Recidivism Official Statistics

By Allen J. Beck, Ph.D.
BJS Statistician
Bernard E. Shipley
BJS Program Manager

Source

Of the 108,580 persons released from prisons in 11 States in 1983, representing more than half of all released State prisoners that year, an estimated 62.5% were rearrested for a felony or serious misdemeanor within 3 years, 46.8% were reconvicted, and 41.4% returned to prison or jail. Before their release from prison, the prisoners had been arrested and charged with an average of more than 12 offenses each; nearly two-thirds had been arrested at least once in the past for a violent offense; and two-thirds had previously been in jail or prison. By yearend 1986 those prisoners who were rearrested averaged an additional 4.8 new charges. An estimated 22.7% of all prisoners were rearrested for a violent offense within 3 years of their release.

These findings were based on a sample of more than 16,000 released prisoners, representing all those released from prison in 11 States during 1983. The 11 States in the sample included California, Florida, Illinois, Michigan, Minnesota, New Jersey, New York, North Carolina, Ohio, Oregon, and Texas. These States accounted for more than 57% of all State prisoners released in the Nation during the year.

Other findings from the survey include the following:

An estimated 68,000 of the released prisoners were rearrested and charged with more than 326,000 new felonies and serious misdemeanors, including approximately 50,000 violent offenses (of which 17,000 were robberies and 23,000 were assaults), more than from 141,000 property offenses (of which 36,000 were burglaries), and 46,000 drug offenses.

The research reported here represents the most ambitious of the studies with data for a large sample of releasees, representing more than half of all persons released from State prisons in 1983. By linking State and FBI criminal-history records, the study assembles for the first time comprehensive criminal-history data both within and outside the States in which the prisoners were released.

It provides the most precise estimates of recidivism available among prisoners of all ages and all types of postrelease supervision.

The Bureau gratefully acknowledges the contribution of the Identification Division of the FBI and officials of corrections departments and criminal history repositories in California, Florida, Illinois, Michigan, Minnesota, New Jersey, New York, North Carolina, Ohio, Oregon, and Texas.

This cooperation of State and Federal officials was vital to the success of this important research project.

Joseph M. Bessette
Acting Director

Office of Justice Programs

April 1989
Revised page 12 2/19/97

Recidivism was inversely related to the age of the prisoner at time of release: the older the prisoner, the lower the rate of recidivism.

The more extensive a prisoner's prior arrest record, the higher the rate of recidivism ¾ over 74% of those with 11 or more prior arrests were rearrested, compared to 38% of the first-time offenders.

The combination of a prisoner's age when released and the number of prior adult arrests was very strongly related to recidivism: an estimated 94.1% of prisoners age 18 to 24 with 11 or more prior arrests were rearrested within 3 years.

More than 68% of the prisoners released for property offenses were rearrested within 3 years, compared to 59.6% of violent offenders, 54.6% of public-order offenders, and 50.4% of drug offenders.

Approximately 40% of the released prisoners had previously escaped from custody, been absent without leave (AWOL), or had a prior revocation of parole or probation. An estimated 73% of these prisoners were rearrested within 3 years of their release.

The amount of time served in prison did not systematically increase a prisoner's likelihood of rearrest. However, those prisoners who had served the longest, more than 5 years in prison, had lower rates of rearrest than other offenders during the followup period.

Released prisoners were often rearrested for the same type of crime for which they had served time in prison. Within 3 years, 31.9% of released burglars were rearrested for burglary; 24.8% of drug offenders were rearrested for a drug offense; and 19.6% of robbers were rearrested for robbery.

Released rapists were 10.5 times more likely than nonrapists to be rearrested for rape, and released murderers were about 5 times more likely than other offenders to be rearrested for homicide. An estimated 6.6% of released murderers were rearrested for homicide.

Nearly 1 in 3 released violent offenders and 1 in 5 released property offenders were arrested within 3 years for a violent crime following their release from prison.

Measuring recidivism Criminal-history data for the sample of released prisoners were obtained from the criminal identification bureaus in the 11 participating States and from the Federal Bureau of Investigation. Information was collected only on felonies and serious misdemeanors. These data included information on arrests, prosecutions, court appearances, and postsentencing statutes, such as incarceration, probation, and parole.

Additional information on prisoners released in 1983, including data on demographic characteristics, sentencing, time served, and postrelease supervision status, was collected from the department of corrections in each State as part of BJS' annual Nation Corrections Reporting Program (NCRP). (See table 1 for a summary profile of prisoners in the sample.) Recidivism rates, as defined here, are estimates of the percentages of released prisoners who commit another offense.

Reconviction refers to a conviction on at least one charge after the date of release from prison. Data on reconvictions and other dispositions were reported by prosecutor's offices, courts, and correctional agencies.

Reincarceration refers to (1) any return to prison or (2) any admission to a local jail with a sentence for a new offense. Data on incarcerations were reported by receiving agencies, typically State or Federal prisons and local jails.

In previous studies of recidivism, criminologists have concluded that in the aggregate rearrest is the most reliably reported measure of recidivism.1 Although some rearrested individuals may be innocent of the crime charged, using only reported convictions would understate the true recidivism rates because not all offenders are prosecuted or go to trial.

Parolees, for example, often do not face prosecution for a new offense, but with their parole revoked they return to prison to serve the balance of their sentence.

Moreover, new convictions were the most underreported of the three measures.

Data on convictions and other dispositions were not reported for approximately 32% of all arrests in the criminal-history files.

For some arrests that contained no court records, convictions were inferred from prison admissions for a new sentence. 2 Recidivism of Prisoners Released in 1983 Table 1. Profile of prisoners released in 1983 and in 11-State recidivism study Number of prisoners released in the 11 States 108,580Percent of all States prisoners released in 1983, nationwide 57.3 %

[table omitted]

Note: Data are based on an estimated 108,580 prisoners who were released from prison in 11 States in 1983 and who were still alive in 1987. 1For a discussion of problems with other measures of recidivism, see M.D. Maltz, Recidivism (Orlando, Fla.: Academic Press, 1984).

