Parolee Success in Iowa:
Individual and Treatment Effects on
Recidivism

Presented to the Iowa Department of Corrections
Directors Meeting
on January 17, 2013 in Des Moines, Iowa.

David Peters, Ph.D.
Andrew Hochstetler, Ph.D.
1

OVERVIEW
What is the project about?
To understand the individual and community factors that drive
recidivism rates among parolees in Iowa.
• Demographics and risk propensities.
• Treatment provisions and quality.
• Distance to DOC and social services.
• Community socioeconomic contexts
Who funded it?

National Institute of Justice grant encouraging …
• Collaboration between criminal justice agency professionals and
university researchers.
• Application of social science research to address practical issues in
the criminal justice system.
2

OVERVIEW
Why does it matter?
Need to serve more
parolees with less state
funding.
Need to focus on areas
where you have control and
can make a difference.
Avoid spending resources on
areas where you have little
control.

3

PLAN OF WORK
Step 1 … Controlling for treatment selection.
• Compile data from DOC, Census, and other sources.
• Estimate propensity scores for statistical model.
Step 2 … Understanding fixed effects of recidivism.
• Predict probability of recidivism at 1-yr and 2-yr intervals.
• Model includes individual risk factors, treatment
provision/quality, distance from services, and
community socioeconomic conditions.
• Regional models – urban-rural areas and CCDs.
• Demographic models – men-women and race.
• Purpose is to understand what factors drive recidivism that are
largely beyond the control of CCD staff.
4

PLAN OF WORK
Step 3 … Understanding amenable effects of recidivism.
• Identify parolees that did better than the model predicted.
• Interviews will be conducted within each CCD to ascertain why
these parolees did not reoffend, when their risk profile suggested
they should have.
• Purpose is to understand the unique factors that cannot be
accounted for by the model, and which CCD staff may have some
control over.
Step 4 … Synthesis and dissemination.
• Based on the results, a series of “best
practices” will be identified and
communicated to corrections professionals.
5

PAROLEE SUCCESS IN IOWA

DEMOGRAPHICS

6

DEMOGRAPHICS
Defining parolees …

N=1,272 persons who …

• were paroled or special sentenced in 2010
• had valid physical address
• lived in Iowa and not institutionalized
Basic Demographics …

Most parolees are white middle-aged men who were born in Iowa.
•
•
•
•
•

Male – 85%
Average age – 38 years.
White, non-Hispanic race – 74% (97% in 4th-District)
Iowa born – 74% (62% in 4th-District)
Hispanic any race – 3% (6% in 3rd, 6th, 7th Districts)
• Only 1% needed interpreters.
7

DEMOGRAPHICS
Parole / Special Sentence Population …

•
•
•
•
•
•
•
•

1-District
2-District
3-District
4-District
5-District
6-District
7-District
8-District

=
=
=
=
=
=
=
=

17.0
10.9
7.7
5.3
30.7
6.5
11.4
10.5
8

PAROLEE SUCCESS IN IOWA

RISK ASSESSMENT

9

RISK ASSESSMENT
Most parolees are high risk who have multiple
previous convictions.
High LSI Risk – 38.91 (last score)
Growing LSI Risk – 1.96 rise (last v. first score)
Previous convictions – 7.7
•

Highest in 7th-District (10.1). Lowest 5.6 4th-District (5.6)

•

Highest in 6th-Dist (30.6 mos). Lowest in 4th-Dist (18.3 mos)

Sentence served – 24.5 months

Paroled from violent offense – 17.8%
•

Highest in 6th-Dist. (21.6%). Lowest in 4th,7th Dist. (13.1%)

•

Highest in 2nd,4th,6th Dist (7.3%). Lowest in 5th-Dist (3.6%)

Paroled sex offender – 5.0%

Drug and other non-property offenses – 49.9%.

