Does personality moderate reaction and adaptation to major life events? Analysis of life satisfaction and affect in an Australian national sample

Journal of Research in Personality 51: 69–77 (August 2014), DOI 10.1016/j.jrp.2014.04.009. Read as the NIH author manuscript (PMC4235663; accepted, not copyedited; PMC text mining permitted, not open access). Michigan State University. A replication and extension of Yap, Anusic & Lucas (2012), summaries/carry_on/yap2012_personality_moderation.md. Provenance: papers/carry_on/anusic2014_hilda_personality.provenance.json.

What was read

All 566 lines of the text derived from the PMC XML:

• abstract, introduction, method, results, discussion, conclusions;
• two footnotes, the funding statement, all 15 references;
• both figure captions;
• every cell of Tables 1–5.

The converter checks that the text keeps every non-whitespace character of the XML. The figures are images and were not inspected. The online supplement (full model specifications; Big Five correlations, means and SDs) was not fetched or read.

Question

• Do the trajectories of life satisfaction around marriage, a first child, widowhood and unemployment found in the British panel replicate in Australia?
• Does positive and negative affect (not measured in the British panel) change with these events?
• Do the Big Five, measured before the event, moderate either?

Method

• Data. HILDA (Household, Income and Labour Dynamics in Australia), waves 1–10: annual from 2001, all household members aged 15 and over, average annual attrition 7%.

• Samples and controls. As in Yap et al. (2012), with propensity-matched controls on age, age², sex, log household income and education.

  Event	N	Women	Age at event, M (SD)	Waves before / after
  Marriage	1,370	51.2%	31.7 (8.9)	3.8 / 4.2
  First child	1,137	51.0%	30.2 (7.1)	4.7 / 4.0
  Widowhood	298	70.8%	72.0 (11.0)	4.6 / 4.3
  Unemployment	1,438	53.5%	31.9 (14.1)	3.0 / 5.2

• Outcomes.

  • Life satisfaction, 0–10: mean 7.91, within-person SD 1.07, between-person SD 1.10 (N = 21,269). The between-person SD is the unit of change.
  • Positive affect, 1–6: four SF-36 vitality and mental-health items (full of life, calm and peaceful, lots of energy, happy person). Mean 4.07, within SD 0.65, between SD 0.78.
  • Negative affect, 1–6: five items (nervous; so down nothing could cheer you up; down; worn out; tired). Mean 2.37, within SD 0.55, between SD 0.66.
  • Both affect scales are "how much of the time during the past 4 weeks".

• Personality. Measured in 2005: 36 adjectives from Saucier's mini-markers, reduced to five scales (α .73–.80). Only people who rated personality before their event enter the moderation models: 548 marriage, 651 first child, 133 widowhood, 430 unemployment.

• Models.

  • For life satisfaction, Yap et al.'s three models: nonlinear basic; nonlinear with controls; and a linear model of baseline, event-year change and post-event change, with one trait at a time as moderator.
  • Affect did not follow nonlinear trajectories, so affect uses reaction–adaptation models: baseline up to two years before; reaction = the year before, of and after the event; adaptation = the later years. These are fitted without and with controls, and with trait moderators.

Results

Life satisfaction (Table 2; points on 0–10; * p < .05):

• Replicated from the British panel: the protective effect of marriage (people are better off than if they had stayed single, because the singles decline), and no lasting effect of a first child or unemployment once normative change is modelled.
• Not replicated: the lasting drop after widowhood. Here it is mostly normative decline (−0.06, not significant, against −0.19* in the British panel).
• People who became unemployed started lower than controls (group 0.25*; the British panel had 0.20).

Affect (Table 4; points on 1–6):

• Positive affect falls after marriage and after a first child, even against controls: −0.06* and −0.08*, or 0.08 and 0.10 SD.
• Widowhood. Positive affect is lower in the reaction period (−0.23* against controls). In the adaptation period it is higher than the controls' course (+0.14*, 0.18 SD), because the still-married controls declined more; the authors call this "clearly counterintuitive". Negative affect rises around the death (+0.16*, 0.24 SD) and does not last.
• A first child raises negative affect against controls in the long run (+0.11*, 0.17 SD), because childless controls declined.
• Unemployment: nothing lasting against controls.
• Timing differs from life satisfaction. The largest affect changes often did not fall in the event year: positive affect kept declining and negative affect stayed flat.

