.. m . . f s. .:.. v 2‘ .,

A k 4...... , mg...“
n mm. P
ma,

3.

.11.!

4a . ‘

.x. .

a 3.29....

z. tar:
K

:3.

i :1 3?-

.nzuwwku. . .
.HLP. .4:

.
db..:......ré?..... . is
.. . . a...
. J‘.9..n_hnr......t .....i:...Jtnw
)IIQV .ttltliiii ‘- .' i‘
. E—g‘l :Iw. I ,g
sis}. 51$
, -..
yawn;
“mi... . . r
i . . ;
05$ 3 at “(.1
tit-Iz,‘ ‘uluil. if it: I
isnrfiu unfluéxl‘illl: . .uhf. {rah-uh. '5‘
13.35:... 91.: = 3“ 1

an
....
LY: .fifptfipfati
I, “InitgivuZI5f.
“.2 ..
3.7!". f. n .
. a «Fawn 3A firs“. fink aIsJuu.
uykdvye Sufi. Ennviufl an"...
. sfi«.u~eflu§- In
3.!- .v

 

 

 

:. Eh:

.‘ u:
1.!
I!

 

 

"£59450... u: H 62.35».

zilltn
Vslfbs".

9. t. l. .
v- ion-5?,x‘rq-97.
.5331.

 

. .xvo 9 Ant“. . t..L,HI.‘V.........r. 11 . v.5
1.1. 4 F

LIBRARY
Michigan State
University

This is to certify that the

dissertation entitled

UNDERSTANDING BUDGETARY PRIORITIES:
APPROPRIATIONS FOR CORRECTIONS AND HIGHER EDUCATION

presented by

Me t: thew Kle iman

has been accepted towards fulfillment
of the requirements for

Ph.D. degree in Political Science

 

 

 

 

Date H 0

MS U is an Affirmative Action/Equal Opportunity Institution 0-12771

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJCIRC/DateDuepSEw. 1 5

UNDERSTANDING BUDGETARY PRIORITIES:
APPROPRIATIONS FOR CORRECTIONS AND HIGHER EDUCATION

By

Matthew Kleiman

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Department of Political Science

200]

ABSTRACT

UNDERSTANDING BUDGETARY PRIORITIES:
APPROPRIATIONS FOR CORRECTIONS AND HIGHER EDUCATION

By

Matthew Kleiman

Budgets themselves represent the endgame of a struggle for scarce resources that reflect
government priorities. This dissertation seeks to identify the factors that are responsible
for determining the level of appropriations for discretionary budgetary categories that are
in competition with one another. To accomplish this the impact of endogenous and
exogenous factors, the size of the budgetary pie, competition from other budgetary
categories, and specific public policies are examined. More specifically the focus is on
appropriations for corrections and higher education in California, Michigan, and Virginia
between Fiscal Year’s 1970-1999. Considerable attention is devoted to outlining a
suitable methodology (281.8 and SUR) for addressing whether there is a tradeoff between
these two budgetary categories. In addition, recent sentencing policy reforms and the
resultant shift away from indeterminate sentencing and rehabilitation towards determinate
sentencing is addressed. It is found that the confluence of exogenous factors and the
implementation of specific sentencing policies (e. g., truth in sentencing, determinate
sentencing, mandatory minimums, and repeat offender legislation) have been responsible
for the rapid increase in appropriations for corrections. In sum, this analysis contributes
to the theoretical understanding of the budgetary process and to the understanding of how
the implementation of sentencing policies have led to increased pressures on state

budgets.

Copyright by
MATTHEW KLEIMAN
2001

To my Mom and Dad: my biggest supporters, without whom I would have never made it
this far.

iv

ACKNOWLEDGEMENTS

Writing this dissertation truly has been a war of attrition. At times I was wholly
convinced that finishing was an impossibility and quitting graduate school was the only
alternative. Fortunately I have several people in my life that never gave up on me and
kept me on the path towards completion. Throughout graduate school these people gave
me guidance, advice, support, encouragement, and endless love. I owe an eternal debt of
gratitude to these individuals and am sure this project would have never been completed
without them.

Thanks must first go to my parents who have been there for me every step of my
life with words of wisdom, a timely joke to put things in perspective, and endless love
and fi'iendship. They have always found a way to convince me that things are actually
not as bad as I think they are and to keep persevering. For this I am grateful. They are
truly a guiding force in my life and most of all have taught me to work hard and play
even harder.

Thanks also go to Liz Ransom who, unlike me, knew that I could do it all along.
Her perpetual optimism and unwavering love and friendship helped me get through many
hard times and fully enjoy the good times. She deserves credit for reading over countless
drafts, listening to me moan and groan, and being there for me in everyway possible. She
truly is the greatest.

Further thanks are extended to Chuck Ostrom, my advisor, mentor, friend, and
now colleague. Throughout this process he has read countless drafts, answered a never-
ending stream of questions, and provided direction and insight without ever hesitating to

assist me. Chuck has taught me how to be a scholar, a policy analyst, and how to view

the world. He has been the quintessential mentor and it is truly an honor that I can call
myself his student. Thanks is also extended to my other committee members: Mike
Mintrom who taught me what it means to be a professional policy analyst and provided
very insightful comments on earlier drafts; Jim Granato who is one of the good guys; and
Brian Silver.

Finally, a special thanks must go out to all those who have given me friendship
and emotional and intellectual support through this long process. This includes: Philip
Alderfer, Chris Bonneau, Rhonda Burns, Chris Butler, Brad Carl, Mike Cichello, Mike
Dann, John U. Davis, Kristin Dawley, Elaine Eschtruth, Charles Finocchiaro, Ada
Finifter, Sean Fogarty, Matt Greevy, John Halpin, Tom Hammond, Mark and Sheralee
Hurwitz, Julie Kleiman, Andre Lavallee, Bryan Marshall, Brian Ostrom, Brandon Prins,
Jeff Riedinger, Kimberly Smiddy, Mica Stark, Scott Szymendera, Merideth Trahan, and
Nic van de Walle. Also thanks must be extended to the many individuals in California,

Michigan, and Virginia who helped in the difficult data collection process.

Four strong winds that blow lonely
Seven seas that run high
All those things that don’t change
Come what may
If the good times are all gone
Then I’m bound for moving on
I’ll look for you if I‘m ever back this way (Neil Young, 1978)

vi

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................. ix
LIST OF FIGURES ........................................................................................................... xi
CHAPTER 1 ........................................................................................................................ 1
Expansion of the State Prison System ................................................................................. 1
An Introduction to Sentencing Reforms .............................................................................. 4
A Changing Social Policy Trajectory .................................................................................. 6
Overview of Dissertation ..................................................................................................... 8
CHAPTER 2 ...................................................................................................................... l l
Tradeoffs ............................................................................................................................ 12
Budgetary Decisionmaking ................................................................................................ 16
Exogenous Factors ............................................................................................................. 17
Policy Changes in Corrections ........................................................................................... 20
Stages of the Criminal Justice System ............................................................................... 26
Overview of Study ............................................................................................................. 32
CHAPTER 3 ...................................................................................................................... 35
Theoretical Hypotheses ...................................................................................................... 35
Budgetary Tradeoffs .............................................................................................. 36
Tradeoff Hypothesis ................................................................................... 37
lncrementalism ....................................................................................................... 37
lncrementalism Hypothesis ........................................................................ 37

Exogenous Factors ................................................................................................. 39
Exogenous Hypothesis ............................................................................... 40

Budgetary Pie ......................................................................................................... 4O
Budgetary Pie Hypothesis .......................................................................... 42

Public Policies ........................................................................................................ 42

Policy Hypothesis ...................................................................................... 42

Model and Statistical Hypotheses ...................................................................................... 43
Data Source ........................................................................................................................ 44
Operationalization and Data Sources ................................................................................. 46
Appropriations ....................................................................................................... 48
Exogenous Factors ................................................................................................. 52
Violent Crime Rate .................................................................................... 52

Enrollment .................................................................................................. 53

Cohort ........................................................................................................ 53

Political Ideology ....................................................................................... 53
Unemployment ........................................................................................... 54

Overview of Budgeting Patterns In California, Michigan, and Virginia ........................... 55
California — Appropriations ................................................................................... 55

vii

Michigan — Appropriations .................................................................................... 57

Virginia — Appropriations ...................................................................................... 60
Estimation Strategy ............................................................................................................ 63
Time Series Issues - Lags .................................................................................................. 68
Time Series Issues — Stationarity ....................................................................................... 70
Estimation Strategy — Overview ........................................................................................ 75
Anticipated Conclusions .................................................................................................... 82
CHAPTER 4 ...................................................................................................................... 84
California — Policies ........................................................................................................... 84
California — Prison Population ........................................................................................... 89
California — Independent Variables — Governors .............................................................. 90
California — Independent Variables — Unemployment Rate .............................................. 90
California — Independent Variables — Violent Crime Rate ................................................ 91
California — Independent Variables —- Cohort .................................................................... 92
California — Independent Variables - Enrollment ............................................................. 93
California — Correlation Matrix ......................................................................................... 94
California - Two Stage Least Squares ............................................................................... 96
California — Seemingly Unrelated Regressions ............................................................... 100
CHAPTER 5 .................................................................................................................... 1 11
Michigan — Policies ......................................................................................................... l 1 1
Michigan — Prison Population .......................................................................................... l 15
Michigan — Independent Variables — Governors ............................................................. 1 16
Michigan — Independent Variables — Unemployment Rate ............................................. 1 17
Michigan — Independent Variables — Violent Crime Rate ............................................... l 18
Michigan — Independent Variables — Cohort ................................................................... I 19
Michigan - Independent Variables — Enrollment ............................................................ 120
Michigan - Correlation Matrix ........................................................................................ 121
Michigan — Two Stage Least Squares .............................................................................. 123
Michigan — Seemingly Unrelated Regressions ................................................................ 128
CHAPTER 6 .................................................................................................................... 135
Virginia — Policies ........................................................................................................... 135
Virginia — Prison Population ............................................................................................ 140
Virginia — Independent Variables — Governors ............................................................... 141
Virginia — Independent Variables -— Unemployment Rate ............................................... 142
Virginia - Independent Variables - Violent Crime Rate ................................................. 143
Virginia — Independent Variables — Cohort ..................................................................... 144
Virginia — Independent Variables - Enrollment .............................................................. 145
Virginia — Correlation Matrix .......................................................................................... 146
Virginia - Two Stage Least Squares ................................................................................ 148
Virginia — Seemingly Unrelated Regressions .................................................................. 153
CHAPTER 7 .................................................................................................................... 161
Policy Windows and Changes in Correction’s Policy ..................................................... 164

viii

Policies as a Pendulum ..................................................................................................... 168
Intermediate Sanctions ..................................................................................................... 170

REFERENCES ................................................................................................................ 173

ix

LIST OF TABLES

Table 3.1: Statistical Hypotheses ....................................................................................... 44
Table 3.2: Operationalization ............................................................................................. 47
Table 3.3: Desideratum of Appropriations ....................................................................... 49
Table 3.4: Appropriations Data — Title, Source, and Notes About the Data ..................... 51
Table 3.5: Augmented Dickey-Fuller Test: Unemployment Rate in California ................ 74
Table 4.1: California Determinate Sentencing terms (SB 42) ........................................... 86
Table 4.2: Correlation Matrix for Variables included in the Budget Model,

California .......................................................................................................................... 95
Table 4.3: Unit Root Tests (ADF) ..................................................................................... 97
Table 4.4: Two Stage Least Squares — Levels and Differenced, California ...................... 98
Table 4.5: Seemingly Unrelated Regression, Corrections in California .......................... 101
Table 4.6: Seemingly Unrelated Regression, Higher Education in California ................ 102
Table 4.7: Seemingly Unrelated Results (Breusch-Pagan Test), California .................... 104
Table 4.8: Triangulation of Estimates in California ........................................................ 105
Table 4.9: Predicted Values for Appropriations for Corrections (in billions) ................. l 10
Table 5.]: Correlation Matrix for Variables included in the Budget Model,

Michigan .......................................................................................................................... 122
Table 5.2: Unit Root Tests (ADF) ................................................................................... 123
Table 5.3: Two Stage Least Squares (robust) — Levels and Differenced, Michigan ....... 126
Table 5.4: Seemingly Unrelated Regression (differences), Michigan ............................. 129
Table 5.5: Seemingly Unrelated Results (Breusch-Pagan Test), Michigan ..................... 129
Table 5.6: Triangulation of Estimates in Michigan ......................................................... 130
Table 5.7: Predicted Values for Appropriations for Corrections (in millions) ................ 134

Table 6.1: 1975 Virginia Sentencing Schema .................................................................. 136

Table 6.2: Correlation Matrix for Variables included in the Budget Model,

Virginia ............................................................................................................................ 147
Table 6.3: Unit Root Tests (ADF) ................................................................................... 149
Table 6.4: Two Stage Least Squares (robust) — Levels and Differenced, Virginia ......... 151
Table 6.5: Seemingly Unrelated Regression, Virginia .................................................... 154
Table 6.6: Seemingly Unrelated Results (Breusch-Pagan Test), Virginia ....................... 155
Table 6.7: Triangulation of Estimates in Virginia ........................................................... 156
Table 6.8: Predicted Values for Appropriations for Corrections (in millions) ................ 160
Table 7.1: Summary of State Reforms (dates implemented) ........................................... I64

xi

LIST OF FIGURES

Figure 1.1: State Prison Population, 1930-1999 .................................................................. 2
Figure 2.1: Budget Line and the Tradeoff Between Goods ............................................... 14
Figure 2.2: Stages of the Criminal Justice Process and Policies that Impact State

Incarceration Rates ............................................................................................................. 27
Figure 2.3: Supply and Demand for Prison Beds .............................................................. 30
Figure 2.4: Supply and Demand for Higher Education ..................................................... 32
Figure 2.5: The Budgetary Process .................................................................................... 34
Figure 3.1: Budgets as an expanding-sum game ............................................................... 41

Figure 3.2: State of California: Corrections and Higher Education Appropriations

(GF):FY 1969-1999 ........................................................................................................... 56
Figure 3.3: State of California: Percent of Appropriations from Total State
Appropriations (GF): FY 1969-1999 ................................................................................. 57
Figure 3.4: State of Michigan: Corrections .md Higher Education Appropriations
(GF/GP): FY l969-2000 .................................................................................................... 58
Figure 3.5: State of Michigan: Percent of Appropriations from Total State

Appropriations (GF/GP): FY 1969-2000 ........................................................................... 59
Figure 3.6: State of Virginia: Corrections and Higher Education Appropriations

(GF): FY 1971-1999 .......................................................................................................... 61
Figure 3.7: State of Virginia: Percent of Appropriations from Total State

Appropriations (GF): FY 1971-1999 ................................................................................. 62
Figure 3.8: An Illustration of Time and the Budgetary Cycle, Appropriations

and Expenditures ................................................................................................................ 69
Figure 3.9: Illustration of Stationarity via Differencing: Enrollment in California ........... 74
Figure 3.10: Overview of Estimation Strategy .................................................................. 76
Figure 4.1: Number of Prisoners Under State Jurisdiction in California: 1971-1997 ....... 89
Figure 4.2: State of California: Unemployment Rate, FY 1970-1999 ............................... 91

xii

Figure 4.3: State of California: Violent Crime Rate (Part I offenses known to

police per 100,000 inhabitants), FY 1968-1999 ................................................................ 92
Figure 4.4: State of California: 18-24 year old cohort, FY 1968-1999 ............................. 93
Figure 4.5: State of California: Enrollment in Higher Education, FY 1970-1999 ............ 94
Figure 5.1: Number of Prisoners Under State Jurisdiction in Michigan:

1971-1997 ........................................................................................................................ 116
Figure 5.2: State of Michigan: Unemployment Rate, FY 1970-1999 .............................. 1 18
Figure 5.3: State of Michigan: Violent Crime Rate (Part I offenses known to

police per 100,000 inhabitants), FY 1970-1999 .............................................................. I 19
Figure 5.4: State of Michigan: 18-24 year old cohort, FY 1970-1999 ............................ 120
Figure 5.5: State of Michigan: Enrollment in Higher Education, FY 1970-1999 ........... 121

Figure 6.1 : Number of Prisoners Under State Jurisdiction in Virginia: 1971-1997 ........ 14]

Figure 6.2: State of Virginia: Unemployment Rate, FY 1970-1999 ................................ 143
Figure 6.3: State of Virginia: Violent Crime Rate (Part I offenses known to

police per 100,000 inhabitants), FY 1970-1999 .............................................................. 144
Figure 6.4: State of Virginia: 18-24 year old cohort, FY 1970-1999 .............................. 145
Figure 6.5: State of Virginia: Enrollment in Higher Education, FY 1970-1999 ............. 146

xiii

CHAPTER 1

State legislators and governors are engaged in a continuous process of defining
what priorities state governments should focus on and address. States may choose to
focus on or emphasize issues relating to criminal justice, education, health care,
transportation, etc... This process of setting priorities manifests itself in the realm of
budgeting where certain programs and departments are given greater state resources to
address what are deemed as problems. Given a limited set of resources, states cannot
address all problems; therefore, they must choose which areas will be given priority.
This dissertation will focus on how budgetary appropriations are determined and more
specifically examine the recent increases in budgetary appropriations for corrections and
the potential tradeoff that might exist between this budgetary category and higher

education.

Expansion of the State Prison System

In recent years states have seen a radical grth in the size of the prison
population and in the incarceration rate. In 1930 the state prison population of the United
States stood at 117,000 and grew at a relatively constant rate through 1970, when the
prison population stood at roughly 175,000 (See Figure 1.1). However, between 1970
and the turn of the century the prison population grew rapidly and swelled to over 1.1
million prisoners (Langan, Fundis, and Schneider 1988; Beck and Karberg 2001).1 This

rapid growth is mirrored in the incarceration rate, which is the number of inmates in jail

 

' The Postsecondary Education Opportunity (1997) notes that between 1925 and 1973 the prison population
of the US. increased on average 2,333 prisoners per year, while between 1973 and 1995 there was an
average increase of 48,900 prisoners per year.

and prison per 100,000 US. residents. In 1930 the incarceration rate stood at 146 and
rose to 167 in 1970. This rate increased to 461 in 1990 and 703 in 2000 (Ziedenberg and
Schiraldi 2000). Despite the large aggregate increase, the incarceration rate varies
considerably across the different states with Texas having the highest rate of 1,014
inmates per 100,000 residents, California a rate of 721, Virginia 713, Michigan 628, and
Vermont the lowest among the states at 203 in midyear 1999. Additionally, by years end
2000, the United States had almost 2 million people incarcerated in its prisons and jails

and had the highest incarceration rate of any nation in the world (Sentencing Project

1400

 

1200 -

1000 ~-

800 -

600 -

400 4»

Prison Population (thousands)

200

 

 

1 930 1 940 1 950 1960 1970 1980 1990 1999

Year

Figure 1.1: State Prison Population, 1930-1999

 

2001). In comparison, Russia had an incarceration rate of 675, South Afiica 400, the
United Kingdom 125, France 90, and Japan 40.2 It has been projected that the US,
which has 5% of the world population, will have a quarter of the world’s prisoners at
years end 2000 (2 million of the world's total of 8 million) (Ziedenberg and Schiraldi
2000).

This unprecedented expansion can be seen in California where the number of state
prisoners grew from approximately 17,000 in 1971 to over 157,000 in 1997. To
accommodate these increases California built 21 new prisons over this period and hired
almost 25,000 new employees to help manage and maintain prison operations. This rapid
expansion led California and other states to dramatically increase the portion of the
budget designated for corrections. For example, in Fiscal Year (FY) 1969 California
appropriated 161 million dollars from the general fund for corrections, which represented
approximately 3.7% of the entire general fund state budget. In contrast, in FY 1999
California appropriated 4.7 billion dollars for corrections or 7.4% of the entire general
fund budget (California Postsecondary Education Commission). Similarly, Donziger
(1996) notes that between 1979 and 1990, state government expenditures rose 325
percent for prison operations and 612 percent for prison construction (48). These major
changes, that were mirrored nationwide, led William Gregory to comment that “the most
frequently cited concern in state government of late is the ‘crisis in corrections.’ The

‘crisis’ is the dramatic increase in prison population coupled with soaring prison

 

2 The Russian rate of incarceration is calculated for September 2000, while for the other nations they were
collected on varying dates within the period 1997 through 1999 (Sentencing Project, p. 3).

construction and operating costs that are now destabilizing state budgets” (Gregory, 1).3
The questions that arise from the discussion above are what factors have contributed to
this unprecedented growth and what impact have these changes had on state allocations

of monies for corrections and other budgetary categories?

An Introduction to Sentencing Reforms

Prior to 1970, indeterminate sentencing characterized the approach to sentencing
prevalent in all of the states. Alan Dershowitz states that indeterminate sentencing
represented “a continuum of devices designed to tailor punishment, particularly the
duration of confinement, to the rehabilitative needs and special dangers of particular
criminals” (Miller, Roberts, and Carter 1981, 1). Under most indeterminate sentencing
procedures, the trial judge and parole board shared responsibility for determining the
actual sentence served by a convicted offender. However, beginning in the early 1970’s
this approach to sentencing came under attack. Indeterminate sentencing was viewed as
according too much discretion to decision makers that could result in disparity in the time
served by offenders based on racial, ethnic or class biases. Miller, Roberts and Carter
(1981) suggest that “indeterminate sentencing practices based on the principle of
rehabilitation appear to lack objective standards or principles to guide decisionmakers
and are thus believed to be related to unjustified sentencing disparity and ineffective
sanctions” (6). In addition, many critics felt that the actual time under indeterminate
sentencing was too short and the punishment often did not meet the crime (von Hirsch

1976). Tonry and Hatlestad (1997) note that indeterminate sentencing was commonly

 

3 The rapid increase in the number of prisoners led to overcrowding in the prisons. For example, at the end
of 1992 75% of all state prisons were operating over their maximum capacity. California operated at 91%
over capacity, Ohio 77%, Michigan 44%,and Virginia 39%.

disparaged as ‘bark-and-bite sentencing,’ where the judge’s ten year bark often resulted
in a one-to three-year bite (6). Finally, the idea that offenders could be rehabilitated was
challenged by a 1974 report by Robert Martinson, that reviewed 200 prison programs,
and found ‘little value’ in rehabilitation (Clark, 109). The end result was a national
movement that shifted away from indeterminate sentencing towards retribution, just
deserts, and determinate sentencing.

Deterrninate sentencing schemes establish standards or guidelines that dictate how
much punishment convicted offenders should receive. They generally establish a set
amount of time that offenders must serve for particular offenses while removing the
discretionary element of sentencing that characterized indeterminate sentencing. The
acceptance of determinate sentencing resulted in the abolition of parole, the adoption of
sentencing guidelines, mandatory minimums, truth in sentencing, and three strikes and
you’re out policies in the states. These changes represented a radical departure from past
practice. Messinger and Johnson note that

law is constantly changing, but the change is usually evolutionary and

incremental. Occasionally, a statute or judicial decision breaks abruptly

with the past, announcing not only a set of new rules but also a new

philosophical approach, indicating a change in the way the opinion leaders

of society are thinking about a longstanding problem (13).

The implementation of determinate sentencing was just such a shift and signaled an era of
‘tough on crime’ policies and harsher penalties for offenders.

The shift towards determinate sentencing significantly altered the time served by

offenders and the size of the prison population. The new policies directly led to increases

in the number of offenders committed to prison, the length of imposed sentences, and the

time served. The innovations in sentencing and the adoption of numerous sentencing

reforms led to a large increase in the number of offenders in prison and hence the size of
corrections budgets necessary to build and maintain prison facilities.

The growing enthusiasm for ‘tough on crime’ policies that favored incarceration
significantly increased the amount of monies allocated towards corrections. The
Postsecondary Education Opportunity (1994) argued that the policy changes not only
increased the share of budgets going to corrections but also adversely impacted spending
on other budget areas. They suggest that the shift to determinate sentencing

led to an ever-growing share of state budgets allocated to the construction

and operations of prisons. The increased share of state budgets for prisons

means that some other state activity is receiving a smaller share of state

budgets the share of social resources allocated to higher education has

shrunk while the share allocated to prisons has increased. And as state

financial support for public higher education has shrunk, tuition charges to

students have increased to offset the loss of state support (7).

The concern for potential tradeoffs between increases in state spending on corrections

and other budgetary categories, especially higher education, has been echoed by other

scholars.

A Changing Social Policy Trajectory

The movement towards determinate sentencing and ‘tough on crime’ policies
represents a reprioritizing of concerns for state governments. This is reflected in the
increased share of state budgets that are designated for corrections. According to Layzell
and Lyddon “a budget turns out to be a political document that reflects the values, vision,
mission and purpose of an organization” (xix). Viewed in this light public budgeting
systems are systems for making choices of ends and means, where the budget represents

an audit of the activities and priorities of government (Lee and Johnson 1998). In this

sense, the increases are a reflection of the growing importance placed on the punishment
of crimes. Donziger (1996) notes that “politicians at every level — federal, state, and local
— have measured our obsession, capitalized on our fears, campaigned on ‘get tough’
platforms, and won” (2). As a result, politicians have passed laws requiring harsher
sentences that have directly led to an expansion of the prison system. These changes
have not come without significant consequences.

Macallair, Taqui-Eddin, and Schiraldi (1998) note that in 1995 states spent more
building prisons than colleges for the first time. They state that there was

nearly a dollar-for-dollar tradeoff between corrections and higher

education, with university construction funds decreasing by $954 million

(to 2.5 billion) while corrections funding increased by $926 million to (2.6

billion). Around the country, from 1987 to 1995, general fund

expenditures for prisons increased by 30%, while general fund

expenditures for universities decreased by 18% (1).
The decline in monies for higher education can also be seen in California where general
fund appropriations declined from 16.8% ($733 million) of the total state general fimd
budget to 12.6% ($8 billion). Gottfredson (1995) suggests that the relationship between
increases for corrections and decreases in higher education could be viewed as ‘pretty
direct.’ He cites that in California state spending for corrections increased 25 percent
from FY 1991 through FY 1994, while funds for higher education declined by 25 percent.
Furthermore, the Postsecondary Education Opportunity (1994) note that “[a]bout half of
the increased costs of incarceration are being financed by states shifting resources out of
higher education and into prisons.” They report that between 1980 and 1992 the

increased share of state taxes allocated to corrections meant that $9.5 billion had been

shifted from other state activities into corrections. During this same period of time, there

was a approximately a $4.6 billion reduction in public higher education’s share of state

taxes (13).

Overview of Dissertation

This dissertation examines the factors that have had a direct impact on the
budgetary expansion in corrections. It also investigates whether a tradeoff exists between
appropriations for corrections and appropriations for higher education. In order to
accomplish these tasks, issues relating to the construction of budgetary theories will be
explored and outlined. This dissertation is fashioned around two complimentary
perspectives. The first perspective approaches the issues at hand as a student of
budgeting, with the aim of building upon and advancing previous theoretical constructs
of budgeting. The second perspective approaches the issues as a policy analyst with the
aim of examining the impact that the implementation of new sentencing policies has had
on state budgets. This dissertation is organized as follows:

Chapter 2 provides a general overview of budgetary theory and presents the
context within which budgeting for corrections will be examined. The chapter begins by
examining the theoretical literature surrounding budgeting. This includes literature
relating to tradeoffs (guns/butter), incrementalism, and the influence of exogenous
factors. Discussions of both bottom-up and top-down perspectives on budgeting are
covered. Additionally, the shift from indeterminate sentencing to determinate sentencing
is covered. This includes a historical look at the genesis of various determinate
sentencing policies in different states. From there the influence of the policy changes on

sentencing length and incarceration is discussed. It is shown that these new policies

directly impact the sanctioning phase. Furthermore, a supply/demand framework is
explicated that provides a construct within which to understand how an increase in
demand for prison space leads to greater appropriations for corrections. Finally, a
diagram of the budgetary process is presented that links the various theoretical
frameworks of budgeting into one holistic framework that serves as a springboard for the
research questions that will be explored in this dissertation. These include questions
about tradeoffs between appropriations on corrections and appropriations on higher
education at the state level; the extent that this year’s budget is a function of the previous
year’s budgetary base; the role of exogenous factors in the final budgetary allocations to
various categories of expenditure; how traditional incremental models need to be adjusted
to accommodate for an expanding or declining budgetary pie; and the extent to which
specific policies that are formulated outside of the budgetary realm have an impact on
budgetary allocations.

Chapter 3 provides an overview of the research design constructed to answer the
questions laid out in the previous chapter. This includes the construction of theoretical
and statistical hypotheses, the specification of an empirical model, and a discussion of the
Operationalization of concepts, data sources, and case selection. Furthermore, a summary
of the budgeting patterns in California, Michigan, and Virginia is outlined. This
discussion centers around the size of general fund appropriations for corrections and
higher education and the size of the budgetary pie for FY’s 1970-1999. From there the
various issues and technical procedures surrounding estimation are covered. This
includes a focus on issues of simultaneity, lagged variables, stationarity,

multicollinearity, heteroskedasticity, and autocorrelation. Two paths for estimation are

selected. The first treats the data in levels and the other in differences. The estimation
procedures for each path are laid out and a discussion of the rationale for competing
approaches and how to choose between the two is provided. The chapter concludes by
listing the anticipated conclusions of the dissertation.

Chapter 4 utilizes California as a case study to test the tractability of the research
design outlined in the previous chapter. The chapter begins with an overview of
correctional policy changes and reforms that occurred in California beginning in 1917.
This discussion is followed by a brief review of the contextual background in California
during the time period of investigation (1970-1999). This includes focusing on the
number of prisoners under state jurisdiction in California, the unemployment rate, the
violent crime rate, the size of the 18-24 year old cohort, the enrollment rate in higher
education, and a listing of governors. From there the estimation procedures as outlined in
the previous chapter are implemented and the various hypotheses tested. The chapter
concludes with a discussion of the implications of the statistical results. Chapter 5 and
Chapter 6 follow the same format as Chapter 4, but focus on Michigan and Virginia,
respectively.

The concluding chapter, Chapter 7, summarizes the results found in the empirical
chapters 4-6. In addition, some of the recent state responses to rising prison populations
and their potential impact on state budgets are discussed. The primary focus will be on
the national movement towards intermediate sanctions and Virginia’s recent experiment
with risk assessment and the diversion of nonviolent offenders to alternative sanctions

away from prison.

10

CHAPTER 2

In 1940, VD. Key asserted that a theoretical understanding of budgeting did not
exist. He proclaimed that scholars needed to break from the confines of an institutional
framework and generate a general understanding of budgetary behavior (Key 1940). Key
recognized that budgeting operates within a realm of scarce resources and wished to
understand the process whereby scarce means are allocated to alternative uses (1138).

Since Key’s proclamation, there has been a vast array of studies aimed at filling
this theoretical void (e.g., Wildavsky 1964; Davis, Dempster, and Wildavsky 1966).
Wildavsky’s seminal 1964 study focused on budgeting within the national government
and sought an understanding of how the budget was made and how Congress
appropriated funds. Wildavsky viewed budgeting as a process that was concerned with
the translation of financial resources into human purposes. He stated that “[s]ince funds
are limited and have to be divided in one way or another, the budget becomes a
mechanism for making choices among alternative expenditures” (2). Therefore, the
budgetary process can be viewed as an expression of politics and characterized by a
struggle for funding between departments, agencies, and appropriation committees.
Agencies would make a request for funding, budget bureaus would recommend how
much to appropriate, and finally the appropriation committee would decide how much to
allocate. Wildavsky was concerned with the strategies that various actors might employ
to get the funding they wanted and suggested that “[t]aken as a whole the federal budget
is a representation in monetary terms of governmental activity. If politics is regarded in

part as conflict over whose preferences shall prevail in the determination of national

11

policy, then the budget records the outcomes of this struggle” (4). As such, it can be seen
that a primary function of budgeting is to make decisions about how resources will be
allocated. Naomi Caiden suggests “[b]udgeting is more than technique. Budgets express
societal decisions regarding the levels and sources of public revenue and mobilization
and the purposes supported by public expenditure” (Caiden 48). Thus, a positivist study
of budgeting is important because it can shed light on explaining and allowing for
predictions of budgetary decisions and an understanding of how resources are allocated to
diverse interests and claimants.

In order to gain an understanding of the budgetary process and attempts at
generating a theory of budgeting, a review of several diverse sets of literatures will be
conducted. The first strand of literature deals with tradeoffs between budgetary
categories (defense and social welfare). The second set of literature relates to bounded
rationality and the influence of cognitive limitations on establishing future budgets
(incrementalism). Finally, literature regarding exogenous influences on the budgetary

process will be explored.

Tradeoffs

Tradeoff relationships assume interdependence between categories so that a gain
in one category results in a loss in another category. Examples of tradeoffs that have
been posited include: equality and efficiency (e.g., Okun 1975), inflation and
unemployment (e.g., Phillips 1958), timeliness and quality in court performance (e.g.,
Ostrom and Hanson 1999), and guns and butter (Russett 1969). Tradeoff relationships

may also exist within the process of determining budgetary expenditures for broad

12

categories of expenditures (e. g., defense and domestic spending), departments, agencies
within departments, or programs within agencies (Berry 1990, 191). A tradeoff between
two categories of expenditure assumes that “the choice of an expenditure for one
category must affect the amount of money allocated to the other category, such that
spending in one comes at the expense of the other” (Berry and Lowery 1990, 671).
Budgetary tradeoffs occur when the demand for resources for public programs exceeds a
finite pool of public resources that are available for allocation.1 One way to understand
tradeoff relationships is to examine a budget line.

A budget line combines the ideas of resource availability and substitution between
various categories of expenditure (usually two budgetary categories at a time).2 The line
itself represents the total budget that can be allocated among various activities, in this
case government activities, with a downward slope implying that appropriating more
funds to one area means less for other areas. The frontier is based on expected revenues
and financial constraints. An examination of Figure 2.1 reveals that at point A (on the
fi'ontier) more money will be appropriated towards category I than category 11. A point
on the budget line represents the point where policymaker indifference curves are tangent
to the budget line. This point represents the Optimum or preferred policy, where the
utility of the budgetary decision makers are maximized. Similarly, a move to point B and
the appropriation of more money for category 11 mandates that category I will receive less

funds. This fiamework is useful for the present study of budgeting by providing an

 

‘ Governments can avoid budgetary tradeoffs by providing a limitless supply of resources for public
programs. Demands can be met by raising taxes and/or absorbing deficits (Pryor 1968). However, political
realities place a cap on both of these sources, guaranteeing a finite pool of resources.

2 This author recognizes that there may be more than two competing forces for limited resources in a
society. Nevertheless, this framework provides a rudimentary way to understand the process of the
competition over limited resources and the resultant gains at the expense of other areas.

13

overview that illustrates that a budget is a limiting resource that necessitates that the
allocation of resources to one budgetary category comes at the expense of other
categories in any fiscal year (e.g., spending on corrections that could be used for higher
education, or vice versa).3 The most analyzed budgetary tradeoff in the political science

literature is the relationship between guns and butter.

Category I

 

 

 

 

 

 

Category II

Figure 2.1: Budget Line and the Tradeoff Between Goods

Studies focusing on a tradeoff between guns and butter attempt to assess the

existence of a tradeoff between defense and social welfare spending (e.g., education

 

3 An alternative way to conceptualize this would be to focus on the production possibility curve. This
approach allows one to understand tradeoffs and opportunity costs (the goods and services that must be
given up in order to acquire more of another good) within the light of efficient production.

14

and/or health). These studies have been cross-national (Russett 1969; Caputo 1975;
Domke, Eichenberg, and Kelleher 1983; Looney and Frederiksen 1996) and domestic
(Peroff and Podolak-Warren 1979; Russett 1982; Mintz 1989; Mintz and Huang 1991) in
focus and have reached mixed conclusions in their ability to confirm the existence of a
tradeoff between guns and butter.4

The mixed results have led many scholars to believe that defense and welfare
expenditures are driven by separate sets of determinants (Caputo 1975; Domke,
Eichenberg, and Kelleher 1983; Russett 1982). These authors argue against the idea of
interdependence and feel that spending decisions are characterized less by a “systematic
comparison of alternatives than a disjointed aggregation of spending decisions reached
largely in isolation” (Domke, Eichenberg, and Kelleher, 20). Domke, Eichenberg, and
Kelleher (I983) extend this line of reasoning by asserting that “students of budgetary
decisionmaking have discovered that direct priority setting - the trading off of one ‘
program against the other - is rare or nonexistent in democratic governments” (20). To
further explore this issue of interdependency it is necessary to examine the literature

surrounding budgetary decisionmaking.

