Denying parole at
first eligibility:
How much
public safety
does it
actually
buy?

A study of prisoner release and recidivism in Michigan

August 2009

Denying parole at
first eligibility:
How much public safety does it
actually buy?

A study of prisoner
release
and recidivism
in Michigan

August 2009

Additional copies of this report may be purchased for $8, including shipping and handling. The full report is
also available on the CAPPS website: www.capps-mi.org.

Citizens Alliance on Prisons and Public Spending
403 Seymour Avenue, Suite 200
Lansing, MI 48933
Tel: 517.482.7753
E-mail: capps@capps-mi.org
© August 2009

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
A Study of Prisoner Release and Recidivism in Michigan
Executive Summary 1
I. Framing the question: does denying parole keep the public safer? 8
A. Length of stay drives Michigan’s prison growth 8
B. Parole denials impact length of stay 10
C. Past harm and future risk: a critical distinction 12
II. Overview of the research 15
A. Defining recidivism 15
B. The data 16
C. The frequency of offense groups 17
III. The significance of crime type 19
A.

B.

C.

Crime type and re-offense rates: homicide and sex offenders do best 19
1. Overall differences 19
2. Returns for new crimes: offenses against people are rare 20
3. Return Rates for max outs vs. parolees 22
4. Placing the re-offense data in a national context 23
Crime type and parole denial: success rates and release decisions often
don’t match 27
1. Release differences by offense group 27
2. Differences between the old and new boards 29
3. The disconnect between success rates and release on the ERD 30
The relationship of release trends to success over time 32

IV. Other key factors: age, prior prison terms and misconduct 35
A. Older age at release, first prison terms and few misconducts help predict
success 36
1. Age 37
2. Prefix 37
3. Misconducts 38
a) Nonbondable misconducts 38
b) Total misconducts 39
B. Impact of age, prefix and misconducts on release time 40

V. The irrelevance of length of stay: more time in prison does not increase success 43
A. The Michigan data on length of stay 44
B. The national research 46
C. Increased length of stay under the new board 48
VI. How much does denying parole increase public safety? 50
A.
B.
C.
D.

Success rates of those kept an extra year or two (or three or four) 50
Impact of parole denial on crimes against persons 55
The minimal impact of parole decisions on crime rates 57
One and two-year parole denials: the large cost of small gains 59

VII. Implications of the research 62
VIII. Conclusion 64
Appendix A: Offense Categories 65
Appendix B: Methodology 67
Appendix C: Michigan and National Prison Commitment Rates 69
Appendix D: Additional Tables 70
Table A.
Table B.
Table C.
Table D.

Effect of Release Year Change on Four Offense Groups 70
Success & Failure by Mean Age 70
Success and Failure by Prefix 71
Failure Rates by Nonbondable Misconducts 72

Acknowledgments 73
Tables and Figures
Figure 1

Parole Approval Rates by Offense Group 11

Figure 2

Parole Decisions for Cases with High Parole Guidelines Scores – 2006 14

Table 1

Frequency of Offense Groups 18

Figure 3

Success/Failure Rates by Offense Group – All Cases 19

Table 2

New Offense for Which People Returned to Prison 20

Table 3

Returns to Prison for New Offenses by Original Offense Group 21

Table 4

Returns to Prison with New Sentences: Parolees & Max Outs 22

Table 5

Returns for New Conviction of Same Offense Type –
New York: Released 1985-2003 24

Table 6

Sex Offender Recidivism Rates 27

Table 7

Release Time by Offense Category 28

Table 8

Release Time Differences – Old and New Boards 29

Figure 4a

Success and Release Rates Compared by Offense – Old Board 31

Figure 4b

Success and Release Rates Compared by Offense – New Board 31

Figure 5

Success, Failure and Release Rates Over Time 32

Figure 6

Comparing PVNS with Property Crime Rate, 1986-1999 34

Table 9

Frequency Within Offense Groups of Characteristics Associated with Success 36

Figure 7

Success and Failure by Nonbondable Misconducts 39

Figure 8

Success and Failure by Total Misconducts 40

Table 10

Proportion Released on ERD by Key Factors 42

Figure 9

Zero Nonbondable Misconducts by Offense Groups – Parolees and Max Outs 43

Table 11

Overall Success Rate and Average Years Served, by Offense 44

Table 12

Minimum Sentence & Time Served by Success and Failure 45

Table 13

Average Years Served Relative to Minimum and Maximum Terms – Old Board and
New Board Compared 49

Figure 10

Percent of People Who Succeed or Fail by Release Time 51

Table 14

Success and Failure by Release Time and Offense Group 53

Table 15

Repeat Crimes Against Persons by Release Past ERD 56

Table 16

14-Year Impact of Releasing on ERD People Who Were Released
1 or 2 Years After 61

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
A Study of Prisoner Release and Recidivism in Michigan
Executive Summary
Widely-held assumptions about incarceration and parole are contradicted by Michigan data and
national research.
Assumption

Since Michigan’s overall rate of sending convicted felons to prison is below the
national average, Michigan does not incarcerate enough people.

Fact

Michigan’s overall prison commitment rate is below the national average because
the state imprisons fewer drug and other nonassaultive offenders. Its commitment
rate for assaultive and sex offenders is very high and, because it keeps those
offenders incarcerated for a long time, Michigan’s average length of stay in prison is
much higher than the national average.

Assumption

Longer prison sentences keep the public safer.

Fact

Increased length of stay does not reduce recidivism and may actually increase it.

Assumption

Half of all released prisoners commit new crimes and return to prison.

Fact

The majority of former prisoners do not return within four years for any reason,
even without the support services now provided through reentry programming. Of
those released for the first time from 1986-99, whether paroled or discharged on
the maximum, 63% did not return at all; fewer than 18% were returned with new
sentences for new crimes.

Assumption

Assaultive and sex offenders present a high risk to public safety.

Fact

People who commit assaultive or sex offenses are much less likely to reoffend than
those who commit financially-motivated crimes. Homicide and sex offenders rarely
commit new crimes against persons. Characterizing these offenders as “high risk”
confuses the harm from their past crimes with the likelihood they will commit new
ones.

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Assumption

Parole is a form of “early release” that was unintended and unanticipated by the
sentencing court.

Fact

In Michigan, judges set the minimum sentence according to legislative guidelines
that determine how much punishment is appropriate for the offense and the
offender. The minimum is often agreed to by the prosecutor during plea
negotiations. While the parole board can deny release until the prisoner has served
the maximum sentence, it cannot grant release until the person has served the
minimum. Parole at the first eligibility date indicates the board has found no
reason, based on current information, to extend incarceration beyond what was
imposed by the sentencing court.

Assumption

Decisions to deny parole are based on objective assessments of the risk of
reoffending.

Fact

Parole has routinely been denied based on the nature of the offense to people who
have served their judicially imposed minimum punishment and posed a low risk of
reoffending.

Assumption

People who are repeatedly denied parole have terrible institutional records.

Fact

While poor behavior in prison reduces the likelihood of parole for everyone,
assaultive and sex offenders are less able to earn release through good conduct.
Many of them “max out” despite having no serious history of institutional
misconduct.

Assumption

If parole decisions were made “correctly” and more prisoners were required to
serve their maximum sentences, all crime by parolees would be prevented.

Fact

The total prevention of new crimes by parolees would require the ongoing
incarceration of tens of thousands of people who would not, in fact, commit a new
offense if released. The fiscal cost would be astronomical and, because most crimes
are not committed by parolees, the impact on overall crime rates would be minimal.

Assumption

Denying parole for an extra year or two keeps the public substantially safer.

Fact

If, from 1986-99, everyone denied parole for up to two years had been released
when first eligible, 2,300 fewer beds per year would have been needed but returns
with new sentences would have increased only 1.7%.

2

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Purpose of the Research
The Citizens Alliance on Prisons and Public Spending examined 76,721 cases of Michigan prisoners
sentenced to indeterminate terms after 1981 and released for the first time from 1986 through
1999. The primary goal was to answer the following questions:
Does continuing to incarcerate people who have served their minimum sentences
actually improve public safety and, if so, to what extent and at what cost?
Specifically, does denying parole at the minimum only to release a person a year or two
thereafter have a substantial impact on re-offense rates?
The analysis also shed light on a number of other important questions regarding the nature and
cost-effectiveness of parole decision making. In particular, because offense groups were analyzed
separately, the impact of the nature of the crime on both release decisions and actual recidivism
could be identified.
The Data
The data was drawn from the management information system of the Michigan Department of
Corrections (MDOC). Nearly 47% of the people were released between 1986 and 1992 by the “old”
civil service parole board and 53% were released between 1993 and 1999 by the “new” appointed
board. Nearly 6% were discharged on their maximum sentences; the rest were paroled. They were
divided into nine offense categories: homicide, sex, assault, robbery, larceny, burglary, drugs,
weapons, and motor vehicles. They were similarly situated to the extent that all were released
before the abolition of disciplinary credits and pre-parole community placements, dramatic
revisions in drug sentences and the implementation of the Michigan Prisoner ReEntry Initiative
(MPRI). Recidivism was defined as returning to prison within four years of release either as a
technical parole violator (PVT) or with a new sentence for a new crime (PVNS).
Underlying Premises
Long length of stay caused Michigan prison growth.
•
•

From 1990-2005, Michigan’s average length of stay was 16 months longer than the average
of other Great Lakes states.
The Citizens Research Council (CRC) estimates that if Michigan’s average throughout that
period had been one year shorter, by 2005, Michigan would have had roughly 14,000 fewer
prisoners.

More parole denials are a primary cause of longer prison stays.
•
•
•

A 1992 change in the composition of the parole board from civil servants to appointees with
a mandate to “get tough” caused parole approvals to decline substantially.
CRC estimates that if pre-1992 release rates had continued through 2006, an average of
2,500 more people would have been released per year.
The proportion of prisoners who were past their earliest release date (ERD) and thus
eligible for parole grew from 16% (5,687 people) in 1991 to 31% (15,950 people) in 2006.

Citizens Alliance on Prisons and Public Spending

3

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
•

Increased parole denials reflected a conscious choice to keep assaultive and sex offenders
longer based on their crimes, not their risk of re-offending. The proportion of people who
scored low risk on the MDOC’s own parole guidelines who were granted parole fell from
81% in 1996 to 55% in 2006. For low risk sex offenders and people whose crimes involved
a death, 2006 grant rates were were only 13% and 28%, respectively.

The Findings
The research supports seven key findings and numerous related ones.
1. Success rates varied greatly by offense type. Homicide and sex offenders re-offended the
least while those whose crimes were financially motivated returned to prison the most.
a. Overall, about 63% of people released did not return to prison within four years. Nearly
20% were returned for technical parole violations and nearly 18% returned with new
sentences for new crimes.
b. Homicide (80.1%) and sex offenders (77.6%) had the highest success rates and their rates
of return with new sentences were below 8%.
c. At the other extreme, success rates for the larceny and burglary groups were below 55%
and returns with new sentences exceeded 24%.
2. New crimes against persons by released prisoners were rare. Among those who returned
with new sentences, larceny was the most common new crime.
a. Larceny, drugs and burglary accounted for nearly 63% of all the new convictions for which
people were returned.
b. Only 4.5% of more than 76,000 people released during the 14-year period were returned
for a new crime against a person.
c. Larceny and motor vehicle offenders were most likely to repeat their offenses.
d. Of 2,558 homicide offenders, 69 (2.7%) were returned for any new crime against a person
and 14 (0.5%) were returned for another homicide.
e. Of 6,673 sex offenders, 280 (4.2%) were returned for any new crime against a person and
204 (3.1 %) were returned for a new sex offense.
f. Returns with new offenses were nearly 50% greater for people who maxed out than for
parolees.
o Re-offense patterns among max outs were similar to those of parolees, with sex
(13.6%) and homicide (16.3%) offenders returning least often and larceny (35.6%)
and robbery (32.3%) offenders returning most often.
o Even among more than 4,000 max outs, three-quarters did not return in four years.
g. Numerous studies confirm that Michigan re-offense data is like that of other states.
3. The impact on parole decisions of characteristics well known to predict success varied
widely by offense.
a. Greater age at release, first incarceration and little or no institutional misconduct, were
associated with greater success within every offense group.
b. These characteristics occurred most frequently in the groups with the lowest reoffense
rates but that did not explain why success rates differed so much among groups.

4

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
c. While having an “A” prefix (no prior Michigan prison term) and a good institutional record
increased the chance of release when first eligible, they by no means guaranteed it,
especially for people convicted of assaultive and sex offenses.
4. Parole was commonly denied based on the nature of the past crime, not the actual
likelihood the person would commit a new one.
Overall
a. Overall, 61.4% of released prisoners were paroled when they first became eligible.
However, this ranged from 78.8% of motor vehicle offenders and 75.5% of drug offenders
down to 32.6% of sex offenders.
b. Overall, an additional 15.2% were released one year after their ERD and another 14.6%
were released within two years of their ERD. Thus, on average, more than 91% were
released within two years of first eligibility.
c. In the sex offender group, fewer than 74% were released within two years. Nearly 14%
were required to max out.
Differences between the “old” and “new” boards
After the parole board was restructured in late 1992, disparities in release time increased
dramatically for people convicted of assaultive and sex offenses.
a. Overall, 2,560 people who would have been released on or within one year of their ERD by
the old board were continued for two, three, four or more years by the new board.
b. An additional 899 people were held to their maximums. The proportion of sex offenders
required to max out nearly tripled.
c. Increased parole denials by the new board caused the average length of stay to grow by 4.8
months, but the amount of change varied greatly by offense.
d. Changes in time served due to parole board decisions (as opposed to longer sentences)
were minimal for drug, motor vehicle, larceny and burglary offenders. However, the board
increased length of stay by nearly 10 months for robbery offenders, 12 months for homicide
offenders and 16 months for sex offenders.
o Overall, under the old board, people served 92.6% of the average minimum
sentence while under the new board they served 106%. Sex offenders went from
serving 94% of their average minimum to 124%.
o For the seven-year period from 1993-1999, the new board’s policies resulted in
15,601 additional years served, requiring 2,229 more beds per year than would
have been required under the old board’s policies. Nearly 30% of the additional
beds were for sex offenders.
5. Sheer length of time served bears no relationship to success.
a. There is no magic number of years in prison that will guarantee success. People in different
offense groups served similar amounts of time but had very different outcomes.
b. Serving more time in prison does not improve success upon release. On the contrary, in the
majority of offense groups, serving more time was associated with failure.
c. Within offense categories there is not a lot of difference in the amount of time served by
those who succeed and those who return to prison, either as technical violators or with new
sentences for new crimes.
Citizens Alliance on Prisons and Public Spending

5

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
d. Numerous studies done over several decades confirm that keeping people in prison longer,
especially for no more than a few additional years, does nothing to enhance public safety
and may be counterproductive.
6. Incarcerating people for an additional year or two after they reached their earliest
release date had very little impact on success rates in general and returns with new
sentences in particular.
On average, the people continued for one year were somewhat less successful than those released
when first eligible, but keeping people for two, three or four years made virtually no additional
difference. However, variations among offense groups were substantial.
a. Among those released one year after their ERD, the decline in success rates ranged from
1.8% for motor vehicle offenders to 14.2% for the homicide group.
b. Among those released two years after their ERD, the additional decline in success ranged
from zero for homicide and assault offenders to 5.9% for the robbery group.
c. The extent to which continuances for a year or two reflected accurate risk prediction also
varied by offense group. For five groups (homicide, sex, motor vehicles, assault, drugs),
those held for an additional year still had success rates from 63%-78% and those continued
for two years still had success rates of 60-75%.
To avoid increasing returns for new crimes against people from 4.5% to 6.9%, 9,664 assaultive and
sex offenders who would not have returned with any new offense were imprisoned from one to
four additional years after they became eligible for parole.
a. In the homicide group, 849 of 885 people who were kept up to four years past their ERD, or
95.9%, were not returned to prison for any new crime against a person, compared to 98.4%
of people released when first eligible.
b. In the assault group, 2,456 of 2,646 people who were kept up to four years past their ERD,
or 92.8%, were not returned to prison with a new sentence for any new crime against a
person, compared to 95.7% of those released when first eligible.
c. In the sex offense group, 3,807 of 3,998 people who were kept up to four years past their
ERD, or 95.2%, were not returned to prison within four years with a new sentence for any
new crime against a person, compared to 97.1% of those released when first eligible.
d. In the robbery group, 3,521 of 3,894 people who were kept up to four years past their ERD,
or 90.4%, were not returned to prison within four years with a new sentence for any new
crime against a person, compared to 93.7% of those released when first eligible.
7. If everyone denied parole for up to two years had been released when first eligible, it
would have saved nearly 33,000 beds over 14 years, or more than 2,300 beds a year, on
average. The impact on overall arrest rates would have been virtually imperceptible.
a. The overall success rate of people released on their ERD would have declined by fewer than
three points.
b. The overall rate of returns with new sentences would have increased by only 1.7 points.
c. Annual arrests would have increased by less than 0.4%.
d. The cost savings, in today’s dollars, would have exceeded $1 billion for the entire period and
averaged nearly $74 million a year.

6

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?

Implications of the findings
1. Relying on lengthy incarceration as a crime control strategy is not a cost-effective policy.
2. Michigan can safely reduce its incarceration rate to the levels of comparable states while
continuing to adequately punish the most serious offenders.
3. Sentence length should reflect appropriate punishment but should not assume that exceedingly
long sentences are necessary to prevent reoffending. Sentencing guidelines for assaultive, sex
and habitual offenders should be reevaluated.
4. Substantially increasing the rate of parole on the earliest release date would reduce the
prisoner population without threatening public safety. In particular, eliminating the frequent
delay of parole for one or two years would save significant amounts of money without releasing
anyone “early” or causing more than a minimal change in success rates.
5. Establishing a statutory presumption of parole on the minimum for all prisoners, subject to
individualized risk assessments, would avoid unnecessarily imprisoning people who have
served their punishment and are at low risk for reoffending.
6. Current parole guidelines which account for institutional misconduct, prior record and current
age are adequate to identify an individual’s statistical risk for reoffending. However, they
should be adjusted to accurately account for the nature of the offense.
7. Since time served does not relate to success on release, the use of sentence reductions or
community placements as incentives for good behavior would not reduce public safety and
might increase it by encouraging self-discipline and program participation in prison and a
structured reentry to the community.
8. The application of stringent supervision conditions to whole categories of parolees and of
employment and residence barriers to whole categories of people with criminal convictions is
not necessary to protect the public. The selective use of parole conditions and of work and
residence related prohibitions, based on actual risk, would be fairer and more cost-effective.
9. Since many former prisoners will not reoffend in any event, reentry efforts will be most costeffective if targeted at the people who are at highest risk for reoffending, such as by providing
job training and placement for those who committed financially motivated crimes and
treatment for those suffering from mental illness or substance abuse.
10. In order to make realistic and responsible policy choices at every stage of the criminal justice
system, we need to better understand what distinguishes those former prisoners who do not
reoffend from those who do. Research should focus not only on which post-conviction
programs are effective but on which offender characteristics correlate with success even in the
absence of programs.

Citizens Alliance on Prisons and Public Spending

7

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?

REPORT
I. Framing the question: does denying parole keep the public safer?
With the nation’s seventh largest state prison system1, Michigan’s prisoner population has grown
from less than 15,000 in 1984 to more than 47,000 in 2009. The budget of the Michigan
Department of Corrections (MDOC) has reached $2 billion, diminishing the state’s ability to provide
other important services. As the pressure to rein in corrections spending mounts, so does debate
over whether the number of prisoners can be reduced while maintaining public safety. The answer
depends on two factors: how big the risk to public safety actually is and how much risk the public is
willing to tolerate.
The present research was designed to clarify the actual risk of paroling people who have served
their minimum sentences. In the 14 years from 1986-1999, nearly 77,000 people were released
from Michigan prisons. Roughly 61% were paroled when they first became eligible. The rest were
required to serve one, two or in some cases many years beyond their earliest release date. If those
held for one or two additional years had been released when first eligible, roughly 2,300 fewer beds
would have been needed per year at a savings, in today’s dollars, of more than $1 billion for the
entire period. MDOC data was analyzed to determine who was denied parole, for how long, and
what impact these decisions had on public safety.
A. Length of stay drives Michigan’s prison growth
Michigan’s prison growth is part of the national “tough on crime” trend that led to 1.6 million
people being in state or federal prisons in 2008.2 Faced with violent crime rates that grew steadily
from 1961 until they peaked in 1991, Americans rejected concepts of rehabilitation in favor of more
and longer prison terms. Criminal justice policies were adopted that resulted in more and longer
prison terms. Ironically, prison growth was also fueled by a policy meant to be rehabilitative, or at
least, benign -- the deinstitutionalization of the mentally ill. Lacking adequate community-based
treatment, a great many people with mental health needs end up in the criminal justice system. A
recent analysis concludes that 80-85% of prison expansion over the last 25 years is not the result of
more crime but of public policy choices about who goes to prison and how long they stay.3
Michigan’s choices have been particularly punitive, especially compared to similar states. At 505
prisoners per 100,000 citizens, in mid-2008 Michigan was tied with Virginia for the twelfth highest
incarceration rate in the country. The average of the nine states in the Northeast was 305. The 12
states in the Midwest averaged 395. Most states with higher rates were in the South. 4
Heather C. West and William Sabol, Prison Inmates at Midyear 2008 – Statistical Tables, U.S. Department of Justice,
Bureau of Justice Statistics (Washington, D.C. March 2009), Table 2.

1

2

Id.,

Steven Raphael and Michael A. Stoll, “Why Are So Many Americans in Prison?” in Raphael and Stoll (eds.), Do Prisons
Make Us Safer? The Benefits and Costs of the Prison Boom (New York, 2009), pp 27-72, at p 65. See also, Alfred Blumstein
and Allen J. Beck, “Reentry as a Transient State between Liberty and Recommitment” in Jeremy Travis and Christy Visher
(eds.), Prisoner Reentry and Crime in America (Cambridge, 2005), pp 50-79 at p 50 (except for drug offenses, none of
increased incarceration rate from 1980-2001 is attributable to increased crime rates).
3

4 West and Sabol, note 1 supra, Table 10. The states with higher incarceration rates than Michigan’s are Louisiana (858),
Mississippi (749), Texas (668), Oklahoma (668), Alabama (619), Arizona (565), Florida (548), Georgia (542), S. Carolina
(537), Missouri (515) and Arkansas (507).