10

RISK ASSESSMENT
LSI-R Scores (last) …

•
•
•
•
•
•
•
•

1-District
2-District
3-District
4-District
5-District
6-District
7-District
8-District

=
=
=
=
=
=
=
=

39.43
38.97
37.67
38.38
38.73
38.90
39.31
39.26
11

RISK ASSESSMENT
Change in LSI-R Scores (last v. first) …

•
•
•
•
•
•
•
•

1-District
2-District
3-District
4-District
5-District
6-District
7-District
8-District

=
=
=
=
=
=
=
=

2.45
2.11
1.42
1.97
1.75
2.03
2.65
1.19
12

PAROLEE SUCCESS IN IOWA

TREATMENT COMPLETION

13

TREATMENT
Most parolees are currently in treatment, but very few have actually
completed any treatment.
Treatment In-Progress or Completed (Needs 1-2) …
• 40.1% in no treatment
• 28.8% in 1 treatment
• 22.9% in 2 treatments
• 8.3% in 3 or more treatments
Treatment Completion (Needs 1-2) …
• 65.5% no completions
• 21.5% 1 completion
• 11.3% 2 completions
• 1.7% 3 or more completions

14

TREATMENT
What drives treatment completion?
Treatment more likely to be completed if …
•
•
•
•
•
•
•
•
•

Male parolee
3 or more offenses
Added second offense
Added charges while under supervision
Past school expulsion or suspension
Did not complete 12th grade
Unfavorable attitude towards conviction
Absence & increasing anti-criminal acquaintances
Increasing unsatisfactory relationship with relatives
Low Treatment Completion Bias = 26.4%
No race/ethnic bias in treatment completion.
15

TREATMENT
What drives treatment non-completion?
Treatment less likely to be completed if …
•
•
•
•
•
•
•

Convicted from Dubuque courts
Paroled from prison
Three of more address changes last year
Added adult conviction
Severe interference from active psychosis
Psych assessment needed (on-going)
Increasing drug problem
Low Treatment Completion Bias = 26.4%
No race/ethnic bias in treatment completion.

16

TREATMENT
Treatment Completion (Needs 1 & 2) …

•
•
•
•
•
•
•
•

1-District
2-District
3-District
4-District
5-District
6-District
7-District
8-District

=
=
=
=
=
=
=
=

43.72
28.78
23.71
32.35
31.03
34.94
36.55
42.11
17

TREATMENT
Under-Treatment and Over-Treatment …

•
•
•
•
•
•
•
•

1-District
2-District
3-District
4-District
5-District
6-District
7-District
8-District

=
=
=
=
=
=
=
=

1.65
-2.71
-8.86
0.92
-2.84
0.39
3.57
10.37
18

PAROLEE SUCCESS IN IOWA

EXPLAINING RECIDIVISM

19

RECIDIVISM
Recidivism at 2-Years (to prison or arrest) …

•
•
•
•
•
•
•
•

1-District
2-District
3-District
4-District
5-District
6-District
7-District
8-District

=
=
=
=
=
=
=
=

38.89
36.69
43.88
41.18
40.26
32.53
32.41
39.85
20

FINDINGS – RURAL PAROLEES
Factors increasing odds of recidivism …
• Not married (51%)
• Less serve conviction crime (15%)
•
•
•
•

Greater
Greater
Greater
Greater

& growing criminal history risk (31% & 16%)
housing/residence risk (41%)
alcohol/drug risk (15%)
& growing emotion/psychological risk (14% & 13%)

• Farther from community college & workforce offices (1% each).
Model explains 27.3% of 2-year recidivism.
52.4% correct classification as recidivist.

21

FINDINGS – RURAL PAROLEES
Factor decreasing odds of recidivism …
• Older age (-2%)
• More dependents (-19%)
•
•
•
•
•

Greater education/employment risk (-12%) – programs available
Greater non-participation/time-use risk (-26%) – more supervision
Greater pro-crime attitudes (-20%) – more supervision
Growing financial risk (-29%) – public assistance
Growing family/martial risk (-17%) – programs available

• Greater distance to DOC office (-1%)
Model explains 27.3% of 2-year recidivism.
52.4% correct classification as recidivist.