 

‘ Berry and Lowery (1990) note that almost all empirical work on the tradeoff between guns and butter has
relied on the same basic methodology, either a single-equation regression model or a multiequation causal
model. These models measure the strength of the relationship between annual percentage change in one
category and change in another, with a negative sign for the relationship indicative of a tradeoff
relationship. Berry and Lowery comment that this approach limits conclusions to a descriptive level and is
incapable of providing insight into the process whereby tradeoff choices are made (Berry and Lowery, 672-
74). The authors suggest that tradeoff relationships can manifest themselves in many forms and each type
requires a different model to illuminate tradeoff processes. In order to understand the determinants of the
degree to which A is rewarded at B’s expense they construct a taxonomy consisting of allocations from a
fixed pool (sequential choice and simultaneous) and floating pool allocations (Berry and Lowery, 677).

15

Budgetary Decisionmaking

Early scholarly works dealing with budgetary decisionmaking processes were
derived from theories of bounded rationality (Simon 1947) and incrementalism
(Lindblom 1959). Lindblom viewed decisionmaking as a series of successive limited
comparisons that served to reduce and simplify decisionmaking. Within this framework,
policy comparisons are limited “to those policies that differ in relatively small degree
from policies presently in effect” (Lindblom 1959, 84). This perspective was generated
from a critique of the rational decision making approach. Lindblom asserted that the
rational-comprehensive (root) model of decision making consisted of a series of ordered,
logical steps where all possible alternatives are identified and the alternative with the
highest payoff (alternative that attains greatest amount of desired values) is chosen. The
assumption was made that complete information about all alternatives is both available
and manageable. Critics of the rational comprehensive model claimed that cognitive
limitations prevented a complete evaluation of all alternatives. For example, Simon
(1947) states that “[t]he limits of rationality have been seen to derive from the inability of
the human mind to bring to bear upon a single decision all the aspects of value,
knowledge, and behavior that would be relevant” (108). In other words, all alternatives
could not be known and even if they could, individuals and agencies could not compare
them all. Thus, individuals and agencies are forced to use rules of thumb and to satisfice.

These foundational principles were incorporated into theories of budgeting by
Wildavsky (1964) and Davis, Dempster, and Wildavsky (1966) who contend that budgets
themselves are never reviewed as a whole. Actors in the budgeting process cannot

consider the value of all preexisting programs compared to all possible alternatives.

l6

Instead, incremental calculations proceed from existing bases, so that “this year’s budget
is based on last year’s budget, with special attention given to a narrow range of increases
or decreases” (Davis, Dempster, and Wildavsky 1966, 529-30).5 In this
conceptualization, the base is considered as a given and is representative of past decisions
or negotiated agreements between competing interests, whereby only the increment of
change is evaluated in the present. This perspective implies that budgeting follows a
‘bottom-up’ process and that each program’s budget is independent of spending pressure
for other expenditure categories (Davis, Dempster, and Wildavsky 1966; Gist 1982) and
remains immune from political and economic considerations (Su, Kamlet, and Mowery
1993, 215). 6 This ‘bottom-up’ conceptualization has received criticism for ignoring the
possible impact of line item interdependency as well as the possibility that exogenous
factors may influence the budget making process (e. g., macroeconomic factors, partisan

control of legislatures, and public opinion).

Exogenous Factors

Kamlet and Mowery (1980) suggest that the incrementalist (‘bottom-up’) model
of budgeting fails to account for the finite resource base that characterizes the resource
allocation process (814). In addition to last year’s spending level, the level of spending

for a budget category may be influenced by “fiscal pressures, the availability of tax

 

5 Allen Schick (1988) critiques incrementalism as being a normative theory that is a statement both about
how budgeting works and how it should work. He states that incrementalism is valued because it
“moderates conflict, reduces search and transaction costs, stabilizes budgetary roles and expectations,
reduces the amount of time that busy officials must spend on budgeting, and facilitates remedial action to
correct mistakes” (Schick 61).

6 The ‘bottom-up’ process involves strategic bargaining and negotiations between bureaus and higher
levels in the budgetary hierarchy (e.g., departments and the Office of Management and Budget), in regards
to changes in the increment (Bozeman and Straussman 1982). Bozeman and Straussman state that “the
incrementalist approach assigns central importance to the role of negotiation strategies” (510).

I7

revenues, or by spending pressures generated by other categories of spending” (Su,
Kamlet, and Mowery 1993, 215, note 2). As such, they argue the notion of independence
is dismissed, as there is interdependence between budgetary, tax, and fiscal priorities.7
These ‘top down’ models (Kamlet and Mowery 1980, 1983, 1987; Auten, Bozeman, and
Cline 1984; Su, Kamlet, Mowery 1993) once again open the door for explicit tradeoffs to
occur between budgetary categories, by suggesting that resource pools are indeed finite.
Ultimately, budgetary outcomes are sensitive to pressures for growth in other programs
and various exogenous fiscal and political variables.

Davis, Dempster, and Wildavsky (1974) suggest that an agency’s budget is
primarily incremental in nature, but is occasionally responsive to specific political,
general economic, or social events. For these authors, the inclusion of exogenous factors
into their models represented an expansion of a previous specification that was comprised
solely of factors internal to the budgetary process (1966). Similarly, Allen Schick
suggests that “[b]udgeting is the process that prescribes how, when, and by whom claims
are to be made and resources distributed” (Schick, 63). The core characteristics of this
process are claiming forces (requests for resources made according to the rules and
procedures of the process) versus conserving forces (activity intervening between claims
and allocation that authoritatively rations resources). Schick states that in “a perfect
budget process, rations reflect the balance between claims and conservation. Anything
that alters the relative strength of these functions will change budget outcomes” (66).

The strength of these functions is often influenced by exogenous factors. The multitude

 

7 Fiscal and tax policies, in conjunction with economic trends, establish a fixed pool of resources for each
budgetary period that act as a constraint on total outlays (Kamlet and Mowery 1980). The ‘top-down’
process reflects both fiscal and political aspects that operate in accord with budgetary policy in establishing
government expenditures.

l8

of studies, following the work of Davis, Dempster, and Wildavsky, that focus on
exogenous factors and their impact on budgetary priorities have focused on the role of
factors such as: deficit levels, unemployment, inflation, increases in target populations,
and partisan control of the executive and legislative branches. This conceptualization has
been incorporated into studies focusing on criminal justice expenditures, spending for
corrections and spending for higher education.

William Taggart (1989) and Harriman and Straussman (1983) both focus on the
impact of court orders regarding overcrowding on state expenditures for corrections.
Taggart (1989) attempts to develop a general model of state expenditures for corrections
that accounts for judicial interventions as shift points that can alter the incremental nature
of the budget. Similarly, Calderia and Cowart (1980) focus on the responsiveness of
government institutions to policy relevant conditions (e. g., crime rates for criminal
justice). Caldeira and Cowart (1980) theorize that “policymakers in criminal justice
decide on appropriate levels of expenditure for their units by observing these policy-
relevant conditions” (416). Likewise, Layzell and Lyddon (1990) suggest that political,
economic, and demographic factors “serve as immediate indicators of supply and demand
for state services” (80). Finally, Marlow and Shiers (1999), in a recent article attempt to
understand if expenditures on law enforcement ‘crowd-out’ public education
expenditures in state governments.8 They state that “[t]he total budget is then the sum of
its component parts, where various demographic (e. g., age distribution), economic (e. g.,
unemployment rates), and cultural (e.g., past crime rates) characteristics then influence

demand for individual spending programmes” (257). Thus, these policy relevant

 

3 It should be noted that these authors do not attempt to generate a theory of the budgetary process and only
focus on data from two discrete time points (1985 and 1992).

19

conditions help establish a ‘window of opportunity’ for policy entrepreneurs (Kingdon,
1995). Policy entrepreneurs utilize these indicators and conditions to focus attention
towards problems where solutions can be attached (Kingdon 1995; Mintrom 1997).9
However, these authors and others fail to consider the impact of specific policies on the
budgetary process.

Specific policies have the ability to impact the budgetary process since they
necessitate the commitment of resources to certain programs and governmental
departments. Recent state policy changes relating to corrections and sentencing provide a

fertile area to study the impact of policies on budgetary choices.

Policy Changes in Corrections

Beginning in the early 1900’s and continuing through the 1970’s the medical or
rehabilitative model of corrections dominated state sentencing practices. This model
operated under a framework whereby legislatures set broad sentencing parameters, judges
imposed minimum and maximum prison terms within these parameters, and parole
boards determined the time in prison (Griset 1994, 533). The rehabilitative ideal placed
an emphasis upon the reformation of the offender and attempted to ensure that the
punishment fit the criminal and not necessarily the crime (Davies 1985, 3). This
approach focused on the individual and tried to achieve sentences that were just for each
individual defendant (Ostrom et a1. 1998, 8). Therefore, this practice was indeterminate

in nature and allowed for case-by-case decisionmaking and tremendous discretion for

 

9 Policy entrepreneurs are individuals who seek dynamic policy change and are advocates for proposals or
for the prominence of an idea (Mintrom 1997, 739; Kingdon 1995, 122).

20

both judges and parole boards.'0 However, the rehabilitative system was characterized by
its critics as being “replete with unbridled and unprincipled discretion, arbitrary and
capricious, unfair, violative of the notions of proportionality” (Lagoy et al., 386).

In the mid 1970’s a coalition of interests arose in opposition to indeterminate
sentencing. One part of the coalition was concerned with discretionary decisions being
made about individual’s liberty and the excessive disparity among sentences in
presumably similar cases. They were especially concerned with the possibility of
discriminatory sentencing practices based upon the race of the offender. The other side
of the coalition was more concerned with lenient parole boards that would often release
prisoners well before maximum sentences were reached (Blumstein 1984, 130). The
critics of indeterminate sentencing were responding to a series of factors including:
prison uprisings in the 1960’s (e. g., Attica and Tombs riots in New York), a
disillusionment with rehabilitation,ll a concern about individual rights and the control of
discretion, racial and class discrimination, and rising crime rates throughout the 1960’s
(Blumstein et al. 1983, 2). In response, critics called for a new approach that would mete
out fair and fixed punishments that would eliminate sentencing disparities and discretion,
in other words, a move towards presumptive or determinate sentencing.‘2

Deterrninate sentencing schemes are presumptive in nature and establish

standards or guidelines that dictate how much punishment convicted offenders should

 

'0 Judges had a broad authority to set appropriate sentences, while parole boards had the authority to set
release dates and release conditions for those sentenced to prison.

' ' A 1974 report by Robert Martinson that reviewed 200 prison programs found ‘little value’ in
rehabilitation (Clark, 109).

'2 The argument has been made that limiting discretion in one area only leads to increases in discretion in
other areas (hydraulic displacement of discretion). Thus, a reduction of discretion for judges, who have
broad authority to set appropriate sentences, and for parole boards, who have the authority to set release
dates and release conditions, is replaced by increased discretion of prosecutors who decide which offenders
are to be prosecuted and with what charges (Ostrom et a1. 1998, 9).

21

receive.l3 Inclusive within this schema are the creation of sentencing schedules (e.g.,
California Determinate Sentencing Law), sentencing guidelines that create ranges of
sentences for given offenses and offender characteristics (e.g., Minnesota), and the
abolition of parole boards (e. g., Virginia).l4 At the same time as the shift to determinacy,
several states imposed mandatory minimum penalties for offenders who committed very
serious crimes, for offenders who could be classified as repeat or habitual offenders, and
for offenses ranging from sexual assault to drug crimes to the use of a gun (Shane-
Dubow, Brown, and Olsen, v). According to a 1996 study conducted by the National
Council on Crime and Delinquency, all 50 states have some form of mandatory minimum
sentences or habitual offender statutes (Austin, 161). Mandatory minimums force judges
to hand out fixed sentences without parole, to people convicted of certain crimes.” For
example, in 1978 Michigan lawmakers passed the “650 Lifer law” which mandated life
imprisonment without parole for possession, sales, or conspiracy to sell, deliver, or
possess 650 grams or more of cocaine or heroin. An additional example is the 1973
Rockefeller Drug laws established in New York, which mandated minimum penalties for
various drug offenses. A report by the National Institute of Justice suggests that
mandatory minimums serve to deter firture criminal behavior through the threat of

imprisonment and act to enhance public safety via incapacitation (Ku 1980, 2). These

 

‘3 Determinate sentencing schemes have wide variation among ‘operational forms’ among the various
states (Nagin 1979; von Hirsch and Hanrahan 1981; Griset 1994).

'4 Between 1975 and January 1982 eleven states (Alaska, Arizona, California, Colorado, Connecticut,
Illinois, Indiana, Maine, Minnesota, New Mexico, and North Carolina) abolished parole boards (Blumstein
1984, 130).

'5 One unintended consequence of these reforms are that mandatory minimums provide defendants with a
strong incentive to go to trial in order to avoid lengthy sentences and this translates into increases in the
number of court cases and thus higher court costs. For example, prior to the introduction of a three strikes
law in California, ninety percent of all felony cases in California were plea-bargained. Following the
introduction of the law in 1994, up through 1995, only 14 percent of second felony and 6 percent of third
felony cases obtained guilty pleas (Edwards, 468).

22

new policies and philosophical shifts represented a radical departure fi'om previous
sentencing practices.

On May 1, 1976 Maine became the first state to abandon indeterminate
sentencing and the parole system. They were soon joined by several states including
California who adopted the Determinate Sentencing Law in 1977.'6 The law established
a presumptive sentencing scheme that limited prison terms to a very narrow range of
possibilities. Additionally, the law eliminated parole discretion by abolishing the release
decision power of the Adult Authority (California parole board). The new law in
California represented a radical shift away from the open-ended sentences of the
rehabilitation era and a move towards a ‘justice’ system of sentencing (Davies 1985, 2).
The new philosophical approach is expressed in Section 1170a of the California Penal
Code, which states “the purpose of imprisonment for crime is punishment” (Davies 1985,
2). The justice approach views offenses as the responsibility of the offender and
punishment as a ‘just dessert’ (von Hirsch 1976). Hence, rehabilitation is replaced by a
retributive philosophy that views punishment as an end in itself (Davies 1984, 4). Don
Gottfredson, commenting on the move towards determinate sentencing, states that
“sentencing purposes have shifted from an emphasis on utilitarian aims, particularly
treatment, toward a greater focus on deserved punishment, on the proportionality of
sanctions to harms done, and on equality” (Gottfredson 1984).

Another policy aimed at reducing judicial discretion, by mandating severe prison
sentences for second and third time offenders, is Three Strikes and You’re Out laws.
Beginning in 1993, 23 states and the federal government adopted some form of three

strikes legislation targeting repeat violent offenders. California which adopted a three

23

strikes law in 1994, doubles the prison sentence for second strike offenders with a prior '
serious or ‘violent’ felony conviction. Offenders convicted of any felony after two
previous serious or ‘violent’ felony convictions are sentenced to prison for a length of
twenty-five years to life.

Additionally, several states have enacted Truth in Sentencing laws, which require
offenders to serve a substantial portion of their sentences and thus reduce the discrepancy
between the imposed sentence and time served in prison. A 1995 report by the US.
Department of Justice points out the large discrepancy that existed before truth in
sentencing reforms. The report stated that criminals convicted of violent crime serve on
average only 48% of their court-imposed sentences before being released early from
prison (Crimestrike, 1). The act of bringing the time an offender actually serves in
custody closer in line with the length of the sentence imposed by the court can help
restore public confidence in the criminal justice system, help further deterrence, victim’s
rights, predictability, fairness, and consistency in the sentencing process (Ostrom et a1.
1999, 14). By the end of 1998, 27 states and the District of Columbia required violent
offenders to serve at least 85 percent of their prison sentences, compared with only 5
states in 1993 (Ditton and Wilson 1999).'7 The creation of Truth in Sentencing laws was
spurred on by the federal Violent Crime Control and Law Enforcement Act of 1994.
Contained within this act was the Violent Offender Incarceration/Truth-in-Sentencing
(VOI/T IS) incentive grants program that offered $9.7 billion for states that required

offenders convicted of Part I violent crimes to serve not less than 85 percent of their

 

'6 California introduced indeterminate sentencing in 1917.

'7 Truth in Sentencing laws vary across the states with some states requiring 100 percent of sentences to be
served (e.g., Idaho and Nevada), some 85 percent (e.g., California, Michigan, New York, and Virginia),
and others only 50 percent (e. g., Indiana and Texas).

24

sentence. '8 States that complied were eligible to receive grants that could be used to
build and operate new or existing prison facilities (Ditton and Wilson 1999). The state of
Michigan alone received over $55 million between 1994 and 1998 under this program
(Michigan Department of Corrections, 36a). Truth in sentencing initiatives operate in
such a way that policies that acted to reduce the amount of time an offender served, such
as good time, eamed-time and parole-board release, are restricted or eliminated (Ditton
and Wilson 1999, 3). Steven Donziger suggests that this federal incentive scheme
represents evidence of a national ‘get tough’ on crime trend that began in the late 1960’s.
He states that the federal government influences “crime policy by withholding money
from the state unless they adopted certain ‘get tough’ policies favored by the federal
government” (Donziger 1996, 13). '9

The various policies that have been enacted by state governments influence
sentencing lengths from both the ‘front end’ and from the ‘back end.’ Determinate
sentencing, mandatory minimums, sentencing guidelines, and truth in sentencing all
function to establish a predetermined length of commitment for offenders at the time of
sentencing. While the elimination of parole boards and the abolition of other
mechanisms of prison reduction act to eliminate mechanisms for early release from
prison following sentencing and imprisonment. Another way to examine the impact of
various public policies impacting corrections is to focus on their place within the stages

of the criminal justice process that impact growth in incarceration.

 

'8 Part I violent crimes include murder, nonnegligent manslaughter, rape, robbery, and aggravated assault.
'9 Donziger notes that since 1968, six major anti-crime bills have been passed by Congress and signed by
the President (Donziger 1996)

25

Stages of the Criminal Justice System

The stages of the criminal justice system that impact the growth of incarceration
include the commission of a crime, the arrest of an offender, the conviction of the
offender, the commitment of an offender, and time served in prison by the offender
(Blumstein and Beck 1999). Blumstein and Beck (1999) suggest that as “a result of the
‘get tough’ rhetoric that has characterized much political discussion about criminal
justice policy over the past two decades, both significant growth in the likelihood of
commitment to prison for those arrested and in time served by those sent to prison might
be expected” (27). As can be seen in Figure 2.2, the various policies that have been
discussed earlier have an impact on the level of commitment to prison and on the length

of time served by offenders. 2°

 

2° Parole has been added as a stage because it serves as a procedure that can impact time served, by
reducing the sentence of an offender. Additionally, the elimination of this release mechanism by many
states represents the removal of a safety valve for prison overcrowding. Previously, when states did not
meet statutes regarding prison occupancy levels, the threshold of the degree of rehabilitation warranting
release could be lowered. Finally, acts of technical parole violations lead to many parolees being returned
to prison. For example, out of 9,424 convicted offenders committed to prison as new court commitments,
probation violators, parole violators, and escapees with new sentences in Michigan in 1998, 3,109
represented technical parole violators (Factors That Drive Prison Population Growth, State of Michigan).
This process has been facilitated by the advent of Re-entry Courts in several states.

26

Stages Policies

COMIC‘iO“ : Determinate Sentencing
’ . Sentencing Guidelines
, """"""""""" > Mandatory Minimums
Commitment r' , _ ,
' Three Stnkes and You re Out

I, A. v _ _ ____!__-,

‘ Truth In Sentencing

Time Served -----------
, ‘ . , g 7 ““““ Elimination of Parole Boards
r— - ‘ '- ' -* . Adoption/Modification Good Time'
Parole T 7 fl 7 7 7 I F M 1

Figure 2.2: Stages of the Criminal Justice Process and Policies that Impact State
Incarceration Rates

In a previous study, Cohen and Canela-Cacho (1994) examined the growth in
incarceration rate from 1977 to 1988 in six states by focusing on the impact of the five
stages of the criminal justice process.21 While variation existed between the states, the
authors found that the growth in all the states’ incarceration rates was a result of
commitments to prison and time served (Blumstein and Beck 1999, 26, ft. note 5).
Similarly, Blumstein and Beck (1999) suggest that the “growing incarceration rate is a

compounded effect of changes at each of the stages of the criminal justice system” (42).22

 

2' Two of the six states the authors focused on are included in this study (California and Michigan), while
the other four states were Florida, New York, Pennsylvania, and Texas.

22 Blumstein and Beck (1999) note that there is variation among crime types in regards to the affect of
various stages. For example, the growth in incarceration rate for drug crimes is attributable to a rapid
growth in the number of arrests and commitments to arrest, not to time served. Conversely, growth in
incarceration rates for murder is entirely accountable by the growth in time served (42).

27

The authors focus on six offenses (murder, sexual assault, robbery, assault, burglary, and
drugs) between 1980 and 1996 and provide an indication of the contribution of each stage
to the total grth in the incarceration rate. They find that the offense rate accounts for
11.5 percent, arrests per offense .5 percent, commitments per arrest 51.4 percent, and
time served 36.6 percent. These findings indicate that only 12 percent of the growth in
incarceration is attributable to increased offending, while the remaining 88 percent is
attributable to increases in the imposition of sanctions. Finally, the authors examine the
growth in incarceration rate while omitting drug offenses. Blumstein and Beck find that
there is a negative trend for the offense rate, arrests per offense account for .8 percent,
commitments per arrest 41.5 percent, and time served 57.7 percent. Thus, over 99
percent of the growth in incarceration is associated with sanctions (43).23 On a similar
note, Austin (1996) points out that while new court commitments may have grown
slightly, stayed the same, or even declined, prison populations have continued to grow at
a faster rate. For example, in Michigan the number of new court commitments to prison
declined 18.4% from 9,333 in 1990 to 7,614 in 1994, while the prisoner population
increased 17.4% from 34,267 to 40,220. Similarly, new court commitments in Virginia
grew by 11.7%, while the prisoner population grew 45% between 1990 and 1994 (Austin,
159). These findings are important because they suggest that policy changes may have
directly impacted the sanctioning phase of the criminal justice process and resulted in
increased incarceration rates and helped bring about the ‘crisis in corrections.’

As discussed earlier in this dissertation, the ‘crisis in corrections’ derived from

serious prison overcrowding that resulted in the necessity to build several new state

 

23 An additional study by Langan (1991), which focused on increasing prison populations between 1974
and 1986, found that increases were not a function of increased sentence lengths, but rather a function of

28

prisons and increase the number of employees working within the Department of
Corrections. For example, in California alone 21 new prisons were built and 25,864 new
employees were added to the Department of Corrections between 1984 and 1997
(Ambrosio and Schiraldi 1997). In budgetary terms, this would suggest that states would
have to increase appropriations for capital budgets (construction) and operations for
corrections. Thus, policies that have impacted prison incapacitation may have a direct
impact on budgetary appropriations. An illustration utilizing a supply and demand
framework can be used to further demonstrate the impact of ‘tough on crime’ policies on
budgetary appropriations.

In figure 2.3, point A represents the status quo equilibrium. A shift from the
status quo occurs as states adopt ‘tough on crime’ policies that both increase
commitments per arrest and the time served by offenders. Thus, the demand for prison
beds increases (represented by a shift in the demand curve) and a new equilibrium is
established at point B. In the short run, the existing supply of prison beds would remain
constant and the increase in incarceration would be met by double bunking inmates.
Without a simultaneous increase in the supply of prison beds, prisons begin to become
overcrowded. As a result, the price per prisoner increases as more prison guards are
hired, operation staffs are increased, and the possibility of increased costs associated with
overcrowding (lawsuits and inmate deaths) becomes a reality. In response to these
increased costs, the legislature appropriates more money to corrections for both capital
outlays (to build more prisons) and for operations (for increasing Department of

Corrections employees). Graphically this is illustrated in the shift to point C as the

 

more prison sentences per arrest.

29

number of beds increase to meet the increase in supply of inmates and the per inmate cost

associated with ‘get tough’ policies seems to fall.24

Price per Prison Bed

 

 

Number of Prison Beds

Figure 2.3: Supply and Demand for Prison Beds

A similar framework can be adopted to understand the potential tradeoff that
might exist between appropriations for corrections and appropriations for higher
education. Figure 2.4 shows the supply and demand curves for higher education. Similar
to corrections (prisons), the demand curve shifts outward to the right to point B as
demand for higher education increases. Demand is increasing as college degrees have

become more of a prerequisite for employment and as the benefits associated with

 

2‘ It is beyond the scope of this illustration to make conjectures about the elasticity of both the supply and
demand curves and to offer up a prediction about where equilibrium C resides in relation to equilibrium A
when considering price. The graphical aid and discussion above is intended to provide elementary insights

30

attaining a college degree have risen (e.g., salary).25 To meet this demand, universities
are forced to hire new faculty members, build new classrooms, and increase staffs, all at a
cost. In the short run, increased financial pressures are met by raising the price per
student (tuition). However, faced with pressure from various interest groups the
legislature responds by increasing appropriations for higher education (shift to point C)
and thus increasing the supply of desks and decreasing the cost per desk (student).
However, the question still remains: given the fact that state budgets are limited by scarce
resources and an increasing demand for both corrections and higher education exists, how
are the demands for these two differing state priorities met?

As the overall population continues to increase, more ‘tough on crime’ policies
are implemented, and the incentives for the continuance of post-secondary education rise,
demand for correction and higher education resources will continue to grow. It is
possible that an expanding budgetary pie will be able to accommodate resource requests.
However, the question becomes whether or not appropriations for corrections are rising at
a rate sufficient to keep the cost per prisoner relatively constant compared to whether the
amount appropriated for higher education is rising at a rate sufficient to keep the cost per
student relatively constant. If scarce government resources are being appropriated for
corrections in large enough sums that they crowd out resources available for higher

education, then the price of attending college will rise to offset the increasing supply.26

 

into the impact of policy changes upon the appropriation of monies for corrections and the expansion of
correctional activities.

25 Between 1970 and 1998 the number of students age 14 to 34 years old enrolled in four-year colleges
increased from approximately 4,910,000 students to 8,794,000 (US. Census Bureau 1999).

2" The ideas discussed in this section have evolved out of discussions this author had with Dr. Brian
Ostrom.

31

D!

.y. S
D S.
c:

g / B \

8

'c

u.)

1'5 A C

in

I

83‘

O.

8
'c

0..

 

 

Number of Higher Education Desks

Figure 2.4: Supply and Demand for Higher Education

Overview of Study

This study will not attempt to deterrrrine the impact of policy reforms on prison
populations, the impact of increased imprisonment on crime rates, nor will it be an
evaluation of the effectiveness of particular policy changes on increasing sentencing
lengths or the number sentenced. Instead, this study will focus on the impact of specific
policies associated with corrections, at the state level, and model and evaluate the
budgetary implications of these policy choices. It is hypothesized that these policies will
have a direct impact on both appropriations for corrections and appropriations for higher
education. The implications that follow from this framework are that legislative policy
choices serve as an added exogenous factor that has an impact on budgetary choices.

Additionally, it is assumed that budgetary choices made at the state level, that are based

32

upon a pool of resources that expands (or contracts) in each budgetary period (Berry and
Lowery 1990), may be influenced by both ‘bottom-up’ and ‘top-down’ processes. Thus,
it will be critical to examine not only the explicit budgetary process, but also legislative
activities (creation and implementation of specific policies) that produce budgetary
outcomes. For a visual illustration of the budgetary process, see Figure 2.5.
Using Figure 2.5 as a design template, this study will focus on the following
questions:
1. Does independence or interdependence between categories of spending
characterize the budgetary process? In particular, is there a tradeoff between
appropriations on corrections and appropriations on higher education at the

state level?

2. To what extent is this year’s budget a function of the previous year’s
budgetary base?

3. What is the role of exogenous factors in the final budgetary allocations to
various categories of expenditure?

4. How can traditional incremental models be adjusted to accommodate for an
expanding or declining budgetary pie?

5. To what extent do specific policies that are formulated outside of the
budgetary realm have an impact on budgetary allocations?

33

mmoooE gem—:5 2:. ”Wm 053m

,. re

 

 

 

 

 

 

 

 

 

 

 

«3:398»
waged

 

 

920ml acceptance
ozocomoxm ass

a
<

anionic? teapot. 33-26%? wagon
“mamEowxm EmfiEoEocoS 6233.53

        

 

    
  

cumin 95>
835.:— n.< Eve—canoe

Tffl

 

 

 

 

 

 

 

34

CHAPTER 3

In order to address the questions posed at the end of Chapter 2 a series of
theoretical hypotheses are generated that provide logical explanations for the linkage
between concepts under investigation. Subsequently, a model is constructed that allows
for an assessment of interdependence between budgetary categories. The model is
comprised of various components and parameters to account for the influence of both
incremental and ‘top-down’ processes, a possibly expanding budget, and exogenous
factors, including specific public policies. After the empirical model has been specified,
statistical hypotheses are generated that allow for empirical testing of the designated
relationships that characterize the budgetary process. From there a research design is
explicated that serves as an overview of how the study proceeds. Included is a discussion
of the operationalization of concepts, sources of data, an overview of estimation

strategies, and anticipated conclusions from the empirical analysis.

Theoretical Hypotheses

As outlined in chapter 2, this dissertation focuses on the construction of a
budgetary theory and the determinant of commitments of state budgetary resources for
different budgetary categories. More specifically, focus is placed upon the amount of
monies appropriated for corrections and higher education. Building upon prior empirical
research and theoretical frameworks this study explores if there is a tradeoff relationship
between appropriations for corrections and appropriations for higher education; if

budgetary appropriations can be characterized by an incremental process; if exogenous

35

factors exert significant influence on determining budgetary allocations; how a changing
budgetary pie impacts appropriations; and the role that the implementation of specific
public policies may play in determining budgetary allocations. In order to investigate
these questions a series of theoretical hypothesis are generated that provide a theoretical

linkage between the cause and effect relations between various concepts.

Budgetary T radeojfs. Budgetary categories can be either independent of one
another or be interdependent. In examining corrections and higher education, bottom-up
processes may be the sole determinant of departmental allocations or there may be a
tradeoff relationship between categories. It will be assumed that a finite pool of resources
limits resources available to state governments. Therefore, budgetary gains in one
category must come at the expense of other categories. Consequently, the budgetary
process is characterized by interdependence between expenditure categories. The focus
is placed on the potential tradeoff between corrections and higher education because
these two budgetary categories are discretionary in nature, unlike entitlement programs
such as Medicaid, and other budgetary areas that necessitate by law that a proportion of
the state budget go to fund them (e. g., K-12).l Thus, changes in the amount for
corrections cannot be offset by changes in these other programs that are fixed in nature
and must be met either with an increase in taxation, additional revenue generation, or the

reduction of budgets in other areas that are discretionary in nature.2

 

' For example, California’s Proposition 98 (1988) guarantees a minimum amount of state funding for K-14
education and Michigan’s Headlee Amendment (1978) assures that at least 41% of the state budget goes to
local governmental units.

2 Carroll and Bryton (1996) note that “the legislature is not required by statue or federal mandate to provide
higher education or other services. Under current law, they must do with whatever revenues are left over
after meeting the spending requirements of programs mandated by either federal law or propositions
enacted into the California constitution.”

36

TRADEOFF HYPOTHESIS: As total appropriations for corrections increases,
total appropriations will decline for higher education and vice versa.

lncrementalism. The assumption is made that budgetary actors are constrained by
limited cognitive abilities and thus, prior budgetary choices (t-l) will have a significant
impact on present outcomes (t). In other words, budgetary actors areunable to review the
entire budget each year and cannot consider all possible alternatives, thus budgetary
actors are forced to make successive limited comparisons that reduce and simplify the
decisionmaking process (Lindblom 1959). This implies that budgetary actors make
budgetary decisions based upon the existing budgetary base. This relationship is
captured in the following hypothesis.

INCREMENTALISM HYPOTHESIS: This year’s budget will be influenced by

the previous year’s budget.

While past studies have found evidence that have reinforced the assertion that the
budgetary process is characterized by incrementalism and decisionmaking by bounded
rationality (Davis, Dempster, and Wildavsky 1966, 1974; Bozeman 1977; Kamlet and
Mowery 1985; Kamlet, Mowery and Su 1988), more current studies have argued against
the utility of this notion on both methodological and theoretical grounds.

In a recent article, Achen (2000) discusses some of the potential consequences
associated with the inclusion of lagged dependent variables in time series applications.
He suggests that when a lagged dependent variable is included into a regression equation
with serial correlation the lagged variable effectively soaks up the effects of not only

unmeasured variables, but also the included variables in the model. This results in the

37

incremental component often being highly significant and as a result any conclusions
drawn about budgetary decisionmaking are often a direct result of a statistical artifact.

As such, Achen argues that inclusion of incremental components often lead to large R2
values and incredibly large t-ratios exaggerating the impact of incrementalism and
minimizing the impact of political effects. This finding has major implications for
studies of budgeting. Achen notes that “last year’s budget will predict this year’s budget
very well even if it has little or no real causal impact” (9). In addition, Achen notes that
the intellectual basis for incrementalism has recently come under “serious attack.” He
suggests that advances in games of incomplete information have rendered the
incrementalist perspective as “too limited to explain even the substantial variation in
budget success” (8). He states that “[g]ames of incomplete information show that fully
rational players will misleadingly appear to be merely boundedly rational in the usual
case when the political environment is stable, but not at other times” (10). The author
goes on to note that budgets are indwd quite stable over time and well predicted by the
prior year, but this does not necessitate that decisionmakers are bounded rationalists or
incrementalists. Thus, Achen’s argument provides both a methodological and theoretical
rationale for excluding an incremental component from models of budgeting. From a
theoretical perspective the idea of bounded rationality in budgeting has been debunked by
games of incomplete information and empirically incrementalism has been shown to be
difficult to disentangle. On the basis of these two critiques the incremental component of
budgeting is not included in this study and the incrementalism hypothesis, laid out above,

will not be tested.

38

Exogenous Factors. Beyond an explicit interdependence between categories,
budgets are also responsive to exogenous factors. The budgets for corrections and higher
education may be influenced by factors such as: the political ideology of the governor,
the violent crime rate, changes in the size of the 18-24 year old cohort, the number of
individuals enrolled in institutes of higher learning, and the unemployment rate. These
factors combine to establish policy relevant conditions that create windows of
opportunity and hence may influence the budgetary process. For example, the presence
of a conservative governor may lead to an increase in the percentage of total
appropriations committed for corrections and decreased levels appropriated for higher
education. Charles Thomas, director of the Private Corrections Project at the University
of Florida, states that “[c]onservatives put imprisonment higher on their political agenda
than do their more liberal counterparts” (Mergenhagen 1996). Therefore, conservative
state governors might pursue ‘tough on crime’ agendas that lead to the increase in the
allocation of resources for criminal justice activities, including increases in
appropriations for spending on corrections. Similarly, high violent crime rates might lead
policymakers to respond to what they deem to be an active threat to society by increasing
spending on corrections. Corrections increases might occur to meet the resource demand
that occurs as more individuals are incarcerated3 or as policymakers anticipate a need for
increased funding to meet future growth in prison populations.

In addition, a rise in the 18-24 year old cohort, the age group most prone to
criminal activity, might signal that crime rates might rise and that increased spending for
corrections is needed. The idea is that as more individuals engage in criminal activities,

more individuals will enter into the stages of the criminal justice process and eventually

 

 

Thrs assumes that higher crime rates are assocrated wrth higher arrest, convrctron, and commitment rates.

39

end up increasing the number of individuals incarcerated, thus, increasing the need for
additional resources for corrections. Furthermore, this cohort represents the age that most
individuals pursue some form of postsecondary education. Thus, an increase in the size
of this cohort may lead to more resources being designated for higher education to meet a
growing demand. Additionally, increases in the number of students enrolled in
institutions of higher education might lead to increases in appropriations for higher
education. These appropriations might be needed to meet growing demands for
additional operational resources and infrastructure renovations and expansions as
growing student populations put a strain on existing higher education staffs and facilities.
Finally, the fiscal well being of a state’s economy may influence the amount of monies
appropriated for corrections and higher education. During state fiscal crises, budgets for
discretionary budget categories may be reduced to help balance state budgets and make
additional revenues available. State unemployment rates are utilized in this study as a
proxy for fiscal well being, so that as unemployment rates increase, appropriations for
higher education decline. On the other hand, rising unemployment rates lead to increased
appropriations for corrections. Huang et al., state that “higher unemployment rates lead
directly or indirectly through their positive influence on crime rates, to higher
imprisonment rates” (402) and thus higher appropriations for corrections.

EXOGENOUS HYPOTHESIS: Exogenous factors help in the creation of

windows of opportunity and thereby influence appropriations in an episodic
fashion.

Budgetary Pie. Budgets themselves do not remain constant over time. Expanding

economies help foster grth in budgets and recessions serve to retard this development.

40

Overall budgetary expansion may be responsible for the growth of categorical budgets
and mask tradeoff relationships. An expanding budget allows for the possibility that
negative tradeoffs will not always occur in real terms. Rather than being a zero-sum game
where an increase in one category necessitates a decrease in another category, it can be

viewed as an expanding-sum game (See Figure 3.1).

 

 

20% ($80)
15% ($30)
A
15% ($30)
10% (3401
Budget ($200) at time t-l Budget ($400) at time t

 

 

 

Figure 3.1: Budgets as an expanding-sum game.