Recidivism and time after release Within 3 years after their release from prison in 1983, an estimated 62.5% of the released prisoners had been rearrested; 46.8% had been reconvicted; and 41.4% had been reincarcerated (figure 1).

Recidivism rates were the highest in the first year. Four of every 10 released prisoners were rearrested in the first year; nearly 1 in 4 were convicted of a new crime; and nearly 1 in 5 were returned to prison or sent to jail (table 2).

Of all those who were rearrested during the 3-year followup period, approximately two-thirds were arrested by the end of the first year. Of all those reincarcerated, nearly 45% were reincarcerated in the first year.

An estimated 67,898 of the 108,580 prisoners who were released in 1983 were rearrested and charged with 326,746 new offenses by yearend 1986 (table 3). More than 50,000 of the new charges were violent offenses, including 2,282 homicides, 1,451 kidnapings, 1,291 rapes, 2,626 other sexual assaults, 17,060 robberies, and 22,633 other assaults.

More than 40% of the new charges were for property offenses. The released prisoners were rearrested for an estimated 51,268 larcenies, 36,483 burglaries, and 20,233 fraud offenses.

Nearly 25% of the new charges were for public-order offenses. Of the approximately 80,000 public-order offenses, 12,791 were weapons charges and 15,395 were violations of probation or parole or flight to avoid prosecution.

About 14% of the new charges were for drug offenses. Because sufficient detail was not recorded for more than 40% of the drug offenses, reliable estimates differentiating the number of possession and trafficking charges could not be made.

The released prisoners had been arrested in the past for more than 1.3 million offenses. Before their release from prison, they had been charged with an estimated 214,778 violent crimes, including more than 12,000 homicides, nearly 9,000 rapes, 5,600 kidnappings, and 84,000 robberies. When combined with the number of new arrest charges, these released prisoners had been arrested and charged with approximately 1.7 million offenses, an average of 15.3 charges each since their first adult arrest.

Recidivism of Prisoners Released in 1983 3

Percent of releasedprisoners who were:

Time after release Rearrested Reconvicted Reincarcerated

6 months 25.0% 11.3% 8.4%

1 year 39.3 23.1 18.6

2 years 54.5 38.3 32.8

3 years 62.5 46.8 41.4

Number of arrest charges

Arrest charge

Prior to

release

After release

1983-86

Total 1,333,293 326,746

Violent offenses 214,778 50,121

Homicidea 12,185 2,282

Kidnaping 5,622 1,451

Rape 8,922 1,291

Other sexual

assault 10,335 2,626

Robbery 84,166 17,060

Assault 84,497 22,633

Other violent 9,051 2,778

Property offenses 628,320 141,416

Burglary 184,690 36,483

Larceny/theft 199,450 51,268

Motor vehicle theft 54,157 8,649

Arson 3,294 647

Fraud 82,522 20,233

Stolen Property 60,873 13,738

Other property 43,334 10,398

Drug offenses 149,881 46,382

Possession 69,438 20,684

Trafficking 22,429 5,788

Other/unspecified 58,014 19,910

offenses 307,191 79,773

Weapons 55,539 12,791

Probation/parole

violations 44,962 15,395

Traffic offenses 35,300 5,844

Other public-order 171,390 45,743

Other offenses 12,957 1,111

Unknownb 20,166 7,943

Note: Data are based on an estimated 108,580 prisoners who were released from prison in 11 States in 1983 and who were still alive in 1987. aHomicde includes murder, nonnegligent manslaughter, and negligent manslaugter.

Unknown charge include those that could not be converted to an NCRP offense code and those not coded because only 6 charges were recorded for each arrest.

Prevalence of violence among State prisoners released in 1983 Among persons released from State prison in 1983, an estimated 77% had been arrested at least once in the past or rearrested after their release for a violent offense

Nature of violent record

Percent arrested

at some time for

a violent offense

Prior arrest charge 52.1%

Most serious charge

when released 34.6

Arrest charge within

3 years after release 22.7

Ever charged 77.0

6 12 18 24 30 36

0

20

40

60

80

100

Rearrested

Reconvicted

Reincarcerated

Cumulative percent

Months after release from prison

Total number

of arrest charges*

Percent of all

released prisoners

Cumulative percent

of all released prisoners

Cumulative percent

of all arrest charge

45 or more 5.0% 5.0% 19.4%

33-44 4.4 9.4 30.5

25-34 9.2 18.6 47.8

20-24 7.7 26.3 58.8

15-19 11.7 38.0 71.6

10-14 17.1 55.1 84.9

5-9 26.2 81.3 96.7

1-4 18.9 100.0 100.0

Note: percent may not add to 100% because of rounding.

*Arrest charge include those made prior to release and those between release in 1983 and December 31, 1986

Figure 1

A small fraction of offenders were responsible for a disproportionate number of these 1.7 million charges. An estimated 5% of the prisoners were charged with 45 or more offenses each before and after their release in 1983 (table 4). This group of high-rate offenders accounted for nearly 20% of all arrest charges. Offenders with 25 or more charges represented 18.6% of all offenders but accounted for 47.8% of all charges. Released prisoners with fewer than five arrest charges, however, represented 18.9% of all prisoners but only 3.3% of the arrest charges.

One way of measuring the volume of crime attributable to released prisoners is to express the new charges as a percentage of arrest charges recorded for all offenders in the 11 States during the same time period. Exactly comparable data on arrests of all offenders, however, do not exist. The Uniform Crime Reports (UCR), collected annually by the Federal Bureau of Investigation, record arrests for all offenders and classify the arrests by the most serious charge. Not all offense categories are compiled by the FBI, and definitions of some offenses differ from those in this 11-State study. The offenses that are comparable, the UCR Index crimes, include murder and nonnegligent manslaughter, rape, robbery, aggravated assault, burglary, larceny, and motor vehicle theft; therefore, the number of arrests of released prisoners for these offenses can be compared to the total number of arrests for Index crimes.

