(

ne of the stated goals in President
Barack Obama's crime and law
enforcement agenda is to break
down employment barriers for people who
have a prior criminal record, but who have
stayed clean of further involvement with the
criminal justice system. To understand how
many people are affected by some of these
barriers, we only need look at the widespread computerization of criminal history
records in the United States.

O

According to the Society for Human
Resource Management, more than
80 percent of U.S. employers perform
criminal background checks on prospective
employees' Add two additional factors to
that equation - advances in information
technology and growing concerns about
employer liability - and we can begin to
understand how complicated the issue of
employing ex-offenders has become.

10

The numbers leave no doubt that we have
reached a broad penetration of criminal history records into the fabric of our society:
• In 2006, nearly 81 million criminal records
were on file in the states, 74 million of
which were in automated databases.'
• Another 14 million arrests are recorded
every year. 3
What does this mean for employers? And
what does it mean for ex-offenders who
need a job?
Consider a 40-year-old male who was convicted of burglary when he was 18 years old
and has committed no further crimes. Every
time he applies for a new job, he tells the
potential employer that he was convicted
of a felony; even if he does not state this
up-front, the employer is likely to do a
criminal background check. In either case,

NIJ JOURNAL / ISSUE NO.

2631

Nil

--------

he probably will not get the job because
many employers are unwilling to hire an
ex-offender. 4
This situation prompted us to ask the
question: Is it possible to determine
empirically when it is no longer necessary
for an employer to be concerned about
a criminal offense in a prospective
employee's past?
Most people would probably agree that
there should be some point in time after
which ex-offenders should not be handicapped in finding employment. The question is when, precisely, should this occur?
In the case of our hypothetical 40-year-old,
when should a prospective employer no longer consider a burglary that was committed
more than two decades earlier if the job
applicant has stayed clean since then?
Currently, employers have no empirical
guidance on when it might be considered
safe to overlook a past criminal record
when hiring an ex-offender for a particular
job. Employers generally pick an arbitrary
number of years for when the relevance of
a criminal record should expire: five or 10
years, for example. It goes without saying
that different types of employers will have
different sensitivities about the potential
employee's criminal record. Those sNving
vulnerable populations like children and the
elderly would be particularly sensitive to
a prior record involving violence, while a
bank hiring a teller would be particularly
sensitive to property crimes. A hiring crew
for a construction company might be far
less sensitive to most prior records.
The point is that determining when a
potential employee's criminal record may
no longer be relevant has, to date, been an
arbitrary exercise. Although considerable
research has been done on how to forecast
possible criminal behavior,' no empirical
basis has been found for deciding when
a person's record is stale enough for an
employer to consider it no longer useful
or relevant 6

There should be some point in time
after which ex-offenders should not be
handicapped in finding employment.
The question is when, precisely,
should this occur?
The National Institute of Justice funded
our study to "actuarially" estimate a point
in time when an individua~ with a criminal
record is at no greater risk of committing
another crime than other individuals of the
same age. Although our research is ongoing
- and our findings, discussed in this article
are preliminary - we have created a model'
for providing empirical evidence on when an
ex-offender has been clean long enough to
be considered, for employment purposes,
"redeemed." An in-depth discussion of our
findings and research methods appears in
the May 2009 issue of Criminology.'

What We Have Known for Years
It is well known - and widely accepted by
criminologists and practitioners alikethat recidivism declines steadily with time
cleans Most detected recidivism occurs
within three years of an arrest and almost
certainly within five years 9 But is it possible
to identify when the risk of recidivism
has declined sufficiently to be considered
irrelevant in hiring decisions?
In our study, we obtained the criminal history records of 88,000 individuals who were
arrested for the first time in New York state
in 1980'0 First, we determined whether
they had committed any other crime(s)
during the ensuing 25 years or if they had
stayed clean. Then we compared this data
against two populations:
(1) People in the general population who
were the same age."
(2) People of the same age who had never
been arrested.

Until now.

11

Nil

NIJ JOURNAL / ISSUE NO. 263

We believe that our analysis provides the
criminal justice community with the first
scientific method for estimating how long is
"long enough" for someone with a prior record
to remain arrest-free before he or she
should be considered "redeemed"
by a prospective employer.
Our goal was to determine empirically at
what point in time the risk of recidivism
for people in our study group was no greater
than the risk for our two comparison populations." To do this, we plotted data curves
to determine when the risk of re-arrest for
individuals in our study group:
• Dropped below the risk of arrest for
same-aged people in the general
population.
• Approached the risk of arrest for people
who had never been arrested.
We believe that our analysis provides
the criminal justice community with the
first scientific method for estimating how
long is "long enough" for someone with
a prior record to remain arrest-free before
he or she should be considered "redeemed"
by a prospective employer.

