Justice Modelling Presentations 2008

Fourth National Justice Modelling Workshop

Griffith University and Queensland Government recently hosted the Fourth National Justice Modelling Workshop, 10-11 July 2008.  The two-day workshop was opened by Dr Peter Crossman, the Assistant Under Treasurer and Government Statistician within Queensland Treasury.  Professor Alfred Blumstein, who in 2007 was the recipient of the prestigious Stockholm Prize in Criminology for his research into criminal careers, presented the Keynote Address which was attended by over 100 invitees and workshop participants.  The Keynote Address examined the different approaches that have been used to model the justice system, focusing on the interaction of criminal careers and incarceration policy.  The workshop included 16 presentations that showcased how a range of recently developed models (including trajectory, simulation, resource allocation, economic and spatial) had been used to facilitate evidence based decision making in social and justice organisations.  The workshop had in excess of 50 attendees from New Zealand, interstate, and included representatives from many Queensland Government agencies.  Given the success of the workshop, a fifth workshop is planned for 2010.  

Abstracts from the workshop are available below by clicking the presenter’s name.  Presentations are available for download by clicking the relevant title.  Additionally, four photographs from the workshop have been included for viewing:

Professor Alfred Blumstein. Policy Modelling of the Criminal Justice System
Daniel Nagin. The Relationship between First Imprisonment and Criminal Career Development: A Matched Samples Comparison
Brent Davis. The Akuna Model of the Australian Criminal Justice System
David Harpham. Longitudinal Data and Drawing a Richer Picture of the Criminal Justice Sector
Trang Tang. Simulation Model of the Magisterial Workload in the Kimberley Region of Western Australia
Ben Smith. OOHC Funding Simulation Model
Matthew Manning.  Developing the Analytic Hierarchy Process for  Evaluating Alternative Early-in-Life Intervention Programs
Shane Perkins.  Criminal Justice Sector Transformation – Calculating the Social and Economic Benefits of Projects Delivered in the  Queensland Public Sector
Paul Henderson. A Pipeline Model of the  New Zealand Criminal Justice System
Tony Simmers. Justice System Pipeline: Development Challenges
Jason (Qingsheng) Wang and Ross Edney. Forecasting using Discrete Event Simulation for the NZ Prison Population
Lucy Snowball, Steve Moffatt, Don Weatherburn and Melissa Burgess.  Did the Heroin Shortage Increase Amphetamine use in NSW?
Nadine Smith and Craig Jones. Monitoring Trends in Re-offending Among Convicted Offenders in Adult and Children’s Court
Steve Moffatt and Lucy Snowball. Explaining NSW Long Term Trends in Property and Violent Crime
Dan Birks. Synthesis over Analysis:  Using Agent-Based Models to Examine the Interactions of Crime
Michael Townsley and Susan Donkin. There and Back Again: Putting Journeys Back into Journey to Crime Research
Anna Stewart, James Ogilvie and Troy Allard. Community Transitions and the Spatial Distribution of Crime

Professor Alfred Blumstein

Policy Modelling of the Criminal Justice System
H. John Heinz III School of Public Policy and Management, Carnegie Mellon University, Pittsburgh, PA, USA

Over the past forty years, a number of colleagues and I have been involved in bringing Operational Research (OR) techniques of quantitative modelling, system perspective, and planning to the justice system. The issues addressed have included modelling of criminal careers as a stochastic process, bringing those analyses to the assessment o incapacitation effects of incarceration, review of trends in incarceration  and factors contributing to those trends. We will address some approaches to modelling in the criminal justice system and devote particular attention to the interaction of criminal careers and incarceration policy.

Daniel Nagin

The Relationship between First Imprisonment and Criminal Career Development: A Matched Samples Comparison 
John Heinz III Professor of Public Policy and Statistics Carnegie Mellon University, Pittsburgh, PA, USA
  
Using data from the Netherlands-Based Criminal Career and Life-course Study we examine the effect of first-time imprisonment between ages 18-38 on the conviction rates in the three years immediately following the year of the    imprisonment.  Unadjusted comparisons of those imprisoned and those not imprisoned will be biased because imprisonment is not meted out randomly.  Selection processes will tend to make the imprisoned group  disproportionately crime prone compared to the not imprisoned group.  In this study we combine group-based trajectory modelling with risk set matching to balance a variety of measurable indicators of criminal propensity.  We find that first-time imprisonment is associated with an increase in criminal activity in the three years following release.  The effect of imprisonment is similar across offence types.