8

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
In its June 2008 analysis of Michigan’s prison growth over the last 30 years, the Citizens Research
Council of Michigan (CRC) explained that Michigan’s high incarceration rate occurs not because
Michigan sends proportionally more offenders to prison but because it keeps them incarcerated
longer. In fact, Michigan became a leader in the use of community-based sanctions in 1988, when it
established the Office of Community Corrections.5 In 1998, it implemented sentencing guidelines
that discourage prison sentences for less serious offenders and in 2002, it reformed its drug laws.
Thus, Michigan sends a lower proportion of people convicted of felonies to prison than other states
because it places more lower-level offenders in county jails and other community placements. 6
However, CRC reported, those people who are sentenced to prison in Michigan spend substantially
more time there than prisoners in other states. From 1990-2005, the average length of stay in
Michigan was 16 months longer than the average of other Great Lakes states. By 2005, the average
estimated length of stay was 24.6 months for Great Lakes states, 27.4 months for the United States
as a whole and 43.5 months for Michigan. Michigan’s length of stay increased 57% since 1981,
when it was approximately 28 months, i.e., roughly what the national average is now.7
CRC characterized the impact of increased length of stay on the growth of Michigan’s prison
population as “profound.” It estimated that if, from 1990-2005, Michigan’s annual average length of
stay had been one year shorter, by 2005, Michigan would have had roughly 14,000 fewer prisoners,
an incarceration rate of 351, a corrections workforce with 4,700 fewer employees and annual
expenditures of $403 million less.8
Michigan’s longer length of stay has resulted from a mixture of sentencing, good time, parole and
parole revocation policies. It was initially fueled by several choices made in the late 1970s: a
mandatory two-year penalty for possession of a firearm during the commission of a felony, long
mandatory minimum sentences for drug offenders, the increased use of sentence enhancements for
repeat offenders and the elimination of generous “good time” awards for in-prison conduct. While
good time was replaced by a more limited amount of “disciplinary credit” in 1982, disciplinary
credits were abolished, in turn, by “truth-in-sentencing” legislation in 1998, forcing Michigan
prisoners to serve every day of their minimum sentences, regardless of their conduct or program
participation. The sentencing guidelines that reduced prison commitments for some offenders also
increased minimum prison terms for those committed for assaultive and sex offenses.
Decreasing parole approval rates were another important consideration. The CRC called them a
“primary contributor” to increased length of stay. The Council of State Governments has noted:

Citizens Research Council of Michigan, Growth in Michigan’s Corrections system: Historical and Comparative Perspectives
(Livonia, MI. June 2008) [hereinafter CRC report], at pp 21-22.

5

6Nationally, in 2004, 40% of all convicted felony offenders were sentenced to prison while, in 2007, the rate in Michigan
was 22.8%. However, when jail and prison sentences are collapsed, the national and Michigan incarceration rates are
virtually identical, 70.0% and 70.7%, respectively. Notably, Michigan’s imprisonment rate for drug offenders was 12.1%
compared to the national average of 37.0%. See Michigan Department of Corrections, 2007 Statistical Report, Tables A1
and A1b, at http://mich.gov/documents/corrections/2007_Statistical_Report_277947_7.pdf and Matthew Durose and
Patrick Langan, Felony Sentences in State Courts, 2004 (Bulletin and statistical tables), U.S. Department of Justice, Bureau
of Justice Statistics (Washington, D.C. July 2007), Michigan’s commitment rate for assaultive and sex offenses appears to
be at or above the national average, although differences in the way crimes are categorized make comparisons imprecise.
For more detailed information on Michigan’s prison commitment rates, see Appendix C.
7

CRC report, note 4 supra. at p 11.

8

Id. at p 12.

Citizens Alliance on Prisons and Public Spending

9

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
“[T]he average minimum sentence imposed by Michigan judges for various violent crimes is
comparable to the average length of incarceration for people nationally. Nevertheless, people
sentenced to prison in Michigan for various violent crimes stay in prison considerably longer…”
The Council attributed the “overwhelming difference between the lengths of time served nationally
and in Michigan…to the unique level of discretion available to the state’s parole board.”9
B. Parole denials impact length of stay
Unlike the 17 states that have abolished discretionary parole, Michigan maintains an indeterminate
sentencing scheme. With some exceptions,10 the legislature sets a maximum sentence for the crime
by statute, the judge sets a minimum sentence for the individual defendant, and the parole board
decides whether and when the person will be released at some point in between. Thus, for
instance, a person sentenced today to a prison term of 2-10 years for burglarizing a store cannot be
paroled until s/he has served the two-year minimum, may be released at any point after two years,
but cannot be kept incarcerated after reaching the 10-year maximum.
Parole at the earliest release date (ERD) is not “early release;” it is release after serving the
punishment a judge determined to be appropriate for the crime. The board has the absolute
discretion to make someone serve the maximum sentence, regardless of the judge’s intention, the
understanding embodied in a negotiated guilty plea or the prisoner’s efforts to earn release while
incarcerated. Release on the ERD indicates that the parole board has found no reason, based on
current information, to extend incarceration.
In 1992, after a series of rape-murders by a paroled sex offender, the composition of the parole
board was changed. In lieu of corrections professionals who tended to focus on the prisoner’s
growth while incarcerated, ten appointees were given a mandate to “get tough.” CRC examined the
release decisions made from 1976-2006 and compared those made by the old and new boards. It
found that the average annual parole approval rate declined from 66% in the period before 1992 to
54% in the years after. CRC estimates that “[h]ad the parole rate experienced during the years
before 1992 continued through 2006, an average of 2,500 more prisoners would have been
released per year.”11 Instead, the proportion of prisoners who had served their minimum sentences
and were beyond their earliest release date (ERD) grew from 16% (5,687 people) in 1991 to 31%
(15,950 people) in 2006. 12

9 Council of State Governments Justice Center, Justice Reinvestment in Michigan: Analyses of Crime, Community Corrections
and Sentencing Policies (New York, 2009)[hereinafter CSG report], at pp 8-9.

First-degree murder carries a mandatory sentence of life without parole, committing a felony while in possession of a
firearm carries a mandatory consecutive two-year sentence (for the first offense) and a number of the most serious
offenses, including second-degree murder, armed robbery, first-degree criminal sexual conduct and assault with intent to
commit murder, carry “life or any term” penalties that allow the judge to choose between life with eligibility for parole
and an indeterminate sentence for which the judge sets both the minimum and maximum.

10

11

CRC report, note 4 supra, at p 12

Id. at p. 13. This figure does not include parolable lifers, who technically do not have an earliest release date because
they do not have a minimum sentence. To determine the actual number of people who could be paroled, one must add at
least 800 lifers who became eligible under Michigan’s “lifer law.” See Citizens Alliance on Prisons & Public Spending, The
high cost of denying parole: an analysis of prisoners eligible for release (Lansing, Nov. 2003).

12

10

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
A 1997 MDOC report entitled Five Years After: An analysis of the Michigan Parole Board since 1992
made it clear that these consequences were quite intentional. In a prologue, then Director Kenneth
McGinnis stressed that the new board was “much less willing to release criminals who complete
their minimum sentences,” was requiring more people to serve their maximums, was “clamping
down especially hard on violent and assaultive offenders” and was imposing “significantly higher”
punishment on parole violators. One result of this approach was that in 2005, of 11,770 total
releases, 1,486, or 12.6%, were people who had “maxed out” on an indeterminate sentence.
Mr. McGinnis emphasized that the board was less willing to parole “violent and assaultive”
offenders. As Figure 1 shows, the nature of the offense always played a critical role in parole
decision-making. People convicted of violent and sexual offenses were released at a lower rate than
drug and other nonviolent offenders by the old board as well. But the disparities became far
greater. In 1990, the parole approval rate was 81.3% for drug offenders and 46.5% for sex
offenders, a difference of nearly 35 points. In 2008, the rate was still over 80% for drug offenders,
but for sex offenders it was 21.0%. The difference had grown to nearly 60 points. Violent offenders
saw their approval rate decline from over 60% to 43%.13

It should be noted that the 2008 approval rate for sex offenders was the highest it had been since 1996. That rate was
as low as 10.3% just two years earlier. For violent offenders, the 2008 approval rate was the highest since 1999.

13

Citizens Alliance on Prisons and Public Spending

11

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
C. Past harm and future risk: a critical distinction
The reluctance to release people convicted of violent or sex offenses is rooted in repugnance for the
crimes and the desire not to allow those crimes to be repeated. However, it is not necessarily
rooted in the actual likelihood that the crimes would be repeated. In the public debate over when
prisoners should be released, the terms “high risk” and “violent” or “assaultive” tend to be used
inter-changeably. That is, people who were convicted of crimes against people, like murder,
assault, robbery and sexual assault, are characterized as high risk while people convicted of crimes
against property or public order are considered low risk. This confuses the more serious harm that
assaultive and sex offenders caused by their past crimes with the likelihood that they will commit
new ones.
The degree of harm caused and the culpability of the defendant are the key considerations at
sentencing. The trial judge, in accordance with sentencing guidelines, looks backward to select a
punishment that is proportional to the offense and the offender. In contrast, the parole board’s role
is to look forward and assess the likelihood of future harm if the offender is released. MCL
791.233(1)(a) defines the board’s responsibility in this way:
A prisoner shall not be given liberty on parole until the board has reasonable assurance,
after consideration of all of the facts and circumstances, including the prisoner's mental and
social attitude, that the prisoner will not become a menace to society or to the public safety.
To steer the new board towards objectively assessing the risk of reoffending, legislation enacted in
1992 required the MDOC to develop parole guidelines “that shall govern the exercise of the parole
board’s discretion…” 14 Michigan’s adoption of parole guidelines is part of a national trend towards
making parole decision-making more consistent and more accurate at predicting risk. Academic
research has repeatedly shown that statistical risk prediction, also called actuarial based
assessment, is substantially more reliable than subjective professional or “clinical” judgments,
especially when supplemented with individualized information that cannot be built into general
scoring instruments. 15
The potential for error in subjective risk assessment has multiple causes. These include the
pressure to make decisions quickly, the anxiety associated with highly consequential decisions, the

influence of work group norms and senior officials, the human tendency to disregard or
downplay information not consistent with one’s preconceptions, and a tendency “to attach undue
importance to easily recalled information, such as highly publicized cases, when assessing the risks
with a particular case.” “Excess conservatism” is also caused by “biased feedback.” Parole board
members are much more likely to hear about people who fail than those who succeed, causing them
to become increasingly cautious.16 Because parole boards seek to minimize parole failures,
professional judgments tend to over-predict dangerousness, resulting in fewer releases.17
14

MCL 791.233e (1)

15 Daniel Glaser, Classification for Risk, 9 Crime and Justice 249-291 (1987); Stephen Gottfredson, “Prediction: An
Overview of Selected Methodological Issues” in Gottfredson and Tonry, eds., Prediction and Classification (Chicago, 1987),
pp 21-51; R. Karl Hanson and Kelly Morton-Bourgon, Predictors of Sexual Recidivism: An Updated Meta-analysis, Report
2004-02, Department of the Solicitor General Canada (Ottawa, 2004), at p 3.
16

Glaser, note 15 supra, at pp 252-255.

17

Harry Allen et al, “Granting Parole” in Probation and Parole in America, (New York, 1985), at p 105.

12

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Criminologist James Austin summarizes the situation more bluntly. Characterizing professional
judgment as “by far, the least accurate risk assessment method,” Austin says such judgments are too
often just “gut” reactions that may “vary from expert to expert on the very same offender.” 18
Depending on their parole guidelines scores, Michigan prisoners are classified as high, average or
low probability of parole.19 Administrative Rule 791.7716 (2) requires that guidelines ranges be set
to ensure, to the extent possible, “that prisoners who score in the high probability of parole range
do not exceed an assaultive felony recidivism rate of 5%.” The statute places heavy reliance on the
guidelines as a screening tool to be supplemented by the clinical judgment of parole board
members. MCL 791.233e (6) states:
The parole board may depart from the parole guideline by denying parole to a prisoner who
has a high probability of parole as determined under the parole guidelines or by granting
parole to a prisoner who has a low probability of parole as determined under the parole
guidelines. A departure under this subsection shall be for substantial and compelling
reasons stated in writing.
However, despite the statutory presumption of release, the proportion of people with high
probability scores who are granted parole has steadily fallen, from 81% in 1996 to just 55% in
2006.
Prisoners often maintain that the reasons given for denying them parole despite a low statistical
risk of reoffending are not truly “substantial and compelling.” Rather, they say, even though the
nature of the offense was the critical factor in setting the minimum sentence, the board reacts to the
offense and effectively resentences people to increased punishment. The lack of any process for
reviewing parole denials makes it difficult to systematically assess the board’s compliance with the
statutory standard.20
Recent MDOC data confirm that crime type does have a substantial impact on parole decisionmaking regardless of the person’s current risk for reoffending. In 2006, of 6,700 people with high
probability scores, 2,993, or 44.7%, were denied parole. Figure 2 shows that denial rates varied
dramatically by offense group. Of the drug and nonviolent offenders, 13% or fewer were denied.
However, despite their low statistical risk for reoffending, the board found substantial and
compelling reasons to continue nearly 87% of the high scoring sex offenders and nearly 72% of the
high scoring homicide offenders.

James Austin, The Proper and Improper Use of Risk Assessment in Corrections, 16 Federal Sentencing Reporter 1, Feb.
2004.

18

19 The parole guidelines award both positive and negative points for specific factors in seven broad areas: offense, prior
record, institutional misconduct, statistical risk, age, program participation and mental health. Most areas are weighted
differently depending on how long the person has been incarcerated.
20 Although prosecutors and victims can appeal decisions to grant parole, board decisions to deny parole cannot currently
be appealed by the prisoner to the courts, nor is there any administrative review process. For illustrative cases of people
with high probability scores who were denied parole, see the website of the Citizens Alliance on Prisons and Public
Spending, www.capps-mi.org/parole%20denied.htm

Citizens Alliance on Prisons and Public Spending

13

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?

Figure 2. Parole Decisions for Cases with High
Parole Guidelines Scores - 2006 *
.9

86

.6

.7

90

.8

89

86

100

71

80
55

.6

.4

60

44

Percent

70

50

Approved

40

28

.3

Denied

.4

30
13

13

10

.1

20

.2

10
0
Drug

Non-Violent,
Non-Drug

Other Violent
Crime

Crime Involving
Death

Sex

*Crime type passed on a progression through all active sentences – if sex crime, then sex crime; if not, then crime
involving death; if not, then other violent; et cetera.
Source: Michigan Department of Corrections
(5/30/2007)

Basing parole decisions solely on crime type led directly to the tripling of prisoners past their
earliest release date that CRC described. In a 2005 presentation to the legislature, the MDOC
explained why the population of 48,557 included 15,083 people who were past their ERD. Noting
that 29% of the group had committed nonviolent crimes, the department itemized seven
aggravating factors, including prior record, institutional misconduct and behavior during prior
releases, to explain why these prisoners had been denied parole. As to the other 71%, the
department merely said: “serving sentences for sex offenses, murder or other violent crimes” in the
apparent belief that no other explanation was needed.21
The continuing primacy of crime type over risk assessment was demonstrated most recently by a
much-awaited consultant’s report. On Jan. 22, 2009, the Council of State Governments Justice
Center (CSG) presented a series of policy options for Michigan criminal justice, including several
designed to reduce spending on corrections. One of the CSG proposals was that the parole board be
required to release people who had served their minimum sentences subject to certain exceptions.
Release could be delayed for failure to complete required programs or institutional misconduct but
would be mandatory at 120% of the minimum. Notably, the requirement would not apply at all to
people who scored very high risk of reoffending on a validated risk assessment instrument or who
were serving for crimes with statutory maximum penalties of life or any term. That is, regardless of
their actual sentence, institutional record or low reoffense risk, the parole board would retain the
discretion to “max out” anyone convicted of such crimes as second-degree murder, first-degree

Michigan Department of Corrections, Prison Population & Capacity, presentation at Senate Judiciary Committee Hearing,
Feb. 1, 2005.

21

14

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
criminal sexual conduct and armed robbery. Thus corrections experts ratified the parole board’s
practice of elevating the nature of the offense over the objective risk of re-offending.22

II. Overview of the research
All the policies that lengthen prison stays, so costly in both human and fiscal terms, were
implemented on the assumption that keeping people incarcerated longer keeps the public safer.
This research tests that assumption in the context of parole decision-making by comparing the
recidivism rates of people who were released when they first become eligible with those of people
who were denied release for some number of years. The data also allows for the examination of
important related questions, including the types of crimes for which people returned to prison, the
extent to which low-risk offenders were denied parole, the impact of policy choices by the old and
new boards, and the relationship of factors that correlate with success to offense type and release
decisions. Because offense groups are analyzed separately, the critical impact of the nature of the
crime on release decisions and actual recidivism is not lost in a mountain of undifferentiated data.
A. Defining recidivism
The concept of “recidivism” has been given many meanings and measured in different ways. All
measurements are imperfect; each has its advantages and disadvantages. (For further discussion,
see Appendix B)
Whether they have returned to prison is the most reliable and readily available information about
former prisoners. For this research, recidivism is defined by whether the person returned to prison
in Michigan within four years of release, either as a technical parole violator or with a new
sentence.23 The distinction between these two groups is maintained throughout and only those
who returned with new sentences are assumed to have committed new crimes. Success is defined
by the absence of a return to prison.
Recidivism rates are affected by the method of release. Parolees may be returned either for
violating the rules of supervision (parole violator technical -- PVT) or because they have been
convicted and sentenced for a new crime (parole violator new sentence -- PVNS). People who are
discharged after completing their maximum sentences are never placed on parole and so, by
definition, can only return to prison for new crimes. Similarly, people who have completed parole
and been released from MDOC custody can only return to prison for new crimes.

22 Council of State Governments Justice Center, Justice Reinvestment in Michigan: Policy Options to Deter Crime, Lower
Recidivism, and Reduce Spending on Corrections (New York, 2009).

Because of the date on which the database was run, people released between May and December 1999 had a follow-up
period from 3.3 to 4 years.

23

Most periods of parole supervision are initially set for two years and most parole revocations for technical violations
occur within that time. However, the parole period can be extended up to the maximum sentence and extensions are a
common way of dealing with supervision violations while keeping the person in the community.
Returns for new convictions were counted if they occurred within four years of release whether the person was on parole,
had completed parole or had served the maximum sentence and never been on parole. For convenience, the abbreviation
PVNS is used to refer to all these circumstances.

Citizens Alliance on Prisons and Public Spending

15

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Technical parole violators are not a homogeneous group. Some have engaged in noncriminal
conduct, such as missing treatment sessions, associating with prohibited people or crossing the
boundaries set by electronic monitors. Others have engaged in conduct that could have been
prosecuted but was not.24 The MDOC database did not identify the reasons for technical violations.
Once a violation of a parole condition is found, the parole board has absolute discretion in deciding
whether to revoke parole. Over time, policies about which sanctions to apply to which violations
have varied. At some points, parole revocation was common for violations that today would bring
progressive community-based sanctions. Parole revocation policies tend to vary with the
philosophy of the board, the history of the offender, the availability of alternatives and the
pressures on prison bedspace.
At the other extreme, the board has no control over the number of former prisoners who come back
because they were convicted of new felonies and received new sentences to prison. Where there is
sufficient evidence that a parolee has committed a serious new offense, particularly a crime against
a person, presumably prosecutors do not simply rely on the board to revoke parole. If they can
prove that someone on parole for rape, robbery, murder or assault has hurt a new victim, it is
unlikely that they would not seek a new conviction. Thus, while the category of return to prison
with new sentence does not capture new crimes that did not result in imprisonment at all or that
were treated only as technical parole violations, it is a reliable measure of trends in reoffense rates,
particularly for the most serious crimes.
B. The data
The data was drawn from the MDOC database, known as the Corrections Management Information
System (CMIS), as it was constituted on May 6, 2003. The data consists of 76,721 cases of people
sentenced to prison for indeterminate terms after 1981 and released for the first time from 1986
through 1999.25 It includes 72,054 people who were paroled at some point during their sentences
and 4,667 who were discharged after serving their maximum sentences (maxed out).
As a result of this time frame, all the cases are similar on a number of critical points.
1. All involve sentences imposed after the Michigan Supreme Court adopted sentencing
guidelines but before sentencing guidelines were enacted by the Legislature.
2. All the sentences were imposed after generous awards of “good time” were abolished in
1978, but before “truth in sentencing” completely eliminated both “disciplinary credits” and
pre-parole placements in community programs.
3. All the releases occurred before the MDOC began its current prisoner reentry initiative.
4. All the releases in drug cases occurred before substantial changes to the drug laws.
The burden of proof is lower and the proceedings are less formal at parole revocation hearings than in court. If the
evidence is weak or if the new crime is not very serious, it is easier and cheaper for the county to simply let the board
revoke parole. The board will have complete discretion as to whether and when to reparole. If the prisoner still has many
years left before serving the maximum sentence, he or she can be returned to prison for years as a technical parole
violator for conduct that might have brought only a year or less in the county jail if prosecuted. In some cases where
parolees have received short jail sentences, they have subsequently been returned to prison as technical violators for the
same underlying conduct.
24

It does not include people serving only flat terms for felony-firearm convictions where the board has no discretion to
exercise. It also does not include parolable lifers who became eligible for parole after serving ten years but who
technically do not have a minimum sentence with an “earliest release date” (ERD).

25

16

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
These key internal consistencies, along with the sheer size of the data base and the long period of
years included, make this data uniquely valuable.
An additional advantage is that the dataset permits comparison of two clearly demarcated periods
in Michigan parole policy. The composition of the parole board was changed in late 1992. There
are 36,032 cases where release occurred in the seven years from 1986-1992 under the “old” board
and 40,689 cases where release occurred from 1993-1999, under the “new” board.26 Thus the new
board made 4,600 more decisions to release in a seven-year period than the old board did. The
figures below include only the nine offense groups analyzed in this report. They show that annual
releases began climbing dramatically after 1987, when growth in the prison population began to
increase the number of prisoners who became eligible for parole. The annual number of releases
actually peaked in 1991 and dropped markedly after 1993.
Old Board
1986
1987
1988
1989
1990
1991
1992

2,748
3,184
4,036
5,356
6,246
6,482
6,428
34,480

New Board
1993
1994
1995
1996
1997
1998
1999

6,228
5,297
5,469
5,514
5,092
5,852
5,562
39,014

There are other differences between each board’s cases:
1. More prosecutions and greater incarceration rates for certain offenses, such as drunk
driving, caused the new board to see more of those kinds of cases.
2. The elimination of generous “good time” credits after 1978 meant that people who received
long sentences during the 1980s did not become parole-eligible until the 1990s.
3. In addition, since the database consists only of cases where the sentence was imposed after
1981, anyone with a minimum sentence longer than 13 years would not have become
eligible until after 1992, at the soonest, and thus was pushed into the pool of cases
considered by the new board.
C.

The frequency of offense groups

The cases were divided into nine offense groups, based on the crime for which the person was
serving when released.27 Relatively uncommon miscellaneous offenses that do not fit into the nine

26 No attempt was made to segregate cases actually decided by the new board in Nov. and Dec. 1992. It would have been
impossible to determine which cases were already in progress and had input from both boards. Given the total size of the
database, it is unlikely that shifting the small quantity of cases in those two months would affect the results.

