Subjective Well-Being and Adaptation to Life Events: A Meta-Analysis

Journal of Personality and Social Psychology 102(3), DOI 10.1037/a0025948. Read as the NIH author manuscript (PMC3289759; accepted, not copyedited), whose title adds "on Differences Between Cognitive and Affective Well-Being". FU Berlin, Chicago, Michigan State. It stands in for Lucas et al. 2003, which could not be obtained; Lucas is a co-author and the 2003 study is among those cited. Provenance: papers/carry_on/luhmann2012_adaptation.provenance.json.

What was read

All of the manuscript text: introduction, method, results for each event, general discussion, footnotes, the appendix equations, the full reference list, the figure captions (the figures themselves are images) and Tables 1–3. The online supplement (the list of studies and effect sizes, the curves across events, full moderator estimates) was not available and not read.

Question

How does subjective well-being change around major life events, and does the change differ between:

• cognitive well-being (CWB): life satisfaction and satisfaction with specific areas such as a relationship;
• affective well-being (AWB): pleasant and unpleasant feelings and moods?

Method

• Studies. Longitudinal only: 188 publications, 313 samples, 65,911 people, 802 effect sizes.
  • The search ran in 2008 with an update in winter 2009, across six databases (2,159 publications screened).
  • 170 authors were asked for missing statistics: 135 replied and 66 supplied them.
• Eight events: marriage, divorce, bereavement, childbirth, unemployment, re-employment, retirement, relocation/migration. An event is a time-bounded change of status.
• Effect size. Mean difference from the first measurement, divided by the baseline standard deviation.
  • Deliberately not the mean change divided by the SD of change, which mixes mean change with individual variation in change.
  • No correction for measurement error. Missing retest correlations (39%) were imputed at the median, r = .48.
• Model.
  • Random-effects meta-analysis in a structural equation model, clustered by sample.
  • Log of time since the event as predictor: intercept b0 = the initial reaction, slope b1 = the adaptation rate. Log fitted better than linear by AIC/BIC for most events.
  • A CWB/AWB dummy and its interaction with time.
  • Prospective studies (baseline before the event) and post-hoc studies (baseline after it) are analysed separately.
• Moderators, one at a time: age, age², percentage of men, and whether some participants reversed the event.

Results

By event (prospective designs; effect sizes in baseline SDs):

• Moderators (significant ones only).
  • Age: older people react more positively to marriage (b4 = 0.03) and childbirth; more negatively to bereavement (b4 = −0.02).
  • Men adapt more slowly after bereavement (b9 = −0.49).
  • Marriage samples where some couples separated show more negative change.
  • Re-employment effects are more variable in younger people and more positive for men.
• CWB and AWB differ for almost every event. Most events move CWB more than AWB, and the hypothesis that AWB adapts faster held only partly. Childbirth splits them: life evaluations fall while daily affect rises.
• Desirability doesn't predict the pattern. Divorce drew a milder reaction than retirement or re-employment, and undesirable events were not adapted to more slowly. CWB falls after "positive" events and rises after "negative" ones at similar rates.
• Anticipation. The level just before an event is often not the person's usual level (above it before marriage, plausibly below it before divorce). So "decline after marriage" means a return from a pre-wedding high, not harm. The paper cannot say when adaptation is complete.
• Spread between studies. Effect sizes vary far more for AWB than for CWB. Examples: bereavement from d = −0.94 at 0.5 months to +0.52 five months later; unemployment AWB from d = −1.09 to +0.66. The authors point to personality, coping, emotion regulation and support, which drive individual differences in reaction and adaptation, and to measurement (60% of AWB measures are depression scales). Identifying "the sources of individual differences in adaptation" is named as the main research goal ahead.

Limits

• The estimates are sample averages. The paper shows that reactions differ, but individual differences are inferred, not measured person by person.
• Mostly purpose-recruited convenience samples: 72.5% mean attrition, rarely checked for selective dropout; about a third men; ethnicity mostly unreported.
• Few studies for divorce, re-employment, retirement and relocation. Publication-bias signals for marriage, childbirth and unemployment.
• 97.9% self-report; measures differ by event.
• The event and well-being can run both ways: lower life satisfaction predicts later unemployment and separation.

What it means for Kurisutina

• The pilot's events have known average reactions to check the replica against. For the GSS strata:

  • lost job: large drop in life evaluation, slow recovery (about 3 years);
  • widowed: large drop, then recovery;
  • married: short-lived rise;
  • divorced: mild reaction, then improvement.

  For the replica this is population knowledge. The empty-slot condition shows whether the base model already carries it, which it likely does from text about these events. The person's own slot has to add the individual deviation from these averages.

• Individual differences in reaction are real and large for affect, smaller for evaluations. The GSS pilot's targets split the same way: happiness is closer to affect, financial satisfaction to evaluation. The pilot should report its contrasts by item type. On this evidence, person-specific gains should be easier to find, and noisier, on happiness.

• The pre-event level may already reflect the event. In the GSS the earlier interview can fall during a divorce process or a spouse's illness. A replica loaded then is loaded with an anticipating person, and its "reaction" has to be scored against what the person was already going through. This supports the plan's multiple interviews: a replica loaded from one interview inherits whatever that moment held.

• Where the event came from matters. Life satisfaction predicts later unemployment and separation, so the people in an event stratum differ systematically from others. That justifies matching the "other" donor on the event.

Cross-references

• summaries/carry_on/gss_mr119.md: stability of GSS items; real individual change in 2006–10 concentrated in finances, jobs and marriage.
• summaries/carry_on/eckstein2022_context.md: individual updating that does not transfer across contexts; here, reactions that differ by event and by component.
• docs/research/gss_pilot_design.md: the strata and target items this informs.