Brent Davis

The Akuna Model of the Australian Criminal Justice System  
The Australian Institute of Criminology (AIC), Canberra, Australian Capital Territory

The Australian Institute of Criminology (AIC) has initiated a major research project to model the Australian criminal justice system. The model has been named Akuna, an Aboriginal word meaning ‘flowing water’. The Akuna model is a macro-simulation, stock and flow design, and is intended to provide robust evidence-based input for policy analysis and projection (both ‘what if’s’, and ‘solver’ analyses). Reflecting the nature of the aggregate data currently available, the Akuna model is not intended to be a forecasting platform per se, or directly applicable for operational and resource management and planning purposes at the individual institutional level.
 
Work on the model commenced in early 2008, and has been informed by a range of sources: the Australian Bureau of Statistics’ National Criminal Justice Statistical Framework; and, macro- and micro-simulation modelling undertaken elsewhere in Australia, and in a number of other countries. The Akuna model currently has five linked sectors:  crime; investigation; adjudication; sentencing; and corrections.  The model is currently focused on the stocks and flows of offenders, although there are plans to extend the model to take into account financial costs of the operation of the criminal justice system. The Akuna Modelling work is being undertaken by the AIC’s Modelling and Forecasting team, and the presentation to the Justice Modelling Workshop is a ‘first external exposure’ of work-in-progress, seeking constructive comment from those with experience of similar projects.

David Harpham

Longitudinal Data and  Drawing a Richer Picture of the Criminal Justice Sector
Department of Corrections, New Zealand
 
New Zealand Corrections’ data has been clarified to a set of simplified longitudinal offender histories for all  NZ Corrections’ managed offenders since 1980.  The specification and utility of this data is explored with some examples giving new insights into the evolving NZ offender population.  A simple graphical timeline method for  describing longitudinal offender histories is introduced. For example the micro graph of this offender: shows him to have a long history of sanctions. This compact way of describing an offender’s history is  used to explore some individuals with colourful pasts.  Example collections of timelines of offenders dealt with in a typical day by Corrections are given and graphical exploration of offender typologies is started.

The use of such a rich graphical description of offenders is discussed in relation to understanding intervention, evaluation and complementing and communicating statistical results. A concept for modelling a national offender pool is introduced and a graphical representation of the pool of New Zealand offenders explored.  The offender pool concept is built from the full national set of all offender timelines and approximates all active offenders as being those that have been managed by Corrections within the last 10 years. Conclusions reached include:
•    that the NZ offender population is aging
•    that the  overall number of active offenders is dropping per head of population, but the proportion under current management  increasing
•    that the use of timelines enables improved comprehension of the criminal justice arena

Trang Tang

Simulation Model of the Magisterial Workload  in the Kimberley Region of Western Australia
Department of Attorney General, Perth, Western Australia
 
The Western Australian Department of the Attorney General (DotAG) is actively pursuing the development of a series  of simulation models of court processes to assist in the future planning, financing and management of courts.  However, the approach provides a number of additional ancillary benefits. This paper presents a case study, using a  predictive simulation modelling approach, to assess the impact on court performance if a second magistrate was appointed  to the Kimberley region of Western Australia.  The model demonstrates the ability of simulation to use many  variables (Circuit locations, travel time, judicial sitting times) to produce answers that could correctly guide a  decision maker.  It demonstrates that models become more complex as the number of variables increase.  Hence, it is important to adequately account for the more important variables.  In this case it is clear that variables such as offence type, age, gender and ethnicity play an important role in assessing the implications of policy and legislation changes.

The paper concludes that simulation modelling is capable of demonstrating the effects of policy and many other changes on performance in a complex environment and is, therefore, of great assistance to management and policy decision makers.  Simulation modelling is one of a number of statistically sound methodologies that enable policy makers to make evidence-based decisions.  In the case of this particular model, it was able to support a successful budget bid by DotAG to obtain the additional funds necessary to place the additional magistrate in the Kimberley by demonstrating downstream effects on court performance through a series of scenarios which allowed for variation in future workload and magisterial resources.

Ben Smith

OOHC Funding Simulation Model  
Department of Communities, Ashfield, New South Wales

The New South Wales Department of Community Services is a state child protection agency in Australia. It has developed a stock and flow discrete-event simulation model that forecasts the population of children in out-of-home care. This will allow planning and funding of out-of-home care services, as well as forecasting carer payment expenditure. It utilises estimates of unit costs for different service components and makes some assumptions about the average service packages and carer payment rates to be provided for the different types of children. Lower, medium and high need clients receive services of different intensities and duration, as well as a different mix of contingency spending. For example, high-need clients receive more frequent counselling than medium needs clients, as well as higher allowance and contingency spending and a different mix of contingencies, such as more respite. 