For example, if the budget was $200 at time t-l, and both categories A and B received
fifteen percent of the budget, both categories would receive $30. At time t the budget
might expand to $400 and category A might receive twenty percent ($80) of the budget
and category B might receive ten percent ($40) of the budget. Thus, even though both
categories increased their budgets in real terms, category B experienced a decline in the
percentage of the budgetary pie it was allocated (Hayes 197 5, 28). In regards to the

present study, it is possible that both appropriations for corrections and appropriations for

41

higher education are increasing, but the percentage of the budgetary pie they are
receiving from year to year is declining.

BUDGETARY PIE HYPOTHESIS: An increase (decrease) in the budgetary pie

leads to increases (decreases) in both appropriations for corrections and higher

education.
Public Policies. Finally, it is possible that discrete political events, in the form of public
policies, have a direct impact on budgetary outcomes. These policies will impact the
budgetary process by committing funds to specific programs and departments. For
example, state government policies surrounding truth in sentencing, the adoption of
determinate sentencing, mandatory minimums for weapon and drug crimes, and Three
Strikes and You’re Out will indirectly influence the resources that are needed to maintain
the activities of correctional departments. For example, Blumstein and Beck (1999) note
that the growth in state incarceration rates is due to the sanctioning phases of the criminal
justice stages, which can be related to specific criminal justice policies. An increasing
incarceration rate necessitates an increase in the resources required for both the operation
and maintenance of correctional facilities, including the need to build new facilities. The
resources that are committed to corrections as a result of legislative politics will come at
the expense of higher education.

POLICY HYPOTHESIS: The implementation of specific correctional policies

(e. g., truth in sentencing, Three Strikes and You’re Out) lead to increases in the
percentage of total appropriations for corrections.

42

Model and Statistical Hypotheses

Now that a series of theoretical hypotheses have been established, a model is
constructed that allows for the testing of statistical hypotheses that are generated from the
various theoretical hypotheses. The model of the budgetary process for corrections and

higher education can be expressed as the following system of equations:

K P
C: = bro + b11Ez+ b12(P: — Pr — r) + 2k 161th + Z l7pr1+4211
= p:

K
E1: bzo + b21Cr+ b22(Pr — P: - r) + Ska/(Xe + e21

where C, and E represent appropriations for corrections and higher education,
respectively, (Pt-PH) is the change in the size of the overall budget from the previous
year, X, represents exogenous factors, and Zp represents specific public policies.4

Each of the components and hypotheses discussed above can be assessed utilizing
the specified model. First, the existence of a tradeoff relationship can be determined by
examining the coefficients b“ and b2.. The traditional way to think about a tradeoff
occurring would be to think of one of the coefficients being positive and the other
negative. The other way would be for one coefficient to be greater than one and the other
less than one. This would express the rate at which one of the budgetary categories is
growing in relation to the other category. Second, the impact of exogenous factors can be
determined by examining the magnitude of 01k (k=1, 2, , K). If the coefficients are
zero, then independence can be deduced. Similarly, the impact of specific public policies

can be determined by assessing the magnitude of 7,, (p=l , 2,..., P). Finally, it is possible

43

that changes in budgetary appropriations may be a function of an expanding or
decreasing budget. As a result, the impact of a changing budgetary pie can be examined

by evaluating the magnitude of big and by.

Table 3.1: Statistical Hypotheses

b1] > O Will—2210, OI‘ _
b” > 1.0 while b2. < 1.0 (vice versa)
Independence 01;, = 0

Tradeoffs

Interdependence on, at 0

' blZ > 0,
Budgetary Pre b22 > 0

 

 

Policy Impact

Data Source

In order to test the statistical hypotheses, data was collected fi'om three states:
California, Michigan, and Virginia, for the years 1970 to 1999. These states are
representative of the geographic regions of the United States (West, Midwest, South) and
possess over 19 percent of all prisoners held under the jurisdiction of state correctional
authorities in 1996 (Gilliard and Beck 1998, 3). At year’s end in 1996 California held
142,865 state prisoners, Michigan held 42,349, and Virginia 27,062 out of a total of
1,131,581 state prisoners (Gilliard and Beck 1998, 3). In a 1994 ranking of states with
the largest prison populations, all three states ranked in the top ten states with the most
prisoners, with California ranking first. Additionally, the states all experienced problems

with overcrowding. In 1992, California had the most over crowded prisons in the nation

 

44

at 91 percent over capacity. Michigan and Virginia both were in the top ten and were 44
percent and 39 percent over capacity, respectively. Therefore, these three states are all
states where the issue of prisons and corrections is very salient.

As discussed earlier in this dissertation, there has been a national trend since the
early 1970’s towards the implementation of policies that are both ‘tough on crime’ and
determinate in nature. These reforms have occurred during the period of study and allow
for an interrupted time series framework to be undertaken. The range of these policies
have come under the headings of mandatory minimums, truth in sentencing, sentencing
guidelines, and alterations to the parole system. The three states in this study comprise a
sample that have all experienced a series of reforms that, taken together, include all of the
typologies listed above. For example, mandatory minimums have been implemented in
Michigan, truth in sentencing in Virginia, sentencing guidelines in Michigan, and
modifications to parole mechanisms have occurred in California and Michigan.5 Each of
these states has experienced radical policy reforms that are representative of a national
change in philosophical approaches towards sentencing and a fundamental restructuring
of the way that judicial sentencing is performed. These reforms have impacted three
decision points in the sentencing process: the decision to incarcerate, the selection of
sentence length, and the decision to release. As such, all three states have experienced a
large increase in their prison populations and overcrowding in their prisons as the number

of inmates outgrew prison capacities.

 

5 . . . .
Thrs list rs far from exhaustive.

45

Operationalization and Data Sources

The next step in the research process is to operationalize theoretical concepts into
variables that can be represented by numerical referents. Table 3.2 displays the way that
the various theoretical concepts are measured. In conjunction with this table a discussion
of the Operationalization of concepts is warranted. This discussion provides an overview

of the desired data characteristics, the data measure selected, and the data sources.

46

Table 3.2: Operationalization

 

Operationalization

 

Symbol Concept
Ct Appropriations for Corrections
E, Appropriations for Higher
Education
(Pt-PH) Budgetary Pie
Xk Exogenous Factors
Zp Public Policies

47

Appropriations for corrections from
state appropriations for fiscal year t

Appropriations for higher education
from state appropriations for fiscal
year t

Total state appropriations for fiscal
year t minus state appropriations for
fiscal year t - 1

Violent Crime rate -- Federal Bureau
of Investigation's Uniform Crime
Rate for Index Crimes

Enrollment in institutions of higher
education - Statistical Abstract of the
United States

18-24 year old cohort -- US. Census
Bureau population statistics
Governor’s political party (0 =
Democrat; 1 = Republican)

State Unemployment rate —- US.
Census Bureau, Current Population
Survey

Specific public policies (e.g., Three
Strikes and You're Out) related to
corrections (0 = no policy; 1 =
following passage of policy)

Appropriations:

Appropriations were selected as the unit of measurement for budgeting because it
best represents the culmination of the legislative budgetary decision process. This
process includes the submission of requests for appropriations from governmental
departments and agencies, the establishment of the governor’s executive budget,
negotiations and resultant changes to budgetary proposals by state legislative fiscal
committees, and final approval of appropriations by state legislative bodies. Thus,
appropriations represent an endpoint of a lengthy process whereby funds are made
available to departments and agencies for expenditure. Inclusive within the budgetary
process are decisions about how large the state budget should be and potential allocation
tradeoffs between departments and agencies. In contrast, expenditures are less desirable
for examining the budgetary process because they represent a different set of
decisionmaking processes. Expenditures represent the actual spending decisions of
various departments and agencies that follow decisions about the amount of monies to be
appropriated. The amount of money expended is often more or less than the amount that
has been appropriated. As such, the desired characteristics for appropriations data are

listed in Table 3.3:

48

Table 3.3: Desideratum of Appropriations

General Fund Appropriations for Corrections, Higher Education, and State
Total

0 Exclusive of: federal, special, or other funds

0 Inclusive of: operations and capital outlay

Corrections: Department of Corrections for adult corrections

Higher Education: Four-year universities

 

The desired budgetary unit for this study is general firnd appropriations. The
general fund reflects the major share of a state’s fiscal transactions. Utilizing this
measure ensures that all monies from special funds, federal sources, and other funds are
not included in the appropriations for corrections, higher education and state totals.
Consequently, one is able to focus on the pool of state monies available for general state
purposes. In addition, it is desired that appropriations from the general fund be inclusive
of both monies for operations and for capital outlays. These components of the budget
are desirable because they represent aspects that might be impacted by an increasing
demand for certain state activities. For example, policy changes and changing exogenous
conditions might lead to the expansion of correctional activities. This expansion may
necessitate the increase in appropriations for both the operation (e. g., need to hire more
correctional officers and prison staffs) and the construction of new facilities or the
renovation of older facilities (capital outlay). Thus, it is desirable to contain both fiscal
components in a study of budgeting. Finally, it is preferable to focus solely on

appropriations for adult corrections and appropriations for four-year universities. This

49

allows the level of analysis to remain focused on the state level and not on the
interrelation that might exist between the state and localities.

However, due to difficulties in obtaining the desired data from various legislative
analyst’s offices, budget and management offices, departments of corrections,
departments of education, and a multitude of national organizations (e.g., National
Association of State Budget Officers; National Institute of Corrections) the data that
could be obtained does not fully meet the desideratum of appropriations. Table 3.4
outlines the data that was obtained, the sources of the data, and defining attributes.
Nevertheless, the deviations fiom the desired characteristics of appropriations will not
invalidate the results of this study. Relationships that characterize the budgetary process
can still be examined. However, it does mean that a pooled analysis incorporating cross
sectional variation is not possible since units of measurement across the states are not
comparable, while a direct comparison of empirical results between states must be done

with caution.

50

Table 3.4: Appropriations Data — Title, Source, and Notes About the Data

 

California Appropriations: Corrections, Higher Education, Total

Description: “State General Fund Appropriations for State Government Functions,
State Operations and Local Assistance in Actual and 1999-2000 Constant
Dollars for Fiscal Years 1967-68 Through 1999-2000.”

Sources: California Postsecondary Education Commission, Fiscal Profiles, 1999
www.cpec.ca.gov

Notes: Utilize actual dollar amounts, not constant dollars
All monies for capital outlays are not in display
Corrections is both Youth and Adult Corrections
Higher Education includes local community colleges and 4-year public
colleges and universities

 

Michigan Appropriations: Corrections, Higher Education, Total
Description: Michigan General Fund/General Purpose Appropriations
Source: Executive Budget of the Governor of Michigan

Notes: Includes both capital and operations
Higher Education includes only 4-year public colleges and universities

 

Virginia Appropriations: Corrections

Description: “Approximate Appropriations Related to Operating Expenses for Housing
State Inmates, 1970-2002.”

Source: Dick Hall-Sizemore, Department of Planning and Budget, VA

Notes: General Funds Only; operating expenses

Virginia Appropriations: Higher Education
Description: “Appropriation History for Fiscal Years 1970-71 through 1999-2000.”

Source: A. Fletcher Magnum, State Council of Higher Education, VA

51

Table 3.4 (cont): Appropriations Data -— Title, Source, and Notes About the Data

 

Notes: Total appropriations for four year colleges and universities (excludes two
year institutions) from General Fund

Virginia Appropriations: Total
Description: “History of Appropriations for Operating Expenses”

Source: Mike Barton, Department of Budget and Planning State of Virginia

Notes: Actual appropriations for operations from the General Fund.

 

Exogenous Factors

Violent Crime Rate: The crime rate is measured by utilizing reported figures item
the Uniform Crime Reports as produced by the Federal Bureau of Investigation and made
available by the Bureau of Justice Statistics. For purposes of this study the violent crime
index, which is an estimation of offenses known to police per 100,000 inhabitants, is
utilized. This index is comprised of the following Part I offenses: murder and
nonnegligent manslaughter, forcible rape, robbery, and aggravated and simple assault.
Data was obtained from an annual publication entitled The Sourcebook of Criminal
Justice Statistics, produced by the United States Department of Justice. It should be
noted that the use of data from the Uniform Crime Reports is often subject to criticism.
Critics (Skogan 1974; Siedman and Couzens 1974) claim that this measure represents a

deflated estimate of the amount of criminal activity, because it relies solely on offenses

52

known to police. Thus, it is argued that only a small portion of criminal activity is ever
reported and that the index is susceptible to changes in administrative procedures
(Siedman and Couzens 1974). Nevertheless, this measure is appropriate for this study
since media outlets and politicians often make reference to this statistic and it has become
a part of public discourse surrounding issues of public safety. As such, it is an instrument
that might be interpreted by policymakers as a signal for policy relevant conditions.

Enrollment: The number of students enrolled in institutions of higher education is
measured by utilizing data from the National Center for Education Statistics, Higher
Education General Information Survey and the Integrated Postsecondary Education Data
System. The measure itself represents the total number of students enrolled in the fall
semester, inclusive of public four-year institutions of higher education.

Cohort: The size of the 18-24 cohort is measured by utilizing data from the
United States Census Bureau’s population statistics. Specifically, the data comes from
reports entitled “Population Estimates for the US, Regions, and States by Selected Age
Groups and Sex: Annual Time Series;” “Historical Annual Time Series of State
Population Estimates and Demographic Components of Change 1980 to 1990, by Single
Year of Age and Sex;” and “Intercensal Estimates of the Resident Population of States
1970 to 1980.”

Political Ideology: The issue of the political ideology of the state government is
captured in a dichotomous variable that treats a Republican governor as being
synonymous with a ‘conservative’ state government and a Democratic governor with a
‘liberal’ state government. The party of the governor was chosen because the governor

represents a strong political force in the budgetary process at the state level; The

53

governor not only appoints agency leaders who make requests for funds, but also is
responsible for preparing the executive budget that establishes priorities and provides
direction for state government and policy. An alternative measure constructed by Berry,
Rinquist, Fording, and Hanson (1998) was considered because it represents a more
complete specification as it accounts for the ideology of the governor and each house in
the state legislature. However, data for this measure only exists through 1995, which
would severely curtail the scope of this study and make it impossible to investigate the
impact of policies that were implemented in the latter half of the 1990’s.6
Unemployment: This measure is being used as a proxy for state economic well-
being. As such, during economic downturns states face revenue shortages and may
redirect resources away fi'om discretionary areas of the budget and move them towards
mandated areas of the budget. Unemployment is operationalized as the unemployment
rate, which represents the number of unemployed as a percent of the labor force, in each
state for each fiscal year. According to the Bureau of Labor Statistics, unemployed refers
to all persons who had no employment during the reference week, were available for
work, except for temporary illness, and had made specific efforts to find employment
some time during the 4-week-period ending with the reference week. Persons who were
waiting to be recalled to a job which they had been laid off need not have been looking
for work to be classified as unemployed. Data comes from the Bureau of Labor
Statistics, using data from the Current Population Survey (CPS) as collected by the

Bureau of the Census.

 

6 Correlations between the dichotomous variable for partisanship of the governor and Berry et al.’s (1998)
measure were run for California between 1969 and 1995. The results produced a correlation coefficient of
.944, confirming the reliability of the alternate construct.

54

Overview of Budgeting Patterns In California, Michigan, and Virginia
Before the estimation strategy is introduced the three key budgetary variables,
appropriations for corrections, apprOpriations for higher education, and the size of the

budgetary pie, are outlined for each of the three states in the study.

California -- Appropriations

An examination of Figure 3.2 reveals that state General Fund appropriations for
corrections and higher education both rose significantly between FY 1969 and FY 1999.
Over this time period, appropriations for corrections increased from $161 million to $4.7
billion equating to a 2,843 percent increase. Similarly, appropriations for higher
education rose from $733 million to $8 billion over this same time period, for a percent
increase of 993. As such, it can be inferred that corrections was growing at a faster rate
than higher education during this time span. In addition, an inspection of the two
budgetary series shows that the budget for corrections is less volatile than for higher
education. Appropriations for corrections grew slowly and steadily between FY 1969
and FY 1976. In FY 1977 the rate at which appropriations for corrections was increasing
began to rise and this rate increased even more following FY 1983. In contrast, higher
education grew fairly steadily between 1969 and 1979, and then experienced a period of
little growth between FY 1980 and FY 1983. Following this period of limited growth,
appropriations for higher education continued to grow in FY 1984 until FY 1992 and FY
1993 when the amount appropriated declined from the previous year. Finally, a
resurgence in grth occurred between FY 1994 and FY 1999. Nevertheless, viewing

appropriations for each budgetary category independent of the impact of the size of the

55

state budget gives an incomplete picture. In order to get a better understanding of
changes to departmental appropriations over time, it is necessary to examine the
budgetary allocations in the light of the entire state budget. Thus, the level of

appropriations are converted into a percentage of the total state appropriations.

SL000
8,000+-
'LOOOl-
(i000—
4,000,-
1&OOO---_- -.- - -_
2,«--~-~000...- - '_, ,
1,000«-_.'_,_e _____ _ ____________

 

Appropriations (millions)

 

 

 

 

1969
1971
1973
1975
1977 .
1979
1981
1983
985.
1987
1989
1991
1993 1
1995
1997
1999

Year
EFI- - ifigherEducafion

 

 

Coneflhmsj

 

Figure 3.2: State of California: Corrections and Higher Education Appropriations (GF):
FY 1969-1999

A visual inspection of Figure 3.3 reveals that between FY 1969 and FY 1993 the percent

of the state budget appropriated for corrections has increased while the percent allocated

towards higher education has declined. Corrections grew from 3.7 percent of the state

budget in FY 1969 to 8.7 percent of the budget in FY 1993. Over this same time period

the percentage for higher education declined from roughly 16.8 percent to 12 percent. In

contrast, between FY 1994 and FY 1999 higher education grew to 12.6 percent of the

56

state budget, while corrections declined to 7.4 percent. While the existence of an
apparent tradeoff between corrections and higher education may manifest itself in a
depiction of state budgetary categories as a percentage of the state budgetary total, an

examination of empirical results is warranted before such conclusions are reached.

 

 

 

 

 

Q

C

.2

‘6

'C

3

‘2‘

2'!

O

.—

‘6

g

Q.

0% rrfi *Yt‘lf
Greecesassaragrg
FPFPPFFFPFFFI—FFP

Year

 

 

{- - - Higher Education Corrections];

 

 

Figure 3.3: State of California: Percent of Appropriations from Total State
Appropriations (GF): FY 1969-1999

Michigan — Appropriations
An examination of Figure 3.4 reveals that General Fund/General Purpose
(GF/GP) appropriations for both corrections and higher education rose between Fiscal

Year (FY) 1969 and FY 2000.

57

Appropriations (millions)

 

 

 

.‘,__..T._E.__.

f l l

s- 8 IO N
8: or 81 83
‘- ‘- s- 1-

f
58‘0" O)
mm888
‘- ‘-

g-

Correctional

1973
1975
1 977
1 979

Fisegl 1391* _
L— - - Higher Education

 

 

 

Figure 3.4: State of Michigan: Corrections and Higher Education Appropriations
(GP/GP): FY 1969-2000
During this time period, appropriations for corrections rose over 5,400 percent from
approximately $26.9 million in 1969 to $1.49 billion in FY 2000, while appropriations
for higher education rose over 445 percent from $307.8 million to $1.68 billion over the
same period. Thus, it can be seen that appropriations for corrections were growing at a
faster rate than appropriations for higher education. However, to gain an understanding
of the impact of the size of the entire state budget on growth, for specific departments, it
is useful to examine the level of appropriations as a percentage of the total GF/GP

appropriations.

58

25%

 

20%---- _,

15% -— ~ -

10% -

Percent of Total Appropriations

5%

 

 

 

 

0
0 O . r r r 1 . ' rrrrrr

r ' r r t V r r r Y fl

excessesaaarasrs

gammmmmmmmmgmmgim

FFFPPFv—PPFPPFPFF
FiseaIYear

 

l? - - Higher—Education Corrections 1

 

 

Figure 3.5: State of Michigan: Percent of Appropriations from Total State Appropriations
(GF/GP): FY 1969-2000
A visual inspection of Figure 3.5 reveals that between FY 1969 and FY 2000, the percent
of the state budget allocated towards corrections grew significantly fi'om roughly two
percent to 16.1 percent of the entire state budget. Over this same time period the
percentage for higher education declined from roughly 22.6 to 18.2 percent. Despite the
presence of a negative correlation between the two trends, one cannot infer that there is a
tradeoff between appropriations for corrections and higher education. Nevertheless, the
Michigan Council on Crime and Delinquency, citing studies from fiscal analysts within

the Michigan House of Representatives, assert that “the dramatic expansion of the

59

corrections industry in Michigan has resulted in destabilization of other state budgets as
monies are taken from other areas to fund corrections” (Michigan Council on Crime and
Delinquency, 10).7 This apparent change in priority is one issue to be explored in the

data analysis that follows.

Virginia — Appropriations

General Fund appropriations for corrections grew from roughly $13.5 million in

FY 1971 to over $573 million in FY 1999. This equates to a grth rate of 4,148%. :-
Over this same time period general fund appropriations for four-year colleges and
universities (higher education) grew from $85.6 million to $819 million, or a percentage
increase of 856%. Appropriations for corrections grew at a slow and steady rate until FY
1976 when the rate of grth increased. Another increase in the rate of growth occurred
in the late 1980’s and early 1990’s. Overall, the series for corrections seems to increase
without much fluctuation. In contrast, the series for higher education appropriations is
characterized by a period of decline. Appropriations for higher education increases at a
fairly steady rate between FY 1971 and FY 1978, at which point the rate of grth
increases. This growth hits its peak in FY 1990 when appropriations were roughly $640
million. This was followed by a period of decline where appropriations fell to under
$550 million in FY 1993. The next three fiscal years were characterized by limited

growth until FY 1997 when appropriations for higher education increased at a very rapid

 

7 In 1978 Michigan adopted the Headlee amendment to the Michigan Constitution. Article IX, Section 29
of the Constitution implies that the state cannot reduce financing to any existing activity or service required
of units of local government by state law. It mandates that the proportion of total state spending paid to all
units of Local Government taken as a group, shall not fall below 41.61%.

60

pace. Between FY 1996 and FY 1999 appropriations for higher education increased

from $588 million to $819 million, or an increase of 40% in the amount appropriated.

 

800- _____ _ - - - - -.-_.___5

700 .e

600-

400 4»

200 -

Appropriations (millions)

100«

 

 

 

 

 

1973
1975 -
1977
1979
1981
1983
1985
987
1989
1991
1993
1995 ,
1997
1999

1971

Fiscal Year

 

 

 

l;- - Higher Educafio‘n Correction;

 

Figure 3.6: State of Virginia: Corrections and Higher Education Appropriations (GF): FY

1971-1999

In order to get a sense of the potential tradeoff that might exist between
appropriations for corrections and appropriations for higher education, a visual inspection
of Figure 3.7 is warranted. Figure 3.7 illustrates appropriations for corrections and higher
education as a percentage of the total state budget. The percent of the state budget
appropriated for corrections rose gradually between FY 1971 and FY 1999 from 1.6% to
6.5%. In contrast, higher education declined from 9.9% in FY 1971 to 9.2% in FY 1999.
However, the percent appropriated for higher education hovered between approximately

10-11 percent between FY 1971 and FY 1990, with a high of 11.7% in FY 1990.

61

Beginning in FY 1991, the percent of appropriations for higher education declined to a
low of 7.8% in FY1996. Finally, the percent appropriated for higher education rose at the
end of the 1990’s. The end result is that the gap between the two budgetary categories

closed, and as visually expressed in Figure 3.7, a bottleneck shape characterizes the two

 

 

 

 

 

series.
14%
a 12%...- . -. .-,_ -.-_..- _ __.__.___
g ’~va"-‘~—~_
“-1 -‘
:100/0L.f._-’ ,__ \ —» _
O. s
O \ I
ii. \
2 gin/0.--..- -. - -- - -. - _-.-.--\.-_'-r.£'...
g 6%.
‘6
E 4°/o~
11’ 2%-
0% I I r , 7 [ . T r , r l’ r r r frfir
: a a 1: e a a a a a '- a .0 c g
or c» or or or or or or a: or Si Si Si 83
v- 1- s- v- 1- v- P I- 1" v- F 1- v- F ‘-
FlsealYear

 

l: - :Higher Echation Corrections/’71

 

Figure 3.7: State of Virginia: Percent of Appropriations from Total State Appropriations
(GF): FY 1971-1999

62

Estimation Strategy

The model of the budgetary process for corrections and higher education
represents a system of simultaneous equations. The equations are simultaneous because
endogenous variables in one equation appear as explanatory variables in another equation

of the system (e.g., C. and E.).

C: = bro + b11E1+ b12(Pr - Pr — 1) + br3Unemp :— r + b14Cohort : — r + brsGovernor z —— 1 +

P
b16ViolentCr r— 1 + z lyprz +811
p:

E: = b20 + b21Cr + b22(Pz — Pr - I) + b23Unempz -1+ b24Cohort t -1 + bszovernor r-1+
b26Enrollment r- l + e2:

As a result, the endogenous variables are correlated with the stochastic disturbances (e)
and applying ordinary least squares (OLS) results in biased estimators (Kennedy 1993,
151).8 Therefore, the equations must be estimated simultaneously. However, before the
estimation procedure can be initiated the problem of identification must be addressed.
The identification problem asks if one can obtain unique numerical estimates of
the structural coefficients. The identification problem is a modeling problem and can be
resolved by specifying an appropriate set of restrictions. Gujarati notes that “[flor an
equation to be identified it must be shown that the given set of data will not produce a
structural equation that looks similar in appearance to the one in which we are interested”
(659). In other words, additional information is needed in each equation to set it apart

from the other, in order for parameter estimates to be obtained. This can be

 

8 OLS assumes that causation is unidirectional, an assumption that is violated when there is endogeneity or
feedback in a system.

63

accomplished by utilizing the order condition9 which states that in order for an equation
to be identified, the number of predetermined variables (exogenous) excluded from the
equation must not be less than the number of endogenous variables included in that
equation less one (Hanushek and Jackson 1977, 665). Thus, there needs to be enough
exogenous variables excluded from the equation to provide an instrumental variable for
each of the endogenous variables appearing as regressors in that equation (Kennedy

1992, 165). In addition, the instrumental variables need to be uncorrelated with the error
term. A failure to find instruments that are not correlated with the error term could lead
to estimates that are biased and inconsistent. Furthermore, it is possible that collinearity
may arise because the instrumental variables act as a linear combination of the other
variables. As such, Wooldridge (2000) notes that multicollinearity can be more
problematic for 2SLS than OLS. Once the model is identified and potential pitfalls with
selecting an instrument have been avoided, the model can be estimated. The instrumental
variables that are used in the present study are enrollment for the estimation of
corrections and the violent crime rate and the policy variables for the estimation of higher
education.

The simultaneous equation model is estimated by utilizing two-stage least squares
(ZSLS).lo By applying two separate applications of OLS, the correlation between the
endogenous variables and the error terms is eliminated. The estimation procedure works
by regressing the endogenous variable being used as a regressor on all exogenous
variables in the system. The predicted value of the endogenous variable is then used to

estimate the structural equation. This process ensures that the parameter estimates are

 

9 The order condition is a necessary but not sufficient condition.
'0 zsLs will be accomplished by utilizing Stata and the ivreg function.

64

unbiased and consistent and that the instrumental variables are not correlated with the
error term (Kennedy 1992, 159). An example of the procedures surrounding 2SLS
follows.

In order to estimate the higher education equation, that contains the endogenous
corrections variable, identification must first be achieved. This is accomplished by
utilizing variables not included in higher education equation, the violent crime rate and
the corrections policies, as instrumental variables. These instruments need to be
correlated with the endogenous corrections variable and not correlated with the error term
of the higher education model. Once the instruments have been selected and the order
condition satisfied, estimation can proceed. The estimation procedure occurs in two
steps. In this example, the first step estimates the fitted values of the reduced form of the
corrections equation. This stage of estimation purges Q of its correlation with the error

term in the higher education equation. The reduced form is as follows:

C: = I'Iro-i- I'Ir 1E: — 1 + H12Enrolh-1+ I'Ir3(P: — Pr - 1) + 1'114VCr-1 + I'ItsGov: -1+
IIréUnemr - r + I'Ir7Cohr - r + HrsDeterr - 1 + 1119817er -1 + III IoThreer — r + e1:

In the reduced form equation, the endogenous variable higher education has been
removed and the exogenous variable enrollment has been added. For estimation to be
unbiased and consistent, unemployment and the policy variables (Deter, BillR, and
Three) must be correlated with C,. In the first stage, OLS regression produces fitted
values for Ct. These fitted values are then substituted into the higher education equation
in place of C,. The second stage estimates appropriations for higher education using
OLS. For each state under study in this dissertation, results of two separate regressions

of the second stage of the 2LS estimation are presented. The first regression includes

65

corrections as the dependent variable and the second includes higher education as the
dependent variable. If all previously stated conditions are met, all aspects of OLS
interpretation apply to the second stage of ZSLS.

Once the 2SLS results have been generated, separate OLS regressions are run for
both the corrections and higher education models with the endogenous explanatory
variables removed from the equation. Estimating OLS equations absent the endogenous
explanatory variables allows for a specification test to be conducted that explicitly tests
for simultaneity. As such, the Durbin-Wu-Hausman test is used to examine if the
difference in the coefficients between the instrumented and OLS model are in fact
systematic. If the null hypothesis (difference in coefficients are not systematic) can be
rejected, then it can be inferred that there is a simultaneous equation bias and the standard
OLS estimates are inconsistent. If the null cannot be rejected then the OLS estimates are
consistent and explicit interdependence between appropriations for corrections and
appropriations for higher education can be abandoned.

The Durbin-Wu-Hausman test could lead to three potential outcomes in the
present study. The first scenario is where it is found that there is systematic bias in the
estimation of both corrections and higher education using OLS. In this case, ZSLS is the
most applicable procedure. The second scenario is when there is no systematic difference
in the estimation of appropriations for corrections and appropriations for higher education
using OLS when compared to 2SLS estimation. Finally, the third scenario is one in
which one of the OLS equations (e.g., higher education) is found to have systematic
differences and the other equation does not (e. g., corrections). This recursive system is

one in which there is unidirectional dependency among the endogenous variables. In the

66

example provided, appropriations for corrections is found to be a determinate of
appropriations for higher education, but appropriations for higher education is not found
to be a determinate of appropriations for corrections. For scenarios two and three explicit
interdependence can be ruled out and ZSLS is no longer applicable. However, the
equations in scenarios two and three might still be connected through their error terms.
Therefore, when 2SLS is forsaken, an alternative view of interdependence, seemingly
unrelated regressions, is pursued.

When explicit interdependence between two series has been ruled out it is still
possible that the error terms of the two equations are correlated. Equations that possess
these conditions are called seemingly unrelated regression (SUR) systems, because at
first look they appear to be unrelated, but in fact the equations are related through the
correlation in the errors. The presence of contemporaneous cross-equation error
correlation leads to least squares estimates that are biased and no longer efficient.
Efficiency can be improved by estimating the equations as a set, using a single large
regression.ll Beyond asymptotically more efficient estimates, the SUR procedure
generates a correlation matrix of the residuals between the two series and the Breusch-
Pagan test. This test is asymptotically distributed chi-square and is used to test the null of
no correlation between the residuals and to see if the variance-covariance matrix of the
residuals is diagonal. Since the residuals of a regression analysis represent the
unobserved or stochastic element of the model, one could assume that if the residuals of
the two series were significantly correlated then what is driving the unexplained portion

of the models are a function of similar processes or variables. If they are not correlated,

 

” Zellner and Huang (1962) point out that when both equations contain the same variables there is no gain
relative to OLS.

67

we can then assume that the two series are unrelated and being driven by separate
processes (e. g., corrections does not influence higher education and vice versa) (Simon,
Ostrom, Marra, 1991). As such, the seemingly unrelated regression approach provides an
unobtrusive way of looking at the relationship between corrections and higher education
and the impact of policies and exogenous factors on the budgetary process.

Now that an estimation procedure has been laid out that allows for the various
hypotheses to be tested, a discussion follows that provides an overview of issues
associated with using time series data. The first issue to be covered is the lag structure of
variables in time series followed by issues of stationarity, spurious regression, and unit

roots.

Time Series Issues — Lags

This study makes use of time series data to investigate dynamic causes of the
budgetary process12 and as such necessitates that the lag structure of the variables be
discussed. When considering the impact of various factors on budgetary appropriations
the dependent variables (corrections and higher education) are treated
contemporaneously. As far as the independent variables are concerned it makes no sense
to treat them as contemporaneous, thus the variables must be correctly specified in
regards to time.13 In other words, the appropriate lag structure must be determined.
Therefore, Tucker (1982) asserts that it is important to think about the relation variables

may have in regards to the budgetary cycle. An example, adopted fi'om Tucker (1982) is

 

'2 Utilization of time series data allows for moments of the distribution to change over time.
'3 This holds for all independent variables except for the endogenous variables of corrections and higher
education.

68

utilized to show how variables from earlier time periods (t -— n) affect budgetary

appropriations at time t.

 

 

 

 

 

 

 

 

 

FY I991 FY I993 FY 1994
Apprapr APP"!!! Appropr
FY I99I Eapeadllurea FY I991 Expenditures FY I993 Expenditures
n... + r : : : : I: I“

 

July January July January 1", January July
I990 I99I I99I I992 199: I993 I993

Figure 3.8: An Illustration of Time and the Budgetary Cycle, Appropriations and
Expenditures
The figure above shows the periods of budgetary negotiations surrounding
appropriations and the periods of government expenditure. The period of budgetary
negotiations and requests for appropriations for a specific fiscal year pre-dates the
subsequent fiscal year expenditures. Tucker notes that
approximately 12 months before the fiscal year or biennium begins,
individual agencies start to develop their budget requests. A few months
later these requests are forwarded to officials — usually agents of the
governor — who are authorized to develop a comprehensive budget. The
comprehensive budget is presented to the state legislature early in its
session, and legislative appropriations are passed before the fiscal year or
biennium begins... Expenditures take place throughout the fiscal year or
biennium (177).
Therefore, when considering the impact of environmental stimuli (exogenous and
political factors) on the budgeting process one must decipher what information is

available at the time of the negotiations. Ostrom and Marra (1986) note that

“[i]nformation having utility for decision makers must be available at the time decisions

69

are made” (827). In the figure above it can be seen that the earliest that information
would be available for legislators and policymakers when determining appropriations for
FY 1992 is 1990. So for example, increases in the unemployment rate in 1990 may
impact budgetary appropriations in FY 1992. Similiarly, appropriations in FY 1993
would be impacted by factors in 1991. Thus, the exogenous variables and impact of
public policies in this study will be lagged two years to capture their impact on the

budgetary process. '4

Time Series Issues — Stationarity

An additional issue to consider when using time series data is the concept of
stationarity. Before modeling and estimation of a time series can take place one must
achieve stationarity. Stationarity means that the properties of the distribution are
unafi'ected by changes in time origin (Mills 64). This implies that the sample moments of
the mean and variance must be unconditional, finite, and independent of time and finally
that shocks are transitory. '5 The presence of nonstationary data makes the use of
statistics such as the t and R2 inappropriate and the running of regressions with this data

may produce spurious results (Kennedy 252).

 

:4 An alternative way to determine the lag structure would be to utilize reduction theory to move fiom a
general to a specific framework. This framework employs various diagnostics in order to drop lags to
obtain models that are theoretically and statistically defensible. Granato (1991) notes that “reduction
theory affords the analyst a rigorous methodological basis for reducing the number of right-hand-side
variables with no loss of relevant information to the parameters of interest” (125). While it is desirable to
include a multitude of lags for each variable, the small number of data points in this study creates a degrees
of freedom problem and thus makes this impossible in this study.

'5 Stationarity that focuses solely on the first two moments is referred to as second-order or weak
stationarity. In contrast, strict stationarity is characterized by a joint distribution that is the same
throughout the series. When normality is assumed, weak and strict Stationarity are identical. Nevertheless,
meeting the conditions of weak stationarity is sufficient for most time series analysis.

70

Spurious regression results whenever two or more variables are regressed on one
another that are not 1(0). 1f the dependent and explanatory variables possess a stochastic
trend, the fit of the estimated equation appear to be meaningful, even if the variables are
actually unrelated. The large t-values and R2 are therefore spurious. Granger and
Newbold (1974) regressed two variables that were independent random walks on one
another and found via simulation that the slope coefficient had a significant t-value 77
times out of a hundred when significance was only anticipated five percent of the time
(Type I error). The authors concluded that regressions attempted with nonstationary data
yield invalid standard errors and test statistics. Therefore, in order to begin a process of
model generation, it must be ensured that the stochastic properties of the dependent
variable are invariant with respect to time, which enables certain asymptotic properties of
regression estimates to hold.

