1W! .

l
!

     

l

llW!WHUIWWIUIWIIWWW

   

913 W

   

835
THS

     

THESIS

l

:Cbl

 

LIBRARY
Michigan State
University

 

 

 

This is to certify that the

thesis entitled

IMPACT OF SENTENCING REFORMS
ON PRISON POPULAlTON

presented by

Yan Zhang

has been accepted towards fulfillment
of the requirements for

MS. a . Criminal Justice
egree 1n

 

 

' ”0074580!
Date .LQfll/QQ

0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

 

 

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
05 “m H 74R

MEN

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

moo chlM.p65p.14

 

IMPACT OF SENTENCING REFORMS ON PRISON POPULATION

By

Yan Zhang

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE
School of Criminal Justice

2000

ABSTRACT
IMPACT OF SENTENCING REFORMS ON PRISON POPULATION
By

Yan Zhang

The United States prison population began to increase rapidly in the early 19705.
While research has failed to Show that changes in demographics factor or crime rates
influenced the growth of prison population, other factors such as ‘get tough’ sentencing
reforms are believed to be related to increased prison populations. This thesis tests
hypotheses derived from the formal legal perspectives of sentencing reforms: the
implementation of sentencing reforms should increase the sentencing severity which in
turn should account for the increased prison population. To test these hypotheses, the
rate of new court commitment, average time served in prison, and prison populations
rates in all fifty states and the District of Columbia, from 1973 to 1996 are examined
using hierarchical multivariate linear models (HMLM). The research results Show that
sentencing reforms are not directly related with the change in state prison populations.
Only two sentencing reforms, three-strikes laws and sentencing commission’s
consideration of prison capacity, are indirectly related to prison population changes.
Each of these factors, however, has opposite influences on the two mechanisms
(admissions and time served) of total prison populations. These outcomes do not
consistently support the formal legal rationality argument, but to some extent support the

substantive rationality perspective of sentencing reforms.

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................ IV
LIST OF FIGURES ............................................................................................................ V
INTRODUCTION ............................................................................................................... 1
LITERATURE REVIEW .................................................................................................... 2
US. Prison Population .................................................................................................... 6
Research Explaining Prison Population Growth ............................................................. 8
RESEARCH PURPOSE ................................................................................................... 15
RESEARCH DESIGN ...................................................................................................... 17
Methods ......................................................................................................................... l7
Variables ........................................................................................................................ 19
Modeling ....................................................................................................................... 27
RESULTS ......................................................................................................................... 34
DISCUSSION ................................................................................................................... 37
REFERENCES .................................................................................................................. 45

iii

LIST OF TABLES

Table 1. Measures and Variables ...................................................................................... 20
Table 2. Definition of sentencing policies ....................................................................... 22
Table 3. Sentencing practices. .......................................................................................... 23

Table 4. ANOVA for New Court Commitment, Time Served in Prison, and Prison

Populations ................................................................................................................ 29
Table 5: Sentencing Policies Effects on New Court Commitments .................................. 41
Table 6: Sentencing Policies Effects on Time Served in Prisons ..................................... 42
Table 7: Sentencing Policies Effects on Prison Population .............................................. 43
Table 8: Admission and Length of Stay Effects on Prison Population ............................. 44

iv

LIST OF FIGURES

Figure 1. US Prison Population Rates ................................................................................ 7
Figure 2. US state prisons new court commitments rates ................................................ 28
Figure 3. Estimated time served in state prisons .............................................................. 32
Figure 4. US states prison population rate ....................................................................... 33

INTRODUCTION

The United States criminal justice system has undergone numerous reforms over
the past several decades to minimize police officers’, judges’, and correction officials’
discretion while making decisions about individual cases (Walker 1993; Gottfredson and
Gottfredson 1988). These reforms have occurred at nearly every decision point in the
system, including recommending or mandating arrest at the scene of every domestic
violence incident (Sherman and Cohen 1989) and partial release decision that are based
upon empirical measures of risk rather than a judges intuition (Maxwell 1999). While
these and other similar reforms represent substantial policy changes, the one set of
reforms that has likely had the widest effect in terms of the number of cases applied to as
well as in the nature of the outcome is in the area of sentencing reform. In general, these
sentencing reforms were enacted to reduce sentencing disparity or increase sentencing
fairness, to control prison populations, and to increase the deterrent value of criminal
sanctions (BJA 1998).

While many of the penal changes have come under serious investigation as to
their impact on system discretion or other planned changes (Miethe 1987; Savelsberg
1992; Cohen and Canela-Cacho 1994; Levitt 1996; Marvell and Moody 1996; Engen and
Steen 2000), most studies are methodologically limited in several ways that prevent a
national assessment of how these changes have affected penal policies and outcomes
specifically. To address some questions left by the earlier research this paper will assess
whether the changes in sentence policies have changed penal outcomes across the entire

nation. More specifically, this thesis tests whether the various sentencing reforms have

influenced the aggregate of decisions to imprison and for how long. To test these
hypotheses, the rate of new court commitment, average time served in prison, and prison
populations rates in all fifty states and the District of Columbia, from 1973 to 1996, are

examined using hierarchical multivariate linear models.

LITERATURE REVIEW

For nearly 100 years (1870 to 1970), the dominant sentencing model in the United
States was generally one based upon indeterminate and back-end sentencing decisions
that featured broad judicial and parole discretion, and case-by-case decision-making.
Under the indeterminate sentencing model, most sentences were individualized so that
the punishment fit the offender’s needs rather than his or her crime. This model was
chosen over other possible models because, in part, the main goal of imprisonment was to
rehabilitate the offender, which seemed to require that the exact sentence length be
determined long after adjudication and reflect the offender’s dynamic rehabilitation
needs. However, by the early 19708 there were a growing number of negative
accusations about this sentencing model and its attached rehabilitation goals. For
instance, the American Friends Service Committee published a report stating that
indeterminate sentences were used to repress inmates (AFSC 1971). In addition, the idea
that the prison system could rehabilitate offenders was also seriously questioned after
Martinson ( 1974) wrote that “with few and isolated experience, the rehabilitative efforts

that have been reported so far have had no appreciable effect on recidivisms”. This

report was hailed subsequently and strongly by both politicians and academicians as the
final word on the lack of rehabilitation efficiencies of the US. prison system.

With the increased cynicism directed toward the rehabilitative model, increased
pessimism concerning prison officials’ and judicial discretion, and the perception of
rising violent crime rates, the stage was set for sentencing reform in the spirit of more
rational and predictable punishments. Andrew von Hirsch, for instance, argued that
punishment should be based primarily upon the seriousness of the crime and prior
criminal record, not upon the perceived amenability of the offender rehabilitation (von
Hirsh 1976). This argument, unlike many other suggested changes, appealed to both
conservatives and liberals who had voiced reservations over the current state of
imprisonment and punishment in the United State. This argument was accepted by many
in part because von Hirsh did not exclude the idea of trying to rehabilitate people while
they were sanctioned. What von Hirsh and other similar scholars did reject was the idea
that the type and length of a sanction should be determined by rehabilitation needs, or for
that matter, by any other utilitarian rationales such as deterrence.

By the end of the 19705, many state legislatures and the US. Congress began
passing various versions of von Hirsh’s model, and in theory, judges and other
practitioners began implementing and experimenting with these diverse, and far reaching
sentencing reforms. By the end of 1990S, the twenty years of penal code and policies
changes resulted in no single American sentencing system. Some states replaced
indeterminate sentencing with structured sentencing schemes, and others retained
indeterminate sentencing. About 20 percent of the states moved to presumptive

sentencing guideline systems, another 18 percent passed voluntary guidelines, and one

state enacted mandatory guidelines. All of the states were also affected in diverse ways
by three or two strikes rules, mandatory minimum sanctions, or truth-in-sentencing
requirements. Most states retained some form of good time but some states did abolished
parole altogether (Tonry 1999). Even as states claimed to have indeterminate sentencing
structures, these laws have been integrating determinate characters by use of mandatory
minimums, truth in sentencing provisions, and reduction in the amount of good time
credits (BJ A 1998). While these changes are numerous and diverse in how and where
they are implemented, their goals are quite common: increasing sentencing fairness,
reducing unwarranted disparities in the decision to imprison and in sentence length, and
institutionalizing principles of just deserts and deterrence (BJA 1998; Engen and Steen
2000).