The model provides a broad indication of the mix of program funded services needed, including casework,  psychological services, tutoring, residential and intensive foster care in each region over the next five years.   This is assisting DoCS to fund the right services to help promote the best outcomes for children and young  people in OOHC. It was recently used to allocate notional funding for the OOHC Expression of Interest and is  currently being used to project the impact of future service improvements. For example, early intervention strategies may cause a reduction in high-need clients later on, changing future caseloads and service needs.  Children have been found to stay longer in relative/kinship care compared to foster care. The biggest cause of out-of-home care population increases is the greater use of relative/kinship placements since 1998. The Model  incorporates only five parameter shifts but its out-of-home care population projections closely predict observed  past data, despite excluding all random changes. Forecasts are being continually updated as relevant new information  becomes available.

Matthew Manning

Developing the Analytic Hierarchy Process for  Evaluating Alternative Early-in-Life Intervention Programs
Key Centre for Ethics, Law, Justice  and Governance, Griffith University

Criminology scholars and practitioners have come to recognise the  advantages of programs aimed at early intervention compared with those dealing with problems once they present  themselves (Brooks-Gunn, Fuligni & Berlin, 2003; Farrington and Welsh, 2002). A plethora of early intervention  programs have emerged such as the Perry Preschool Project (Schweinhart, 2004), the Elmira Prenatal/Early Infancy  Project (Eckenrode, Olds, Henderson, Kitzman, Luckey, Pettitt, Sidora, Morris, Powers, & Cole, 1998; Olds, 2002)  and the Seattle Social Development Project (Hawkins, Catalano, Kosterman, Abbot, & Hill, 1999); Each of these programs impacts differentially on outcome domains of interest to policy makers. For example, some early intervention programs have strong positive effects in relation to educational outcomes (e.g. intellectual and academic scores) and less impact in relation to family outcomes (e.g. family well-being) (Manning, 2008). This results in policy makers having to make complex decisions involving the weighing up of differential performances of alternative options in relation to a set of highly desirable criteria (Manning, 2008). This paper presents the results of an economic study which developed a methodology for assisting policy makers confronting resource allocation decisions with respect to alternative policy options for early childhood  intervention. The focus was on the impact of these intervention options on non-health related quality of life outcomes  in the adolescent years. The methodology that emerged involved an adaptation of an operations research tool (namely the   analytical hierarchy process) and use of the Expertchoice software package to assist in demonstrating its usefulness based on responses obtained from surveys conducted with academics, policy makers and practitioners’ experiences in the early childhood intervention field.

Shane Perkins

Criminal Justice Sector Transformation– Calculating the Social and Economic Benefits of Projects Delivered in the Queensland Public Sector
Department of Justice and Attorney General, Brisbane, Queensland

The Integrated Justice Information Strategy (IJIS) is a Queensland Government (Australia) initiative to enhance community safety by improving information sharing and collaboration among criminal justice agencies.  The budget for the State government program is $35.7 million to deliver eleven projects over seven years commencing in 2003-2004.  This whole-of-government initiative aims toimprove business processes and ence the efficiency and effectiveness of Queensland’s criminal justice sector and provide better communication and information systems.  These outcomes will contribute benefits back to the community as well as enhance the capability of all criminal justice agencies.  The original IJIS consideration was that internal efficiency benefits within and across multiple Queensland Government agencies would justify the expenditure required to implement the initiative.  However, many of the expected benefits are anticipated to be public benefits, in that they accrue to the community rather than provide direct financial benefits for the government.  Previously, on the productivity benefits of IJIS had been valued and there was now a need to value the social and economic benefits for the community as a whole.

In order to identify realistic and justifiable values for the social and economic benefits the criminal justice sector would need to be modelled and benefit indicators agreed and quantified in an environment of limited research data and great sensitivity.  By working collaboratively the Queensland Government has obtained a comprehensive valuation of the social and economic benefits that can be achieved through IJIS.  The social and economic values, in conjunction with agency focused returns, can then be used to guide development of IJIS projects to ensure the maximum benefits are realised for the Queensland community.  Benefits management methodologies acknowledge that economic and social benefits are important for public sector investments, but provide little guidance on how to value such benefits.  This case study shows that the benefits can be estimated using a multi-disciplinary approach comprising strong stakeholder involvement, economic expertise and simulation modelling. 

Paul Henderson

A Pipeline Model of the New Zealand Criminal Justice System
New Zealand Ministry of Justice

A model of the New Zealand criminal justice system is presented: the Justice Sector Pipeline model.  This has been constructed to provide a cross-agency view of the whole system allowing the assessment of the impact of legislative or operational changes on downstream agencies.  The model was developed using the simulation package Extend.  The first stage was the production of a prototype, with greater detail and functionality being provided by the subsequent addition of modules describing specific agencies (such as the police or the courts) or functions  (such as sentencing).  The presentation will describe the history to date, the present status of the model, and  some expectations for the future.