Determining the Hazard Rate
Our analysis was based on a statistical
concept called the "hazard rate." The
hazard rate is the probability, over time,
that someone who has stayed clean will
be arrested. For a person who has been
arrested in the past, the hazard rate
declines the longer he stays clean.
To determine the hazard rate for our study
group, we looked at two factors:

• Age at the time of the 1980 (first) arrest.
• Type of crime.
We then compared these hazard rates, as
they declined over time, to people of the
same age in'the general population. For
these data, we used the arrest rate (the
age-crime c~rve) from the Uniform Crime
Reports, maintained by the Federal Bureau
of Investigation.
In the figure on page 13, we show the
hazard rate for 18-year-olds when they were
arrested for a first offense of one of three
crimes: robbery, burglary and aggravated
assault. The figure shows that for robbery,
the haeard rate declined to the same arrest
rate for the general population of sameaged individuals at age 25.7, or 7.7 years
after the 1980 robbery arrest. After that
point, the probability that individuals would
commit another crime was less than the
probability of other 26-year-olds in the
general population .
The figure also shows our analysis for burglary and aggravated assault. The hazard
rates people who committed burglary at
age 18 declined to the same as the general
population somewhat earlier: 3.8 years
post-arrest at age 21.8. For aggravated
assault, the hazard rates of our study group
and the general population of same-aged
individuals occurred 4.3 years post-arrest
or at age 22.3.
Individuals who were arrested for robbery at
age 18 had to stay clean longer than those
who were arrested for burglary or aggravated assault to reach the same arrest rate as
same-aged people in the general population.
We also looked at the effect of the
arrestee's age at the time of his first
arrest in 1980. We examined the hazard
rates for three ages of people in our study
group - 16, 18 and 20 years old - who
were arrested for robbery in 1980. Based
on the criminal histories of these people,
we found that individuals who were first

NIJ JOURNAL / ISSUE NO. 263

Nil

Hazard Rate for 18-Year-Olds: First-lime Offenders Compared
to General Population
The probability of new arrests for offenders declines over the years and
eventually becomes as low as the general population.
25%..,---------------------

20%++--------------------

Probability .1
first arrest for
general population
or subsequent
arrest for
previous
offenders

15%

10%

........

\\

""'",,---

~.

... ~~~,
5%

. \,

.._-- ........

.........- ...------.... General population

+--------"'",---"...- \ - - - - - - - - - - Robbery
;~~~'-. Aggravated assault

Burglary

2

4

6

8

10

12

14

16

18

Years since first arrest

arrested when they were 18 years old had
the same arrest rate 7.7 years later as a
same-aged individual in the general population. In contrast, those whose first arrest
occurred at age 16 crossed the curve for a
same-aged individual in the general population 8.5 years later, and individuals who

were first arrested at age 20 crossed their
curve 4.4 years after their .first arrest.
Thus, our analysis showed that the younger
an offender was when he committed
robbery, the longer he had to stay clean
to reach the same arrest rate as people

13

Nil

Our findings could play an important role
in policy discussions about the maintenance
of and access to criminal record databases.
his same age in the general population. We
also performed the same analysis for the
first offenses of burglary and aggravated
assault and found similar results.

Comparing Hazard Rates
to the Never-Arrested
As noted earlier, our study also compared
hazard rates to people who had never been
arrested. Needless to say, the hazard rates
for people in our study group (because they
had been arrested) would never be the
same as the hazard rate for people who had
never been arrested. But it is reasonable
to expect that an ex-offender's hazard rate
gets close enough - the longer he stays
clean - for an employer performing a criminal background check to determine acceptability for a particular position.
The higher an employer's risk tolerancethat is, the closer a prospective employer
would have to get to the hazard rate of the
never-arrested - the longer an ex-offender
would have to stay clean.

How Robust Were Our Results?
Our preliminary results are limited to people
who were arrested in New York state in
1980. Our next step will be to determine
if the data hold true at other times and in
other places. For example, we want to see
whether we get similar results if we draw
upon a sample of people who were arrested
for the first time in 1985 and in 1990
because these years were quite different
from 1980 in a number of important ways:
• 1980 was a peak crime year due.to demographic shifts of baby boomers aging out
of the high-crime ages.