All groups include attempts and conspiracies to commit the underlying crime and people sentenced as habitual
offenders for convictions of these offenses. The data reflects the most serious offense for which the person was convicted.
It does not reflect additional convictions of another type for which a shorter sentence was imposed.

27

The data also does not reflect initial charges that may have been reduced or dismissed, whether as the result of plea
negotiations or for other reasons. The original charges are not part of the MDOC database. Even if they were, they would

Citizens Alliance on Prisons and Public Spending

17

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?

groups were collapsed into a tenth group called “other.” 28 As Table 1 indicates, for some
offense groups the frequencies were substantially different for the old board and the new board.
Drug and motor vehicle cases each constituted a larger share of the new board’s releases.
Conversely, the proportions of larceny and burglary cases were both markedly smaller.
Table 1. Frequency of Offense Groups
Old Board

All

Larceny
Drugs
Burglary
Robbery
Sex
Assault
Motor Vehicle
Homicide
Weapons
Other
TOTAL

Number
of Cases

Percent
of Total

Number
of Cases

Percent
of Total

16,194
16,144
10,601
9,010
6,675
6,529
3,199
2,558
2,523
2,607
76,040

21.3
21.2
13.9
11.9
8.8
8.6
4.2
3.4
3.3
3.4
100.0

8,779
6,647
5,865
4,401
3,005
2,581
738
1,171
1,243
1,283
35,713

24.6
18.6
16.4
12.3
8.4
7.2
2.1
3.3
3.5
3.6
100.0

New Board
Number
of Cases
7,415
9,497
4,736
4,609
3,670
3,948
2,461
1,387
1,280
1,324
40,327

Percent
of Total
18.4
23.6
11.7
11.4
9.1
9.8
6.1
3.4
3.2
3.3
100.0

Some offense groups reflect a broad spectrum of behavior, ranging from offenses carrying
maximum sentences of four years to those carrying up to parolable life. Other groups contain a
concentration of offenses of similar degree of seriousness.
For instance, three-quarters of the drug convictions were for delivery or possession of quantities
less than 50 grams. Nearly 80% of the burglary convictions were for breaking and entering a nonresidential structure. More than 70% of the robberies were armed. In the sex offense group, 63% of
the cases were criminal sexual conduct in the second or third degree, both of which carry a 15-year
maximum sentence. More than 75% of the assaults were either felonious assault (a 4-year
maximum) or assault with intent to commit great bodily harm (a 10-year maximum). Nearly all the
motor vehicle convictions were for driving under the influence. Nearly all the weapons offenses
carry a 5-year maximum. Notably, nearly two-thirds of the homicide convictions were for
manslaughter, not murder. This is attributable, at least in part, to the large number of people
convicted of murder who are serving life terms that are not included in the data base. 29
(For more detailed information about each offense category, see Appendix A.)
not be a reliable indicator of the person’s actual conduct, since they may have been reduced or dismissed because they
were unsupported by the evidence.
28

The total does not include 681 cases with missing data.

While a parolable life sentence may be imposed for various crimes, including armed robbery and first-degree criminal
sexual conduct, such sentences are common for second-degree murder. For example, in 1992, mid-way through the study
period, 3,181 people were serving life terms. Of these, 75% were convicted of either first or second-degree murder.
Parolable life terms had been imposed in 25% of the second-degree murder cases, but only 8.7% of the criminal sexual
conduct cases and 5.3% of the armed robberies.
29

18

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?

III. The significance of crime type
The data demonstrate how parole policies have tended to focus on past harm. Although, as is true
nationally, most Michigan prisoners incarcerated for assaultive and sex offenses presented a
relatively low risk of reoffending, they nonetheless had a lower chance of being released on their
ERD than did people convicted of property and public order offenses.
A. Crime type and re-offense rates: homicide and sex offenders do best
Overall, the success rate for the 14-year period was 62.9%. That is, more than six out of ten people
released from Michigan prisons from 1986-1999 did not return within four years for any reason. A
total of 19.5% were brought back as technical parole violators; 17.6% returned with new sentences
for new crimes.
1. Overall differences
Figure 3 shows that different offense groups succeeded at very different rates.

At more than 80%, success was greatest for those convicted of homicide; at nearly 78%, sex
offenders were a close second. In both these groups, fewer than 8.0% returned to prison with new
convictions. Motor vehicle, assault and drug offenses form the next group, with success rates
between 69.1 and 70.8%. For these offenders, returns for new crimes and for technical violations
each ranged between 13.3 and 16.5%.
Weapons and robbery offenders occupied a middle ground. Their success rates were 62.3% and
57.1%, respectively, but their returns with new convictions were almost identical at roughly 18%.
Robbery offenders were brought back more often for technical violations.
At the low end of the spectrum, larceny and burglary offenders had success rates below 55%.
Nearly one-quarter of those released in these groups were returned with new convictions. Another
one-fifth to one-quarter returned with technical violations.

Citizens Alliance on Prisons and Public Spending

19

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
The lower success rates for robbery, burglary and larceny reflect the fact that these economically
motivated crimes are more likely to be part of a calculated pattern of activity that may be disrupted
by prison, then resumed, with incarceration seen as a cost of doing business. The people who
commit them may be less likely than other offenders to have employment histories and job skills
that will help them re-enter the community successfully upon release. Notably, people who commit
crimes of impulse -- a decision to fight someone, to have sex or to drive drunk -- seem least likely to
reoffend.
2. Returns for new crimes: offenses against people are rare
The public image is that people released from prison go back to the community and commit the
very same crimes again. Thus it is assumed that, at least without extensive supervision or
intervention programs, robbers will rob, rapists will rape and those who killed will kill again. The
data present a very different picture.
Roughly one in six people released for the first time from a Michigan prison during the 14-year
study period returned to prison in Michigan within four years with a conviction for a new offense.
That is, of 76,040 people for whom the data is available, 13,416 returned with a new sentence for
an offense of any type and 3,394, or 4.5%, came back with an offense against a person, including
homicide, assault, sex or robbery.
Table 2 shows that the overwhelming majority of the new convictions were for property or public
order offenses. Just as they constituted 56% of the crimes for which people were released, the
three most common offenses -- larceny, drugs and burglary -- accounted for nearly 63% of all the
new convictions for which people were returned. For every offense group except motor vehicles,
the first or second most common new offense was larceny.
Table 2. New Offense for which People Returned to Prison
NEW OFFENSE TYPE
Larceny
Drugs
Burglary
Robbery
Assault
Motor Vehicle
Sex
Weapons
Homicide
Other
TOTAL

NUMBER
4,039
2,492
1,867
1,268
1,110
697
673
652
343
275
13,416

PERCENT
30.1
18.6
13.9
9.5
8.3
5.2
5.0
4.9
2.6
2.0
100.0

Of those who returned with new sentences, two-fifths were convicted of the same type of crime for
which they had originally been released. Return for the same offense type was far less common for
assaultive crimes. Table 3 examines these trends by offense group.

20

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Table 3. Returns to Prison for New Offenses by Original Offense Group30
New Offense Type
Original Offense
Larceny
(N=16,192)
Burglary
(N=10,602)
Robbery
(N=9,009)
Weapons
(N=2,524)
Motor Vehicle
(N=3,197)
Assault
(N=6,530)
Drugs
(N=16,146)
Sex
(N=6,673)
Homicide
(N=2,558)
TOTAL
(N=73,431)

None
12,173
75.2.%
8,033
75.8%
7,337
81.4%
2,059
81.6%
2,688
84.1%
5,645
86.4%
14,004
86.7%
6,170
92.5%
2,380
93.0%
60,489
82.4%

Crime of
Any Type
4,019
24.8%
2,569
24.2%
1,672
18.6%
465
18.4%
509
15.9%
885
13.6%
2.142
13.3%
503
7.5%
178
7.0%
12,942
17.6%

Crime of Same
Type as Original
2,016
12.5%
879
8.3%
440
4.9%
44
1.7%
402
12.6%
189
2.9%
1,111
6.9%
204
3.1%
14
0.5%
5,299
7.2%

Crime Against
Person*
727
4.5%
513
4.8%
727
8.1%
144
5.7%
54
1.7%
367
5.6%
394
2.4%
280
4.2%
69
2.7%
3,275
4.5%

*Homicide, sex, assault, robbery

At 7.0%, homicide offenders had the lowest rate of return for a new crime of any group. At 0.5%,
they also had the lowest rate of returning for the same type of crime. That is, people who served
time for murder or manslaughter were almost never convicted of killing again. Only 2.7% of the
homicide offenders -- 69 of nearly 2,600 people -- returned for any offense against a person.
At 7.5%, sex offenders had the second lowest return rate for new crimes. While they were more
likely to be returned for a new sex offense than for any other crime, only 204 of the nearly 6,700
released, or 3.1%, were imprisoned for new sex offenses within four years. Only 4.2% were
returned with a conviction for any new crime against a person. 31
At the other extreme, nearly one-quarter of the larceny and burglary offenders returned for new
offenses -- usually another theft. Of those who committed new crimes, half the larceny offenders
returned for a new larceny and one-third of the burglary offenders returned for burglary. In
addition, 3.1% of the larceny offenders returned for burglary and 7.3% of the burglary offenders
returned for larceny. Fewer than 5.0% of these property offenders returned to prison for a new
crime against a person.
This table includes only the nine original offense groups that are analyzed throughout this report, as nothing
meaningful could be concluded about the catch-all tenth category of “other offenses.” However, all the new offenses for
which people returned to prison were counted, no matter what the type, so that data regarding people who originally
committed a crime in one of the nine primary offense groups but returned with some other type of new conviction was
not discounted.
30

31 These findings are consistent with the results of the MDOC’s current effort to better identify the reoffense risk of sex
offenders being considered for parole. In 2009, the department began administering the Vermont Assessment of Sex
Offender Risk or VASOR. Although the analysis is still preliminary, by June 7, 2009, of 3,079 people tested, 75.0% scored
low risk for committing a new sex offense, 19.3% scored moderate risk and 5.7% scored high risk. Data provided to
CAPPS by MDOC Office of Research and Planning by e-mail dated July 27, 2009.

Citizens Alliance on Prisons and Public Spending

21

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
The lowest return rates for crimes against people belonged to the drug and motor vehicle groups.
More than half the drug offenders who returned for a new crime came back for another drug
offense. Only 2.4% had a new crime against a person. Motor vehicle offenders were even more
consistent. Eighty percent of those who returned with a new crime had a new motor vehicle crime;
only 1.7% -- 54 of nearly 3,200 releases -- had a new crime against a person.
The offense groups with the highest rates of return for crimes against people were robbery,
weapons and assault. For robbery and assault offenders, this was primarily because of returns for
new crimes like their original ones. However, weapons offenders presented a different situation.
While only 1.7% returned for weapons offenses, 2.2% returned for assault, 1.7% returned for
robbery and 1.1% returned for homicide -- the highest return rate for homicide of any offense
group. Thus, while carrying a weapon is in itself not an assaultive offense, it appears that people
who carry them risk graduating to assaultive crimes.
In sum, the vast majority of people first released from prison from 1986-1999 -- 82% -- did not
return to a Michigan prison within four years for any new crime. Of those that did, most returned
for property or drug offenses. Fewer than 5% had new convictions for crimes against people.
3.

Return rates for max outs vs. parolees

Similar differences among offenses existed whether people were denied parole or released only
after serving their maximum terms. Overall, 5.5% of the cases were “max outs.” In every offense
group, max-outs actually had higher success rates than parolees -- an average of 74.4% Because
they were not supervised on parole and thus could never be returned as technical parole violators,
one avenue of failure did not exist for them. If max outs did not return to prison with new
sentences, they had succeeded.
However, when the PVNS rates are compared for parolees and max outs, we see in Table 4 that, on
average, max outs returned with new sentences at a rate nearly 50% higher than that of parolees.
The disparity was greatest for homicide and lowest for burglary. The pattern of returns by offense
group is similar, albeit not identical, to that of people who were paroled, ranging from 13.6% for sex
offenders to 35.6% for the larceny group.
Table 4. Returns to Prison with New Sentences: Parolees & Max Outs
Total
Cases
Burglary
Larceny
Robbery
Weapons
Motor Vehicles
Drugs
Assault
Sex
Homicide
TOTAL

22

10,444
14,933
8,859
2,344
3,127
15,863
5,727
5,737
2,430
69,464

Parolees
No.
Percent
PVNS
PVNS
2,532
3,564
1,624
409
493
2,065
702
383
158
11,930

24.2
23.9
18.3
17.4
15.8
13.0
12.3
6.7
6.5
17.2

Total
Cases
170
1,277
158
181
72
291
805
947
129
4,030

Citizens Alliance on Prisons and Public Spending

Max Outs
No.
Percent
PVNS PVNS
46
454
51
55
17
75
182
129
21
1,030

27.1
35.6
32.3
30.4
23.6
25.8
22.6
13.6
16.3
25.6

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Presumably the parole board made people serve their maximum sentences because of their
perceived risk. Thus it seems particularly noteworthy that three-quarters of the max outs did not
return at all.
4.

Placing the re-offense data in a national context

Although the published research on recidivism varies substantially in how reoffending is measured,
the results on offense type are highly consistent. Reoffending is highest for property offenses and
lowest for homicide and sex offenses.
The Bureau of Justice Statistics (BJS) conducted the largest study.32 It looked at 272,111 people
released from prisons in 15 states (including Michigan) in 1994. The rate of return to prison with a
new sentence for any offense was 10.8% for homicide offenders, 12.6% for those convicted of rape
and 10.5% for those with other sex offenses. The rate was nearly triple for larceny offenders
(32.6%) and burglars (30.8%). Drunk drivers, at 16.6%, also had a low rate of returns with new
offenses. In commenting on the rearrest rates of various offense groups, the authors of the BJS
report noted that those with the highest rates had been convicted of crimes motivated by money.
BJS also examined the extent to which people were arrested for committing crimes of the same type
for which they been in prison. (Rearrest rates, it should be noted, range from two to four times
higher than rates of return to prison with new sentences.) Only 2.5% of released rapists were rearrested for rape; only 1.2% of homicide offenders were rearrested for another homicide. By
comparison, 13.4% of released robbers were rearrested for robbery; 23.4% of released burglars
were rearrested for burglary; 33.9% of larceny offenders were rearrested for larceny.
A Washington State study defined recidivism as a conviction occurring during the first five years
after release to the community.33 The database included nearly 70,000 people placed in the
community from 1994-1998. Researchers found that 13.0% of sex offenders, 31.5% of other
violent offenders and 33.7% of nonviolent offenders were reconvicted of any new felony. The
extent to which people repeated their original offenses also varied greatly. Only 2.7% of the
Washington sex offenders were convicted of new sex offenses, while 16.6% of other violent
offenders committed new violent offenses and 25.2% of non-violent offenders (which included
property and drug crimes) were convicted of a new non-violent offense.
The New York Department of Correctional Services analyzed data on 441,297 people who were
released from prison in New York from 1985-2003. 34 It defined recidivism as a return to prison
within three years. While, overall, 16.8% of the total returned with a new commitment and 24.8%
were returned for a technical parole violation, recidivism rates varied substantially by offense type.
Homicide (7.5%), sex (7.8%) and drunk driving offenders (9.3%) clustered together with low rates
of reincarceration for new commitments; robbery (19.6%), larceny (20.5%) and burglary (21.7%)
clustered together with rates that were two to three times higher.
Patrick Langan and David Levin, Recidivism of Prisoners Released in 1994, U.S. Department of Justice, Bureau of Justice
Statistics (Washington, D.C., 2002).

32

Robert Barnoski, Sex Offender Sentencing in Washington State: Recidivism Rates, Washington State Institute for Public
Policy (2005).

33

Leslie Kellam and Michael Hayes, 2003 Releases: Three Year Post Release Follow-up, New York State Department of
Correctional Services, Division of Program Planning, Research & Evaluation, at pp 14-15.

34

Citizens Alliance on Prisons and Public Spending

23

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?

Table 5. Returns for New Conviction of Same Offense Type
New York: Released 1985-2003*
Release
Offense
Homicide
Sex
Robbery
Burglary

Total
Released

Number.
Returned

Percent
Returned

13,458
13,890
85,824
53,485

57
291
5,548
6,050

0.4
2.1
6.5
11.3

*See Leslie Kellam and Michael Hayes, 2003 Releases: Three Year Post Release Follow-up, New York State
Department of Correctional Services, Division of Program Planning, Research & Evaluation, at p 16.

The New York researchers also looked at selected crime groups to determine the extent to which
people returned to prison for committing new crimes of the same type as their original convictions.
As Table 5 shows, they found sex offenders committed new sex offenses at one-third the rate at
which robbers committed new robberies and one-fifth the rate at which burglars committed new
burglaries. Homicide offenders almost never killed again.
Three-year reincarceration rates for 1,786 men released from Massachusetts prisons in 2002 also
varied greatly by offense type. Property offenders had the highest rates by far and were nearly
three times as likely to return as sex offenders, who had the lowest rates. Noting that their results
were consistent with national studies, the authors concluded: “The misconception that serious
violent offenders, including sex offenders, will reoffend may be due in part to misrepresentations in
the media and a general fear of these particular offenders. 35
An analysis of Illinois arrest data produced similar results. Of people arrested in 1990, those
charged with robbery had the highest rates of rearrest for any offense within five years, followed by
burglary, assault, larceny, sex and homicide. 36 When rearrest for the same type of offense was
examined, sex and homicide offenders were again at the bottom of the list. Noting that DNA
collection, registration requirements and notification procedures make sex offenders a highly
visible group for law enforcement to question, the authors found it surprising that 93% of sex
offenders were not rearrested for another sex crime. 37

35 Rhiana Kohl, Hollie M. Hoover, Susan M. McDonald and Amy L. Solomon, Massachusetts Recidivism Study: A Closer Look
at Releases and Returns to Prison, Urban Institute, Justice Policy Center (Washington, D.C., 2008) at p 31. This study did
not separate technical violators from people with new convictions by offense type or examine the types of new offenses
for which people returned.

Lisa Sample and Timothy Bray, Are Sex Offenders Dangerous? 3 Criminology and Public Policy, No. 1, 59-82 (2003).
Time served in prison was taken into account in estimating arrestees’ opportunities to reoffend during the analysis
period.

36

37 A study of people released from Pennsylvania prisons from 1997 through 2003 showed similar results. Recidivism was
defined as a return to prison for any reason. Property offenders and robbers returned at twice the rate of sex offenders.
Robert Flaherty, Recidivism in Pennsylvania State Correctional Institutions, 1997-2003, Pennsylvania Department of
Corrections (2005).

24

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Because of public concern about sex offenses, a great deal of literature is focused specifically on this
category. In a report on sex offender policies in the U.S. published recently by the Vera Institute of
Justice, the author noted:
Many sex offender policies are predicated on the assumption that re-offense rates for sexual
offenses are higher than those for other felonies… However, there is a significant body of
research that appears to contradict this proposition.38
In fact, numerous researchers have commented on the low rate of recidivism among sex offenders.
A frequently cited meta-analysis of 61 recidivism studies with over 23,000 subjects from six
countries showed that, on average, 13.4% of sex offenders were charged with or convicted of a new
sex offense during an average follow-up time of 4-5 years. Rapists showed higher rates of sexual,
violent and general recidivism than did child molesters. The authors concluded: “The present
findings contradict the popular view that sexual offenders inevitably re-offend.”39
A subsequent study analyzed 10 individual sub-samples from Canada, the United Kingdom and the
United States for a combined sample size of 4,724. It also found that most sex offenders do not reoffend sexually, but that rates vary depending on the nature of the offense and the offender’s age
and prior record. The researchers suggested: “Rather than considering all sexual offenders as
continuous, lifelong threats, society will be better served when legislation and policies consider the
cost/benefit break point after which resources spent tracking and supervising low-risk sexual
offenders are better re-directed toward the management of high-risk sexual offenders, crime
prevention and victim services.”40
Several studies of formerly incarcerated American sex offenders confirm their low recidivism rates.
The Bureau of Justice Statistics conducted a separate analysis of the 9,691 male sex offenders who
were within its three-year follow-up study of prisoners released in 1994 from prisons in 15
states.41 It found that 24.0% were reconvicted and 11.2% were returned to prison with a new
sentence for any type of crime. However, only 3.5% or 339 released sex offenders were convicted
of another sex offense. Of 4,295 categorized as child molesters, 150 were reconvicted of a new
sexual offense against a child or adult.
A 2003 study by the Florida Department of Corrections analyzed the factors that affect recidivism. For offense type, it
considered not the offense at release but the offender’s total history. Holding other factors constant, it found that a male
whose most serious offense ever was homicide was 21.6% less likely to return to prison than other offenders. Male sex
offenders were 19.8% less likely to return. Conversely, males whose worst offense was robbery were 13.7% more likely
to return and those whose worst crime was burglary were 28.7% more likely to be reimprisoned. Glen Holley and David
Ensley, Recidivism Report: Inmates Released from Florida Prisons, July 1995 to June 2001, Florida Department of
Corrections (2003).
Tracy Velazquez, The Pursuit of Safety: Sex Offender Policy in the United State, Vera Institute of Justice (New York, 2008),
at p 6.

38

R. Karl Hanson and Monique T. Bussiere, Predicting Relapse: A Meta-Analysis of Sexual Offender Recidivism Studies, 66
Journal of Consulting and Clinical Psychology, No. 2, 348-362 (1998), at p 357. A meta-analysis aggregates the findings of
numerous studies and applies advanced statistical techniques to estimate their combined effects.

39

40 Andrew J.R. Harris and R. Karl Hanson, Sex Offender Recidivism: A Simple Question, Report 2004-03, Department of the
Solicitor General Canada (Ottawa, 2004), at p 12.
41 Patrick Langan, Erica Schmitt and Matthew Durose, Recidivism of Sex Offenders Released from Prison in 1994, U.S.
Department of Justice, Bureau of Justice Statistics, (Washington, D.C., 2003).