Several processes are undertaken to ensure that the series are stationary. The first
is accomplished through a visual inspection of the data. Plotting the individual series
against time in levels allows one to inspect for trends in the mean or variance, seasonal
components, or long swings associated with unit roots. A stationary series is one that
visually is flat without a time trend, lacks a change in variance, and has no periodic
fluctuations. Another method undertaken is the Box-Jenkins approach, which relies on
the inspection of correlograms and resultant autocorrelations and partial correlations to
deduce whether series are stationary and error terms are white noise. For example, a

stationary AR process possesses autocorrelations that die out rapidly over time.'6 If

 

'6 Autocorrelation functions (ACF’s) measure the extent to which one value of the process is correlated
with previous values. It focuses on the strength and memory of the process (Mills 78). Partial
autocorrelation functions (PACF’s) measures the correlation between time series observations that are k
periods apart after controlling for correlations at intermediate lags (Gujarati 739).

71

stationarity is not deduced then a series of transformations are undertaken. Generally
speaking, taking the natural log of a series is a way to stabilize variance and differencing
the data is a way to remove trends and eliminate systematic changes in the mean. First
differencing is accomplished by subtracting the lag of a series from the contemporaneous
series.

AYt = Y, - YH
If data are integrated of the first order, then the operation of first differencing transforms
an 1(1) series into a stationary series.

A specific case of nonstationarity is the unit root. A unit root is characteristic of a
series that possesses a random walk with drift and is integrated of the first order. It
implies that the series of the dependent variable does not possess an unconditional mean
and variance. Additionally, unit roots are characterized by long stochastic swings
without returning to a constant mean level and the effect of shocks to these processes
persists indefinitely. An example of a unit root can be seen when viewing the following
first order autoregressive process (AR(1)).

Yt = pYH + ut
A unit root exists when p = 1 and stationarity exists when |p| < 1 and thus 1(0).'7 The
presence of a unit root process is problematic for statistical inference in that the standard t
statistics do not apply. Viewed in this light, tests for stationarity are tests for p = l and
thus tests for unit roots (Kennedy 253). A formal testing for unit roots are conducted

through an application of the Augmented Dickey-Fuller (ADF) test.

 

'7 When p > 1, the series will exhibit explosive behavior.

72

The ADF test statistically tests the data generating process for nonstationarity
versus stationarity. This test is a modification of the original Dickey-Fuller (DF) test that
is incapable of generating accurate results in the presence of serial correlation.
Conversely, the ADF test absorbs serial correlation that might be present in the data by
incorporating a lag of the variable under study plus lags of the first difference of the
variable. This addition to the DF test allows the autocorrelated error term to remain free
of the effects of serial correlation. Therefore, when utilizing the ADF with MacKinnon
critical values for the rejection of the null hypothesis, one can make an accurate
assessment of whether or not a unit root exists. The ADF test applies the following
restrictions:

1) no intercept, no trend

AYg = ¢Yt-| + szYI-i + 8: Where:
Ho: 4) = 0; unit root (random walk)
HA: 4) at 0; no unit root (stationarity)
2) intercept, no trend
AY, = or + ¢YH + ZpAYH + at where:
Ho: 4) = 0; unit root (random walk with drifi)
HA: 4) at 0; no unit root (stationarity)
3) intercept and trend
AYt = ¢YM + ZpAYH + a, where:
1-10: (1) = 0; unit root (random walk with drift around a deterministic trend)
HA: 4) at 0; no unit root (stationarity)
Unit root tests are applied to all continuous variables to ensure that the variables are 1(0)
in order to achieve balance and avoid spurious regressions.

In order to make the procedures associated with stationarity clear, an example

utilizing enrollment in public four-year institutions of higher education in California is

outlined. Table 3.5 illustrates the output for the ADF test under three restrictions.

Enrollment in levels is significant at the .10 level under restriction two and .05 under

73

restriction three. This suggests the presence of a unit root cannot be rejected and the
series needs to be transformed to induce stationarity. When the first difference is taken,
the ADF statistic is significant at the .01 level for all three restrictions suggesting that the

null of a unit root can be rejected.

Table 3.5: Augmented Dickey-Fuller Test: Unemployment Rate in California

 

No Trend, No Intercept, No Intercept and
Intercept Trend Trend
Enrollment .901 -2.644* -3.948**
Enrollment (d1) -4.44*** -4.504*** -4.366***

 

The effect of first differencing can be seen in Figure 3.9. By first differencing the
emollment rate becomes stationary. This series is visually flat, fluctuates around a mean

of zero, lacks a time trend, and has no periodic fluctuations.

 

 

 

 

 

600.000
Pu .
500.000 “ ** " A“ -- —- ~—-'§-...{ _. ‘... ._‘ —_' _
'I’ §."-‘.-'

400,000.71-— . -- -._
isoopoorw—m ,1 s- .......
2200,0004- 4 ~ ~ —»- ~ 7—- ~~~~~
Ill

100,000«L —————— , H __________________________

M A
0 f k/Av Av V“

-100,000 ,

O N V (D co 0 N (D N ‘0 Q
FlscalYear

'- - - Enrollment

 

 

Enrollment difl‘ferenced .

 

Figure 3.9: Illustration of Stationarity via Differencing: Enrollment in California

74

Inducing stationarity via the transformation of variables by differencing is
undertaken to avoid spurious regression. Nevertheless, some scholars have claimed that
differencing can also lead to potential problems or limitations. For example, King (1989)
notes that “differencing can cancel out the systematic component [relating one series to
another]...Thus, as a nearly general rule, models based on differenced series tend to fit
less and have higher standard errors and less stable coefficient values” (181). In addition,
Durr claims that many critics of ‘prewhitening’ techniques argue that transforming
variables leads one to “throw out the baby with the bathwater” (Durr, 164). Finally,
Gujarati suggests that “[m]ost economic theory is stated as a long-term relationship
between variables in level form and not first differenced” (Gujarati, 725). To account for
these concerns and to aid in the interpretation of the estimated coefficients, models are
estimated using data in differences and in levels. The results of both estimations are
presented and a discussion of interpretation surrounds each, while the primary focus is

placed on the outcome that is most plausible.

Estimation Strategy - Overview

In conclusion, the estimation strategy proceeds by following two paths. The first
path represents the estimation of the system of equations in levels and the second path the
estimation of a transformed system of equations to deal with the issue of stationarity.
The overall estimation strategy is presented in Figure 3.10. As can be seen path two

begins by checking for stationarity and the application of unit root tests. Where

75

14-14 .A..w ~._-... _...- .
r. ;.L

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Path One Path Two
.R i
4 N0 UM oot Unit Root Test
Differencing
v i
ZSLS 2SLS
+ __ +
Durbin Wu Hausman Test for Sirnultaneity Bias
t i
6 + + + J i
Reject Accept Accept/ RCJCCI Accept Accept/
Null Null Reject Null Nu" Null Reject Null
i i i i i i
ZSLS SUR SUR ZSLS SUR SUR
1 l (recursive) l (recursive)

 

Collinearity Diagnostics

 

i

l + + l + l +, l y l _
Correction Correction Correction Correction Correction Correction
i l i y i i i

Autocorrelation

/\ /\ /\ /\ /\ /\
N Y N Y N Y N Y N Y N Y
+ + + l t. l + l t
Correction Correction Correction Conectron Correction Correction

1 i l i l i i i i

Heteroskedasticity

/\ /\ /\ /\ /\ /\
N Y N Y N Y N Y N Y N Y
i i l vi 1 i l i l i
Robust Robust Robust Robust Robust Robust
l i l i i i i i
Bruesch — Pagan Test

Bruesch — Pagan Test

Figure 3.10: Overview of Estimation Strategy

76

necessary differencing is undertaken to remove unit roots and avoid spurious regression.
Once stationarity has been confirmed in path two, both paths proceed by estimating a
system of equations using 2SLS. These results are compared to separate OLS estimates
using the Durbin-Wu-Hausman test. If the OLS estimates are found to be systematically
different, then the OLS estimates are considered inconsistent and in order to avoid
simultaneity bias, ZSLS is used as the estimation strategy. If the results are found not to
be systematically different, seemingly unrelated regressions are utilized to examine
potential correlations that exist among the residuals for either recursive or non-recursive
systems.

Following the initial estimation the models will be investigated for
multicollinearity. Multicollinearity is a term that refers to correlation among the
independent variables in a multiple regression model. Collinearity itself has to do with
specific characteristics of the data matrix and as such is a data problem, not a statistical
problem.‘8 In the presence of multicollinearity, the measured variables are too
intercorrelated to allow precise analysis of their individual effect (Greene 418). As the
degree of multicollinearity increases, the estimates of the coefficients become unstable
and the standard errors of the coefficients become inflated. While the estimates remain
unbiased, the tests of significance are impacted as the confidence intervals tend to be
larger and t-ratios tend to be insignificant. Belsley, Kuh, and Welsch (1980) comment
that “collinearity is a nonstatistical problem that is nevertheless of great importance to the
efficacy of least-squares estimation” (86). Testing for the severity of collinearity is

accomplished by utilizing the col/in command in Stata. This diagnostic procedure

 

'8 Collinearity implies that two variables are near perfect linear combinations of one another. When there is
a perfect relationship among the predictors, the estimates of a regression model cannot be uniquely

77

generates both VIFs (variance inflating factor) for each regressor and a condition
number.19 Following the recommendations of Belsley, Kuh, Welsch collinearity is
considered a problem if any variable possesses a VIP value in excess of ten or the
condition number is large (30 or higher). Potential solutions to multicollinearity include
increasing the sample size or omitting the problematic variables (Gujarati 1995).
Following any corrections to account for multicollinearity, the issue of autocorrelation is
addressed.

One of the classical regression assumption holds that the disturbance term relating
to any observation is not influenced by the disturbance term relating to any other
observation (Gujarati). This implies that Corr(ut,us)=0, for all t not equal to 8. When this
assumption is false, the errors suffer from serial correlation or autocorrelation. In the
presence of autocorrelation the estimated coefficients remain unbiased, but the standard
test statistics are no longer valid as variances are inflated. Kennedy (1992) suggests that
autocorrelation may arise due to prolonged influence of shocks, inertia, or
misspecification (l 19). The most common form of this problem is first order
autocorrelation. This occurs when the disturbance in one time period is a proportion of
the disturbance in the previous time period plus a spherical disturbance. Testing for
autocorrelation is accomplished by examining the Ljung-Box Q statistic at various lags
for the residuals. Corrections to account for autocorrelation are achieved by estimating

the equation with the Cochrane-Orcutt iterative procedure.20 Once the issue of

 

computed.

'9 The condition number of a matrix is the square root of the ratio of the largest to the smallest
characteristic root.

2° Cochrane-Orcutt estimation proceeds by regressing the OLS residuals on themselves lagged one period.
This produces an estimate of rho which is used in an iterative procedure to transform the variables until
autocorrelation is mitigated.

78

autocorrelation has been addressed one additional post estimation issue,
heteroskedasticity, is examined before inferences can be made.

Heteroskedasticity refers to a condition where there is non-constant error
variance. This implies that the error variance changes across observations and the
assumption that the variance of each disturbance term 11, is some constant equal to 02 is
violated. 1n the presence of heteroskedasticity the estimators remain unbiased, yet they
are no longer efficient and the standard errors are overestimated. This may result in
statistically insignificant coefficients, which may in fact be significant if the assumption
of homoskedasticity is met. Testing for heteroskedasticity is accomplished by utilizing
the hettest command in Stata, which is an application of the Cook-Weisberg test for
heteroskedasticity using fitted values, and examining a plot of the residuals for non-
constant variance. If non-constant error variance is found, heteroskedatic consistent
estimators are used by applying robust estimation.

Finally, when all of the diagnostics tests have been met, inferences are made
deriving from the estimation of equations both in differences and in levels. As such,
hypotheses are tested using the results from the estimation of the models in differences
and in levels. These results are interpreted and directly compared and discussed in terms
of their plausibility. Generating two sets of estimates using different procedures leads to
several questions that need to be addressed: what is to be gained by using two different
paths? What does it mean if the results generated from the two procedures generate
similar results? Different results? On what grounds does one choose between competing

results?

79

Similar to inquiry into any phenomenon, when conducting empirical studies there
exists the possibility of equivocality, or multiple interpretations and explanations for a
single phenomenon (Weick 1995). Cherryholmes states that it “is possible to describe
and interpret any event and object in as many ways as there are describers and
interpreters,” however, “[i]t is not possible to treat all beliefs as equally valid” (44).
Accordingly, the role of a researcher as an interpreter of phenomenon is to ascertain the
meanings and implications of a set of materials (Kritzer 2) that are acceptable,
approximating, translations (Mailloux). 2' Kritzer asserts that in order to be useful for
social inquiry, data must be interpreted. In this study of budgeting two paths of inquiry
have been prescribed. The two approaches are rigorous in nature, have been chosen as
the most applicable for the present research, and are potentially equally valid for the
reasons outlined above. However, the question still remains why would one choose to
undertake more than one approach?

To answer the question why multiple approaches would be pursued Kritzer (1996)
suggests that one way to improve interpretation is to use a triangulation approach. This
notion is built on the idea of pinpointing a location by sighting it from different
perspectives (12). He goes on to note that “by stretching the limits of the data or the
techniques it is possible to arrive at new insights and perspectives that aid in interpreting
the results of the overall analysis” (23). Therefore, the two estimation paths in this study
represent a design for triangulation to make sense of the factors that influence budgetary

appropriations.

 

2' Foucault argues that “each society has its regime of truth that is, the types of discourse which it
accepts and makes function as true; the mechanisms and instances which enable one to distinguish true and
false statements the techniques and procedures accorded value in the acquisition of truth”
(Cherryholmes 15).

8O

Karl Weick (1995) argues that sensemaking is a social process oriented around
making sense of what we see. He argues that “[s]ensemaking is about plausibility,
coherence and reasonableness. Sensemaking is about accounts that are socially
acceptable and credible” (61). Weick argues that sensemaking is not built around
accuracy rather it focuses on explanations that are plausible. The research design as
presented represents a vehicle to make sense of the budgeting process. If the two paths
produce results that support one another then they will triangulate or point us towards a
plausible explanation. When convergence occurs between the two paths, one’s
confidence in the conclusions increases. However, if they are not complimentary how
will the plausibility of one set of results be established? Or as Menand asks “what makes
us decide to do one thing when we might do another thing instead” (35)? The answer to
this question can be found in the teachings of pragmatism.

Pragrnatists are first and foremost interested in results and consequences of an
action. Pierce’s maxim states that the meaning of a concept is found by “trace[ing] out in
the imagination the conceivable practical consequences of the affirmation or denial of
the concept” (Cherryholmes 15). For pragmatists, the meaning of a concept or policy
rests solely on the conceivable practical consequences. Pragrnatists choose between
different alternatives based on the value of the practical consequences and on the basis of
satisfaction and fulfillment (52). Richard Rorty proclaims that for the pragrnatist “the
pattern of all inquiry — scientific as well as moral — is deliberation concerning the relative
attractions of various concrete alternatives” (Cherryholmes 117). Similarly,

Cherryholmes suggests that “[p]ragmatists are interested in consequences that are

81

satisfying and desirable. They act as artists as they contemplate outcomes and critics as
they assess them” [emphasis removed] (124).

Therefore, when faced with competing interpretations this dissertation seeks
guidance from the teachings of pragmatism and scholars who have studied sensemaking.
Operating within the contextual framework of political science, policy analysis, and prior

studies of budgeting, this dissertation seeks out interpretations that are most plausible and

.«f.

‘provide the greatest satisfaction.’ The empirical methods and triangulation process that g;
characterize the analysis are a means to an end. They are tools to make sense or interpret if:
phenomenon. Utilizing two paths could potentially lead to four outcomes: both

approaches agree in which variables are statistically significant; only one approach

produces a significant coefficient for a particular variable; both approaches agree a

variable is not significant; and the two approaches lead to opposite conclusions. In the

first three cases, the results will be evaluated and conclusions about the hypotheses made

using the lens of triangulation. In the case where the results do not firlly converge and

one path produces significant coefficients, the variables are viewed as important

predictors. For the final case, plausibility will be the divining rod that guides the effort to

provide an understanding and explanation of the impact of various policies and

endogenous and exogenous factors on the budgetary process.

Anticipated Conclusions
An examination of the state budgetary process will reveal:

. A tradeoff relationship exists between appropriations for corrections and
appropriations for higher education.

82

Exogenous factors create policy relevant conditions that impact budgetary
allocations. Specifically, (I) an increasing violent crime rate leads to increased
appropriations for corrections; (2) increases in the size of the 18-24 year old
cohort leads to increased appropriations for corrections and higher education; (3)
the presence of a Republican governor leads to increased appropriations for
corrections; (4) increasing rates of unemployment lead to increases in the amount
appropriated for corrections and a decrease in the amount appropriated for higher
education; and (5) rising enrollment leads to increases in appropriations for
higher education.

An expanding budgetary pie leads to increases in appropriations for both
discretionary budgetary categories (corrections and higher education).

The implementation of specific public policies has the most impact on changing
budgetary priorities. Therefore, policies that are negotiated and implemented
outside of the formal budgetary process are most instrumental in causing changes
in state government appropriations.

83

CHAPTER 4

In order to test the various budgetary hypotheses and to examine the tractability of
the research program as outlined in the proceeding chapter, California is utilized as a case
study. This chapter investigates the appropriateness of two stage least squares, seemingly
unrelated regression, and issues of model selection. In order to accomplish such a task,
an overview of policy changes to corrections and a brief review of the contextual
background in California are necessary. This is followed by an extensive look into the

various methodological issues outlined above.

California — Policies

California has a long tradition of indeterminate sentencing dating back to 1917,
when laws were enacted granting discretion to the parole board (Adult Authority) and the
Women’s Board of Terms and Parole (Shane-Dubow, Brown, and Olsen, 33). Under this
system, judges sentenced offenders to very broad terms prescribed by law, which were
typically zero to life or five years to life in prison (Griset 1991, 47). Thus, the parole
board was the ultimate arbiter in deciding when offenders were deemed rehabilitated and
hence when they should be released. Pamela Griset states that the parole authority had
“unfettered discretion in determining sentence length” (47). The practice of
indeterminacy was radically altered by a series of reforms that occurred in the mid to late
1 970s.

In 1976 the Determinate Sentencing Bill (SB 42) was passed and became

effective on July 1, 1977. The Determinate Sentencing Bill represented a radical

84

--_ ‘ffi '

sex-h."

departure fiom early sentencing practices. The bill stated that “[t]he sole purpose of the
goals of the sentencing system should be the elimination of sentencing disparity and the
promotion of sentence uniformity” (Shane-Dubow, Brown, and Olsen, 33). The bill
effectively curtailed judicial discretion through the establishment of a narrow range of
prison terms set by the legislature. For each felony offense category A through D, three

sentences were established, a lower term, a middle term, and an upper term (see Table

1*-=‘~:g-

4.1). According to the 1977 law, the court had to impose the middle term (presumptive)

'3“!

unless there were aggravating or mitigating circumstances that might lead to the lower
term or the upper term being selected. Thus, a triad of sentencing choices was
established, with egregious conduct receiving the upper term, mitigating cases the lower
term, and average cases the middle term (Shane-Dubow, Brown, Olsen, 35). In addition,
enhancements from one to three years could be added to the base. Enhancements
pertained to either specific factors relating to the offense such as weapon use or the
infliction of great bodily harm or general factors relating to prior record and offenses
causing great property losses (34). For example, committing an offense while armed
with or use of a weapon would lead to the addition of one or two years (Ku 1980, 4).
Furthermore, the bill provided for good time credits up to three months for good
institutional behavior and another month for participation in prison programs (Cohen and
Tonry 1983, 357). Good time could be lost only for serious misconduct (Griset 1991,
57). Finally, SB 42 eliminated the Adult Authority and the Women’s Board of Terms
and replaced it with the Community Release Board that was responsible for the
revocation of parole, the review of lengths and conditions of parole, and the denial of

good time (Shane-Dubow, Brown, and Olsen, 33).

85

Table 4.1: California Determinate Sentencing terms (SB 42)

 

 

Offense Category Lower Term Middle Term Upper Term
A (e.g, murder second degree, rape
with force of violence) 5 years 6 years 7 years
B (e.g., robbery first degree,
burglary, rape) 3 years 4 years 5 years
C (e.g., robbery second degree, i
arson, assault with deadly ii.‘
2 years 3 years 4 years .
weapon) [
D (e. g., burglary second degree,
forgery, auto theft) 16 months 2 years 3 years

 

SB 42 was immediately amended in 197 8 by SB 709 that became effective
January 1979. The bill focused on increasing middle and upper terms for violent
felonies. Murder in the second degree was increased from 5, 6, 7 years to 5, 7, 11 years;
rape was increased from 3, 4, 5 years to 3, 6, 8 years; robbery in the second degree was
increased fiom 2, 3, 4 years to 2, 3, 5 years; and burglary in the second degree was
increased from 16 mos., 2, 3 years to 2, 3, 4 years (Shane-Dubow, Brown, Olsen, 36).

A second sentencing reform, the Victim’s Bill of Rights that was passed in 1982
as a referendum (Proposition 8 of 1982), represented an extension of the previous system
of enhancements. ' This law required an extra five-year term for each serious felony
conviction where the defendant previously stood convicted of a serious felony. Serious

crimes included: armed or unarmed burglary of a residence; robbery; felonies in which

 

' Since 1977, there have been several revisions to the level of enhancements. For example, SB 13 passed in
1979 and effective January I, 1980 required severe enhancements for sex offenders. The bill mandated a
ten-year enhancement for each previous incarceration for a violent sex crime when the defendant had two
or more prior incarcerations for a sex crime and the current offense was for a violent sex crime (Griset
1991,56)

86

the defendant inflicted great bodily injury on a person; felonies in which the defendant
used a firearm or a dangerous weapon; and the sale of drugs to minors (Griset 1991, 57).
Additionally, judges were restricted in choosing probation or suspended sentences for
persons who were convicted of murder, robbery, kidnapping, first-degree murder, or who
intentionally inflicted bodily injury. In effect, the Victim’s Bill of Rights forbade the
plea bargaining of any serious felony (F eeley and Kamin, 144).

The final major sentencing reform occurred in 1994 with the passage of Three
Strikes and You’re Out legislation m both houses of the state Legislature and the passage
of Proposition 184. The movement for this mandatory sentencing scheme for repeat
felony offenders grew out of public sentiment in favor of harsher penalties for recidivistic
acts following the murders of 12-year-old Polly Klaus and the daughter of Fresno
photographer Mike Reynolds. The text of the Proposition 184 states “[i]t is the intent of
the People of the State of California in enacting this measure to ensure longer prison
sentences and greater punishment for those who commit a felony and have been
previously convicted of serious and/or violent felony offenses.” The law states that if a
person has one previous serious or violent felony conviction, the mandatory sentence for
a new felony conviction (not just violent or serious) is twice the term otherwise required
under law for the new conviction. If a person has two or more serious or violent felony
convictions, the mandatory sentence for any new felony conviction (not just serious or
violent) is life imprisonment with the minimum term being the greater of 1) three times
the term otherwise required under the law for the new felony conviction; 2) 25 years; 3)
the term determined by the court for the new conviction (California Criminal Law

Observer). Projections evaluating the impact of Three Strikes and You’re Out have

87

suggested that the law will lead to increased prison populations and the need to increase
spending for corrections.2

The Department of Corrections estimated in 1994 that changes due to the
implementation of Three Strikes and You’re Out would result in additional state
operating costs of about $3 billion per year by the year 2003 and about $6 billion
annually by 2026. These amounts assume that a three strike’s policy will add about
270,000 more inmates to the state’s prison population than would have otherwise
occurred. To accommodate this large increase in prison population, the Department of
Corrections estimated that twenty-five new prisons would have to be built at a cost of
over $20 billion by 2026 (California Secretary of State’s Office). Similarly, a recent
study by Rand projects that corrections as a percentage of state funds (General Fund
appropriations) would increase from 9% in FY 1994 to 18% in FY 2002 as a result of
three strikes. The authors suggest that these increases would be offset by large decreases
in the percentage of state funds for higher education, which represented 12% of state
firnds in FY 1994 (Greenwood et al., 84). 3 Finally, a 1995 report by the California
Legislative Analyst’s Office states that

[w]ith the enactment of the Three Strikes legislation, the state is expected

to incur unprecedented growth in its prison population. If the state is to

accommodate this growth, several billion dollars of state funding will be
needed to finance the construction and operation of the additional prisons

 

2 As of July 30, 1998, 36,043 second strike offenders and 4,468 third strike offenders were convicted under
the new law (Campaign for an Effective Crime Policy 1998).

3 Accompanying an increase in corrections would be an increase in the percent of state funds for K-12
education. The authors suggest that funds will increase from 36% of the state budget in FY 1994 to 46% in
FY 2002. This growth would be driven by Proposition 98, which is written into the state constitution, and
sets the minimal levels for K-12 (Greenwood et al., 84). Propositions 98 and 11 define a minimum fending
level for K-14 education depending on enrollments, growth in personal income, and prior-year spending.
They also require that state support for K-14 education be at least enough to achieve the specified minimum
funding level when added to K- 14’s local property tax revenues (Carroll and Bryton 1996). Additionally, it
is projected that health and welfare spending would remain relatively constant since it is tied to federal aid
requirements (Greenwood et al., 84).

88

the state’s costs for expanding and operating its prison system will

require an increasing share of the state’s budget (Nicol 1996).
California — Prison Population

Between 1971 and 1997 California’s prison population grew dramatically from
17,474 prisoners to 157,547 prisoners. Over this same time period the incarceration rate
per 100,000 inhabitants grew from 87.4 to 475. However, an examination of Figure 4.]
reveals that up until 1979 the population remained fairly constant hovering around 20,000
prisoners. It was not until roughly 1980 that California experienced its meteoric rise in

prison population, with a rate of change of 541% between 1980 and 1997. As discussed

 

180,000
160,000 7» ~
140,000 r
120,000 «~
100,000
80,000 —

Prisoners

60,000
40,000 ~

 

20.000 .

 

 

 

o ......

1971
1973
1975
1977
1979
1981
1983
1985:
1987
1989
1991
1993
1995
1997

Year

Figure 4.1 : Number of Prisoners Under State Jurisdiction in California: 1971-1997

earlier, California dealt with this rapid increase by building 21 new prisons and adding
25,864 new employees to the Department of Corrections between 1984 and 1997

(Ambrosio and Schiraldi 1997). Even this increase in resources was not enough to deal

89

with the growth. In 1995, California operated 29 prisons and operated at 182 percent of

capacity (Edwards, 480).

California — Independent Variable — Governors

During the period of study California had five governors, three of whom were
Republicans. Between 1970 and 1975, Ronald Reagan, a Republican was governor. The
next governor was the Democrat Edmund G. Brown, who held office between 1975 and
1983. Two successive Republican governors succeeded Governor Brown. These were
George Deukmejian (1983-1991) and Pete Wilson (1991-1999). Finally, the Democrat
Gray Davis held office between 1999 through the present. As such, it can be seen that

California had a Republican governor during 21 years of the time under study.

California — Independent Variable — Unemployment Rate

During the time period under study the rate of unemployment in California was
very volatile, fluctuating between a rate of 5 percent and 10 percent. The peak years of
unemployment occurred in 1971 (8.8), 1975 (9.9), 1982 (9.9), and 1993 (9.4). As can be
seen in Figure 4.2 the rate of unemployment is very cyclical, with sharp upswings
followed by periods of sharp decline. The periods of growing unemployment occurred
between 1973 and 1975; 1979 and 1982; 1989 and 1993. It is during these periods of
rising unemployment that California faced fiscal pressure on their budgets. For example,
surging state economies in the 1980’s were followed by large state budget deficits. In
1991 California had a budget deficit of 6 billion dollars and between 1990 and 1994

California lost 868,000 jobs. It is hypothesized that rising unemployment rates will lead

90

to decreases in appropriations for higher education and increases in appropriations for

corrections.

 

 

 

 

 

a:

a

E

>.

.2

g

o -« L

c 4

D
2 - ,,,,,,,,,,,,,,,,,,,,,
0 ‘I I r r r I Y—rfi
o N v to co 0 N g (D 8 8 N w
is r» is N N co co co 8 g
a) a: a: a: a) a: or c» 03 a: a:
u- u- v- v- v- 1- 1- 1- ‘- I- r- a- 1- 1- ‘-

Flacal Year

Figure 4.2: State of California: Unemployment Rate, FY 1970-1999

California — Independent Variable - Violent Crime Rate

The violent crime rate in California rose steadily between 1970 and 1980. During
this period of time the rate of violent offenses known to police per 100,000 inhabitants
increased from 475 in 1970 to 894 in 1980 (see Figure 4.3). Between 1980 and 1984 the
violent crime rate declined to 763 only to increase again through the mid 1990’s. In
1992, the violent crime rate reached its peak at 1,120 only to begin a rapid decline in the
late 1990’s to a rate of 627 in 1999. The decline in the latter half of the 1990’s mirrors

nationwide trends in the decline of the crime rate.

91

1 .200

 

 

 

 

 

1.000.

‘3 800--~—---

K

i

U 600%

E

g 4004
200 _, . - -- ......
o s - . ......
§§§5§§§§§§§§§§§§

Fiscal Year

Figure 4.3: State of California: Violent Crime Rate (Part I offenses known to police per
100,000 inhabitants), FY 1968-1999

California - Independent Variable - Cohort

The size of the 18-24 year old cohort has increased from approximately 2.1
million individuals in 1968 to over 3.3 million individuals in 1999 (see Figure 4.4). The
increase in this cohort gradually inclined between 1968 and 1990, increasing by 62
percent over this time period. Between 1990 and 1996 the size of this cohort dropped

below 3 million, only to increase to 3.3 million in 1999.

92

 

4.000.000

 

 

 

 

 

3.500.000 ., ~ -- ——
3,000,000 « ~
182.500.0001
‘5'
02,000,000?
«1-
81,500,000er 7 ~ iiiii
1.000,000+
500,000»
0 . - ......... 4 .
eeezsssaasasaa§§
22929222222922FF
FlacalYear

Figure 4.4: State of California: 18-24 year old cohort, FY 1968-1999

California — Independent Variable — Enrollment

Enrollment in the California system of higher education grew roughly 25 percent
between 1970 and 1999. Fall enrollment in public 4-year institutions of higher education
increased significantly between 1970 (429,397 students) and 1975 (516,096), only to
decline through the late 1970’s. By 1990, enrollment grew to over 536,000 students,
followed by a period of decline in the early 1990’s to a low of roughly 490,000 students
in 1995. Finally, in the latter part of the 1990’s enrollment once again rose to a high of
537,357 students in 1999. As such, the enrollment series is characterized by periods of
growth and decay, but overall there is an upward trend in enrollment in 4-year public

institutions of higher education in California.

93

600.000

 

 

 

 

 

5500.000.

3

3400000. - ~ ,, _- ~ — - w- -

m 0

3

‘5

£300,000- ~ ~ .........

.5

£200,000 .. - .. . _ . t

3

I“100.000—.v.r.....,_.
or . WC L
O N V to m o N (O Q N w

FlscalYear

Figure 4.5: State of California: Enrollment in Higher Education, FY 1970-1999

California - Correlation Matrix

In order to assess the strength of bivariate relationships and to diagnose the
possibility of collinearity, which could lead to a biased estimation of the effects of the
independent variables on appropriations for higher education and corrections, a

correlation matrix is constructed.

94

Table 4.2: Correlation Matrix for Variables included in the Budget Model,
California

 

Corrections Higher Ed Budgetary Cohort Enroll ‘39:"! Unemployment

Appropriations Appropriations Pie (t-2) (t-2) 04) (t-2)

Corrections -- .922 .410 .486 .662 .781 -. l 74

Appropriations .000 .025 .007 .000 .000 .377

Higher Ed .922 __ .512 .688 .582 .791 -.261

Appropriations .000 .004 .000 .001 .000 . l 79

. .410 .512 .146 .009 .140 .236

Budgetary P": .025 .004 " .441 .963 .400 .227

.486 .688 .146 .496 .779 -.265

COM“ (“2) .007 .000 .441 " .005 .000 .156

.662 .582 .009 .496 .516 -.231

Em" “'2’ .000 .001 .903 .005 " .003 .220

Violent Crime .781 .791 .140 .779 .516 -- -.045

(t—2) .000 .000 .460 .000 .003 .8 l 4
Unemployment -.1 74 -.261 .236 -.265 -.231 -.O45 __

(t-2) .3 77 .032 .227 .156 .220 .814

 

Note: Pearson correlation statistic followed by significance level

The results shown in Table 4.2 illustrate that all of the continuous independent variables
are significantly correlated and correlated in the hypothesized direction with
appropriations for corrections and appropriations for higher education, except for
unemployment with corrections which is not significant and correlated in a direction that
is opposite of the hypothesized relationship. For example, appropriations for corrections
is positively correlated with the size of the budgetary pie, the size of the 18-24 year old
cohort, and the violent crime rate. Furthermore, appropriations for higher education is
positively correlated with the size of the budgetary pie, enrollment, and negatively

correlated with the unemployment rate. These results are in line with previous theoretical

95

expectations. However, for estimation purposes the issue of multicollinearity must be
addressed.

Multicollinearity is a term that refers to correlation among the independent
variables in a multiple regression model. In the presence of multicollinearity, the
measured variables are too intercorrelated to allow precise analysis of their individual
effect (Greene 418). Estimation in the presence of multicollinearity leads to regression
coefficients that possess large standard errors in relation to the coefficients themselves,
implying that the coefficients cannot be estimated with much precision or accuracy.
Nevertheless, the coefficients remain unbiased and efficient. One preliminary method to
check for any potential problem is to examine the correlation coefficient between
independent variables. Large correlations may indicate the presence of a potential
problem.4 In the matrix above, high levels of correlation (> 0.70) are found between
appropriations for corrections and appropriations for higher education, the lag of the
violent crime rate and appropriations for corrections, the lag of the violent crime rate and
appropriations for higher education, and the lag of the violent crime rate and the lag of
the size of the 18-24 cohort. Further insights into multicollinearity will be conducted at

the estimation phase by examining collinearity diagnostics.

California — Two Stage Least Squares
Utilizing the data from California, I initially estimate a system of simultaneous
equation models to test the various hypotheses. As discussed earlier in this dissertation,

estimation is conducted for the data in levels and in differences. Differences are taken to

 

4 Belsley, Kuh, and Welsch (1980) suggest that utilizing a correlation matrix is insufficient for detecting
multicollinearity. They argue that correlation matrices are incapable of detecting when three or more
variates are collinear while it is possible that no two of the variates taken alone are highly correlated.

96

mitigate the effect of unit roots and spurious regression. Decisions regarding the
appropriate transformations were made by referencing the Augmented-Dickey Fuller
(ADF) test and visual inspections of the series. Table 4.3 presents the results of the ADF

tests.

Table 4.3: Unit Root Tests (ADF)

No Trend, No Intercept Intercept, No Trend Intercept and Trend
Corrections 3.675 2.56 -l.644

 

Corrections d1 -.892 -2.307 -3.734"
Corrections d2 5805*" -5.788*“ -5.6508***
Higher Ed -1 .6348 -.0594 -3.022
Higher Ed d1 2056'” 6.207" -3. 1901
Higher Ed d2 4.991?" 4914"” 4829*"
Cohort .664 -2.661‘ -2.94
Cohort d1 -3.34*"”" 8.448" -2.193
Cohort d2 -6.097*“ 4.61"" 4804"”
Violent Crime -.366 -l .695 -.097
Violent Crime d1 -l.82* -l.72 -2.54
Violent Crime d2 -4.7402*** 425"" 4691""
Enroll .901 -2.644* -3.948”
Enroll d1 444*“ 4504*“ 4366*”
Unemployment -l .04 -3 .502" 6.626”
Unemployment d1 -3.855*** -3.79*" -3.800"
Pie -.8745 -2.443 -2.78
Pie d1 4874*" 4877*“ 4886*"

 

Note: "‘ = .10; *"' = .05; "* = .01 significance level

It was determined that appropriations for corrections, appropriations for higher education,
the size of the 18-24 cohort, enrollment, unemployment, and the budgetary pie all need to
be first differenced to make these series 1(0). In addition, violent crime is differenced
twice in order to induce stationarity. Furthermore, due to multicollinearity, found using

collinearity diagnostics in Stata, the size of the 18-24 cohort is removed fi'om the

97

corrections equation.5 Finally, all non-endogenous variables are lagged two time periods
to more accurately reflect the impact of exogenous factors and public policies on the
budgeting process. Now that issues relating to the data have been taken care of the
results from the 2SLS estimation are presented.

Table 4.4 includes results of the simultaneous equation estimation for both data in
levels and the transformed data. The results in levels illustrate that the coefficients
associated with the violent crime rate and the political ideology of the governor are
significant at the .05 level, while the implementation of determinate sentencing is

significant at the .10 level.