Overall, between midyear 1983 and yearend 1986, prisoners released in 1983 accounted for 2.8% of the Index crime arrests in the 11 States (table 5). The arrests of these released prisoners comprised 3.9% of the Index crime arrests in the 6 months after their release in 1983, 3.4% of the arrests in 1984, 2.6% in 1985, and 2.1% in 1986. It should be noted that by the end of the first year after release nearly 1 in 5 prisoners had been reincarcerated and were not liable for rearrest, and by the second year nearly 1 in 3 had been returned to prison or jail.

For the entire 3½ year period, the percentage of UCR arrests attributable to these released prisoners was highest for robbery (5.0%) and burglary (4.8%) and lowest for larceny (2.0%) and rape (1.8%) (table 6). In 1984, the first full year after release, these prisoners accounted for an estimated 6.4% of the arrests for robbery and 5.9% for burglary.

The new offenses occurred not only in the States in which the prisoners were released from prison but in other States as well. More than 1 of every 8 rearrests were made in States other than the State in which the prisoner was released. An estimated 5.5% of the released prisoners were rearrested only in States other than those in which they were released. An additional 4.7% of the prisoners were rearrested both in their State of release and in another State.

If data on arrests had been restricted only to the State of release, which is typical of most recidivism studies, the overall rearrest rate would have been an estimated 57.0% rather than the actual 62.5%. In addition, the number of prior arrests would have been underestimated: 26.5% of the prisoners had been arrested in more than one State in the past, and nearly 13.0% of all prior arrests had occurred in States other than the State of release in 1983.

4 Recidivism of Prisoners Released in 1983

Number of new charges

Number of arrests

for UCR Index

crimes among

released prisonersa

Total number of

arrests for Index

crimes in 11 Statesb

Percent of

all arrests

for Index crimes

Total, 1983-86 326,746 93,574 3,308,496 2.8%

1983c 56,892 18,494 469,315 3.9

1984 101,024 30,571 904,951 3.4

1985 86,613 23,572 921,877 2.6

1986 82,219 20,937 1,012,353 2.1

Note: Subcategories may not add to total because

of rounding.

aIncludes only arrests in the State in which the

prisoner was released. For arrests involving multiple

charges, the most serious charge was selected.

Index crimes include murder and nonnegligent manslaughter,

rape, robbery, aggravated assault burglary,

larceny, and motor vehicle theft. Arson was excluded.

bIncludes only adult arrests in the 11

States participating in the National

Recidivism Reporting System (NRRS).

cBecause on average there were only 6

months of exposure to rearrest, the estimated

total number of Index crime arrests for 1983

was divided by 2.

Years of arrest

Rearrest charge,

Index crime

Total

1983-86 1983 1984 1985 1986

Murder and nonnegligent

manslaughter 2.3% 3.1% 2.5% 2.0% 1.9%

Rape 1.8 3.8 2.4 .9 1.1

Robbery 5.0 6.6 6.4 4.1 3.7

Aggravated assault 2.4 3.0 2.7 2.5 1.7

Burglary 4.8 6.8 5.9 4.3 3.4

Larceny/theft 2.0 2.8 2.3 1.8 1.5

Motor vehicle theft 3.3 4.6 3.7 3.3 2.3

Note: For each percent the numerator is the number of

arrests for the Index crime among prisoners released in

1983, and the denominator is the estimated number of

arrests for Index crimes among all offenders in the 11

States. Percents for 1983 were adjusted

for partial-year exposure to rearrest.

Men were more likely than women to be

rearrested, reconvicted, and reincarcerated

after their release from prison

(table 7). The rate of rearrest was 11

percentage points higher among

men than among women.

Blacks had slightly higher recidivism rates

than whites, approximately 5 to 8 percentage

points higher for each measure. Released

prisoners of Hispanic origin also

had recidivism rates that were about 6

percentage points higher than those

among non-Hispanics.

Recidivism was inversely related to the

age of the prisoner at the time of release:

the older the prisoner, the lower the rate

of recidivism. More than 75% of those

age 17 or younger when released from

prison were rearrested, compared to

40.3% of those age 45 or older. However,

rearrest rates declined by less than

5 percentage points among prisoners

between the age of 18 and 34. (Prisoners

in this age group comprised nearly 80% of

all those released in 1983.) The largest

declines in recidivism were found among

prisoners age 35 or older, but even those

age 45 or older had rearrest rates of 40%.

The amount of prior education was also related

to recidivism among released

prisoners. Prisoners who had graduated

from high school or had some college

education had somewhat lower rates of

rearrest, reconviction, and reincarceration

than those who failed to complete high

school.

Prisoners released for property offenses

had higher recidivism rates than those

released for violent, drug, or public-order

offenses (table 8). An estimated 68.1%

of the property offenders released in 1983

were rearrested within 3 years, compared

to 59.6% of the violent offenders, 54.6%

of the public-order offenders, and 50.4%

of the drug offenders. Property offenders

also had higher rates of reconviction and

reincarceration than other types of

offenders.

Prisoners who had served time for motor

vehicle theft had the highest recidivism

Recidivism of Prisoners Released in 1983 5

Percent of released prisoners

who within 3 years were:

Prisoner characteristics

Percent of all

released prisoners Rearrested Reconvicted* Reincarcerated*

All released prisoners 100.0% 62.5% 46.8% 41.4%

Male 94.1% 63.2% 47.3% 41.9%

Female 5.9 51.9 38.7 33.0

White 54.1 58.7% 44.2% 38.0%

Black 45.1 67.1 49.9 45.3

0ther .8 58.7 50.6 45.3

Hispanic 12.1% 68.5% 52.4% 47.0%

Non-Hispanic 87.9 61.7 46.0 40.6

17 or younger .5% 75.6% 65.4% 50.6%

18-24 35.0 68.0 51.2 44.9

25-29 26.5 65,0 48.9 43.2

30-34 17.7 63.4 47.9 43.0

35-39 9.3 56.9 40.8 36.5

40-44 4.8 48.9 36.1 30.7

45 or older 6.2 40.3 28.6 25.7

8th grade or less 19.3% 61.9% 46.0% 38.4%

Some high school 48.0 65.1 46.9 40.9

High school graduate 25.8 57.4 39.8 35.0

Some college or more 6.8 51.9 36.1 30.4

Note: Data on sex were reported for 100% of 108,580

releases date on race for 99.6%, Hispanic origin for

99.9%, age at time of release for 96.8%, and education

for 46.6%. Subcategories may not add to total because

of missing data.

