Personality prior to disability determines adaptation: agreeable individuals recover lost life satisfaction faster and more completely

Psychological Science 22(11): 1397–1402, DOI 10.1177/0956797611421790. Paris School of Economics and University of Manchester. Read as the accepted manuscript ("Accepted for Publication at: Psychological Science") posted on the first author's own website (journeyforhappiness.co.uk, publication list). The published version was not read and may differ in wording. Provenance: papers/carry_on/boyce2011_disability_agreeableness.provenance.json.

What was read

All 694 lines of pdftotext -layout output (16 pages): abstract, introduction, method, results, discussion, references, Table 1 (both regressions, both steps) and Figure 1 (rendered and read; its values are plot readings, "(fig)").

Question

Does personality measured before a disability predict how life satisfaction reacts to it and recovers afterwards?

Method

• Data. SOEP, 2004–2009.
  • 11,680 people with personality and life satisfaction in every year.
  • Of these, 307 became disabled between 2005 and 2009 (162 men; age 17–86, mean 56.8) and stayed disabled for at least two years. Data after any recovery were dropped.
  • That gives 1,479 person-years.
• Disability. Officially certified reduced capacity to work, or severe handicap (independent medical assessment).
• Measures.
  • Life satisfaction: a single 0–10 item, standardized (mean 6.23, SD 2.03).
  • Personality: the SOEP 15-item BFI-S, standardized. The text dates it to 2004 in one place and 2005 in another; see Limits.
• Model. Multilevel growth of life satisfaction on years disabled (linear and squared), controlling for 2004 life satisfaction, with random slopes. Then personality × years-disabled interactions.
  • Regression 2 adds age, gender, education, marital status and pre-disability income, with their interactions, plus current income and employment.

Results

• Average course. Years disabled b = −0.20, squared b = 0.04, both p < .01.

  • Life satisfaction falls about 0.16 SD in year 1 and bottoms around −0.23 SD in years 2–3, with partial recovery to about −0.13 SD by year 4 (fig).
  • No personality trait predicts the level once 2004 life satisfaction is controlled.

• Agreeableness moderates recovery.

  • Adding the ten interactions improves fit (χ²(10) = 19.25, p = .037 by my computation).
  • Agreeableness × years b = −0.10, squared b = 0.03 (β ±.20, p < .05). It holds in Regression 2 with demographic, income and employment controls (−0.11, 0.03).

  Group	Year 0	Year 2	Year 4
  +1 SD agreeable (fig)	≈ +0.05	≈ −0.26	≈ +0.05 (back to start)
  −1 SD agreeable (fig)	≈ −0.05	≈ −0.20	≈ −0.27, still falling

  The gap at year 4 is 0.32 SD.

  • Both groups drop about equally in the first two years. The difference is in what happens after.

• Neuroticism also interacted in Regression 1 (b = 0.10 and −0.02) but not with controls or in the robustness checks. It was set aside as unstable.

• Robustness of the agreeableness interaction:

  • years 0–3 only: −0.16 and 0.06;
  • all 11,373 never-disabled people added as controls: −0.16 and 0.06;
  • 2003 life satisfaction as the baseline: −0.10 and 0.03, all p < .05.

Limits

• A single-item outcome.
• 307 people, with thin data at year 4: only those disabled in 2005 can reach it. The "complete adaptation" rests on that last point, with SE bars of about ±0.09 (fig). The years 0–3 check keeps the interaction but not that claim.
• Ten interactions were tested, and one (neuroticism) was dropped as unstable. There is no correction for testing five traits.
• Only people who stay disabled; recoveries are censored.
• Inconsistencies found:
  • When personality was measured.
    • The Participants section says "personality measures in 2004". Measures says "2005 pre-disability personality levels". Table 1 labels personality "T = 1", with 2004 as T = 0.
    • The SOEP BFI-S was first fielded in 2005 (consistent with the Specht 2011 summary). So 2005 is right.
    • Disability onsets start in 2005, so for people disabled in 2005 personality was measured in the onset year, not strictly before it. The "prior to disability" claim holds fully only for 2006–2009 onsets.
  • Plotted curves against printed coefficients. The printed Regression 1 coefficients give a year-4 gap of 0.26 SD and +1 SD agreeable at −0.03 (my computation, including the main effect 0.05). The figure shows 0.32 and about +0.05. The difference is within what the two-decimal rounding of the squared-term coefficient can produce.
• Checked:
  • 11,680 − 307 = 11,373 controls.
  • χ²(10) = 19.25 gives p = .037, consistent with "p < .05".
  • The average-course coefficients reproduce the left panel of Figure 1 to within about 0.03.
  • The 0.32 SD gap matches the plot.

What it means for Kurisutina

• Q1: a trait measured before the event changes the course after it, not the first hit. Agreeable and disagreeable people fell alike for two years. Then the agreeable recovered and the disagreeable kept declining.
  • "Reacting like her" is not a fixed offset. It is a different trajectory shape, and it only shows after about two years (verified).
  • With Haehner 2024, the picture is:
    • appraisal and concrete facts set the depth of the dip;
    • here, a trait sets whether recovery happens;
    • most people recover on a common fast-then-slow course.
• This is a positive moderation result, but a weak one (one trait out of five, n = 307, the key point at year 4). It fits the earlier caution that trait moderation findings are few and fragile (Anusic 2014 found little moderation in HILDA). A replica should not hard-code "agreeable people recover" (inferred).
• For LISS T1 (proposal). Disability, or health-limitation onset, is a candidate event type.
  • LISS has personality (Big Five in the personality module) measured before events. The same interaction can be declared as a secondary analysis before any outcome is loaded.
  • It needs at least three post-event waves, because the effect appears late.
• For the GSS pilot. The pilot's strata and 2-year GSS spacing sample one post-event point. This paper shows that person differences can be invisible at year 1–2 and large at year 4, so a null pilot result on reaction differences would not rule them out (inferred).

Cross-references

• summaries/carry_on/boyce2010_conscientiousness_unemployment.md: the same authors' unemployment moderation.
• summaries/carry_on/anusic2014_hilda_personality.md: little personality moderation in HILDA.
• summaries/carry_on/haehner2024_event_characteristics_swb.md: event characteristics and individual change.
• summaries/carry_on/specht2011_personality_life_events.md: SOEP BFI-S fielded in 2005.
• summaries/carry_on/luhmann2012_adaptation.md: average adaptation to disability and other events.
• docs/research/liss_q1_design.md: the T1 design.