A general assessment of von Hirsh’s and other similar sentencing models is one
that has attempted to “reverse the substantive rationalization of criminal law and to return
to law based on a neoclassical formal rationality” (Savelsberg 1992; Engen and Steen
2000). As Weber (Weber 1968) described, formally rational law is law “guided by norms
intrinsic to the legal system, bound by objectively verifiable and legally relevant facts,
and comprising a gapless set of rules from which decisions in concrete cases can be
deduced” (Savelsberg 1992). Substantively rational law, in contrast, “takes into account
economic, sociological, and ethical criteria and represents a desire for substantive
expedience, a concern with impact, effect, and ends served by legal action, and
adaptation to the concrete case” (Savelsberg 1992). Under the neoclassical formal
rationality model, the expected goals of sentencing reform require sentencing decisions to

be based solely on legally relevant facts, which are characteristics of offenses, and

disregard the substantive characteristics of the offender. In addition, while many of the
models do not explicitly call for more severe sentences, the belief among many was that a
return to the formal rationality would increase the severity of sentencing decisions (i.e.,
people would no longer be released early due to good behavior while in prison or upon
successful completion of a treatment program). In turn, many have argued that a primary
outcome of these sentencing policies would be a substantial change or growth of the US.
prison population following implementation of the reform. However, the formal legal
rationality hypothesis of sentencing reform has been questioned by the alternative
substantive rationale supposition. The substantive hypothesis argues that the criminal
justice system in the United States is firmly entrenched in substantive concerns
(Savelsberg 1992). Sentencing reform such as sentencing guidelines have to be
implemented through networks of decision makers in complex administrative
environments and by actors within those organizations (Savelsberg 1992). Several
dilemmas of substantivation policies such as due process and sentencing discretions
continue to be stressed by the criminal justice profession and public and thus
circumventing many of the sentencing reforms. In arriving at appropriate dispositions,
concerns related to social inequality, social justice, perceived causes of criminal behavior,
and the anticipated consequences of sentencing will still be taken into account
(Savelsberg 1992). Furthermore, some have argued that a hydraulic displacement of
discretion exists in the sentencing process (Alschuler 1978; Tonry and Zimring 1983).
According to the hydraulic displacement hypothesis, gains in sentencing neutrality and
uniformity would be eroded by greater disparities in the use of prosecutorial discretion in

charging and plea bargaining practices, and the ability to achieve the reform goals would

be impeded (Miethe 1987). Thus, sentencing outcomes will remain substantively rational
and not change according to sentencing reform. However, by 2000, it is not clear
whether either of these two hypotheses is supported by research, particularly in regard to

understanding prison populations.

U.S. Prison Population

The Federal Government began keeping annual records about the United States
prison population in 1926 (Langan 1991). During the period of 1926 to 1970, the prison
population rate was reasonably constant. Research on the imprisonment rate for the
period 1930 to 1970 found that the average prison population was about 110 prisoners per
100,000 US. residents, with only a standard deviation of nine (Blumstein and Cohen
1973). Because of this rather stable trend, Blumstein and Cohen (1973) theorized that
while there have been substantial changes in society, the imprisonment system adopted to
these changes to keep a fairly stable number of citizens incarcerated. However, this
theory seems to have had a relatively short life span. Since 1973, the US. aggregate
prison population has experienced almost constant growth. The average annual increase
of the prison population from 1973 to 1979 was about 7%. During the 1980s, the average
annual rate grew to almost 12% per year. After the 19805, the rate of the annual increase
began to decrease, but still increased at an averaging of about 8% annually (figure 1)’.
By midyear 1998, 1,277,866 prisoners were under Federal and State jurisdiction, more
than four times the number incarcerated in 19805 (315,974) and six times the number
incarcerated in 19735 (204,211) (Gilliard 1999). Certainly, some of this increase is due

to general population growth over this period (Cohen and Canela-Cacho 1994),

 

’ Criminal Justice Source Book, 1998.

nevertheless, the annual inmate rate adjusted for total population rose from the rate of 96
per 100,000 residents in 1973 to 452 per 100,000 residents as of June 1998, an annual
average increase of over 22 persons per 100,000 residents.

There is little doubt that the increased incarceration rate had some impact on
crime reduction and had some deterrent effect (Blumstein and Beck 1999). However, the
rapid growth of the prison population also led to the problem of prison overcrowding. As
of 1997, the average state prison system was operating at 115% of capacity and the
Federal system was running at 119% of capacity (Gilliard and Beck 1998). Three states
(California, Pennsylvania, and Virginia) were operating in excess of 150% of capacity
(Gilliard and Beck 1998). The effects deriving from overcrowding are many: a rise in
correctional costs, delays before imprisonment, inequality in the treatment of inmates, a
greater likelihood of violence and victimization, and especially the deterioration in living
and working conditions for inmates and prison officers.

Figure 1. US Prison Population Rates (per 100,000 resident populations)

500

 

450 -
400 d
350 .
300 4
250 d
200 t
150 '4
100 -

50 -‘

 

 

 

O I I I I I I I I I I I I

‘3\’\§9@5b\6\1\")¢\9%‘3\’\
dd$ee$eeeee§§

The growth of the prison population has attracted a considerable amount of
research into why, after 50 years of stability, has the population grew more than 10 times
in size in just 25 years. Besides the sentencing reform, factors proposed by
criminologists include changes in crime and arrests rates (Langan 1991; Cohen and
Canela-Cacho 1994; Greenfeld and Beck 1996); demographic effects (Langan 1991;
Mackenzie and Tracy 1988); the war on drugs (Langan 1991); and prosecutorial policies
(BJ S 1996). While empirical evidence has failed to show that changes in population
demographics and in crime rates since the 19705 has a direct and significant impact on
prison population, a common strain of thinking emerged that stated that sentencing
reforms are the principal cause of the vastly increased numbers of people in prison

(Fumiss 1996; Sabol and Lynch 1997; Wicharaya 1995).

Research Explaining Prison Population Growth

Two basic forces drive prison population growth: the admissions to prison and the
length of time served for those offenders committed to prison. Thus, the increase of
prison population could be the result of an increase in certainty, while the length of time
served remains stable or decrease, an increase in the length of time served while the
certainty of committed to prison remains stable or decrease, or a simultaneous increase in
certainty and the length of time served. As argued later these three situations have very
different insinuations when it comes to assessing sentencing policies (Kensey and
Toumier 1999).

The proposed sentencing practices that potentially effect the length of prison

terms include: mandatory minimum sentences, determinate sentencing laws, and truth in

sentencing laws. The mandatory minimum sentences are generally directed at certain
violent offenses such as crime involving weapons, crimes committed by repeat offenders
or habitual criminals, and certain drug crimes. The minimum sentences require inmates to
serve a fixed amount of prison time before becoming eligible for discretionary release or
for release without the approval of a parole board (depending on within an indeterminate
sentencing structure or a determinate sentence). It was believed that the mandatory
minimum laws would result in more offenders going to prison for longer periods and
deterring other potential offenders from committing crime. By the year 1996, 40 states
applied mandatory sentencing for repeat or habitual offenders; 38 states and the District
of Columbia sanctioned crimes committed using a deadly weapon; 36 states and the
District of Columbia used it for drug possession/trafficking; and 31 states for drunk
driving. As a means of mandatory sentencing enhancement, three-strikes laws impose
longer prison terms than earlier mandatory minimum sentencing laws. A study about
Califomia’s three-strikes law indicated that through long-term incarceration, the three-
strikes law would triple Califomia’s prison population over the next 25 years (Parent and
Dunworth 1996).

Another sentencing change that involves removing correctional discretion is the
implementation of determinate sentencing policies. Under determinate sentencing
structures, sentence length is determined mainly by the judge, with no discretionary
release by a parole board (Reitz and Reitz 1993). Since determinate sentencing laws
remove the power of parole boards to reduce large numbers of sentences at the end of the
offenders’ prison terms (Zimring and Hawkins 1991), the determinate sentencing laws

(DSL) affect prisons because of sentence length. For example, under Illinois determinate

sentencing law, an inmate must serve 50% of his or her sentence prior to being eligible
for release (Joyce 1992). The Illinois DSL also increased the length of time served for all
serious crimes. An analysis of Illinois UCR index crime offenses population and
admission data from 1984 to 1991 showed that the State experienced a large increase in
the average daily prison population because the long-term offenders were crowding in
prison (Joyce 1992).

A recent version of the determinate sentence that has won wide approval from
politicians and the public at large are the Truth-in-Sentencing requirements. Like the
earliest determinate sentencing laws, these new versions similarly reduce discrepancy
between the sentences imposed and actual time served in prison and are considered a
major reason for more recent inmate growth problems. The first truth-in-sentencing law
was enacted by the State of Washington in 1984. By 1998, 27 states and the District of
Columbia met the Federal Truth-in-Sentencing Incentive Grants program criteria, which
require persons convicted of a Part 12 violent crime to serve not less than 85% of the
prison sentence (Ditton and Wilson 1999). As a result of truth-in-sentencing practices,
some believe state prison populations should continue to increase well into the future
because new offenders will continue to arrive, while older offenders who would normally
be released because of good behavior must now serve sentences that may exceed 50%
more of the current expected length.

These simple predictions are supported by several studies that modeled state
prisoners population data from 1990 to 1997. Ditton and Wilson (1999) found that while
the number of inmates held in state prisons increased 60% since 1990, the number

admitted to prison increased only 17%; the number of admissions relative to the number

 

2 UCR part 1 violent crimes, including: murder, rape, robberies, and aggravated assault.

10

of inmates in prison dropped from 73 per 100 State prisoners in 1990 to 52 per 100 in
1997. These longer sentencing explanations are also backed by other research.
Wooldrege (1996) studied the relationships between state level sentencing policies and
inmate crowding, and found that sentencing practices with longer minimum sentences,
mandatory prison terms, and parole guidelines were significantly related with a higher
prison population of long-term inmates. Marvell’s (1995) time series analysis of nine
states that maintain presumptive sentencing guidelines also suggested that the sentencing
guidelines have little impact on the number of defendants sentenced to prison, changes in
prison population growth are due mainly to changes in prison term length.