*Because of the underreporting of court and

custody data in Ohio, the percents reconvicted

and reincarcerated exclude data from Ohio.

Percent of released prisoners

who within 3 years were:

Most serious offense

for which released

Percent of all

released prisoners Rearrested Reconvicted* Reincarcerated*

All offenses 100.0% 62.5% 46.8% 41.4%

Violent offenses 34.6% 59.6% 41.9% 36.5%

Murder* 3.1 42.1 25.2 20.8

Negligent manslaughter 1.4 42.5 27.9 21.8

Kidnaping .6 54.5 35.7 31.3

Rape 2.1 51.5 36.4 32.3

Other sexual assault 2.1 47.9 32.6 24.4

Robbery 18.7 66.0 48.3 43.2

Assault 6.4 60.2 40.4 33.7

Other violent .4 50.1 33.2 31.4

Property offenses 48.3% 68.1% 53.0% 47.7%

Burglary 25.8 69.6 54.6 49.4

Larceny/theft 11.2 67.3 52.2 46.3

Motor vehicle theft 2.6 78.4 59.1 51.8

Arson .7 55.3 38.5 32.3

Fraud 5.5 60.9 47.1 43.3

Stolen property 1.7 67.9 54.9 50.5

Other property .8 54.1 37.3 33.9

Drug offenses 9.5% 50.4% 35.3% 30.3%

Possession 1.2 62.8 40.2 36.7

Trafficking 4.5 51.5 34.5 29.4

Other/unspecified 3.9 45.3 34.5 29.1

Public-order offenses 6.4% 54.6% 41.5% 34.7%

Weapons 2.2 63.5 46.7 38.1

Probation/parole violations 4.2 49.9 38.9 33.0

Other offenses 1.1% 76.8% 62.9% 59.2%

Note: The offense distribution and percents rearrested

are based on 106,216 released for whom most serious

offense at release was known. Percents of those re-

convicted and reincarcerated are based on

99,103 releases, after data from Ohio were

excluded.

*Includes nonnegligent manslaughter.

were rearrested, 59.1% reconvicted, and

51.8% reincarcerated. Other released

prisoners with relatively high recidivism

rates included those classified as "others"

unspecified felons (76.8%), burglars

(69.6%), those released for possession or

sale of stolen property (67.9%), larcenists

(67.3%), and robbers (66.0%). Those

released for murder or for negligent manslaughter

had the lowest rates of recidivism

(42.1% and 42.5%, respectively).

Released prisoners were often rearrested

for the same type of crime for which they

had served time in prison (table 9). For

example, an estimated 33.5% of released

larcenists, more than any other group,

were rearrested for another larceny. The

same pattern was observed for every

other type of offender. Thus, released

murderers were more likely than other

prisoners to be rearrested for a new

homicide (6.6%), released rapists were

more likely than other prisoners to be

rearrested for rape (7.7%), released

robbers to be rearrested for robbery

(19.6%), and so forth.

The relative likelihood of rearrest for

a similar crime was highest among

prisoners released for rape, sexual

assault, homicide, or fraud and lowest

among those released for public-order or

drug offenses (table 10). Released rapists

were 10.5 times more likely than nonrapists

to have a subsequent arrest

for rape. Prisoners who had served time

for other sexual assaults were 7.5 times

more likely than those who had not served

time for sexual assault to be arrested for

a new sexual assault. Those released

after serving time for murder or nonnegligent

manslaughter were nearly 5 times

more likely than other prisoners to be

rearrested for homicide.

Despite the tendency of released prisoners

to be rearrested for the same type of crime

for which they were released from prison,

they were often rearrested for other crimes

as well. Nearly a fifth of the released

property and public-order offenders were

rearrested for violent crimes during the

followup period (table 9). Released

property offenders were, however, much

more likely than violent offenders to be

rearrested for a property offense (49.8%

compared to 32.1%). Except for released

murderers, 20% or more of each type of

released prisoner were subsequently

arrested at least once for a public-order

crime.

6 Recidivism of Prisoners Released in 1983

Percent of prisoners rearrested within 3 years of release

whose most serious offense at time of release was:

Violent offense Property offense

Rearrest charge

Total ,

all

offenses Total Murdera Rape Robbery Assault Total Burglary

Larceny/

theft

Motor

vehicle

theft Fraud

Drug

offense

Public-order

offense

All charges 62.5% 59.6% 42.1% 51.5% 66.0% 60.2% 68.1% 69.5% 67.3% 78.4% 60.9% 50.4% 54.6%

Number of

released prisoners 106,216 36,769 3,258 2,214 19,815 6,756 51,332 27,416 11,896 2,785

5,809 10,104 6,826

Note: The numerator for each percent is the number

of persons rearrested for a new charge, and the

denominator is the number released for each type

of offense. Detail may not add to totals

because persons may be rearrested for

more than one type of charge.

aIncludes negligent manslaughter.

bIncludes murder, nonnegligent manslaughter,

and negligent manslaughter.

Rearrest charge

Relative

likelihood

of rearrest

Violent offenses 1.9

Homicidea 4.9

Rape 10.5

Other sexual assault 7.5

Robbery 2.9

Assault 2.1

Property offenses 2.3

Burglary 3.0

Larceny/theft 2.1

Motor vehicle theft 4.2

Fraud 4.6

Stolen property 2.4

Drug offenses 1.8

Public-order offenses 1.2

Note: For each type of rearrest charge,

the numerator is the odds of rearrest for that

charge among prisoners released for the same

type of offense; the denominator is the odds of rearrest

for that charge among prisoners released for

a different type of offense. Each ratio expresses

the odds of rearrest among prisoners released on

a similar offense relative to the odds of rearrest

among those released on a different type of

offense.

