Systems analysis and planning for the criminal justice system

Systerris-Analysis and Pl’anning for the Criminal Justice System Alfred

Blumstein

School of Urban and Public Carnegie-Mellon University

Affairs,

Systems analysis, which has made important contributions in dealing with technological systems, also has an important role to play in dealing with a variety of social system problems that occur within the city. The problem of crime, and especially of the operation of the criminal justice system is one of the most fundamental of these problems, and one which in many respects typifies the operation of a wide class of social systems. This paper first presents a brief description of the criminal justice system, a discussion of some attempts to model that system, an enumeration of some of the difficulties inherent in developing and using such models, and then an indication of some approaches to dealing with those difficulties. This leads to a description of JUSSIM, an interactive computer model for criminal justice planning and system design. Finally, the author indicates how that model can be applied to other social systems.

T

HE

CRIMINAL

JUSTICE

SYSTEM

*Based on a paper presented at the Case Western Reserve Symposium on “Systems Approach and the City”.

1972

Figure

1. The

convicts him (and only a small portion of court cases involve a trial) the court may convict him and sentence him to corrections. In the corrections subsystem. a small

Criminal

Justice

System.

Society

I

Crimes I

I

r

--e-m-

--m----_

I

-1

I

(CJS),

comprising the agencies of police, courts, and corrections, is charged by society with trying to deal with the problems of crime through a process of arrest, adjudication, and ‘correction’. A general summaryofthissystem isshown in Figure 1. Here, a society gives rise to crimes, some of which get reported to the police, the normal entry point to the CJS. When the police solve the crime (and that happens

DECEMBER,

most frequently for murders. least frequently for property crimes like larceny and burglary) and arrest a suspect, he may then be brought to court. If the court

I

I

I

1

I , I

I

I I_ ---I--------~-~

I

I

I Corrections

I I

L

I I

I

I The

Criminal

Justice

SystemA --

61

portton 01. those convicted are-sent to an Institution for imprisonment. Most tind thetr sentences suspended, pay a fine. or arc placed on probatton in the custody of a typically overworked (caseloads of IOO~I 50 are common) probation officer Thus. we see a system charactertzed by a downstream flow from reported crimes to corrections. wtth the flow through any stage being a subset ofthe flow at a previous stage. And we also note that in virtually all cases (except those in which an individual dies while in the process). the individuals leave the control of the criminal justice system and return IO society. perhaps to commit further crimes (as ‘recidivists’). The traditional operation of the CJS has called for independent and autonomous operation of the individual subsystems, subject to independent judicial review. This process has been developed principally to provide checks and balances within the system. and to prevent any single ‘system manager’ from being able alone to decide the fate of an individual who flows through the process. An important contribution of the President’s Crime Commission’ in 1967 was the recognition that even though autonomy in handling individual cuscs was necessary, the operating policies of the separate parts of the system interacted intimately, and so must be examined and dealt with in an integrated way. The arrests by police represent the principal input into the courts. the convictions by the courts provide the input to corrections. and the failures of corrections provide the bulk of the subsequent input to the police. There are important interactions and tradeoffs among the parts of the system that should be considered in any examination of the total system. For example. in many jurisdictions. arrested offenders without bail money spend extended periods in detention cells awaiting trial. III many of these cases, it may be very wise to put additional resources into the courts to speed up the court processing of these cases. thereby reducing the time spent in pre-trial detention, and thereby saving more money than it would cost to speed the processing through the courts. This. of course. would be in addition to the other social values of speedier trial&avoiding undue imprisonment of the innocent. excessive detention of the guilty. and increasing deterrence of potential criminals. In view of the recognition of the need to deal with this total system, the Crime Commission’s recommendations were effectuated in the Omnibus Crime Control and Safe Streets Act of 1968. which led to the creation in each state of a ‘State Criminal Justice Planning Agency’ (SPA)

