Criminal Justice Simulation Model (CJSIM):
Technology and the Flow of Criminals in the Criminal Justice System
Rod MacDonald
Mohammad Mojtahedzadeh
Corresponding Author:
Rod MacDonald
Initiative for System Dynamics in the Public Sector
300G Milne Hall
Rockefeller College of Public Affairs and Policy
University at Albany
135 Western Avenue
Albany, New York 12222
Tel. 518-442-5772
rmacdonald@albany.edu
Funding for this project was provided through a grant from the National Institute of Justice.
Grant Number 1413683
SD Conference Version 1
Criminal Justice Simulation Model (CJSIM):
Using a Simulation Model to Examine the Allocation of Technology to Im-
prove the Criminal Justice System
Abstract
This paper examines what would happen in the New York’s criminal justice system if two pri-
mary changes were made. Those changes are an exogenous doubling of the number of new
criminals entering the system (new people becoming criminals) and a doubling of the productiv-
ity of police officers. These inputs were selected because they represent changes on the front or
upstream end of the criminal justice system, and the effects of these changes on the whole system
could be observed. Changes in police capacity, number of people in prison and the total number
of active criminals are examined to understand the implications that technological improvements
that increase police productivity may have on other parts of the criminal justice system.
Introduction
The criminal justice system in New York State is large, complex and diverse. Although referred
to as a system, most entities in the criminal justice system operate independently. Separation of
functions, such as those of the courts from the DA or police departments, is built into the system
to ensure that courts operate independently from the rest of the system. Although cooperation
between different parts of the system occurs, the lines of command, budgets and structures of the
various entities within the criminal justice system have developed independently over time. As a
result, decisions and actions taken in one part of the system have implications for other parts of
the system. At the Symposium celebrating the 30th anniversary of the 1967 President’s Com-
mission on Law Enforcement and Administration of Justice publication titled “The Challenge of
Crime in a Free Society,” Henry Ruth said that the two greatest contributions made by the Com-
mission were “(1) recognizing that the criminal justice agencies and processes should be viewed
as a system, and (2) stimulating data collection and analysis” (p. 6). This paper examines the
influence of resources and technology on the flow of criminals through the criminal justice sys-
tem in upstate New York from a system dynamics perspective.
This paper discusses the development of a system dynamics computer simulation model that was
built to examine the influence of technology on the flow of criminals through the criminal justice
system (the model will be referred to as CJSIM). The basis for this model is the criminal justice
system in New York State. The model tracks criminals from arrest to adjudication, from their
time in jail or prison to parole and probation. Furthermore, the model tracks criminals after re-
lease in order to capture and examine the concept of recidivism. The model specifically exam-
ines the system-wide implications of technological improvements in different parts of the system
and whether improvements in one area will result in performance problems in other parts of the
system.
Page 2 of 21
SD Conference Version 1
Background on Computer Simulation Modeling in Criminal Justice
The application of quantitative techniques in the Criminal Justice System goes back to the 1950s
and 1960s (Blumstein, 2002). Scholars have long recognized the need for tools and techniques
to aid decision-makers in understanding the interrelationships between different agencies in the
Justice System. Cassidy argues that decisions made in one agency will affect other agencies
within the system, (Cassidy, 1985, Cassidy and Turner, 1978). This reason alone justifies the
need to take a system wide perspective, enabling each agency to take into account the impact of
its own decisions as well as the decisions of others on the whole system.
Blumstein (1971, 2002) reports extensively on the development of JUSSIM, an interactive simu-
lation model of the criminal justice system that contains three basic elements: process stages,
flow paths and resources utilized. JUSSIM served as a pedagogic device for understanding the
ripple effects of decisions and actions, as well as a planning tool for the justice system that is
“largely fragmented among the police, court and correction agencies” (Jacqueline Cohen et al
1973). One of the insights from JUSSIM was the relationship between the prevalence of arrest
and the probability of recidivism. Blumstein and Graddy (1982) found significant race differ-
ences in prevalence of arrests. Similar rates of recidivism between races was also found, indicat-
ing that similarities among individuals in “criminal careers” exists regardless of race.
Jonathan Bard of the Aerospace Corporation (1977) described the use of a system dynamics
simulation approach in “assessing the merits of alternative criminal justice policies and proce-
dures” and concluded that a continuation of current practices would lead to a gradual decline in
the ability to deliver services and control crime. Brad’s repeated experiments with the simulation
model led him to uncover the notion that a reduction in the grand jury delay may cause a short-
term imbalance between felony trial queue flow rates. Moreover, he found that it could also lead
to a trial delay fractionally greater than that first observed.