Recidivism rates were strongly related to

the number of prior adult arrests: the more

extensive a prisoner's prior arrest record,

the higher the rate of rearrest after release

from prison (table 11). Prisoners with a

greater number of prior adult arrests were

also arrested more quickly than those with

fewer prior adult arrests. Moreover,

regardless of how long prisoners stayed

away from rearrest following their release,

those with longer prior records had higher

rates of rearrest in subsequent time

periods than those with shorter records.

More than a quarter of all prisoners

released in 1983 had 11 or more prior

adult arrests. (An adult arrest is one that

occurred when the individual was of adult

age, as defined by State law, or when the

individual was a juvenile but charged or

tried in court as an adult.) Nearly 75%

of the prisoners with 11 to 15 prior arrests

and 82.2% of those with 16 or more prior

arrests were arrested again following their

release from prison. Approximately threequarters

of those rearrested who had 11

or more prior arrests were rearrested

within the first year.

In contrast, among those prisoners who

had one previous arrest (9.1% of all

released prisoners), approximately 38.1%

were rearrested within 3 years. Nearly half

of those rearrested among prisoners with

one prior adult arrest were rearrested

within the first year.

The percents of those rearrested among

released prisoners were systematically

related to the extensiveness of the prior

records. For each 3-month interval during

the followup period, the cumulative

percent of those rearrested was higher for

persons with more prior adult arrests

(figure 2). Within the first 6 months, for

example, released prisoners with 11 or

more prior arrests were nearly 4 times

more likely than those with 1 prior arrest

and more than twice as likely as those

with 2 or 3 prior arrests to have been

rearrested for a new offense.

Failure rates, defined as the number

rearrested within a 3-month period divided

by the number not yet rearrested at the

beginning of the period, were highest in

the first several months following release

from prison (table 12). Failure rates in the

first 3 months after release were more

than 3 times higher than those in the last 3

months of the followup period (14.2%

compared to 4.5%). In addition, in almost

every 3-month period those prisoners with

longer arrest records had higher failure

rates. Even after 33 months, prisoners

with 11 or more prior arrests were more

than twice as likely as those with 1 prior

arrest to be rearrested within the next 3

months (7.0% compared to 2.8%).

Recidivism of Prisoners Released in 1983 7

Percent of releases

who were rearrested

Number of adult arrests

prior to release*

Percent of

all releases Within 3 years Within 1 year

All released prisoners 100.0% 62.5% 39.3%

1 prior arrest 9.1 38.1 19.0

2 10.8 48.2 25.5

3 10.8 54.7 30.1

4 9.7 58.1 35.5

5 8.0 59.3 33.4

6 7.0 64.8 38.2

7-10 18.8 67.7 42.0

11-15 11.9 74.9 53.3

16 or more 14.0 82.2 61.5

Note: The percents were based on 108,309

weighted cases with valid data on the number

of prior adult arrests.

*An adult arrest is one that occurs when an

individual is of adult age, as defined by

State law, or when the individual is a

juvenile but is charged or tried in court

as an adult.

0 12 24 36

0

20

40

60

80

100

Cumulative percent

2-3

4-6

7-10

Months after release from prison

when rearrested

1

11 or more

prior ar rests

Of all prisoners not previously rearrested, the percent

rearrested for the first time during each 3-month

period after release, by number of prior arrests

Time period

after release

All released

prisoners

1 prior

arrest

2-3

arrests

4-6

arrests

7-10

arrests

11 or more

arrests

1-3 months 14.2% 5.1% 8.1% 11.7% 15.3% 24.1%

4-6 12.6 6.4 8.9 10.9 13.2 20.6

7-9 11.2 5.4 7.1 10.8 10.7 17.7

10-12 8.9 3.6 7.2 8.3 11.6 14.7

13-15 7.9 4.1 6.7 7.8 10.5 10.4

16-18 7.3 4.4 6.1 6.9 8.9 10.6

19-21 6.6% 3.4% 5.1% 6.9% 7.4% 9.4%

22-24 6.0 3.2 4.8 6.2 6.8 9.1

25-27 5.3 3.0 4.7 5.8 5.9 6.8

28-30 4.6 3.0 4.6 4.6 5.2 6.5

31-33 4.4 2.8 3.3 4.2 5.0 6.2

34-36 4.6 2.8 3.6 4.6 6.4 7.0

Note: For each percent the denominator is the

number of released prisoners who had not been

rearrested before the 3-month period, and the

numerator is the number who were

rearrested during the period.

Figure 2

The number of prior arrests remained

a strong predictor of recidivism among

released prisoners for both males and

females, for each race or ethnic group,

and for all age groups (tables 13 and 14).

Within each sex and racial category, the

percent rearrested among released

prisoners increased with the number of

prior adult arrests. The percentage

rearrested was slightly Iower among

Hispanics with 7 to 10 prior arrests than

those with 4 to 6 prior arrests, but the

difference was not statistically significant.

Within each prior arrest group, recidivism

rates varied by sex, race, ethnicity, and

age. Except for prisoners with 7 to 10

prior arrests, men had higher rates of

rearrest than women. Black prisoners had

rearrest rates that were 8 to 14 percentage

points higher than those for whites, depending

on the number of prior arrests.

Hispanic prisoners, who could be white or

black, also had rates higher than those for

non-Hispanic whites within every category

of prior arrests.

Within each prior arrest category, older

prisoners had lower rates of rearrest than

younger prisoners. Released prisoners

who had one prior adult arrest and were

age 35 or older had the lowest recidivism

rates of all prisoners: 6.8% of those age

35 to 39 and 12.1% of those age 40 or

older were rearrested within 3 years. In

contrast, an estimated 94.1% of the

released prisoners who were age 18 to 24

and who had 11 or more prior arrests were

rearrested during the followup period.

The age at which a released prisoner was

first arrested and charged as an adult was

inversely related to recidivism: the

younger the age at first arrest, the higher

the rate of recidivism (table 15). An estimated

72.2% of the prisoners first arrested

before the age of 18 were rearrested

within 3 years of their release, compared

to 39.2% of those first arrested between

age 25 and 29 and 26.6% of those first

arrested at age 30 or older.