62

charged with planning for the total criminal justice system. Furthermore. these planntng agencies have the financial power to do more than merely plan. for they are the disbursers of the Federal funds appropriated under the Act. Thus. here was an extremely valuable opportunity to deal with the total CJS effectively and in an analytical way. The effort of these state planning agencies can be characterized much better as ‘grant administration’ than ‘planning’. This lack of planning was due in part at least to the Inadequacy of the methodology available to the SPA’s for undertaking such planning. This was a surprise and a disappointment. since there is in the literature a paper2 that appeared to have laid out the methodology for undertaking such planning. The models in that paper can be used to allocate costs by subsystem or by type of crime. to project future resource requirsments (i.e. for judges, cells, probation officers, etc.) based on a projection of future crimes or arrests. or to change the system parameters (branching ratios. unit costs. etc.) to see the impact on flows. costs. and resource requirements elsewhere in the system. All these considerations are important parts of the planning process. Nevertheless. they have not yet been extensively applied by the SPA’s

CHARACTERISTICS OF SOCIAL SYSTEMS To explore the reasons for limited use of such models and approaches, it is necessary to examine some of the basic characteristics of decisions. decision making, and decision makers in social systems like the criminal justice system. Measure of Effectiveness A fundamental aspect of all such systems is that their measures of effectiveness are complex vectors rather than the simple scalars much more characteristic of industrial systems. This problem of dealing with a multi-dimensional criterion is characteristic of most problems in the public sector. where government tries to perform some social good at some cost. In most cases, the social good (e.g. reduce crimes) is not easily measured in dollar terms (what is the ‘value’ of a rape avoided?). even though economists often try to force some such fit. Thus, there is at least one effectiveness component reflecting this social objective of the system. In addition, there are typically several cost components reflecting governmental costs and costs to various sectors of the society, and these various costs are not easily compared. nor are they easily related to

the etfecttveness measures. In these circumstances. \*~th at least t\\o components in the objective functton. all the mathemattcs of opttmtzatton becc>nles particularly difficult to apply. If one could specify all but one component as constraints. then he could nptimtre the rematntng one. But that tmpltes ;I zero-one loss function at the constratnt. which is rarely the case. Typically. a decision maker is much more concerned wtth e\plortng the rich collection of trade-otfs avatlable to him. There is iarsly a natural specihcatton of the constraint set that v+ill permtt the formulatton of a straightfor\r,trd oprimizatton problem. Not only is the quantiliable portton of the measure of ctTectivencss large and cornpIe\. but for 11~0s~ social svstems the /rtj/r-quantifiable considerations tend to he very significant in ;I dectston. It is perhaps because of this that many public decisions have been declared to be ‘trr‘tttonal’. The apparent irrationality of course dertvcs from inability of the observer to comprehend the rtch utilttv function of the decision maher. An oft&al uho weights some of the components that arc’ not readily quantified (e.g. protectton of privacy. prospects for rs-election. defusing a tense situation). may appear to be ‘Irrational on simple cost or even cost-elrectiveness grounds. but making such a judgment requires a much deeper knowledge of his goals than can possibly be subsumed in any model. Uncertainty of Causal Relationships Fundamental difficulties tn dealing with social systems arise from the widespread uncertainty about cause-and-effect relationships. This derives. in large part. because the systems arc based on a large (but not wJ’ large) number of human actions and dectsrons. and such decisions are typically characterized by caprictousness. adaptation to changed environment. and uniqueness to the sp:ctfics of any particular sttuatton. Available and reliable social science theory typically covers only a limited set of circumstances in the highly multi-variate envtronment. but the theoretical formulatton rarely provides adequate guidance to the boundaries of the space whet-t it is applicable. Furthermore, it can reasonably be expected that any specific case of concern is not specifically accounted for in any available theories. Thus, a decision maker in a social system who looks to social sctence theory (as opposed to the often valuable insights of social scientists) rarsly finds the theory adequate to answer his question. and so he must make his own intuitive guess of the consequences of actions he is considering. The problem of estimating effects is further compltcated by the fact that the

