Separations of romantic relationships are experienced differently by initiators and noninitiators

PNAS 119(23): e2020901119 (June 2022), DOI 10.1073/pnas.2020901119, PMC9191343, open access (CC BY-NC-ND 4.0). Maastricht University; single author. Provenance: papers/carry_on/bruning2022_separation_initiators.provenance.json.

What was read

All 607 lines of the text derived from the PMC XML: significance statement, abstract, all sections, footnotes, data statement and all 86 references. The converter checks that the text keeps every non-whitespace character of the XML. Figures 1–6, which hold the trajectories, are images and were not inspected; numbers below are those in the text. The SI appendix (exact coefficients, tests of initiator/noninitiator differences, the extreme-split and repartnering analyses, gender splits) was not read.

Question

Divorce theory says the partner who expects to gain initiates. Do initiators and noninitiators of the same event, a separation, follow different well-being trajectories?

Method

• Data. German Family Panel (pairfam), 11 annual waves 2008–2019, cohorts born 1971–73, 1981–83 and 1991–93.
• Sample. First separations of cohabiting or married couples: 1,150 (670 women, 480 men). Only the anchor respondent is followed after a separation.
• Initiator status. "How strongly did you take the initiative to end the partnership?" on a scale of 1–5, asked at the first interview after the separation. 4–5 counts as initiator (662); 1–3 as noninitiator (488), with 3 counted as noninitiator because people over-attribute initiation to themselves. Nearly 60% of women answered 5.
• Outcomes. 12, including life satisfaction (0–10) and a depression indicator (T-STDS sum ≥ 25 of 40), plus partnership, social and economic outcomes.
• Model. Event study with individual fixed effects and cohort × sex × wave effects. Time bins run from 4.5 years before to 4.5 years after; the base is the wave just before the separation.
  • Control group: 4,937 people who stayed in a union (32,084 observations).
  • Initiators and noninitiators are estimated separately.

Results (as stated in the text)

• Life satisfaction.
  • Initiators: −0.25 before the separation. They stay lower in the short run, then rise "by even more than the initial loss".
  • Noninitiators: no pre-trend, then −0.73 immediately after. Still lower at 1.5–2.5 years; full recovery by 3.5–4.5 years.
• Depression.
  • Initiators: +5 percentage points before, back to baseline right after.
  • Noninitiators: +9.1 points right after, a 71% rise from 12.8%; then back to baseline.
• Extreme split (5 against 1; SI): noninitiators −1 point and +16 percentage points; both groups still recover fully.
• Relationships.
  • Initiators repartner faster: 30% have a partner 0.5 years after and 75% in the long run. The gap is 15 points right after and 8.5 in the long run.
  • 22.5% of initiators report infidelity in the year before, against 7.4% of noninitiators.
  • Relationship satisfaction fell about 1 point before the separation for initiators.
• Repartnering explains much of the difference. Those with a new partner right away have near-identical trajectories (life satisfaction +0.52 and +0.56). Without one, both lose, noninitiators more.
• Gender matters little beyond this.
• Economic outcomes. Initiation predicts them poorly. Both groups lose household income, driven by women; in the long run, initiators are less likely to be employed.

Limits

• Initiator status is reported after the separation, at the same interview as the first post-separation outcome. It may be coloured by how the person is doing. The paper discusses recall bias only for retrospective evaluations, not for this measure (my inference).
• Only anchors are followed, so men's and women's reports concern different relationships. Cohorts are born 1971–1993 only. The base period is the wave just before the separation, which for initiators already contains the pre-separation decline.
• Inconsistencies found:
  • The significance statement and results say initiators become "better off" and rise "by even more than the initial loss". The conclusion says their well-being shows "a rebound to initial levels". Without the figures, which of the two holds relative to 4 years earlier cannot be checked.
  • One sentence on noninitiators' career satisfaction is garbled ("Compared to the situation 3.5 to 4.5 y before … an increase of 0.29 points above their level 0.5 y before").
  • Arithmetic checked: 662 + 488 = 1,150 = 670 + 480, and 9.1 / 12.8 = 71%.

What it means for Kurisutina

• Question 1: one binary fact about the person's role splits the same event into opposite reactions.
  • Whoever wanted the separation declines before it and recovers after. Whoever was left is stable before and drops sharply after.
  • Averaged over both, "separation" looks like a moderate dip. That average is wrong for nearly everyone.
  • This is the external-control and valence appraisal of Kritzler and Haehner in concrete form, and it explains why breakup valence is the most variable within-event rating (SD 0.90).
  • A replica must know, or infer, its person's agency in an event.
• Timing follows agency. For the initiator, the change is anticipatory; for the other, it is reactive. This is Luhmann et al. (2014)'s distinction (identity change anticipates, role demands follow), shown within one event type. A replica's time course after an event should depend on whether the person chose it (inferred).
• What follows the event carries much of the effect. Repartnering nearly erases the difference (as in Krämer et al. 2024). A replica carrying on after a breakup should be scored with its later life in view, not on the breakup alone.
• Tests.
  • For scenario-based tests of "reacts like her": present the same event with agency varied ("you ended it" / "you were left"). The own-slot replica's asymmetry should match what the person reports or predicts; the empty-slot replica gives the population asymmetry. Proposal only.
  • For LISS T1: separations will look inconsistent across a person's repeated events unless their role differs systematically. Whether LISS records who initiated is unchecked; the background variables do not.
  • For the GSS pilot: the separation strata mix both roles, and nothing in the prompt says which applied.

Cross-references

• summaries/carry_on/kritzler2022_event_perception_profiles.md, summaries/carry_on/haehner2022_event_perception.md: appraisal dimensions (external control, valence).
• summaries/carry_on/luhmann2014_its_about_time.md: anticipation, identity against role demands.
• summaries/carry_on/kramer2024_multiple_events.md: separation and new partners; control for later events.
• summaries/carry_on/mancini2011_hedonic_treadmill.md: divergent trajectories after divorce.
• docs/research/liss_q1_design.md, docs/research/gss_pilot_design.md.