JFA Associates/The Institute, Washington, DC/Austin, Texas Risk Assessment Facts, Myths and Trends...
-
Upload
jocelyn-stevens -
Category
Documents
-
view
219 -
download
0
Transcript of JFA Associates/The Institute, Washington, DC/Austin, Texas Risk Assessment Facts, Myths and Trends...
JFA Associates/The Institute, Washington, DC/Austin, Texas
Risk Assessment
Facts, Myths and Trends
James Austin, Ph.D.
2008
JFA Associates/The Institute, Washington, DC/Austin, Texas 2
The Need to Manage Risk
• Public Safety – Reduce Recidivism Rates and Allows for Better Investments
• Increases Credibility with the Public and Legislature
• Reduces the potential for disproportionate use of incarceration by Gender, Race and Ethnicity
• Better use of public resources
JFA Associates/The Institute, Washington, DC/Austin, Texas 3
The Key Decision Points
1. Initial Parole Eligibility Hearinga. Paroleb. Continue (why and for how long)c. “Serve All”
2. Rehearing (more of the initial hearing)
3. Mandatory Parole
4. Imposition of Supervision Conditions
5. Parole Revocationa. Detain or Releaseb. Period of Incarceration
JFA Associates/The Institute, Washington, DC/Austin, Texas 4
Sources
• Statistics and Trends -- Bureau of Justice Statistics, http://www.ojp.usdoj.gov/bjs/pubalp2.htm#Prisoners
• Research -- National Institute of Justice, http://www.ojp.usdoj.gov/nij/
• Technical Assistance – National Institute of Corrections
• Program Money and TA – Bureau of Justice Assistance, http://www.ojp.usdoj.gov/bja/
JFA Associates/The Institute, Washington, DC/Austin, Texas 5
Adult Correctional Populations 1980-2005
Population 1980 2005 % Change
Probation 1,118,097 4,162,536 272%
Jail 163,994 747,529 359%
Prison 329,821 1,446,269 339%
Parole 220,438 784,408 256%
Total Adults Under Corrections 1,832,350 7,056,000 285%
Adult Population 162.8 Million 214.8 million 32%
% of Adults Under Corrections 1.1% 3.3% 188%
JFA Associates/The Institute, Washington, DC/Austin, Texas 6
Past and Current Projected Prisoner Population
JFA Associates/The Institute, Washington, DC/Austin, Texas 7
Key Concepts in The Prediction of Risk
1. Variance
2. Probabilities
3. Independent Variables – The Predictors
4. Dependent Variable – What We Are Trying to Predict
5. Static Predictors – Things That Do Not Change
6. Dynamic Predictors – Things That Do Change
JFA Associates/The Institute, Washington, DC/Austin, Texas 8
Variance in the Use of Imprisonment
JFA Associates/The Institute, Washington, DC/Austin, Texas 9
Variance in the Use of Probation and Parole
JFA Associates/The Institute, Washington, DC/Austin, Texas 10
Variance in Life Chances of Being Imprisoned in USA1974-2001
1974 1991 2001
Total 1.9% 5.2% 6.6%
Males
White 3.6% 9.1% 11.3%
Black 13.4% 29.4% 32.2%
Hispanic 4.0% 16.3% 17.2%
Females 0.3% 1.1% 1.8%
White 0.2% 0.5% 0.9%
Black 1.1% 3.6% 5.6%
Hispanic 0.4% 1.5% 2.2%
JFA Associates/The Institute, Washington, DC/Austin, Texas 11
Variance in Key Criminal Justice Decision Points
Racial Group
Population Offender Arrested Convicted Prison/Jail
White 75% 64% 69% 54% 39%
Black 12% 25% 29% 44% 47%
JFA Associates/The Institute, Washington, DC/Austin, Texas 12
Variance in Average Sentences and Time ServedBy Race – 2002 Prison Releases
Offense Group
White Black
Sentence Time Served
Sentence Time Served
All Offenses 63 mos 27 mos 69 mos 32 mos
Violent 85 mos 45 mos 95 mos 53 mos
Property 56 mos 22 mos 58 mos 25 mos
Drug 61 mos 19 mos 64 mos 23 mos
Public Order 44 mos 18 mos 45 mos 21 mos
Source: Bureau of Justice Statistics, National Correctional Reporting Program 2002
JFA Associates/The Institute, Washington, DC/Austin, Texas 13
Indicator White Black
Married 57% 35%
Less than High School Degree 11% 21%
Unemployed 5% 11%
Under $25,000 Income 26% 39%
Below Poverty Level 8% 23%
Under 18 years 10% 30%
Central City Residence 21% 52%
Residence Ownership 83% 53%
Variance in Crime Risk Factors by Race
JFA Associates/The Institute, Washington, DC/Austin, Texas 14
7,000.0
CCrime and rime and WWelfare 1931-2003elfare 1931-2003
6,000.0
5,000.0
4,000.0
3,000.0
2,000.0
1,000.0
0.0
Cri
me R
ate
16,000.0
14,000.0
12,000.0
