Critical Examination of Actuarial Offender-Based Prediction Assessments

384 Pages·2003·1.1 MB·English

by  Michele M. Connolly

The author(s) shown below used Federal funds provided by the U.S.
Department of Justice and prepared the following final report:
Document Title: Critical Examination of Actuarial Offender-
Based Prediction Assessments: Guidance for
the Next Generation of Assessments
Author(s): Michele M. Connolly
Document No.: 202982
Date Received: 11/21/2003
Award Number: 2001-IJ-CX-0003
This report has not been published by the U.S. Department of Justice.
To provide better customer service, NCJRS has made this Federally-
funded grant final report available electronically in addition to
traditional paper copies.
Opinions or points of view expressed are those
of the author(s) and do not necessarily reflect
the official position or policies of the U.S.
Department of Justice.
Copyright
By
Michele Moczygemba Connolly
2003
The Dissertation Committee for Michele Moczygemba
Connolly certifies that this is the approved version of the
following dissertation:
A Critical Examination of Actuarial Offender-Based Prediction
Assessments: Guidance for the Next Generation of Assessments
Committee:
___________________________
William R. Kelly, Supervisor
___________________________
Patricia Harris
___________________________
Dan Powers
___________________________
Art Sakamoto
___________________________
Mark C. Stafford
A Critical Examination of Actuarial Offender-Based Prediction
Assessments: Guidance for the Next Generation of Assessments
by
Michele Moczygemba Connolly, B.A., M.A.
Dissertation
Presented to the Faculty of the Graduate School of
the University of Texas at Austin
in Partial Fulfillment
of the Requirements
for the Degree of
Doctor of Philosophy
The University of Texas at Austin
August 2003
Dedication
This is dedicated to three very special individuals without whom this document
would not exist.
Dimitria D. Pope and Marty M. Martin
Without your daily support and encouragement over the past ten years, this
would never have happened.
Matthew Moczygemba Connolly
Without your love, understanding and advice, this would never have been
finished.
Acknowledgements
I would like to acknowledge the assistance given to me by my
dissertation committee: Dr. William R. Kelly, Dr. Patricia Harris, Dr. Dan
Powers, Dr. Art Sakamoto, and Dr. Mark C. Stafford. Dr. Kelly inspired me
to conduct policy relevant research without loosing focus of the importance of
sound theoretical grounding. He provided considerable advise and support
throughout this project. Dr. Harris exposed me to the world of criminology
over twelve years ago and hooked me on this life-altering path. Her
knowledge and passion for this field has and continues to be a source of
learning and strength. Dr. Powers and Dr. Sakamoto were my research
methodology and statistics instructors. Through their expert instruction they
opened new worlds of knowledge for me. Dr. Stafford spent endless hours
with me instructing me in criminological theory. Through our hours of
independent study, he instilled in me a strong appreciation for applying theory
in addressing policy research questions.
Sam Field provided statistical consulting and confirmation. Andy
Johnson, Graduate Coordinator, has patiently assisted me in getting through
the bureaucracy of the university.
I must also acknowledge all of the individuals who read and
commented on various versions of this dissertation. Mike Eisenberg offered
his expertise in assisting me on clearly communicating the distinction between
classification and prediction and how they interact. Dimitria D. Pope and
Marty Martin, dear friends and coworkers for over ten years, have advised,
counseled and supported me through my academic career.
v
They read and commented on numerous drafts of this dissertation, but more
importantly encouraged and assisted me to not give up. Matt Connolly, my
husband and editor in chief, has been invaluable in too many ways to list in
assisting with the completion of this project.
Dr. Todd Clear provided guidance in the data collection form used in
this project and shared his vision on where he saw risk prediction going in the
future. Ray Gingerich hired me for my first real job, which was developing a
risk prediction instrument. His wisdom and knowledge inspired me to pursue
this dissertation topic.
Finally, I must thank my family and friends who have supported me
and been there to share my good times as well as my bad. Rocky, Mark and
Rachel all I can say is thank you, and yes I’m finally done.
vi
A Critical Examination of Actuarial Offender-Based Prediction
Assessments: Guidance for the Next Generation of Assessments
Publication No._______________
Michele Moczygemba Connolly, Ph.D.
The University of Texas at Austin 2003
Supervisor: William R. Kelly
This study critically examines the prediction and classification aspect
of the community supervision process. Probation departments across the
United States, Canada and Europe use assessment instruments to attempt to
predict who is likely to continue to engage in criminal behavior so that they
can be classified and supervised accordingly. This study focuses on four
fundamental questions: What is prediction and classification of offenders?
Why are prediction and classification important? What do we know about the
reliability and validity of prediction and classification applications? How can
prediction and classification be improved?
The methods of the study consists of constructing risk prediction
models to compete against one of the most commonly used risk assessments in
the field of community supervision: the Wisconsin risk and need assessment.
Over thirty logistical regression models are constructed in an attempt to
improve upon existing technology. Models are constructed for the outcomes
rearrest, probation revocation and probation success.
vii
The findings of this study in no way diminish the need for accurate prediction
and appropriate assessment. They do show that the predictive power of the
most commonly used assessment instruments and instruments based on current
data and methods is negligible and therefore should not be relied on as a sole
factor in classification.
Concluded is that significant improvement in offender risk prediction
instruments will likely only be made if the specifications of the instruments
become more closely linked with criminological theory. Utilizing a battery of
assessments grounded in theory that take into account the offender’s
characteristics and the community in which they reside, may be the only way
we make progress in predicting their likelihood of future offending.
viii
Table of Contents
List of Tables……………………………………………………………..….x
List of Figures………………………………………………………………xiii
List of Appendices………………………………………………………….xiv
Chapter 1: Introduction………………………………………………… 1
Chapter 2: Overview of Prediction and Classification…………………. 5
Chapter 3: Research Design and Methodology……………………… 38
Chapter 4: Descriptive Analysis of the Felony Cohort……………….. 63
Chapter 5: Testing the Validity and Reliability of Predictor Variables.. 81
Chapter 6: The Construction of Rearrest, Revocation and Successful
Probation Prediction Models………………………………108
Chapter 7: Discussion and Conclusion………………………………..143
Appendices…………………………………………………………….… 169
Bibliography………………………………………………………………351
Vita………………………………………………………………………..369
ix