Citizens Alliance on Prisons and Public Spending

25

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
The California Sex Offender Management Board released two studies in June 2008. One followed,
for a period of ten years, 3,577 sex offenders who were paroled in 1997.42 Only 3.4% had been
returned to prison for a new sex offense; an additional 3.8% had been returned for a new
conviction for some other kind of offense. The second study followed, for five years, 4,204 sex
offenders who were paroled in 2002.43 The results were nearly identical. Only 3.2% had been
returned to prison for a new sex offense; an additional 4.7% had been returned for a new
conviction for some other kind of offense.44
The Minnesota Department of Corrections examined recidivism among 3,166 sex offenders first
released from prison between 1990 and 2002. 45 The follow-up period ranged from 3 to 16 years,
with an average of 8.4. Any sex offense, including misdemeanors as well as felonies, was used as
the measure of reoffending. After three years, 5.7% had been reconvicted of a sex offense and 3.2%
had been re-incarcerated. Nearly 9% had been returned to prison for any offense. By the end of the
follow-up period, 10% had been reconvicted of a sex offense and 7% had been reincarcerated.
The Indiana Department of Correction examined the recidivism rates of people released from
prison in 2002-2005. Of 3,615 sex offenders, 405 (11.2%) returned within three years for any new
offense; only 70 (1.9%) returned for a new sex offense. 46
The Ohio Department of Rehabilitation and Correction reviewed 879 sex offenders who were
released in 1989.47 During a 10-year follow-up period, 8% returned to prison for a new sex

offense, 14.3% returned for a non-sex offense and 11.7% returned as technical violators,
including 3.1% whose violations involved sexual behavior. More than half the returns occurred
within two years of release. Nearly three-quarters occurred within four years. Offenders whose
victims were adults returned for new sex offenses more than twice as often as those whose victims
were children.48
The Ohio researchers noted that although returns to prison may under-represent the actual amount
of new sex offenses to some extent, someone who served a prison term for a sex offense would be
unlikely to receive a non-prison sentence if convicted of a new sex offense. After reviewing a
number of other studies, the Ohio report observes that the low rate of sexual reoffending in that
state is similar to other research findings. It concludes: “Contrary to the popular idea that sex
42

California Sex Offender Management Board, Recidivism of Paroled Sex Offenders – A Ten (10) Year Study (2008).

43

California Sex Offender Management Board, Recidivism of Paroled Sex Offenders – A Five (5) Year Study (2008).

Notably, in both studies, sex offenders had substantially lower rates of return for technical parole violations than other
offenders did. Although California returns parolees for technical violations at the rate of 60-70%, far above the national
average, fewer than 50% of sex offenders were returned. The studies point out that sex offenders are intensely
supervised. Thus it does not appear that they were returned less often for new convictions because they were being
returned for technical violations instead.

44

45

Minnesota Department of Corrections, Sex Offender Recidivism in Minnesota (2007).

46 Data for releases from 2002-2004 was reported in Indiana Department of Correction, Recidivism Rates Compared: 20052007 (2009). Data for 2005 releases was provided by Aaron Garner, the research analyst who prepared that report.
47 Ohio Department of Rehabilitation and Correction, Bureau of Planning and Evaluation, Ten-Year Recidivism Follow-Up of
1989 Sex Offender Releases (2001).
48

Lower offense rates for sex offenders with child victims were also found in the Washington State study, note 33 supra.

26

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
offenders are repeatedly returning to prison for further sex crimes, in this population a sex offender
recidivating for a new sex offense within 10 years of release was a relatively rare occurrence.”49
Table 6 summarizes the results of the sex offense research.
Table 6. Sex Offender Recidivism Rates
Total
Cases

New
Sex
Crime

Any
New
Offense

Follow-up
Period

Recidivism
Measure

Michigan

6,673

3.1%

7.5%

4 years

Return to prison

Bureau of Justice Statistics

9,691

3.5%

24.0%

3 years

Reconviction*

Washington

4,091

2.7%

13.0%

5 years

Reconviction

California

3,577

3.4%

7.2%

10 years

Return to prison

California

4,204

3.2%

7.9%

5 years

Return to prison

8.0%**

14.3%

10 years

Return to prison

Study

Ohio

879

Minnesota

3,166

5.7%
3.2%

25.4%
8.6%

3 years
3 years

Reconviction*
Return to prison

New York

13,890

2.1%

7.8%

3 years

Return to prison

3,615

1.9%

11.2%

3 years

Return to prison

Indiana

*includes misdemeanors
**also found that 1.4% had parole violations for behavior constituting a sex offense

B. Crime type and parole denial: success rates and release decisions often don’t match
The frequency with which people were released on their earliest release date (ERD) or continued
for some number of years varied widely by offense group. The extent of those variations was
markedly different for the old and new parole boards.
1. Release differences by offense group
As Table 7 shows, for the entire 14-year period, 61.4% of all released prisoners were paroled when
they first became eligible. More than 15% were released within a year of their ERD and another
14.6% were released within two years. Thus, on average, 91.2% of the cases in the database were

49

Ohio Department of Rehabilitation and Correction, note 47 supra, at p 15.

Citizens Alliance on Prisons and Public Spending

27

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
released within two years of their first possible parole date.50 At the other extreme, 1.9% were
released more than four years after their earliest date.
Table 7. Release Time by Offense Category
(in percentages)

Motor Vehicle
Drugs
Larceny
Weapons
Homicide
Burglary
Assault
Robbery
Sex
Total

Rel’d on
ERD
78.8
75.5
63.5
62.7
61.1
58.6
57.2
53.6
32.6
61.4

1 yr. past
12.4
13.8
17.4
16.4
11.0
17.9
14.0
14.8
13.8
15.2

2 yrs past
7.4
8.0
13.4
13.5
14.1
15.6
17.8
18.2
27.5
14.6

2-4 yrs past
1.4
2.3
4.8
6.5
9.6
6.4
8.9
10.2
18.8
6.9

>4 yrs
past
0.0
0.3
0.9
0.9
4.1
1.5
2.2
3.1
7.2
1.9

Max
Out
2.2
1.8
7.6
7.1
4.7
1.6
12.1
1.7
13.9
5.5

Table 7 also indicates that, overall, 5.5% of the people released from 1986-1999 had served their
maximum sentence or had “maxed out.” The proportion of people who max out reflects the parole
board’s determination to keep those people incarcerated as long as the law allowed. 51
Motor vehicle (78.8%) and drug offenders (75.5%) were well above average in the proportion of
people paroled as soon as they became eligible. And if not released on their ERD, they were
generally released soon after. Roughly 90% were paroled on or within one year of their first
eligibility date. Almost none were continued for more than four years and only about 2% were
required to serve their maximums.
Five offense groups -- larceny, weapons, homicide, burglary and assault -- largely mirrored the
overall average, especially in the proportion of people released on their ERD. For those not paroled
when first eligible, homicide and assault offenders were less likely to be released within one year
and more likely to be continued for more than two years. Weapons and larceny offenders were
maxed out more often, presumably because their maximum sentences were relatively short.52 A
larger share of assault offenders, 12.1%, served their maximum sentences, but this also reflects, in
part, the fact that in over 40% of the assault cases the maximum was four years or less.
Somewhat further below the average were robbery offenders. Fewer than 54% were paroled on
their ERD and only 68.4% were released within a year of first eligibility. However, at the two-year
mark, the robbery group was virtually identical to the homicides, with 86% released.
It must be remembered that these figures do not reflect the overall parole grant rate. Since the database does not
include people who could have been released but were not, it does not reveal the proportion of all eligible prisoners who
were paroled in any given year.

50

51

The number of years served past the earliest release date and “maxing out” are distinct concepts. Someone with a short
minimum but a long maximum sentence may serve many years after first becoming eligible for parole, even if not
required to max out. Conversely, someone with a two-year minimum and a four-year maximum who maxes out will have
served only two years past the first parole date.
52

Virtually all of the weapons cases and at least 60% of the larceny cases carried maximums of five years or less.

28

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
The most striking differences occurred with sex offenders. Only one-third were released on their
ERD; fewer than half were paroled within one year. Nearly 28% were continued for two years,
almost twice the average. The proportion continued from two to four years (18.8%) was nearly
three times the average for all offenders and the proportion continued for more than four years
(7.2%) was nearly four times the average. Almost 14% of sex offenders were made to serve their
maximum sentences, which for the vast majority of this group was at least 15 years.
2. Differences between the old and new boards
The figures for the entire database mask significant differences between the old board’s release
decisions and those of the new board. As Table 8 shows, when all offenses are averaged together,
the proportion that the old board released on their ERD was not quite three points greater than that
of the new board. Overall, the proportion released up to two years after their ERD declined from
over 94% for the old board to about 88% for the new board. However, the extent of the difference
varied greatly by offense type.
Table 8. Release Time Differences – Old and New Boards
(in percentages)
New Board

Old Board

Motor Vehicle
Drugs
Homicide
Assault
Weapons
Larceny
Burglary
Robbery
Sex
Total

Rel’d
on ERD

1 yr past

2 yrs past

Max
out

Rel’d
on ERD

1 yr past

2 yrs past

Max
out

76.4
75.8
69.8
64.2
63.9
62.9
57.0
56.1
47.4
62.8

15.1
14.8
11.8
14.8
18.2
19.1
20.5
17.1
20.5
17.7

6.9
7.8
10.9
15.4
13.2
13.7
16.1
17.6
20.9
14.0

2.9
1.6
5.4
7.3
5.1
7.2
1.2
1.2
7.1
4.1

79.5
75.3
53.8
52.6
61.5
64.0
60.6
51.3
20.5
60.1

11.6
13.1
10.3
13.5
14.7
15.5
14.7
12.5
8.3
13.1

7.5
8.1
16.8
19.4
13.8
13.2
15.0
18.9
33.1
15.1

2.1
1.9
4.8
15.6
9.2
8.8
2.1
2.3
20.0
6.7

For motor vehicles, drugs, larceny and burglary, the differences between the two boards were
small. The proportion released on or within one year of the earliest release date declined by 2.5%
or less. The new board’s stricter tendencies were more pronounced for weapons offenders, where
the proportion released on or within one year of the ERD declined by nearly six points and the
proportion of people maxed out grew from 5.1% to 9.2%. Nonetheless, for these five groups,
release came within two years of the ERD for at least 90% of the people released by both boards.
For robbery, assault, homicide and, above all, sex offenders, the change in release policies by the
new board was dramatic. Both the time past ERD and the proportion of people required to max out
increased substantially. The proportion of homicide offenders released when first eligible dropped
16 points; the proportion of assault offenses dropped nearly 12. The proportion of assault
offenders who maxed out more than doubled.

Citizens Alliance on Prisons and Public Spending

29

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Sex offenders experienced the most extreme change, with release on the ERD dropping from more
than 47% under the old board to less than 21% under the new board. The proportion released
within a year dropped from 67.9% to 28.8%; the proportion released within two years dropped
from nearly 89% to less than 62%. The proportion of sex offenders held to their maximum nearly
tripled to a stunning 20%.
If the new board had made release decisions for these four groups in the same fashion as the old
board, over a seven-year period it would have released 1,811 more people on their ERD and maxed
out 899 fewer. Nearly 2,600 people would have been released on or within one year of their ERD
instead of being continued for two, three, four years or even more. 53
3. The disconnect between success rates and release on the ERD
How did release decisions relate to success rates? Were the people who were most likely to
succeed also most likely to be paroled when they became eligible? Did delaying release cause
success rates to rise? Figures 4a and 4b compare each group’s overall success rates with the extent
to which the old and new parole boards granted parole on the earliest release date. Comparing the
overall rates, as opposed to those only of the people released on their ERD, accounts for any impact
of delaying release on success. The graphs make several points.
Under both the old and new boards, there was no clear correlation between the rate of release on
the ERD and the rate of success. That is, people were not paroled at their earliest eligibility date in
proportion to their success upon release. The disparity between success and release rates varied
among offense groups and between the two boards. Sometimes success rates were higher and
sometimes it was release rates.
For the homicide, assault and sex groups, success rates exceeded release rates under both boards.
Under the old board, the disparities were 8.8, 5.7 and 31.7 points, respectively. Under the new
board, the gap between success and release rates widened profoundly to 27.5, 17.4 and 56.0 points.
Thus the graphs demonstrate the extent to which, in these groups in particular, the denial of parole
was unrelated to the actual likelihood of success.
For four other offense groups (motor vehicles, drugs, larceny and burglary), under both boards the
disparities were smaller and their direction was reversed. That is, the release rates exceeded the
success rates by roughly five to ten points.
For robbery offenders, the gap between success and release rates did not only grow under the new
board, it changed directions. Under the old board, the 56.1% release rate was slightly higher than
the 54.1% success rate. However, under the new board, while the success rate grew to 59.7%, the
release rate decreased to 51.3%, causing releases to fall behind by 8.4 points.
Weapons offenders came closest to having an identity of success and release rates under both
boards, although they also saw a small reversal of direction. Under the old board, 60.6% succeeded,
but 63.9% were released on their earliest date. Under the new board, 63.8% succeeded, but only
61.5% were released.
53

It cannot be determined from the data exactly how many additional years individuals served. Someone who would
have been released on his ERD by the old board might have been continued for two, three or four years by the new board
while someone who the old board would have continued for two years might still have been released two years after their
ERD by the new board. For further details, see Appendix D, Table A.

30

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?

Percent

Figure 4a. Success and Release Rates Compared by Offense
Old Board
100
90
80
70
60
50
40
30
20
10
0

.6
78

.1
79

.8
69

4
6.
.5 7
71

.9
69

8
5.
.8 7
69

.2
64

.9
.0 6 3
60

.4
47

Homicide

Sex

Motor
Vehicle

Assault

Drugs

Weapons

1
.4 56.
54

Robbery

.9
62
.5
52

.8
48

Larceny

.0
57

Burglary

54
.0

60
.6

64
.0
51
.3

20
.5

56
.8

63
.8

52
.6

59
.7

Assault

61
.5

Motor
Vehicle

68
.5

70
.0

75
.3

79
.5
70
.6

76
.5

81
.3

100
90
80
70
60
50
40
30
20
10
0

53
.8

Percent

Figure 4b. Success and Release Rates Compared by Offense
New Board

Homicide

Sex

Drugs

Success

Weapons

Robbery

Larceny

Burglary

Released on ERD

While one would not expect release decisions to mirror success rates precisely, if the primary
purpose of parole decision-making is to release people who have served their minimum sentences
unless there is reason to believe they pose some risk to the public, it is reasonable to expect some
relationship between known outcomes and release on the ERD. That is, the offense groups with the
highest success rates would have the highest release rates; the offense groups with the lowest
success rates would have the lowest release rates. The data make it apparent that parole release
decisions, especially those made by the new board, were based on grounds other than
demonstrated risk.

Citizens Alliance on Prisons and Public Spending

31

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
The data also allow for the testing of another hypothesis. If release decisions accurately reflect the
likelihood of success, one would expect more selective release decisions to produce higher success
rates. That is, as release rates decline, success rates should increase. However, Figures 4a and b
show that changes in release rates by the new board had virtually no relationship to changes in
success rates.
For the homicide group, although the release rate declined by 16 points, the success rate increased
by fewer than three. For the assault group, the release rate dropped by nearly 12 points, but
success increased by just one-tenth of a point. And for sex offenders, although the release rate
declined by nearly 27 points, the success rate also declined, albeit by fewer than three points.
Among the offense groups that saw release rates increase under the new board, success rates
dropped by about one point for the motor vehicle group but increased by 4.3 points for larceny and
5.2 points for burglary offenders. Only the weapons and robbery groups saw a roughly
proportional increase in success as release rates declined.54 Thus it cannot be concluded that
releasing fewer people led to increased success because the “right” people were being paroled or
that releasing more people led to higher failure rates.
C. The relationship of release trends to success over time
If releasing fewer people when they first become eligible kept the public safer, we would expect the
success rate to climb as the rate of release on the ERD falls. Conversely, success rates would decline
when releases at first eligibility grow. As Figure 5 shows, success and release rates both varied
over time, but not in ways that prove this connection.
Figure 5. Success, Failure and Release Rates Over Time
80

70

60

Percent

50
Success
PVNS

40

PVT
Released on ERD

30

20

10

0
86

87

88

89

90

91

92

93

94

95

96

97

54

98

99

Release rates dropped by 2.4 points for weapons offenders and 4.8 points for robbers while their success rates
increased by 3.2 and 5.2 points, respectively.

32

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Notably, the rates of success and release on the ERD are virtually identical in 1986, 56.2% and
57.2%, respectively. They are virtually identical again in 1999, when 63.6% of those released
succeeded and 63.1% of releases occurred on the ERD. If both rates had climbed straight up
together, parallel all the way, it would appear that increasing releases at first eligibility actually
improved success rates. That could be readily explained if the increased releases were of people in
the offense groups with higher success rates. That is, if more of the people most likely to succeed
are released on their ERD, quicker releases will actually drive success rates higher. In fact, the two
rates did climb essentially in tandem through 1992, the last year of the old board. However, after
that, they diverged markedly.
The release rate dropped eight points in 1993, climbed a bit, then dropped again by nearly six
points. It hit a low of 55.5% in 1995, then rose steadily for the next three years. The success rate,
on the other hand, changed much more modestly. It peaked at 67.0% in 1994, then declined by four
points over the next four years. That is, not only did the success rate fail to grow in proportion to
the decline in releases on the ERD, the success rate itself declined when the release was lowest.
Clearly, the substantial decrease in releases at first eligibility during the early years of the new
board did not result in greater success overall.
Release

Success

1986
1996

57.2%
57.1%

56.2%
65.4%

1987
1997

60.3%
60.4%

56.5%
64.2%

1989
1994

61.2%
61.3%

60.7%
67.0%

As the box at left shows, the lack of
relationship between success and
release timing can also be seen by
comparing years when the rates of
release on the ERD were identical. For
each pair of years, the success rates
were markedly different. This further
suggests that success is not a function
of when people are released.

Figure 5 also allows us to examine the
two types of failure over time. Returns with new sentences and for technical violations followed
different patterns. New sentence returns showed an occasionally halting but generally steady
decline throughout the period, from 24.2% in 1986 to 12.2% in 1999, with a brief uptick in 1995.
The biggest decline in new offenses, a drop of 9.3 points, occurred from 1990-94.
Technical violations, on the other hand, were more erratic. From 1986-88, they increased
modestly, from 19.6% to 21.4%, then declined markedly. The PVT rate bottomed out in 1991 at
15.3%, rose one point in 1992, then rose steadily during the new board’s tenure, peaking in 1999 at
24.2%.
The relationship between the two types of failure is complicated. Unlike returns with new
sentences, revocations for technical violations are a matter of parole board discretion. The stated
purpose of returning technical parole violators to prison is to protect the public. Some unknown
proportion of technical violators engaged in criminal behavior that was not prosecuted or that
resulted in sanctions other than a prison sentence. The evidence may not have been strong enough
to support a conviction in court or the new offense may only have warranted jail time or even
probation. Since it was already determined that these parolees were not going to be returned to
prison with new sentences, bringing more of them back as technical violators would increase the
PVT rate without reducing the PVNS rate.

Citizens Alliance on Prisons and Public Spending

33

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
The rest of the technical violators broke conditions of their supervision that were believed to
increase their risk of reoffending. To the extent that new crimes that would have resulted in new
prison sentences were prevented, one would expect an increase in returns for technical violations
to produce a decrease in returns with new sentences. That is, if increasing revocations for technical
violations actually increases public safety, this should be reflected at least to some extent by a
decline in the proportion of parolees who receive new prison terms for new felony convictions.
We know that the new board increased the technical violation rate steadily. In 1986, the PVNS rate
was 4.6 points higher than the PVT rate. By 1993, for the first time, the PVT rate was the higher
one. Six years later, the disparity had grown to 12 points, with returns for technical violations
occurring at twice the rate of returns for new offenses.55 The question is whether this means the
new board’s policy worked. Did re-offending by parolees decrease because the board acted
preemptively, returning people to prison for technical violations before they had the chance to
reoffend? Or did reoffense rates stay stable but returns for new offenses declined because more
people who committed new crimes were returned as technical violators instead of being
prosecuted? Or is there a third possibility?
The answer is unclear. If more parole revocations for technical violations actually prevented new
crimes, one would expect a steady correlation between decreasing PVNS rates and increasing PVT
rates. However, as we saw in Figure 5, from 1988-91, both rates declined. In addition, there is
another strong correlation to consider. Figure 6 shows that Michigan’s property crime rates
declined steadily throughout the 14-year period, except for a small uptick in 1992. The 34% drop
in property crime paralleled the 50% drop in parolee returns for new offenses, which we have seen
are disproportionately for property offenses.

55

This was despite the fact that the new board increased the number of people who maxed out and thus, by definition,
could not become technical parole violators.

34

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Violent crime rates also dropped 28% overall, although the trend was more erratic, with upticks in
1990, 1991, 1993 and 1998. Thus it appears that the decline in new sentences for former prisoners
may be a function of declining crime rates generally. In particular, the economic prosperity of the
1990’s, when jobs were available even for people with criminal records, may have had at least as
much to do with a reduction in returns for new offenses as the re-incarceration policies of the new
parole board.

IV. Other key factors: age, prior prison terms and misconduct
Whether someone returns to prison will depend on numerous factors in addition to offense type.
These may be static, that is, incapable of change, or dynamic. They may include the stability of
family and other social support systems, physical and mental health, substance abuse, educational
level, and access to stable housing and adequate employment. The relative importance of these
factors may vary by individual and change over time. Strengths in some areas may compensate for
weaknesses in others. A growing national focus on those dynamic factors that can be positively
impacted by in-prison programs and post-release community-based services is exemplified by the
Michigan Prisoner Reentry Initiative (MPRI).
Some important factors are quite basic. Criminologists have long known that characteristics related
to success upon release include older age at release, the lack of a prior criminal record and good
institutional conduct. 56 The MDOC’s research division has duplicated those findings with Michigan
prisoners and has weighted them heavily in scoring Michigan’s parole guidelines. Therefore we
must consider the possibility that it is not the nature of the offense itself that is associated with
success rates but the extent to which offense groups reflect other characteristics that relate to
success. For instance, it may not be that sex offenders have high success rates because of
something about their common behavior but because they tend to be older first offenders with
good institutional conduct. Burglars may have lower success rates not because of the nature of
theft offenses but because they are younger, and have more prior convictions and more
institutional misconducts. 57
The data allowed for analysis of three other factors: age at release, prior prison terms and
institutional misconduct citations. It revealed that the offense groups with the highest success rates
had a larger proportion of people who were older at release, were serving their first Michigan
prison terms and had few, if any, misconducts. It also revealed that, within offense groups, the
people with these characteristics tended to do better upon release. However, these factors do not

56 See, for example, Flaherty, note 37 supra, at pp 9 and 10 (recidivism rates are lowest for people with no misconducts
one year before release and older age); Garner, note 46 supra, at p 15 (the younger the age at release, the greater the
likelihood of return to prison); Harris and Hanson, note 40 supra, at p 11 (factors associated with higher recidivism rates
for sex offenders are a) male victims, b) prior sexual offenses and c) young age); Holley and Ensley, note 37 supra, at p 21
(the four most important factors affecting the reimprisonment of male offenders for new offenses are, in rank order:
prior recidivism, age at release, education level and disciplinary reports); Kohl, Hoover, McDonald and Solomon, note 35
supra, at p 31 (men returned to prison were, on average, younger when released, single and had longer criminal
histories); Kellam and Hayes, note 34 supra, at pp 18 and 24 (returns to prison within three years decline steadily as age
at release increases while returns increase with more prior felony convictions); Langan and Levin, note 32 supra, at pp 7
and 9 (younger age at release and prior prison sentences increase likelihood of rearrest).

It is also logically possible, of course, that the offense could have the primary causal effect and other characteristics are
dependent on it. That is, older first offenders with good institutional records might have higher success rates because so
many are from offense groups, like homicide and sex, with higher success.