Shaffer (1976) used system dynamics modeling to examine the criminal justice system in Massa-
chusetts with an emphasis on the courts. However, he did develop a smaller model that included
all the components of the criminal justice system and found that increased prison capacity had
the greatest impact on reducing crime. At the time Shaffer was doing this work, prison sentences
had been substantially reduced due to overcrowding in prisons. Shaffer found that the reason for
the impact of prison capacity was due to the deterrent effects of sentences and that by keeping
people in prison longer reduced the cycle time for criminals.
The model discussed in this paper was designed to build upon the work that has been done over
the last 50 years and begin to develop system-wide maps and models that portray the complexity
of the criminal justice system. The final outputs of the model, as well as the process of building
the model, should help decision-makers gain a better understanding of the criminal justice sys-
tem, enabling them to formulate more effective policies for better performance.
Page 3 of 21
SD Conference Version 1
tom
on Beg
spedxy Aq padojanag dey MaIMaAQ Moj4 pue yO0}S =} aunbi4
afinere mewn
Ea)
pupa Bap
Bese poi
Baran Ro
voi pois plas d
40 00s 9
vo spawn
228 poms
pl pos (stay
oH 0) sow
peas (shay 04)
pay seem
sau 4
ei a pe spay
rea pam
apc sean ns amy
er
la exmepipy Basan
vow
adi y sea
Jor st ep poe oN
Pay sy Cn
ro belg in botany rng Beg sus l
ssapony pe say pe Say
x png spunsy vy ws =
id (io
‘ou Peg paren
29.94 bane, ENE
sea
kay ay pes
et wpe po
ay nog wa, Kbeway mproy ined
Suen som. ne a
(oo '28'Pe 10)
PRY pom
ep
Page 4 of 21
SD Conference Version 1
Modeling Process
The modeling process was initiated through open-ended interviews with experts working in the
system. These experts worked in the system as police officers, sheriffs, DAs, judges, probation
officers, parole officers, and those working in the New York State Division of Criminal Justice
Services. The primary purpose of these interviews was to develop a stock and flow map of how
criminals moved through the system, from the time a crime was committed until they were re-
leased from any and all sanctions after capture and conviction. Data for the model were obtained
from publicly available information sources published by different government agencies in-
volved in the criminal justice system.
At each interview, participants examined the current stock and flow diagram. An explanation of
what they were looking at was provided, followed by a discussion of whether the places crimi-
nals could flow to were accurately captured in the diagram. Interviewees were then asked if, in
their opinion, there were other places that criminals could flow. The changes related to their
suggested additions and/or “subtractions” were then physically drawn on the map. For the boxes
(i.e., stocks) added, the interviewees were then asked where the criminals in those boxes could
flow to. For example, upon release from prison where do criminals go? Criminals can be re-
leased from prison and go to parole or they can move into a stock labeled Pool of Potential Re-
cidivists. Furthermore, the participants were also asked to identify the sources where criminals
came from before flowing into each box. The participants were also asked to identify what re-
sources were needed in order to facilitate or control the flow of criminals from one stock to an-
other. For example, in order for criminals to move from jail awaiting adjudication, resources
from the DA’s office and the courts would have to be used. Focusing on the area in which they
worked, the participants were then asked about the types of decisions they made and what infor-
mation they looked at to make those decisions.
This process was followed with all participants interviewed. As the process progressed, fewer
and fewer changes were made to the map. By the end of the interview process, a consensus on
the stock and flow structure of the model was reached, with discussions confirming that no fur-
ther structural changes to the map were needed.
Figure | contains the completed stock and flow map! of the flow of criminals through the crimi-
nal justice system in New York State, as identified by experts in the criminal justice system. The
left hand side of the diagram shows criminals not yet known to the system. These are people
committing crimes who have never been caught. Once criminals are caught, they move through
the system from left to right in Figure 1. After arrest, criminals await adjudication in one of three
places. These boxes capture criminals Awaiting Adjudication (In Jail), Awaiting Adjudication
While on Bond - Released on Own Recognizance (ROR), or Awaiting Adjudication (House Ar-
rest). From there, people can move to criminals Conditionally Discharged, Criminals in Jail,
Criminals in Prison, Criminals on Parole, and Criminals on Probation. All of these folks are un-
der some form of sanction from the criminal justice system. While criminals in jail and prison
' This is the map that experts came to a consensus on. Acquittals may occur regardless of where people are held
after arrested. However, acquittals are not shown as an outflow from criminals on House Arrest or Awaiting Adjudi-
cation in Jail. This was done in order to simplify the map.
Page 5 of 21
SD Conference Version 1
cannot commit any additional crimes while incarcerated, criminals on probation, parole and on
conditional discharge can commit additional crimes or violate the conditions of their sanctions
and can move immediately to jail or prison. ?