However, the lengthening prison stay explanation is questioned by prison
population trend studies. These trend studies indicate that growth of the state prison
populations are not the result of longer sentences and there has not been any significant
change in the time served by offenders (Langan, 1991; Greenfeld and Beck 1996).
Compared with the pre-1973 period, persons admitted since 1973 were less likely to have
a life sentence, had a shorter median sentence length, and were less likely to have
sentences 10 to 19 years in length (Langan 1991). Between 1985 an 1992, the average
maximum sentence of prisoners actually declined from 78 months to 67 months (Perkins
1994). The median sentence length of prisoners admitted from court remained constant
at 48 months. Moreover, despite the increasing use of mandatory minimums and
sentencing enhancements during the period, the proportion of inmates who received a
maximum sentence of 10 years or longer actually declined from 19.7 percent in 1985 to
17.7 percent in 1992 (Greenfeld and Beck 1996). Instead, these trend studies suggest that

rise in prison population since the middle 19705 has been driven by increases in prison

11

admissions. A national survey on the impact of sentencing guidelines pointed out that
dispositional policies (i.e., rules specifying which offenders should or should not be
imprisoned) generally have had a greater impact on prison population levels than policies
about the amount of time to be served. The reason is that sentencing commissions have
generally chosen to change the categories of offenders who are to be imprisoned instead
of changing the current average duration of imprisonment (Parent and Dunworth 1997).

While sentencing reforms are criticized as the key reason fOr prison crowding,
one of the goals of these reforms in some states has also been to moderate the rapid
growing trend. In fact, several guideline commissions were required to consider the
impact of changing guidelines on the need for additional correctional resources (BJA
1998). For instance, Minnesota’s guidelines, which took effect on May 1, 1980, require
that sentencing practices stay within the state’s correctional resources, especially
available prison capacity. As a result, Minnesota’s state prison population increased very
slowly in the early 19805 and stayed well below their capacity. Between 1980 and 1984,
the prison population increased only by 8%, while the total national prisoners increased
41%. However, by the later 19805, the slow rate of increase began to change.
Minnesota’s prison population rose by 29% from 1984 to 1988 and then increased by
another 37% from 1988 to 1992. Nevertheless, by the end of 1992 Minnesota still had
the second-lowest per-resident imprisonment rate in the country (Tonry 1997).

Marvell’s (1995) work also supports the notion that sentencing guidelines can
control increases in prison populations. He found that presumptive sentencing guidelines
were associated with declines in prison population growth in six states (Delaware,

Florida, Minnesota, Oregon, Tennessee, and Washington) where legislators decreed that

12

 

guideline framers consider prison capacity when establishing recommended prison
sentence lengths, at least in relation to nation-wide trends. Marvell estimated that prison
populations were approximately fifteen to thirty-five percent lower than would be
expected in the six states. Other evaluations conducted by individual state sentencing
guideline commissions have similarly revealed positive evidence concerning the
effectiveness of presumptive sentencing guidelines on controlling increases of prison
populations (Tonry 1997). The State of Washington’s enacted presumptive guidelines
linked the correctional resources in 1984 and found that between 1984 and 1988, their
incarceration rate declined from 156 per 100,000 populations to 124 per 100,000
populations (Tonry 1997). Similarly, Oregon as well as Louisiana and Tennessee have
passed guidelines that include consideration of capacity. Oregon’s found that after the
law was enforced incarceration rates decreased from 235 per 100,000 residents to 166 per
100,000 residents (Tonry, 1997).

Besides changes in sentencing policies, there are other factors that scholars have
argued may account for some of the prison population growth. The first of these
explanations is the one centered on changes in population demographic characteristics,
which is often referred to as the “demographic change hypothesis” (MacKenzie and
Tracy 1988). According to this hypothesis, populations can be classified into subgroups,
some of which are more apt to be incarcerated than other groups. As Langan (1991)
figured out “relative to their representation in the general population, males (48% of the
population and 95% of all prisoners), blacks (11% of the population and 48% of
prisoners), and persons in their twenties (24% of the population but 50% of prisoners) are

highly over-represented in prisons.” Such subgroups form the “prison-prone” population.

13

A change in the population demographics can be expected to result in a change in the
incarceration rate. Around 1986 the numbers of persons ages 20 to 29, both black and
white, climbed to their highest levels ever (42 million altogether) and the largest birth
cohort from the baby boom, those born in 1961, reached a peak age of imprisonment, age
25. The demographic shifts that occurred between 1974 and 1986 were significant, and
account for about 20% of prison population admission growth (Langan, 1991). This
demographic change hypothesis was questioned by other researchers as well.
Observations of time series data of males in their twenties from 1970 to 1990 did not
show the variance in population concentration a major influence on the increase of
prisoners (Zimring and Hawkins 1991).

A second alternative explanation for the changes in the prison population is that
crime and arrest may be two important factors impacting the imprisonment process. This
model proposes that increases in crime and arrest, especially for violent crimes that are
more likely to result in sentences to prison following conviction, may cause the prison
population to rise (Cohen and Canela-Cacho 1994). Researchers studying UCR index
crimes from 1980 to 1996, however, have found no evidence that adult offending rates
for murder, robbery, burglary and sex assault except aggravated assault showed a clear
upward trend (Blumstein and Beck 1999). Furthermore, estimates from the National
Crime Victimization Surveys show that crime declined slightly between 1973 and 1991,
from 35.7 to 34.7 million (Marvell and Moody 1994. The UCR arrest rates for all the
crime types mentioned above are also stable between 1980 and 1996 (Blumstein and
Beck 1999). Indeed, one researcher estimated that changes in crime and arrest rates

explained only 9% of admission growth (Langan 1991). Thus, there is little evidence

14

that at the national level aggregate demographic changes or changes in crime patterns
affected the growth of national prison population rate. However, prior research has not
yet systematically asked the question of whether demographic changes or changing crime

patterns at the state level may jointly affect state level prison population changes.

RESEARCH PURPOSE

Many policymakers and advocates, as well as some criminologists, consider the
increases in the prison population a direct result of sentencing reforms, such as
mandatory penalties, determinant sentencing, and guideline-based sentences (see, for
example, Sabol and Lynch 1997; Fumiss 1996; Langan, 1991). With this presupposition
in mind, prior research about the impact of sentencing reform on prison population has
typically been based on time trend statistics, before and after the reform became
effective. The usual interpretation is that the observed post-reform changes are credited
to the sentencing reforms enacted contemporaneously (Wicharaya 1995). Actually, three
types of relationships could exist between sentencing policy reforms and the change of
prison populations, which are causal, contributory, and coincidental (Zimring and
Hawkins, 1991). This overly simple deduction may result in a false relationship or at
least an exaggerated correlation. Some other research have either examined bivariate
relationships between specific sentencing practices and inmate growth with short post-
law periods or just checked a specific time (Langan, 1991; Wooldredge, 1996). These

research designs suffered from limitation of short post-reform observation periods, which

15

does not allow for exploration of evolutionary or delayed types of reform effects, and
limited control variables.

Recently, some multiple time series designs have been used to estimate the impact
of presumptive sentencing guideline, and determinate sentencing laws on imprisonment
rates (Marvell and Moody 1996). These research designs provide a large sample size and
allow one to consider a greater breadth of control variables. The assessment of individual
state laws, however, is still based upon a comparison of one state to all other states
combined. This aggregation ignores the heterogeneous nature of American society and
the fact that these other states are similarly experimenting with penal changes as well.