Table 4.4: Two Stage Least Squares — Levels and Differenced, California

 

 

 

 

 

 

 

Levels Differenced

Corrections Higher Ed Corrections Higher Ed

Coef t-value Coef t-value Coef t-value Coef t-value
Constant -4376.5 "-3. 18 -1641.4 -0.68 -44.36 -0.42 360.53 ”2.34
Corrections -- -- 0.9281 "7.71 - -- -0.3301 -0.49
Higher Education .4689 0.77 -- -- 0.029 0.09 -- ~-
Pie -0.025 -0.17 0.2483 ”4.44 0.023 0.68 0.0929 ”2.64
Cohort -- -- 0.0023 "5.94 -- -- -0.0006 -0.54
Enrollment -- -- -0.004 -1 .04 -- -- -0.0043 - l .36
Violent Crime 4.54 ”2.64 -- - -0.590 -1 .44 -- --
Unemployment 93.207 0.56 -l75.0 ”-2.72 -13.27 -0.37 -87.788 -1.60
Governor 833.30 ”3.23 9.993 0.04 107.0 1.86 -52.6 19 -0.34
Determinate Sent. -435.43 -1.73 -- -- 89.33 “3.08 -- --
Victim’s Bill -272.05 -0.24 -- -- 39.55 0.32 -- --
Three Strikes 1190.4 1.20 -- -- -l82.0 -l.36 -- --
R2 0.973 0.965 0.613 0.4151
Hausman
(ch12 prob) 0.5351 0.0001 0.927 0.6096

Note: Corrections and Higher Education in millions; ** significant at .05

 

 

5 After removal of the cohort variable, the condition number is below a value of 10.

98

 

All of these coefficients are in their hypothesized direction, except for the
implementation of determinate sentencing. Furthermore, the coefficients of corrections,
the budgetary pie, the size of the 18-24 cohort, and the unemployment rate are all
significant and in the hypothesized direction for higher education. When the data is
differenced, only the political ideology of the governor and the implementation of
determinate sentencing are significant for corrections and the size of the budgetary pie for

higher education. In the differenced model, the signs associated with the significant

W'TT“.""ITr 67.7"? “-

coefficients are in the hypothesized direction.

It is important to note that results from the 2SLS estimation show that in levels an
increase in appropriations for corrections leads to increases in appropriations for higher
education. However, the .92 coefficient indicates that corrections is growing at a faster
rate than higher education. When the data is differenced there is no significant
relationship. Before a tradeoff relationship between the two series can be ruled out it is
necessary to test and see if in fact 2SLS is an appropriate methodology. As such, OLS
models are estimated and these results are compared to the ZSLS estimates using the
Durbin-Wu-Hausman test. This test examines whether the difference in the coefficients
between the instrumented and OLS model are in fact systematic. These results are
reported in Table 4.4. The test reveals that when the data is treated in levels the OLS
estimates for corrections are not systematically different from the estimates produced by
ZSLS. Furthermore, when the data is differenced the OLS estimates for corrections and
higher education are also found not to be systematically different from the estimates
derived from the simultaneous equation model. This implies that there is in fact no

simultaneity bias in the OLS estimates of these equations. However, the test reveals that

99

in levels the higher education estimates are systematically different. Since it was found
that the corrections equation does not suffer from simultaneity bias, it can be inferred that
the relationship between corrections and higher education is recursive. This means that
there is unidirectional dependency among the endogenous variables. Appropriations for
corrections is found to be a determinate of appropriations for higher education, but
appropriations for higher education is not found to be a determinate of appropriations for
corrections.

The determination that there is not an explicit tradeoff between appropriations for
corrections and appropriations for higher education does not rule out the possibility that
the error terms of the OLS models used to estimate corrections and higher education are
correlated. As such, appropriations for corrections and appropriations for higher
education may be seemingly unrelated. In order to assess this possibility and to achieve
more efficient estimates the technique of seemingly unrelated regressions are applied to

both the transformed data and the recursive model in levels.

California — Seemingly Unrelated Regressions

Tables 4.5 and 4.6 show the results of the seemingly unrelated regressions. When
the differenced data is utilized, the model explains 61 percent of the variance in the
corrections model. In addition, diagnostic tests rule out potential problems related to
autocorrelation.6 Coefficients associated with the size of the budgetary pie, the political
ideology of the governor, the implementation of determinate sentencing and Three

Strikes and You’re Out are found to be significant deterrninates of appropriations for

 

6 At the sixth lag, the Ljung-Box Q stat is insignificant (.l 173) and the Q stat generated from the
Portmenteau test for white noise is insignificant (.4095). Furthermore, since SUR is a GLS procedure
heteroskedasticity is not a problem.

100

corrections. These results imply that corrections receives on average 2.5 percent of the
change in the budgetary pie, holding all other variables constant. This suggests that as
the budgetary pie expands, appropriations for corrections increase. Additionally, the
presence of a Republican governor in office leads to an average increase in the change in
appropriations for corrections of $104.7 million per year. This implies that Republican
governors are more likely to favor ‘tough on crime’ policies and are willing to expand the

criminal justice system while in office. Furthermore, the implementation of determinate

,_.__..._ .._,-_
i - ‘ -.- - ~

sentencing leads to an average increase in the change in appropriations for corrections of
approximately $90.1 million per year, while the implementation of Three Strikes leads to

a decrease in the change in appropriations for corrections of $181 million per year.

Table 4.5: Seemingly Unrelated Regression, Corrections in California

 

 

Differenced-Differenced Levels-Recursive

Coefficient z-value Coefficient z-value
Constant -37.04567 -0.86 -3860.97 “-7.8
Pie 0.025685 ”2.67 0.055123 1.65
Violent Crime -0.595021 - l .77 6.065963 " 12.2
Unemployment -16.1331 1 -0.098 8.692519 0.21
Governor 104.6619 "2.55 751.3223 "4.32
Determinate Sentencing 90.11306 ”3.94 -309.5103 "-2.31
Victim’s Bill of Rights 40.34203 0.68 521.6149 "2.06
Three Strikes -l8l.0394 ”-2.54 2769.183 "11.47
RT 0.6093 0.9742

Note: Corrections and Higher Education in millions; ** significant at .05

When estimation of the higher education model is performed, on the differenced
data, the model is able to capture 42.6 percent of the variance (See Table 4.6). The
coefficient associated with the size of the budgetary pie is significant at the .05 level and

the unemployment rate is significant at the .10 level. The results imply that on average

101

higher education receives 8.6 percent of the change in the size of the budgetary pie and
that a change in the unemployment rate of one percent leads to a decline in the size of

appropriations for higher education by $84.5 million per year.

 

 

Table 4.6: Seemingly Unrelated Regression, Higher Education in California [

Differenced-Differenced Levels-Recursive i
Coefficient z-value Coefficient z-value l

Constant 321.0487 "2.92 -l707.409 -0.86 i q

Corrections N/A N/A 0.919928 “9.60 ____

Pie 0.085865 "3.14 0.250779 ”5.35

Unemployment -84.49526 -l.82 -l76.2135 l”-3.17

Governor -79.28214 -0.63 21.20963 0.10

Cohort 0.000431 -0.49 0.002288 "6.96

Enrollment -0.041318 -1.53 0004517 -1.21

R7 0.4263 .9655

Note: Corrections and Higher Education in millions; " significant at .05

The alternative specification, using data in levels, estimates the models for
corrections as a recursive system. The model for corrections is able to account for 97
percent of the variance. In addition, diagnostic tests reveal that autocorrelation is not a
problem.7 Furthermore, coefficients associated with the violent crime rate, the political
ideology of the governor, the implementation of Three Strikes, the Victim’s Bill of
Rights, and determinate sentencing are significant at the .05 level. The results suggest
that on average an increase in the violent crime rate by one percent leads to an increase in
appropriations for corrections by $6.1 million, holding all other variables constant. In

addition, the presence of a Republican governor leads to an average increase in

 

7 At the sixth lag, the Ljung-Box Q stat is insignificant (.9040) and the Q stat generated from the
Portrnenteau test for white noise is insignificant (.6465).

102

appropriations of corrections of $751 million. Furthermore, the implementation of Three
Strikes leads to an average increase in appropriations for corrections of $2.8 billion and
the implementation of the Victim’s Bill of Rights leads to an average increase in
appropriations for corrections of $521 million, suggesting that the state government is
responding to the demands being placed on the criminal justice system by these ‘tough on
crime’ policies. Finally, the implementation of determinate sentencing leads to an
average decrease in appropriations for corrections of $309.5 million. The decline in
appropriations associated with the implementation of determinate sentencing may be a
function of shorter durations of time served by offenders. While this policy mandated
specific sentences for various crimes and removed discretion, the actual time served
under determinate sentencing may have been less than the time served when judges and
parole boards had discretion in determining the time served by offenders.

When estimation of the higher education model is performed on the data in levels
for the recursive system, 96.6 percent of the variance is captured. The coefficients
associated with corrections, the budgetary pie, the unemployment rate, and the size of the
18-24 cohort are all significant at the .05 level. The results suggest that on average
higher education receives 25 percent of the budgetary pie, holding all other variables
constant. In addition, as appropriations for corrections increases, appropriations for
higher education increase. However, a one-dollar increase in corrections is met with only
a 92-cent increase for higher education. This suggests that corrections is increasing at a
larger rate than higher education. Furthermore, an increase in the unemployment rate by
one percent leads to an average reduction in appropriations for higher education by

$176.2 million. Finally, an increase in the size of the 18-24 population by one individual

103

P"?
A

leads to an average increase of approximately $2,300 in appropriations for higher
education.

The previous discussion has provided insights into the relationship between public
policies and exogenous factors on the budgets of corrections and higher education.
However, the question if there is a tradeoff between these two budgetary categories has
not been resolved. Previously it was found that an explicit tradeoff could be dismissed,
however, it is still possible that these budgetary categories are seemingly unrelated.
Table 4.7 presents the results of the Breusch-Pagan test. The results show that for
estimation of the differenced data the errors are not related, suggesting that there is no
tradeoff between corrections and higher education. In contrast, the test results from the
series in levels suggest that the errors of the two series are correlated at a value of 007.
While this suggests that they are seemingly unrelated, the relationship is not significant.
As such it can be inferred that there is in fact no tradeoff relationship between
appropriations for corrections and appropriations for higher education in California
between 1970 and 1999. In fact the results from estimation of the recursive model show
that increases in appropriations lead to almost equivalent increases in appropriations for
higher education. These finding confirm Domke, Eichenberg, and Kelleher’s (1983)
assertion “that direct priority setting — the trading off of one program against the other —

is rare or nonexistent in democratic governments” (20).

Table 4.7: Seemingly Unrelated Results (Breusch-Pagan Test), California

 

Correlation Chi-square Probability
Differenced-Differenced 0.0826 0. 184 0.6676
Levels-Recursive -0.0702 0. 138 0.7102

 

104

This chapter has presented a framework within which to test the hypotheses laid
out in Chapter 2. Estimation was conducted for transformed data and for data in levels.
A comparison of the results generated from the two paths can be seen in Figure 4.8. This
figure shows the sign and the p-value (in parentheses) of those coefficients that were
found to be significant for corrections and higher education. When both paths produce

significant coefficients with the same sign, the results have converged. Convergence

cum
. a: I
1

occurs for the size of the budgetary pie and the partisan ideology of the governor for

.33

corrections and the budgetary pie and the unemployment rate for higher education. In
addition, only one of the paths for corrections produces significant coefficients for the
implementation of the Victim’s Bill of Rights. For higher education, only one of the
paths produces significant coefficients for corrections, the size of the cohort, and
enrollment. Finally, the signs of the coefficients associated with the violent crime rate,
the implementation of determinate sentencing, and Three Strikes and You’re Out are in
conflict with one another. In order to understand these results, the findings will be

directly related to the statistical hypotheses.

Table 4.8: Triangulation of Estimates in California

 

Corrections Difference — Difference Levels-Recursive
Pie + (.008) + (.100)
Violent Crime - (.077) + (.000)
Unemployment

Governor + (.011) + (.000)
Determinate Sentencing + (.000) - (.021)
Victim’s Bill of Rights + (.039)
Three Strikes - (.011) + (.000)

 

105

Table 4.8 (continued): Triangulation of Estimates in California

 

Higher Education Difference — Difference Levels-Recursive
Corrections + (.000)

Pie + (.002) + (.000)
Unemployment - (.068) - (.002)
Governor

Cohort + (.000)
Enrollment - (.125)

 

Note: +/- symbols are the signs of coefficients and p-values are in parenthesis (S .125).

The results in California reveal: (1) that there is not a tradeoff relationship
between corrections and higher education appropriations. Instead the relationship is
recursive with higher education growing at a slower rate than corrections. (2) Exogenous
factors create policy relevant conditions that impact budgetary allocations. Specifically,
the presence of a Republican governor leads to increased appropriations for corrections.
Additionally, an increasing violent crime rate leads to a growing demand for corrections
resources and thus increased appropriations for corrections. Finally, appropriations for
higher education increase as the unemployment rate declines and as the size of the 18-24
year old cohort increases. (3) An expanding budgetary pie leads to increases in
appropriations for both corrections and higher education. (4) The implementation of
specific public policies is instrumental in causing changes in state government
appropriations. The implementation of the Victim’s Bill of Rights leads to an increase in
appropriations for corrections. Furthermore, the results surrounding the implementation
of determinate sentencing and Three Strike’s and You’re Out are mixed. The more
plausible of the two results is that the implementation of determinate sentencing actually

led to shorter lengths of time served and decreased appropriations and the longer

106

 

sentences associated with the implementation of Three Strikes led to increased
appropriations for corrections.

To gain further insights into the relative impact of these factors on appropriations,
predicted values are generated for the data in levels. The table that follows (Table 4.9)
shows the estimated coefficients for the constant term and the independent variables for
corrections. In addition, the predicted values associated with each fiscal year are
generated by multiplying the coefficients by the value of the corresponding independent
variable in billions of dollars. Calculating predicted values allows for an inspection into
the relative contribution that each factor has on the amount appropriated. For example, in
FY 1972 the coefficient for the violent crime rate is $006 billion and the violent crime
rate itself is 474, resulting in 2.8 billion dollars (.006*474) appropriated for corrections.
Finally, the predicted appropriations for corrections for each fiscal year are shown as well
as the actual amount appropriated.

Table 4.9 shows the predicted values generated for corrections in levels. The
constant term suggests that absent any other influence corrections would receive —$3.9
billion. Two factors play a large role in offsetting the negative coefficient. These are the
violent crime rate and the implementation of Three Strikes. The growing violent crime
rate in the 1970’s led to an increased demand for correctional resources. This demand
was met with greater resource allocation in the form of increased appropriations for
corrections. This can be seen in the predicted values for corrections attributable to the
violent crime rate, which increased from $2.9 billion in 1972 to $5.4 billion in 1982.
Following a short period of decline the appropriations attributable to violent crime rose to

a high of $6.8 billion in 1994. In the latter half of the 1990’s the violent crime rate in

107

California and the rest of the nation declined. This meant that there was less of a demand
being placed on the correctional system and appropriations attributable to the violent
crime rate declined to $4.8 billion in 1999. In conjunction, one would expect overall
appropriations for corrections to decline. However, appropriations for corrections
continued to rise in the late 1990’s. This increase was a function of the implementation
of Three Strikes and You’re Out.

The implementation of Three Strikes dramatically influenced the amount of
monies appropriated for corrections. This ‘tough on crime’ policy was implemented at
roughly the same time that the violent crime rate began to decline. Three Strikes
necessitated that additional resources be allocated to corrections. Appropriations
attributable to Three Strikes measured $1.4 billion in 1997, $2.1 billion in 1998, and $2.8
billion in 1999.8 Thus, despite a decline in the violent crime rate, appropriations for
corrections continued to rise primarily due to ‘tough on crime’ policies. Furthermore, it
is interesting to note that if the violent crime rate had not declined in the late 1990’s the
amount appropriated for corrections would have significantly increased. For example, if
the number of violent offenses known to police per 100,000 inhabitants in 1996 and 1997
had been roughly the same rate as the early 1990’s (1,000), the amount appropriated for
corrections would have risen from current predicted value of $4.5 billion to $5.3 billion
in 1998 and $4.7 billion to $6.0 billion in 1999, ceteris paribus. Such a large increase

would have most likely placed a tremendous fiscal strain on the state of California. The

 

8 The dummy variable used for Three Strikes is coded as .5 in 1997, .75 in 1998, and 1 in 1999. This
weighting system is utilized to more accurately capture the implementation process of this policy. Many
offenders being tried after the implementation of this policy in the first two years were in fact not eligible to
be sentenced under its’ guidelines. This is because many offenders were arrested prior to its
implementation, but due to the lengthy amount of time associated with trials and criminal processing they
were not sentenced until afterwards.

108

state may have had to respond by increasing taxes or funding corrections at the expense
of other programs. This implies that California presently finds itself in a very tenuous
place. If the violent crime rate were to increase inthe coming years the growing demand

for additional resources for corrections might lead to a statewide fiscal crisis.

“a" "
5.1

‘—

:‘me‘q Q

109

.. Mel“: l.~l.|n.. bk

 

 

 

110

85. ~45. 553 -55 565. 3.5.4 855 E5 8~5 65.5- 555.
~85 .53. :3 -55 55.5. 5-5 855 .85 535 .85- 82
83 ~55 58.? -55 58.5. 585 855 E5 8~5 .85- ~55.
8~5 8~5 555.5 ~85 565. 5.35 2.5.5 E5 8~5 55.5- 85.
85.5 ~85 555.5 -55 565. 855 ~85 P85 555 55.5- 82
85.5 ~85 555.5 -m5 555. 8~5 85 E5 85 .85- 85?
58.5 85.5 555.5 -55 585. 855 855 55 85. 55.5- 82
585 85.5 555.5 -55 585. 585 585 85 3.5. 55.5- ~52
58.5 ~8.~ 555.5 -55 555. 58.5 355 E5 525 55.5- 55’
~8~ :5.~ 555.5 ~85 55.5. 58.5 855 .85 ~85 8.5- 552
SE v8~ 555.5 -m5 555. ~85 585 E5 525 .85- 58.
-5.~ ~5m.~ 555.5 ~85 585. 8.5 855 E5 585 .85- 58?
53.. 83 555.5 ~85 55.5. 545.4 855 E5 595 8.5- ~55,
85; ~85 555.5 ~85 58.5. 85.4 855 .25 535 55.5. 58.
25.. 8: 555.5 ~85 56.5. 58.4 v8.5 .85 52.5 .85- 82
85.? -~.. 555.5 -55 55.5. 85.5 58.5 555.5 $5 85- 48?
85.5 855 555.5 555.5 56.5. 5-5 «8.5 555.5 ~85 .85- 82
85 5555 555.5 555.5 58.5. :45 58.5 555.5 85.5 85- ~55.
85 855 555.5 555.5 55.? 53... 48.5 555.5 58.5 .85- .8.
555.5 58.5 555.5 555.5 555. 55.4 ~85 555.5 535 85.5- 58?
5~m5 8~5 555.5 555.5 555. t? :55 555.5 RE .85- 22
~85 85.5 555.5 555.5 555.5 85... 585 555.5 8~5 8.5- 85,
855 8~5 555.5 555.5 555.5 25.5 85.5 555.5 585 55.5. E:
~85 85 555.5 555.5 555.5 8~5 855 E5 585 v8.5- 85?
5555 :45 555.5 555.5 555.5 5%. F555 .85 855 .85- £9
8~5 .5~.5 555.5 555.5 555.5 8~5 855 E5 585 .85- £5?
8~5 8~5 555.5 555.5 555.5 ~35 855 55 ~85 .85. 25,
8~5 85. 555.5 555.5 555.5 83 855 E5 855 55.5. -5F
83 -m5 555. 85.5 5555 E5 585 55.5-
.532 88.58.. «9....» :5 552358 2.3 sum 2..

3E0>oo revenge 282.30 E

«co—«cotoo 20:02.30 02:... «.8305 8.5555200 ace—05 2.083352:

A8553 a: 20:09:00 .2 «2033595.? .8 mo:_n> 38:55 6.5. min...

CHAPTER 5

The second state examined in this dissertation is Michigan. Michigan
experienced several policy reforms between 1970 and 1999 that shifted the nature of

sentencing from indeterminacy to deterrninacy. Reforms included the implementation of

g

mandatory minimums for firearms and drugs, truth in sentencing, sentencing guidelines,

‘ff““!"f

and the elimination of good time for certain offenses. This chapter proceeds by providing
an overview of these policy changes and a brief review of societal conditions in Michigan
during the period of study. This includes a look at the prison population, the
unemployment rate, the violent crime rate, size of the 18-24 cohort, and enrollment in 4-
year public institutes of higher learning. This contextual overview is followed by the
estimation of models for corrections and higher education appropriations. The various
budgetary hypotheses in this dissertation will be tested using the estimation strategy laid

out in Chapter 3.

Michigan — Policies

Until 1977, sentencing in Michigan was almost exclusively indeterminate in
nature and operated under a structure where each sentence carried a minimum and
maximum term (set by penal statute) with the exact date of release at the discretion of the
parole board. Following years of broad judicial and parole board discretion, Michigan
enacted a series of reforms that represented a shift towards determinate sentencing. The
reforms included the introduction of mandatory minimum sentences, sentencing

guidelines, truth in sentencing, and alterations to prison release mechanisms.

111

The first major reform, M.C.A.A., Sec 750.227 (Michigan Felony Firearm
Statute), adopted in 1976 and effective January 1, 1977, mandated a two-year prison term
for any person convicted of possession of a firearm during the commission of a felony.I
This term could not be suspended or shortened by release on parole (Cohen and Tonry
1983). A second conviction mandated a five-year term and a third conviction a ten-year
term. These sentences were to be served prior to serving their sentence on the underlying
offense.2

The second major reform was established via statewide referendum in 1978
(Ballot Proposal B). This law mandated that prisoners convicted of any of a long list of
offenses must serve the minimum sentence imposed by the court. The offenses included:
arson related crimes, various types of assaults, breaking and entering, firearm offenses,
forms of criminal sexual conduct, and armed and unarmed robbery (Shane-Dubow,
Brown, and Olsen, 153). Thus, the minimum sentence could not be reduced by good
time or special parole. A report prepared by the Michigan House Fiscal Agency suggests
that Proposal B served to increase average sentences for these offenses by 1.8 years
(Gregory p. 6). Proposal B was subsequently amended in 1982 by Public Acts 442 and
458. These acts allowed prisoners, who were no longer eligible for good time, to receive
a five-day per month reduction in their sentence as disciplinary credits and an additional
two days per month from a prisoner’s disciplinary committee. Under the new system, the

first five days were awarded automatically, unless a prisoner misbehaved, and the second

two days were at the discretion of prison officials. In total, prisoners could have had their

 

' Tonry (1987) recounts that one Michigan prosecutor promised on billboards and bumper stickers that
“One With a Gun Gets You Two” (25).

2 A similar law in Massachusetts, the Bartley-Fox Gun Law (1975) mandates a one-year prison sentence for
the first conviction, a five-year sentence for the second conviction, and seven years for the third. However,
unlike in Michigan, offenders may be paroled.

112

ill'n': u.|“ Lq ""‘l'°. ~‘— ‘

sentences reduced by up to 84 days per year. Finally, in 1984, good time credits, which
automatically granted credits ranging from five to 22 1/2 days per month depending on the
length of sentence, were removed and the entire inmate population was placed under the
disciplinary credit structure, at a lower and flat rate. Gregory suggests that “this measure
had the effect of adding an additional 500 prisoners to the annual population count” (6).

The third reform, the “650 Lifer Law,” was also established in 1978. This law
mandated life without parole for possession of 650 or more grams of cocaine or heroin.
The law was aimed at large drug dealers and represents a ‘one-strike and you’re out’
policy. As of 1998 there were approximately 220 offenders in Michigan prisons serving
life terms under this law, a majority of who can be viewed as nonviolent. Prisoners
sentenced to life represent a fixed cost for the Department of Corrections since they are
not eligible for parole and will continue to fill beds and impact costs of operation.3

A fourth reform was the passage of Public Act 519 of 1980 (signed into effect on
January 21, 1981), which established the Prison Overcrowding Emergency Powers Act
(EPA). The EPA gave the governor the power to reduce the sentences of all prisoners
(except life and gun sentences) by ninety days every time the states’ prisons and
correctional camps were overcrowded for thirty days in a row and all other administrative
remedies had been exhausted. This mechanism acted to accelerate the parole process by
reducing a prisoner’s earliest parole eligibility and was seen as a short-term solution for
dealing with the problem of prison overcrowding associated with the ‘crisis in
corrections’ (Gregory, 7). The EPA was employed nine times between its inception and

1984 (once in 1981; three times in 1982; twice in 1983; and three times in 1984) and

 

3 In July of 1998, Michigan passed bills changing the "650 lifer" law to "life or any term of years, not less
than 20" and allowing parole for current 650 lifers who had already served twenty years.

113

 

benefited an estimated 10,000 inmates and served to reduce many sentences by as much
as 810 days. However, the use of this mechanism as a ‘safety-valve’ for overcrowding
did not endure.

On October 25, 1984 two people were murdered in the Lansing area (one of the
victims was an East Lansing police officer). The murders were committed by a male
parole violator and a female halfway house escapee. The male had benefited from

previous enactments of the EPA and soon thereafter the release mechanism came under

ML; aw-..‘ w

intense scrutiny.4 As a result, in September of 1984, Governor James Blanchard refilsed
to invoke the EPA to help alleviate prison overcrowding and since then the mechanism
has not been utilized again. The decision to not employ the EPA arose at the same time
that Michigan faced a 1984 court consent decree that grew out of action brought against
the state by the United States Department of Justice. The consent decree stipulated that
medical, sanitation, fire safety, and overcrowding conditions be rectified in Michigan’s
prisons. The decree declared that inmates were required to have at least 60 square feet in
the dormitories as their own (Gregory, 6). The implication that follows is that Michigan
had to either build additional capacity or release some offenders. Despite an obvious
incentive to continue using the EPA, the State of Michigan chose to build new prisons.
Warren Gregory states that Blanchard’s decision implied that “the state would expand
capacity; it would build its way out of the crisis” (Gregory, 8). Thus, between January

1985 and December 1991, the State of Michigan undertook a six-year prison construction

 

4 For example, between the period of September 22 — 25, 1985 the Detroit Free Press ran a series entitled
“Revolving Door Prisons” that highlighted the early release of prisoners and the subsequent offenses these
early released offenders committed. In response to the increased criticisms, the Michigan House of
Representatives Correction’s Committee released a report entitled the Report of the House Standing
Committee on Corrections Investigation of Michigan Department of Corrections Release Policies and
Practices (1985). Among other things, the report focused on the use of the EPA, good time credits, and
disciplinary credits.

114

program that was estimated to cost over $800 million. The project increased state prison
capacities from 12,930 to 31,548 beds (Gregory, 1).

The final sentencing reforms in Michigan have occurred in the late 1990’s. In
1998 Michigan implemented truth in sentencing (created in 1994) for assaultive offenses
committed on or afier December 15, 1998. This was later expanded to include all
felonies beginning on December 15, 2000. These provisions eliminated disciplinary

credits for violent and assaultive crimes and ended the practice of the placement of

_ ‘ -mr‘.w

offenders in community corrections programs. The National Council on Crime and l .
Delinquency expects these changes to add 5,400 more offenders to the state prison
system by 2007 (Michigan Department of Corrections, 47). Additionally, Public Act
317, enacted on January 1, 1999, established a set of sentencing guidelines that covered
all felonies and replaced guidelines imposed by the Michigan Supreme Court in 1988.
This was the fulfillment of an effort to create and implement sentencing guidelines that

began in 1979.

Michigan — Prison Population

Over the time period of study, Michigan has dramatically increased its prison
population. Between 1979 and 1999, Michigan’s prison population has increased by
230% from 13,330 to 44,191 inmates. As can be seen in Figure 5.1, the total number of
prisoners under state jurisdiction in Michigan remained relatively stable between 1971
and 1974 with a population hovering below 10,000 prisoners. Between 1975 and 1978
there was an increase in the number of prisoners to about 15,000. This remained

relatively stable until 1985 when Michigan saw a dramatic and continuous increase in its

115

 

prison population, towards 45,000 prisoners.5 In addition to growth in the number of
prisoners, Michigan’s incarceration rate per 100,000 inhabitants grew from 106.4 in
l 971, to 440 in 1997. During this same time period, appropriations for corrections began

to increase and the percentage of the state budget designated for corrections began to

 

 

 

 

 

 

increase.
50.000
45.000 «
40,000 J
35.000 +
2 30.000
3
0 25.000 .
O
1:
“- 20.0009
15.000
10.000 1
0 r . . . . . , 5
v- (‘0 10 IN G) 1'- CO I!) IN C) v- 00 IO N
IK N N N N (D co co 0 (I) O) O) 8 O)
92 92 92 92 92 92 92 92 92 92 92 92 .. 92
Year

Figure 5.1: Number of Prisoners Under State Jurisdiction in Michigan: 1971-1997

Michigan - Independent Variable — Governors
During the period of study Michigan had three governors. The first governor,

William G. Milliken, a Republican, held office between 1969 and 1983. Governor

 

5 In addition to building new prisons to meet increases in the state’s prison population, Michigan made
arrangements to transfer approximately 1,500 prisoners to prisons in the state of Virginia in the mid 1990’s.
Virginia had a surplus of prison beds following a massive prison construction project implemented under
the auspices of Governor George Allen in 1995.

116

Milliken was followed in office by a Democrat, James J. Blanchard. Governor Blanchard
held office from 1983 to 1991. Finally, Governor John M. Engler, a Republican, held
office from 1991 to the present. Thus, Republicans held the office of governor in

Michigan for all but eight years of the time under investigation.

Michigan — Independent Variable - Unemployment Rate

Between 1970 and 1999 the unemployment rate in Michigan fluctuated
dramatically between 3 and 15 percent. These lows and highs reflect fiscal good times
and fiscal bad times for the state of Michigan. In the mid 1970’s and the early 1980’s
Michigan experienced periods of extreme unemployment. Between 1973 and 1975 the
unemployment rate rose from 5.9 to 12.5 and between 1978 and 1982 it rose from 6.9 to a
high of 15.5. In 1982 the unemployment rate in California (9.9) and the nation as a
whole (9.7) were significantly lower than that found in Michigan. Following the fiscal
crisis in the early 1980’s in Michigan, the unemployment rate declined to 7.1 in 1989
only to increase again to 9.3 in 1991. The remainder of the 1990’s was characterized by

a steady decline to a low of 3.8 in 1999.

117

r—-‘-I-*-z‘—-—-W-

 

18

Unemployment Rate

 

 

 

 

 

2 l.
0 . 2 5
O N V (D 00 O N a CO co 0 N v (D d)
tx ts Ix Ix rx co co co co 0) CD 0) O) O)
O) O) CD CD 0’) O) O) 0') O) O) O) 0') CD 0') O)
v- 1- v- v- ‘- s- 1— ‘- 1- ‘- ‘— 1- ‘- 1- v-
Fiacal Year

Figure 5.2: State of Michigan: Unemployment Rate, FY 1970-1999

Michigan — Independent Variable — Violent Crime Rate

Similar to California, the violent crime rate in Michigan rose between 1970 and
the early 1990’s. However, the increase in Michigan was not as steady in the mid 1970’s.
In Michigan the violent crime rate rose to a rate of 686 in 1975 only to decline to 577 in
1979. Between 1979 and 1986 the rate of violent offenses known to police per 100,000
inhabitants increased fairly steadily to a rate that exceeded 800. While not as dramatic an
increase as in California, the violent crime rate increased by roughly 40 percent between
1970 and 1986. The violent crime rate in Michigan remained above 700 through the
early to mid 1990’s. However, in the second half of the 1990’s the violent crime rate,

mirroring national trends, declined to a rate of 574.9 in 1999.

118

900

 

800
700

 

600
500 ~—- - — . , ~ , fl , 2 .2 2 2. , ,, , _ ..----
400 7
300 1

Violent Crime Rate

200 9
100 J.

 

 

 

1970
1972
1974
1976
1978
1980
1982

g 00
s-

P

1988
1990
1992
1994
1996
1998

Fiscal Year

Figure 5.3: State of Michigan: Violent Crime Rate (Part 1 offenses known to police per
100,000 inhabitants), FY 1970-1999

Michigan — Independent Variable — Cohort

The size of the 18-24 cohort in Michigan increased from slightly over one million
in 1970 to roughly 1.3 million in 1980. After this period of expansion in the age group
most prone to criminal activity and most likely to attend institutes of higher learning, the
size of the cohort declined to a low of 917,000 in 1997, only to rise slightly through 1999.
Therefore, the majority of the growth in appropriations for corrections and higher
education occurred during a period of time when the 18-24 cohort in Michigan was

waning.

119

1 ,400.000

 

 

 

 

 

1,200,000«
1,000,000 ~
‘5
.5 800.0002
0
§. 600.000-
Q
1-
400.000
200.0007
0 . . . 2 . , . . 2. T .2 .
C) N v (D co 0 N (D Q N no

Fiscal Year

Figure 5.4: State of Michigan: 18-24 year old cohort, FY 1970-1999

Michigan — Independent Variable — Enrollment

Enrollment in public four-year institutions of higher education in Michigan grew
roughly 20 percent between 1970 and 1999. Between 1970 and 1980 enrollment
increased from 217,700 to 242,100 students. After a slight decline through 1984,
enrollment increased to a high of 278,800 students in 1997 only to decline again slightly

to roughly 260,000 in 1999.

120

300.000 -

 

 

 

 

 

C

£250,000

N

U

a 1

m200,000~

b

O

‘5

$150000

.5

$100,000

250.0002 22 2 2 2222 2 22 22 2

m
0 ..
eczeeeaasasa§sa
99.998299382992292

FlscaIYear

Figure 5.5: State of Michigan: Enrollment in Higher Education, FY 1970-1999

Michigan — Correlation Matrix

Now that an overview of policy changes and contextual conditions in Michigan
have been provided, a preliminary look at the relationship between continuous variables
in the model for corrections and higher education appropriations is provided in the form
of a correlation matrix. By examining correlations between variables the strength of

bivariate relationships and a preliminary diagnosis of multicollinearity can be conducted.

121

Table 5.]: Correlation Matrix for Variables included in the Budget Model, Michigan

Violent

 

Corrections Higher Ed Budgetary Cohort Enroll C . Unemployment

Appropriations Appropriations Pie (t-2) (t-2) (t-2) (t-2)

Corrections -- .967 -.130 -.881 .889 .466 -.498

Appropriations .000 .495 .000 .000 .009 .006

Higher Ed .967 __ -.135 -.824 .873 .582 -.409

Appropriations .000 .478 .000 .000 .00 1 .028

B d t P' -.l30 -.135 __ .069 -.208 -.053 .092

“ 3“” 'e .495 .478 .727 .289 .783 .640

-.881 -.824 .069 -.693 -.287 .634

C°h°n (“2) .000 .000 .727 ” .000 .123 .000

.889 .873 ~.208 -.693 .296 -.477

Em" ("2) .000 .000 .289 .000 " .1 12 .008

Violent Crime .466 .582 -.053 -.287 .296 __ .261

(t-2) .009 .001 .783 .123 .l 12 .163
Unemployment -.498 -.409 .092 .634 -.477 .261 __

(t-2) .006 .028 .640 .000 .008 .163

 

Note: Pearson correlation statistic followed by significance level

The results shown in Table 5.1 illustrate that only a few of the continuous independent
variables are significantly correlated and correlated in the hypothesized direction with
appropriations for corrections and appropriations for higher education. These include the
violent crime rate with corrections and the enrollment rate and the unemployment rate for
higher education. For both corrections and higher education the size of the 1824 cohort
is significantly correlated in a direction that is counter to the hypothesized direction. In
addition, the size of the budgetary pie is correlated in a direction that is opposite of the
hypothesized relationship and is not statistically significant with both corrections and
higher education. Finally, there are no correlations between independent variables that

exceed the magnitude of .7. This preliminary investigation suggests that multicollinearity

122

is not a problem. Results of diagnostic procedures will be presented later in this chapter

that allow for a more formal test for the presence of multicollinearity.

Michigan — Two Stage Least Squares

The hypotheses laid out in Chapter 3 will be tested using the data from Michigan
in both levels and in differences. Before estimation can be undertaken the data itself is
diagnosed and adjusted to account for several methodological issues. First, to avoid
problems associated with spurious regression, the Augmented Dickey-Fuller (ADF) test
is referenced to inform decisions regarding the transformation of data to induce
stationarity. Second, tests for multicollinearity are conducted to help ensure that the
standard errors of the coefficients do not become inflated. Finally, the data itself is

adjusted to the appropriate lag structure. Table 5.2 presents the results of the ADF tests.