LONG

RANGE

PLANNING

behaviour of the participants in social systems changes over time in response to changes in public policy, changes in public values, changes in what is regarded as good professional practice. adaptation to differing capacity constraints, and simply new fads. Thus, even if there were a good theoretical construct that would describe, say, how the plea bargaining process changes with the introduction of additional public defenders, it is not clear that this response would be the same in separate jurisdictions or even that the response in one jurisdiction would be consistent over a period of several years. Here again, the decision maker is forced usually to rely on his own judgment and intuition. Technical Backgrounds It is rare for managers and planners of public systems to have a strong technical or analytical background. Much more typically, they reach their positions from backgrounds of politics, law, social work, or possibly one of the social sciences pursued no further than a master’s degree. Thus, in view of these backgrounds, elaborate mathematical models would be treated at best with a kind of distant and uncertain respect and at worst with a xenophobic hostility. These public officials can reasonably be expected to relate only to a model which they can comprehend and for which they understand and help develop the assumptions. They know very well all the ignorance about the operation of the system, and so are properly sceptical of any model that would presume to function as an oracle, dispensing conclusions and recommendations from a complex and mysterious ‘black box’. In this setting, then, it would appear that many of the classical models involving optimization or even batch-processed simulation might have only limited applicability. The fact, however, that decisions are necessary in these large social systems provides a need for such models if they can be properly coupled to the decision makers. The models must be comprehended by individuals who know very well the systems they run and operate but who have no extensive technical background. The many assumptions involved in the models must be understood by the decision maker, and he must participate in making them. Then, if the model provides him outputs, he can use it easily to gain insight into some of the quantifiable aspects of his decision. and then it could make a valuable contribution to his operations. THE JUSSIM MODEL It is with these considerations we developed the JUSSIM

DECEMBER,

1972

in mind that model of the

CJS.* This model is intended to permit criminal justice system managers and planners to test the downstream resource and cost implications associated with contemplated changes within their CJS. JUSSIM is an interactive computer programme, which represents the flow of ‘units’ (crimes, arrestees, cases, offenders. etc.) through the CJS and the application of resources to process these units as they flow through the system. The CJS is characterized by a flow diagram and the associated branching ratios p,j representing the proportion of the flow from stage i going to stagej. The resources are characterized by the stage or flow paths at which they are applied, an associated annual availability, and a cost per unit of workload. Each processing operation is characterized by a unit processing time. Furthermore, multiple parallel flow paths are each path having its own provided, parameter values, so that each crime type can be treated as a separate parallel channel of flow. Single Stage Analysis We can describe the basic notions of the JUSSIM model by considering a single stage, say the Circuit Court. If Ni (i= 1.2. . m) is the number of defendants charged with crime type i appearing in Circuit Court for disposition in one year, and if p,k is the proportion receiving disposition ii (k = 1,2. . ., q), and if 1,~ is the judge-days used to dispose of a case of type i by disposition k. the Nip;~ defendants receive disposition /i, and this requires (Ntpikltk) days of judge time. If judges work TI days per year. and if the cost of a judgeship is CI dollars per day, then 4 Z Nipirtir k=l

WI; =

days of judge

time, and

judges are required for dealing with crime type i, and the cost of these judgeships in Circuit Court is Cli = Then, required judgeship

Cl

cfl

Ni/riA!,>,ollars

per year.

the total number of judges for all crime types, NI, and the costs, CZ, are found from:

*The author is pleased to acknowledge the many contributions of Jack Belkin and the valuable programming of William Glass in this development.

111,

1?1<

N,=

t

hk;,

i=l

=

x i=l

Clr=cI

IfI,

q

Xh’,/u/,~ =

111,

q

1

z

i=l,4=I

t II

I,

PI

i= I

x=1

i=l

l?l, CI

Ii

111. A,p,rr,h=

LI i=

II

I,.

I

Similarly. we can calculate the number ol prosecutors required. Nz. and the prosecutorial cost. Cr. and the total Circuit Court cost is CI - CL Basic Operation The operation of the JUSSIM model begins with a ‘base case’ rstlecting the current operation of the system. All the data on the base case parameters must be collected and stored. The user 01 J USSI M-the criminal justice system planner himself-then creates a ‘test case’ h! making changes in any of the base cast parameters. The programme then reports to him the changes in flows. costs. worhloads. and resource requirements resulting from the changes he introduced. He thus uses the model as a very flexible design tool by making contemplated changes. rapidly getting an assessment of the effects of those changes. and then trying another change suggested by the feedback from the previous try. If the designer does not lihe the implications of a proposed change. then he can reject it immediately and try another. Operating the model to assess the consequences of a system change, a user necessarily has to make assumptions about the detailed micro-consequences of the change as they are reflected in changes in the system’s parameters. For those changes that initially app:ar attractive. he may then want to explore the assumed consequences more carefully. Several system planners can each explore the same system changes, each using his own assumptions. If a threshold of acceptance or rejection lies outside the range of calculated macro-consequences by this group, then acceptance or rejection is clear and further exploration is not necessary. If it lies within the range, then closer examination is required to assess the validity of the various assumptions made. In other terms, if E (possibly a vector) is a measure of effectiveness, and if E,, and E,; ( Eui 2 Erj) are acceptance and rejection values of E for a proposed system ,j (i.e. accept the change if E, 2 EC,,and reject it if Ej 5 E,i), and if planner i in evaluating change ,j makes assumptions leading to a value E;j, then accept if Eoj -< Mi” Eij, reject if Erj L M,?x E,,. and explore further if either of these inequalities is reversed. One of the virtues of this process is that it forces the planners into a debate on their