14

(

NIJ JOURNAL / ISSUE NO. 263

• 1985 saw a "trough in crime rates" before
young people were recruited to sell crack
as older crack sellers were sent to prison.
• 1990 was near a peak before the beginning of the ,crime drop in the 1990s'"
If we find that the hazard rates for exoffenders in these years are similar to what
we have found in our preliminary analysis,
the usefulness of our hazard-rate analysis
method would be strengthened.
Note that our analysis looked at any crime
as the marker for when a second arrest
occurs; we would also like to examine the
relative risk of a specific second crime
becausil, as we stated earlier, different
types of employers have different risk
tolerances for particular crimes.
We also want to test our risk-analysis
model with data from different states.
Although it is possible that variations in
local populations and arrest practices may
affect the results, we anticipate that they
would be reasonably close.
Another aspect of future research will
explore the possibility that some of the individuals in our study group who looked clean
in New York state might have been arrested
in another state. We will access FBI records
to determine if an individual with no further
arrests in New York may have been arrested
in New Jersey or Florida, for example.

Public Policy Implications
We believe that our preliminary findings
and ongoing research offer an opportunity
to think about when an ex-offender might
be "redeemed" for employment purposes
- that is, when his or her criminal record
empirically may be shown to be irrelevant
as a factor in a hiring decision.
People performing criminal background
checks would find it valuable to know when
an ex-offender has been clean long enough
that he presents the same risk as other

N I J J 0 URN A L / ISS U END. 263

people in the general population. Employers
also might be more likely to use this type
of analysis if there were state statutes protecting them against due diligence liability
claims when they adhered to reasonable
risk-analysis findings.
We also believe that our findings could
play an important role in policy discussions
about the maintenance of and access to
criminal record databases. Considerable
policy control rests with those who oversee
state criminal history repositories. These
decision-makers could establish policies
that prevent repositories from distributing
records that are determined by hazardrate analysis to be no longer relevant. Or
repositories could seal or even expunge
old records if they are deemed, based on
such an analysis, to be no longer relevant to
assessing future risk. Such policy decisions
would inevitably vary from state to state and
be driven by other relevant considerations,
but policymakers may find valuable guidance in our research findings and methods
for considering such decisions.
For example, officials who manage repositories of criminal records could inform prospective employers (and others who access
criminal history records) when such records
are "stale" - that is, when a recidivism risk
analysis demonstrates that a prior arr,est
or conviction is no longer meaningfully relevant. Pardon boards, too, could use this
type of analysis to decide when to grant a
pardon to an applicant.

Where to From Here?
At a meeting of the American Society of
Criminology in the early 1970s, one of the
panelists argued against computerization that was just then beginning - of criminal
history records. Computers, he maintained,
didn't understand the Judeo-Christian concept of "redemption." Another panelist
challenged him, stating that paper records
certainly did not understand that concept ...
but at least computers could be "taught."

INI)

We believe that these findings represent the first
empirical evidence on "redemption times" and how
these could affect policies aimed at enhancing
employment opportunities for ex-offenders.
Our research is looking at what we might
"teach" those computers.
As we said at the beginning of this article,
our research is ongoing and needs much
further robustness testing to ensure that
findings apply more universally, beyond
our study group of first-time 1980 arrestees
in New York. Nonetheless, we believe
that these findings represent the first
empirical evidence on "redemption times"
and how these could affect policies aimed
at enhancing employment opportunities for
ex-offenders.
NCJ 226872

About the Authors
Alfred Blumstein, Ph.D., is the J. Erik Jonsson University Professor ..
of Urban Systems and Operations Research and former dean atthe Heinz
College of Carnegie Mellon University. In 1987, Blumstein received the
American Society of Criminology's Sutherland Award for his contributions to research; he was president of the ASC from 1991 to 1992. In 2007,
Blumstein was awarded the Stockholm Prize in Criminology. His research
started when he was director of the task force on science and technology forthe 1965-1967 President's Crime Commission and has covered
many aspects of criminal justice, including crime measurement, criminal
careers, sentencing, deterrence and incapacitation, prison populatio~s,
demographic trends, juvenile violence, and drug-enforcement policy.'
Kiminori Nakamura is a doctoral student at the Heinz College of Carnegie
Mellon University. His research interests include the dimensions of a criminal career, life-course (developmental) criminology, recidivism, collateral
consequences of criminal history records, and quantitative methods such
as social network analysis. Nakamura received his M.A. in demographic
and social analysis in 2005 and his B.A. in criminology, law and society in
2004 from the University of California, Irvine.