Previous studies have found a similarly

strong relationship between the age at

which a criminal career began and the

probability of recidivism.2

Because records of arrest and prosecution

of juveniles were not reported in State or

Federal criminal files, unless the offender

was charged or tried in court as an adult,

an exact measure of the age at first arrest

was not available. Nevertheless, more

than 1 of every 4 released prisoners in

1983 had a record of an arrest before they

were 18 years old. It should be noted that

some arrests may have occurred in States

in which the age of majority is under 18.

Age at first adult arrest and recidivism

were related regardless of the number of

prior arrests. Within almost every category

of prior arrests, the older the prisoners

had been when first arrested, the lower

the rate of rearrest following their release

from prison. Among prisoners with one

prior arrest, those first arrested at age 17

or younger were about 4.3 times more

likely to be rearrested than those first

arrested at age 30 or older. Among those

with 11 or more prior arrests, those age 17

or younger when first arrested were nearly

1.3 times more likely than those age 30 or

8 Recidivism of Prisoners Released in 1983

Percent of State prisoners released in 1983,

who were rearrested within 3 years

Sex Race/ethnicity*

Number of prior

adult arrest Male Female

White non-

Hispanic

Black non-

Hispanic Hispanic

All released prisoners 63.2% 51.9% 56.1% 67.1% 68.4%

1 prior arrest 39.5% 21.4% 31.1% 45.1% 40.9%

2-3 52.6 32.9 46.3 56.2 53.3

4-6 61.1 47.1 54.8 64.1 69.6

7-10 67.6 69.2 63.0 71.6 67.8

11 or more 79.0 76.5 73.2 81.4 84.5

Number of prisoners 101,902 6,392 46,205 47,854 13,079

*Too few cases of other racial or ethnic groups existed to provide reliable estimates.

Percent of State prisoners released in 1983, who

were rearrested within 3 years, by age at release

Number of adult

arrest prior to release

Age 17 or

younger 18-24 25-29 30-34 35-39

Age 40

or older

All released prisoners 75.6% 68.0% 65.0% 63.0% 56.8% 43.7%

1 prior arrest 76.7% 48.6% 29.2% 24.8% 6.8% 12.1%

2-3 ... 61.8 42.4 38.2 38.2 14.6

4-6 ... 72.8 57.9 51.0 41.6 30.1

7-10 ... 81.0 72.5 64.8 54.5 39.0

11 or more ... 94.1 87.6 80.5 76.0 61.2

Number of prisoners 523 37,932 28,712 19,281 10,083 11,972

Note: Cases with missing data on the number

of prior adult arrests or age at time of release

were excluded.

...Too few cases to provide a reliable

estimates.

Percent of State prisoners released in 1983, who

were rearrested within 3 years, by age at first arrest

Number of adult

arrest prior to release

Age 17 or

younger 18-24 20-24 25-29

Age 30

or older

All released prisoners 72.2% 66.4% 55.1% 39.2% 26.6%

1 prior arrest 65.0% 53.5% 30.6% 15.9% 15.1%

2-3 62.1 56.5 46.4 35.5 22.8

4-6 68.6 61.2 55.7 45.4 32.4

7-10 69.3 69.3 65.3 60.2 41.9

11 or more 82.7 79.8 71.4 50.9 62.6

Number of prisoners 35,837 38,058 22,470 6,653 4,783

Percent of all released prisoners 33.2% 35.3% 20.8% 6.2% 4.4%

Note: Cases with missing data on the number of prior adult

arrests and age at first adult arrest were excluded.

2M.E. Wolfgang, R. Figlio, and T. Sellin, Delinquency in

a Birth Cohort (Chicago: University of Chicago Press,

1972); Recidivism of Young Parolees, BJS Special Report,

NCJ-104916, May 1987.

older to be rearrested. Overall, both age

at first adult arrest and the number of prior

arrests were related to the likelihood of

rearrest among prisoners released in

1983.

The length of a prisoner's criminal history,

defined as the time between the first adult

arrest and most recent prison admission

prior to release in 1983, was also

associated with recidivism (table 16). An

estimated 51% of the prisoners who had

been arrested for the first time in the year

before their most recent admission to

prison were rearrested within 3 years,

compared to 62% or more rearrested

among prisoners with longer criminal

histories.

The intensity of prior record rather than

length, however, was more strongly

associated with the likelihood of a

prisoner's rearrest. Prisoners with a large

number of prior arrests in a short period

of time were more likely to be rearrested

than those with fewer prior arrests in a

longer period of time. Within almost every

prior arrest category, the longer the

period of time over which the prior arrests

occurred, the lower the percentage rearrested

among released prisoners. Among

prisoners with two or three prior arrests,

for example, nearly 59% of those with a

criminal history of a year or less were

rearrested, compared to 26% of those with

a criminal history of over 10 years.

The amount of time served by prisoners

on their most recent offense before

their release in 1983 was not associated

with an increased likelihood of their

rearrest (table 17). Only the prisoners

who had served the longest, the estimated

4.1% who had been in prison for more

than 5 years, had lower rates of rearrest

during the 3-year followup period. An

estimated 48.3% of those who had served

more than 5 years in prison were

rearrested, compared to 59.0% or more

of those who had served less time.

Recidivism of Prisoners Released in 1983 9

Percent of State prisoners released in 1983, who were

rearrested within 3 years, by length of criminal history*

Number of adult

arrest prior to release

12 months

or less

13-24

months

25-60

months

61-120

months

More than

120 months

All released prisoners 50.9% 61.6% 66.0% 65.3% 64.3%

1 prior arrest 39.0% 37.5% 35.1% 22.8% ...

2-3 58.9 57.7 50.6 39.4 26.0

4-6 71.9 71.0 68.3 55.6 40.5

7-10 ... 72.9 77.7 72.1 54.8

11 or more ... ... 92.2 83.8 75.7

Number of prisoners 16,921 9,779 22,437 27,537 31,541

Percent of all

released prisoners 15.6% 9.0% 20.7% 25.4% 29.1%

Note: Cases with missing data on the number of prior adult arrests, date of birth, and date of most recent admission before release in 1983 were excluded.

...Too few cases to provide reliable estimates.

*Length of criminal history is defined as the number of months from the first adult arrest to the most recent admission to prison before release in 1983.