Tony Simmers

Justice System Pipeline: Development Challenges
New Zealand Ministry of Justice

Creating then deploying and maintaining a complex simulation model presents a number of challenges to those developing the model.  The JS Pipeline model is designed to simulate the implications  of changing policy settings in one part of the justice system on other parts of the system. Output from the model will be used to inform policy development and ensure that costs and system capacity issues are considered as part of the   policy development process.  The presentation will discuss how the development team has tackled technical issues   such as:
•    ensuring new features do not detract from existing work
•    establishing processes    for updating the data behind the model
•    keeping track of versions of the model that are run to address particular policy questions.

Jason (Qingsheng) Wang and Ross Edney

Forecasting using Discrete Event Simulation for the NZ Prison Population
New Zealand Ministry of Justice

Conventional methods of forecasting the population are not able to fully capture the processes that drive the prison population. Time-series forecasting models do not allow the detailed relationships between the prison population and its drivers, prisoner inflows, the sentence imposed, and the proportion of sentence served. Parametric models that attempt to establish structural relationships between the underlying drivers and population do not provide accurate forecasts when the relationships between drivers of the population change due to, for example, policy changes.

To forecast the prison population, a logic based dynamic “discrete event simulation” (DES) method is used to model the relationship between the population and its underlying drivers. The first model is built using the underlying drivers’ mean values. Historic data are fed into the DES model. Time-series and other forecasting methods are then employed to determine the future values for these drivers of the model. Using this approach, consensus forming discussions about the likely path of forecast drivers are more easily understood by policy managers and justice sector experts. The second DES model is employed for risk management purposes and uses the distributions of the population drivers to carry out a Monte Carlo analysis. The Monte Carlo technique produces a population forecast with an empirical distribution and confidence intervals. Confidence intervals are important to establish risks of bed shortages and to plan capacity. The analysis reveals that these plans need to allow for empirical distributions with heavy non-normal tails.

Lucy Snowball, Steve Moffatt, Don Weatherburn and Melissa Burgess  

Did the Heroin Shortage Increase Amphetamine use in NSW?
NSW Bureau of Crime Statistics and Research (BOCSAR), Sydney

Over the last ten years there has been a substantial growth in the use of amphetamine type substances (ATS). A number of studies have found evidence that the growth in ATS use may have been stimulated or exacerbated by the heroin shortage that began around Christmas 2000. One limitation of these studies is that they mostly involved interviews with groups of street or treatment-based former or current heroin users. There is no guarantee that the ATS-using population contains a large proportion of former heroin users, or that the patterns of drug switching evident in street or treatment based drug users are typical of heroin users in general.

This study examines whether there is a statistical relationship between the increase in ATS use and the decrease in heroin use as a result of the heroin shortage. Arrests for use/possess amphetamines and use/possess narcotics were used as a proxy for ATS and heroin use. Emergency department (ED) admissions for amphetamines and heroin were used to validate the results. We used time series techniques including vector auto regression (VAR) and auto regressive integrated moving average (ARIMA) modelling to test this relationship. Different time points and locations were modelled.

Nadine Smith and Craig Jones

Monitoring Trends in Re-offending Among Convicted Offenders in Adult and Children’s Court
NSW Bureau of Crime Statistics and Research (BOCSAR), Sydney

Internationally, governments are making renewed efforts to reduce rates of re-offending. Measuring progress against this objective is difficult because officially recorded reconviction rates are determined not only by how effective the justice system is in dealing with offenders but also by the characteristics of offenders coming to court in the first place. The current research describes the development of a technique known as the Group Risk Assessment Model (GRAM), which adjusts for the characteristics of offenders coming before the courts in order to obtain more accurate estimates of trends in re-offending over time. Separate logistic regression models were developed for juvenile and adult offenders given non-custodial sanctions in 2002. For juvenile offenders, age, sex, Indigenous status, prior convictions and concurrent convictions were found to be highly predictive of subsequent reconviction. For adult offenders, these same offender  characteristics, in combination with the jurisdiction in which the offender was dealt with and the offenders’ most serious index offence,  were found to provide a good model of reconviction likelihood. An application of the models in relation to the 2003 and 2004 adult and  juvenile offender cohorts revealed that there had been a statistically significant decrease in rates of reconviction among juveniles  convicted in 2004 but not among adults convicted in 2004. The observed reconviction rates among the 2003 adult or juvenile cohorts were  not significantly different from the expected rates given the characteristics of those cohorts.