57

Citizens Alliance on Prisons and Public Spending

35

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
fully explain why some offense groups are more successful than others and they clearly do not
outweigh offense type in the parole board’s decision-making process.
A. Older age at release, first prison terms and few misconducts help predict success
The data confirm that several characteristics associated with success upon release are also
associated with the offense groups that are most successful. These characteristics are: older age at
time of release, serving first Michigan prison term, having no “nonbondable” (serious) misconducts
in the preceding three years and having no more than one misconduct of any kind in that time.
Table 9. Frequency within Offense Groups of Characteristics Associated with Success

Homicide
Sex
Motor Vehicle
Assault
Drugs
Weapons
Robbery
Larceny
Burglary
TOTAL

Success
(percent)
80.1
77.6
70.8
70.0
69.1
62.3
57.1
54.5
51.1
62.9

Average age
at release
(years)
35.5
36.1
37.5
31.1
30.5
28.3
28.9
29.9
27.9
30.8

“A” prefix
(percent)
85.5
90.7
88.7
85.7
87.6
82.4
78.4
75.1
70.4
81.1

Zero nonbond
misconducts
(percent)
88.9
89.2
87.7
78.6
80.6
77.1
77.8
76.5
73.9
79.4

Zero or one
misconducts.
(percent)
65.8
74.0
75.4
56.7
54.7
52.2
46.6
50.1
43.4
54.2

As Table 9 indicates, the offense groups with the highest success rates -- homicide, sex and motor
vehicles -- had an average age at release that was at least 4.4 years greater than the next closest
group. While the association is not perfectly direct, the average age for the three high success
groups was from 35.5 -37.5 while for the four lowest groups it was less than 30.
The most successful groups were more likely to have “A” prefixes on their prison numbers,
indicating that this was their first Michigan prison term. While again, the association is not perfect,
the most successful groups were less likely to have been in prison before. Thus nearly 91% of the
sex offenders but fewer than 71% of the burglars had “A” prefixes.
Institutional conduct was also markedly better for the groups with higher success rates. Nearly
90% of those in the three highest groups had none of the more serious non-bondable misconducts
and at least two-thirds had no more than one misconduct of any kind. By comparison, 74% of
burglars had no non-bondable misconducts and fewer than 44% had 0-1 misconducts of any kind.58
A closer look reveals how these characteristics relate to success within individual offense groups.

The motor vehicle group is somewhat anomalous. Its success rate is much like that of assault and drug offenders but its
misconduct history is markedly better.

58

36

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
1. Age
Average age at release reflects both the age when people entered prison and how long they stayed.
Motor vehicle offenders are the oldest at release because drunk drivers are not sent to prison until
they have been convicted several times and thus are older when they first go to prison. Although
they are typically facing their first conviction, sex offenders are also older when they are sentenced.
Homicide offenders are also older at release but are much more likely to have been young when
they committed their crimes and to have served long terms. At the other extreme, burglars are the
youngest when released because breaking and entering is often a young man’s crime.
There was a strong tendency for those who succeeded upon release to be older and for those who
fail most seriously, by being convicted of a new offense, to be markedly younger. Technical parole
violators tended to fall in between. However, there were noteworthy differences among offense
groups. The age ranges were widest for sex and homicide offenders, but it was the homicide group
that followed the pattern exactly. Sex offenders who failed were the same average age, whether
they returned with new sentences or technical violations. The robbery and burglary groups
showed little age difference no matter what the outcome. In those groups, it was actually the
technical violators who were oldest. Moreover, the average age of those who succeeded in some
groups was similar to the age of those who failed in other groups. Thus there was no specific age
that, independently of offense type, predicted success on release. See Appendix D, Table B.
2. Prefix
Each time a person is returned to a Michigan prison with a new sentence for a new conviction, s/he
retains the same prisoner number but the letter prefix is changed. Someone with an “A” prefix is
serving a sentence in a Michigan prison for the first time; someone with a “D” prefix is serving a
fourth. While the structure of the database did not permit analysis of all prior convictions, it did
allow examination of the relationship to success of prior incarceration.
Prefix is only a rough surrogate for prior criminal record. While a higher prefix clearly shows more
sentences to prison, the prefix does not reflect convictions that did not result in Michigan prison
terms. People with a history of less serious offenses are likely to have received probation or jail
terms before finally being imprisoned while those convicted of assaultive crimes are more likely to
receive a prison term even if they have no prior record. Thus an “A” prefix may belong to an older
person finally sent to prison for a third drunk driving offense or a 16-year-old whose first adult
conviction was for murder.
Prefix does focus on an important characteristic that prior record does not reflect – the impact of
the prison experience. By distinguishing those who have been released after serving their first
term from those who had been to prison before, it is possible to draw some inferences about the
deterrent effect of prison on different groups of offenders.
Eighty-one percent of people released from prison during the study period were serving their first
prison term. The proportion was highest for sex (90.7%), motor vehicle (88.7%) and drug (87.6%)
offenders. However, even for burglary, the group with the relatively fewest “A” prefixes, the
proportion was 70%.
People serving their first prison sentences tended to succeed more than people who had been in
prison before. Thus the offense groups with the highest success rates had the highest share of “A”

Citizens Alliance on Prisons and Public Spending

37

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
prefixes. The offense groups with the lowest success rates had relatively more people who had
been to prison before.
It is impossible to know how much people succeed because they have been deterred by prison
itself. They may simply have matured while incarcerated. Some may have committed situational
offenses that were unlikely to recur in any event. For some, the shame of being caught and
convicted, or the burden of the guilt they carry, would have been deterrent enough. However, we
can see that people who were not deterred by prior imprisonment from committing another
offense do not do as well when released from the subsequent sentence. On average, success rates
were 13.3 points lower for people with a “B” prefix. Notably, there was little difference in success
rates between those who had served one prior prison term and those who had multiple prior terms.
The extent of the relationship between prefix and success varied substantially among offense
groups. Most showed a difference in success rates of 8.5 to 10.5 points between those with “A” and
“B” prefixes. However, for the homicide group the disparity was 20 points while for the weapons
group it was less than five. Only the homicide and robbery groups showed the pattern one might
intuitively expect of a steady, substantial decline in success and a concomitant increase in returns
with new sentences as people move from “A” prefix to “B” prefix to “C” or higher. In other offense
groups, either the decline in success rates from “B” to “C” prefix was minimal or success actually
increased. Apparently, serving one’s first prison term increases the chance of success but serving a
third or fourth term does not necessarily decrease it. Indeed, after returning to prison enough
times, people may eventually get tired and stop.
While success rates declined with more prison terms, offense type appeared to have a strong
independent impact. Even for people with a “C” prefix or higher, success rates were 70% or greater
for motor vehicle and sex offenders; they were 60% or greater for assault and drug offenders.
However, for the robbery, larceny and burglary groups, even those with “B” prefixes had success
rates below 50%. For further detail, see Appendix D, Table C.
3.

Misconducts

In Michigan, prison rule infractions are characterized as either major or minor misconducts. Major
misconducts are divided into two types: “nonbondable” and “bondable.”
a) Nonbondable misconducts
Nonbondable misconducts are those that are so serious that the person is immediately placed in
disciplinary segregation pending a hearing. They include assaults on staff or other prisoners,
threatening behavior and escape attempts -- often actions that would violate the criminal law.
Nonbondable misconducts are relatively uncommon. In nearly 80% of the cases, people went for at
least three years without receiving a citation or “ticket” for one. There was no more than one
nonbondable misconduct in more than 96% of the cases.
As Figure 7 indicates, the relationship between nonbondable misconducts and success or failure on
parole was strong. As the number of nonbondables went up, success declined sharply and returns
with new sentences increased. Among people with no nonbondables, more than 65% succeeded
and 16% returned with new sentences. Among people with two nonbondables, fewer than 49%
succeeded and new sentence returns increased by more than 12 points to 28.6%. Thus it appears

38

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
that people who engage in assaultive or other serious misbehavior in the highly structured
environment of prison have a more difficult time avoiding new criminal behavior upon release.

59

The relationship between nonbondable misconducts and returns with new sentences appeared in
every offense. However, it did not explain why some groups had so much higher return rates than
others. At one extreme, roughly 6.5% of homicide and sex offenders who had no nonbondable
misconducts returned with new sentences. At the other extreme, 23% of burglary and larceny
offenders who had no nonbondable misconducts returned with new sentences. That is, when the
effect of differing misconduct histories is eliminated, the disparities in failure rates among offense
groups remains. See Appendix D, Table D.
The relationship between nonbondable misconducts and technical parole violations was much
more erratic. Except for the assault and weapons groups, those who had no nonbondable
misconducts had a lower rate of return for technical violations than those who had one or two.

70

65
.3

Figure 7. Success and Failure by
Nonbondable Misconducts

48
.5

54
.7

60

Success

40

20

.4
18

21
.4

30

.9
23

28
.6

PVNS

16
.3

Percent

50

.9
22

PVT

10
0
0

1

2

Nonbondable Misconducts

However, for most groups there was little difference in the PVT rates for people with one
nonbondable misconduct and those with two. Thus a history of nonbondable misconducts is not as
reliable a predictor of technical parole violations as it is of new criminal conduct.
b) Total misconducts
“Bondable” misconducts are far more common than nonbondables. The most frequent citations are
for disobeying a direct order, being out of place and insolence. Given the multitude of prison rules,
the large number of staff with whom prisoners interact and the breadth of staff discretion, even
prisoners with excellent institutional records sometimes cannot avoid an occasional “ticket.” Thus,
as Table 9 showed, only a little over half the people went for at least three years with no more than
one misconduct. Table 9 also shows that the offense groups with the highest success rates had the
greatest proportion of people with zero or one misconduct of any kind.
59 Because fewer than 1% of the people in every offense group had three or more misconducts, the numbers were too
small to analyze.

Citizens Alliance on Prisons and Public Spending

39

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Figure 8 indicates that the relationship between total misconducts of any kind and success on
parole is similar to the relationship between nonbondable misconducts and success but much less
marked. People who had no misconducts of any kind had an even higher success rate (69.7%) than
those with no nonbondables (65.3%). This holds true in every offense group except larceny,
suggesting that people with the skills necessary to adapt to prison routines, such as patience, selfcontrol and flexibility, can apply these skills in readapting to free society.
Figure 8. Success and Failure by
Total Misconducts

56
.0

51
.0

54
.9

60

58
.4

59
.4

61
.0

65
.7

70

69
.7

80

Success

40
.6
22

.5
23

.9
22

26
.8

.1
22

21
.1

.2
21

18
.5

.3
19

17
.8

.4
16

15
.0

20

13
.9

30

21
.6

PVNS

19
.0

Percent

50

PVT
.2
22

10
0
0

1

2

3

4

5

6

7+

Total Misconducts

As Figure 8 shows, the success rate dropped four points with one misconduct and four more with
two misconducts. As the number of total misconducts increased, the decline in success was more
gradual. It actually ticked up at six misconducts before falling to 51% for seven citations or more.
The re-offense rate, which began at 13.9% for people with no misconducts, rose in small increments
with each additional citation to 21.6% at five. It fell back slightly at six misconducts, then rose to
nearly 27% for people with seven tickets or more.
Changes in the technical violator rate were less dramatic. That began at 16.4% for people with no
misconducts, peaked at 23.5% for people with five, and essentially flattened out once people had
three citations or more.
While this pattern varied among offense groups, overall, total misconducts were not as strong a
predictor of success as nonbondable misconducts and their relationships were less consistent
across offense groups.
B. Impact of age, prefix and misconducts on release time
Table 9 showed that the offense groups with the highest success rates tended to be older, to have
overwhelmingly “A” prefixes and to have excellent institutional conduct. These characteristics
make members of those groups tend to score “high probability of release” on the MDOC’s parole
40

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
guidelines. However, as Figure 2 demonstrated, achieving that score and actually being released
are two very different propositions.
Clearly, when the parole board decides whether to release someone, the nature of the offense
outweighs both the likelihood of success and the characteristics associated with success. Larceny
offenders rank eighth in success, age, prefix and nonbondable misconducts but third in release on
ERD. Sex offenders, who rank second in success, second on age, first in proportion with “A” prefix
and first in proportion with no non-bondable misconducts, nonetheless rank a very distant ninth in
release on ERD. Nonetheless, it is still useful to examine whether, within offense categories, age,
prefix and misconduct correlate with time of release.
Age. It does not appear that age has a strong impact on the timing of release. Overall, the
relationship of average age to release date was:
On ERD
30.5

1 Yr Past
30.0

2 Yrs Past
31.4

2-4 Yrs Past

5 Yrs Past

>5 Yrs Past

32.4

35.4

37.2

The dip in age for people who were one year past their ERD suggests that being older had some
influence on the board’s decision to grant parole when people first became eligible. However, the
steady increase in age for each year thereafter seems to just show that people got older as they
awaited release. That is, if people released on their ERD were 30 and people released five years
after their ERD were 35, the entire age difference is accounted for by the five-year delay. Certainly
there is no pattern of older prisoners being released more quickly.
Even the pattern of people held one extra year being the youngest did not hold for every offense
group. For assault, weapons and burglary offenders, those released on their ERD and those
released one year past were virtually the same age. Only the homicide and sex offenders showed a
dip in average age greater than seven months.
Prefix. There is no question that having served a prior prison terms affects the timing of release
decisions. Overall, the proportion of people released on or within one year of their ERD declined
from roughly 80% to 69% to 56% as it moved from “A” to “B” to “C” or higher prefix.
As Table 10 shows, having had just one prior incarceration caused a marked drop in release at first
eligibility for every offense group. However, the extent of the impact varied substantially. The
proportion of homicide, assault and robbery offenders released on their ERD was at least 25 points
lower for those with a “B” prefix than for those with an “A”. Even though they were not then
serving for an assaultive crime, for weapons offenders the drop from “A” to “B” prefix was still more
than 22 points, indicating the board’s concern that people who have been to prison once already
and return for a weapons violation are at greater risk for committing assault in the future. (There
may also be an association with the nature of the prior offense that the data does not reveal.) Less
explicably, drug offenders with “A” and “B” prefixes showed a difference in release when first
eligible of nearly 19 points. For the larceny, burglary and motor vehicle groups, the decline ranged
only from 12 to 15 points.
Whether a person has ever been in prison before is the most critical distinction. Once there is any
history of returning to prison, the frequency of return has somewhat less impact.

Citizens Alliance on Prisons and Public Spending

41

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?

Table 10 also shows that while serving a first prison term is a very important factor in obtaining
release sooner rather than later, it provides no guarantee. Although more than three-quarters of
motor vehicle and drug offenders with an “A” prefix were released when first eligible, for most
groups the proportion was roughly two-thirds. For sex offenders, it was one-third. In fact, a sex
offender serving his or her first prison term was less likely to be released on the ERD than most
other offenders who were serving their third.

Misconducts. It is not surprising that misconduct history correlates strongly with the time of
release. Board members intuitively believe that someone who cannot control his or her behavior in
the structured environment of prison will be unable to exercise control if released. And, as
discussed above, misconducts, especially nonbondable misconducts, do in fact have some value in
predicting returns to prison with new convictions.
Parole denial is also used to punish poor institutional behavior. Although a range of sanctions is
used to punish misconduct when it occurs, parole boards do not want to be seen as rewarding
misbehavior with favorable release decisions. Thus the relationship between misconduct and
release is about institutional management and public perception as well as risk assessment.
Table 10 illustrates the relationship clearly. For every offense group, release on the ERD declined
as the number of both nonbondable misconducts and total misconducts increased. The impact of
misconduct is even stronger than the impact of prior incarceration. On average, 76% of people with
no misconduct citations of any kind for the preceding three years were released on their ERD. By
comparison, fewer than 42% of those with one nonbondable were released when first eligible.
Having two nonbondables pushed the chance of release on the ERD below 20%
The differences among offense groups were similar in pattern to those for prefix but more
dramatic. Having no misconducts of any kind brought release on the ERD to more than 90% of the
motor vehicle and drug offenders and to 81% of the weapons and larceny groups. For the assault,
burglary, homicide and robbery groups, the proportion dropped to 70-75%. And, of course, good
institutional behavior did little to help sex offenders win release. Only 40% of those who had no
misconducts were paroled on their ERD.

42

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Of particular interest are the differences between people who paroled and those who maxed out.
Figure 9 makes two points quite clearly. First, in every offense group, people who maxed out were
much more likely to have had at least one nonbondable misconduct in the preceding three years.
The proportion of parolees who had zero nonbondables ranged from 15 points (sex offenders) to
nearly 38 points (drug offenders) more than the proportion of maxouts.
Second, while in most offense groups 50% or fewer of the people who maxed out had zero
nonbondables, for homicide and sex offenders, more than 70% had zero. That is, having
nonbondable misconducts could not have been the reason for denying release to seven of ten sex
and homicide offenders who were required to serve their maximum sentences.

Figure 9. Zero Nonbondable Misconducts by Offense Groups
Parolees and Max Outs

88

.2

89

91

.6

.4

100

7

79

8.
2

.1

.4
79

81

3
4.
7

.9

68

.1
*

72

76

.2

80

.2

82

.6

90

.5

45

.5

43

43

.6

47

50

.4

50

.4

.0

60
50

Percent

70

Parolees
Max Outs

40
30
20
10
Sex

Homicide

Motor
Vehicle

Assault

Drugs

Weapons

Larceny

Robbery

Burglary

In sum, in every offense group a poor conduct history substantially contributed to parole denial.
However, the extent to which good or even perfect conduct brought release varied. While most
prisoners could reasonably expect to earn their release by complying with institutional rules, this
was less true for those convicted of assaultive crimes and barely true at all for sex offenders.

V. The irrelevance of length of stay: more time in prison does not increase success
Success on parole is not related to the sheer number of years served in prison. Serving more time
does not decrease the chances of reoffending and may actually increase it.
Before considering the figures on length of stay, a caveat is in order. It must be remembered that
the database includes only the people who were released during the 14-year study period.
Particularly for the most serious offenses, both the average minimum sentence and the average

Citizens Alliance on Prisons and Public Spending

43

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
time served shown in Tables 11, 12 and 13 are shorter than Michigan prisoners actually serve.
These averages do not include people sentenced to life without parole for first-degree murder or to
parolable life terms for various crimes, including second-degree murder, assault with intent to
murder, armed robbery and first-degree criminal sexual conduct. In addition, people sentenced
after 1981 to very long minimum terms for such crimes may not have been released within the
study period. Since the people with very long sentences tend to be high on the factors associated
with success, their exclusion in no way undermines the findings.
A. The Michigan data on length of stay
The specific amount of time served bears no relationship to success. Table 11 indicates that
homicide and sex offenders, who have the highest success rates, also rank first and second in
average time served. But motor vehicle offenders, who rank third in success, are last in time
served. Robbery and sex offenders both served, on average, approximately 4.2 years but the
success rate of the sex offenders was 20 points higher. Similarly, drug and larceny offenders both
served, on average, 2.2 years, but the success rate of the drug offenders was nearly 15 points
higher. Conversely, assault and motor vehicle offenders had virtually identical success rates, but
they served, on average, 3.3 and 1.8 years, respectively.
Table 11. Overall Success Rate and
Average Years Served, By Offense
Success
Rate
Homicide
Sex
Motor Vehicle
Assault
Drugs
Weapons
Robbery
Larceny
Burglary

80.1%
77.6%
70.8%
70.0%
69.1%
62.3%
57.1%
54.5%
51.1%

Avg. Yrs.
Served
5.5
4.3
1.8
3.3
2.2
2.0
4.1
2.2
2.9

Time served is most directly a function of sentence
length, which is in turn a function of offense type and
prior record. That is, time served is a more accurate
reflection of the punishment imposed than it is a
predictor of success. This is readily demonstrated by
Table 12, which shows the relationship of time
served to success and to the minimum sentence.

If incarcerating people longer made them more likely
to succeed upon release, in each offense group the
average years served would be longest for those who
succeeded and shortest for those who failed. If the people most likely to succeed were released the
soonest, the average years served would be shortest for those who succeeded. Table 12 shows that
neither of these hypotheses is true. Only motor vehicle offenders had the fewest years served by
those who succeeded. Among assault, burglary and drug offenders and for the database overall,
those who succeeded served the longest. For the robbery, sex and larceny groups, those who
succeeded and those who returned with new sentences served the same amounts of time.
The more consistent trend was that time served tended to track the minimum sentence. In the
homicide, robbery, sex, and drug groups and in the database as a whole, those who served the most
time also had the longest minimums. In the robbery, burglary, drug and motor vehicle groups and
the database overall, those who served the shortest time also had the shortest minimums.
What is striking about Table 12 is how small the differences in time served were within offense
groups, regardless of the outcome. In six groups, the difference between the longest and shortest
average time served was 0.2 years, or 2.4 months. In two groups, the difference was 0.3 years, or
3.6 months. Only homicide offenders showed a greater disparity.

44

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Table 12. Minimum Sentence & Time Served* by Success and Failure
Success

Homicide
Robbery
Sex
Assault
Burglary
Drugs
Larceny
MotorVeh
Weapons
Total

Avg.
Min.
6.3
4.5
3.9
3.2
3.0
2.3
2.0
1.9
1.7
3.0

PVT
Years
Serv’d
5.4
4.1
4.3
3.3
3.0
2.3
2.2
1.7
2.0
3.0

Percent
of Min.
Serv’d
85.7
91.1
110.3
103.1
100.0
100.0
110.0
89.5
117.6
100.0

Avg.
Min.
7.1
4.6
4.0
3.2
2.8
2.0
2.0
2.0
1.7
2.9

PVNS
Years
Serv’d
6.0
4.2
4.5
3.1
2.8
2.1
2.0
1.8
1.9
2.8

Percent
of Min.
Serv’d
84.5
91.3
112.5
96.9
100.0
105.0
100.0
90.0
111.8
96.6

Avg.
Min.
6.7
4.5
3.8
2.8
3.0
2.0
2.0
2.0
1.7
2.7

Years
Serv’d
5.3
4.1
4.3
3.2
2.9
2.1
2.2
2.0
2.2
2.8

Percent
of Min.
Serv’d
79.1
91.1
113.2
114.3
96.7
105.0
110.0
100.0
129.4
103.7

*Includes consecutive two-year term for committing felony with a firearm, when imposed.