10,000.0
8,000.0
6,000.0
4,000.0
2,000.0
0.0
Welf
are
Recip
ien
ts
Year
JFA Associates/The Institute, Washington, DC/Austin, Texas 15
More Key Concepts on Risk
• Measurement Error Triangulation (Multiple Sources) Public Records Interviews Questionnaires Observation
• Reliability Inter-Reliability (Do we all do it the same way) Intra-Reliability (Do I do it the same way)
• Validity Internal Validity (does it work in my place) External Validy (does it also work in other places)
JFA Associates/The Institute, Washington, DC/Austin, Texas 16
Still More Concepts
• An Instrument Can Be Reliable but Not Valid
• An Instrument Cannot Be Unreliable and Valid
• False Positives (should have recidivated but did not)
• False Negatives (should have not recidivated but did)
• Multi-collinearity (independent predictors)
JFA Associates/The Institute, Washington, DC/Austin, Texas 17
And More Concepts
• Clinical Judgments Alone Are In-Effective in Predicting Risk
• Statistical Models Are Effective In Predicting General Recidivism
• Statistical Models Are In-Effective in Predicting “Rare Events”
1. Violent Crimes
2. Career Criminals
3. Sex Crimes
• The Problem of “Low” Base Rates
JFA Associates/The Institute, Washington, DC/Austin, Texas 18
Current State of Risk Assessment
• There is no superior or better risk assessment instrument or system
• There are commercial and public risk systems that can work
• The differences are in costs and staff skill requirements
• Few states have risk assessment systems that have been properly developed and implemented
• Lack of reliability and validity
JFA Associates/The Institute, Washington, DC/Austin, Texas 19
Some Basics About Criminal Behavior
• Criminal Behavior (Like Other Behavior) Is Learned
• Criminal Careers Have Starting And Ending Points
• Most Criminal Behavior is Episodic in Nature
• Very Few “Criminals” Are Career Criminals
• Places and Other People Impact Behavior
JFA Associates/The Institute, Washington, DC/Austin, Texas 20
What About Treatment?
1. Good treatment in prison is rare
2. The “market share” problem
3. Most one can expect is 10% reduction in the expected recidivism rate
4. Education and vocational training should be priorities
5. Wrong Use of Treatment Increases Risk
JFA Associates/The Institute, Washington, DC/Austin, Texas 21
Percent of Arrests Attributed to Released Prisoners
Type of arrests N %
Total Arrests in Seven States 1994-97 2,994,868 100%
Arrests of Prison Releases 1994-97 140,534 5%
Percent that are Violent Crimes 36,000 1%
JFA Associates/The Institute, Washington, DC/Austin, Texas 22
1983 and 1994 Recidivism Rates
Recidivism
Measures
1983 Prison Releases
1994 Prison Releases
Re-Arrested 63% 69%
Re-Convicted 47% 47%
Re-Imprisoned
41% 40% – 52%
JFA Associates/The Institute, Washington, DC/Austin, Texas 23
Method of Release and Re-Arrest
Re-Arrest
Rate
Unconditional
Releases
Mandatory
Releases
Discretionary Paroles
Unadjusted 62% 61% 54%
Adjusted 61% 61% 57%
JFA Associates/The Institute, Washington, DC/Austin, Texas 24
Success on Parole and Probation – 1995-2003
Probation Parole
Successful Completions
1995 62% 45%
2000 60% 43%
2003 59% 47%
Reason for Failures
Re-incarcerated
New Conviction/Sentence 5% 11%
Revocation 7% 26%
Other 4% 0%
Absconded 4% 9%
Other 22% 6%
JFA Associates/The Institute, Washington, DC/Austin, Texas 25
Length and Stay and Recidivism - Louisiana
Length of Stay in Prison Releases % Return to Prison
Return With New Charge
Up to 6 months 2,844 20.1% 36.1% 20.2%
6 to 12 months 2,724 19.3% 36.7% 17.9%
12 to 24 months 3,351 23.7% 39.0% 14.8%
24 to 48 months 2,614 18.5% 39.5% 16.6%
48 months or more 1,062 7.5% 32.4% 13.7%
JFA Associates/The Institute, Washington, DC/Austin, Texas 26
The Impact of Treatment by Risk Level
Level of Treatment
Study Risk Level Minimal Intensive
O’Donnell et al (1971)
Low 16% 22%
High 78% 56%
Baird et al (1979) Low 3% 10%
High 37% 18%
Andrews & Kiessling (1980)
Low 12% 17%
High 58% 31%
Bonta et al (2000) Low 15% 32%
High 51% 32%D.A. Andrews and James Bonta. 2003. The Psychology of Criminal Conduct (3rd ed.). Cincinnati: Anderson Publishing. p. 260.