Besides the methodological limitations mentioned above, many researchers also
ignore the theoretical sides of sentencing reforms. The purpose of this study is to test the
~ hypotheses derived from the formal legal perspectives of sentencing reforms: sentencing
policies reform should impact the aggregate sentencing outcomes. Specifically, this
thesis examines. three hypotheses: (1) changes in sentencing policies impact the annual
change in the rate of new court commitments; (2) changes in sentencing policies impact
annual change in the average time served among incarcerated prisoners; and, (3) changes
in sentencing policies impact the total annual change in prison population rates. At the
same time, this thesis also examines several alternative hypotheses that are consistent
with the substantive factors theories, such as political and social diversity will circumvent
the influence of sentencing reforms on sentencing outcomes. In sum, by jointly testing
these theories, this thesis seeks to improve our current understanding of how changes in

penal policies can impact penal practices within the context of a changing social context.

l6

RESEARCH DESIGN

Methods

This analysis uses a quasi-experimental design to examine changes of prison
population status and sentencing practices from 1973 to 1996 for 50 states and the
District of Columbia. Nineteen seventy three is the first year that prison population in
United States began to increase following a decade of declining prison populations. The
sentencing reforms also began in the middle of 19705. So, the starting point of observing
was selected as 1973. Time series data of prison populations, new court commitments,
and total prison admissions from 1973 to 1996 for each state and the District of
Columbia, which were collected by BJ S, are used to reflect the sentencing outcomes
companying with sentencing reforms. Changes in crime rates, demographic components,
and economic situations at the same periods for each state and the District of Columbia
are also tested for the purpose of increasing the fitness of research models.
Hierarchical linear modeling is employed in this analysis. Hierarchical linear models
(HLM) can be used for data that have a multi-level, nested structure (Bryk and
Raudenbush 1992). With hierarchical linear models, each level in this structure is
formally represented by its own sub-model. These sub-models express relationships
among variables within a given level, and specify how variables at one level influence
relations occurring at another (Bryk and Raudenbush 1992). One application of
hierarchical models involves repeated observations nested within individuals. Since the
time series data of imprisonment status including prison population, year’s new court

commitments, and year’s admission of inmates from 1973 to 1996 are available on a

17

series repeated observing time points (the end of every year), it makes it possible to
utilize hierarchical modeling in this analysis. Under HLM, the multiple observations of
each state can be considered as nested within the individual state rather than as the same
fixed set for all states. The treatment of multiple observations as nested allows
proceeding even when the number and spacing of time points vary across states. This
characteristic becomes more important when there is missing data, which is a barrier to
use time series design and some other multivariate repeated measures (Bryk and
Raudenbush 1992). Another advantage that is much stronger than those used in the
earlier research is, by using hierarchical design one can include specific characteristics of
each state, and examine not only the effects between states but also within states, which
is not able to be obtained in a simple multiple regression model or time-series design.
Furthermore, in this study, as the number of observing points are the same for each state,
which is 24 (1973-1996), and the spacing between time points are equal for every state, it
is actually a fixed data set, and the analysis can be reconceived as a multivariate repeated
measures analysis. Consequently, instead of using HLM2, hierarchical multivariate
linear model (HMLM) is used to accomplish this multivariate analysis. HMLM is
flexible in allowing a wide variety of assumptions about the variation and co-variation of
the repeated measures, and allows some missing data points. Within the framework of
HMLM, it is possible to estimate models with first-order auto-regressive errors, which is
likely to happen when there exists 24 repeated observing points.

As discussed earlier in this paper, imprisonment sanctions can be distinguished in
terms of the certainty of going to prison following arrest and the length of time served for

those offenders who are committed to prison. The observation level (level 1) outcome

18

variables are identified as every year’s new court commitments to prison, which is used
to reflect the certainty of going to prison, estimated average time served by each
committed offender, and the state prison population rates.

Each of these dependent variables is assumed a function of a systematic growth
trajectory. Besides the time predictor, there are six time-varying independent variables
used to describe the sentencing practice status of each state, and three other control
variables reflecting the crime, economic, and demographic situation distinctly for each
state. Another 24 indicator variables are also included. The observation level individual
state growth parameters become the dependent variables in the state level (level 2)
models. The level-2 predictors are identified as regional variables (Table 1), which are

used to present diversities of the political and social difference for each state.

Variables

Prison population rates are counted as the number of adult inmates3, who have
been sentenced to prison terms of more than one year under the jurisdiction of state
correctional authorities, divided by state population and multiplied by 100,000. The new
court commitments" variable is the number of prisoners with sentences of one year or
more who have entered prison in a given year, divided by state population and multiplied
by 100,000. The third level 1 dependent variable, the average time served per

commitment is not available directly, it is calculated as the ratio of the total number of

 

3 BJS. Prisoners under State or Federal jurisdiction, Federal and State-by-State, 1977-96. [online] '
Available http://www.oir).usdoi.gov/bis/prisons.htm.
‘BJS. New court commitments admitted to State or Federal jurisdiction, 1977-96. [online] Available

http://www.ojp.usdoj.gov/bjs/prisonshtm.

19

Table 1. Measures and Variables

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variables Measures Source
LN [(prison population/total
Prison population population)*100,000] BJS
LN[(new court
New court commitments/total
commitments population)* 100,000] BJ S
Level-1 (prison population/admission of
dependent prisoners at that year)* 12
variables Time served in prison (months) B] S
Voluntary sentencing
guidelines
Presumptive
sentencing guidelines
Parole abolishment
Truth in sentencing
law
Three strikes law 0 — before the enactive of this
Considering prison policy
capacity 1 - after enactive of this policy B] A
LN [(UCR index crime report
Crime rate total/U CR population)* 100000] UCR
(Unemployed population/civil Department
Level-1 Unemployment rate labor forces)* 100 of Commerce
independent (male 20-29/total population) US Census
variables Male at age 20 to 29 * 100 Bureau
East region
,LCVCI'Z Middle West region Dummy variables .
independent 0 — state not in this region
variables West region 1 — state in this region

 

20

 

inmates at end of a year and the year’s total admissionss (Cohen and Canela-cacho 1994),
and this ratio multiplied by 12 as the number of months.

The time variable in levell model is the number of years from the first analyzed
year. At the first year, the time variable is set as zero. Since the visual examination of
the dependent variables’ curves indicated some nonlinear patterns, the quadratic and
cubic time variables are also included in the case (Figure 3, Figure 4).

The six independent sentencing practice variables are identified as (1) voluntary
sentencing guidelines, (2) presumptive sentencing guidelines, (3) two or three strikes
laws, (4) abolishment of discretionary release by parole board, (5) legislative direction
regarding consideration of prison capacity, and (6) truth in sentencing laws. Because the
criteria used to identify determinate sentencing practices are “explicit standards specify
the amount of punishment and a set release date with no review by an administrative
agency (parole board)” (BJ S, 1998), the “abolishment of discretionary release by parole
boar ” is used in this analysis instead of determinate sentencing policy. Another reason
for not using the determinate sentencing variable is the lack of information about when it
was enacted in each determinate sentencing state is available. Definitions of these six
sentencing practices are listed in Table 2. According to the 1996 national survey of state
sentencing structure (BJ S, 1998), there are seven voluntary sentencing guideline states,
and ten presumptive sentencing guideline states. Twenty-four states have two or three
strike laws. Twelve states have abolished discretionary parole release. Fourteen State
legislatures direct their sentencing commissions to consider prison capacity as they

develop guidelines. According to a 1998 General Accounting Office report about the

 

5 BJS. Sentenced prisoners admitted to State or Federal jurisdiction, 1977-1996. [online] Available
http://www.ojp.usdoj.gov/bjs/prisonshtm.

21

Table 2. Definition of sentencing policies

 

Voluntary sentencing guidelines

Recommended sentencing policies that are not
required by law. Usually based on past sentencing
practices, they serve as a guide to judges. The
legislature has not mandated their use. Voluntary
guidelines may employ determinate or indeterminate
sentencing structures.

 

Presumptive sentencing
guidelines

Sentencing that meets the following conditions: (1) the
appropriate sentence for an offender in a specific case
is presumed to fall within a range of sentences
authorized by sentencing guidelines that are adopted
by a legislatively created sentencing body, usually a
sentencing commission; (2) sentencing judges are
expected to sentence within the range or provide
written justification for departure; (3) the guidelines
provide for some review, usually appellate, of the
departure. Presumptive guidelines may employ
determinate or indeterminate sentencing structures.

 

Parole abolishment

States abolished the parole board that adjusts actual
time served in prison.

 

Truth in sentencing law

Truth-in-sentencing laws requiring violent offenders to
serve at least 85 percent of their imposed sentences.

 

Three strikes law

Some states enacted the two or three strikes laws,
which are intended to increase the prison sentences of
persons convicted of specific crimes that have been
previously convicted o a violent or serious felony
offense.

 

 

Considering prison capacity

 

State legislatures direct their sentencing commissions
to consider prison capacity as they develop guidelines.

 

Note: these definitions are obtained from “BJ A, 1996. National Assessment of
Structured Sentencing,” and “BJA, 1998. 1996 National Survey of State Sentencing

Structures.”

22

 

Availability of Federal Grants Influenced Truth in Sentencing Laws in Some States
(GAO, 1998), there is twenty-seven states who have enacted truth in sentencing law
conforming to the federal TIS eligibility requirements. The years that these sentencing
policies enacted are not consistent through different sources. They are identified based
on the 1996 and 1994 BJ S national sentencing structure survey, Tonry’s (1997)
Sentencing reform in overcrowded times, a recent repOrt about the truth in sentencing in
state prisons (Ditton and Wilson 1999), and the GAO report (1998). Each type of
sentencing policy is represented by a dummy variable: the variable is equal to zero before
the policy went into effect; it is equal to one after the policy went into effect in a state
(Table 3).

Besides the sentencing variables, other levell control variables include several
factors that are considered as associated with sentencing outcomes. The first one is crime
rates. The crime rate data used in this study is the UCR index crime reports total6 per
100,000 UCR populations7. The reason in choosing this crime report data is simply the
difficulty in obtaining the crime arrest total data by state, by year, through the whole
study periods. An important limitation in using the UCR index crime report data is the
lack of drug offenses, which are considered one major driving force of the rapid growth
of prison populations. Despite its limitations, UCR index crime report data are useful as
indicators of crime data, especially the violent crime data over time. The second control

variable is demographic changes. The selected demographic variable is the percent of

 

6 Data are available in UCR crime reports datasets 1977-1996. Data from 1973 to 1976 data are obtained
from Moody’s crime rate datasets.