[table omitted]

Figures on time served are based on 55,263 cases.

In general, despite controls for the effects of prior record, age when released, age at first adult arrest, and type of offense for which released, no relationship was found between recidivism and length of time served in prison (table 18). Within each of the groups considered, differences in the likelihood of rearrest among those who served varying amounts of time in prison were inconsistent and often statistically insignificant. Though prisoners who had served more than 5 years had lower rates of rearrest, in order to make reliable comparisons, they were grouped with all those who had served more than 3 years.

Regardless of the length of time prisoners had served, the number of prior arrests, age when released, and age when first arrested as an adult remained strong predictors or recidivism. Further, with the exception of prisoners who had served between 19 months and 24 months, released property offenders had higher rates of rearrest than violent offenders. In every category of time served, released drug offenders had the lowest rates of rearrest.

Other characteristics of prior record Prior violent arrests Recidivism rates were related to the prevalence of violence in the prior record (table 19). An estimated 52.1% of all prisoners released in 1983 had been arrested for a violent crime before the crime for which they were released in 1983. Regardless of the offense category for which they were released, prisoners with a prior arrest for a violent offense had a greater likelihood of rearrest than other released prisoners. More than 68% of those with a prior violent arrest, compared to 56.2% of other prisoners, were rearrested within 3 years of their release.

Approximately 2 of every 3 released prisoners had previously been incarcerated in prison or jail for a crime other than the one for which they had served time and were then released in 1983.

Rearrest rates among prisoners who had been incarcerated before their most recent admission to prison were nearly 20 percentage points higher than the rates among prisoners who had been incarcerated for the first time.

10 Recidivism of Prisoners Released in 1983

Prisoners whose most serious offense when released was a drug offense were coded "no" if there was no prior drug arrest.

Prior escape or revocation Almost 40% of all prisoners released in 1983 had at some time in their past escaped from custody, been absent without leave (AWOL) or had a prior parole or sentence to probation formally revoked.

Approximately 73.1% of such prisoners were rearrested within 3-years followup period. Depending on the type of crime for which they had served time in prison, these prisoners had rearrest rates that were about 15 to 26 percentage points higher than the rates among prisoners with no prior revocation or history of escape or AWOL.

Prisoners with one or more prior drug arrests were more likely than those without a prior drug arrest to be rearrested within the 3-year followup period (68.6% compared to 58.8%). In addition, within each offense category for which prisoners were released, a prior drug arrest was associated with a higher rate of rearrest.

Numerous factors have been found to be related to the likelihood of rearrest. These factors were considered independently.

With the exception of controls for the number of prior adult arrests and most serious offense for which released, the relationships among these risk factors and their relative contribution to the likelihood of rearrest have not been examined. To avoid misinterpretation of the findings, the impact of each risk factor should be evaluated relative to that of other factors. Such a multifactor examination might or might not reveal,

for instance, that the higher rate of rearrest among certain types of property offenders than among violent offenders and drug offenders is the result of differences in the criminal histories or age composition of these offender groups rather than any unique attributes of the property offenders.

In addition, the numbers of prior arrests, a prior revocation or escape, a prior incarceration, a prior arrest for a violent offense, or a prior drug arrest may not be uniquely related to the likelihood of rearrest.

The effect of these factors may overlap, and some effects may be more important than others.

Logit analysis separates the effects of the risk factors by simultaneously controlling for the relationship between each factor and the likelihood of rearrest. In logit equations, one variable is considered dependent upon variation induced by others. The odds of rearrest (that is, the ratio of the number rearrested to the number not rearrested) for each combination of categories among risk factors is a multiplicative function of the effects of these factors. When transformed by logarithms, the effects of these factors are additive.3

Eight of nine risk factors examined here have independent net effects on the odds of rearrest (table 20). The length of criminal history has no statistically significant effect, primarily due to its close relationship to age when released and number of prior arrests. The effects of the remaining eight factors are reported in descending order of their overall net impact: age when released is found to have the largest impact, followed by the number of prior arrests, prior escape or revocation, most serious offense, prior incarceration, age at first arrest, prior violent arrest, and prior drug arrest. The effects of age at first arrest, prior arrest for a violent offense, and prior drug arrest, though statistically significant, are relatively weak (adding between 0.5% and 1.5 to the overall fit to the data) and may be excluded from the final equation.

The estimated effects, or logit coefficients, show more precisely the magnitude and direction of the effect that each category among the risk factors has on the likelihood of rearrest. A positive coefficient implies an increase in the predicted logarithm of the odds (or log odds) of rearrest; a negative coefficient implies a decrease. The contribution to the predicted log odds of rearrest by prisoners who were age 24 or younger (.721) is larger than that by those with 7 or more prior arrests (.694), which in turn is larger Recidivism of Prisoners Released in 1983 11 Table 20. The relationship of selected risk factors to the odds of rearrest within 3 years

[table omitted]

For an elaboration of logit analysis as well as other log-linear techniques, see S.E. Feinberg. The Analysis of Cross-Classified Data (Cambridge: MIT Press, 1977; or D. Knoke and O.J. Burke, Log-linear Models (Beverly Hills: Sage Publications, 1980). than that by those with a prior escape or revocation (.224), those released after serving time for a high-risk offense (.178), and those with a prior incarceration (.175).

Taken together the coefficients imply that individuals with all five of the high-risk characteristics have more than a 10-to-1 odds of rearrest. To calculate these odds exactly, the coefficient for the overall level of rearrest and for each category are summed (for example, .721 + .694 + .224 + .178 + .175 + .317 (a constant) = 2.309). The inverse of the natural logarithm of this sum provides an estimate of the odds, or 10.064 to 1. In terms of the predicted percent rearrested, these odds imply that 90.4% of the prisoners with these five characteristics will be rearrested within 3 years.

Estimates for released prisoners with other combinations of these five risk factors may also be computed. Individuals with the lowest risk of rearrest, for instance, are those who are age 35 or older, who had three or fewer prior arrests, who have no prior escape or revocation, who served time for a low-risk offense, and who had not been previously incarcerated. The estimated odds of rearrest for this low-risk group is .206 to 1, or 17.1%. The odds of rearrest for prisoners with other combinations of risk factors will vary between 17.1% and 90.4%.

