Mental health dips for many in mid-life. Class at birth predicted who.

Cohort study. 6,992 people. distress measured at ages 26, 34 and 42; main outcome = change from age 34 to 42; predictors from birth, age 10, 26 and 34.

People were watched for years. Nobody was told what to do. A clue, not proof.

Who: 6,992 people born in Britain in one week of 1970 who answered the distress questions at 34 and 42 (51.5% female; cohort ~96% white)

• Born working classhigherodds of a dip
• Top fifth income at 34lowerodds of a dip
• Marriage or kidsnoneno clear link

More to know, from our read of the paper:

• Observational: linked, not proven. Authors stress random-forest importance is predictive, not causal
• Outcome is self-reported: 9-item Malaise Inventory (yes/no symptoms, 0-9); 'decline' = rising at least 2 points (1 SD) from 34 to 42, a threshold the authors chose (no established cut-off); 17.8% met it
• Effects are odds ratios: say 'higher/lower odds', never 'X times more likely' or '% of people'. The 28% vs 11% figures are model-predicted probabilities for extreme combined profiles (other variables at the mean), not observed group rates
• Social class at birth = father's occupation (Registrar General's classes); income = self-reported take-home pay quintile at 34
• Job security/satisfaction ranked high in the random forest but were not significant after adjustment (e.g. job security OR 0.90, 0.69-1.18); marriage and children not associated

6,992 people born in Britain in one week of 1970 answered a self-reported distress checklist at 34 and again at 42, and 17.8% got clearly worse. People born into working-class families had higher odds of this mid-life dip, and people in the top fifth of earners at 34 had lower odds. Marriage and children showed no clear link, and working-class roots did not predict a dip between 26 and 34. In the model, the chance of a dip was 28% for the least favourable mix of factors versus 11% for the most favourable. Observational: linked, not proven.

We read the abstract and the full text of this paper and checked every number on this page against it (16 facts, checked on 2026-10-29). The study design is named above. Cohort and other observational studies are described as linked, not proven. Found an error? Email theis@onestudyaday.com.