7 The population from UCR reports is not the total population of state, but the population of the relative
jurisdiction that reports the crime data to FBI.

8 The UCR crime arrests data are only available after 1980.

23

 

o o o
o mmo_ o
whm_ Nwm_ o
o c o
vmmfi o o
o o vom~
c who_ 0
hwo_ o vmmfi
o o o
mma~ o woo"
o o o
o o vmofi
o whafi o
o o o
o o o
o o mmm_
mamw mm¢~ mam~
o O o
hwm_ oma~ o
o o voa_
c c vam—
o o vom~
voo_ o mam~
o vam— o
o o o
o o o

o
mmo_
waa~
o
ommfi
moo~
o
vao_
o
vom~
o
o

o

OOOOOOOv-‘Ov-‘OOOOOO—‘OOOOOOF'OO

COO—‘OOOOOOOOOOOOOO—‘OF‘OOOOF‘

oo~0~oo~o~ooc~ooooo~ooo~~o

383:2
mammmmmmzz
980::va

sags:
820882
3352
02.82
«5683
5.03ch
macaw
«32

e565
mesa

BEE
wasps:
S30v0
auto—m

o 6:35
033209
50:02:80
8.8200
«8.8.5.6
unmade—2
«:35
8.32
«:8an

 

.0688 .8th
we couacoEmcoU 88—2 208m

520% 32 8wa @6588

So» @6558 @6588

36533.

36558

32,—. 5 52B 250an 338305 93:23» Secateuoufi 82:50qu 08m: 23m

 

A: moosofim wEocoEom .m 053.

24

 

2 N_ cum

0 $2
9 o

53 $2 33

O\
OO
O\
v—t
Ch
00
O\
'—

\O

OO

O\

.—
OOOOOOOO

V3
0‘
a
'—

vam~
nmm_
mam_
o
mom—
o
mom~
o
mmo_
o
o

O

0mg 00% o

o mag

33 $3 .39

o
vam—
moi
c
mag
o
32

OOOOOOO

VN
o

o
swam
mag

33

moa—
coa—

5o—
39

82
mag
.33
33

OOOOOO

m—

39

$3
53

83

$3

awa—
awo—

voo—

OOOOOOO

OOOOOOO—‘O—‘O—‘F‘OOOv-‘OOOO—‘OOOE

OOOOOOOOOO—‘OOOOOOO—‘OF‘OO—‘OOO

v—Iu—av—nv—nv—iOv—oOv—tOv—rOv—«w—tv—n—uv—Av—nv—nv—IOOv—nv—tv—th
m

g

OOOOO—‘O—‘Ov—Ov—OOOOOOOO—‘F‘OOOV

38H
maize»?
5308;?
95 so?
acacia?
2.:qu
EOE?»
533

£38k
33253.
8:5 520m
030 Snow
«2 80nd
53.35.05
:omBO

eeefizo
e20

35 532
86 582
Stow 302

8:82 302

>342. 302
:3: 302
«352
8.8502
«5:82

 

@638 .8th
.8 53802200 0322 28mm

520% 32 wee—Em wcmocoEOm

So» @8558 wEocoEow

mEoEcEOm

wEocoEOm

cog E 58% 053230 "5:92.395 ESE—:5 822:“:qu 0358335 25: 88m

 

$0:an mcmoccfiom €253.58 .m 2an

25

male youth aged twenty to twenty-nine yearsg, who are deemed the prison-prone
subgroup populations. One economic variable is the unemployment rate, which is the
unemployed civil labor force10 divided by the total civil labor force, then multiplied by
one hundred.

Since the prison populations measure and new court commitment measure are
both taken at the end of the year, and usually on average, there exists a half-year distance
between the commission of a crime and sentencing (Marvell and Moody 1996), so the
sentencing practices variables are lagged for one year. The mean of estimated average
time served variable is about twenty-two months, which implies a two-year lag of
sentencing practices variables. All levell continuous variables will be expressed as
natural logs of per capita data to moderate the impact of outliers.

Indicator variables are created to facilitate use of HMLM. Each indicator variable
is the indicator for the measure at each observing time point. For example, indicator one
is indicator at the first time point. If a state’s outcome data is available at this point, it is
coded one, otherwise zero. In the year’s new court commitments model and the prisoner
rate model, there are 23 indicator variables (because of a lag of one year, the examined
series is from 1974 to 1996). In the time served model, there are 22 indicator variables
(two years lag).

At the second level, predictors, which represent state characteristics, should be

stable across the study period. The measure is identified as geographical regional

 

9 US. Census Bureau, Population Division, Population Distribution Branch.
http://wwwcensus.gov/population/www/estimates/statepop.html. The 1980’s population data only include
by sex by age combination, the by race by sex by age group is estimated through using the total race
percentage multiply the relative sex age group population.

0 Unemployment data and civil labor force data was obtained from the Department of Commerce from
1975 to 1986, and from 1988 to 1996. Parts data from 1973 to 1974 are gotten from Moody’s datasets.

26

variables. States may be clustered into regional groups, which reflect the social and
cultural diversity across different regions and simulation within the same region. A
former (Zimring and Hawkins, 1991) research indicated that such grouping is merited on
imprisonment policies as well. The states within a region tend to have similar prison
population policies, but different policies from states in other regions. The ranks of
1980’s prison population showed that sixty-five percent of all southern states were found
in the top third of the national prison population distribution, while seventy-eight percent
of the northeastern states were clustered in the bottom third of the national distribution
(Zimring and Hawkins 1991). In this study, the regional patterns are classified as
Northeast; Middle West; South; and West, symbolized by dummy variables separately.
The sentencing practices are not included directly in the state level models because of the
difficulties of factorizing these sentencing policies variables. For example, all states have
some form of mandatory minimum sentencing laws, but there exists enormous diversity
in types of offenses and time of enacting. To code this variable, further studies are

needed on each mandatory sentencing provision.

Modeling

As noted above, hierarchical multivariate linear models are used to examine the
effects of sentencing practices, which change among the observing period, on individual
state prison population increase rates, and they examine the effects of state level variables
on individual prisoner change curve outcomes. To demonstrate the certainty of using
hierarchical models, the one-way AN OVA models with random effects were examined

primarily.

27

Level-1 model
Y1: BO) + rj
Level-2 model
1301' = r00 + llol‘
Three outcome variables were new court commitments, time served in prison, and prison
populations separately. The results of ANOVA models are provided in Table 4.
The state level variance for new court commitments, time served in prison, and prison

population are 58.97%, 29.03%, and 52.17% respectively, which means that substantial

differences did exist among states.

Figure 2. US state prisons new court commitments rates

4.60 .,

New court commitments (LN)
a
8
_Wlw

1980 1985 1990 1995

Year

28

Table 4. ANOVA for New Court Commitment, Time Served in Prison, and Prison

 

 

 

 

 

 

 

 

 

 

Populations

Estimated Coefficients and Errors

Coefficient SE t-Value
Parameters
New Court Commitment 4.22 0.07 62.13***
Average state mean, r00
Error Variance Components Variance X2 d.f.
Between states M 0.23 1623.49 50
Within states r,- 0.16

Estimated Coefficients and Errors

Coefficient SE t-Value
Parameters
Time Served in Prison
Average state mean, r00 21.69 0.54 39.86***
Error Variance Components Variance x2 d.f.
Between states poj 13.59 496.72 50
Within states rj 33.22

Estimated Coefficients and Errors

Coefficient SE t-Value
Parameters
Prison Population
Average state mean, r00 5.08 0.07 72.92***
Error Variance Components Variance x2 d.f.
Between states pm 0.24 1288.20 50
Within states rj 0.22

 

Note: *p<0.1 **p<0.05 ***p<0.01

29

Four full models are set up to explain the relationship between sentencing policies
and prison population. The first model is the new court commitments model. A visual
examination of the grand mean of new court commitments curve (Figure 2) and each
state’s new court commitments curve as well indicated a nonlinear growth pattern. It
suggested an approximate quadratic change model.