Table 12 also shows the extent to which people served more or less time than the minimum
sentence. In addition to parole board policies, time served was affected by disciplinary credits. In
the homicide and robbery groups, the time served was consistently less than the minimum
sentence regardless of the outcome on release, reflecting the award of disciplinary credits. This
was also true for most motor vehicle offenders. In other groups, the time served was equal to or
greater than the minimum. Disciplinary credits may have been unavailable for the offense (as was
the case for most drug offenders and those convicted of being habitual offenders) or denied for
misconduct or parole may have been denied even though credits were earned (as was undoubtedly
true for many sex offenders).
While differences among offense groups in the proportion of time served can be explained, within
offense groups these differences do not relate to success or failure. That is, just as success is not
consistently associated with the largest or smallest number of years served, it is also not
consistently associated with having served the largest or smallest proportion of the minimum
sentence. In homicide and burglary cases, those who returned with new sentences had served the
smallest proportion of their minimums before being released. In the sex, drug, motor vehicle and
weapons groups, those who succeeded had served the smallest proportion. In the assault and
larceny groups, those who came back as technical parole violators had served the smallest
proportion. And among released robbers, the proportion of the minimum served was virtually
identical, regardless of the outcome.
Table 12 makes three points quite clearly.
1. Serving more time in prison does not improve success upon release. On the contrary, in the
majority of offense groups, serving more time was associated with failure.
2. There is no magic number of years in prison that will guarantee success, for people in
general or for particular groups of offenders. Two years was the average time served by
weapons offenders who succeeded, larceny offenders who returned as technical violators
and motor vehicle offenders who returned with new sentences.
3. Within offense categories there is not a lot of difference in the amount of time served by
those who succeed and those who return to prison, either as technical violators or with new
sentences for new crimes.
Citizens Alliance on Prisons and Public Spending

45

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
B. The national research
Substantial research exists on the question of whether the experience of incarceration deters
people from reoffending. Studies have attempted to test several different theories.60
Classic “specific deterrence” theory is based on a model of rational choice. It assumes that
individuals choose to act in ways that maximize pleasure and minimize pain. According to this
theory, because prison is a highly unpleasant experience, prison sentences deter crime. The longer
the sentence, the greater deterrence value it will have. Thus, according to this theory, there should
be a direct correlation between longer lengths of stay and lower rates of recidivism.61
Another theory is that the deterrent effect of prison is offset at some point by the negative
consequences of incarceration on the individual’s ability to readjust upon release. The longer
someone sits in prison, the more their employment prospects and support networks erode. As the
prospects for conventional success diminish, a criminal life style may be viewed as the only rational
alternative. Thus, according to this theory, the relationship between time served and recidivism is
U-shaped. Prison sentences provide a net deterrent effect up to some optimum point, after which
they actually increase the probability of future crime.
A third theory, sometimes referred to as prisonization, views prisons as “schools of crime” that
reinforce antisocial norms and attitudes. According to this theory, the longer people are
incarcerated, the more likely they are to internalize the norms of prison culture. Thus, it is
suggested, there is a direct correlation between longer length of stay and increased recidivism.62
The theoretical explanations can be debated, but the empirical evidence is overwhelming. Increased
length of stay does not reduce recidivism and may actually increase it.
Song and Lieb reviewed four studies of whether, after controlling for other factors, length of
sentence affects recidivism.63 While one showed slightly lower recidivism rates for armed robbery
and drug offenders who had longer sentences, overall either there was no substantial relationship
between time served and recidivism rates or rates increased with longer sentences. In addition,
For more detailed discussion of these theories see Thomas Orsagh and Jonh-Rong Chen, The Effect of Time Served on
Recidvism, 4 Journal of Quantitative Criminology, No. 2, 155-171 (1988) and Lin Song and Roxanne Lieb, Recidivism: The
Effect of Incarceration and Length of Time Served, Washington State Institute for Public Policy (1993);
http://www.wsipp.wa.gov.
60

The literature does not address the issue of pain being relative. The deterrent effect of prison could vary depending on
how pleasant or painful someone’s life was in the community.

61

62 Whatever the merits of the prisonization theory are, a variation on it may be worth exploring. Anti-social attitudes may
be encouraged or reinforced when length of stay is extended by parole denial. Prisoners, particularly those who pled
guilty in exchange for a specific minimum sentence, may feel themselves to be a party to a contract with authorities. They
accept their obligation to serve the minimum sentence, conduct themselves appropriately and complete required
programs while waiting to become eligible for parole. If they feel they have met these expectations and are denied parole
nonetheless for reasons they do not see as justified, resentment of authorities who have not upheld the contract may
negatively affect their willingness to play by the rules in the future. Bukstel and Kilmann make a related point when they
note that indeterminate sentencing may contribute to adjustment problems because of the uncertainty of the release date.
Lee Bukstel and Peter Kilmann, Psychological Effects of Imprisonment on Confined Individuals, 88 Psychological Bulletin,
No. 2, 469-493 (1980).
63

Note 60 supra.

46

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
three studies of early release programs undertaken to relieve prison overcrowding in California,
Washington State and Illinois concluded that releasing people anywhere from 3.5 to 6 months
before their scheduled parole dates did not increase recidivism.
In 1999, Gendreau, Goggin and Cullen conducted a meta-analysis that included 23 studies of the
impact of more versus less time in prison.64 In 2002, Smith, Goggin and Gendreau published a
follow-up meta-analysis that included 26 additional studies. 65 Neither analysis found any deterrent
effect. On the contrary, longer length of stay was associated with a small increase in recidivism.
The authors note that even slight increases in recidivism can be very costly when they occur on a
large scale and suggest that excessive incarceration may be fiscally irresponsible for that reason.
An August 2001 report on parole decisionmaking in Hawaii analyzed numerous factors to
determine which ones best predicted success on parole. It found that the total length of stay in
prison did not significantly affect the odds of parole revocation.66
The New York Department of Correctional Services looked at return rates by time served for people
who were first released in 2003 and returned for any reason within three years. It concluded that
“parole releases who served the most [5+ years] and least [12 months or less] amount of time…had
the lowest returns rates.” Those who served 13-18 months had the highest rates. Among
conditional releases, return rates “tended to vary.” Those who had served at least five years had the
lowest rate; those who served 3-4 years had the highest.67
There were similarly mixed results from a Bureau of Justice Statistics (BJS) study that tracked
272,111 former prisoners from 15 states for three years after they were released in 1994.68 For
people who had served anywhere up to 60 months before being released for the first time, re-arrest
rates ranged from 62.6% to 68.3%. Within that range, as time served increased, the rates moved
down, then up, then down again, peaking for the 25-30 month group. Thus there appears to be no
clear relationship between length of stay and re-arrest.69
However, for those serving more than 60 months, rearrest rates dropped substantially to 54.2%.70
There are several possible explanations. It may be that longer length of stay in itself does reduce
recidivism but the time needed to obtain that effect is far longer than the median length of stay for
all prisoners in the study, which was 13.3 months. It may be, as Marc Mauer of The Sentencing
Paul Gendreau, Claire Goggin and Francis T. Cullen, The Effects of Prison Sentences on Recidivism, Solicitor General of
Canada (Ottawa, 1999).

64

Paula Smith, Claire Goggin and Paul Gendreau, The Effects of Prison Sentences and Intermediate Sanctions on Recidivism:
General Effects and Individual Differences, Solicitor General of Canada (Ottawa, 2002).
65

66 Gene Kassebaum, Janet Davidson-Coronado, Paul Perrone and Joe Allen, Parole Decision Making in Hawaii, Social
Science Research Institute, University of Hawaii at Manoa and Research and Statistics Branch, Crime Prevention and
Justice Assistance Division, Dept. of the Attorney General (August 2001), www.cpja.ag.state.hi.us. , at p 3.
67Kellam
68

and Hayes, note 34 supra, at p 28.

Langan and Levin, note 32 supra.

69 The report does not compare length of stay to any of the other three recidivism measures it uses, including returns to
prison with a new sentence.

70

Since this time period is open-ended and no further interval breakdowns were given, the extent of variation within this
group and the potential effect of a small number of very long sentences is unknown.

Citizens Alliance on Prisons and Public Spending

47

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Project suggests, that the improved rearrest rate is due to the greater age of people who have
served longer.71 Or, as the Michigan data suggests, it may also be a function of offense type. That is,
the offenses for which people serve the most time include those which have the lowest recidivism
rates for other reasons.
Rosenfeld, Wallman and Fornango further analyzed the BJS data and found no significant
association between the number of months served and the incidence of rearrest. The authors
concluded this result “carries an important policy implication: Reducing time served in prison
evidently would not elevate recidivism, as measured by arrests for new crimes.”72
Canadian researchers focused on 627 sex offenders who were assessed between 1983 and 1995.
More than 60% had been incarcerated while the remainder had been sentenced to community
supervision. The average time of opportunity to reoffend was 8.11 years. About 13% of the sample
had a new charge or conviction for a sex offense. There was no significant association between
recidivism and either the fact of incarceration or the length of stay. The authors found no evidence
to support prison as a deterrent or as a school of crime and no suggestion “that there is some
optimal length of incarceration for sexual offenders that maximally reduces recidivism compared to
shorter or longer periods of incarceration.”73
Overall, studies of various types, done over several decades by numerous researchers, leave little
doubt that keeping people in prison longer, especially by a few additional years or less, does
nothing to enhance public safety and may well be counter-productive.
C. Increased length of stay under the new board
The obvious result of keeping more people for longer periods after their first parole eligibility date
is that their total length of stay increases. For every offense group, people released by the new
board served more time than those released by the old board. Overall, the difference was 0.8 years,
or 9.6 months. However, as Table 13 indicates, half the difference was explained by the fact that
new board releases in every offense group had longer average minimum sentences imposed by the
courts. Since the board cannot release anyone who has not served his or her minimum sentence
(less applicable credits), longer minimums must result in more time served. The average minimum
increased by 0.4 years, or 4.8 months. For most offenses the increase in the minimum ranged from
0.1 to 0.6 years. For homicides, the increase was 1.4 years, or nearly 17 months.
The other half of the increase in time served, another 4.8 months overall, was attributable to
discretionary decisions by the new board. Overall, under the old board, people served 2.5 years on
average, or 92.6% of the average minimum sentence of 2.7 years. Under the new board, people
served 3.3 years on average, or 106% of the average minimum sentence of 3.1 years. As we would
expect, the extent of the difference between the minimum sentence and actual time served varied
by offense.

71

Marc Mauer, The Hidden Problem of Time Served in Prison, 74 Social Research, No. 2, 701 (Summer, 2007), at p 703..

72

Richard Rosenfeld, Joel Wallman and Robert Fonango, “The Contribution of Ex-Prisoners to Crime Rates,” pp
80-104, at p 95, in Travis and Visher, note 3 supra.
Kevin Nunes, Philip Firestone, Audrey Wexler, Tamara Jensen and John Bradford, Incarceration and Recidivism among
Sexual Offenders, 31 Law and Human Behavior, No. 3, 305-318 (2007), at p 314.

73

48

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
There was relatively little difference between the two boards’ treatment of burglary, larceny, drug
and motor vehicle offenders. These groups served, on average, an additional few months as the
result of board decisions. The difference was more marked for weapons offenders, who went from
serving 2.4 months beyond their minimum sentences to six months beyond.
Table 13. Average Years Served Relative to Minimum Terms*
Old Board and New Board Compared
Old Board

New Board

Impact of Increased Years Served

Av.
Years
Served

Av.
Min.

Min.
Minus
Years
Served

Av.
Years
Served

Av.
Min.

Min.
Minus
Years
Served

Diff. in Min.
Minus
Years
Served

New Bd.
Cases
(N=)

Add’l.
Years
Served

Av.
Add’l.
Beds/
Year

Homicide

4.2

5.6

1.4

6.6

7.0

0.4

1.0

1,387

1,387

198

Robbery

3.5

4.3

0.8

4.7

4.7

0.0

0.8

4,609

3,687

527

Sex

3.4

3.6

0.2

5.1

4.1

-1.0

1.2

3,670

4,404

629

Assault

2.7

2.9

0.2

3.6

3.3

-0.3

0.5

3,948

1,974

282

Burglary

2.7

2.8

0.1

3.2

3.1

-0.1

0.2

4,736

1,947

135

Larceny

2.0

2.0

0.0

2.3

2.1

-0.2

0.2

7,415

1,483

212

Drugs

1.9

1.9

0.0

2.5

2.4

-0.1

0.1

9,497

950

136

Weapons

1.7

1.5

-0.2

2.4

1.9

-0.5

0.3

1,280

384

55

Motor Vehicles

1.6

1.7

0.1

1.8

2.0

0.2

0.1

2,461

246

35

Total

2.5

2.7

0.2

3.3

3.1

-0.2

0.4

39,003

15,601

2,229

*Sentences include additional 2 years for felony firearm conviction, where applicable.

The disparity between the two boards was increasingly pronounced for the assault group. Under
the old board, the time served was 0.2 years (2.5 months) less than the minimum; under the new
board, the time served was 0.3 years (3.6 months) more. Thus the proportion of the minimum
sentence served went from 93% to 109%.
The robbery group shows a still larger disparity between the two boards. Under the old board,
these offenders served, on average, 81% of the minimum while under the new board they served
the entire minimum. An increase in time served of 0.8 years (9.6 months) was attributable to the
new board’s policies.
The biggest change in the proportion of the minimum served occurred for homicide and sex
offenders. Under the old board, homicide offenders served, on average, 1.4 years less than their
minimum sentences or 75%. The new board reduced the gap between time served and the
minimum term to 0.4 years, so that homicide offenders served 94% of their minimums. For sex
offenders, the new board dramatically reversed the proportion of the minimum sentence served.
Under the old board, sex offenders served 94% of their average minimum. Under the new board,
sex offenders were kept for 1.2 years beyond their minimum, thus serving, on average, 124%.
Citizens Alliance on Prisons and Public Spending

49

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Table 13 shows the impact of the changed policies on prison bed space. The column “Min. Minus
Years Served” shows the extent to which the average length of stay was longer or shorter than the
average minimum sentence. The column “Diff. in Min. Minus Years Served” shows the increased
length of stay under the new board that was not due to longer minimum sentences. Multiplying the
number of cases the new board had in an offense group by the increased amount of time served
yields the total additional number of years served. For the seven-year period, the new board’s
policies resulted in 15,601 additional years served. Thus, on average, 2,229 more beds were
needed each year than would have been required under old board policies. Nearly 30% of the
additional beds were for sex offenders.

VI. How much does denying parole increase public safety?
The Michigan statutory standard for granting parole, found in MCL 791.233(1)(a), is:
A prisoner shall not be given liberty on parole until the board has reasonable assurance,
after consideration of all of the facts and circumstances, including the prisoner’s mental and
social attitude, that the prisoner will not become a menace to society or to the public safety.
Thus, when the parole board denies parole, it is concluding that releasing the person would pose
too great a risk to public safety and that the cost of continued incarceration to the prisoner and the
public is warranted. Since the parole board’s role is to assess each prisoner’s risk of reoffending,
one would expect the people released at first eligibility to do better than those the board has
continued for some period. If the board’s predictive abilities were perfect, all the people released
when they first became eligible would succeed and all those paroled after their earliest release date
would fail. If the two groups had identical success rates, it would indicate a total inability on the
part of the board to distinguish those who present a risk from those who do not. The question,
then, is how effectively does the board make this distinction?
A. Success rates of those kept an extra year or two (or three or four)
More than 61% of all prisoners were released on their earliest date; 15.3% were released after
serving one additional year; another 14.6% were released after serving two additional years. Thus,
more than 91% of all prisoners were released within two years of first eligibility. Nearly 7% more
were released within four years. Fewer than 2% were kept more than four years beyond their ERD.
Figure 10 confirms that, in general, people paroled after their earliest release date tend to be less
successful than those who are released on their ERD. They have higher rates of returning with both
new sentences and technical violations. Since success does not improve with the denial of parole,
increased rehabilitation cannot be the justification for denials. In fact, repeated parole denials may
actually decrease success as time erodes people’s support systems, their ability to function
independently and the opportunities available to them in the community. Thus the justification for
parole denial has to be the extent to which further incapacitation protects the public.
Figure 10 shows a striking pattern. The success rate began at 66.1% for those released on their
ERD, then dropped 8.5 points to 57.6% for those released one year after. The technical violator
rate increased by 3.6% and the new sentence return rate increased by 4.9%. However, after that
initial decline, all the rates flattened out. For those released anywhere from one to four years after
their ERD, each type of outcome stayed constant. Then, for people kept more than four years past
50

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
their ERD, the success rate climbed by 8.3 points, returning to virtually the same level achieved by
those who were released when first eligible.

57
.9

60

56
.7

57
.6

66
.1

70

66
.2

Figure 10. Percent of People Who Succeed or Fail by Release Time

Success

40

PVT
30
.4
21

21
.6

6

.6
20

.7
15

17
.7

18
.2

20

.
20

21
.9

PVNS
21
.8

Percent

50

.1
16

10
0
ON ERD

1 YR. Past 2 Yrs. Past

2-4 Yrs.
Past

>4 Yrs.
Past

Release Time

These patterns make two points quite clearly. First, the parole board was only modestly adept at
predicting who was likely to succeed upon release. Overall, the people it chose to parole at first
eligibility had success rates almost nine points higher and returns with new sentences about five
points lower than the people it continued for at least a year. But for the people it decided to keep
anywhere from one to four years past their ERD, parole denials were much like a coin toss. If the
goal was to anticipate the likelihood of failure, the board did not do it accurately more than half the
time.
Second, keeping people two, three or four years beyond their first eligibility date made virtually no
difference. On the one hand, it did nothing to increase success rates. That is, people generally did
not do better because of the additional years served. On the other hand, it did not provide the
public with any greater measure of protection. That is, since success rates were not lower for
people who were denied parole for several years, it does not appear the public was safer than if
parole had been denied for just one year.74
The fact that people released on their ERD and those released more than four years after had
virtually identical rates for every outcome is more difficult to interpret, particularly since the latter
was a relatively small group of people who were denied parole for widely varying lengths of time.
It may indicate that those kept the longest gained something positive from their continued
incarceration, or at least “aged out” of criminal behavior. It may also simply indicate that these
people would have been just as successful if they had been released when first eligible. Notably,
It is theoretically possible that the people denied parole would have had worse outcomes if they had been released
sooner. However, there is no evidence to support this speculation and all the research on the lack of a relationship
between length of stay and success upon release suggests such speculation would not be justified.
74

Citizens Alliance on Prisons and Public Spending

51

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
more than half of them were homicide, sex and assault offenders who had higher success rates than
other offenders no matter when they were released. 75
Table 14 shows each offense group’s success and failure rates according to how long parole was
denied. It reveals substantial differences among the groups that the overall pattern masks.
Homicide. There was a marked difference between homicide offenders released on their ERD and
those kept for one year, but virtually no difference between those kept for one year and those kept
for two. There was another marked decline in success for those kept two to four years past their
ERD. As a result, the homicide group showed the greatest difference in success between those
released when first eligible and those kept up to four years -- a decline of 21.5 points. Nonetheless,
even those homicide offenders paroled one or two years after their ERD had success rates above
70%. PVT and PVNS rates both roughly mirrored the changes in success at each point. The PVNS
rate more than tripled for people kept up to four years past their ERD but still never reached 15%.
Sex. Sex offenders kept any amount of time past their ERD showed very little difference in success
and PVNS rates. Their success rates were 75% or higher and their PVNS rates were 10% or lower
no matter when they were released. Even those sex offenders released from two to four years past
their ERD had higher success and lower PVNS rates than the people released when first eligible in
every other group except homicide. Thus, for sex offenders, repeated parole denials seemed to
have little relationship to increased public safety.
Motor Vehicle. The nearly 80% of motor vehicle offenders released when first eligible and the
additional 12% kept for one additional year had virtually identical outcomes. However, there was
nearly a 10 point decline in success rates between those kept for one additional year and those kept
for two. The decline was almost wholly attributable to an increase in returns with new sentences
as the returns for technical violations increased very little. Thus parole denials for people kept two
years past their ERD appear to have been related to the risk of reoffending.
Assault. There was a moderate difference between assault offenders released on their ERD and
those kept for one year. For those kept longer, PVNS rates showed small but steady increases, but
the success rate flattened and then actually increased, due to a decline in returns for technical
violations. At least two-thirds of assault offenders succeeded no matter when they were released.
Thus, the benefit to public safety of repeated parole denials for assault offenders was modest.

It must also be recalled that, as the following table shows, the proportion of people who max out increases as time past
the ERD increases. Since people who max out cannot be returned for technical parole violations, they have higher success
rates. But for the greater number of max outs, the number of technical parole violators might have been greater and the
success rate proportionally lower for those people who were held three, four and more years past their ERD. However,
the pattern of returns with new sentences would have been unaffected.