63

assumptions and micro-consequences rather than on the generalized goodness of a possible change. Thus, the model serves the same functio.n as any other model, it lowers the level of argument to more fundamental and empirically testable issues. The operation of the JUSSIM model can best be described by enumerating its inputs, the outputs, and the relationships between them. Inputs The basic inputs for JUSSIM are enumerated below: 1. A vector V of crime types considered (V = VI, 1’2, . . , v,). 2. A vector S of system stages (S = SI,SZ ,...) s,). 3. An n L n matrix P of branching ratios characterizing the proportion of flow from stage to stage s, (i = 1, 2, . . . n-l) t0 stage Sj (,j = 2, 3, . . ., n) which succeeds it in the CJS flow diagram. Here, p,j = 0 if there is no flow from stage s; to s,, and pij = 0 if i Lj. (,2,,1=

1)

4. A vector

R of resources (R = rl, r2, . . . , rk), and associated vectors A of the annual unit availability of the resources (A = al, UZ, . . , ax). and C of the cost per unit time (C = Cl, CL!,. . . , CL). 5. A X- x jr matrix T of the unit workloads, or times for processing a unit of flow at stage sj by resource r, represented by to. 6. A reference flow, (NI), typically the number of reported crimes or the number of arrests, that sets an absolute level of flow throughout the rest of the system when the branching ratios are specified. The branching ratios, the unit processing times, and the reference flow are all functions of crime type, and so each element in these is a vector with a component for each crime type. outputs The outputs of the JUSSIM model are presented to the user in whatever order and organization he specifies. The potential outputs include the following variables: 1. Flow through each processing stage, a vector N (N = NI, Nz, . . .,Nn). D, 2. Costs at each stage, a vector (D, = D,,,. Sr2,. .,D,,). Costs can be given for any aggregation of stages into specified subsystems. including a complete aggregation into single total system. 3. Resource costs, a vector D, (D, =

64

the D,,, SG, ., DA) indicating costs associated with each of the k resources. 4. Resource workloads. a vector W (W = IV,, U’Z, . . U’A) of the man-hours per year of workload imposed on each of the resources. required. a vector Q 5. Resources (Q = Ql, Qz. . . . QA) indicating the numbers of each of the specified resources. that would be required to handle the workload. All of these output variables are functions of crime type, and so each of the components of the output vector can be presented as a vector with a component for each crime type, or as a scalar summed over the crime types. Basic Relationships Assuming knowledge of all the input parameters, one can then calculate the output variables by the following relationships. Knowing the basic input flow (N1) as a reference level and the branching ratio iteratively matrix (P), one can then calculate the flow at each stage by: N, = X pij Ni i<.j

and create the flow vector: N= (NI, Nz, . . ., N,,). The i,, stage processing costs are:

Dri= Ni z

Ck

tik

(i

=

I,2 9 . . ., 4

or, in vector notation, D, = (C T) * N where the vector operation* is the component-by-component product, i.e. if C = A *B, then c; = aibi. The subsystem costs are simply the sum over the constituentstage costs. The resource costs for the use or resource i-k are:

DA =

ok

t

Nitik

or: D, = (N T’) * C. The workload on resource

or, as a vector: W = NT’. The number of resource year is then: Q:

=

ut

:

Ni

n is:

rk required

per

tik

or Q = (N T’) f A where the vector operation + is the component-by-component ratio, i.e. if K = A + B, then C; = ai/bi for bi # 0. Operation of a Run In the operation of a JUSSIM run, the user’s basic role is to create a ‘test case’ to compare with a ‘base case’ already stored. Initially, the test case is made identical to

the base case. so the user enters only changes to the base case. The user. setting at a terminal. is asked a sequence of questions about what changes he wants to make. Each of these questions is an entry gate to a ‘phase’. A separate phase is provided for changing each of the following parameters:

1. Branching ratios (P): 9 Unit workloads (T): _. 3. Annual unit resource availability (A): 4. Resource unit costs (C). Once a phase is entered. further detailed questions permit the user to specify precisely which parameters he wants to change and the crime type(s) for wjhich he wants to make the change. All the questions are in clear language. and the answers regarding stage numbers. resource numbers, crime-type codes, and other codes are based on a code used in creating a base-case data tile. The programme then displays the base-case value for the parameters the user identified and asks him to type in the new, or test-case values. In dealing with the multiple parallel channels of flow for the nz crime groups,* most users. at least initially, do not want to sit through the complete detailed output for each crtme type. In a separate phase, JUSSIM permits them to choose one of a number of standard complete partitions of the crime types (e.g. into felonies and misdemeanors. Part 1 and Part 11 crimes, etc.), or to specify his own crime groupings. Thus, he may want the complete details on one or a few specitic crime types, and aggregate the remainder. The complete crime-type vector and the standard groupings are specified in the data file. In another phase, the user specifies the output tables to be displayed. The output tables present calculated results on flows, costs, workloads, and resource requirements for the base case, the test case, and the absolute and percentage change in going to the test case. These results can be presented by crime group or summed over all crime groups. The choices to be made available are specified in the data file, reflecting various aggregations of stages into subsystems as well as resource partitions or aggregations. Presumably, in the early stages of an exploration, he will want to conserve time and will examine results only for the total system or for a few critical subsystems. At the end of an exploration, the user is more likely to want more detail. The user is also asked if he wants to specify a reference flow. This is usually the number of reported crimes or arrests, although specification of any flow variable

“The

data

FBI’s Uniform Crime Reporfs organizes into 29 separate crrme types.

LONG

RANGE

PLANNING

its

TEST CASE I RESULTS SUMMARY OF RESULTS FOR POLICE

RUN CJS WELCOME TO VERSION 1 OF THE CJS MODEL ENTER PHASE I-SPEC.IFlCATlON OF CRIME GROUPING ENTER CODE NUMBER OF DESIRED GROUPING -2 -DO

_V

YOU

INTER ENTER

WISH

PHASE STAGE

TO

SPECIFY

NEW

~--SPECIFICATION NUMBER, CRIME

TEST CASE RATIOS.. .

BRANCHING

OF BRANCHING GROUP

RATIOS

FLOWS PATROLMAN DETECTIVE

GROUP

SUMMARY

GROUP

GROUP

SUMMARY

ENTER NEW VALUE -2.00 ENTER INDEX NUMBER, CRIME GROUP _* DO YOU WISH TO SPECIFY NEW TIME RESOURCES.. -N TO

SPECIFY

DO YOU

WISH

TO SPECIFY

NEW

SUMMARY

.

DESIRED

OUTPUT..

.

-Y ENTER PHASE ‘I-SPECIFICATION OF DESIRED OUTPUT ENTER SUBSYSTEM NUMBERS OF DESIRED TABLES 3, 4, 15 ENTER WORKLOAD INDEXES OF DESIRED TABLES

-DO YOU -N DO YOU CRIMES.. -N DO YOU -N

WANT WISH

WISH

A BREAKDOWN TO

TO

SPECIFY

RE-DO

BY CRIME

NEW

ANY

LEVELS

GROUP..