Level-l Model

Y = ind1(Y1*)+ ind2(Y2*)+ ind3(Y3*)+ ind4(Y4*)+ ind5(Y5*)+ ind6(Y6*)+
ind7(Y7*)+ ind8(Y8*)+ ind9(Y9*)+ ind10(Y10*)+ ind11(Y11*)+ indl2(Y12*)+
ind13(Y13*)+ ind14(Y14*)+ ind15(Y15*)+ ind16(Y16*)+ ind17(Y17*)+ ind18(Y18*)+
ind19(Y19*)+ ind20(Y20*)+ ind21(Y21*)+ ind22(Y22*)+ ind23(Y23*)

Y*= BO+B1(time)+[32(time)2+p3(voluntary sentencing guidelines)+
+0.,(presumptive sentencing guidelines) +B5(parole release abolish) +B6(truth in
sentencing laws ) +B7(three strikes law) +Bg(consideration of prison capacity) +pg(crime
rates) +Blo(male at age 20-29) +Bn(unemployment rates) + r
Level-2 Model

[30=Goo+ 601(northeast)+ Goz(middle west)+ 603(west)+uo

[31: 610+ G, .(northeast)+ G,2(middle west)+ G,3(west)+p,

[32: 620+ GZI(northeast)+ G22(middle west)+ 623(west)+u2

l33= G30 134= G40
PF 050 [36: G60
B7= G70 138: 680
139: G90 1310= G100

[31F GIIO 1312= Gno-

30

The most frequent assumption in the standard HIM is that the within-state
residuals r’s are independent with a constant variance, 02. Considering there were 23
repeated observing occasions, it is likely that the variances are different at each occasion,
012; or there exist first-order auto-regressive level-1 errors, which means a level-1
residual is a function of the immediately preceding level-1 residual. The model for the
residuals is:

rtj=peH j+(1- p)et,- e”- _ independently N(0, 0'2)

At state level (level-2), the intercept, which measures the first occasion (1974)
new court commitments; time parameter (slope), which measures the increased rate of
each year’s new court commitment; and time squared parameter, which measures the
acceleration of each year’s new court commitments, are random and influenced by the

regional variables. The p,- is an error assumed normally distributed with mean zero and

variance 1;. All the other level-1 control variables coefficients are considered having fix
effects at level-2.

The second model is the time served in prison model. The estimated average time
served in prison curve displayed a complex trend (Figure 3). The average time served in
state prisons kept increasing before 1979, and then decreased to the lowest point at 1981
(during the study periods). From 1981 to 1984, the length of time served in state prisons
has again been increasing. After a short stable period, it began to decrease until 1990.
Since 1990, the average length of time served has been increasing once more. Thus, a
cubic growth model is used in level-l. Since the sentencing variables were lagged two

years in predicting the time served in prison, there were twenty-two observing occasions.

31

Figure 3. Estimated time served in state prisons

3.20 -i

l

3.1o-;

Time served in prison (LN)

ml

"""WWI'W" WI IWI 'I ' WI—T I" ' 'TW
1978 1980 1982 1984 1986 1988 1990 1992 1994 1996

Year

Level-1 Model

Y = ind1(Y1*)+ ind2(Y2*)+ ind3(Y3*)+ ind4(Y4*)+ ind5(Y5*)+ ind6(Y6*)+
ind7(Y7*)+ ind8(Y8*)+ ind9(Y9*)+ ind10(Y10*)+ ind11(Y11*)+ ind12(Y12*)+
indl3(Yl3*)+ ind14(Yl4*)+ ind15(Y15*)+ indl6(Y16*)+ indl7(Yl7*)+ ind18(Y18*)+
ind19(Y19*)+ ind20(Y20*)+ ind21(Y21 *)+ ind22(Y22*)

Y*= [30-1-81(time)+82(time)2+p3(time)3+[34(voluntary sentencing guidelines)+
+Bs(presumptive sentencing guidelines) +fl6(parole release abolish) +B7(truth in
sentencing laws) +Bg(three strikes law) +89(consideration of prison capacity) +Bm(crime
rates) +Bn(male at age 20-29) +fllz(unemployment rates) + r

The level-2 model will be the same as in the new court commitment model.

32

The third model is the prison population model. A visual examination of the
grand mean of prison population change curve, displayed in Figure 4, indicated a linear
growth pattern. There is a one-year lag for control variables, so there are twenty-three
indicator variables. The level-1 model and level-2 model are the same as in the new
court commitment model, except the level-1 model is a linear model.

Considering the influence of sentencing policies on the total prison population
may effect through the direct influence on the two components, the last model is set up to
examine the relationship between total prison populations at the end of a year with this
year’s new court commitments and time served in prison. This model includes twenty-
four years observations, which are from 1973 to 1996.

Figure 4. US states prison population rate

5.75 -

5.50 -

..___ “mml

5.25 -;

5.00 -

4.75 -
i

!

Prison population rate (LN)

 

1978 1980 1982 1984 1986 1988 1990 1992 1994 1996

Year

33

Level-l Model

Y = indl(Yl*)+ ind2(Y2*)+ ind3(Y3*)+ ind4(Y4*)+ ind5(Y5*)+ ind6(Y6*)+
ind7(Y7*)+ ind8(Y8*)+ ind9(Y9*)+ ind10(Y10*)+ indl 1(Y1 1*)+ ind12(Y12*)+
ind13(Yl3*)+ ind14(Y14*)+ ind15(Y15*)+ ind16(Y16*)+ ind17(Y17*)+ ind18(Y18*)+
ind19(Y19*)+ ind20(Y20*)-l- ind21(Y21*)+ ind22(Y22*)+ ind23(Y23*)+ind24(Y24*)

Y* = 80+Bl(time)+82(new court commitments)+83(time served in prison)+r
Level-2 Model

Bo=Goo+ Gm(northeast)+ 602(middle west)+ 603(west)+p.o

01: 610+ G. .(northeast)+ G,2(middle west)+ G.3(west)+p..

B2= (320

[33: G30

RESULTS

The results of the HMLM analyses for each of the outcome measures—new court
commitments, time served in prison, prison population, and the relations between these
three variables—are given in tables 5, 6, 7, and 8 respectively.

Each table contains two models. Model-1 assumes a homogenous level-l
variance, and model-2 addresses possible first-order auto-regressive level-1 variance.
The model deviances are employed to evaluate the fits of the two models. The results
show that the first-order auto-regressive models fit better than do the homogeneous

models. As in the new court commitments model, the first-order auto-regressive model

34

fits better than does the homogeneous sigma squared model, xz = 410.89, d.f. = l. The
maximum-likelihood estimate of correlation p in the first-order auto-regressive model
equals 0.775; its standard error is 0.044, which suggests a significantly positive auto-
correlation. This phenomenon happens in the three other models. Therefore, the reported
results use data from models with first-order auto-regressive level-1 variance.
Examination of the model indicates that state’s average new court commitment at
1974 was 38.86 (e3'66) per 100,000 populations. States in the Eastern region had
significantly less new court commitments than as in the South region at that time. States
in the Middle West and West region had the same propensity as comparing with Southern
states. The mean increase rate of new court commitments is 0.07. Nevertheless, the
region does not make any differences for the increase rates. The mean acceleration of the
new court commitments is negative 0.001 , which means that with the time passed, the
change of increase rate slowed down. There is no significant difference between
different regions. Only the acceleration of Eastern states is a little slower than that of
Southern states and significant at 0.1 levels. Sentencing policies do not influence the new
court commitments directly and significantly, except sentencing commissions who are
required by legislative direction to consider about the prison capacity. As a state has
sentencing commission, and it is required by legislatures to consider prison capacity as
they develop guidelines, the new court commitments reduced by 1.17 (e016) per 100,000
populations comparing with no prison capacity consideration. Another sentencing
practice that may influence the new court commitments is the three strikes law. The three

strikes law has a coefficient as negative 0.068, and a p-value as 0.098. Crime rates,

35

unemployment rates, and percent of male at age twenty to twenty-nine populations are
not direct factors that account for the shifts of new court commitments.

The findings in the time served in prison model indicate some different
relationships both within and between states. The average time served in prison at 1975
was 17.3 months. No significant difference was found between Eastern and Southern,
Middle- West and Southern, or Western and Southern. Time, time squared, and time
cubical are significant covariates. Similarly, the regional variables at state level do not
make any difference for the increased rate of time served in prison, slope of time squared,
and .slope of time cubical. The three strikes law is the only sentencing practice that is
significantly associated with time served in prison. When three strikes laws are enacted,
the state prison stays for inmates increase 3.98 months. The unemployment rates also
significantly correlate with the time served in prison. As unemployment rates increased
by one unit (1 person per 100 labor force), the time served in prison increased 0.17
months. No other sentencing policies are found significantly covariate with length of
time served in prison.

None of the sentencing practice variables shows any direct and significant effects
on the total prison populations. Crime rates, unemployment rates, and demographic
shifts do not have an obvious and significant impact on total prison population either.
Although the 1974’s prison populations in the East area states and in the Middle West
area states are significantly less than prison populations in Southern states, the increase
rates of prison populations do not significantly differ from each region. The average

increase rate of prison population was 0.07.

36

The prison population as an effect of new court commitments and length of stay
in prisons model shows some expected results. Both new court commitments and the
length of stay in prisons significantly correlate with total prison populations. As new
court commitments increase by one unit, the prison populations increase by 0.24 units
(approximately one person increase in new court commitments results in 0.47 increase in
total prison population). A One-month increase of the length of a stay in prison results in
0.01 units increase of prison populations (approximately 1.01 prison populations).
Therefore, changes in the length of stay in prisons may have more effects on the total
prison population than changes of new court commitments. Even though the evidence
does not show that sentencing reforms have the expected strong influences on the
aggregate imprisonment outcomes, one may think likely that sentencing reforms focused
on the prison terms drive the growth of prison population more powerfully than

sentencing reforms focused on prison admissions.