Steve Moffatt and Lucy Snowball

Explaining NSW Long Term Trends in Property and Violent Crime
NSW Bureau of Crime Statistics and Research (BOCSAR), Sydney

Over the last 7 years, recorded property crime in New South Wales (NSW) has been dropping dramatically, on the back of a steady increase from 1995 to 2000. From 1995 there has been an increasing long term trend in recorded violent crime which has only stabilised in recent years. Some previous research has examined explanations for these trends and tested statistical correlates of the series however this has primarily focused on property crime.

This study examines the statistical relationship between levels of recorded crime (both property and violent) in NSW and possible explanatory factors. These include economic (such as the consumer sentiment index and quarterly rental vacancies index), law and justice (including imprisonment rates and police activity measures), drug and alcohol use, seasonal and other temporal factors. Time series techniques including autoregressive moving average (ARMA) and autoregressive integrated moving average (ARIMA) modelling were employed to test the relationship. Forecasts have  been developed in order to predict future crime levels as well as the impact on these levels of a simulated change in one of  the explanatory variables.

Dan Birks

Synthesis over Analysis:  Using Agent-Based Models to Examine the Interactions of Crime
Justice Modelling @ Griffith, Griffith University

This presentation aims to provide an introduction to the use of simulation techniques within crime analysis. It explores the methodology through presentation of a prototype agent-based simulation of residential burglary which incorporates existing theories of offending and several real GIS datasets. Current discourse in environmental criminology posits that macro-level crime patterns are best described as an aggregation of micro-level interactions between potential offenders and their environment. However, conventional crime analysis has typically adopted a top-down analytical strategy, focusing on the outcome of these interactions: the crime itself. In comparison, simulation techniques offer a novel opportunity to examine emerging crime patterns using bottom-up approaches, formalising and subsequently simulating theoretical concepts from the domains of environmental criminology, cognitive science and ecology. Such simulations can act as complex thought experiments which allow researchers to explore the ramifications of their theoretical assumptions in an attempt to assess validity. The first section of this presentation will introduce simulation techniques by describing a simple model of victimisation and detection, as well as the assumptions and formalisms associated with its development. The second will focus on a more complex agent-based model of volume crime which attempts to more closely replicate the individual level interactions of the crime event. Initial results from this research will be described, demonstrating that plausible macro-level crime patterns can be observed as the cumulative output of numerous micro-level agent interactions. Finally, the potential applications of such simulations to both examine and refine current theories, as well as their potential utility as practical crime prevention tools, will be discussed.

Michael Townsley and Susan Donkin

There and Back Again: Putting Journeys Back into Journey to Crime Research
School of Criminology and Criminal Justice, Griffith University

Journey to crime research is the body of knowledge that links the spatial  distributions of offenders and the spatial distributions of where offences are committed. However, much of the journey to crime literature focuses exclusively on the distance travelled, largely ignoring the directional component of the crime trip. We examine the “direction to crime” through the exploration of two key questions, namely whether there is a relationship between trip origin and trip direction; and whether there is a relationship between trip destination and trip direction. Our findings were congruent with previous research, showing that the closer offenders live the more similar their mean travel directions are. Results for destination and direction vary by crime type, showing consistency for burglary offences, with robberies displaying more contrasting patterns. There is also evidence of considerable variation in patterns by area. A number of crime reduction implications, both preventative and investigative in nature, are outlined.

Anna Stewart, James Ogilvie and Troy Allard

Community Transitions and the Spatial Distribution of Crime
Justice Modelling @ Griffith, Key Centre for Ethics, Law, Justice and Governance, Griffith University

Available empirical evidence indicates that the links between neighbourhood social and economic conditions and crime are complex, with a range of processes postulated to mediate these links including; social disorganisation, collective efficacy, parenting practices and deviant peer associations.  While many studies have examined the spatial correlates of crime few have examined the impact of community change on crime.  Currently Queensland is experiencing unprecedented economic and population growth. Between the 2001 and 2006 census, there was a 10.7% growth in population and a 28% growth in median household income.  However, this growth is not consistent across the state, some urban areas have increased by as much as 200% while many rural and remote areas are experiencing population loss.  

Between 2001 and 2006 Queensland experienced a drop in the rate of offending but again this drop was not experienced in all areas or all crimes. In this paper we examine the impact of social and economic change on crime rates across Queensland.  Different crime profiles were identified for both urban and rural/remote areas.  The results highlight the importance of understanding the dynamic nature of communities when examining crime. These results are discussed in relation to spatial theories  as well as the implications for community based crime prevention strategies.