1972

88522.3 -17542.1 395768.3 71176.4

-16.5 21.9

52.1 i’32.8 284.9

-10.3 41.9 31 .5

-16-5 21 ,9 12.5

77691.8 77691.8

4451.9 4451.9

6.1 6.1

COURT CHANGE 36.3 97 2 133.5

4.8 15.1 9.6

59805.0 1650.6

62683.3 1900.0

2878.3 249.4

4.8 15.1

7.3

8.4

1 .I

15.1

59805 0 59805.0

62683.3 62683.3

2878.3 2878.3

4.8 4.8

FOR SUPERIOR TEST

COURT

CHANGE

O/O CHANGE

945.4 538.2 1483.5

1235.3 759.3 1994.6

289.9 221.2 511.1

30.7 41 .l 34.4

3468.0 748.6

4531.4 1056.2

1063.4 307.7

30.7 41 .l

5.2

7.3

2.1

41 .l

3468.0 3468.0

4531.4 4531.4

1063.4 1063.4

30.7 30.7

FOR TOTAL

SYSTEM

OF RESULTS

TEST

CHANGE

COSTS IN THOUSANDS POLICE 4032.7 4651.8 COURT 2880.5 3525.1 CORECTIONS 6936.6 8945.3 TOTAL 13849.9 17122.2 RESOURCE REQUIREMENTS POLICE 253.3 284.9 COURT 12.5 15.7 TOTAL 265.8 300.6 FLOWS POLICE 73239.9 77691.8 COURT 63273.0 67214.7 CORECTIONS 43714.6 46715.2 DO YOU WISH TO RERUN THE PROGRAM.. . -Y

PHASES..

DO YOU WANT BASE CASE.. -Y Run with the JIJSSIM

O/O CHANGE

789.8 740.7 1530.5

OF REPORTED

Figure 2. An Illustrative

DECEMBER,

-16.5 21.9 15.4

TEST

BASE

COSTS..

-114.0 733.1 619.1

575.4 4076.4 4651.8

753.5 643.5 1397.0

OF RESULTS

COSTS IN THOUSANDS PROSECUTOR JUDGE TOTAL WORKLOADS PROSECUTOR CASE JUDGE DAYS RESOURCE REQUIREMENTS JUDGE FLOWS PROSECUTOR JUDGE

O/O CHANGE

FOR LOWER

BASE

-N

-1,

OF RESULTS

COSTS IN THOUSANDS PROSECUTOR JUDGE TOTAL WORKLOADS PROSECUTOR CASE JUDGE DAYS RESOURCE REQUIREMENTS JUDGE FLOWS PROSECUTOR JUDGE

STAGE 17-S FOR DISP CRIME GROUP 1 CURRENT VALUES ARE.. . 17.8 NOLLED G PLEA 72.1 BENCH 3.9 JURY 6.2 ENTER NEW VALUES -15, 74.1, 4.1 ENTER STAGE NUMBER, CRIME GROUP _* DO YOU WISH TO SPECIFY NEW WORKLOADS.. . -Y ENTER PH 4SE 4-SPECIFICATION OF WORKLOADS ENTER INDEX NU:dBER, CRIME GROUP -2.2 CRIME GROUP 2 CURRENT VALUE FOR PATROL ARREST IS 1.00 HRS. ENTER NEW VALUE -0.7 ENTER INDEX NUMBER, CRIME GPOUP -3. 1 PRIME GROUP I CURRENT VALUE FOR DETECT REPORT IS 1.54 HRS.

WISH

73239.9 73239.9

BASE

-17,l

DO YOU

689.4 3343.3 4032.7

WORKLOADS PATROLMAN HOURS 106064.4 DETECTIVE HOURS 324591.9 RESOURCE REQUIREMENTS PATROLMAN 62.4 DETECTIVE 190.9 TOTAL 253.3

-1 ,l STAGE l-REPORTED CRIME GROUP 1 CURRENT VALUES ARE.. . ARRESTED 22.6 NO ARREST 77.2 ENTER NEW VALUES -30, 70 ENTER STAGE NUMBER. CRIME -5, 1 STAGE 5-L ARRAIGNED CRIME GROUP 1 CURRENT VALUES ARE. . . JUVENILE CT. 10.7 BIND OVER 21.2 L FOR DISP 66.2 ENTER NEW VALUk -10.7, 25, 64.4 ENTER STAGE NUMBER, CRIME -5,1 STAGE 6-L FOR DISP CRIME GROUP 1 CURRENT VALUES ARE.. . NOLLED 56.9 G PLEA 41 .a BENCH 1.3 JURY 0.0 VIOL. BUREAU 0.0 ENTER NEW VALUES -50, 48.5, i .5 ENTER STAGE NUMBER, CRIME

COSTS IN THOUSANDS PATROLMAN DETECTIVE TOTAL

CHANGE

TEST

BASE

YOUR TEST CASE

TO BECOME

O/O CHANGE

619.1 644.6 2008.7 3272.4

15.4 22.4 29.0 23.6

31.5 3.2 34.8

12.5 25.9 13.1

4451.9 3941.7 3000.6

6.1 6.2 6.9

TEST CASE THE NEW

II

Model.