The finding reveals that the odds of rearrest are predictable from each of the five factors. However, the finding do not suggest that the risk factors excluded from the equation are either statistically or substantively insignificant. Each is strongly related to the likelihood of rearrest but also related to one or more of the factors in the final equation. Age at first arrest, for instance, is related to both current age and the number of prior arrest.

No attempt has been to specify further the relationships among these factors. The final equation simply identifies the factor with the largest net effect on arrests.

Overall, these findings should not be used to predict the future behavior of any individual; however, they may be used to predict the likelihood of rearrest for specific groups of released prisoners that are identified by particular characteristics.

When interpreting these finding, the difference between risk assessment and the prediction of individual behavior must be kept in mind.4 The statistical test of the final equation was not based on the prospective number of "right" or" wrong" predictions from the classification of individuals on these characteristics. The equation merely provides an assessment of risk posed by released prisoners with specific characteristics.

A sample of prisoners released in 1983 was obtained from records submitted by participating States in the National Corrections Reporting Program (NCRP).

Individual corrections records were linked with records of arrests and prosecutions (rap sheets) maintained by the criminal identification bureaus in the 11 States.

Rap sheet data on offenders who were arrested in more than one State were obtained from the Federal Bureau of Investigation.

The States represented in the sample were California, Florida, Illinois, Michigan, Minnesota, New Jersey, New York, North Carolina, Ohio, Oregon, and Texas. State and Federal rap sheets were found for 16,355 of the 18,374 prisoners in the original sample. Excluding the 159 prisoners who died during the followup period, complete records were obtained for 90% of the original sample. Most of the sampled prisoners with incomplete records did not have an FBI identification number in their corrections record or on the State rap sheet. Without this number, FBI rap sheets could not be obtained. There was no evidence of any systematic difference between those persons with complete records and those lacking either a State or FBI rap sheet.

Findings in this study are representative of an estimated 108,580 prisoners who were released in the 11 States and who were alive in 1987. Only released prisoners with sentences to State prison of greater than 1 year are included. Administrative releases, prisoners who were absent without leave (AWOL), escapees, transfers, releases on appeal, and those who died in prison are excluded from the sample.

A separate, self-representing sample of male and female prisoners was drawn within each of the participating States, except Minnesota, in which all released prisoners were selected. Within each gender group in the 10 sampled States, prisoners were grouped into 24 strata that were defined by categories of race, age, and type of offense. Prisoners were selected systematically from each strata to yield independent samples of males and females within each State.

To adjust for differences in the sampling rate among State and for differences among strata in the coverage of rap sheets, a series of weights were introduced. The weights were applied so that individuals in each State and stratum were properly represented in the combined 11-State sample.

Eight of the 11 selected States, excluding Florida, New Jersey, and New York, were among the 29 States participating in the 1983 National Corrections Reporting Program. Released prisoners in the 11 States closely resembled all those reported in the NCRP (Appendix table).

The sex, race, age, and offense distributions were nearly identical in both groups of States. A slightly higher percentage of prisoners in the NCRP States than in the 11-State sample had served 6 months or less before their release in 1983 (16.8% compared to 14.5%). However, the difference may reflect longer time served by prisoners in Florida, New Jersey, and New York rather than sampling error.

Overall, the 95% confidence interval for the percent of all released prisoners who were rearrested within 3 years (62.5%) was approximately plus or minus 1 percentage point. The precision of other estimates varied by item, size of the estimate, and sample size for each group.

The precision of estimates of the percent rearrested based on 1,000 sampled 12 Recidivism of Prisoners Released in 1983 4See L.W. Shannon, "Risk Assessment vs. Real Prediction: The Prediction Problem and Public Trust, " Journal of Quantitative Criminology (June 1985) 1(2):159-189. Revised page 12 2/19/97 prisoners, for example, varied between 2% and 3.5%, depending on the percent rearrested.

Additional details on the sampling

procedures and precision of the sample are available upon request. Unless otherwise noted, differences cited in the text between groups of released at the 95% confidence level.

Coverage of criminal-history files Criminal-history information maintained by the State identification bureaus and the FBI includes all felonies and serious misdemeanors. These files exclude arrests and court actions, involving charges such as drunkenness, vagrancy, disturbing the peace, curfew violation, loitering, false fire alarms, unspecified charges of suspicion or investigation, and traffic violations (except manslaughter, driving under the influence of drugs or liquor, and hit and-run, which are included in the files).

Information on offenses committed by juvenile offenders is not reported in the rap sheets unless the offender was charged or tried in court as an adult. Consequently, all figures presented in this report refer to adult arrests only.

Arrests for serious offenses are not always recorded in the criminal-history files, largely because of the absence of readable fingerprint cards. To correct for this underreporting, incarceration records lacking prior arrest records were counted as arrests in the calculation of rearrest rates, time to first rearrest, and the number of prior adult arrests.

The offenses reported in the criminalhistoryfiles were recoded following definitions outlined in BJS Crime Definitions, which is available upon request. For each arrest in the files, the total number of charges and counts was recorded. However, the type of charge, disposition, and sentence were coded on a maximum of six charges per arrest. For prisoners released in 1983 after serving time in prison for more than one offense, the offense with the longest sentence was defined as the most serious.

[table omitted]

Shipley wrote this report. It was edited by Thomas Hester. John Dawson, Christopher Innes, and Jacob Perez provided statistical Dorothea Proctor assisted in survey production. Marilyn Marbrook, publications unit chief, administered report production, assisted by Sophie Bowen, Jeanne Harris, Yvonne Shields, and Jayne Pugh. The Regional Justice information Service (REJIS) of St. Louis, Missouri, processed the data.

April 1989, NCJ-116261

The Assistant Attorney General, Office of Justice Programs, coordinates the activities of the following program offices and bureaus: the Bureau of Justice Statistics, National Institute of Justice, Bureau of Justice Assistance, Office of Juvenile Justice and Delinquency Prevention, and the Office for Victims of Crime.