Skip to content

Do Arrests Predict Future Behavior

THE PREVALENCE of criminal history background checks as part of the hiring process prompted a recent study by researchers at Carnegie Mellon University in Pittsburgh, Pennsylvania. They analyzed recidivism rates over time in an effort to determine how long it takes before a criminal is no more likely to commit crimes than the rest of the population. The authors examined New York State arrest records for more than 88,000 individuals who were first arrested in 1980 and followed them through 2007.

“We were able to follow their records and find that a large fraction did indeed get rearrested in those first few years, but those who did stay clean for five years or more, they were pretty good risks,” says Alfred Blumstein, a Carnegie Mellon professor and co-author of the study, which he wrote about in the journal Criminology.

The researchers examined what is known as the “hazard function,” according to Blumstein, “which is the probability that someone will get arrested after staying clean for a reasonable period of time. And that probability declines the longer he stays clean.”

Another standard the study looked at is how the arrested individual compares with people in the general population who have never been arrested. Blumstein says that, over time, the previously arrested individual’s risk can get very close to that of the nonarrested individual but is very unlikely to dip below it. The researchers looked at crimes that were “sufficiently numerous” for statistical analysis, such as burglary and assault; they did not look at more serious but less common crimes, such as murder.

The U.S. Equal Employment Opportunity Commission (EEOC) hopes “that employers use these findings in crafting their criminal records policies,” says Carol Miaskoff, EEOC assistant legal counsel.

Lester S. Rosen, president of Employment Screening Resources, says that while the study is very important and “for the first time, attempts to measure some of the statistics and…how we actually judge whether or not a person is a danger,” the downside of the study is that people may “try to extrapolate from it things that the study doesn’t say and can’t say.” He stresses that this research should instead be viewed as “just the start of a long road.”

Rosen points out that one of the study’s weaknesses is that it looks at New York data only, when people could have been arrested in another state after their New York arrest. “Without those statistics, their results are, in my mind, very skewed in favor of a lower [recidivism] number,” says Rosen.

Blumstein acknowledges that weakness, saying one of his next steps will be to compare the full arrest records of these individuals to see if they have been arrested in other states. The researchers have already contacted the FBI to make those checks. Based on very preliminary estimates, this additional research might lead to a 10 percent increase in the estimated risk that a person once arrested will again commit a crime, Blumstein says.

This study may help to spur the discussion on access to records and the use of criminal records, says Arthur Cohen, vice president of operations and general counsel of Concorde, Inc., an employee background screening firm. There are laws regarding the availability of criminal records in background checks (read more about this in the May 2009 issue of Security Management).

Cohen says that access to the FBI database is what the researchers “really need because they have to slice and dice all the data, and they’ve got to do it for all offenses.” Blumstein says that he was required to obtain a National Institute of Justice grant in order to gain access to the FBI data for the research.

Another weakness pointed out by Rosen was that the study looked at arrests and not convictions. Looking at convictions only would greatly reduce the starting sample size, which the report acknowledges. Blumstein hopes to examine convictions in the future and rule out arrests that do not lead to convictions.

The report suggests that the findings could perhaps influence employer liability with regard to charges of negligent hiring. For example, if a company hired an ex-convict who was clean for five years but then committed a crime that resulted in charges of negligent hiring, the employer could argue that the company had reason to believe the person did not pose a risk.

Blumstein also suggests that state record repositories limit dissemination of certain records. In other words, perhaps records of certain crimes predating five years would not be made available for everyone searching that person’s background on the theory that it’s no longer relevant and would only tarnish the person unfairly.

Rosen cautions about basing laws on the research. “Our concerns are mathematical rules imposed on employers that replace the employers’ ability to make a judgment. In other words, should the government be able to tell employers, you have to take somebody, because some mathematical model says they’re now safe?” says Rosen.

It may depend on the job, Rosen notes. “The point that we make in the background screening industry is that there’s a job for everybody, but not everyone is entitled to every job.”