15

NIJ

NIJ JOJRNAL / ISSUE NO. 263

For More Information
• Blumstein, A., and K. Nakamura,
"Redemption in the Presence of
Widespread Criminal Background
Checks," Criminology 47 (2) (May 2009).
• See Criminology & Public Policy 7 (3)
(August 2008): Uggen, C., "Editoria[
Introduction to 'The Effect of Criminal
Background Checks on Hiring
Ex-offenders:" 367-370; Freeman, R.,
"[ncarceration, Criminal Background
Checks, and Emp[oyers in a Low(er)
Crime Society:' 405-412; Stoll, M.A.,
and S.D. Bushway, "The Effect of
Criminal Background Checks on Hiring
Ex-offenders:' 371-404; and Western,
B., "Crimina[ Background Checks and
Emp[oyment Among Workers With
Criminal Records:' 413-417.
• See also Ho[zer, H.J., S. Raphael, and
M.A. Stoll, "Perceived Criminality,
Criminal Background Checks, and the
Racia[ Hiring Practices of Emp[oyers:'
Journal of Law and Economics 49
(October 2006): 451-480.

Notes
1.

Burke, M.E., 2004 Reference and
Background Checking Survey Report: A
Study by the Society for Human Resource
Management, Alexandria, VA: Society for
Human Resource Management, 2006.

2. Bureau of Justice Statistics, Survey of
State Criminal History Information Systems,
2003, Criminal Justice Information Policy
Report, Washington, DC: Bureau of Justice
Statistics, February 2006 (NCJ 210297),
available at www.ojp.usdoj.gov/bjs/pub/pdfl
sschis03.pdf.
3. Federal Bureau of Investigation, Crime in
the United States, 2007, Washington, DC:
Federal Bureau of Investigation, September
2008, available at www.fbi.gov/ucr/cius2007.

16

4. Holzer, H.J., S. Raphael, and M.A. Stoll,
"Perceived Criminality, Criminal Background
Checks, and the Racial Hiring Practices of
Employers," Journal of Law and Economics
49 (2) (October 2006): 461-480.
5. Bushway, S.D., "The Impact of an Arrest on
the Job Stability of Young White American
Men," Journal of Research in Crime and
Delinquency 35 (4) (1998): 454-479.
6. Research has shown that the risk of offending for those with criminal records converges
toward the risk for those without a record
as substantial time passes. See Kurlychek,
M.C., R. Brame, and S.D. Bushway, "Scarlet
Letters and Recidivism: Does an Old
Criminal Record Predict Future Offending?"
Criminology & Public Policy 5 (3) (September
2006): 483-504; and Kurlychek, M.C., R.
Brame, and S.D. Bushway, "Enduring Risk?
Old Criminal Records and Predictions of
Future Criminal Involvement," Crime &
Delinquency 53 (1) (2007): 64-83.
7.

See Blumstein, A., and K. Nakamura,
"Redemption in the Presence of Widespread
Criminal Background Checks," Criminology
47 (2) (May 2009).

8. Maltz, M.D., Recidivism, Orlando, FL:
Academic Press, August 1984.
9. Beck, A.J., and B.E. Shipley, Recidivism
of Prisoners Released in 1983, Special
Report, Washington, DC: Bureau of Justice
Statistics, April 1989 (NCJ 116261), available at www.ojp.usdoj.gov/bjs/pub/pdf/
rpr83.pdf; and Langan, P.A., and D.J. Levin,
Recidivism of Prisoners Released in 1994,
Special Report, Washington, DC: Bureau of
Justice Statistics, June 2002 (NCJ 193427),
available at www.ojp.usdoj.gov/bjs/pub/pdfl
rpr94.pdf.
10. Data were provided by the New York State
Division of Criminal Justice Services; we
thank David van Alstyne, a research manager
in that office, for his support in this research
study. All data were provided with no individual identifiers - that is, all names and other
identifying information were removed before
the data were given to us.

NIJ JOURNAL / ISSUE NO. 263

11. "General population" included people with
no arrests as well as ex-offenders who had
served their time and were back in the
general population.
12. All of the findings reported in this article are
based on arrest records. As our research
continues, we will address case disposition. We anticipate that hazard rates in our

NIJ

ongoing analyses will be somewhat higher
because they will not include individuals who
were not charged or who were found to be
not guilty.
13. Blumstein, A., and J. Wallman, eds., The
Crime Drop in America, 2nd ed., Cambridge:
Cambridge University Press, 2006.

17