Prometheus 6

Sentencing by the Numbers

By EMILY BAZELON

Imagine that you could stop a crime before it happened. Not by zapping a murderer seconds before he bloodied his victim, like the future cop played by Tom Cruise in ''Minority Report,'' but by sitting calmly on the bench in judge's robes and perusing a single sheet of data. Armed with the stats, you could predict the likelihood that a convicted drug dealer or thief standing before you would be arrested again in the near future if you let him go free.

For decades, the science of predicting future criminality has been junk science -- the guesswork of psychologists who were wrong twice as often as they were right. But today, the detailed collection of crime statistics is beginning to make it possible to determine which bad guys really will commit new offenses. In 2002, the Commonwealth of Virginia began putting such data to use: the state encourages its judges to sentence nonviolent offenders the way insurance agents write policies, based on a short list of factors with a proven relationship to future risk. If a young, jobless man is convicted of shoplifting, the state is more likely to recommend prison time than when a middle-aged, employed woman commits the same crime.

Virginia's new sentencing method was born of a budget crunch. Faced with the prospect of building new prisons after passing a tough-on-crime measure in 1994, the Legislature asked the state sentencing commission to figure out which nonviolent offenders could be kept out of prison without posing a risk of committing new crimes. The commission's director, Richard Kern, and his staff members tracked 1,500 nonviolent drug, larceny and fraud offenders for three years after their release from prison. The researchers found that men were 55 percent as likely to be rearrested as women, and that offenders in their 20's were a much higher risk than those older than 40. Being unemployed made offenders more likely to commit another crime. So did being single.

Using these factors and a few others, including a defendant's adult and juvenile criminal records, Kern designed a simple 71-point scale of risk assessment as an aid for judges. If he scores 35 points or less, a defendant who would have otherwise gone to prison under Virginia sentencing guidelines is recommended for an alternative sanction like probation or house arrest. Anything above 35 means a recommendation of jail time. ''Judges make risk assessments every day,'' Kern said. ''Prosecutors do, too. Our model brings more equity to the process and ties the judgments being made to science.''

Kern tested his model on prisoners released five years earlier and found that his ratings correctly predicted who would be reconvicted in three out of four cases. Of the felons who scored at or below the 35-point cutoff, 12 percent committed new crimes, compared with more than 38 percent for those who scored higher. After calculating that only a slight increase in recidivism would result, the state raised the 35-point cutoff to 38 points last July. Meanwhile, the growth of the state's prison population -- which used to be more than twice the national average -- has slowed nearly to a halt.