75

Proportion of Max Outs per Year Past ERD

Parolees
Max Outs

52

ON

1

2

2-4

>4

Total

45,001
100.0%

10,537
94.1%

9,387
87.8%

3,668
72.8%

817
58.6%

69,410
94.7%

656
5.9%

1,301
12.2%

1,372
27.2%

578
41.4%

3,907
5.3%

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Table 14. Success and Failure by Release Time and Offense Group

Homicide
Success
PV Tech
PVNS
TOTAL
Sex
Success
PV Tech
PVNS
TOTAL
Motor Vehicle
Success
PV Tech
PVNS
TOTAL
Assault
Success
PV Tech
PVNS
TOTAL
Drugs
Success
PV Tech
PVNS
TOTAL
Weapons
Success
PV Tech
PVNS
TOTAL
Robbery
Success
PV Tech
PVNS
TOTAL
Larceny
Success
PV Tech
PVNS
TOTAL
Burglary
Success
PV Tech
PVNS
TOTAL
Grand Total
Success
PV Tech
PVNS
TOTAL

Rel’d
on ERD
No.
%
1,342 86.1
150
9.6
67
4.3
1,559

1 year
past ERD
No.
%
202
71.9
49
17.4
30
10.7
281

2 years
past ERD
No.
%
258 71.9
67 18.7
34
9.5
359

2-4 years
past ERD
No.
%
159 64.6
52 21.1
35 14.2
246

>4 years
past ERD
No.
%
81 77.1
14 13.3
10
9.5
105

1,771
273
127
2,171

81.6
12.6
5.8

719
106
95
920

78.2
11.5
10.3

1,375
319
140
1,834

75.0
17.4
7.6

937
209
106
1,252

74.8
16.7
8.5

358
75
42
475

75.4
15.8
8.8

1,812
323
382
2,517

72.0
12.8
15.2

278
59
60
397

70.0
14.9
15.1

143
38
55
236

60.6
16.1
23.3

28
4
12
44

63.6
9.1
27.3

1
0
1
2

50.0

2,690
629
403
3,722

72.3
16.9
10.8

597
172
141
910

65.6
18.9
15.5

766
189
205
1,160

66.0
16.3
17.7

399
72
106
577

69.2
12.5
18.4

105
15
21
141

74.4
10.6
14.9

8,722
2,005
1,453
12,180

71.6
16.5
11.9

1,407
474
347
2,228

63.2
21.3
15.6

768
285
235
1,288

59.6
22.1
18.2

203
82
94
379

53.6
21.6
24.8

44
10
10
64

68.8
15.6
15.6

1,025
307
247
1,579

64.9
19.4
15.6

42
96
76
414

58.5
23.2
18.4

193
65
82
340

56.8
19.1
24.1

97
18
48
163

59.5
11.0
29.4

13
4
7
24

54.2
16.7
29.2

3,114
986
731
4,831

64.5
20.4
15.1

705
343
284
1,332

52.9
25.8
21.3

772
488
384
1,644

47.0
29.7
23.4

412
302
206
920

44.8
32.8
22.4

143
72
67
282

50.7
25.5
23.8

5,847
2,125
2,255
10,227

57.2
20.8
22.0

1,397
655
764
2,816

49.6
23.3
27.1

1,054
429
690
2,173

48.5
19.7
31.8

394
133
250
777

97
21
29
147

66.0
14.3
19.7

3,419
1,412
1,384
6,215

55.0
22.7
22.3

901
485
509
1,895

47.5
25.6
26.1

733
460
461
1,654

44.3
27.8
27.9

287
215
180
682

81
36
38
155

52.3
23.2
24.5

29,742
8,210
7,049
45,001

66.1
18.2
15.7

6,448
2,439
2,306
11,193

57.6
21.8
20.6

6,062 56.7
2,340 21.9
2,286 21.4
10,688

2,916
1,087
1,037
5,040

923
247
225
1,395

66.2
17.7
16.1

Citizens Alliance on Prisons and Public Spending

50.7
17.1
32.2
42.1
31.5
26.4
57.9
21.6
20.6

50.0

53

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Drugs. The drug offenders held past their ERD showed a steady decline in success and a steady
increase in PVNS rates for each year up to four. Success rates dropped by a total of 18 points;
returns with new sentences grew by nearly 13. Notably, PVT rates increased for people kept one
additional year, then held steady. With more than 75% of drug offenders released when first
eligible, for this group it appears that parole denials were directly related to the risk of reoffending.
Weapons. Success rates declined moderately for weapons offenders held for one or two years past
their ERD, then increased for those kept two to four years. The drop in success rates was tempered
by returns for technical violations which rose moderately for people kept an additional year, then
declined steadily. However, returns with new sentences grew steadily every year, peaking at 29.4%
for those kept up to four years. With 13.8 points difference between those released on their ERD
and those denied parole for up to four years, weapons offenders had the largest increase in PVNS
rates of any group. Thus it appears that parole denials were directly related to the risk of
reoffending.
Robbery. As with the drug group, robbery offenders showed a steady decline in success rates.
There is a difference of nearly 20 points between those released when first eligible and those
released up to four years after. However, unlike the drug group, the decline in success was due
much more to an increase in technical violations than to returns with new sentences. While both
types of failure increased roughly 6 points for those kept one additional year, PVT rates continued
to increase steadily, reaching nearly 33% for those kept up to four years, while PVNS rates rose
another two points, then dropped back one. While keeping people for more than one year past
their ERD had little relationship to returns with new sentences, there was clearly a relationship to
failure overall.
Larceny. The larceny group followed the general pattern for success rates. Success dropped 7.4
points for those kept one additional year but decreased only slightly more for those kept two years
and then began to rise. However, unlike the general pattern, failure rates for the larceny group did
not increase moderately for people kept one additional year and then plateau. Instead, both types
of failure increased moderately, but PVT rates then declined to 17% for those kept up to four years
while PVNS rates continued to grow to 32%. This pattern of declining PVT rates is similar to the
assault and weapons groups.
Burglary. Although the burglary group started out with success rates 10 points lower than the
robbery group, it showed a very similar pattern of changes in success and failure for people
released after their ERD. Success rates dropped the most for people kept one additional year but
continued to decline steadily for people kept up to four years past their ERD, creating a total
difference of nearly 13 points. Two-thirds of the decline in success was due to increased returns for
technical violations, which grew steadily and reached nearly 32% for those kept up to four years.
The PVNS rate rose only two points for people kept two years past their ERD, then dropped back
nearly two points. At 42%, burglars kept up to four years past their ERD had the lowest success
rate of any group at any point. Thus, keeping burglars past their ERD had a modest relationship to
returns with new sentences but a marked one with failure overall.
These summaries permit four conclusions.
1. Denying parole for up to four years did not increase success rates or decrease PVNS rates
for any group compared to the rates of those who were released when first eligible.

54

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
2. There was no consistent relationship between parole denial and success or failure that
applied to all offense groups.
a. For some groups, success declined steadily with each year of delay while others saw
virtually no difference at particular points.
b. For some groups, especially drug and weapons offenders, each delay in release was
related to increased rates of reoffending. For several groups, this relationship was
more modest and for sex offenders it barely existed.
c. Repeated parole denials showed steady increases in technical violations for some
groups (homicide, motor vehicle, robbery, burglary), decreases in other groups
(assault, weapons, larceny), an irregular pattern for sex offenders and virtually no
change for drug offenders.
3. There was no relationship between each year a person was kept past the earliest release
date and a particular amount of change in success or failure rates. For example, the
difference in success rates for people denied one year ranged from 14 points for the
homicide group to two points for motor vehicle offenders. Between people held one
additional year and those held for two, the change in PVNS rates ranged from a drop of
nearly three points for sex offenders to an eight point increase for motor vehicle offenders.
4. The rank order of offense groups by success stayed roughly the same at each release point.
That is, whether released on their ERD or up to four years after, homicide, sex and assault
offenders had much higher success rates than robbery, larceny and burglary offenders.

B. Impact of parole denial on crimes against persons
The public’s greatest concern is that people imprisoned for assaultive or sexual crimes will commit
similar crimes again when released. Table 2 demonstrated that, in fact, only 4.5% of people who
were released during the 14-year period returned to prison for new crimes against people. The
question that remains is how this rate is affected by parole board decisions. Table 15 takes a closer
look at the offense groups involving crimes against persons and the extent to which denying parole
for a specific number of years prevented further crimes against persons.
The results for every group are striking.
Homicide
•
•
•

Of 281 people who were incarcerated for one additional year, 12 returned to prison with a
new sentence for a crime against a person, including two new homicides.
Of 359 people who were incarcerated for two additional years, nine returned with a new
crime against a person, including three new homicides.
Altogether, 849 of 885 homicide offenders who were kept from one to four years past their
earliest release dates, or 95.9%, were not returned to prison for any new crime against a
person, compared to 98.4% of people released when first eligible.

Citizens Alliance on Prisons and Public Spending

55

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?

Assault
•
•

56

Of 910 people who were incarcerated for one additional year, 55, or 6.0%, returned to
prison with a new crime against a person, including seven homicides and 35 assaults.
Of 1,159 people who were incarcerated for two additional years, 87, or 7.5%, returned to
prison with a new crime against a person, including nine homicides and 46 assaults.
Notably, the return rates for those kept two to four and more than four years past their ERD
were nearly identical to the two-year group.

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
•

Altogether, 2,456 of 2,646 assault offenders who were kept up to four years past their
earliest release dates, or 92.8%, were not returned to prison with a new sentence for any
new crime against a person, compared to 95.7% of those released on their ERD.

•

Of 919 people who were incarcerated for one additional year, 51, or 5.5%, returned to
prison with a new crime against a person, including 40 new sex offenses.
Of 1,832 people who were incarcerated for two additional years, 80, or 4.3%, returned to
prison with a new crime against a person, including 55 new sex offenses. The return rates
for those kept two to four and more than four years past their ERD were very similar to the
two-year group.
Altogether, 3,807 of 3,998 sex offenders who were kept from one to four years past their
earliest release dates, or 95.2%, were not returned to prison within four years with a new
sentence for any new crime against a person, compared to 97.1% of those released on their
ERD. Only 3.5% returned for a new sex offense.

Sex

•

•

Robbery
•
•
•

Of 1,330 people who were incarcerated for one additional year, 110, or 8.3%, returned to
prison within four years with a new sentence for a crime against a person, including nine
homicides and 70 robberies.
Of 1,644 people who were incarcerated for two additional years, 165, or 10.0%, returned
with a new crime against a person, including 12 homicides and 107 robberies. Return rates
were nearly identical for those released two and two to four years past their ERD.
Altogether, 3,521 of 3,894 robbery offenders who were kept from one to four years past
their earliest release dates, or 90.4%, were not returned to prison within four years with a
new sentence for any new crime against a person, compared to 93.7% of those released on
their ERD.

In sum, to avoid increasing returns for new crimes against people from 4.5% to 6.9%, 9,664 people
from four offense groups who apparently would have committed no new offense of any kind were
incarcerated up to four years after they became eligible for release. Of these, 2,893 were continued
for one year, 4,230 were continued for two years and 2,541 were continued for three to four. An
additional 1,001 people, 86% of whom did not return for new offenses when they were ultimately
released, were continued for more than four years.
C. The minimal impact of parole decisions on crime rates
Concern that increasing parole grant rates would threaten public safety rests on the fact that nearly
one-fifth of prisoners return to prison with a new sentence within three years of their release. It is
also true that, compared to the general population, parolees as a group commit a disproportionate
share of total crime. These facts make it highly desirable to improve both risk prediction, to the
extent possible, and support services that reduce the likelihood of recidivism.
However, when the question is what impact on total crime will result from paroling more people
when they first become eligible, two other facts must be remembered. First, the vast majority of
crimes are not committed by parolees. Second, people denied parole for a few extra years do not
commit crime at a much higher rate than those released when first eligible.

Citizens Alliance on Prisons and Public Spending

57

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
1. Parolees are a very small proportion of the total population. In 2008, the Michigan population
was 10,003,422. With 12.7% under age 18, the state’s adult population was 8,732,987. The
average number of parolees under supervision that year was 17,435 or 0.2% of the total adult
population. Thus, even though they commit more crimes, on average, than citizens in general,
parolees constitute too small a portion of the total population to be responsible for more than a
small fraction of total crime.
a. The Bureau of Justice Statistics (BJS) examined all the people released from prison
in 1994 from 13 states, including Michigan. It found that they accounted for 4.7% of
arrests for serious crime (including misdemeanors) from 1994-1997.76
b.

Rosenfeld, Wallman and Fornango looked at the BJS data by state and examined the
Michigan prisoners released in 1994 who were rearrested during a one-year period
from 1994-1995. They found that these former prisoners accounted for 2.5% of all
adult arrests for violent crimes, 2.7% of adult arrests for property crimes and 1.1%
of adult arrests for drug crimes.77

c. The Council of State Governments conducted its own analysis of more recent
Michigan data. It found that parolees accounted for 3.0% of arrests in 2007 for
index crimes, which include murder and non-negligent manslaughter, rape, robbery,
aggravated assault, burglary, larceny, motor vehicle theft and arson, but not drug or
weapons offenses or driving under the influence.78
While these data vary as to the crimes included and whether they count “max outs” as well as
parolees, it is reasonable to infer that parolees account for no more than 4% of annual felony
arrests in Michigan.
2. The data examined in this report show that people whose parole was delayed did not return to
prison for new crimes at a much different rate than people released when first eligible. People
on their ERD had a new sentence return rate of 15.7% while those released one to four years
after their ERD had a rate of 20.9%. Although this difference resulted from convictions not
arrests, it is reasonable to assume that the difference in arrests for the two groups would be
similar.
When these two facts are taken into account, it is possible to roughly estimate that the increase in
arrests from releasing on their ERD the people who are held for an additional year or two would be
less than 0.4%. This estimate was derived as follows.
In 2006, there were 54,183 arrests for index crimes. If parolees accounted for 4%, that was 2,167
arrests.79
In 2006, there were 332,521 total arrests, which included nearly 12,000 for disorderly conduct,
nearly 22,000 for liquor law violations, more than 47,000 for driving under the influence and nearly
76Langan
77Richard
78CSG

and Levin, note 32 supra, at p 6.
Rosenfeld, Joel Wallman and Robert Fonango, note 72 supra, at p 95

report, note 9 supra, at p 5.

79Michigan

58

State Police, Criminal Justice Information Center, Crime in Michigan: 2006 Uniform Crime Report, at p 13.

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
107,000 for “all other,” which includes public drunkenness and vagrancy. If parolees accounted for
4% of the total arrests, that was 13,301.
The average number of people on parole in 2006 was 16,018. This includes people who were
released both on and after their first eligibility dates but not people who discharged on their
maximum sentences and therefore were not subject to supervision.80
The average number of arrests per parolee for index crimes was: 2,167 ÷ 16,018 = .135.
The average number of total arrests per parolee was: 13,301 ÷ 16018 = .830.
The total number of people released one or two years after ERD during the 14-year study period
was 21,881 or 1,563 per year.
If 1,563 additional people had been on parole in 2006, the number of additional arrests for index
crimes would have been: 1,563 x .135 = 211. The percentage increase in arrests for index crimes
would have been: 211 ÷ 54,183 + 211) = 0.39%.
If 1,563 additional people had been on parole in 2006, the number of additional total arrests would
have been: 1,563 x .830 = 1,297. The percentage increase in total arrests would have been:
1,297 ÷ 332,521 + 1,297 = 0.39%.
While this exercise produces only rough estimates, it strongly suggests that, standing alone,
releasing more people on their first parole eligibility dates would have virtually no impact on
overall crime. It appears that the increase would be so subtle it would be effectively undetectable
on a trendline chart showing the number of yearly arrests for index crimes, the number of total
arrests or their associated rates.
D. One and two-year parole denials: the large cost of small gains
To individual victims and their loved ones, any new crime against a person by a parolee is
unquestionably one too many, especially if it is an inexplicable tragedy like the murder of a total
stranger, the rape of a child or an assault causing permanent injuries. However, the total
prevention of new crimes by parolees would require the ongoing incarceration of tens of thousands
of people who would not, in fact, commit a new offense if released. Not only would the expense be
astronomical, all risk of crimes by former prisoners would not be eliminated because most people
would eventually max out. Thus, in determining what level of risk is tolerable to society as a whole,
policymakers must assess the relative costs and benefits of continuing to incarcerate people who
are eligible for parole.
The data support a number of conclusions that are relevant to that assessment:
•
•

The risk of new crimes against persons generally and of repeat crimes by homicide, assault,
robbery and sex offenders is less than 5%, but the risk of new property or public order
offenses by former prisoners is 13%.
Reoffense risk is not related to the sheer length of time served and is not decreased by
denying parole for anywhere from one to four years.

80Michigan

Department of Corrections, 2006 Statistical Report, Table E5 at
http://mich.gov/documents/corrections/MDOC_2006_Statistical_Report_255590_7.pdf

Citizens Alliance on Prisons and Public Spending

59

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
•
•

Thousands of people are denied parole each year who would not have returned to prison
for a new offense.
The extent to which repeated parole denials prevent the release of people who would
reoffend varies greatly by offense type.

The data also permit an informed estimate of what the increased risks and the cost savings would
have been if, during the 14-year study period, everyone released one or two years after their ERD
had been released when first eligible.
Table 16 is derived from the raw numbers reported in Table 14. It assumes, based on the preceding
findings, that serving one or two additional years had no impact on success or failure, i.e., that
people would have had the same outcome if released on their ERD as they had when they were
actually released one or two years after their ERD. For instance, drug offenders released on their
ERD had a success rate of 71.6%. Table 16 assumes that the drug offenders denied parole for one
year would have had a 63.2% success rate if they had been released when first eligible and that
those denied parole for two years still would have had a 59.6% success rate.
The people in each offense group who were released on their ERD established the base rate for each
outcome. All the people who were released one and two years after their ERD were then added to
the base and the outcome rates were calculated for the new totals. For example, 12,180 drug
offenders were paroled on their ERD and 8,722 (71.6%) of them were successful. Of the 2,228 who
were released one year after their ERD, 1,407 (63.2%) were successful. Of the 1,288 who were
released two years after their ERD, 768 (59.6%) were successful. When the groups are combined,
there were 15,696 drug offenders released on or within two years of their ERD and 10,897 (69.4%)
were successful. Thus, if all drug offenders denied parole for an additional one or two years had
been released when first eligible, the success rate for drug offenders released on their ERD would
have been 2.2 points lower. Using the same method of calculation, the rate of returns with new
sentences and the rate of technical violations of drug offenders released on their ERD would each
have been just 1.1 points higher.
If released without having to wait an additional year or two, the people who would have succeeded
in any event would have succeeded sooner; those who were going to fail would have failed sooner.
But the differing success and failure rates of the smaller groups who were denied for a year or two
would have had a minimal impact when combined with the rates of the larger group that was
paroled on their ERD. Thus, if an additional 3,516 drug offenders had been released when they first
became eligible, there would have been only 412 fewer successes, 197 more technical violators and
215 more returns with new sentences attributable to the changed release dates. Since 1,288 people
served two additional years, the total number of beds saved would have been 4,804.
Table 16 shows, for each offense group, what the impact would have been if everyone released
either one or two years after their ERD had been released when they first became eligible. The
table shows the extent to which success rates would have declined, return rates would have
increased, additional people would have been released and prison beds would have been saved. It
then shows how much money would have been saved for the entire 14-year period. Dollar savings
are based on $31,729 per bed per year, the 2008 combined average cost of a minimum and low
medium (Level 1 and 2) security level bed. The vast majority of parolees are released from a facility
at one of these levels.

60

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Table 16. 14-Year Impact of Releasing On ERD People
Who Were Released 1 or 2 Years After

Homicide
Assault
Sex*
Robbery
Burglary
Larceny
Drugs
Weapons
Mtr Veh
Grand
Total

Base
Success
On ERD
86.1%
72.3%
81.6%
64.5%
55.0%
57.2%
71.6%
64.9%
72.0%

Decrease
in
Success
4.2%
2.3%
3.1%
5.7%
3.2%
2.7%
2.2%
2.3%
1.1%

Base
PVNS
Rate
4.3%
10.8%
5.8%
15.1%
22.3%
22.0%
11.9%
15.6%
15.2%

Increase
in PVNS
1.7%
2.1%
1.6%
2.8%
1.8%
2.4%
1.1%
1.8%
0.6%

Base
PVT
Rate
9.6%
16.9%
12.6%
20.4%
22.7%
20.8%
16.5%
19.4%
12.8%

Increase
in PVT
2.5%
0.1%
1.6%
2.9%
1.4%
0.3%
1.1%
0.6%
0.5%

Add’l
People
Released
640
2,070
2,754
2,976
3,549
4,989
3,516
754
633

Beds
Saved
999
3,230
4,588
4,620
5,203
7,162
4,804
1,094
869

31,697,271
102,484,670
145,572,652
146,587,980
165,085,987
227,243,098
152,426,116
34,711,526
27,572,501

66.1%

2.9%

15.7%

1.7%

18.2%

1.2%

21,881

32,569

1,033,381,801

Dollars Saved

* If 1,242 sex offenders released 3 and 4 years past their ERD are included, decrease in success =3.9%, increase in
PVNS = 1.8 %, increase in PVT = 2.1%.

One can utilize these figures in various combinations, such as limiting the offense groups to those
with the highest success rates or including people whose outcomes when released two to four years
after their ERD were stable or actually improved. However, if everyone who was denied parole for
up to two years had been released at their first eligibility date, for the entire 14-year period, 21,881
additional people would have been released on their ERD. That is, instead of 45,001, the total
released when first eligible would have been 66,882. The overall success rate would have been
decreased by only 2.9%. For seven offense groups, the difference would have ranged from 1.1% to
3.2%. Only the robbery group would have seen success decrease by as much as 5.7%.
The overall rate of returns with sentences for new crimes would have increased only 1.7%. For six
of nine offense groups, the rate increase would have been less than 2.0%. The robbery group would
have had the greatest increase 2.8%.
Overall returns for technical parole violations would also have increased very little at 1.2%. For
seven offense groups, the PVT rate increase would have been less than 1.5%. Once again, robbery
offenders would have had the greatest increase at 2.9%.
With 11,193 people who were denied parole for one year and 10,688 who were denied parole for
two years, the total number of beds saved by releasing everyone on their ERD would have been
32,569. In today’s dollars, the savings over the entire 14-year period would have been
$1,033,381,801. The average savings per year would have been $73,812,985 for 2,326 beds, the
equivalent of two prisons.
While these figures are necessarily estimates, they illustrate the cumulative effect of past parole
policies. They also suggest the possible scope of future savings if those policies are changed to
reflect the actual risk of releasing people who committed various types of crimes. 81

Since the cost per prisoner increases each year, the actual savings, unadjusted for inflation, would have been less. This
also means, however, that the annual future costs of not changing these policies would be even more.

81

Citizens Alliance on Prisons and Public Spending

61

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?

VII. Implications of the research
The research findings have numerous implications for public policy. Overall, the lack of a
relationship between reoffense rates and the sheer length of time served suggests that relying on
lengthy incarceration is not a cost-effective crime control strategy. Michigan’s longer average
length of stay stems in part from the fact that it incarcerates fewer drug and non-violent offenders.
Thus, its average length of stay is not drawn down by a large number of less serious offenders who
may be serving relatively shorter sentences. However, Michigan prisoners convicted of assaultive
and sex offenses have high rates of commitment to prison and serve a relatively long time. The data
suggest that Michigan can safely reduce its incarceration rate to the levels of similar states while
continuing to adequately punish the most serious offenders.
Sentencing. Length of stay is a function of both the minimum sentence imposed by the court and the
parole board’s decision as to whether and when to grant parole. Sentencing guidelines used by
judges should reflect appropriate punishment but should not assume that exceedingly long
sentences are necessary to prevent reoffending. Effective alternatives to incarceration should be
pursued and the guidelines for assaultive, sex and habitual offenders should be reevaluated. In
addition, allowing prisoners to earn sentence reductions for good behavior (as most were able to do
throughout the study period) or program participation would not reduce public safety and might
increase it by encouraging self-discipline and achievement.
Parole decision making. Denying parole based on the seriousness of past harm rather than the
actual likelihood of future offenses created an anomaly. The very people who were least likely to
reoffend were repeatedly denied parole after serving their minimum sentences or even required to
serve their maximum sentences. As a result, thousands of people who were at very low risk of
committing a new crime against a person were forced to serve additional years in prison at great
expense to taxpayers.
The data suggest that substantially increasing the rate of parole on the earliest release date would
reduce the prisoner population without threatening public safety. In particular, eliminating the
frequent delay of parole for one or two years would save significant amounts of money without
releasing anyone “early” or causing more than a minimal change in success rates. This could be
achieved by establishing a statutory presumption of parole on the minimum for all prisoners,
subject to individualized risk assessments. Current parole guidelines that account for institutional
misconduct, prior record and current age are adequate to identify an individual’s statistical risk for
committing a new assaultive offense. However, they should be adjusted to accurately account for
the nature of the offense. That is, factors routinely associated with assaultive offenses (such as
physical injury or the use of a weapon) and the sheer fact that the offense involved sexual behavior
should not be scored as increasing risk.
Parole conditions and reentry services. The majority of former prisoners did not return to prison
within four years, even when parole grant rates were higher, parole conditions were less strict,
highly structured re-entry planning did not exist and pre-parole community placements were
common. While recidivism rates are unacceptably high, and many undoubtedly struggle with
poverty, substance abuse and mental illness, the fact remains that most people released from
prison manage to avoid committing new felonies. This suggests that resources devoted to
supervision and reentry support will be most cost-effective if they are carefully targeted at the
people who are at highest risk for reoffending. In particular, the much higher reoffense rates for
financially-motivated crimes suggest that employment-related services would be most productively
aimed at people convicted of larceny, robbery and burglary. It also suggests that the reintroduction
62

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
of pre-parole community placements, in conjunction with current reentry programming, would not
increase failure rates.
At the other extreme, intensive supervision of people who are at very low risk does not seem useful.
The application of stringent supervision conditions to whole categories of parolees and of
employment and residence barriers to whole categories of people with criminal convictions
complicates their ability to function normally in the community while doing very little to enhance
public safety. The selective use of parole conditions and of work and residence related prohibitions
for former offenders generally, based on actual risk, would be fairer and more cost-effective.
Need for further research. More research is needed to understand what distinguishes those who
will not reoffend in any event from those whose reoffense risk can be reduced through
programming and those who are least likely to benefit from even the best-targeted efforts. Can
some people simply be allowed to finish their punishment and then be left alone? Are some people
inevitably going to reoffend no matter what? Does the strength of particular characteristics
distinguish people who return to prison repeatedly? Can we separate people likely to reoffend
from those likely to commit technical violations?
It is important to measure not only the existence of relationships but the extent to which specific
factors affect reoffense rates and how they interact with each other. How much does the nature of
the offense explain differences in reoffense rates? How much of the variance is explained by each of
numerous other variables? How much does the importance of some factors depend on the
existence of others? How much does the relationship of factors to reoffending change over time?
How many crimes against people are responses to circumstances that are impossible to predict
and/or unlikely to reoccur, making prevention efforts futile either way?
Unfortunately, limitations in the database did not permit the current research to examine how such
variables as employment history, academic and vocational education, family support, medical and
mental health problems, and substance abuse affected either release decision making or success
upon release. Because they are all susceptible to being changed, such dynamic factors have
enormous implications for crime prevention, sentencing, prison programming, parole decisions,
parole supervision and re-entry services.
Also worth exploring is the prison experience itself, for example, whether people spent their time
locked down in segregation or working on crews in the community, the nature and quality of the
programs they completed, and how much they received incentives for positive behavior as opposed
to punishment for negative behavior. The impact on reoffense rates of criminal justice practices
should also be objectively assessed. These should include not only alternatives to incarceration
(such as diversion programs and “problem-solving” courts), but the nature, intensity and length of
probation and parole supervision, sex offender registries, and the automatic application of
licensing, employment and residence restrictions to people with criminal convictions.