Parole is also considered a sanction, but criminals can only be on parole after serving time in
prison; thus they move from prison to parole. From parole, criminals can be sent back to prison
or they can successfully complete parole and move into the Pool of Potential Recidivists. Al-
though not explicitly shown in Figure 1, the Pool of Potential Recidivists is broken down into
two pools of recidivists based on time. Recidivism rates for those criminals making it through
the first five years of release is lower than for those within the first five years of release. Disag-
gregating these in the model allows for the tracking of these two groups and for the assigning of
different recidivism rates for these groups. Furthermore, once people have been arrested, con-
victed, sanctioned and flow into the pool of potential recidivists, they are considered potential
recidivists for the rest of their lives.
With the map of the system complete and agreed upon, the process of taking the feedback loops
identified by experts and incorporating them into the model began. Information obtained through
the open-ended interviews was combined with knowledge gained from the literature to develop
the feedback loops incorporated in the model. Data was collected from a variety of publicly
available sources to set the initial conditions. The formal model was tested at each stage in order
to understand the behavior generated from sub-sectors before they were added to the overall
model.
The Formal Model
CJSIM3 is a system dynamics computer simulation model developed from the sources described
above. The formal model captures the flow of criminals through the criminal justice system. For
the purposes of this paper, one scenario and three tests of technology were used to examine the
implications on different parts of the criminal justice system. The scenario test increased the
number of new criminals entering the system by 100 percent at time 20. Although extreme, it
was felt that this shock to the system could be performed under conditions when resources were
fixed and when resources were flexible, which would provide information related to time delays
in the system, from the time of arrest until the criminal ended up in the pool of recidivists from
which they have the opportunity to recidivate. An additional test focused on improvements in
technology at different points in the system, of which the initial influence should be to increase
the number of people in the system through police enhanced technology or improvements in DA/
Court processes or supervision of criminals on probation and/or parole. New technologies that
work should result in more arrests, convictions, or violations until criminals learn about the ef-
2 Crime can be committed while criminals are incarcerated, but these crimes were considered to be outside the
model boundary for the purposes of this model.
3 Complete stock and flow diagrams and the equation listing for the model can be obtained by contacting the
authors.
Page 6 of 21
SD Conference Version 1
fectiveness of the new technology and change their behavior to either evade detection or stop the
criminal activity.
Equilibrium Model
A system dynamics model is in equilibrium when all the inflows equal all the outflows for all the
stocks in the model. Each stock is unchanging and, as a consequence, all model variables are
unchanging over time (Richardson and Pugh, 1981). This is not to say that criminals do not con-
tinue to flow through the CJSIM model, rather, new criminals moving into a stock are equal to
those moving out of that stock. System dynamics models are put into equilibrium for the pur-
pose of analysis and understanding (Sterman, 2000). By starting in equilibrium behavioral
changes that result from changes in parameters and/or structure of CJSIM can initially be ana-
lyzed in isolation and behavior and structure can be linked more efficiently. Although simulating
the model in equilibrium means that the historical data for New York State will not be replicated,
it should be noted that the initial numbers used in the model were obtained by taking the average
number of criminals in jail, prison, on probation and on parole over the 1999-2004 time period,
the most recently available data at the time.
Simulation of CJSIM in Equilibrium
Graph | contains simulated output for Criminals in Prison, Total Number of Criminals in Jail*,
Criminals on Parole, Criminals on Probation, Criminal Conditionally Discharged and New Un-
known Criminals. These variables were selected as indicators of what is happening at each of
these points in the criminal justice system where criminals are under some form of sanction and
consume system resources. Any variable in the model may be examined to observe its behavior
over time. Furthermore, by starting in equilibrium, a base run from which changes can be com-
pared is available.
4 The Total Number of Criminals in Jail is made up of criminals in jail awaiting adjudication and criminals sentenced
to jail after conviction.
Page 7 of 21
SD Conference Version 1
250,000
187,500
125,000
62,500
0 12 24 «936048 sisi
Time (Month)
108 120
Prison
Criminals
Total Number of Criminals in Jail
Criminals on Parole
Criminals on Probation
Conditional Discharge : Base Run
Total Criminals : Base Run
Criminals
Scenario 1
Graph 1: Criminals Under Sanction
The first scenario involves doubling the number of new criminals entering the system at month
20. The simulation results for this scenario are shown in Graph 2. The number of Criminals on
Probation> increases as does the number of Criminals on Parole, and the number of Criminals
Conditionally Discharged. This is expected. As more criminals enter the system, the number of
Total Criminals overwhelms the capacity of the police to make arrests and grows off the scale.