JFA Associates/The Institute, Washington, DC/Austin, Texas 27
Key Points About Offender Risk
• Prisoners are slowing down or ending their criminal activities
• Probation versus prison is slightly more effective
• Extending or reducing prison terms is not related to recidivism
• Extending or Reducing parole/probation supervision is not related to recidivism
• Prisoners who “max out” do better than those paroled
• Small Percent (5%-10%) of all crimes are committed to persons released from prison.
• Very small % of released prisoners are re-arrested for murder or rape (less than 1%) and very small % of released murders or rapists are re-arrested for these crimes (under 2%).
• Treatment for Low Risk Persons Increases Recidivism while treatment for high risk persons reduces recidivism
JFA Associates/The Institute, Washington, DC/Austin, Texas 28
Factors that Predict
Static Factors
Age at First Arrest
Gender
Prior Supervision Failures (recent)
Mental Health Problems
Crimes of Economic Gain
Substance Abuse History
Prior Gang/Peer Associations
Dynamic Factors
Current Age
Current Education Level
Current Employment
Marital/Family Status
Gang/Peers Associations
Residency
Treatment (Good versus Bad)
Institutional Conduct
JFA Associates/The Institute, Washington, DC/Austin, Texas 29
Vermont Risk Level Results
Risk N % Recidivism Rate
Low 151 23% 26%
Moderate 293 45% 49%
High 200 31% 67%
JFA Associates/The Institute, Washington, DC/Austin, Texas 30
Kentucky Parole Board Risk Assessment Decision by Risk
Risk Level Cases Heard Grant Rate
Level I (0-6 pts) 1,475 45.7 Level II (7-11 pts) 4,754 37.0 Level III (12-14 pts) 1,522 29.3 Level IV (15+ pts) 183 18.6
JFA Associates/The Institute, Washington, DC/Austin, Texas
Kentucky Parole Board Risk Assessment Decision by Offense Severity
Offense Severity Cases Heard Grant Rate
Low 2,394 42.7 Moderate 2,335 35.9 High 2,759 34.4 Highest 445 21.0
32
Texas Expected Levels of Parole Grants
Offense Severity Class
RISK LEVEL
Highest(12+)
High(9-11)
Moderate(6-8)
Low(0-5)
Highest0-5%
15-15%
25-15%
216-25%
3
High5-15%
216-25%
321-35%
421-35%
4
Moderate5-15%
221-35%
436-50%
551-75%
6
Low16-25%
321-35%
451-75%
676-100%
7
33
Texas FY 06 Actual Versus Expected Grant Rates
Guideline Expected Actual
1 0%-5% 3%
2 6%-15% 13%
3 16%-25% 17%
4 26%-35% 20%
5 36%-50% 33%
6 51%-75% 41%
7 76%-100% 48%
JFA Associates/The Institute, Washington, DC/Austin, Texas 34
Do’s and Don’ts
• Must Be Tested On the Your Correctional Population
• Must Conduct Objective and Independent Inter-Reliability and Validity Tests
• Must Allow for Dynamic and Static Factors that Have Been Well
Accepted and Tested in a Number of Jurisdictions
• Must Be Compatible With the Staff’s Skill Level
• Must Be an Opportunity to Depart from Scored Risk Levels Based on a System of Structured Clinical Judgments
• Must Have “Face Validity” with Staff, Offenders and Policy Makers
JFA Associates/The Institute, Washington, DC/Austin, Texas 35
Strategic Steps for Building Risk Assessment Instrument
Agree on the Need for Risk Assessment
Conduct Recidivism Study of Released Prisoners
Build Risk Instrument Based on Recidivism
Conduct Reliability Study
Implement/Monitor
Finalize Risk Instrument
JFA Associates/The Institute, Washington, DC/Austin, Texas 36
Summary Points
• You cannot afford not to use risk assessment in release and supervision decisions
• Not using risk assessment worsens public safety• Key Areas of Concern
1. Imposition of treatment and conditions to low risk prisoners 2. Excessive periods of supervision (more than 12 months)3. Re-incarceration for non-criminal behavior or misdemeanor crimes4. Excessive periods of confinement – the diminishing return problem
• The release decision• The revocation decision
5. The lack of information and its contribution to mythology• DUIs• Sex Offenders• Public safety
6. Gender and Racial Bias