Cambridge-Somerville Youth Study

Summary

The Cambridge-Somerville Youth Study is arguably the most consequential single experiment in the history of social-policy evaluation, and not because it succeeded. Initiated by Harvard physician Richard Clark Cabot in 1937, it was designed to test whether a generous, well-funded prevention program could redirect the lives of at-risk boys. Cabot died in 1939 before learning the answer. Three decades later, Joan McCord — using death records, court records, and survey instruments — tracked down 506 of the original participants and reported the result: men in the treatment arm were *worse off* on essentially every measured dimension. The harm was small but consistent, statistically significant on multiple outcomes, and impossible to reconcile with the intuitive theory that 'more support is better.' McCord's findings forced a generation of social scientists to take seriously the possibility that intervention itself could cause harm — what is now called the iatrogenic effect. The leading explanation involves peer contagion at the summer camps treatment boys were sent to (mixing high-risk peers together appeared to consolidate antisocial norms), but the precise mechanism remains debated. The study's most important legacy is methodological: it demonstrated that a benevolent program designed by competent professionals, evaluated against a no-treatment control, can produce reliable harm — and that without rigorous long-term evaluation, no one would have known.

Research question

"Does intensive mentoring, counseling, and structured support for at-risk boys reduce later delinquency, criminal involvement, and adverse life outcomes?"

Methodology

Result

Negative

Treated men were *more* likely than controls to have committed multiple crimes (relative risk ~1.4), to have died younger, to be diagnosed with serious mental illness, alcoholism, or stress-related illness, and to report job dissatisfaction. On essentially every outcome examined, the treatment arm did worse.

Evidence strength

Strong

Randomized controlled trial with large sample.

Replication status

Partially replicated