> The convention that variable names are capitalized
Page 8 of 21
SD Conference Version 1
Graph 2: Dou-
bling the
Number| 250.000 of New
Crimi- nals
187,500 ST eT TT
Graph 3} 125,000 contains
infor- Imation
about) 62,500 ae the
number lof — ar-
rests in ‘ tt h e
model. 0 12 24 36 48 60 72 84 = -96-—=—S «108-120 A E
though Time (Month) t hoe
number lo f
ser Prison Criminals 1g
CYIM1| Total Number of Criminals in Jail Criminals [Mls en-
i Criminals on Parole Criminals S-
Cerin g Criminals on Probation Criminals tthe sys
tem as Conditional Discharge : Double Criminals 5 $- Criminals n © W
crimi- Total Criminals : Double Criminals Criminals nals
doubles, t he
number of arrests also doubles. The number of Total Criminals (this is active criminals under no
sanctions in the system) also continues to increase. This is due to a number of factors. One fac-
tor is resource constraints in the police sector. Although police resources remain the same, addi-
tional arrests are made for two primary reasons. First, there are more active criminals (the Total
Criminals variable and the Crime Solving Loop in Figure 2), and thus crime committed in-
creases, and it becomes easier for police to arrest criminals as there are more of them. Second,
the increase in crime generates command pressure (The Command Pressure Loop in Figure 2) on
police and that leads to additional arrests. This command pressure could take the form of over-
time or the shifting of resources so that more police are patrolling and undertaking investiga-
tions. However, there are limits to how much can be accomplished, and there is not enough
slack in the system to accommodate the additional criminals and crimes generated. Hence, when
the flow of new criminals into the system is greater than the ability of the police to make arrests,
the number of Total Criminals grows.
Furthermore, when a new Unknown Criminal is arrested for the first time, they become known to
the system, serve some form of sanction, and then move into the pool of potential recidivists
from which they are likely to be arrested again. Therefore, the introduction of a new criminal
into the system results in police resources being required to deal with the initial crime and, then
due to recidivism, they will deal these people again later. The problem of criminal behavior is
not solved for most criminals with one arrest. They become known to the system, but continue
to commit crimes when they are not under the sanction of jail or prison in the model.
Page 9 of 21
SD Conference Version 1
20,000
15,000
10,000
Criminals/Month
0 12 24 36 = 4860 724 96 «108 =120
Time (Month)
Total Arrests : Criminals Increased
Total Arrests : Base Run
Graph 3: Number of Arrests
Figure 2 contains two negative feedback loops, the Command Pressure Loop and the Crime
Solving Loop, described above. The negative feedback loops compensate for the increased
number of criminals in the system by placing pressure on the police to do more with the re-
sources they have.
Police Workload
Ratio
. Total
Effect of Pressure to Make Criminal
+ More Arrests on Police Police
Effect of Crime on Police Productivity ss Capacity
Officer Productivity (B) Command
H New Unknown
(B) Crime Pressure Loop crinincte
Solving Loop
¥ Total
+ Arrests
Police Officer Skee
Productivity
Figure 2: Feedback Loops Around Police Workload
Graph 4 shows changes in the prison population in the base run and when criminals in the system
are doubled (Scenario 1). As expected, the increase in criminals in the system results in addi-
tional criminals being sentenced to prison and the prison population increases. However, over
Page 10 of 21
SD Conference Version 1
time the number of criminals in prison falls and, after 90 months, is below the equilibrium condi-
tion for the number of criminals in prison. The reason for this, in the model, is that resources
throughout the system are not increased. The same number of police, DAs, judges, jail and
prison cells, and parole and probation officers remain constant in the system. The compensating
feedback loops in Figure 2 results in more arrests, but the DA / Judicial sector of the model can-
not keep up with the additional work. Figure 3 contains feedback loops associated with the DA /
Judicial sector of the model.
As the DA / Judicial Workload Ratio increases, it increases pressure for the DA to take action.
The DA thus increases the number of plea bargains by offering shorter sentences to criminals.
The shorter sentences work as an inducement to criminals to resolve the case quickly, thus reduc-
ing the DA / Judicial Workload Ratio. More cases are resolved and in the short-run prisons end
up with more prisoners, but the sentences are not as long and the shortened sentences result in
the overall decrease, after some time, in the prison population.