22

FINDINGS – URBAN PAROLEES
Factors increasing odds of recidivism …
• Male (189%)
• More dependents (9%)
• Greater criminal history risk (28%)
• Greater emotional/psychological risk (19%)
• Growing housing/residence risk (20%)
Model explains 21.3% of 2-year recidivism.
52.1% correct classification as recidivist.

23

FINDINGS – URBAN PAROLEES
Factors decreasing odds of recidivism …
• Older age (-4%)
• Drug conviction (-23%)
• Full-time employment (-29%)
•
•
•
•

Greater education/employment risk (-6%) – programs available
Greater financial risk (-17%) – public assistance
Greater housing/residence risk (-35%) – more supervision
Growing family/marital risk (-8%) – programs available

• Treatment completion (-47%) – treatment works
Model explains 21.3% of 2-year recidivism.
52.1% correct classification as recidivist.

24

FINDINGS – SURVIVAL FUNCTION
Survival Function ‐ Rural Parolees

For rural parolees …
Treatment has worsens survival.
Effects are small and non-significant.

1.0
0.9

Survival Function

0.8
0.7
0.6
0.5
0.4

6 mos

12 mos

18 mos

0.3
0.2
0.1

Rural ‐ No Treatment
Rural ‐ Treatment

0
29
61
78
113
123
144
157
182
213
238
272
299
336
351
384
409
443
526
563
631
667

0.0
Days to Recidivism

Survival Function ‐ Urban Parolees
1.0
0.9
0.7
0.6
0.5
0.4
0.3
0.2

12 mos

6 mos

18 mos

Urban ‐ No Treatment
Urban ‐ Treatment

0.1
0.0

0
33
66
88
106
124
144
161
188
203
230
256
282
326
357
384
424
472
540
573
616
707

For urban parolees …
Treatment has improves survival.
Effects are large and significant.

Survival Function

0.8

Days to Recidivism

25

FINDINGS – SURVIVAL FUNCTION
Effect of treatment on survival for rural parolees …
• 6 mos – 2.2% decrease in survival
• 12 mos – 3.8% decrease in survival
• 18 mos – 5.2% decrease in survival
• 24 mos – 5.7% decrease in survival
• Overall – 1.3% decrease in survival
Effect of treatment on survival for urban parolees …
• 6 mos – 4.7% increase in survival
• 12 mos – 8.7% increase in survival
• 18 mos – 11.0% increase in survival
• 24 mos – 12.8% increase in survival
• Overall – 6.9% increase in survival

26

SUMMARY
Treatment only works for urban parolees.
Employment matters only for urban parolees.
Distance to services important for rural parolees.
Dependents reduces rural recidivism, but increases urban rates.
Traditional risk factors increase recidivism for all.
Risks amenable to community treatment reduce recidivism for all.
Community factors only have small impact on offender recidivism.

27

NEXT STEPS AND COMMENTS
Next Steps …
• Refine regression models.
• Conduct interviews this February and March.
• Analysis and write-up this summer.

Comments and Questions
For more information
David Peters or Andy Hochstetler
515-294-1122
dpeters@iastate.edu
www.soc.iastate.edu
28

PLEASE GIVE US FEEDBACK!

•

s
uaijQ

(

l

,
JaAaN

S
uaYQ

- • -•
'" ii

JaAaN

~

IOWA STATE UNIVERSITY

0
c

.0>

Extension and Outreach

•

~

Session : Parolee Success in Iowa

0

~

,•

~

~

0

,

•uc

Presenter: Peters and Hochstetler

~

~

•

0

-

-•

Please circle appropriate response .
Write comments on back.

-

,

J:: 0

Knowledge of subject before session:

A k>t

None
1

N

!t

3
•

<is.

~

0

0

.

--

~

0
u

Date: January 17, 2013

"
,-

~

J::

~

w

,

l

:uO! leWJoJU! 5! 41 asn II!M 5JB410

:UO!leWJOJU ! S!41 asn IIiMI

:;-

•

(

2

3

4

5

-

]

Knowledge of subject aher session :

A k>t

None
1

2

3

4

5
29