65

anywhere in the system completely specifies all the other flow variables. At the end of a run. the user is asked if he wishes to re-run the model. In addition to using this to explore a new issue, he will do this iteratively if he is dissatisfied with the implications of some of the assumptions he has made and would like to reconsider some of those. In a re-run, he is given considerable flexibility in respecifying his new base case (e.g. calling a data file, using his latest test case), and then begins again to create a new test case. An

Illustrative

and

in Values

1. Changes

Program improvement

Additional

System

detectives

Extension

The most fundamental limitation of the JUSSIM model is that it can now provide only downstream consequences of an upstream change. As a result. the output measures are all in terms of flows, resources and costs. However important these considerations may be they are unquestionably secondary to an ultimate concern over crimes as an output measure. That model would have to deal with feedback and recidivism. Unfortunately, however, so little is known quantitatively of the impact of almost any CJS actions on crimes. In operating the feedback model, it is necessary to make important distinctions between the recidivist arrests and the virgin a distinction not easily nor arrests, normally made by most users. We are now extending the model in these directions to provide this added capability.

of System

Parameters

Parameters

Changed

* Detective investigation per report

for more intensive investigation

time

for Test

Case.

Base Case Values (hours)

lest Case Values (hours)

1.54

2.0

0.228

0.300

0.569 0.418 0.013 0.212

0.500 0.485 0.105 0,250

0.178 0.721 0,039 0.062

0,150 0.147 0.041 0 062

1 .oo

0.70

(Part I)

1 Proportion

of arrests per report (Part I) _ Lower Court disposition type . -nolle * -guilty plea * -bench trial * -bindovers Superior Court disposition type * -nolle * -guilty plea 1-bench trial --jury trial

Run

To illustrate the operation of JUSSIM, we indicate in Table I two potential improvements in police operations, more intensive detective investigation of Part I crimes and the use of summons for street arrests of Part II crimes. The user must identify the system parameters affected, and then specify the specific parametric changes. These are shown in Table I. Figure 2 presents a reproduction of the computer terminal printout developed in running this case. In the figure, the user inputs to the programme are indicated by a mark in the left-hand margin. The total computer charges for performing this run at prime-time commercial rates was under S4.00. Limitation

Table

Use of summons street arrests An increase

in the

for

- Patrolman (Part II)

parameter

time per arrest

is indicated

Implementation

The response to the interactive JUSSIM model was, as expected, far more positive than to a batch-process model. In teaching criminal justice planners how to use the model, it was impressive to note how quickly and easily they learned to sit at the terminal and respond to the programme’s interrogations, to translate their project ideas into judgments about model parameter changes, and to operate the model iteratively as a design tool. As a result of the introduction to the model in September 1970, two states and two cities have already organized data collection efforts to describe their systems as a base case for the JUSSIM model. and plan to use a version of the model for their own planning. SUMMARY

This paper has tried to emphasize the virtues of an interactive comouter model as a planning and design tool that could be very valuable to decision makers involved with social systems. Their direct involvement in making the necessary assumptions assures both that they become familiar

by

*

and

a decrease

is indicated

by

,

with the model and its limitations. and that the analysis draws on their judgment and experience in the many social-system areas where no scientific knowledge exists. We have described here one such model for the criminal justice system. and have indicated how it has been used. It should be clear to readers concerned with other social systems that there IS nothing in the basic operation of JUSSIM that confines it to the CJS. All of the inputs and outputs are described generally. and so the model can be used for any system describable by a non-feedback flow diagram where concern is over issues of flows, costs. and resources. Hopefully. the feedback limitation will also be removed shortly. n

REFERENCES (1) The Challenge of Crime

(2)

in a Free Society. President’s Commisslon on Law Enforcement and the Admlnistration of Justice. Government Printing Office, Washington, D.C. (1967). A. Blumsteln and Rtchard C. Larson, Models of a Total Criminal JustIce System, OperaLions Research, Vol. 17, No. 2 (1969).

LONG

RANGE

PLANNING