Graph 4: . Prison
Popula- Prison tion in
Base 80,000 Run and
with In- creased
Crimi- nals
70,000 TESS
b
60,000 eu
50,000
40,000
0 12 24 3648 60 72 84 96 108 120
Time (Month)
Prison : Scenario 1. +——+—-+—_+—_+_+—_+4+_+_ +++ Criminals
Prison : Base Run Criminals
Page 11 of 21
SD Conference Version 1
Pressure to Offer
Plea Bargains a," a Length of Prison /
+
Plea Jail Sentence
Bargains
DA / Judicial -
| (Reduce
Woneats. Rauly DA/Judicial in
Backlog Loop) DA Fadicial Recidivism
udicial
Productivity
Total Cases (More Crime
Awaiting i Loop)
dentate s Cases Bs
Adjudiciation~____adjudiciated
wt
Crime
Arrests
Figure 3: DA/ Judicial Sector
Technological Improvements
Technological improvements in the criminal justice field can take various forms, ranging from
improved police communication, better evidence handling, new monitoring techniques in jail or
prison, and/or better monitoring techniques or treatment programs when criminals are on proba-
tion or parole. For the purpose of model analysis, technological improvements have been added
to the model in the form of productivity improvements®. This captures those technological im-
provements that will result in increased movement of criminals from one place in the system to
another. For example, technological improvements that allow police officers to become more
efficient will allow them to make more arrests in a given time period. This will result in more
criminals needing adjudication, and more criminals going to prison, jail, probation and parole, if
all other things are held constant.
Improved Police Productivity
The model was structured so that police productivity could be increased at month 20 by changing
one model parameter. Graph 5 contains the output from increased police productivity to what is
® A second model has been developed that specifically addresses the issue of technology aiding productivity. In in-
terviewing folks from different agencies, it became apparent that some technologies generate more work, not only in
the short-run due to a learning curve, but in the long-run. Monitoring devices can now be placed on individuals and
parole, probation or those supervising people under house arrest will be notified when violations occur. This re-
quires additional work on the part of those monitoring. Furthermore, forms of technology, although electronically
recorded, require the review of people in order to capture violators. These reviews were not previously required as
the technology did not exist (interlock devices for felony DWI offenders is an example). Although this will be re-
ported in a different paper, it is worth noting here.
Page 12 of 21
SD Conference Version 1
assumed to be technological changes that will allow the police to catch more criminals. Police
productivity was increased by 50 percent in the simulation run below. The results are not as
dramatic, given the 50 percent increase in police productivity, as observed in the increases in the
number of criminals in prison, on probation, on parole and conditionally discharged. The num-
ber of criminals in jail remained the same since jail capacity was at a maximum in the base run.
So additional arrests do not lead to more criminals in jail due to capacity constraints. However,
the number of Total Criminals (these are active criminals not under any sanctions) decreases and
then starts to come back, leading to a decrease over the time of the simulation of approximately
28,000 or 18 percent (Graph 6 shows only Total Criminals for the Base Run and Police Produc-
tivity 1 run). The decrease and then increase in Total Criminals is due to additional arrests,
which leads to more criminals being under some form of sanction. However, criminals eventu-
ally get released and some of these criminals recidivate, becoming active criminals, and thus
push up the number of Total Criminals near the end of the simulation run.
Graph 5: Criminal
Locations After In-
creased | 75%:000 Police
Productiv- ity
187,500
ne ee ee |
125,000 -——
62,500
0
0 122 24 «49360-48728 HCSs—siOBSs«*1220
Time (Month)
Prison Criminals
Total Number of Criminals in Jail Criminals
Criminals on Parole Criminals
Criminals on Probation + Criminals
Conditional Discharge : Productivity 1 + + Criminals.
Total Criminals : Productivity 1 Criminals.
Page 13 of 21
SD Conference Version 1
Total Criminals
200,000
175,000
+.
150,000 [
125,000
100,000
Time (Month)
Total Criminals : Productivity 1 ———+——+—_+_+__+—_++ Criminals
Total Criminals : Base Run Criminals
Graph 6: Total Number of Criminals in the System
It was anticipated that a 50 percent increase in the productivity of police would lead to a substan-
tial decrease in the number of Total Criminals with a subsequent and offsetting increase in those
held in prison, parole, probation and conditionally discharged. There are a number of reasons in
the model for the changes observed in Graph 6. First, resources in the system were allowed to
vary’. Additional arrests by the police created stress downstream, but downstream resources
were allowed to increase. The DA/judicial sector were allowed to increase capacity as were pa-
role, probation, jails and prisons. The oscillation of the prison population observed in Graph 7 is
generated by the long time delays in bringing brick and mortar resources on line as with the con-
struction on new prison cells.
7 The desired resources for DAs, jails, prison, probation and parole were structured based on a normal-
ized workload ratio. If the workload ratio were equal to 1 then the amount of work to do was considered
to be normal. If the workload ratio were above 1 then the amount of work to do was above normal and
would require additional resources. If the workload ratio were below 1, then there were too many re-
sources for the work to be performed. Varying resources means that resources in the model were al-
lowed to change, after an appropriate time delay, so that the workload ratio would move back to 1.