Table 5.2: Unit Root Tests (ADF)

 

 

No Trend, No Intercept Intercept, No Trend Intercept and Trend

Corrections 8.163 4.186 -1.528
Corrections d1 -.8 17 -2.02 -3.633"”"
Corrections d2 -7.825"‘"‘* -7.804"'** -7.616”*
Higher Ed 3.584 .272 -2.835
Higher Ed d1 -5.259“”""' -8.842“* -8.901"'"
Cohort -.764 .206 -5.538'”""
Cohort d1 -.3.881*** 4393"” -3.101**
Violent Crime .062 -l.954 -.949
Violent Crime d1 4585*" -4.475"'*“ 4727*"
Enroll 1.454 -.892 -2.424
Enroll d] -5.068"‘" 4475‘” 4727*"
Unemployment -.730 -1.451 -1 .973
Unemployment d1 4833‘“ 4766"“ 4811*"
Pic 4468*" -9.382"* 9378*"

 

Note: "‘ = .10; ** = .05; *** = .01 significance level

The results of the ADF tests reveal that all of the continuous variables in the
model need to be differenced to induce stationarity, except for the size of the budgetary
pie. Appropriations for higher education, the size of the 18-24 year old cohort, the
violent crime rate, and the unemployment rate all need to be first differenced to make
these series 1(0). Stationarity for appropriations for corrections is achieved by second
differencing the serics. In addition, collinearity diagnostics in Stata reveal that the
condition number for all the independent variables is 8.27 and none of the VIF’s
associated with the independent variables exceed ten. As such, potential problems
associated with multicollinearity can be ruled out. Finally, all non-endogenous variables
are lagged two time periods to ensure the correct specification of their relationship with
the budgetary process in regards to time. Now that data issues have been addressed
estimation is conducted on the data in levels and differences utilizing two stage least
squares.

Table 5.3 presents the results of robust ZSLS estimation for the data both in
levels and in differences. The model for corrections in levels is able to explain 99
percent of the variance and diagnostics tests reveal that autocorrelation is not a problem.6
The results in levels illustrate that the coefficients associated with the political ideology

of the governor, the implementation of mandatory minimums,7 and the implementation of

 

6 At the sixth lag, the Ljung-Box Q statistic is insignificant (.0987) and the Q stat generated from the
Portrnenteau test for white noise is insignificant (.1532). Furthermore, heteroskedastic consistent
estimators are used by applying Huber-White robust estimates of the standard errors.

7 Mandatory minimums is coded as zero’s predating the implementation of reforms, 1 in 1977 (felony
firearms), and 2’s from 1978 on (Proposal B and 650 Lifer in 1978). As such, it represents the cumulative
impact of various mandatory minimum policies ranging from minimum sentences for firearm possession to
drug possession.

124

 

the Prison Overcrowding Emergency Powers Act8 are significant at the .05 level and in
the hypothesized direction. The results of the estimation imply that corrections receives
on average 104.8 percent of the amount appropriated for higher education. In other
words, for every dollar appropriated for higher education, corrections receives one dollar
and an additional 5 cents. Thus, while controlling for other explanatory factors it can be
seen that corrections is growing at a faster rate than higher education. While not fitting
the classic model of a tradeoff relationship where an increase in one budgetary category
leads to a decrease in another category, budgetary appropriations in Michigan expresses a
form of tradeoff where corrections experiences gains at the expense of higher education.
In addition, the presence of a Republican governor leads to an average increase in
appropriations for corrections of $127.7 million. This implies that while in office
Republican governors in Michigan have led to the expansion of the criminal justice
system. Furthermore, the implementation of a series of policies imposing mandatory

minimum sentences on certain offenses led to an average increase in appropriations of

 

8 It is hypothesized that the implementation of the EPA by governors in Michigan would lead to decreased
appropriations for corrections by removing some of the fiscal demands placed on corrections via early
release.

125

Table 5.3: Two Stage Least Squares (robust) — Levels and Differenced, Michigan

 

 

 

 

 

 

 

 

Levels Differenced

Corrections Higher Ed Corrections Higher Ed

Coef t-value Coef t-value Coef t-value Coef t-value
Constant 884.53 "3.14 - 144.26 -0.46 172.98 1.30 92.32 "2.52
Corrections -- -- 0.8837 “8.51 -- -- -0.0185 -0.27
Higher Education 1.0482 l""‘10. 13 -- -- -0.160 -1.12 -- --
Pie 0.0494 0.97 -0.0244 -.045 -0.075 -1.87 -0. 1021 1.56
Cohort -0.0013 ”-6. 17 0.00058 1.93 -0.001 -0.77 -0.0001 -0.10
Enrollment -- -- 0.0003 1 0.20 -- -- -0.0001 0.03
Violent Crime -0.0634 -0. 18 -- -- -0.232 -0.81 -- --
Unemployment 6.31 18 0.74 2.3879 0.52 -4.805 -1.07 -11.333 -1.32
Governor 127.656 ”2.90 -141.47 "-6.24 5.238 0.14 -16.115 -0.29
Mandatory Mins. 74.966 "2.77 -- -- -43. 16 -l.31 -- --
Discipline Credit -184.84 “-2.84 -- -- -29.85 -.055 -- --
EPA -157.68 "-2.56 —- -- -10.77 -0.21 -- --
R2 0.991 0.981 0.2881
Hausman
(chiz prob) 0.000 0.0008 0.4393 0.9995

Note: Corrections and Higher Education in millions; ** significant at .05

 

$75 million for corrections, while the release of offenders under provisions of the EPA

led to an average decrease of $158 million holding all other factors constant. Finally, the

coefficients associated with the size of the 18-24 year old cohort and the implementation

of the disciplinary credit structure are significant at the .05 level but opposite the

hypothesized direction. When the data is differenced none of the coefficients are

significant predictors of appropriations for corrections.9

For higher education in levels the model is able to explain 98 percent of the

variance and diagnostic tests reveal that problems associated with autocorrelation can be

ignored. At the sixth lag, the Ljung-Box Q statistic is insignificant (.2454) and the Q stat

generated from the Portrnenteau test for white noise is insignificant (.3326).

 

9 Negative R-squared values are possible with the ivreg command in Stata (Sribner and Wiggins 2001).
When negative the R-square value is not reported by Stata.

126

Furthermore, heteroskedastic consistent estimators are used by applying Huber-White
robust estimates of the standard errors. The results reveal that only the coefficients
associated with appropriations for corrections and the political ideology of the governor
are significant. The results suggest that holding all other factors constant, appropriations
for higher education grows at a slower rate than appropriations for corrections. In fact for
every dollar appropriated for corrections, higher education receives 88 cents. In addition,

the presence of a Republican governor leads to an average decrease of $ 141.5 million per

 

year. This suggests that in Michigan, Republican governors are increasing funding for IL j
corrections while at the same time decreasing funding for higher education. Governors in
Michigan may in fact be trading off dollars for higher education for dollars for
corrections. Finally, when the data is differenced none of the coefficients are significant
predictors of appropriations for higher education.

Before moving on to talk about the triangulation of the results the issue of
simultaneity must be addressed. To achieve this task OLS models are estimated for both
corrections and higher education and these results are compared to the ZSLS estimates
using the Durbin-Wu-Hausman test. This test examines whether the differences between
the models are systematic. The results reported in Table 5.3 test the null hypothesis that
there is not a systematic difference between the ZSLS and OLS results. The test reveals
that the null hypothesis is rejected for both corrections and higher education in levels,
implying that a simultaneous equations model is the appropriate methodology. In
contrast, the null hypothesis is not rejected for both corrections and higher education in
differences. This means that there is no simultaneity bias in the OLS estimates of these

equations. However, it is still possible that the error terms of the OLS models used to

127

estimate corrections and higher education are correlated. As such, a seemingly unrelated

regression is undertaken to estimate both corrections and higher education in differences.

Michigan — Seemingly Unrelated Regressions
Table 5.4 shows the results of the seemingly unrelated regression for the data in

differences. The model for corrections captures 26 percent of the variance, while the

..1"-"—

model for higher education captures 29 percent of the variance. In addition, diagnostic

lunar-1.5m arr-r.

tests rule out potential problems associated with autocorrelation. The test generates an
insignificant (.0629) Q statistic equal to 18.889. For both corrections and higher
education only the coefficients associated with the size of the budgetary pie are
significant. Since appropriations for corrections is twice differenced and appropriations
for higher education first differenced, the coefficients are difficult to interpret in a
meaningful fashion. What is important to note is that that the seemingly unrelated
regression model in differences does not do a good job in predicting appropriations for
either corrections or higher education. Furthermore, the Breusch-Pagan test (Table 5.5)
reveals that the two series are positively correlated. However, the relationship is
insignificant suggesting that there is not a seemingly unrelated tradeoff relationship when

the data is in differences.

128

Table 5.4: Seemingly Unrelated Regression (differences), Michigan

 

 

Corrections Higher Education

Coefficient z-value Coefficient z-value
Constant 203364 1.53 81.0430 “2.46
Pie -0.04685 "-2.74 -0. 102398 "-2.66
Cohort 0.00026 0.99 0.000027 0.96
Enroll -- -- 0.000947 0.33
Violent Crime -0.02129 -0. 16 -- --
Unemployment -2.64107 -0.83 -1 1.2776 - l .73
Governor -10.9977 -0.75 -14.8655 0.68
Mandatory Minimums -4.81645 -0.37 -- --
Disciplinary Credit 17.6559 0.62 -- -- .
EPA 30.0148 0.75 -- —- l‘
R2 0.2580 0.2863 1.

 

Note: Corrections and Higher Education in millions; Data differenced; *"‘ significant at .05

Table 5.5: Seemingly Unrelated Results (Breusch-Pagan Test), Michigan

Correlation Chi-square Probability
Differenced-Differenced 0.2732 1 .940 0. 1637

 

This chapter has provided the results from the output of estimation conducted on
the data in levels and in differences. A summary of the results for the two estimation
paths is shown in Table 4.6. This table shows the sign and the p-value (in parentheses) of
those coefficients that were found to be significant for corrections and higher education.
The results reinforce the finding that for Michigan estimation in levels produce results
that are more robust. As a result, none of the variables converge between the two paths.
Instead, when the data is in levels higher education, the size of the cohort, the partisan
ideology of the governor, the implementation of mandatory minimums, the EPA, and the
statewide implementation of the disciplinary credit structure are significant for
corrections, while only the budgetary pie is significant when the data is differenced. For

higher education, corrections, the size of the 18-24 year old cohort, and the partisan

129

 

ideology of the governor are significant when the data is in levels, while only the size of
the budgetary pie is significant when the data is differenced. To give these findings more

meaning, they will be directly related to the statistical hypotheses.

Table 5.6: Triangulation of Estimates in Michigan

 

 

 

Corrections ZSLS (levels) SUR (differenced)

Higher Education + (.000)
Pie - (.014)
Cohort - (.000)
Violent Crime
Unemployment
Governor + (.009)
Mandatory Minimums + (.013)
Discipline Credit - (.011)
EPA - (.020)

Higher Education ZSLS (levels) SUR (differenced)
Corrections + (.000)
Pie - (.008)
Cohort + (.067)
Enrollment
Unemployment - (.084)
Governor - (.000)

 

Note: +/- symbols are the signs of coefficients and p-values are in parenthesis (S .125).

The results in Michigan reveal: (1) that there is a tradeoff relationship between
appropriations for corrections and appropriations for higher education. Appropriations
for corrections grow at a faster rate than for higher education. (2) Exogenous factors help
create policy relevant conditions that impact budgetary allocations. More specifically,
the presence of a Republican governor leads to increases in appropriations for corrections
and declines in appropriations for higher education. Between 1970 and 1999 Republican
governors in Michigan placed a primacy on being tough on crime that led to increases in

prison populations and ultimately to increases in appropriations for corrections.

130

 

Additionally, increases in the size of the 18-24 cohort leads to increases in the amount
appropriated for higher education and decreases the amount appropriated for corrections.
(3) An expanding budgetary pie does not lead to increases in appropriations for either
corrections or higher education. (4) The implementation of specific public policies plays
a big role in determining the amount of monies appropriated for corrections. The
implementation of mandatory minimums leads to an increase in appropriations for
corrections. Furthermore, the use of the Prison Overcrowding Emergency Powers Act by
the governor leads to decreases in appropriations for corrections.

To obtain a richer view of these results predicted values are generated for fiscal
years 1972-1999 for the data in levels. Table 5.7 presents the relative contribution of
each of the independent variables towards the amount of monies appropriated for
corrections in each of the fiscal years. The predicted total appropriations from the
empirical model are presented as well as the actual amount appropriated.lo Examining
the predicted values gives a more complete picture of how the different predictors impact
total appropriations. This approach provides a way to look at the impact of exogenous
factors as they vary over time and the implementation of specific public policies.

The constant term suggests that absent any other influences corrections would
receive $884.5 million. Beyond the constant, the factor most responsible for the size of

appropriations for corrections is higher education. As discussed earlier, appropriations

 

'0 An inspection of the predicted values in comparison to the actual values reveals that the empirical model
does a good job in estimating appropriations for corrections. The only year in which the residual is
noticeably large is FY 1976. This corresponds to a year in which actual appropriations for corrections
increases from $50.9 million in the previous fiscal year to $100 million only to decline to $99 million in the
next FY. To account for this anomaly a model was run with a dummy variable inserted for FY 1976.
Estimation of the model with the dummy variable did not change the results in any meaningful way. The
same coefficients that are significant in the model without the dummy variable are significant in the model
with the dummy variable and the magnitudes are roughly the same. Since there is no theoretical rationale
or event that can account for this deviation, the results from the model without the dummy variable are
presented.

131

 

for corrections exceeds appropriations for higher education by an average of 4.8 percent
each year. Therefore, as appropriations for higher education rose from approximately
$378 million in FY 1972 to $1.60 billion in FY 1999, appropriations for corrections
attributable to higher education rose from $396.3 million to $1.68 billion.” The factor
that is most influential in mitigating the large magnitude of the constant and
appropriations attributable to higher education is the size of the 18-2-4 cohort. As the
cohort size increases, appropriations for corrections decline. In addition, the presence of
a Republican governor leads to increases in the size of appropriations in corrections of
$127.7 million. Furthermore, as the unemployment rate increases the amount of monies
appropriated for corrections increases. For example, when the unemployment rate was at
its highest in the early 1980’s it resulted in appropriations for corrections that approached
$90 million. At the end of the 1990’s when unemployment was on the decline
appropriations attributable to the unemployment rate amounted to $26.5 million. Finally,
the impact of the size of the budgetary pie and the violent crime rate are somewhat
marginal. Unlike California where the violent crime rate accounted for upwards of $5
billion, the violent crime rate in Michigan is associated with a reduction of appropriations
for corrections of a maximum of $50 million.

The implementation of correctional policies in Michigan had a mixed impact on
appropriations for corrections. The implementation of mandatory minimums accounted
for increases of roughly $150 million, while the implementation of disciplinary credits
led to a decline of $184.8 million. Finally, the use of the Prison Overcrowding

Emergency Powers Act by the governor, over a four-year period, led to a decline in

 

" Since corrections and higher education are viewed as a system of equations, higher education acts as both
an endogenous factor and an explanatory factor in the corrections equation.

132

appropriations of $157.7 million. Therefore, in the aggregate the impact of these policies
on the budget seems to cancel one another out. Nevertheless, recent sentencing policies
and changes in the state economy may lead to increased appropriations in the future.

The implementation of truth in sentencing comes at a time when corrections
receives roughly 16 percent of the entire state budget. This percentage is up dramatically
from the 2 percent that it received in the early 1970’s. While policies themselves have
not been accountable for these growths, this policy change will necessitate that more
monies be allocated towards corrections to accommodate increases in the prison
population. Likewise, this policy is implemented at a time when the unemployment rate
is at low for the period of time under study. As Michigan and other states again begin to
struggle with fiscal crises, appropriations for corrections may continue to rise in

Michigan at the expense of appropriations for higher education.

133

E

 

 

 

134

555.555. .858. 555.5 ~88: ~85... 85.5- 5.5.5~ 555.-. 5.5.5-: 8~5. ..8.-5. 55585 555.
5-.58. 58.555. 555.5 ~88: ~85... 58: 8555 555.-. ~8.~5~.- 55...~- 555.8. 55585 555.
555.55~. 55~.55~. 555.5 ~88: ~85: 5555... 85.8 555.-. 58.58.- 5-.~. 35.555. 555.85 ~55.
555.55~. 8558. 555.5 ~88: ~85: ~85... 5..~.~5 555.-. ~.5..-.. 5555. .8555. 55585 555.
5~5.-.. .855: 555.5 ~88: ~85... 55.55- 5.5.... 555.-. 5882- 58.~ 858... 55585 555.
58.5.. ~5...55. 555.5 ~88: ~85: ~5~.5: 5~..55 555.-. 5~..~.5.- ~55.5 .5585. 5885 555.
58.55 85.58. 555.5 ~88: ~85... 555.55- 55~.55 555.-. 5.558.. 5555- 555.5~5. 5885 555.
555.555 555.-5 555.5 ~88: ~85: 8555- 5~5.: 555.5 5....58: 58.5. ....555. 55585 ~55.
5~5.55~ 8558 555.5 ~88: ~85... ~88. 5.5.3. 555.5 58.55.. .5~.5~ 55..5-. 55585 .55.
5555.~ -55- 555.5 ~88: ~85... 555.? 5~5.: 555.5 3.5555: 555. 3.58. 55585 555.
5555.5 55558 555.5 ~88: ~85... 85.5... ~5~..5 555.5 855...: ~85. 555.55.. 55585 555.
55.55 8 .58 555.5 ~88 .- ~85: .855- 35.55 555.5 58.58: 5~5.- 5~5. .5. . 55585 555.
5.55.8. 8.555 555.5 ~88: ~85... ~55: ~5...~5 555.5 85.5%.- 5~55~ .5555. 55585 ~55.
8.58 555.5... 5~5.~5.- 555.5 ~85... 55.55. ~555~ 555.5 58.5.5: 555.5 85.58. 55585 555.
5555.5 ~555- 5~5.~5.- 555.5 ~85... ~35: 5~5.55 555.5 -5.555.- 5.~.- 55~.~55 55585 555.
838 5~58~ 5~5.~5.- 555.5 ~85: 85.: 85.~5 555.-. 58555.- 555..- ~55.~5~ 55585 85.
55...5~ .885 5~5.~5.. 555.5 ~85... 555: 58.- 555.-. 85.58: ~5~.- 8.58 55585 85.
5.5.-~ :88 555.5 555.5 ~85... 855... 55~5~ 555.-. 5-.5~5: ~35: 55...5~ 55585 ~55.
555.55. 58.5: 555.5 555.5 ~85... .855- 8~5. 555.-. .~..555.. .8.- 55~.555 55585 .55.
555.2. 55~.55~ 555.5 5555 ~85: 5~555- .558 555.-. 8.55.. .85. 55~55~ 55585 555.
5.5.5... ~88. 555.5 555.5 555.: 555.~5- ~5~..5 555.-. 85.555: 5~5.- .~..5..5 55585 5~5.
8~5... 555.55 555.5 555.5 555.5 5555: 55~.55 555.-. 555.58: 85.5.- 85.~.~ 55585 5~5.
5.5.55 ~55.5~. 555.5 555.5 555.5 ~55: ~555~ 555.-. ~5..~.5.- 58.5.. -..~8 55585 -5.
5855. 555.~.~ 555.5 555.5 555.5 85.... ~5~.5.. 555.-. 555.5~5.- 58.5 55~.5- 55585 5~5.
555.55 58.- 555.5 555.5 555.5 5~5.~5- 58.8 555.-. 555.35.. ~55.5. 58.58 55585 5~5.
58.55 5.5..5 555.5 555.5 555.5 8.55- 8...... 555.-. 5555.5: ~55.~. -..5~5 55585 5~5.
5555.. 85.. 555.5 555.5 555.5 .....55- 5~5.~.. 555.-. 58.55.... ~55.5. .5553. 55585 5~5.
5555.. .85.. 555.5 555.5 555.5 -5.55. 55~.~.. 555.-. 555555.. ~55.~. -~555 55585 -5.

5~5.~5: ~88: 555.: 855. ~.5.5 555.-. .555. 585 58.. 55585

Eugen. g0 <55 Lfiwso “Hugh.“ 25.0 :50; swung: 85260 .850 a... 8.85:5 gunman .5828 >5

.32—=8 .5 30:09:00 .0. 8.0.8.5032 .0. oc:.m> 030.025. Us.» 035...

CHAPTER 6

The third and final state examined in this dissertation is Virginia. Similar to
California and Michigan, Virginia experienced several policy reforms that shifted the
practice of sentencing from indeterminate to determinate. This chapter proceeds by
providing an overview of these policy changes and a brief review of the contextual
background in Virginia. This overview is followed by the estimation of models for
corrections and higher education appropriations. The estimation strategy laid out in

Chapter 3 is utilized as a framework to test the various budgetary hypotheses.

Virginia — Policies

In 1974, Virginia’s Judicial Conference passed a resolution stating judicial
opposition to the principles of mandatory sentencing. The conference spoke out against
the various forms of determinate and presumptive sentencing and asserted that judicial
discretion needed to be maintained (Shane-Dubow, Brown, and Olsen, p. 257). In spite
of the urging of the judiciary, the legislature soon thereafter passed legislation that would
begin a movement away from indeterminacy. In 1975, Virginia adopted a reorganization
of its criminal code by implementing a classification scheme that divided most criminal
offenses into six felony classes.l Each class prescribed both a minimum and maximum
sentence within which a trial judge (or jury) would set the minimum and maximum terms
of imprisonment. For example, offenders convicted of class I offenses would receive life

imprisonment or a sentence of death, while class V] offenses carried a penalty of one to

 

' Shane—Dubow, Brown and Olsen (1985) claim that the new code resulted in little substantial change from
previous Virginia law (257).

135

five years’ imprisonment or confinement in jail up to one year and fine up to $1,000. In
all cases where parole was a possibility, the actual release time from prison fell on the
parole board (See Table 6.1). The aim of this particular reform was to eliminate “the
unduly wide variations between minimum and maximum penalties and facilitating
rational grading of crimes by comparative seriousness” (257). Following the imposition
of this classification scheme, Virginia would undertake a series of reforms that would

include mandatory minimums and a move towards determinacy.2

Table 6.1: 1975 Virginia Sentencing Schema’

 

 

Class Minimum and Maximum Sentence
1 Life imprisonment or death
11 Twenty years to life imprisonment
III Five to twenty years imprisonment
IV Two to ten years imprisonment

One to ten years imprisonment or confinement in jail up to
one year and a fine up to $1,000

One to five years’ imprisonment or confinement in jail up to

V1 one year and a fine up to $1,000

 

 

2 In 1978 and 1979 a Presumptive Sentencing Act was introduced and later defeated in the General
Assembly. The acts called for the abolition of parole, the establishment of a council to generate sentencing
guidelines, and the elimination of programs designed to accelerate release dates from prison (Shane-
Dubow, Brown, and Olsen, 257).

3 Table adopted from Shane-Dubow, Brown, and Olsen (258).

136

In addition to establishing a set of maximum and minimum sentences for felony
classes, legislation in 1975 introduced mandatory minimums for offenses involving
firearms. The legislation made the use or display of a firearm in the commission of a
felony a separately punishable felony offense. First time offenders were to receive a one-
year sentence and a three-year sentence for subsequent offenses. These sentences were to
be served consecutively to any other sentence and could not be suspended or reduced by
parole (Shane-Dubow, Brown, and Olsen, 258). This law was subsequently amended in
1982, increasing first convictions to two years and subsequent convictions to four years.

Additionally, changes were made to the statutes relating to recidivistic behavior.
Prior to 1979, additional penalties were imposed upon habitual offenders in a
supplementary court proceeding. Offenders with one prior conviction would receive a
one year sentence enhancement and offenders with two and three previous convictions
would receive three and five year enhancements, respectively. The enhancements were
to be served consecutively with other penalties, the offender was not eligible for parole
during the enhancement period, and the sentence could not be suspended. Legislation
enacted in 1979 repealed previous recidivist statutes and incorporated penalties for repeat
offenders into the parole system.

Before the legislative changes most offenders were eligible for parole after
serving one-fourth of their sentence or twelve years (whichever was smaller). Under the
new laws these provisions remained the same for first time offenders, but parole
eligibility for repeat offenders was made harsher. Second time offenders were now
eligible for parole after serving thirteen years or one-third of the term; third time

offenders were eligible for parole after serving fourteen years or one-half of their term;

137

and offenders with four or more prior convictions were eligible for parole after serving
fifteen years or three-quarters of the term. For each of these, the smaller of the two terms
would be selected (Shane-Dubow, Brown, and Olsen, 25 8).

An additional reform was enacted in 1995 with the adoption of truth in

sentencing. Prior to the reform, indeterminate sentencing procedures allowed violent

criminals to be released early from prison by parole boards. For example, criminals l
convicted of first-degree murder served on average about 10 years of a 35-year sentence 1
and rapists sentenced to 9.2 years on average were released after 4.4 years. The Virginia i

Governor’s Commission on Parole Abolition and Sentencing Reform argued that if
serious offenders had been required to serve more time in prison, between 1986 and
1994, 4,375 crimes would not have been committed (Edwards, 479). In an attempt to
restore public confidence in the criminal justice system and to ensure certainty and
predictability in sentencing, the state of Virginia adopted a determinate sentencing format
in the form of sentencing guidelines and truth in sentencing.

Sentencing guidelines trace their origin to a series of newspaper articles and
reports that claimed that sentencing decisions in Virginia were inconsistent and disparate.
In response to these claims, Governor Charles Robb appointed a Task Force on
Sentencing, in 1982, to study sentencing policies and recommend any changes deemed
necessary to improve sentencing practices. In 1983, the Task Force concluded that
variation existed in the use of incarceration as a sanction and in the length of prison terms
for “similarly situated offenders.” They found that factors associated with disparity

included offender race and socioeconomic status (extralegal factors). As such, the Task

138

Force recommended a move towards deterrninacy in the form of historically based
sentencing guidelines (Ostrom et a1. 1999, 10).

The original guidelines that were established were based on historical sentencing
patterns across the state of Virginia. The Virginia Criminal Sentencing Commission
analyzed data from 33,573 felony cases sentenced between February 1985 and June 1987
and constructed a set of voluntary sentencing guidelines that were initially pilot tested in
six judicial circuits in 1988 (Ostrom et a1. 1999, 14). It was found that the voluntary
guidelines improved the extent to which similarly situated offenders who committed
similar crimes received similar sentences (consistency) and reduced variation in sentence
lengths due to extralegal factors (neutrality). As such, the voluntary guidelines were
adopted statewide in 1991.4 These guidelines were monitored annually and adjusted over
the next three years to ensure a correspondence between the guidelines and judicial
sentencing practice (Ostrom et a1. 1999, 16).

Additionally, in September 1994, the state of Virginia passed truth in sentencing
legislation (HB5001 and SB3001).5 The policy banned discretionary parole for all felony
offenses and stated that offenders convicted of violent crimes in Virginia must serve at
least 85 percent of their sentences.6 Governor George F. Allen stated that this legislation

allowed Virginia to restore “integrity and honesty and accountability to our criminal

 

’ The voluntary sentencing guidelines structure required judges to fill out a sentencing guidelines
worksheet for each sentenced offender. Under this system, judges still maintained discretion in their ability
to adjust the recommended sentences up or down depending upon circumstances, without having to explain
their departure. Between 1995 and 1997 the compliance rate with the guidelines was 75.2% with a
departure rate of 24.8% (Ostrom et a1. 1998, 13).

5 In 1994, Virginia also enacted a three strikes law. The law provided mandatory life without parole for
offenders convicted of a third felony involving designated acts of violence. Those sentenced under the law
who reached 65 years of age and who had served five years were allowed to petition the parole board for
conditional release (Edwards, 470). Shortly after the law was passed Virginia eliminated parole for all
sentences and zero offenders have been sentenced under this statute.

6 The truth in sentencing law defines a long list of crimes as violent, including: murder, rape, and robbery
and several offenses not traditionally viewed as violent (burglary, child pornography, and computer
crimes).

139

justice system” (Edwards, 479). On a similar note, Ostrom et a1. (1999) point out that
this legislation grew out of the perception among the public that the parole board’s
release decisions during the early 1990’s were too lenient (17). Thus, parole was
replaced by a system that made punishment more certain, predictable, and severe.

In conjunction with the new truth in sentencing legislation, the sentencing
guidelines were adjusted to reflect the new focus on deterrninacy and a ‘get tough’ policy
towards violent and repeat offenders. Under the new guidelines, the sentences for violent
criminals and offenders with prior felony convictions were increased so that repeat
violent offenders would serve sentences as much as five times longer than before, while
incarceration rates and time served levels remained consistent with past practices for non-
violent, non-repeat offenders. For example, someone convicted of first-degree murder
with no violent prior record would receive a sentence of 40 years and have to serve
around 34 years (up from 10 years). Additionally, an offender convicted of first-degree
murder with a serious violent record would receive a sentence of 84.6 years, up from 14.7
years under the statutes followed in 1988 through 1992. State criminal justice officials
calculated that these increased sentences would prevent 119,000 felonies, including
26,000 violent crimes over a 10-year period (Allen, 6). The state legislatures estimated
that these changes would necessitate the construction of 26 new prisons at a cost of about

$2 billion and entail additional billions for prison operations (Mauer, S6).
Virginia — Prison Population

Virginia’s prison population has grown from 4,981 inmates in 1971 to 28,385

inmates in 1997, representing an increase of 470%. Over this time period the

140

W157“ ‘H
.2 ._

incarceration rate per 100,000 inhabitants increased fi'om 108.9 to 412. The rapid
increase in the number of prisoners began in 1975, following a brief period of stability.
The prison population grew steadily until 1989, when a period of rapid grth began,

culminating with an 18% growth between 1993 and 1994.

 

30.000

25.000

20.000

15.000

Prisoners

10.000

 

 

 

 

5,000 <
0 5
v- 00 In N O) 1- (‘O l!) IN 0) v- 00 IO N
O) O) C) C) O) O) O) O) O) O) C) 0)
v- v- :- v- ‘- s- :- v- v- v- v- 1- a- :-

Year

Figure 6.1 : Number of Prisoners Under State Jurisdiction in Virginia: 1971-1997

Virginia — Independent Variable - Governors

During the period of study Virginia had eight governors, five of whom were
Republicans. Between 1970 and 1982 a Republican was governor. During this span the
following Republicans were governor: Linwood Holton (1970-1974), Mills Goodwin, Jr.
(1974-1978), and John Dalton (1978-1982). Over the next decade and a half Democrats

held the govemorship in Virginia. The governors during this time period were Charles

141

 

Robb (1982-1996), Gerald Baliles (1986-1990), and Douglas Wilder (1990-1994).
Finally, two successive Republican governors succeeded Governor Wilder. These were
George Allen (1994-1998) and Jim Gilmore (1998-present). As such, it can be seen that

Virginia had a Republican governor during 18 years of the time under study.

Virginia - Independent Variable — Unemployment Rate

During the time period under study the rate of unemployment in Virginia
fluctuated between a rate of 3 percent and 8 percent. The peak years of unemployment
occurred in 1975 (6.4), 1982 (7.7), and 1992 (6.4). These highs are significantly lower
than the unemployment rate found in Michigan in the mid 1980’s. As can be seen in
Figure 6.2 the rate of unemployment in Virginia is characterized by periods of increases
followed by periods of decline. The periods of growing unemployment occurred between
1973 and 1975; 1979 and 1982; 1989 and 1992. These periods of growth correspond to
those experienced in California. During these times of high unemployment states face
increased fiscal demands and hence added pressures on their budgets. Finally, after 1992
the unemployment rate declined to a low of 2.8 in 1999. It is hypothesized that rising
unemployment rates will lead to decreases in appropriations for higher education and

increases in appropriations for corrections.

142

IT fijfi .-.. _

9.0

 

 

 

 

 

.fl -.. ‘1“-nn-e _
4

0:
fl
8
>.
2
a.
E
e
C
D
2.0 29 2 9 2-- 22222
1.0..-. ._ .. 2 - ...... -- 2
0.0 Y 1 1 1 r 1 v.1 n
O N V to (D O N (0 CD 00
O) O) O) O) O) O) O) O: O’) O) O) O) O)
1'- 1- :- 1- :- 1- s- s- s- 1- ‘- 1— 1- 1- ‘-
FlecaIYear

Figure 6.2: State of Virginia: Unemployment Rate, FY 1970-1999

Virginia - Independent Variable — Violent Crime Rate

The violent crime rate in Virginia rose slightly between 1970 and 1975. During
this period of time the rate of violent offenses known to police per 100,000 inhabitants
increased from 302 in 1970 to 380 in 1975 (see Figure 6.3). Following a sharp decline
between 1975 and 1978, the violent crime rate remained fairly steady at 300 up until
1989. However, in the early 1990’s (1990-1993) the violent crime rate rose to 375. It
was in response to this growth that Governor Allen implemented many of the ‘tough on
crime’ policies in the mid 1990’s. In the remainder of the 1990’s the violent crime rate
experienced a steady decline to a rate of 315 in 1999. Between 1970 and 1999 the violent
crime rate in Virginia was significantly lower than the rate found in California (high of
1,200 in 1992) and Michigan. Nonetheless, all three states experienced an increase in the

early 1990’s followed by a decrease in the late 1990’s.

143

 

 

150 9 -2 __,--_-~ 1

Vlolent Crime Rate
N
8
w

 

 

 

1
50 __ _._- _ ___ - _._ .__.._- 5 ----.---._-_-_- k"
0 a ‘1 1' 1' T
O N V (D 00 O N (D N
O) O) Q G) O) 0') O) O) O) O) O) O)
1- !- v- 1- 1- :- ‘- 1- 1- v- 1- :- 1- I- 1-

Flecal Year

Figure 6.3: State of Virginia: Violent Crime Rate (Part I offenses known to police per
100,000 inhabitants), FY 1970-1 999

Virginia — Independent Variable — Cohort

The size of the 18-24 year old cohort increased from approximately 615,000
individuals in 1970 to 750,000 individuals in 1984 (see Figure 6.4). Following this
period of steady growth, the size of this cohort began to diminish. By 1997 the size of
this cohort was 648,000. At the end of the century this cohort increased again to 673,000

in 1999.

144

800.000

 

 

 

 

 

700.000
600.0005
5 500.000 22 2 . .....
§400,00022 , 2
Q
‘1‘
:2 300.000
200.000
100.000 -2 2 22 222 .- _________ _ ______
O N V (O 00 O N (O N

Fiscal Year

Figure 6.4: State of Virginia: 18-24 year old cohort, FY 1970-1999

Virginia — Independent Variable — Enrollment

Unlike California and Michigan, fall enrollment in public 4-year institutions of
higher education in Virginia increased steadily between 1970 (94,028 students) and 1999
(175,275). During this period enrollment in the Virginia system of higher education grew
roughly 85 percent. Increasing enrollments place additional demands on Virginia’s

fourteen four-year public institutions of higher education.

145

17“?

 

200.000

 

 

 

 

 

 

180,000«

€160,000

3140.000

I”

7,120,000

‘51

£100,000;

5 80.0002~2 - 2 2 -------

I:

£60,000. 2-2- l

240.000- g.

'" 20.0002-2 -- 2 ,, _ _- - -_ -. l
09 a :i :2 92181931878828 8 818 9§ ‘3
99999959929299.2933- ‘

FlacalYear

Figure 6.5: State of Virginia: Enrollment in Higher Education, FY 1970-1999

Virginia — Correlation Matrix

In order to assess the strength and directionality of bivariate relationships between
the variables in the corrections and higher education models a correlation matrix is
constructed. These correlations can also be used to diagnose the possibility of
collinearity, which could potentially lead to biased estimation of the effects of the

independent variables on appropriations for corrections and higher education.

146

Table 6.2: Correlation Matrix for Variables included in the Budget Model,
Virginia

 

Corrections Higher Ed Budgetary Cohort Enroll \golent Unemployment
. . . . . nme

Appropriations Appropnauons P1e (t-2) (t-2) “-2) (t-2)
Corrections __ .945 .519 -. 136 .919 .408 -.046
Appropriations .000 .004 .491 .000 .031 .817
Higher Ed .945 __ .628 .019 .942 .163 -. 150
Appropriations .000 .000 .924 .000 .397 .439
. .519 .628 -. 140 .51 1 .084 -.070
Budgetary P'“ .004 .000 " .468 .005 .663 .719
-. l 36 .019 -. 140 .100 -.504 .600
““0" ("2) .491 .924 .468 " .600 .005 .000
.919 .942 .51 1 . 100 .272 -.024
Em" ("2) .000 .000 .005 .600 " .147 .899
Violent Crime .408 . 163 .084 -.504 .272 -- .066
(t-2) .031 .397 .663 .005 .147 .730

Unemployment -.046 -. 150 -.070 .600 -.024 .066 __

(t-2) .817 .439 .719 .000 .899 .730

 

Note: Pearson correlation statistic followed by significance level

The results shown in Table 6.2 illustrate that all of the continuous independent variables
are significantly correlated and correlated in the hypothesized direction with
appropriations for corrections and appropriations for higher education, except for the size
of the 18-24 cohort and unemployment with corrections. These variables are both
correlated in a direction that is opposite of the hypothesized relationship and are not
statistically significant. Significant correlations in the hypothesized direction include:
appropriations for corrections being positively correlated with the size of the budgetary
pie and the violent crime rate. Furthermore, appropriations for higher education is
positively correlated with the size of the budgetary pie, enrollment, and negatively

correlated with the unemployment rate. These results are in line with previous theoretical

147

 

expectations. In addition, there are no correlations between independent variables that
exceed the magnitude of .7. This suggests that multicollinearity is not a problem.
Diagnostic procedures later in this chapter will more formally test for the presence of

multicollinearity that may lead to imprecise estimates.