Personality as moderator:

• Life satisfaction (Table 3): 1 of 40 trait × change terms reaches p < .05. Agreeableness predicts a larger decline after marriage (−0.11*). This is the one moderation that replicates the British panel (−0.14* there). The authors suggest it "may be a reliable finding".
• Affect (Table 5): 12 of 80 trait × change terms reach p < .05 (my count). The authors: none are "consistent with previous findings and theories" across desirable and undesirable events. Examples:
  • openness predicts a larger drop in positive affect in the year of marriage and of a first child (−0.06* each);
  • neuroticism predicts larger declines in negative affect around and after marriage and a first child (−0.06* to −0.08*), the opposite of vulnerability;
  • after widowhood, neuroticism predicts a larger fall in positive affect (−0.13*) and extraversion a larger rise (+0.19*).
• Baseline. Personality predicts levels as expected: neuroticism lower, extraversion, agreeableness and conscientiousness higher.
• Conclusion. The trajectories mostly replicate, but "few consistent patterns of moderators emerge across these studies".

Limits

• Ten waves; affect scales are SF-36 items, not a dedicated affect instrument. The model details are only in the supplement, which was not read.
• About 120 moderation tests with no correction. With 13 significant where about 6 are expected by chance, the scattered effects are compatible with noise, which the authors half acknowledge ("advise caution").
• The widowhood sample for moderation is small (133).
• Inconsistencies found:
  • The text says neuroticism was "associated with greater long-term declines in well-being following unemployment". Table 5's only significant unemployment term for neuroticism is −0.06* on negative affect after the event, i.e. a larger decline in negative affect, which is an improvement in well-being.
  • The recap of Yap et al. (2012) lists four significant unpredicted moderations. The 2012 paper's Table 5 also has neuroticism × widowhood event-year change (+0.18*), which the 2012 authors explained by baseline (see the yap2012 summary): five in all.
  • Table 5, first child: agreeableness × post-event negative affect (+0.05*) is significant but not mentioned in the text.

What it means for Kurisutina

• Question 1: the null for traits holds in a second national panel, now with affect. Across two large samples, 1 of 40 life-satisfaction terms and 12 of 80 affect terms reached significance, scattered and often against theory. The one replicated effect (agreeableness after marriage) is small (−0.11 to −0.14 points per trait unit).
  • A slot holding Big Five scores would not tell the replica how this person reacts to marriage, a child, a death or job loss.
  • That strengthens the case for observing the person's own reactions (LISS test T1) rather than inferring them from trait contents.
• Evaluations and affect move on different clocks.
  • Life satisfaction reacts in the event year and returns.
  • Affect drifts: positive affect keeps declining after marriage and children, and negative affect rises after a child relative to the childless.
  • A replica asked "how satisfied are you" and one asked "how often did you feel down" need different models of change after the same event. The Q3 and LISS tests should score both kinds of item.
• Caveat on the pilot's X2 expectation (declared before any score, and left unchanged): it predicted larger person-specific differences on happiness than on financial satisfaction, citing Luhmann's affect-versus-evaluation split. The GSS item "Taken all together, how would you say things are these days … very happy, pretty happy, or not too happy?" is a global judgement. It is closer to life satisfaction than to the affect-frequency scales here. So the expectation's grounding is weaker than stated, and a null on X2 should not surprise.
• Normative decline is most of some "event effects". Unemployment's long-term drop and widowhood's lasting deficit here are largely the decline everyone shows. The GSS pilot's population tables and its no-event pairs play the controls' role. A replica should be credited only for change beyond what people with the same earlier answers showed anyway.
• For LISS. The Personality module has affect and mood measures alongside life satisfaction and happiness, so T1 can separate evaluation from affect, as HILDA does here.

Cross-references

• summaries/carry_on/yap2012_personality_moderation.md: the British panel study this replicates.
• summaries/carry_on/luhmann2012_adaptation.md: the meta-analytic evaluation–affect distinction.
• summaries/carry_on/infurna2016_resilience.md: heterogeneous individual trajectories after the same events.
• docs/research/gss_pilot_design.md: expectation for X2, the no-event pairs (X3).
• docs/research/liss_q1_design.md: T1 and the outcome set.