Page 14 of 21
Graph 7:
lation
The model}
tured so that
lice Capac-
termined by|
nals, Police
and an Ini-
Time to Ar-|
captured the
resource al-|
the police
be based on
SD Conference Version 1
took, on av-
[Prison Popu-
Prison
80,000
75,000
lwas struc-
[Desired Po-
70,000 E
lity was de-
(Total Crimi-
65,000 f |Productivity,
tial Average
60,000 Irest. The
0 12 24 36 486072, 8496108120 concept that
‘Time (Month) locations in
Prison : Productivity 1 Criminals {sector would
Prison : Base Run Criminals
lhow long it
solve crimes.
technology.
rage, to
Figure 4 captures a simplified version of this negative feedback loop. The in-
creased police productivity results in two effects. First, as police become more productive, fewer
officers are needed to do the same job. Second, as police become more productive, Total Arrests
increase and, in turn, Total Criminals and thus Crime decrease, resulting in a decrease in the De-
sired Police Capacity. The police sector in the model can capture the technological improve-
ments that generate increased productivity by reducing police capacity (manpower). This raises
a caveat about how different sectors of the system will react to technological improvements that
increase productivity. There will be more people in the system, but resource allocation decisions
will push back in order to capture savings as a result of the productivity improvements due to
Desired Police
Police
Productivity
+
Police
. Total
CObEITY Arrests
(B) Police
Capacity Capacity Loop :
+
Total
Criminals
Ler
Crime
Police Technology
Innovation
Page 15 of 21
SD Conference Version 1
Figure 4: Police Capacity Loop
After it was realized that the Police Capacity was being reduced in the model, structure was
added so that Police Capacity would remain constant regardless of productivity improvements
and the Desired Police Capacity’. The number of Total Criminals in this simulation run were
substantially reduced when the police sector was not allowed to reduce Police Capacity? (these
simulation runs are labeled Productivity 2). Police were able to make 50 percent more arrests, as
long as criminals existed and crime continued. Police productivity improvements resulted in
more arrests since the productivity improvements were not translated into a reduction in Police
Capacity. The results for this simulation are shown in Graph 7. This time there is a substantial
reduction in the number of Total Criminals in the system and an increase then a decrease for
Criminals on Probation and on Conditionally Discharged. Criminals in jail (Total Number of
Criminal in Jail) and on Parole increase slightly and the criminal in Prison show a slight oscilla-
tion (This can also be observed in Figure 7 under the Productivity 1 run), and the Total Number
of Criminals in Jail.
One of the things occurring in the model is that stress in the downstream part of the system is
increased. Productivity increases in the police sector result in more arrests and more criminals
are fed into the criminal justice system. This creates strains in the various parts or sectors of the
system, including the DA / courts, prisons, jails, parole and probation. Each of these sectors has
compensating feedback mechanisms whereby plea bargains, early release policies, etc. are put
into effect in order to reduce workload stress in that part of the system. Because of this effect,
the numbers of Criminals on Probation and Criminals Conditionally Discharged increase sub-
stantially compared to the base run.
8 This was done in the form of a switch that allowed us to turn off the ability to change the number of police. Police
became a constant.
° Experts from the New York State Division of Criminal Justice argued that this made sense as most police in New
York State are union members and that police unions are strong and that reduction in force due to increased Produc-
tivity or a reduction in overall crime would occur gradually as police retired and were not replaced.
Page 16 of 21
SD Conference Version 1
250,000
187,500
125,000 -~——
62,500 |
—— = z
0
0 12 24 36 48 60 2 84 96 108 120
Time (Month)
Prison Criminals
Total Number of Criminals in Jail Criminals
Criminals on Parole Criminals
Criminals on Probation : : <— Criminals
Conditional Discharge : Productivity 2 Es & S & 5 Criminals
Total Criminals : Productivity 2 Criminals
Graph 7: Increased Police Productivity with Police Capacity Held Constant
Graphs 8 and 9 capture the number of Criminals on Probation and Criminals Conditionally Dis-
charged for the base run, the police technology run when Police Capacity was allowed to vary,
and for the police technology run when Police Capacity was not allowed to vary. There are sub-
stantial differences in the simulation runs (labeled Productivity Runs | and 2) under the con-
straints described. Criminals on Probation increase before declining; Criminals Conditionally
Discharged increase and then begin to decline toward the end of the simulation run. The addi-
tional arrests generated by increased police productivity move more people into the system and
under some form of sanction. The majority of these criminals end up on probation (Criminals on
Probation) or are conditionally discharged (Criminals Conditionally Discharged) as they do not
have the resource constraints of jail and prisons. Furthermore, only criminals sentenced to prison
can end up on parole, so prison ends up as a buffer that limits the number of Criminals on Parole.