Virginia — Two Stage Least Squares

Following the estimation strategy laid out in Chapter 3, estimation is conducted
for the Virginia data in both levels and in differences. Before estimation can occur the
data must be tested for unit roots, multicollinearity, and adjusted for the appropriate lag
structure. The decision to transform the data via differencing is made by referencing the
Augmented Dickey-Fuller (ADF) test. This test provides a way to look for unit roots and
hence avoid the problems associated with spurious regression. Table 6.3 presents the

results of the ADF tests.

148

 

Table 6.3: Unit Root Tests (ADF)

No Trend, No Intercept

Intercept, No Trend

Intercept and Trend

 

 

Corrections 7.69 3.808 -.292

Corrections d1 -l.799* -3.851*" -6.265*"

Higher Ed 4.852 1.525 -.525

Higher Ed d1 -1.113 -2.075"‘ -2.283

Higher Ed d2 -7.521"‘** -7.525"* -7.618*“

Cohort .725 -2.044 -2.555

Cohort d1 2682“" -2.639* -2.485 _ ;
Cohort d2 -8.147*" -7.979*** 4934*" l '
Violent Crime .034 -2022 -1959 .1
Violent Crime d1 4649*" 4542*" -4.450"'** 1 L
Enroll 4.440 -2.636"‘ -2.722

Enroll d1 2733*” -3.921”* 4255*"

Unemployment -.533 -1.975 -2.073

Unemployment d1 -4.41 1*" 4334*” 4468"" ;
Pie -942 -2724" 4019*"
Pie d1 -8.738*" -8.819”* -8.937*" .

 

Note: "' = .10; ** = .05; *** = .01 significance level

The results suggest that all of the continuous variables included in the model, except for

the budgetary pie, need to be transformed. It was determined that appropriations for ‘
corrections, the size of the 18-24 cohort, the violent crime rate, enrollment, and the
unemployment rate, all need to be first differenced to make these series 1(0). In addition,

appropriations for higher education is differenced twice in order to induce stationarity.

Furthermore, collinearity diagnostics in Stata reveal that the condition index between the

independent variables is 4.5 and none of the VIF’s associated with the independent

variables exceed 10. Therefore, potential problems associated with multicollinearity can

be ruled out. Finally, all non-endogenous variables are lagged two time periods to more

accurately reflect the impact of exogenous factors and public policies on the budgeting

149

process. Now that issues relating to the data have been addressed estimation is
performed on the data in levels and differences utilizing ZSLS.

Table 6.4 includes results of the two-stage simultaneous estimation for both data
in levels and the transformed data. This approach is utilized to prevent biased estimators
that would result from the use of OLS. The model in levels for corrections is able to
explain 99 percent of the variation and diagnostic tests rule out potential problems related
to autocorrelation.7 The results in levels illustrate that the coefficients associated with
higher education, the political ideology of the governor, and the implementation of
sentencing guidelines are significant at the .05 level. All of these coefficients are in their
hypothesized direction. The results of the estimation imply that corrections receives on
average 72 percent of the amount appropriated for higher education, holding all other
factors constant. Put another way, for every dollar higher education receives, corrections
receives 72 cents. Thus, while appropriations for corrections have been increasing from
roughly $13.5 million in FY 1971 to over $573 million in FY 1999, the gain is not
coming at the expense of higher education, when controlling for the influence of the other
explanatory factors. In addition, the presence of a Republican governor leads to an
average increase in appropriations for corrections of $33.4 million per year. This implies
that Republican governors in Virginia are willing to expand the criminal justice system
while in office. Furthermore, the implementation of sentencing guidelines and the
resultant tougher sentences for violent and repeat offenders leads to an average increase

in appropriations for corrections of approximately $70.9 million per year, holding all

 

7 At the sixth lag, the Ljung-Box Q stat is insignificant (.1050) and the Q stat generated from the
Portrnenteau test for white noise is insignificant (.2510). Furthermore, heteroskedastic consistent
estimators are used by applying robust estimation.

150

other factors constant. Finally, when the data is differenced none of the coefficients in

the corrections model are significant.

Table 6.4: Two Stage Least Squares (robust) — Levels and Differenced, Virginia

 

 

 

 

 

 

 

 

Levels Differenced

Corrections Higher Ed Corrections Higher Ed

Coef t-value Coef t-value Coef t-value Coef t-value
Constant 51.329 0.23 -639.11 "-5.86 6.410 0.04 -32.70 -1.69
Corrections -- -- 1.289 ”7 .70 - - 0.0653 1.04
Higher Education 0.7176 "3.89 -- -- -.0246 -.003 -- -- -,
Pie -0.0339 -0. 18 0.0561 1.44 -0.041 -0.29 0.0439 1.37
Cohort -0.0005 -1 .69 0.0016 "5.51 .0011 1.61 0.0008 1.02 I:
Enrollment -- -- -0.0013 -0.76 -- -- 0.0020 0.87 3 ,
Violent Crime 0.4615 1.24 -- -- 0.100 0.21 -- -- ,
Unemployment 2.7097 0.18 -26.298 "-2.74 -1.991 -0.49 -2.826 -0.35 '
Governor 33.442 ”2.60 —45.896 "-2.66 -12.98 -0.19 0.9212 0.06
Mandatory Min 15.682 0.40 -- -- 4.180 0.05 -- --
Repeat Offender -8.236 055 -- -- 25.70 0.33 -- --
Sentencing Guide 70.930 “2.43 -- -- 26.83 0.43 -- --
Truth in Sentencing 33.442 -0.59 —- -- 21.01 0.13 -- --
it2 0.989 0.978 0.644 0.175
Hausman (chi2 prob) 0.0018 0.0016 0.972 0.3683

Note: Corrections and Higher Education in millions; “ significant at .05

The model for higher education in levels is able to explain 98 percent of the

variance and diagnostic tests for autocorrelation reveal that potential inflated variances

associated with this condition do not exist.8 The coefficients associated with corrections,

the size of the 18-24 cohort, the unemployment rate, and the political ideology of the

governor are all significant and in the hypothesized direction for higher education. These

results suggest that for every dollar corrections receives, higher education receives one

dollar and an additional 29 cents. Second, higher education can be seen to be responsive

to the demand of an increasing cohort of individuals (1 8-24) who are most likely to

pursue some form of postsecondary education. For every additional individual in this

 

 

8 At the sixth lag, the Ljung-Box Q stat is insignificant (.1986) and the Q stat generated from the
Portrnenteau test for white noise is insignificant (.3071). Furthermore, heteroskedatic consistent estimators
are used by applying robust estimation.

151

cohort $1,600 is appropriated for higher education. Third, appropriations for higher
education are responsive to fluctuations in the fiscal well being of the state. As the
unemployment rate increases by one point the amount of monies appropriated for higher
education declines by $26.3 million. Fourth, while Republican governors can be seen to

be supportive of corrections, their presence leads to a decline in appropriations for higher

education of $45.9 million. Finally, when the data is differenced none of the coefficients I
in the higher education model are significant. ;
E

The results from the 2SLS estimation suggest that there is in fact not a tradeoff P ‘

between appropriations for corrections and appropriations for higher education. In fact
the results in levels show that appropriations for higher education grow at a faster rate
than corrections holding all other factors constant. However, before a tradeoff
relationship between the two series can be ruled out it is necessary to test and see if in
fact simultaneity between these two series is an appropriate assumption using the Durbin-
Wu-Hausman test. The test results are reported in Table 6.4. The test reveals that when
the data is in levels the null hypothesis that there are no systematic differences between
OLS estimates and 2SLS estimates is rejected for both corrections and higher education.
This implies that the two series are indeed a system of equations and that 2SLS is the
appropriate methodology. Furthermore, when the data is differenced the OLS estimates
for corrections and higher education are found not to be systematically different from the
estimates derived from the simultaneous equation model. This implies that there is in
fact no simultaneity bias in the OLS estimates of the equation for corrections and for
higher education and explicit interdependence can be ruled out between the series. As

such, seemingly unrelated regression is undertaken for the differenced data to determine

152

if the error terms of the two series are correlated. Thus, the relationship between
appropriations for corrections and appropriations for higher education will be tested for a

seemingly unrelated relationship.

Virginia - Seemingly Unrelated Regressions
Table 6.5 presents the results for the seemingly unrelated regression for the

differenced data. In the model for corrections the coefficients associated with the size of

”PM “.—.L_._

the 18-24 cohort, the implementation of repeat offender legislation, and the
implementation of sentencing guidelines are all significant and in the hypothesized
direction. The model is able to account for 61 percent of the variation and diagnostic
tests reveal that autocorrelation is not a problem.9 The results suggest that on average an
increase in the size of the 18-24 year old cohort by one from the previous year leads to an
increase in the amount of appropriations for corrections from the previous year of $1,050.
In addition, the implementation of harsher penalties for repeat offenders leads to an
increase in appropriations of $21 million from the previous year. Finally, the
implementation of sentencing guidelines leads to an average increase of $22.8 million
from the previous fiscal year, holding all other factors constant. For higher education,

only the budgetary pie is found to be a significant factor.10

 

9 At the sixth lag, the Ljung-Box Q stat is insignificant (..0615) and the Q stat generated from the
Portrnenteau test for white noise is insignificant (.3071). Furthermore, since SUR is a GLS procedure
heteroskedasticity is not a problem.

'0 Since higher education is second differenced, the interpretation of the coefficient of .0054 is difficult to
interpret in any meaningful fashion.

153

Table 6.5: Seemingly Unrelated Regression, Virginia

 

 

Corrections Higher Education

Coefficient z-value Coefficient z-value
Constant -0.30556 -0.03 -20.7354 -l.54
Pie -0.00338 -0.33 0.053967 "2.04
Cohort 0.00105 ”3.75 0.000331 0.57
Enroll -- -- 0.001720 0.88
Violent Crime 0. 10400 0.81 -- --
Unemployment -2.4945 -O.84 -3.49238 -0.5 1
Governor -8.6817 -1.08 1.872707 0.14
Mandatory Minimums 3.16106 0.44 -- --
Repeat Offender 20.9648 "2.29 -- --
Sentencing Guidelines 22.8119 ”3.27 -- --
Truth in Sentencing 13.7597 1.41 -- --
R2 0.6673 0.1641

Note: Corrections and Higher Education in millions; Data differenced; " significant at .05

In levels it was found that there is no tradeoff between appropriations for
corrections and appropriations for higher education. In differences an explicit
interdependence has been ruled out, but it is still possible that the residuals of the single
large regression analysis are significantly correlated. If the residuals are significantly
correlated it can be concluded that what is driving the unexplained portion of the models
are a function of similar processes or variables. If they are not correlated, we can then
assume that the two series are unrelated and being driven by separate processes. Table
6.6 presents the results of the Breusch-Pagan test for the SUR. The test shows that the
residuals from the two equations are correlated at a value of .2887. Not only is the
correlation positive, suggesting that there is not a seemingly unrelated tradeoff
relationship, but the relationship is also insignificant. In conjunction with the ZSLS

results in levels, it can be inferred that there is not a tradeoff between appropriations for

154

an.” i—.\

corrections and appropriations for higher education in Virginia between fiscal year 1970

and fiscal year 1999.

Table 6.6: Seemingly Unrelated Results (Breusch-Pagan Test), Virginia

Correlation Chi-square Probability
Differenced-Differenced 0.2887 2.2151 0.1335

 

 

This chapter has explored the impact of a variety of exogenous factors and
policies on appropriations for corrections and appropriations for higher education in
Virginia. Estimation procedures followed two distinct paths, one for the data in levels
(2SLS) and one for the data in differences (SUR). A comparison of these two paths is
shown in Table 6.7. This figure shows the sign of the coefficient and the p-values in
parentheses for all those coefficients that have p-values less than .125. Convergence
between the results occurs when the two paths generate significant coefficients with the
same signs. For corrections convergence occurs for the implementation of sentencing
guidelines. Both paths show that implementation led to an increase in appropriations for
corrections. Only one of the paths produces coefficients with p-values less than .125 for
higher education, the budgetary pie, the political ideology of the governor, and the
implementation of repeat offender legislation. In addition, the signs of the coefficients
associated with the size of the 18-24 year old cohort are in conflict with one another.
Finally, for higher education, only one of the paths produces coefficients with p-values
less than .125. These factors are corrections, the budgetary pie, the size of the 18-24

cohort, the unemployment rate, and the political ideology of the governor. In order to

155

TI

make sense of these results, they will be directly related to the statistical hypotheses laid

out earlier in this dissertation.

Table 6.7: Triangulation of Estimates in Virginia

 

‘1 9.0.) run. 2. LL")
.
r
.

 

 

Corrections ZSLS (levels) SUR (differenced)
Higher Education + (.001)

Pie - (.079) I
Cohort - (.109) + (.000)
Violent Crime

Unemployment

Governor + (.019)

Mandatory Minimums

Repeat Offender + (.022)
Sentencing Guidelines + (.026) + (.001)
Truth in Sentencing

Higher Education ZSLS (levels) SUR (differenced)
Corrections + (.000)

Pie + (.041)
Cohort + (.000)

Enrollment

Unemployment - (.012)

Governor - (.015)

 

Note: +/- symbols are the signs of coefficients and p-values are in parenthesis (S .125).

The results in Virginia reveal: (1) that there is not a tradeoff relationship between
corrections and higher education appropriations. Instead, appropriations for corrections
grows at a rate that is slower than higher education holding all other factors constant.

(2) Exogenous factors create policy relevant conditions that impact budgetary allocations.
First, the presence of a Republican governor leads to increased appropriations for
corrections and decreased appropriations for higher education. Second, an increase in the
size of the 18-24 year old cohort leads to increases in appropriations for both corrections

and higher education. Due to the fact that this variable is not significant at the .10 level

156

for corrections, it is more plausible that increases in the size of the cohort lead to
increases in appropriations. Third, appropriations for higher education increase during
times of fiscal well being (as the unemployment rate declines). Finally, several factors
found to be significant predictors of appropriations in other states are not significant in
Virginia. This includes the violent crime rate and unemployment for corrections and the
size of enrollment for higher education. (3) An expanding budgetary pie leads to
increases in appropriations for higher education. The budgetary pie is not a significant

predictor of appropriations for corrections. (4) The implementation of specific public

fit ‘5”. a. my?“
A ‘

policies significantly impacts appropriations for corrections. The implementation of
sentencing guidelines and resultant ‘harsher’ sentences leads to an increase in
appropriations. Furthermore, the implementation of a repeat offender policy leads to
increases in appropriations for corrections. While truth in sentencing was not a
significant predictor of appropriations, the true impact of this recent policy change may
be felt sometime in the future. Assuming that commitment rates remain relatively
constant and offenders are mandated to serve longer sentences, appropriations for
corrections will have to increase to meet a growing demand in the firture.

To gain further insights into the meaning of the empirical results, predicted values
are generated for each independent variable for FY 1972-1999. Table 6.8 presents the
relative contribution of each independent variable and the predicted and actual
appropriations for corrections in millions of dollars. These calculations are made using
the data in levels.

The constant term suggests that absent any other influences corrections would

receive $51 million. Beyond the constant those factors that play a large role in

157

determining the size of appropriations are higher education, the size of the 18-24 year old
cohort, the violent crime rate, and the implementation of sentencing guidelines. The
majority of the grth in appropriations for corrections is a function of corrections
responding to increases in appropriations for higher education. Furthermore, while the
size of the 18-24 cohort and the violent crime rate have a large impact, their values
remain relatively constant over the range of years presented. Between FY 1972-1999 the
range of values associated with the cohort is -$300 million to -$369 million, or a

difference of $69 million. Similarly, the range for the violent crime rate is $132 million

to $176 million, or a difference of $44 million dollars. Thus, while these factors have a
large magnitude the two series do not fluctuate very much over the time period under
study and their effect on changes in the size of appropriations for corrections is
statistically insignificant. In contrast, the implementation of sentencing guidelines led to
increases in appropriations for corrections in the 1990’s.

Policies implemented in the 1990’s dramatically influenced the amount of monies
appropriated for corrections. Appropriations for corrections attributable to sentencing
guidelines are $71 million and for truth in sentencing $33 million. In conjunction with a
Republican governor, these three factors are responsible for over $130 million of the
predicted $618 million in FY 1999. The implementation of these policies brought with
them longer sentences and hence placed an additional burden on correctional resources.
Similar to California, the implementation of these policies corresponds to a period when
the violent crime rate was on the decline. For example, in FY 1998 and 1999 the violent
crime rate was 326 and 314 per 100,000 inhabitants, respectively, in Virginia. If this rate

was to increase and all other factors were to remain constant, the impact on

158

appropriations would be a dramatic rise in appropriations for corrections. For example, if
the rate in FY 1999 were to have been at Virginia’s high (380) during the time of study,
appropriations for corrections would have been $633 million rather than $618 million.
Likewise if the rate in FY 1999 had been equivalent to the rate found in Michigan in the
late 1980’s (800), appropriations would have risen to $818 million and if the rate was
equivalent to the rate found in California in the early 1990’s (1,000), appropriations !
would be $919 million. These scenarios reflect that the sentencing policies have been 3 I

implemented in Virginia during a time when some of the factors that may lead to further

 

increases in appropriations for corrections have been declining. This is true for both the
violent crime rate and for the unemployment rate. Therefore, Virginia faces a potential

problem similar to that found in California. If the crime rate and/or unemployment were
to increase, an additional strain would be placed not only on the correctional system, but

on the state budget as well.

159

 

 

 

160

353.5~5 53~.~.5 ~38 555.5~ 58.? 355.5 .3555. 5555. ~38 5.3.5.5- 55~.~5- 55585 5~5..5 555.
5~5...5 5~..5~5 ~38 555.5~ 55~.5- 355.5 5.5.~5. -5... ~38 5.~.5.5- ~3..5- 8~.5.5 5~5..5 555.
555.~53 85.-5 ~38 555.5~ 55~.5- 355.5 5~5.55. 55..~. ~38 53558- 5~5.5.- 555.5~3 5~5..5 ~55.
~55.353 35~.353 555.5 555.5~ 55~.5. 355.5 .5555. -~.5. ~38 55~.5~5- 553.5.- 35..-3 5~5..5 555.
3~55~3 -..353 555.5 555.5~ 58.? 355.5 55~..~. 5.5.5. 555.5 ~5~.555- .~5.~- ~55.3.3 5~5..5 555.
-5.555 858 555.5 555.5~ 55~.5- 355.5 5~5.5~. ~35.~. 555.5 35..~55. 5~5.5.- 35.~55 5~5..5 355.
55.58 553.55~ 555.5 555.5 58.5. 355.5 55~.-. ~85. 555.5 38.35. 5-.~.- 33.555 5~5..5 555.
555.38 55~.55~ 555.5 555.5 58.5. 355.5 5.5..5. ~55... 555.5 5~...55- 555.5- 833 5~5..5 ~55.
.~5.-5 553.-5 555.5 555.5 58.5. 355.5 ~5~.3. 555.5. 555.5 55~.555- 555.5 555.3 5~5..5 .55.
.5~.55~ 53~.5.5 555.5 555.5 8~5. 355.5 555.55. 555.5. 555.5 85555. 5~5.... 35~.553 5~5..5 555.
-555~ ~3..~5~ 555.5 555.5 55~.5. 355.5 8.8. .55... 555.5 5.5.~55- 5555.- 85553 5~5..5 555.
555.55~ 3.5.58 555.5 555.5 55~.5- 355.5 5-..3. 535.5. 555.5 53~.~55- ~55.-- 355.55 5~5..5 555.
~.5.35~ 5~3.5.~ 555.5 555.5 58.5. 355.5 .5555. 3~..5. 555.5 555.555- -5..: ~53.555 5~5..5 ~55.
53..5~. 33.55. 555.5 555.5 55~.5. 355.5 -3.55. 535.5. 555.5 .5585- 55.5.- 55.5.5 5~5..5 555.
35555. ~5~.53. 555.5 555.5 58.5. 355.5 5.5.55. 5~5.5. 555.5 ~55.555- ~35- 555.55~ 5~5..5 555.
5~5.55. 5~..55. 555.5 555.5 8~5. 355.5 555.~3. 3555~ ~38 5~5.555. 58.5.- ~53.~5~ 5~5..5 355.
.5553. ~5553. 555.5 555.5 55~.5. ~85. 5~3.53. 5~5.5. ~38 3~585- 535.5- 858 5~5..5 555.
~55.~5. 5.~.~: 555.5 555.5 58.5. ~85. 35~..3. 535.5. ~38 .5~.~55- 555.5. 535.35~ 5~5..5 ~55.
5.5.55 .835 855 555.5 58.5. ~85. :58. 55~.~. ~38 38.85- 55~.~- .55..- 5~5..5 .55.
55~.5~ ~55..5 555.5 555.5 555.5 ~85. 5~..~5. ~83. ~38 5~3.~55- 8~5.- ~5~.~5. 5~5..5 555.
35~5~ 5~5.55 555.5 555.5 555.5 ~85. 555.8. .55.... ~38 58.555- 553.~- 5~5.5~. 5~5..5 5~5.
5~5.55 -5..3 555.5 555.5 555.5 ~85. 5~5.~3. ~85. ~38 55~.535- 3~5. .55..3. 5~5..5 5~5.
555.5 5~..55 555.5 5555 555.5 ~85. 3-.5~. ~35.~. ~38 555.535- 5555- 55.55. 5~5..5 -5.
.88 53~53 555.5 555.5 555.5 555.5 355.35. 55..~. ~38 555.-5- 5~5.5- .55.5.. 5~5..5 5~5.
53~.- 555.5~ 555.5 555.5 555.5 555.5 ~55.55. 55~.5 ~38 555.5.5- 3~3.3. 5~5.35. 5~5..5 5~5.
~85. -..- 555.5 555.5 85.5 555.5 355.55. 55~.5 ~38 35~.555- 5.5.? 55.55 5~5..5 3~5.
5.5.~. 5.... 555.5 555.5 555.5 555.5 5.5.55. 5~5.5. ~38 5.5.5.5- ~553- ~5555 5~5..5 5~5.
555.5. 55~.~- 555.5 555.5 555.5 555.5 555.55. 5.~.5 ~38 -5.55~- 5.5.5- 8555 5~5..5 -5.
~38 555.5~ 55~.5. ~85. ~535 5.~.~ ~38 555.5 355.5- 5.~.5 5~5..5
.333 8.28... 2.2.355 85.5250 .8..on 2.55.5: 2....0 5.3. o... 5:858
2282.00 58.82.60 c. 52. 552.355 .38... 5.28:... .56... 5.56.552: 36260 :28 5.85855 .23.: 282.60 C

800.5... .5 000302.00 .8 3033.00.93 .0. oe:_n> 080.00... 6.0 0.00...

 

CHAPTER 7

Budgets themselves represent the endgame of a struggle for scarce resources that
reflect government priorities. Public budgeting itself is characterized by a number of
claimants who make requests for resources and conserving forces in the form of political
actors who wish to exercise control over the budgetary process and authoritatively ration !
resources (Schick 1988). Requests for resources are made in attempt to address societal h 1

needs and problems. However, state governments operate in an environment of limited

 

revenue and thus are incapable of addressing all needs and demands for resources.
Potentially states could increase taxes to meet demands or overspend and incur budgetary
deficits, but neither is a popular alternative for elected representatives. Therefore,
budgets represent the politics of choice in an arena of fiscal constraints (Rubin 2000). In
this context budgets represent a culmination of a process of competition whereby choices
are made each year to allocate resources to diverse interests and priorities. This
dissertation has focused on the factors that are determinants of appropriations for
budgetary categories. Specifically, it focused on those factors that determine the amount
of monies appropriated for corrections and higher education in California, Michigan, and
Virginia. These two budgetary categories represent components of a state’s budget
where there is not a legislative or constitutional mandate requiring a particular level of
spending. Studying two discretionary budget categories allows for an examination into
the ways resources are allocated to diverse interests and claimants and a way to look at

potential tradeoffs between budgetary categories.

161

 

 

One of the hypotheses examined in this dissertation was whether or not there was
a tradeoff between corrections and higher education. To test this hypothesis a series of
equations was estimated and where this methodology was deemed inapplicable a
seemingly unrelated regression equation was estimated. Despite the multitude of claims
by scholars and pundits that the recent growth in corrections has come at the direct
expense of higher education, the empirical results in this study generated mixed P

I I . 4
conclusrons.

In California it was revealed that the relationship between appropriations for

 

corrections and appropriations for higher education is not one of a system of
simultaneous equations. Instead, the relationship is recursive with corrections found to
be a determinant of appropriations for higher education, while higher education was
found not to be a determinate of appropriations for corrections. An examination of the
unidirectional relationship did find that corrections is growing at a faster rate than higher
education when all other factors are held constant. Similarly, in Virginia the notion of a
tradeoff between the two budgetary categories was ruled out. Simultaneous equation
estimation revealed that higher education is growing at a faster rate than corrections when
all other factors are constant. These results give some credence to Domke, Eichenberg,
and Kelleher’s (1983) assertion that “direct priority setting - the trading off of one
program against the other - is rare or nonexistent in democratic governments” (20). In
contrast, the results in Michigan, for the data in levels, suggest that there is a tradeoff
between appropriations for corrections and appropriations for higher education. It was
found that corrections and higher education represent a system of equations where

corrections receives on average 104.8 percent of the amount appropriated for higher

162

 

education and higher education receives only 88 percent of that appropriated for
corrections. Therefore, it can be seen that in Michigan appropriations for corrections is
growing at a faster rate than appropriations for higher education.

The results of tests for the existence of an explicit tradeoff between corrections
and higher education are somewhat muddled. Nevertheless, by its very nature the
activity of budget setting is characterized by an implicit tradeoff relationship. This point
is reinforced by Rubin (2000) who states that all “[a]ll budgets contain tradeoffs” (132).
This implies that every dollar spent towards one budgetary category truly represents an
opportunity cost for dollars that could be spent elsewhere. Given that state revenues and
the amount of monies available for state expenditures are limited; state governments are
forced to make decisions about what they consider to be priorities. Those arenas that are
viewed as the most valued priority are given monies that potentially could be used for
other programs and activities. While not necessarily an explicit tradeoff, increases in
appropriations for corrections from a state’s general fund could be re—designated to
alternative means (e. g., higher education). This begs the question, how are priorities set
and what other factors lead to increases in appropriations for specific budgetary
categories? One way to understand these processes is to look at the idea of policy
windows and more specifically at changes that occurred in the sentencing policies and

practices of corrections.

163

 

 

 

Policy Windows and Changes in Correction’s Policy

Policy changes for corrections have come in two separate waves in California,

Michigan, and Virginia. The first wave of changes dealt with determinate sentencing,

mandatory minimums, and repeat offender legislation and occurred between the late

1970’s through the beginning of the 1980’s. The second set of changes introduced Three

Strikes, truth in sentencing, and sentencing guidelines during the 1990’s. The various

policies that were implemented in these three states, and focused on in this dissertation,

are presented in Table 7 . 1. Each of these states implemented a wide range of correctional

reforms that dramatically altered the way that offenders were viewed and sentenced.

These reforms resulted in increased prison populations and the need for additional

resources to support an increasing demand.

Table 7.1: Summary of State Reforms (dates implemented)

 

. Recidivism Truth Good
Mandatory Sentencmg Parole .
Minimums Guidelines (Three 1“ Mechanisms T‘m"
Strikes) Sentencing Modifications
. . 1982
California 1982 --- l 994 --- 1977 ---
Michigan 1977 1999 1998 1980 1985
1 978
Virginia 1975 1991 1979 1995 --- «-

 

The two waves of policy reforms were implemented during times when a

confluence of factors arose that served to open windows of opportunity for change and

reform. Policy windows become open when there are changes in the political stream

164

fimmuaJ-nw

 

(change in administration), or problems arise that need solutions. As such, policy
windows are periods in time that provide an opportunity for action on given initiatives.
These moments present policy entrepreneurs with a chance to match (couple) their
policies to the ideology of the political stream or to problems that need solutions
(Kingdon 1995). Rubin (2000) notes that one way of dealing with budgetary competition
“is to try to raise the priority of one’s spending proposal, and one way of increasing the
priority of projects or programs is to represent them as solutions to crises or emergencies”
(135). The policy reforms to the criminal justice system that occurred between 1971 and
2000 can be explained within this framework.

The first wave of policies was pushed through during a time when the
rehabilitative or medical model came under attack. Martinson’s (1974) seminal work,
which reviewed 200 prison programs, found that ‘nothing works’ (Clark 1994). This led
critics to question the philosophical foundations of indeterminate sentencing, particularly
the notion of rehabilitation. Simultaneously, there was a coalition of interests in
opposition to indeterminate sentencing on the grounds that there was excessive disparity
among sentences in presumably similar cases and that offenders often were released
before their maximum sentences were reached. The first concern centered on the
possibility that discriminatory sentencing practices based upon the race of the offender
were taking place. The second concern focused on the fact that parole boards would
ofien release prisoners well before maximum sentences were reached (Blumstein 1984,
130). The argument was put forth that these offenders did not serve time commensurate

with the offense committed. As such, states moved away from indeterminacy to a system

of determinate sentencing that had at its foundation the idea of retribution. Don

165

mama—m—

 

Gottfredson, commenting on the move towards determinate sentencing, states that
“sentencing purposes have shifted from an emphasis on utilitarian aims, particularly
treatment, toward a greater focus on deserved punishment, on the proportionality of
sanctions to harms done, and on equality” (Gottfredson 1984). These policies were
implemented at a time when policy windows were open as Republican leaders, who were
tough on crime, were in office and a problem with the status quo was identified.

The second wave of policy reforms was implemented during a time when the
violent crime rate was rising in most states and the states under study had a Republican
governor in office. During the 1980’s the violent crime rate began to rise, reaching a
pinnacle in the early 1990’s. In California the violent crime rate rose from under 800
violent offenses known to police per 100,000 inhabitants in the mid 1980’s to over 1,000
by 1994. This fact combined with Republican governors in office in the early and mid
1990’s created an environment for change. Politicians campaigned on ‘tough on crime’
platforms and implemented reforms when elected that increased sentences for repeat
offenders and forced most offenders to serve a significant portion of their mandated
sentences.l Once again a window of opportunity was opened and policies were
implemented that increased the severity of sentences and ultimately increased the demand
for correctional resources.

Correctional policies were being passed at precisely the same time that demand
was increasing for state services. Exogenous factors, in the form of policy relevant

conditions, created an environment for budgetary expansion. It was found that

 

' During this time period politicians painted a picture of society that was under threat from criminals. Since
rehabilitation was not effective the only solution was to lock up offenders and ‘throw away the key.’ This
message was reinforced by intense rhetoric and the use of imagery (e. g., Polly Klaus case) to define the
problem.

166

Winn“ $4.5—fl a..."

 

appropriations for corrections are responsive to certain exogenous factors. In all three
states the presence of a Republican governor led to increases in appropriations for
corrections. Additionally, in California appropriations for corrections were responsive to
increases in the violent crime rate and in Michigan appropriations were responsive to an
increase in the size of the 18-24 year old cohort. Therefore, the same factors that were
instrumental in opening windows of opportunity for new policies were also associated
with creating windows of opportunity for budget actors. The states responded to the
increases in demand by appropriating more for corrections. In this light the same
mechanisms that are used to understand policymaking are useful for understanding
budgeting. Therefore, state budgets were responsive to two forces at the same time.
They were responsive to exogenous demands from the environment and to resource
demands placed upon the system by new policies.

At the end of the 1990’s budgets for corrections continued to rise in California,
Michigan, and Virginia. The policies enacted as ‘tough on crime’ placed additional
burdens on the budgetary resources of the states. However, many of these policies were
implemented during a time when environmental factors that act as policy relevant
conditions for the expansion of the correction’s budgets were declining and states were
experiencing fiscal good times. For example, in California the violent crime rate was
roughly 625 in 1999 in comparison with 1,100 in 1992. In Michigan, the violent crime
rate was at 575 in 1999 down from over 800 in the early 1990’s. Likewise the
unemployment rate was under four at the end of the 1990’s well under the double-digit
rate found in the 1980’s. Finally, in Virginia the violent crime and the unemployment

rate were also on the decline in the late 1990’s. If these policy relevant conditions, which

167

 

 

 

represent a decline in the demand for corrections, were to revert to levels found at the end
of the 1980’s and early 1990’s states potentially could be faced with a new fiscal crisis.
The implementation of corrections’ policies has increased the likelihood of commitment
to prison for those arrested and the time served by those sentenced. This has dramatically
increased the demand for additional resources.2 If the unemployment rate and/or the
violent crime rate were to increase in these states, demand for additional resources would
increase even more. For example, a simulation in this study found that in California if
the violent crime rate in 1997 had been the same rate as in the early 1990’s corrections
would have risen from $4.7 billion to roughly $6.0 billion, ceteris paribus. Therefore, at
the start of the new millennium states find themselves at a crossroads. How will they
respond to pressures of prison overcrowding and rising prison costs? Will states undergo
another round of prison building? Or will states respond by once again changing their

approach to sentencing?

Policies as a Pendulum

As Kingdon (1995) has suggested policies gain prominence at times when three
streams (political, problems, and policy) come together. During the late 1970’s and again
in the mid to early 1990’s criminal justice policies rose on the agenda and determinate
sentencing and ‘tough on crime’ policies were implemented in response to a problem. In

order to gain a better understanding of these changes and link this to the position that

 

2 For many of the policies implemented during this time period it is still too early to realize the full impact
of the reforms. This includes the implementation of Three Strikes in California and truth in sentencing in
Michigan and Virginia. The full effects of these programs may be for an even greater demand for
resources.

168

 

 

state governments find themselves in, a metaphor equating policy change and a pendulum
clock will be constructed.

A grandfather clock is made up of a collection of weights, cogs, pulleys, and a
pendulum. These apparatus work together to keep an accurate account of time. When
the time on a pendulum clock is inaccurate it is adjusted by moving the pendulum bob up
or down to shorten or lengthen the path of the pendulum. If the clock is drastically off
the mark the bob must be moved a great distance, whereas if it is off a little it must only
be moved only a short distance. After the bob has been adjusted it may take up to a full
day to realize that the clock is still not keeping the correct time.

Similar to the bob in a pendulum, policies are often introduced to change the
status quo (present time) when it is perceived that there is a problem. In the case of
corrections, a move away from indeterrninant sentencing was an attempt to correct
perceived deficiencies associated with both the philosophical foundation and practice of
sentencing. This shift was a radical departure from past practices. The question remains
did the policy changes go to far? Like the shift in the bob, the full impact of policy
changes cannot be known for sometime. While policymakers can anticipate the
consequences of changes there are often unintended consequences and projections and
forecasts of impacts are often inaccurate. In the case of sentencing the consequence of
the reforms was a massive growth in the number of inmates in state prisons. Between
1970 and 1999 the number of offenders in state prisons grew from 175,000 to over 1.2
million. This rapid growth placed a tremendous strain on the correctional resources that
existed. In response, states were forced to increase the number of prisons, personnel, and

resources to support an expanding system. During this period states faced overcrowding

169

“will m?n 4

 

of their prisons and what some have called a crisis in corrections. Confronted with an

ever-expanding prison system and the potential for a fiscal crisis many states are moving
away from tougher criminal laws by reducing the stringency of their statutes and moving
back towards the notion of rehabilitation. In a sense, they are readjusting the sentencing

clock.

Intermediate Sanctions

As discussed above recent policies have led to a very rapid expansion of the
prison system. As a potential solution to this problem many reformers are advocating the
use of intermediate sanctions. Morris and Tonry (1990) argue that prison crowding and
the resultant fiscal pressure requires attention. They contend that what is needed is a
range of meaningful sanctions between probation and prison. This point is reinforced by
von Hirsch, Wasik, and Greene (1989) who state that “[w]ith prisons overfilled and the
costs of new prison construction rising, the need has become apparent for credible
sanctions that can be administered in the community” (597). The range of intermediate
sanctions includes: restitution, community service, fines, residence in community
detention centers, house arrest, and inpatient and outpatient treatment for drug and
alcohol. Wood and Gramsmick (1999) maintain that intermediate or alternative sanctions
are attractive because: (1) they are believed to reduce prison overcrowding by channeling
offenders out of prisons and into community based punishments; (2) the level of risk
some offenders represent is too high for probation, but not great enough for prison; (3)
alternative sanctions are typically less expensive than imprisonment; and (4) alternatives

presumably offer offenders a better chance at rehabilitation. Both California and Virginia

170

 

 

are presently experimenting with policies that move sentencing back towards the ideal of
rehabilitation.

