6] months after). This reduces the noise in the estimation of the endpoints (36 months before and 60 months after the event).

For smoothing, Stata’s lpoly-command is used, which also estimates standard errors automatically. Standard errors are estimated by first fitting a higher order local model with a different (1.5 times larger) bandwidth at each point (month) to calculate the variance of the estimates. Standard errors are then derived by taking the square root of these variances, providing a measure of the variability in the smoothed values at each specific point.

3 I calculate the average of the predicted margins for the period from 36 to 18 months before the event; then, I deduct this average value from all monthly predicted margins.