Although the increase in police productivity does not seem to lead to dramatic increases in the
number criminals in the system in terms of criminals in jail, prison, on probation or parole, the
decrease in crime in the model is dramatic (Graph 10). This dramatic decline is due to the num-
ber of new criminals entering the system on a monthly basis. This number is unknown as people
deciding to become criminals do not indicate their intentions. However, estimates can be made
since the number of people in the system in jail, prison, on probation and parole is known, as is
information about their average lengths of stays in each of these locations. The number of new
criminals entering the system in the model is initially 3,318. This number generates the number
of criminals in jail, prison, on probation and parole that are reasonable for New York State.
Other variables in the model, such as the probability or time it takes to capture a criminal, also
influence this number. The point is that this number is reasonable, but further analysis needs to
be conducted. However, it also points out that a careful selection of the variables examined to
Page 17 of 21
SD Conference Version 1
determine the success of productivity improvements is very important, as the number of people
in the system does not seem to increase by much, yet there is a dramatic fall in crime.
Criminals on Probation Conditional Discharge
Criminals on Probation : Productivity | #—#—2—+—2—=- Criminals Conditional Discharge : Productivity | —-—e—2—» == Criminals,
Graph 8: Criminals on Probation Reported Crime — Graph 9: Conditional Discharge
2M
Graph 10: Reported
Crime 15M he — e Compared
IM
D at al 500.000 land
Model 5 ILimita-
tions 0 12 24 36 48 60 2 84 96 108 120
Time (Month)
As previ- Reported Crime : Productivity 2 Crime jo us ly
, | _ Reported Crime : Productivity 1 Crime
stated, | Reported Crime : Base Run Crime System dy-
namics odel rep-
resents a theory about a particular problem. The theory is developed from discussions with ex-
perts in the criminal justice system in New York State, a review of the available data, and other
reports. Although relevant data is available and is collected from a variety of sources in New
York State, there were data limitations.
Since the focus of this study was to explore how criminals flow through the criminal justice sys-
tem and how technology could influence the flow of criminals, the lack of information about the
actual number of new criminals entering the system is unfortunate. This number is backed into
by using information available in other parts of the model. Furthermore, there were discrepan-
cies between what people said and the data. The experts interviewed on the investigation side all
claimed that the vast majority, about 99 percent, of all arrests lead to convictions. However,
when matching the model against actual data, it was discovered that police make approximately
514,000 arrests per year and that the court system adjudicates about half that number of cases. It
is unclear at this point whether people are arrested for more than one offense and the DA/ court
Page 18 of 21
SD Conference Version 1
system combines charges and adjudicates on only one of the charges or whether the DA / court
system declines to prosecute individuals for some reason. This will be explored further in future
analyses and the appropriate model structure will be added.
Discussion and Summary
As noted above, the computer simulation model for this study was developed based on a review
of the available data and the information gathered in discussions with experts responsible for dif-
ferent parts of the system. The model was initialized in equilibrium so that behavioral changes
resulting from changes in inputs could be understood and analyzed. The initial values were se-
lected to represent the average number of criminals in different parts of the New York State cor-
rections system between 1998 and 2004. All changes in inputs were implemented at month 20 so
that the initial equilibrium conditions could be observed and then the effects of changes could be
examined.
This paper examines what would happen in the New York’s criminal justice system if two pri-
mary changes were made. Those changes are an exogenous doubling of the number of new
criminals entering the system (new people becoming criminals) and a doubling of the productiv-
ity of police officers. These inputs were selected because they represent changes on the front or
upstream end of the criminal justice system, and the effects of these changes on the whole system
could be observed.
Doubling of new criminals entering the system does not result in a doubling of arrests made by
police officers. The number of arrests increases substantially as police in the model have the
ability to adjust their level of effort, which is driven by workload pressure. However, their abil-
ity to adjust their level of effort is limited. They cannot keep up with the number of new crimi-
nals entering the system, and thus crime increases as police are unable to clear all the new cases.
Although more criminals are arrested, the number of people in prison (Graph 4) initially in-
creases and then decreases. This occurs because an increase in criminals arrested causes an in-
crease in the workload of the DA/Judicial sector participants. In order to deal with the increased
workload, with fixed resources, plea bargains are increased. The immediate effect of more plea
bargains is to reduce the time it takes to deal with a case. As a result, the DA/Judicial sector
processes more cases with the given resources. However, in order to increase plea bargains, the
DA/Judicial sector must offer better deals to criminals which means shorter sentences and, on the
margins, a movement from incarceration to probation. This results in an initial increase in the
number of people in prison, but a decrease in the long-run as the sentence length was shortened
as part of the increased pressure for plea bargains.