In California a recent referendum was passed that mandates treatment rather than
prison for many drug offenders. The Substance Abuse and Crime Prevention Act
(Proposition 36) was passed by 61% of California voters on November 7, 2000. This
initiative allows first and second time nonviolent, simple drug possession offenders the
opportunity to receive substance abuse treatment instead of incarceration. It has been
projected that this initiative, which went into effect on July 1, 2001, will save California
taxpayers $1.5 billion over the first five years. According to the State Legislative
Analyst’s Office, the Substance Abuse and Crime Prevention Act will divert
approximately 25,000 non-violent drug possession offenders per year into drug treatment
instead of prison. By reducing the number of offenders in the state correctional system
operating costs for corrections will be reduced by between $200 and $250 million.
Furthermore, parolee supervision caseload will be reduced by as much as 9,5000
parolees. Finally, it is projected that this policy change will reduce the need for state
prison beds needed for drug offenders by between 10,000 and 12,000. This will result in
the delay of construction of additional prison space and lead to a one-time savings in
deferred or avoided prison construction costs of $475 and $575 million (California
Campaign for New Drug Policies).3

Virginia has also recently reevaluated their sentencing practices. In conjunction
with the introduction of truth in sentencing, the Virginia General Assembly charged the
Virginia Criminal Sentencing Commission with the task of studying the feasibility of

placing 25 percent of drug, fraud, and larceny offenders in alternative (non-prison)

 

3 Currently the state is expected to pay for construction of another new prison by 2003.

171

sanctions by using an empirically based risk assessment instrument. The instrument is
used to identify offenders who have the lowest probability of being reconvicted of a
felony crime at the time of sentencing. The risk assessment instrument is presently being
pilot tested in six judicial circuits in Virginia. The Virginia Criminal Sentencing
Commission (1999) feels that the “risk assessment instrument can be viewed as an
important component to help maximize the utilization of alternative punishments for
nonviolent offenders while minimizing threat to public safety and reserving the most
expensive correctional space for the state’s violent offenders” (77). Risk assessment and
diversion are an option only for offenders who do not have any current or prior
convictions for violent felonies or offenders who sell an ounce or more of cocaine.

A recent evaluation of the development and impact of the risk assessment
instrument conducted by the National Center for State Courts finds that the instrument is
effective in predicting recidivistic behavior and thus identifying ‘good’ candidates for
diversion. Furthermore, the benefit/cost component of their analysis finds that if the
program were to be implemented statewide the net societal benefit would be between
82.9 and $3.6 million (Ostrom et al. 2001). This program is illustrative of a shift towards
policies that are seeking alternative sanctions that will result in less offenders being

committed to prison and ultimately result in cost savings for state governments.4

 

4 New York State has also recently expressed a desire to move away from harsh determinate sentences.
New York Governor George Pataki has proposed a reform of the Rockefeller drug laws. These laws are
among the most severe in the United States. For example, a person convicted of selling more than two
ounces, or possessing more than four ounces of cocaine or heroine must be sentenced to a minimum of 15
years in prison. Pataki’s proposal would allow an appeals court to reduce prison terms by up to a third for
first-time, nonviolent drug offenders, expand drug treatment alternatives, and allow trial judges, with the
consent of the prosecutor, to divert some drug defendants to substance abuse programs instead of prison.

172

REFERENCES

Abramsky, Sahsa. 1999. “When They Get Out.” The Atlantic Monthly, Junez30-36.

Achen, Christopher. 2000. “Why Lagged Dependent Variables Can Suppress the
Explanatory Power of Other Independent Variables.” Paper prepared for the
Annual Meeting of the Political Methodology Section of the American Political
Science Association, UCLA, July 20-22, 2000.

Allen, George. 1995. “The Courage of Our Convictions: The Abolition of Parole Will
Save Lives and Money.” Policy Review Spring: 4-7.

Alt, James E. and Robert C. Lowry. "Divided Government, Fiscal Institutions, and
Budget Deficits: Evidence from the States." American Political Science Review

88:811-828.

Ambrosio, Tara-Jen and Vincent Schiraldi. 1997. “From Classrooms to Cell Blocks: A
National Perspective.” The Justice Policy Institute. http://www.cjcj.org/jpi/
highernational.html (November 19, 1999).

Austin, James. 1996. “The Effect of ‘Three Strikes and You’re Out’ on Corrections. In
Three Strikes and You ’re Out: Vengeance as Public Policy, eds. David Shicor and
Dale K. Sechrest. USA: Sage Publications.

Auten, Gerald, Barry Bozeman, and Robert Cline. 1984. "A Sequential Model of
Congressional Appropriations." American Journal of Political Science 28:503-
523.

Bartels, Larry M. 1991. “Instrumental and ‘Quasi’- Instrumental Variables.” American
Journal of Political Science 35, 3: 377-800.

Beck, Allen J. and Jennifer C. Karberg. 2001. “Prison and Jail Inmates at Midyear
2000.” Bureau of Justice Statistics Bulletin, US. Department of Justice.

Belsley, David A., Edwin Kuh, and Roy E. Welsch. 1980. Regression Diagnostics:
Identifying Influential Data and Sources of Collinearity. New York: John Wiley
and Sons.

Berry, William D. 1990. "The Confusing Case of Budgetary lncrementalism: Too Many
Meanings for a Single Concept." The Journal of Politics 52:167-196.

Berry, William D. and David Lowery. 1990. "An Alternative Approach to

Understanding Budgetary Trade-offs." American Journal of Political Science
34:671-705.

173

Berry William D., Evan J. Rinquist, Richard C. Fording, and Russell L. Hanson. 1998.
“Measuring Citizen and Government Ideology in the American States, 1960-
1993.” American Journal of Political Science 42, 1:327-348.

Blumstein, Alfred. 1984. “Sentencing Reforms, Impacts, and Incarcerations.”
Judicature 68, 4-51129-139.

Blumstein, Alfred. 1987. “Sentencing and the Prison Crowding Problem.” In
America ’s Correctional Crisis: Prison Populations and Public Policy, eds.
Stephen D. Gottfredson and Sean D. McConville. New York: Greenwood Press.

Blumstein, Alfred and Allen J. Beck. 1999. “Population Growth in US. Prisons, 1980-
1996.” In Crime and Justice: A Review of Research, Volume 26, eds. Michael
Tonry and Joan Petersilia. Chicago: University of Chicago Press, 17-61.

Blumstein, Alfred, Jacqueline Cohen, Susan E. Martin, and Michael H. Tonry, eds. 1983.
Research on Sentencing: The Search for Reform, Vol. 1, Washington DC:
National Academy Press.

Borcherding, Thomas E. 1985. "The Causes of Government Expenditure Growth: A
Survey of the US. Evidence." Journal of Public Economics 28:359-382.

Bozeman, Barry and Jeffrey D. Straussman. 1982. “Shrinking Budgets and the
Shrinkage of Budget Theory.” Public Administration Review 42:509-515.

Caiden, Naomi. 1988. “Shaping Things to Come: Super-Budgeters as Heroes (and
Heroines) in the Llate-Twentieth Century.” In New Directions in Budget Theory,
ed. Irene S. Rubin. Albany: State University of New York Press.

Caldeira, Greg A. and Andrew T. Cowart. 1980. "Budgets, Institutions, and Change:
Criminal Justice Policy in America." American Journal of Political Science
24:413-438.

California Campaign for New Drug Policies. “The California Correctional System.”
http://www.drugreform.org (October 7, 2001).

California Criminal Law Observer. 1998. “The Three ‘Strikes and You’re Out’ Law:
The Proposition 184 Three Strikes Initiative Effective November 9, 1994.”
http://www.silicon-valley.com/star2.html (June 8, 2000).

California Secretary of State’s Office. 1994. “1994 California Voter Information:
Analysis of Proposition 184 by the Legislative Analyst.” http://ca94.e1ection.
Di gital.com/e/prop/ 1 84/analysis.html (May 1, 2000).

Campaign for an Effective Crime Policy. 1998. “‘Three Strikes’ Laws: Five Years
Later.” Executive Summary. http://www.crimepolicy.org/pdf73 StrikesLast.pdf

174

(June 2, 2000).

Caputo, David A. 1975. "New Perspectives on the Public Policy Implications of
Defense and Welfare Expenditures in Four Modern Democracies: 1950-1970."
Policy Sciences 6:423-446.

Carroll, Stephen and Eugene Bryton. 1996. “The Impact of a Federal Balanced Budget
on California’s Budget.” Rand. http://www.rand.org/publications/CF/CF123/
carroll/ (September 11, 2000).

Casler, Stephen D. 1992. Introduction to Economics. United States: HarperCollins
Publishers, Inc.

Chambliss, William J. 1991. Trading Textbooks for Prison Cells. Virginia: National
Center on Institutions and Alternatives.

Cherryholmes, Cleo H. 1999. Reading Pragmatism. New York: Teachers College
Press.

Clark, Charles S. 1994. “Prison Overcrowding.” CQ Researcher 4, 5: 97-120.

Clear, Todd R. and Patricia M. Harris. 1987. “The Costs of Incarceration.” In
America 's Correctional Crisis: Prison Populations and Public Policy, eds.
Stephen D. Gottfredson and Sean D. McConville. New York: Greenwood Press.

Cohen, Jacqueline and Michael H. Tonry. 1983. “Sentencing Reforms and Their
Impacts.” In Research on Sentencing: The Search for Reform, Vol. 2, cds. Alfred
Blumstein, Jacqueline Cohen, Susan E. Martin, and Michael H. Tonry.
Washington DC: National Academy Press.

Cohen, Jacqueline and Jose A. Canela-Cacho. 1994. “Incarceration and Violent Crime:
1965-1988.” In Understanding and Preventing Violence, vol. 4, Consequences
and Control, eds. Albert J. Reiss, Jr. and Jefii'ey A. Roth, Washington D.C.:
National Academy Press.

Connolly, Kathleen, Lea McDermid, Vincent Schiraldi, and Dan Macallair. 1996.
“From Classrooms to Cellblocks: How Prison Building Affects Higher Education
and African American Enrollment in California.” The Justice Policy Institute.
http://www.cjcj.org/jpi/california.html (November 19, 1999).

Crimestrike. “Stopping the Fraud: Truth-in-Sentencing.” National Rifle Association.
http://www.nra.org/crimestrike/cstis.html (May 31, 1998).

Cyert, Richard M. and James G. March. 1963. Behavioral Theory of the Firm.
Englewood Cliffs: Prentice Hall.

175

 

 

Davies, Malcom. 1985. “Determinate Sentencing Reform in California and Its Impact
on the Penal System.” The British Journal of Criminology 25, 1:1-30.

Davis, Otto A., M. A. H. Dempster, and Aaron Wildavsky. 1966. "A Theory of the
Budgetary Process." American Political Science Review 60:529-547.

Davis, Otto A., M. A. H. Dempster, and Aaron Wildavsky. 1974. “Towards a Predictive
Theory of Government Expenditure: US Domestic Appropriations.” British
Journal of Political Science 4:419-452.

Davis, Mike. 1998. Ecologv of F ear: Los Angeles and the Imagination of Disaster.
New York: Metropolitan Books, Henry Holt and Company.

Dilulio, John J ., Jr. 1999. “Against Mandatory Minimums; Drug Sentencing Run
Amok.” National Review 51, 9.

Ditton, Paula M. and Doris James Wilson. 1999. Truth in Sentencing in State Prisons.
Bureau of Justice Statistics Special Report. US. Department of Justice, Office of
Justice Programs mCJ 170032).

Domke, William K., Richard C. Eichenberg, Catherine M. Kelleher. 1983. "The Illusion
of Choice: Defense and Welfare in Advanced Industrial Democracies, 1948-
1978." American Political Science Review 77:19-35.

Donziger, Steven R., ed. 1996. The Real War on Crime: The Report of the National
Criminal Justice Commission. New York: HarperPerennial.

Durr, Robert. 1993. “What Moves Policy Sentiment.” The American Political Science
Review 87, 1:158-170.

Eckl, Corina. 1998. “The Costs of Corrections.” State Legislatures 24, 2: 30—33.
Edwards, Margaret. 1995. “Mandatory Sentencing.” CQ Researcher 5, 20: 467-487.

Feeley, Malcom M. and Sam Kamin. 1996. "The Effect of ‘Three Strikes and You’re
Out’ on the Courts.” In Three Strikes and You ’re Out: Vengeance as
Public Policy, eds. David Shicor and Dale K. Sechrest. USA: Sage Publications.

Fellner, Jamie. 1997. “Cruel and Usual: Disproportionate Sentences for New York Drug
Offenders.” Human Rights Violations in the United States 9, 2 (B). Human
Rights Watch. http://www.hrw.org/hrw/summaries/s.us973.htrnl (June 1, 2000).

Gangi, Robert, Vincent Schiraldi, and Jason Ziedenberg. 1998. “New York State of
Mind?: Higher Education vs. Prison Funding in the Empire State, 1988-1998.”

The Justice Policy Institute. http://www.cjcj.org/jpi/nysom.html (November 19,
1999)

176

Gelman, Arthur M. 1978. Sentencing Guidelines: Structuring Judicial Discretion.
National Institute of Law Enforcement and Criminal Justice.

Gilliard, Darrell K. and Allen J. Beck. 1998. "Prisoners in 1997." Bureau of Justice
Bulletin. US. Department of Justice.

Gist, John R. 1982. "'Stability' and 'Competition' in Budgetary Theory." American
Political Science Review 76:859-72.

Gottfredson, Don M. 1984. “Criminal Sentencing in Transition.” Judicature 68, 4-
5:125-127.

Gottfi'edson, Stephen. 1995. “Fighting Higher Crime at the Expense of Colleges.” The
Chronicle of Higher Education 41, 19:31.

Granato, Jim. 1991. “An Agenda for Econometric Model Building.” Political Analysis
3: 123-154.

Granger, Clive WJ. and Paul Newbold. 1974. “Spurious Regressions in Econometrics.”
Journal of Econometrics 2: l l l-120.

Greene, William H. 1990. Econometric Analysis. Third Edition. New Jersey: Prentice
Hall.

Greenwood, Peter, C. Peter Rydell, Allen F. Abrahamse, Jonathan P. Caulkins, James
Chiesa, Karyn E. Model, and Stephen P. Klein. 1996. “Estimated Benefits and
Costs of California’s New Mandatory-Sentencing Law.” In Three Strikes and
You ’re Out: Vengeance as Public Policy, eds. David Shicor and Dale K. Sechrest.
USA: Sage Publications.

Gregory, Warren G. 1989. “The Crisis in the Michigan Criminal Justice System: Causes,
Consequences, and Options.” Michigan House Fiscal Agency.

Griset, Pamela L. 1991. Determinate Sentencing: The Promise and Reality of
Retributive Justice. Albany: State University of New York Press.

Griset, Pamela. 1994. “Determinate Sentencing and the High Cost of Overblown
Rhetoric: The New York Experience.” Crime and Delinquency 40, 4:532-548.

Griset, Pamela L. 1995. “Determinate Sentencing and Agenda Building: A Case Study
of the Failure of A Reform.” Journal of Criminal Justice 23, 4:349-362.

Gujarati, Damodar N. 1995. Basic Econometrics. Third Edition. New York: McGraw-
Hill, Inc.

177

Hammond, Paul Y. 1968. "A Functional Analysis of Defense Department Decision-
Making in the McNamara Administration." American Political Science Review
62:57-69.

Hanushek, Eric A. and John E. Jackson. 1977. Statistical Methods for Social Scientists.
San Diego, California: Academic Press, Inc.

Harriman, Linda and Jeffrey D. Strausmann. 1983. “Do Judges Determine Budget
Decisions? Federal Court Decisions in Prison Reform and State Spending for
Corrections.” Public Administration Review 43:343-351.

Hayes, Margaret Daly. 1975. “Policy Consequences of Military Participation in Politics:
An Analysis of Tradeoffs in Brazilian Federal Expenditures.” In Comparative
Public Policy: Issues, Theories and Methods, eds. Craig Liske, William Loehr,
and John McCamant. New York: Sage Publications, Inc.

Holahan, John F. and Joel W. Cohen. 1986. Medicaid: The T radeofl between Cost
Containment and Access to Care. Washington DC: The Urban Institute Press.

Hoole, Francis W., David H. Handley, and Charles Ostrom, Jr. 1979. "Policy-Making
Models, Budgets and International Organizations." Journal of Politics 41 :923-
932.

Hossler, Don, et a1. 1997. “State Funding For Higher Education.” The Journal of
Higher Education 68, 2:160-190.

Howard, Kenneth S. 1973. Changing State Budgeting. Chapel Hill: Council of State
Governments.

Huang, W.S. Wilson, et a1. “Individual and Contextual Influences on Sentence Lengths:
Examining Political Conservatism.” The Prison Journal 76, 4:398-419.

Hurwitz, Mark S. 1998. The Nature of Agenda in the United States Supreme Court and
Courts of Appeals. Unpublished Ph.D. Dissertation, Michigan State University.

Jacobs, James B. 1983-1984. “The Politics of Prison Expansion.” The New York
University Review of Social Change 12, 12209-241.

Jacobs, James B. and Laura Berkowitz. 1983. “Reflections on the Defeat of New York
State’s Prison Bond.” In New Perspectives on Prisons and Imprisonment, ed.
James B. Jacobs. Ithaca: Cornell University Press.

Jakes, Lara and John Caher. 1999. “Draconian Drug Laws No Panacea.” Times Union.
http://www.timesunion.com/news/special/druglaws/draconian.stm (June
5, 2000).

178

Johnson, Perry M. 1983-1984. “Response to Jacobs.” The New York University Review
of Social Change 12, 12249-252.

Johnston, Jack and John DiNardo. 1997. Econometric Methods. Fourth Edition. United
States: McGraw-Hill Companies, Inc.

Jones, LR. 1985. University Budgeting for Critical Mass Competition. New York:
Praeger Publishers.

Kaludis, George, ed. 1973. Strategies for Budgeting: New Directions for Higher
Education: 1, 2. San Francisco: Jossey-Bass, Inc., Publishers.

Kamlet, Mark S. and David C. Mowery. 1980. “The Budgetary Base in Federal
Resource Allocation.” American Journal of Political Science 24:804-821.

Kamlet, Mark S. and David C. Mowery. 1983. “Budgetary Side Payments and
Government Growth: 1953-1968.” American Journal of Political Science 27:
636-664.

Kamlet, Mark S. and David C. Mowery. 1987. "Influences on Executive and
Congressional Budget Priorities, 1955-1981." American Political Science Review
81 :155-1 78.

Kelleher, Catherine M., William Domke, and Richard Eichenberg. 1980. "Guns and
Butter: Patterns in Public Expenditure in the United States and Western Europe,
1920-75." In Defense Politics of the Atlantic Alliance, ed. Edwin H. F edder.
New York: Praeger Publishers.

Kennedy, Peter. 1992. A Guide to Econometrics. Third Edition. Cambridge,
Massachusetts: MIT Press.

Key, V0. 1940. “The Lack of a Budgetary Theory.” American Political Science
Review 34, 621137-1141.

King, Gary. 1989. Unifying Political Methodology: The Likelihood Theory of Statistical
Inference. Cambridge: Cambridge University Press.

Kingdon, John W. 1995. Agendas, Alternatives, and Public Policies. Second Edition.
Boston: Little, Brown and Company.

Kopel, David B. 1994. “Prison Blues: How America’s Foolish Sentencing Policies
Endanger Public Safety.” Policy Analysis 208.

Kraan, Dirk-Jan. 1996. Budgetary Decisions: A Public Choice Approach. Cambridge:
Cambridge University Press.

179

“‘ “1515:5112. nix-anon—

 

Kritzer, Herbert M. 1996. “The Data Puzzle: The Nature of Interpretation in
Quantitative Research.” American Journal of Political Science 40, 1:1-32.

Ku, Richard. 1980. American Prisons and Jail, Vol. 4. Supplemental Report Case
Studies of New Legislation Governing Sentencing and Release. National Institute
of Justice, US. Department of Justice.

Lagoy, Stephen P, Frederick A. Hussey, and John H. Kramer. 1978. “A Comparative
Assessment of Determinate Sentencing in the Four Pioneering States.” Crime and
Delinquency 24, 4:385-400.

Langan, Patrick A., John V. Fundis, and Victoria W. Schneider. 1988. “Historical
Statistics on Prisoners in State and Federal Institutions, Yearend 1925-86.”
Bureau of Justice Statistics, US. Department of Justice.

 

Langan, Patrick A. 1991. “America’s Soaring Prison Population.” Science 251: 1568-
1573.

Layzell, Daniel T. and Jan W. Lyddon. 1990. Budgeting for Higher Education at the
State Level: Engima, Paradox, and Ritual. ASHE-ERIC higher Education Report
4. Washington DC: The George Washington University, School of Education
and Human Development.

Lee, Robert Jr. and Ronald W. Johnson. 1998. Public Budgeting Systems, Sixth Edition.
Maryland: Aspen Publishers, Inc.

Legal Aid Society. 2000. “Reform the Rockefeller Drug Laws: Mandatory Sentencing
and Drug Offenders in New York State.” http://www.legal-aid.org/rock.htm
(April 12, 2000).

Lemox, Penelope. 1997. “The End of the Prison Boom.” Governing 10:32-33.

Leslie, Larry L. and Garey Ramey. 1986. “State Appropriations and Enrollments: Does
Enrollment Growth Still Pay?” Journal of Higher Education 56, 1:1-19.

Lindblom, Charles E. 1959. "The Science of Muddling Through " Public
Administration Review 19:79-88.

Looney, Robert E. and Peter C. F rederiksen. 1996. "Defense Expenditures and
Budgetary Patterns in Selected Middle Eastern and Mediterranean Countries: An
Assessment." Public Budgeting and Financial Management 8293-105.

Lowery, David, Thomas Konda, and James Garand. 1984. "Spending in the States: A
Test of Six Models." Western Political Quarterly 37:48-66.

Lynch, James P. and William J. Sabol. 1997. “Did Getting Tough on Crime Pay?” The

180

Urban Institute, Crime Policy Report.

Macallair, Dan, Khaled Taui-Eddin, and Vincent Schiraldi. 1998. “Class Dismissed:
Higher Education vs. Corrections During the Wilson Years.” The Justice Policy

Institute. http://www.cjcj.org/jpi/classdis.html (November 19, 1999).

Mailloux, Stephen. 1990. “Interpretation.” In Critical Terms for Literary Study, eds.
Frank Centricchia and Thomas McLaughlin. Chicago: University of Chicago
Press.

Marlow, Michael L. and Alden F. Shiers. 1999. “Do law enforcement expenditures
crowd-out public education expenditures?” Applied Economics 31:255-66.

Matthews, Dewayne. 1998. “Transforming Higher Education: Implications for State
Higher Education Finance.” Educom Review, 48-57.

Mauer, Marc. 1996. “The Truth About Truth in Sentencing.” Corrections Today 58, 1:
S1(8).

McEleney, Barbara Lavin. 1985. Correctional Reform in New York. United States:
University Press of America.

McKeown, Mary P. and Kern Alexander, eds. 1986. Values in Conflict: Funding
Priorities for Higher Education. Cambridge, Massachusetts: Ballinger Publishing
Company.

Menand, Louis. 2001. The Metaphysical Club: A Story of Ideas in America. New York:
F arrar, Strauss and Giroux.

Mergenhagen, Paula. 1996. “The Prison Population Bomb.” American Demographics.
http://www.demographics.com/publications/AD/96_AD/9602_AD/AD880.HTM
(June 5, 2000).

Messinger, Sheldon L. and Phillip E. Johnson. 1978. “California’s Determinate
Sentencing and Statute: History and Issues.” In Determinate Sentencing: Reform
or Regression? Proceeding s of the Special Conference on Determinate
Sentencing June 2-3, 1977. National Institute of Law Enforcement and Criminal
Justice.

Michigan Council on Crime and Delinquency. 1993. “Trends in the Michigan Criminal
Justice System: From Crisis to Chaos.” Executive Summary.

Michigan Department of Corrections. 1998. “1998 Annual Report.” http://www.state.
mi.us/mdoc/publication/ 1 998annual.html (June 1, 2000).

Michigan House of Representatives Corrections Committee. 1985. “Report to the House

181

Standing Committee on Corrections: Investigation of Michigan Department of
Corrections Release Policies and Practices.”

Miller, J .L., Marilyn McCoy Roberts, and Charlotte A. Carter. 1981. Sentencing
Reform: A Review and Annotated Bibliography. National Center for State Courts.

Mills, Terrence. 1991. Time Series Techniques for Economists. Cambridge: Cambridge
University Press.

Mintrom, Michael. 1997. “Policy Entrepreneurs and the Diffusion of Innovation.”
American Journal of Political Science 41 :73 8-770. l

Mintz, Alex. 1989. "Guns Versus Butter: A Disaggregated Analysis." American
Political Science Review 83: 1285-1293.

 

Mintz, Alex and Chi Huang. 1991. "Guns Versus Butter: The Indirect Link." American
Journal of Political Science 35:738-757.

Morris, Norval and Michael Tonry. 1990. Between Prison and Probation: Intermediate
Punishments in a Rational Sentencing System. New York: Oxford University
Press.

Mullen, Joan “State Responses to Prison Crowding: The Politics of Change.” In
America ’s Correctional Crisis: Prison Populations and Public Policy, eds.
Stephen D. Gottfredson and Sean D. McConville. New York: Greenwood Press.

Nagin, Daniel. 1979. “The Impact of Determinate Sentencing Legislation on Prison
Population and Sentence Length: A California Case Study.” Public Policy 27, 1:
69-98.

National Council on Crime and Delinquency. 1998. 1996 National Survey of State
Sentencing Structures. Bureau of Justice Assistance.

National Institute of Corrections. 1995. “State Legislative Actions on Truth in
Sentencing: A Review of Law and Legislation in the Context of the Violent
Crime Control and Law Enforcement Act of 1994.” US. Department of Justice.

Nicol, Chuck. 1996. “Status: Accommodating Prison Growth.” California Legislative
Analyst’s Office. http://www.lao.ca.gov/sc10695a.html (September 28, 1999).

Okun, Arthur M. 1975. Equality and Efficiency: The Big T radeofl. Washington, DC:
Brookings Institution.

Ostrom, Brian J ., Fred Cheesman, Randall Hansen, Neal Kauder, and Matthew Kleiman.

2001. An Evaluation of the Virginia Risk Assessment Instrument. National
Center for State Courts, National Institute of Justice.

182

Ostrom, Brian J. and Roger A. Hanson. 1999. Efficiency, Timeliness, and Quality: A
New Perspective from Nine State Criminal Trial Courts. National Center for
State Courts.

Ostrom, Brian J ., Neal B. Kauder, David Rottman, and Merideth Peterson. 1998.
Sentencing Digest: Examining Current Sentencing Issues and Policies. National
Center for State Courts. (R-204).

Ostrom, Brian J ., et al. 1999. T ruth-in-Sentencing in Virginia: Evaluating the Process
and Impact of Sentencing Reform. National Center for State Courts, National
Institute of Justice.

Ostrom, Charles W., Jr. 1978. "A Reactive Linkage Model of the US. Defense
Expenditure Policymaking Process." American Political Science Review 72:941-
957.

Ostrom, Charles W., Jr. 1990. Time Series Analysis: Regression Techniques. Newbury
Park, California: Sage Publications.

Ostrom, Charles W., Jr. and Robin F. Marra. 1986. "US. Defense Spending and the
Soviet Estimate." American Political Science Review 80:819-842.

Padgett, John F. 1980. "Bounded Rationality in Budgetary Research." American
Political Science Review 74:354-372.

Perlman, Ellen. 2000. “Terms of Imprisonment.” Governing 13, 7:23-26.

Peroff, Kathleen and Margaret Podolak-Warren. 1979. "Does Spending on Defence
Cut Spending on Health?" British Journal of Political Science 9:21-39.

Petersilia, Joan. 1992. “Crime and Punishment in California: Full Cells, Empty Pockets
and Questionable Benefits.” In Urban America: Policy Choices for Los Angeles
and the Nation, eds. James B. Steinberg, David W. Lyon, and Mary E. Vaiana.
Santa Monica, California: Rand.

Phillips, A.W. 1958. "The Relation between Unemployment and the Rate of Change of
Money Wage Rates in the United Kingdom, 1861-1957." Economica 25:283-99.

Postsecondary Education Opportunity. 1994. “Infatuation with Incarceration.” The
Mortenson Research Letter on Public Policy Analysis of Opportunity for
Postsecondary Education 28:7-13.

Postsecondary Education Opportunity. 1997. “American Infatuation with Incarceration.”
The Mortenson Research Letter on Public Policy Analysis of Opportunity for
Postsecondary Education 64: 1 5-19.

183

Pryor, F.L. 1968. Public Expenditures in Communist and Capitalist Nations.
Homewood: Richard D. Irwin.

Richardson, Lewis F. 1960. Arms and Insecurity. Pittsburgh: Boxwood Press.

Rubin, Irene S. 2000. The Politics of Budgeting: Getting and Spending, Borrowing, and
Balancing. New York: Chatham House Publishers.

Russett, Bruce M. 1969. "Who Pays For Defense?" American Political Science Review
63:412-426.

Russett, Bruce M. 1982. "Defense Expenditures and National Well-being." American
Political Science Review 76:767-777.

Schick, Allen. 1988. “An Inquiry into the Possibility of a Budgetary Theory.” In New
Directions in Budget Theory, ed. Irene S. Rubin. Albany: State University of
New York Press.

Schlosser, Eric. 1998. "The Prison-Industrial Complex." The Atlantic Monthly,
December:51-77.

Sentencing Project. 2001. “US. Surpasses Russia as World Leader in Rate of
Incarceration.” http://www.sentencingproject.org/brief/usvsrus.pdf (July 6, 2001).

Shane-DuBow, Sandra, Alice P. Brown, and Erik Olsen. 1985. Sentencing Reform in the
United States: History Content and Effect. Washington DC: National Institute of
Justice.

Sharkansky, Ira and Augustus B. Tumbull III. 1969. "Budget-Making in Georgia and
Wisconsin: A Test of a Model." Midwest Journal of Political Science 13:631-
645.

Siedman, David and Michael Couzens. 1974. “Getting the Crime Rate Down: Political
Pressure and Crime Reporting.” Law and Society 457-489.

Silver, Sheldon, Joseph Lentol, and Daniel Feldman. 1997. “The Toughest Crime Laws
in a Generation.” New York State Assembly. http://assembly.state.ny.us/
Reports/Codes/ 1 99704/ (July 18, 2000).

Simon, Dennis M., Charles W. Ostrom, Jr., and Robin F. Marra. 1991. “The
President, Referendum Voting, and Subnational Elections in the United
States.” The American Political Science Review 85, 4:1177-1 192.

Simon, Herbert A. 1947. Administrative Behavior. New York: Macmillan.

Skogan, Wesley. 1996. "Crime and Punishment." In Politics in the American States: A

184

Comparative Analysis, Sixth Edition, ed. Virginia Gray and Herbert Jacob.
Washington, DC: CQ Press.

Sprout, Harold and Margaret Sprout. 1972. "National Priorities: Demands, Resources,
Dilemmas." World Politics 24:293-317.

Sribney, William and Vince Wiggins. 2001. “Missing R-squared for ZSLS/IV.”
http://www.stata.com/support/faqs/sta/2sls.html (May 21, 2001 ).

State of Michigan Special Joint Committee on State Prison Construction and Financing.
1985. “Prison Overcrowding in Michigan: Final Report of Joint Committee on
Prison Construction and Financing.”

St. John, Edward P. 1991. “A Framework for Reexamining State Resource-Management
Strategies in Higher Education.” Journal of Higher Education 62, 3:263-287.

Stephan, James J. 1999. “State Prison Expenditures.” Bureau of Justice Statistics, NCJ
172211.

Stone, Deborah A. 1989. "Causal Stories and the Formation of Policy Agendas."
Political Science Quarterly 104:281-300.

Swope, Christopher. 1999. “Truth and Consequences.” Governing Magazine 31, 4:40-
42.

Su, Tsai-Tsu, Mark A. Kamlet, David C. Mowery. 1993. "Modeling U.S. Budgetary and
Fiscal Policy Outcomes: A Disaggregated, Systemwide Perspective." American
Journal of Political Science 37:213-245.

Taggart, William A. 1989. “Federal Judiciary: The Impact of Court-Ordered Prison
Reform On State Expenditures for Corrections.” Law and Society Review 23, 2:
241-271.

Thompson, Joel A. 1987. “Agency Requests, Gubernatorial Support, and Budget
Success in State Legislatures Revisited.” Journal of Politics 49, 3:756-779.

Tonry, Michael. 1987. “Sentencing Reform Impacts.” US. Department of Justice.
National Institute for Justice.

Tonry, Michael. 1999. “Reconsidering Indeterminate and Structured Sentencing.”
Sentencing and Corrections: Issues for the 21'" Century. Papers From the
Executive Sessions on Sentencing and Corrections, No. 2. Washington DC: US.
Department of Justice, Office of Justice Programs, National Institute of Justice.

Tonry, Michael and Kathleen Hatlestad, eds. 1997. Sentencing Reform in Overcrowded
Times: A Comparative Perspective. New York: Oxford University Press.

185

Tucker, Harvey J. 1982. “It’s About Time: The Use of Time in Cross-Sectional State
Policy Research.” American Journal of Political Science 26, 1:176-196.

US. Census Bureau. 1999. “Table A-7. College Enrollment of Students 14 to 34 Years
Old, by Type of College, Attendance Status, Age, and Gender: October 1970 to
1998.” http://www.census.gov/population/socdemo/school/taba-7.txt (November
9,1999)

Virginia Criminal Sentencing Commission. 1999. 1999 Annual Report.

von Hirsch, Andrew. 1976. Doing Justice: the Choice of Punishments. New York: Hill
and Wang.

von Hirsch, Andrew and Kathleen Hanrahan. 1981. “Determinate Penalty Systems in
America: An Overview.” Crime and Delinquency 27, 3: 289-316.

Von Hirsch, Andrew, Martin Wasik, and Judith Greene. 1989. “Punishments in the
Community and the Principles of Desert.” Rutgers Law Journal 20:595-618.

Wanat, John. 1974. "Bases of Budgetary lncrementalism." American Political Science
Review 68:1221-1228.

Weick, Karl E. 1969. The Social Psychology of Organizing. United States: Addison-
Wesley Publishing Company, Inc.

Weick, Karl E. 1995. Sensemaking in Organizations. California: Sage Publications, Inc.

White, Joseph. 1995. “(Almost) Nothing New Under the Sun: Why the Work of
Budgeting Remains Incremental.” In Budgeting, Policy, Politics: An
Appreciation of Aaron Wildavsky, eds. Naomi Caiden and Joseph White. New
Brunswick: Transaction Publishers.

Wicharya, Tamasak. 1995. Simple Theory. Hard Reality: The Impact of Sentencing
Reforms on Courts, Prisons, and Crime. Albany: State University of New York
Press.

Wildavsky, Aaron B. 1964. The Politics of the Budgetary Process. Boston: Little,
Brown.

Wilkins, Leslie T. et a1. 1978. Sentencing Guidelines: Structuring Judicial Discretion.
National Institute of Law Enforcement and Criminal Justice.

Wood, B. Dan, Sean Crotty, and Nick Andrew Theobald. 2001. “Tradeoffs in State
Expenditures: A Theory and Empirical Analysis Across States and Time.” Paper
presented at the Annual Meeting of the Midwest Political Science Association.

186

Wood, Peter B. and Harold G. Gramsmick. 1999. “Toward the Development of
Punishment Equivalencies: Male and Female Inmates Rate the Severity of
Alternative Sanctions Compared to Prison.” Justice Quarterly 16:19-50.

Wooldridge, Jeffrey M. 2000. Introductory Econometrics: A Modern Approach.
Southwestern College Publishing.

Wooldredge, John. 1996. “Research Note: A State-Level Analysis of Sentencing

Policies and Inmate Crowding in State Prisons.” Crime and Delinquency 42,
32456-466.

Young, Neil and Crazy Horse. 1978. “Four Strong Winds.” On Comes A Time [CD].
Reprise.

Zellner, Arnold. 1962. “An Efficient Method of Estimating Seemingly Unrelated
Regression and Tests of Aggregation Bias.” Journal of the American Statistical
Association 57:358-368.

Zellner, Arnold and David S. Huang. 1962. “Further Properties of Efficient Estimators
for Seemingly Unrelated Regression Equations.” International Economic Review

3:300—313.
Ziedenberg, Jason and Vincent Schiraldi. 2000. “The Punishing Decade: Prison and Jail

Estimates at the Millennium.” Justice Policy Institute. http://www.cjcj.org/
punishingdecade/punishinghtml (February 14, 2000).

187