DISCUSSION

The research results obtained from the new court commitment model and the time
served in prison model do not strongly support the hypotheses that sentencing practices
change new court commitment and the time served in prison. As noted above, sentencing
practices, such as truth in sentencing laws, which were considered as the major reason
that drove the rapid growth of prison pOpulation through the influence on the length of
stays in prison, does not show the same effects in this modeling. Presumptive sentencing

guidelines do not moderate or increase the prison population either. No apparent

37

relationships between the determinate sentencing laws with abolishment of parole and the
prison status were been found. Among the six sentencing practices examined, only the
sentencing commission’s requirements of considering prison capacity and the three
strikes laws correlated with new court commitments and time served in prisons. When
prison capacity is required to be taken into consideration by the state’s sentencing
commission, the new court commitment decreases significantly. The three strikes laws
have some moderately negative influence on new court commitments as well. Three
strikes laws influence the length of time served in prison significantly and positively, and
the consideration of prison capacity by sentencing commission has some moderately
positive influence on the length of time served. However, one phenomenon should be
taken seriously is that these two sentencing practices have reversed influence direction to
these two outcome variables. As three strikes laws increase time served in prisons, it
decreases new court commitments moderately. At the same time as requirements of
sentencing commission’s considering about prison capacity decrease new court
commitments, it increased time served in prison moderately as well. As result, the
combined effects will disappear. This research outcome is consistent with the research
results of prison population model, which suggests that the aggregate measurement of
severity of sentencing outcomes, the prison populations do not change in accordance with
the sentencing reforms. These findings, which reject the formal legal rationality
hypothesis of sentencing reforms, did support the substantive rationality perspective of
sentencing reforms in some extent. While the actual substantive circumvent factors are
not clearly identified in this thesis, there are evidences showing that the sentencing

system itself has a trend to balance the influences of sentencing practices. When policies

38

increase one component of prison populations, it may decrease another component of
prison populations. Although the “hydraulic displacement of discretion” is not examined,
the principle behind this theory is verified by the evidence from this study, which could
be described as a sentencing system that is a self controlled system, and changes in one
segment will result in an opposite changes in another segment, thus the sentencing
outcomes keep at a constant level.

It is difficult to tell exactly how many changes there are of time served in prison,
and how many changes there are in new court commitments when analyzing the
component of growth of prison populations. The research results do tell, however, that a
one month increase of time served in prison will have more impact on total prison
population than that of a one person increase of new court commitment. This result is
consistent with other early research (Marvell, 1995; Wooldredge 1996).

The regional differences did exist significantly at the research starting point,
which is a year before major sentencing reforms were enacted, but it does not make
apparent difference in the imprisonment developing processes. This indicates that
imprisonment differences between states in different regions existed before the
sentencing reform periods. We get an impression here that future research will seek
answers about the imprisonment diversities that should focus on the social, economical,
political, governmental, and criminal justice characteristics of each state before 1974.

While the evidence presented here appears to be strong, there are several
limitations to this study. The first is the exclusion of drug factors, which is considered as
one of the major factors that increases the incarceration rate (Sabol and Lynch 1997;

Blumstein and Beck, 1999; Caplow and Simon 1999). As Blumstein (1998) noted, in

39

1980, 6.8% of all new court commitments to prison were drug offenders; by 1992,
however, the drug offenders comprised 30.5% of persons admitted to state prisons. Due
to the difficulty of obtaining drug arrests data before 1985, it was not included in the
control variables.

The second limitation is the use of regional variables as the representation of
diversity of social, political, and criminal justice policies. Even the regional variables can
reflect a few features in some degree, they may be too simple to reflect the individual
influences on imprisonment, especially when the influences have reversed effects, the
aggregate effects may disappeared as result.

Since this thesis tests the formal rationality hypotheses of sentencing reforms at
an aggregate level, the possible substantive factors are not examined in this study. This
makes the explanation of which factors, and how the factors influence the sentencing
outcomes impossible. The fact of the rapid growth of prison populations indicates that
there must be some causations that drive the hasty increase, however, sentencing reforms

are not this kind of drivers.

40

Table 5: Sentencing Policies Effects on New Court Commitmentsll

 

Estimated Coefficients and Errors

 

Model 2 with first-order auto-
regressive level-1 variance

Model 1 with homogeneous level-
1 variance

 

 

Parameters Coefficient SE t-Value Coefficient SE t-Value

Average state

mean 3.69*** 0.07 56.84 3.66*** 0.06 58.97
east ’2 -0.91*** 0.19 -4.79 -0.89*** 0.18 -4.97 '
mid-west -0.68*** 0. 17 -3.92 -0.70*** 0. 16 -4.26
West -0.61*** 0.17 -3.56 -0.59*** 0.16 -3.67

Time

(increase rate) 0.049*** 0.01 5.26 0.07*** 0.01 6.60
East 0.03 0.02 1.10 0.04 0.02 1.58
mid-west 0.00 0.02 0.21 0.01 0.02 0.58
West 0.03 0.02 1.58 0.03 0.02 1 .37

Time2

(accelerate) 0.00 0.00 0.25 -0.00* 0.00 -1 .92
East 000 0.00 -1 . 18 -0.00* 0.00 -1.85
Mid-west 0.00 0.00 0.37 0.00 0.00 0.05
West -0.00 0.00 -0.98 -0.00 0.00 -0.71

Voluntary 0.05 0.07 0.72 0.03 0.07 0.46

sentencing

Presumptive 0.24*** 0.06 3.76 0.04 0.07 0.58

sentencing

Parole

abolishment -0.08 0.06 -1 .29 -0.07 0.06 -l .08

Truth in

sentencing -0.17*** 0.05 -3.66 -0.06 0.05 -1.30

3-strikes law -0.21*** 0.04 -4.75 -0.07* 0.04 - 1 .68

Considering

prison capacity -0.30*** 0.05 -5.60 -0.16** 0.06 -2.87

Crime rates 0.02 0.04 0.60 -0.01 0.03 -0.40

Unemployment

rate -0.00 0.00 -0.39 0.00 0.00 0.09

Male at age 20

to 29 0.01 0.02 0.44 -0.02 0.03 -0.56

#Parameters 28 29

Deviance -189.91 -600.81

Comparison )8 df

Model 1 vs.

Model 2 410.89*** 1

 

 

'1 Three kinds of parameter estimates are available in hierarchical linear model: empirical Bayes estimates
of randomly varying level-l coefficients; generalized least squares estimates of the level-2 coefficients; and
maximum-likelihood estimates of the variance and covariance components.

’2 All these regional variables are vs. southern.

41

Table 6: Sentencing Policies Effects on Time Served in Prisons

Estimated Coefficients and Errors

 

Model with homogeneous level-1

Model with first-order auto-

 

 

variance regressive level—l variance

Parameters Coefficient SE t-Value Coefficient SE t-Value

Average state

mean 16.83*** 0.78 21.70 17.30“" 0.87 19.88
East -4.66** 2.09 -2.24 -3.62 2.37 -1.53
Mid-west -0.60 1.91 —0.32 -1.24 2.17 ~0.57
West -0.21 1.88 -0.1 1 -0.64 2.15 -0.30

Time 1.70*"‘* 0.32 5.28 1.74*** 0.36 4.81
East 0.89 0.76 1.17 0.40 0.84 0.48
Mid-west -0.62 0.70 -0.89 -0.48 0.77 -0.62
West -0.37 0.68 -0.55 —0.39 0.76 -0.51

Time2 -0.16*** 0.04 4.34 019*" 0.04 4.50
East 010 0.10 -0.97 -0.06 0.1 1 -0.54
Mid-west 0.10 0.09 1.08 0.09 0.10 0.88
West 0.07 0.09 0.74 0.07 0.10 0.71

Time3 0.00m 0.00 3.58 0.01m 0.00 4.13
East 0.00 0.00 1 .03 0.00 0.00 0.78
Mid-west -0.00 0.00 -1 . 13 -0.00 0.00 -0.97
West 000 0.00 -0.92 -0.00 0.00 -086

Voluntary

sentencing 1.84 1.18 1.56 0.75 1.26 0.60

Presumptive

sentencing -1.06 1.1 1 -0.96 -1.46 1.16 -1 .26

Parole

abolishment -l.98* 1.13 -1.76 - l . 12 1.19 -0.94

Truth in

sentencing 0.97 1 . 17 0.83 1.17 1.01

3-strikes law 6.63“” 1.26 5.25 3.98*** 3.45

Considering

prison capacity 2.32" 1.02 2.26 1.85* 1.09 1.71

Crime rates -3.52*** 0.73 -4.86 -0.93 0.62 -1.50

Unemployment

rate 0.29*** 0.07 4.14 0.17 ** 0.08 2.22

Male at age 20

to 29 -0.75 0.55 -l.37 -0.54 0.66 -0.82

Number of 36 37

Parameters

Deviance 6149.88 5847.19

Comparison )6 df

Model 1 vs.

