'Scared Straight' Didn't Scare Kids Straight. Nine Randomized Trials Found Prison Visits Raised the Odds of Youth Reoffending by 68%

In 1978, a group of men serving life sentences at Rahway State Prison in New Jersey sat across a table from 17 juvenile delinquents and tried to terrify them straight, with the cameras rolling. The resulting documentary, Scared Straight!, won the Academy Award for Best Documentary Feature, drew a huge television audience, and planted a simple theory in the American mind: show a kid the worst of prison life, and fear will do the rest, and the program's promoters claimed that barely one in ten participants ever reoffended. Three decades later, A&E turned the formula into Beyond Scared Straight, a reality series that ran nine seasons and became one of the network's highest-rated shows. The intuition felt unassailable: who could watch a lifer describe what happens to young men behind bars and believe the teenager across the table walked out unchanged?

Anthony Petrosino and his colleagues decided to find out what happened when the intuition met randomization. Their systematic review hunted down every trial that had randomly, or quasi-randomly, assigned juveniles to either a Scared Straight-style program or to nothing at all, then tracked who reoffended. Nine trials made the cut, run between 1967 and 1992 in eight states, covering 946 juveniles and young adults ages 14 to 20, most of them 15 to 17 and almost all male. The programs followed the classic recipe: organized visits to adult prisons, confrontational sessions with inmates, sometimes a night in a cell wearing a prison uniform. No one was deterred.

The numbers pointed the wrong way: pooling the seven trials that reported reoffending rates, the odds of reoffending were 68 percent higher among the kids who took the tour than among the kids left alone: an odds ratio of 1.68, with a 95 percent confidence interval of 1.20 to 2.36, and a random-effects model gave 1.72. The authors then tried to break their own result, dropping the trial with the weakest randomization (odds ratio 1.47), then the high-attrition trial (1.96), then both at once (1.68). An earlier, simpler pass at the same studies had found program groups offending at rates 1 to 28 percent higher than their no-treatment controls. The authors' conclusion was blunt: these programs “increase delinquency relative to doing nothing at all to similar youths,” and could not be recommended as a crime prevention strategy; the tour backfired.

Deterrence theory has always said punishment deters when it is certain and swift, and a single terrifying afternoon is neither. The researchers and their successors have floated three explanations for the reversal. Deviant peer contagion: gather the most at-risk teenagers in one room, give them a shared experience and each other's company, and the network you build is a criminal one. Labeling: the system stamps the kid as a future inmate, and the kid starts living up to the stamp. Reactance: confronted adolescents do the opposite of what authority demands, a reflex every parent of a teenager will recognize. A companion Campbell review by the same lead author, covering 29 randomized trials of formal juvenile processing versus diversion, found the same pattern: official processing did not deter and sometimes increased crime. The threat, it seems, is not credible to the kids receiving it, and the delivery vehicle carries side effects. The cure was poison.

The institutions eventually listened: the National Institute of Justice's CrimeSolutions database rates juvenile awareness programs “Ineffective.” The Office of Juvenile Justice and Delinquency Prevention refuses to fund them and has warned they may violate the federal law barring contact between court-involved youth and adult inmates. In a 2011 Baltimore Sun essay, the agency's administrator called the approach “scary, and ineffective.” A landmark federal review filed the programs under “what does not work.” And still the model survives: cheap to run, beloved by television producers, and irresistible to parents at their wit's end, who keep calling local sheriffs asking to sign up their kids. The evidence has been public for a quarter century, and the intuition keeps winning.

Now the strongest case against this reading, at full strength. The nine trials are old: the newest data in the review were collected in 1992, and the 2013 update searched through December 2011 without finding a single new eligible trial. That freezes the evidence in amber. Modern programs are not the 1980s Rahway model, and nobody has tested them properly. Nobody is likely to, either, because ethics boards will not randomize children into a program the literature says harms them, which makes the finding almost impossible to update. Second, every outcome came from official records, arrests and court intakes, which measure detection as much as behavior, since a kid who has toured a prison is watched more closely afterward and extra surveillance inflates the treatment group's arrest count; third, the review tested only the confrontational recipe, so gentler variants built around mentoring, like the one-on-one counseling segments A&E appended to every episode, were never isolated and tested. Fourth, nine small trials is a thin foundation, and two of them had real problems, one with inadequate randomization and one with heavy attrition. The sensitivity analyses survive both, but they are doing heavy lifting on little data. Finally, the explanation for the harm is a guess. The review proves the backfire, not the mechanism.

The skeptic's best case is strong, and it still loses on the arithmetic. An odds ratio of 1.68 that survives every sensitivity analysis, in a review whose authors disclosed that two of them had published the same conclusion before and therefore had every reason to be careful about finding it again, is not a statistical hiccup. The famous success rates that launched a thousand programs, 80 to 90 percent staying clean, were uncontrolled anecdotes: no comparison group, no randomization, cameras in the room. The controlled evidence says the prison tour is worse than the couch. Fear is a theory about teenagers. Randomization is a measurement. When they disagree, believe the measurement.