Citizens Alliance on Prisons and Public Spending

63

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?

VIII. Conclusion
There is no dispute that discretionary decisions by the parole board about when people should be
released have a substantial impact on the size of Michigan’s prison population. As long as the state
retains a system of indeterminate sentencing, it must be assumed that some number of people will
be denied parole when they first become eligible. However, where the line should be drawn
between paroling everyone when they have served their minimum and requiring everyone to serve
their maximum is a matter of great debate. Not surprisingly, that debate often rests on assumptions
about parolee crime that are widely held but thinly supported by the facts.
Inaccurate assumptions about the impact of longer prison stays on reoffense rates generally, and
about the future behavior of people who committed assaultive and sex offenses in particular, have
led us to routinely continue the incarceration of thousands of parole-eligible prisoners who would
not have returned to prison in any event. The cost to those prisoners, their families and taxpayers
is enormous.
Understandably, policymakers and parole board members do not want to risk the commission of a
serious crime by someone who could have been kept in prison. But public officials cannot prevent
all such risk, they can only manage it as effectively as possible. Effective risk management depends
on measuring the risk accurately, weighing the various costs of reducing it, and adopting proven
risk reduction strategies. Above all, it depends on realistically identifying what level of risk is
tolerable or perhaps even inevitable.
Reducing Michigan’s prisoner population and effectively managing the risk of serious new crime by
parolees are not contradictory goals. They both require only a willingness to set reasonable
expectations, to abandon assumptions that are not consistent with the evidence and to incarcerate
only those parole-eligible prisoners who are objectively determined to pose a current threat to
public safety.

64

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?

Appendix A: Offense Categories
Drugs
All these offenses involve the possession or distribution of controlled substances. All sentences
were imposed under a mandatory scheme that was in effect until 2003. By far the most common
offense in the group at 63% is delivery under 50 grams, which carried a 20 year maximum. Nearly
12% are possession of less than 25 grams, with a 4-year maximum. Possession of 25-50 grams and
delivery of 50-225 grams each account for another 4% of this group.

Larceny
This group includes a wide variety of property offenses, such as embezzlement, welfare fraud,
misuse of credit cards and malicious destruction of property. The most common are uttering and
publishing (12.4%) which has a 14-year maximum, larceny from a person (8.2%) which has a 10
year maximum, receiving and concealing stolen property (17.4%), unlawfully driving away an
automobile (9.4%), larceny from a motor vehicle (6.4%) and larceny over $100 (6.5%), all of which
have 5-year maximums, larceny in a building (11.7%) with a 4-year maximum and retail fraud
(7.9%) with a 2-year maximum.

Burglary
Nearly all the offenses in this category involve breaking and entering either a dwelling (11.8%), or a
business or other non-residential structure (79%). Most were committed before 1994 when the
statute was redrafted to distinguish between first and second-degree home invasion. Fewer than
1% are convictions for possession of burglar tools. The vast majority carry a 10 year maximum
sentence. Four percent are for entering without breaking, which carries a 5-year maximum.

Robbery
More than 70% of the cases in this category are armed robbery, assault with intent to commit
armed robbery or carjacking. Nearly 25% are unarmed robbery or assault with intent to commit
unarmed robbery. The remainder involve bank robbery or safe breaking.

Sex
Under Michigan law, there are four degrees of criminal sexual conduct (CSC) as well as a number of
much less common offenses involving sexual behavior. CSC1, which carries up to life in prison, is
sexual penetration with one of various aggravating circumstances. CSC2, with a 15-year maximum
is sexual contact with an aggravating circumstance. CSC3, also with a 15-year maximum, is sexual
penetration without additional circumstances and CSC4, with a 2-year maximum, is contact without
additional circumstances. Within the research group, the distribution of CSC offenses is: CSC1 –
20.3%, CSC2 – 32.9%, CSC3 – 29.9%, CSC4 – 4.5%. Another 10% involve assault with intent to
commit some degree of CSC.

Citizens Alliance on Prisons and Public Spending

65

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?

Assault
While this category includes a broad range of offenses, just three account for about 84% of the total.
They are: felonious assault, which has a 4-year maximum (39.2%); assault with intent to commit
great bodily harm, which has 10-year maximum (35.8%); and assault with intent to murder (8.8%),
which carries life or any term of years. Offenses involving fleeing, eluding or resisting police
account for another 4.7%. Child abuse (3.5%), kidnapping (2.1%) and stalking (1.3%) are also in
this group.

Motor vehicle
More than 94% of the cases in this group involve some variation of driving under the influence of
drugs or alcohol, with 77% being OUIL, 3rd offense.

Homicide
The majority of cases in this group, 62.1%, involve convictions for some form of manslaughter and
carry a 15 year maximum. Second-degree murder, for which judges can impose life or any term,
comprise 30.2% of the group. The remaining 7.6% are offenses like negligent homicide and
careless use of a firearm resulting in death, for which the maximum penalty is two years.

Weapons
Nearly 72% of the cases involve convictions for carrying a concealed weapon, which has a five-year
maximum. Most of the rest involve possession of a weapon that is illegal (such as a short-barreled
shotgun) or possession of a weapon under illegal circumstances (such as, by a felon or with the
intent to commit harm). These cases also carry a five-year maximum term.

66

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?

Appendix B: Methodology
Defining recidivism
The ideal definition would be the commission of any new offense, or at least any new felony.
However, this is impossible to measure. So many crimes are undetected, unreported or unsolved
and so many suspicions are unproven that no researcher can ever know what any group of people,
ex-offenders or otherwise, actually has or has not done.
Some researchers use arrest rates to measure whether former prisoners have committed new
crimes. This is also problematic for several reasons. There are far more arrests than convictions, in
part because the evidence is often insufficient to prove guilt. Thus the sheer fact that someone was
arrested does not mean they actually committed a new crime. In addition, felony arrests often
result in convictions for misdemeanors and thus do not accurately represent the amount of serious
crime. Finally, ex-offenders, especially parolees, are more susceptible to being investigated because
they are known to law enforcement and thus may experience more arrests than their actual
behavior warrants. Overall, using arrest rates risks overestimating the amount of new crime
former prisoners have committed. And, as a practical matter, arrest information was not available
for this database. 82
New convictions, identified as either misdemeanors or felonies, would be a more reliable measure
of reoffending than arrests. Unfortunately, this data was also not available if the new conviction did
not result in a new prison sentence.83
Returns to prison have the advantage of accounting for technical violations as well as those that
resulted in new sentences. The data is consistently reliable and requires no speculation or
extrapolation. It is a measure commonly used by corrections department analysts and other
researchers.

Measuring recidivism
This analysis focuses on the number of people who return to prison, not the number of paroles that
are violated or the total number of returns that occur in a given time period. A single individual
may return to prison multiple times. For instance, a person may be paroled on X offense, have that
parole revoked for a technical violation and then be paroled again to either succeed or fail.
Alternatively, a person may be paroled on X offense and then, regardless of the outcome of that
parole, be subsequently convicted of Y offense, imprisoned and paroled, starting the cycle again.
For example, in this data set, 76,721 people were committed at least once after April 1, 1981 and
released from those sentences for the first time during the 14-year study period. However, of that
total, 16,681 (21.7%) were committed and released more than once. Roughly 18% (13,872 people)
For further discussion, see Flaherty, note 37 supra, Appendix A: Methodology (concluding that arrest may be “an
indiscriminate indicator of ex-offender failure”) and Holley and Ensley, note 37 supra, at pp 8-9 (noting that arrest rates
may vary with changes in policing and expressing “doubt that rearrest is a reliable indicator of reoffending by state prison
inmates”).
82

83 It is not uncommon for a former prisoner, even if still on parole, to receive county jail time or some alternative
community-based sanction for a low-level felony conviction. If the new conviction was for a misdemeanor, a prison term
is not even an option.

Citizens Alliance on Prisons and Public Spending

67

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
were committed and released twice; about 3% (2,452 people) were committed and released three
times. Another 357 people were sentenced to prison and released four or more times during the
period. Thus, 76,721 individuals actually had 96,596 prison commitments for which they were
released during the period.
By focusing only on the first commitment and release during the study period, this analysis avoids
counting people with multiple paroles multiple times, i.e., it does not count the same person more
than once. Counting people prevents overall success rates from being diluted by the multiple
failures of individuals. It allows for an accurate focus on the proportion of released prisoners who
fail at all, whether once or repeatedly within a specific time frame, and for an optimal allocation of
resources to reduce recidivism.

68

Citizens Alliance on Prisons and Public Spending

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?

Appendix C: Michigan and National Prison Commitment Rates
Commitment Rates by Offense Category*
(in percentages)
Prison
40.0
54.0
36.4
37.0

All
Assaultive
Non-assaultive
Drugs

National 2004
Jail
30.0
24.0
32.2
30.0

Total
70.0
78.0
68.6
67.0

Prison
22.8
41.0
20.4
12.1

Michigan 2007
Jail
47.9
39.4
50.2
50.8

Total
70.7
80.4
70.6
62.9

*Cases counted by offender.

Comparing state and national commitment rates can be difficult because crimes are not
counted in the same way. Michigan counts fleeing and eluding, resisting and obstructing, and
malicious destruction of police and fire department property, as assaultive offenses. See
Michigan Department of Corrections, 2007 Statistical Report. Federal statistics combine
property, weapons and other nonviolent offenses into the category “non-assaultive.” See
Bureau of Justice Statistics Bulletin: Felony Sentences in State Courts, 2004. The following
table compares Michigan and national commitment rates for comparable crimes.

Prison Commitment Rates
Selected Assaultive and Sex Offenses
National – 2004*

Michigan – 2007**

Murder/non-negligent
manslaughter

89%

Rape

69%

Other sexual assault

57%

1st-degree murder
2nd-degree murder
Manslaughter
TOTAL

Robbery

72%

Aggravated assault

43%

*Cases counted by offender
(Most recent data available)

1st-degree criminal sexual conduct
2nd and 3rd-degree criminal sexual
conduct
4th-degree criminal sexual conduct
Assault with intent: criminal sexual
conduct
TOTAL
Armed robbery
Carjacking
Assault with intent: armed robbery
Unarmed robbery
TOTAL
Assault with intent: murder
Assault with intent: great bodily harm
less than murder
Felonious assault
TOTAL

97.5%
100.0%
81.1%
94.8%
98.0%
69.4%
22.2%
52.4%
65.2%
97.0%
94.6%
92.5%
54.2%
80.7%
98.2%
61.0%
26.0%
38.6%

**Cases counted by disposition

Citizens Alliance on Prisons and Public Spending

69

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?

Appendix D: Additional Tables
Table A. Effect of Release Year Change on Four Offense Groups*
Total No.
Old
New
Board
Board
Robbery

4,401

4,609

Assault

2,581

3,948

Homicide

1,171

1,387

Sex

3,005

3,670

Total

11,158

13,614

Old bd.
New bd.
Change
Effect
Old bd.
New bd.
Change
Effect
Old bd.
New bd.
Change
Effect
Old bd.
New bd.
Change
Effect
Old bd.
New bd.
Change
Effect

ON
ERD

1 year
past

56.1%
51.3%
-4.8%
-221
64.2%
52.6%
-11.6%
-458
69.8%
53.8%
-16.0%
-222
47.4%
20.5%
-26.6%
-976
57.1%
43.8%
-13.3%
-1,811

17.1%
12.5%
-4.6%
-212
14.8%
13.5%
-1.3%
-51
11.8%
10.3%
-1.5%
-21
20.5%
8.3%
-12.2%
-448
16.9%
11.2%
-5.5 %
-749

ON – 1
year
total
73.2%
63.8%
-9.4%
-433
79.0%
66.1%
-12.9%
-509
81.6%
64.1%
-17.5%
-243
67.9%
28.8%
-38.8%
-1,424
74.0%
55.2%
-18.8%
-2,560

2 yrs
past
17.6%
18.9%
+1.3%
+60
15.4%
19.4%
+4.1%
+162
10.9%
16.8%
+5.9%
+82
20.9%
33.1%
+12.1%
+440
17.2%
22.6%
+5.4%
+735

2-4
yrs
past
7.5%
12.8%
+5.3%
+244
4.9%
11.4%
+6.5%
+257
6.1%
12.6%
+6.5%
+90
9.3%
26.6%
+17.2%
+631
7.2%
16.1%
+8.9%
+1,212

>4 yrs
past
1.7%
4.5%
+2.8%
+129
0.8%
3.1%
+2.2%
+87
1.4%
6.4%
+5.0%
+69
1.9%
11.4%
+9.5%
+349
1.5%
6.1%
+4.6%
+626

Max
Out
1.2%
2.3%
+1.1%
+51
7.3%
15.6%
+8.3%
+328
5.4%
4.8%
-0.6%
-8
7.2%
20.0%
+12.8%
+470
4.6%
11.2%
+6.6%
+899

*For each offense group, the effect of changed release practices was calculated by multiplying the number of new
board cases by the percentage change.

Table B. Success & Failure by Mean Age
(in years)

Motor Vehicle
Sex
Homicide
Assault
Drugs
Larceny
Robbery
Weapons
Burglary
TOTAL

70

Total

Success

PVT

PVNS

37.5
36.1
35.5
31.1
30.5
29.9
28.9
28.3
27.9
30.8

38.1
37.4
36.1
31.9
31.2
30.8
29.0
29.2
27.9
31.7

35.9
31.7
33.8
29.5
29.4
29.5
29.1
27.4
28.5
29.6

36.1
31.7
31.5
28.4
28.1
28.5
28.3
26.4
27.3
28.5

Citizens Alliance on Prisons and Public Spending

Difference:
Success
& PVNS
2.0
5.7
4.6
3.5
3.1
2.3
0.7
2.8
0.6
3.2

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Table C. Success and Failure by Prefix
Number

Success

PVT
(Percentages)

PVNS

Homicide
A
B
C or higher

2,189 (85.5%)
272 (10.6%)
98 (3.8%)

83.2
63.2
56.1

11.4
21.3
24.5

5.4
15.4
19.4

Sex
A
B
C or higher

6,065 (90.7%)
413 (6.2%)
206 (3.1%)

78.4
69.7
71.8

14.5
17.7
14.1

7.1
12.6
14.1

Motor Vehicle
A
B
C or higher

2,837 (88.7%)
271 (8.5%)
91 (2.8%)

71.7
61.3
70.3

12.8
18.1
13.2

15.5
20.7
16.5

Assault
A
B
C or higher

5,599 (85.7%)
653 (10.0%)
280 (4.3%)

71.3
62.5
61.1

15.9
19.8
20.7

12.8
17.8
18.2

Drugs
A
B
C or higher

14,147 (87.6%)
1,439 ( 8.9%)
568 (3.5%)

70.2
61.0
60.0

17.0
22.5
23.1

12.8
16.5
16.9

Weapons
A
B
C or higher

2,081 (82.4%)
279 (11.0%)
165 (6.5%)

63.1
58.4
57.0

18.7
22.6
22.4

18.1
19.0
20.6

Robbery
A
B
C or higher

7,073 (78.4%)
1,195 (13.3%)
749 (8.3%)

60.0
48.5
43.3

23.1
28.7
29.1

16.9
22.8
27.6

Larceny
A
B
C or higher

12,175 (75.1%)
2,663 (16.4%)
1,372 (8.5%)

57.1
46.6
46.6

19.3
25.6
23.9

23.6
27.8
29.5

Burglary
A
B
C or higher

7,469 (70.4%)
2,065 (19.5%)
1,080 (10.2%)

54.1
44.4
43.4

23.5
27.5
26.9

22.5
28.1
29.7

Total
A
B
C or higher

59, 635 (81.1%)
9,250 (12.6%)
4,609 (6.3%)

65.5
52.0
50.0

18.3
24.7
24.5

16.2
23.2
25.5

Citizens Alliance on Prisons and Public Spending

71

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Table D. Failure Rates by Nonbondable Misconducts
(in percentages)

Homicide
Sex
Assault
Drugs
Motor Vehicle
Weapons
Robbery
Burglary
Larceny
TOTAL

RETURNS WITH NEW SENTENCES

RETURNS FOR TECHNICAL VIOLATIONS

Number of Misconducts

Number of Misconducts

None
6.4
6.8
12.0
12.2
15.4
17.2
17.8
23.0
23.7
16.3

One
10.8*
12.6
17.3
17.1
19.2*
19.4
20.4
26.5
26.6
21.4

Two
19.2*
24.4*
27.0
21.4
23.7*
32.6*
24.0
34.1
32.7
28.6

None
12.0
14.3
16.8
16.1
12.7
19.1
22.5
23.8
20.3
18.4

*Fewer than 70 cases

72

Citizens Alliance on Prisons and Public Spending

One
20.8*
18.4
15.3
24.6
16.7*
21.5
30.6
26.9
23.0
23.9

Two
19.2*
18.5*
16.2*
23.7
21.8*
16.3*
30.8
26.6
20.8
22.9

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?

ACKNOWLEDGMENTS
This research was initially undertaken with a generous grant from the JEHT Foundation. The
quantity of the data and overall complexity of the subject matter made the project more
complicated and time-consuming than anticipated. Completion and publication of the report
ultimately depended on the support of many.
Barbara R. Levine, executive director of CAPPS, authored the report. She is solely responsible for
the interpretation and presentation of the data. However, reviews of earlier drafts by the Citizens
Research Council, the Michigan Department of Corrections, Kay Perry, William Tregea, Gail Light
and Tom Burkert greatly improved the final product. All provided comments that made the
analysis more accurate and the report more readable.
Jeffrey Anderson of the MDOC’s Office of Research and Planning restructured the database. (Mr.
Anderson’s outside employment with CAPPS was limited to that specific task.) Terrence H. Murphy,
MDOC’s Chief of Research and Evaluation from 1986 until his retirement in 2002, conducted the
data analysis. He performed the initial analysis according to specifications provided by CAPPS and
additional analyses throughout the project as issues became more refined. Without their patience
and commitment to seeing the project through, this report could never have been completed.
Drew Schendel provided invaluable assistance with the literature review.
Gail Light, formerly with the MDOC Office of Public Information and currently CAPPS’ s
communications specialist, designed the report and produced many versions of the graphics.
Assistance with tables was also provided by LeAnn Voigtritter.
When foundation funding was exhausted before the work was done, generous special contributions
from the following individuals and organizations made it possible to finish the project and to print
and distribute the report.
Kirk Acrey
American Association of
University Professors
American Friends Service Committee,
Michigan Criminal Justice Program
Michael Barnhart
Tony Benavides
Blue Cross/Blue Shield of Michigan
Ronald Bretz
Gene and Karen Brewer
Robert Brown, Jr.
Cascade Engineering
Charlotte Cox
Criminal Defense Attorneys of Michigan
Criminal Law Section,
State Bar of Michigan
James Dankovich
Frank Eaman
Janet Findlater

John Freeman
Robert and Joanne Gielow
Grand Rapids Community Foundation –
Systemic Change Fund
Robert Grosvenor
Steve Harry
Robert and Phyllis Hastings
Susan Helton
Annette Hemmings
Rose Homa
Annette and Cal Johnson
Margaret Kanost
Dorothy Levine
Catherine Lloyd
Dov Lustig
Jeanette Marcinowski
Janet Max
Michigan Building and Construction
Trades Council

Citizens Alliance on Prisons and Public Spending

73

Denying Parole at First Eligibility: How Much Public Safety Does It Actually Buy?
Gov. William Milliken
National Lifer Association members at:
Lakeland Correctional Facility
Michigan Reformatory
Southern Michigan Correctional Facility
Thumb Correctional Facility
Stanely Nesen
Penny Pestle
Prisoner Benefit Funds at:
Chippewa Correctional Facility
Hiawatha Correctional Facility
Kinross Correctional Facility
Lakeland Correctional Facility
Macomb Correctional Facility
Mound Correctional Facility
Muskegon Correctional Facility

Southern Michigan Correctional Facility
Scott Correctional Facility
Pine River Correctional Facility
Saginaw Correctional Facility
Proaction Behavioral Health Alliance
Margaret Sind Raben
Carol Reinstra
Martin Reisig
Laura Sager
Kathleen Schaefer
Larry and Honorene Skover
Patricia Streeter
James Thomas
F. Martin Tieber
William Tregea

The continuing cooperation of the Michigan Department of Corrections is gratefully acknowledged.
The author gives special thanks to the members of the CAPPS board of directors for their
unwavering support in general and of this project in particular.

74

Citizens Alliance on Prisons and Public Spending