Model 2 302.68*** 1

 

42

Table 7: Sentencing Policies Effects on Prison Population

Estimated Coefficients and Errors

 

Model with homogeneous level-1

Model with first-order auto-

 

 

variance regressive level-1 variance

Parameters Coefficient SE t-Value Coefficient SE t-Value

Average state

mean 4.33""""I 0.05 82.23 4.32""""l 0.13 34.24
East -0.95*** 0.15 -6. 17 -0.88** 0.37 -2.37
Mid-west -0.70*** 0. 14 —4.96 -0.78** 0.34 -2.29
West -0.52*** 0.14 -3.76 -0.54 0.33 -1.62

Time 0.07*** 0.00 32.73 0.07*** 0.00 25.86
East 0.01" 0.01 2.29 0.01 0.01 0.83
Mid-west 0.01 0.01 1.00 0.01 0.01 1.26
West 0.01 0.01 1.76 0.01 0.01 1.15

Voluntary

sentencing -0.00 0.04 -0.05 -0.03 0.04 -0.75

Presumptive

sentencing -0.02 0.03 -0.45 -0.03 0.03 -0. 19

Parole

abolishment -0.11*** 0.03 -3.55 -0.02 0.03 -0.61

Truth in

sentencing 0.00 0.02 0.14 -0.00 0.02 -0.05

Three strikes

law 0.01 0.02 0.30 -0.02 0.02 -1. 10

Considering

prison capacity -0.08** 0.03 -2.83 -0.02 0.03 -0.72

Crime rates -0.00 0.02 -0.16 0.01 0.01 0.79

Unemployment

rate 0.00 0.00 0.08 0.00 0.00 1.43

Male at age 20

to 29 0.04*** 0.01 6.03 0.02 0.01 1.63

Number of

Parameters 21 22

Deviance -1427.70 -2358.66

Comparison )8 df

Model 1 vs.

Model 2 930.96*** 1

 

43

Table 8: Admission and Length of Stay Effects on Prison Population

 

Estimated Coefficients and Errors

 

Model with first-order auto-
regressive level-1 variance

Model with homogeneous level-l
variance

 

 

Parameters Coefficient SE t-Value Coefficient SE t-Value

Average state

mean 4.56"‘M 0.05 90.32 4.42*** 0.1 1 39.97
East -0.89*** 0.15 -6.15 -0.82** 0.32 —2.53
Mid-west -0.68*** 0. 13 -5. 10 -0.74** 0.30 -2.51
West -0.47*** 0. 13 -3.64 -0.52* 0.29 -1 .79

Time(increase

rate) 0.04“"II 0.00 22.66 0.05*** 0.00 31.36
East 0.01* 0.00 1.72 0.00 0.00 0.55
Mid-west 0.00 0.00 0.56 0.01 0.00 1.49
West 0.00 0.00 1.12 0.01 0.00 1.5

New court

commitments 0.42*** 0.02 24. 10 0.24*** 0.01 25.77

Time served in

prisons 0.02*** 0.00 22.76 0.01““ 0.00 20.75

Number of '

Parameters 14 15

Deviance -1842.31 ~2140.75

Comparison x2 df

Model 1 vs.

Model 2 298.44*** 1

 

REFERENCES

American Friends Service Committee. Struggle for Justice. New York: Hill and Wang,
1971.

Blumstein, Alfred, and Jacqueline Cohen. "A theory of the stability of punishment."
Journal of Criminal Law and Criminology 64 (1973): 198-207.

Blumstein, Alfred. “Criminal Justice Conundrum: Rising Prison Populations and Stable
Crime Rates.” Crime & Delinquency 44 no.1 (January 1998): 127-135.

Blumstein, Alfred, and Alfred Beck. "Population Growth in US. Prison, 1980-1996." In
Prisons, edited by Michael Tonry and Joan Petersilia, 17-61. Crime and Justice: A
Review of Research 26. Chicago: University of Chicago Press, 1999.

Bryk, Anthony S., and Stephen W. Raudenbush. Hiearchical Linear Models:
Applications and Data Analysis Methods. Advanced Quantitative Techniques in
the Social Sciences 1. Newbury Park, CA: Sage Publications, 1992.

Bureau of Justice Assistance. 1996 National Survey of State Sentencing Structures.
Department of Justice, Office of Justice Program, B] A. Washington, DC: B] A,
September 1998.

—. National Assessment of Structured Sentencing. Washington, DC: BJ A, 1996.

Cohen, J ., and J. A. Canela-Cacho. "Incarceration and Violent Crime: 1965-1988." In
Understanding and Preventing Violence, edited by A. J. Reiss and J. A. Roth.
Washington DC: National Academy Press, 1994.

Ditton, P. M., and D. J. Wilson. Truth in Sentencing in State Prisons. US Department of
Justice, Bureau of Justice Statistics. 1999.

Engen, R. L., and S. Steen. "The Power to Punish: Discretion and Sentencing Reform in
the War on Drugs." America Journal of Sociology 105, no. 5 (2000): 1357-95.

Fumiss, J. R. "The Population Boom." Corrections Today 58, no. 2 (1996): 38-43.

Gottfredson, Michael R., and Don M. Gottfredson. Decision Making in Criminal Justice:
Toward the Rational Exercise of Discretion. Edited by J oel Feinber, et a1. 2nd ed.
Law, Society and Policy 3. New York: Plenum Press, 1988.

Greenfeld, W.A., A Beck, and D. Gilliard. "Prisons: Population Trends and Key Issues
for Management." Criminal Justice Review 21, no. 1 (1996): 4-20.

45

Kensey, A., and P. Toumier. "Prison Population Inflation, Overcrowding and
Recidivism: The Situation in France." European Journal on Criminal Policy and
Research 7 (1999): 97-119.

Langan, P.A. "America's Soaring Prison Population." Science 251 (1991): 1568-73.

Levitt, Steven D. "The Effect of Prison Population Size on Crime Rates: Evidence from
Prison Overcrowding Litigation." Quarterly Journal of Economics 3, no. 2
(1996): 319-51.

MacKenzie, D. L., G. S. Tracy, and G. W. Willianms. "Incarceration Rates and

Demographic Changes: A Test of the Demographic Change Hypothesis." Journal
of Criminal Justice 16, no. 3 (1988): 241-53.

Martinson, Robert. "What Works? -- Questions and Answers About Prison Reform." The
Public Interest 35, no. 22 (1974).

Marvel], T. B., and C. E. Moody. "Deterrninate Sentencing and Abolishing Parole: The
Long-Term Irnpactes on Prisons and Crime." Criminology 34, no. 1 (1996): 107-
20.

Maxwell, Sheila Royo. "Examining the Congruence Between ROR and Failures-to-
Appear." Journal of Criminal Justice 27, no. 2 (1999): 127-42.

Miethe, Terance D. "Charging and Plea Bargaining Practices Under Deterrninate
Sentencing: An Investigation of the Hydraulic Displacement of Discretion."
Criminology 78, no. 1 (1987): 155-76.

Reitz, K. R., and C. R. Reitz. "The American Bar Association's New Sentencing
Standards." In Federal Sentencing Reporter, vol. 6, 170-71, 1993.

Sabol, William J ., and James P. Lynch. "Crime Policy Report: Did Getting Tough on
Crime Pay?" Www.urban.orglcrime/crime.htm. The Urban Institute, August,
1997. 24/09/97.

Savelsberg, J. J. "Law That Does not Fit Society: Sentencing Guidelines as a
Neoclassical Reaction to the Dilemmas of Substnativized Law." America Journal
of Sociology 97, no. 5 (1992): 1346-81.

Sherman, Lawrence W., and Ellen G. Cohen. "The Impact of Research on Legal Policy;

the Minneapolis Domestic Violence Experiments." Law &. Society Review 23, no.
1 (1989): 117-44.

Tonry, Michael. "Building Better Policies on Better Knowledge." In The Challenge of

Crime in a Free Society: Looking Back; Looking Forward.
www.ojp.usdoj.gov/guide/lblf/panel4c.htm. Symposium on the 30th Anniversary

46

of the President's Commission on Law Enforcement's and Administration of
Justice. Washington, DC: Office of Justice Programs, US. Department of
Justice, 19-21 June, 1997.

U. S. General Accounting Office, Truth in Sentencing: Availability of federal grants
influenced laws in some states (1998).

M Tonry, National Institute of Justice/Corrections Program Office, US. Department of
Justice, The fragmentation of sentencing and corrections in America. (1999).

D. Parent, et al., Office of Justice Programs, National Institute of Justice, US.
Department of Justice, Key legislative issus in criminal justice: The impact of

sentencing guidelines. (1996).

D. K. Gilliard, Bureau of Justice Statistics, US Department of Justice, Prison and jail
inmates at midyear 1998 (1999).

Walker, S. Taming the System. New York: Oxford University Press, 1993.

Wicharaya, T. Simple Theory, Hard Reality: The Impact of Sentencing Reforms on
Courts, Prisons, and Crime. Albany: State Universtiy of New York Press, 1995.

Wooldredge, John. "Research Note: A State-Level Analysis of Sentencing Policies and
Inmate Crowding in State Prisons." Crime and Delinquency 42, no. 3 (1996): 456-
66.

Zimring, F, and G. Hawkins. The Scale of Imprisonment. Chicago: University of Chicago
Press, 1991.

47

lllllllllllilllllllil '