The doubling of police productivity had a number of effects. In the model, police capacity was
structured as a goal setting activity in that the number of police was determined by the police
workload, which is the amount of crime divided by the number of police and this in turn was di-
vided by a normal workload. This resulted in an increase or decrease in the number of police
based on the amount of unsolved crime reported. When police productivity was increased, and
police were allowed to change, the number of police was reduced. The gain in police productiv-
Page 19 of 21
SD Conference Version 1
ity resulted in additional arrests, but this decreased the amount of unsolved crime reported and
the number of police was reduced. Improved productivity can result in one of two basic out-
comes. If the number of resources allocated to a given task can be kept constant, the productiv-
ity gains will result in more being accomplished or the benefits of the improved productivity can
be consumed by the affected area through reduced resources while maintaining the same level of
output. These possible outcomes point to the concept that in all areas of the criminal justice sys-
tem there may be a locally determined set of goals about acceptable productivity. An unan-
swered question (unasked during this project) is how would increased productivity change the
goals of acceptable productivity in the short and long-run? Will the savings from productivity
increases result in smaller staff, people accomplishing more, or increased activities not currently
undertaken?
The doubling of police productivity did not have a significant impact on the overall number of
people under some form of supervision in the model. This was due to the feedback effects of the
downstream entities compensating for the increase in their workload due to increased arrests.
Furthermore, increased police productivity substantially reduced the number of new criminals in
the system committing crime. In turn, this reduced police productivity as there was less crime to
solve. This finding raises additional questions. One of the unknowns in the system is how many
unidentified criminals (not recidivists) are entering the system for any given time period. Once
the police arrest more of these people than are entering, a reduction in the number of unsolved
reported crimes will result. Further work needs to be undertaken to determine whether the num-
ber of new criminals entering the system is realistic or a result of the analytic equilibrium.
Bibliography
Bard, J. F. (1977), “Criminal Justice Dynamics: A Planning Model,” Proceedings of the Winter
Simulation Conference 1, 258-268, Gaithersburg, Maryland, USA
Blumstein, Alfred, (2002). “Crime Modeling.” Operations Research, Vol. 50, No. 1, January-
February 2002; pp. 16-24.
Blumstein, Alfred and E. Grady (1982). “Prevalence and recidivism in index arrests: A feedback
model approach.” Law Society Review, 16(2), 265-290.
Blumstein, Alfred, Jacob Belkin, William Glass. 1971. JUSSIM, an interactive computer pro-
gram for analysis of criminal justice systems. Report of the Urban Systems Institute, School of
Urban and Public Affairs, Carnegie Mellon University, Pittsburgh, PA.
Cassidy, R. and R. Turner (1978). “Criminal Justice System Behavior.” Behavioral Science,
23:99-108.
Page 20 of 21
SD Conference Version 1
Cassidy, R. (1985). “Modelling a Criminal Justice System.” In Prediction in Criminology, ed.
Farrington and Tarling. State University of New York Press, Albany.
Cohen, J., et. al. (1973). “Implementation of the JUSSIM Model in a Criminal Justice Planning
Agency.” The Journal of Research in Crime and Delinquency, 10(3), pp. 117-131.
De Keijser, J. W. (2000). Punishment and purpose: From moral theory to punishment in action.
Amsterdam: Thela Thesis (diss. Leiden).
Maltz, Michael D. (1994). “Operations Research in Studying Crime and Justice: Its History and
Accomplishments” in Operations Research and the Public Sector, ed. S. M. Pollock, M.H. Both-
kopf and A. Barnett. North-Holland; Amsterdam.
McEwen, J.T., and R. Guynes (1990). Criminal Justice system simulation model (CJSSIM), Pa-
per presented at the Annual Meeting of the American Society of Criminology, Baltimore, MD.
McCould, P. (1992) The Role of Fiscal Policy in Producing Prison Population Dynamics: A
Trend Analysis and Dynamic Simulation of Felony Offender Processing in New York State. Dis-
sertation, University at Albany.
Morgan, P. (1985). “Modelling the Criminal Justice System.” Home Office Research and Plan-
ning Unit, Home Office, London.
Pullinger, H. (1985). “The Criminal Justice System Model: The Flow Model.” Home Office Re-
search and Planning Unit, Home Office, London.
Rice, S. (1984). “The Criminal Justice System Model: Magistrates Courts Sub-Model.” Home
Office Research and Planning Unit, Home Office, London.
Shaffer, W. A. (1976). Court Management and the Massachusetts Criminal Justice System. Ph.D.
Thesis, Massachusetts Institute of Technology.
Sterman, J. (2000). Business Dynamics: Systems Thinking and Modeling for a Complex World.
Boston, Irwin McGraw-Hill.
Page 21 of 21
