Professional Documents
Culture Documents
net/publication/229454237
CITATIONS READS
50 2,183
1 author:
Jeffery T. Ulmer
Pennsylvania State University
114 PUBLICATIONS 4,953 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Jeffery T. Ulmer on 30 October 2017.
Social scientists have a longstanding concern with the relationship between crimi-
nal sanctions and offenders’ future behavior. This paper uses data from a sample of 528
adult felony offenders to compare the relative probability of rearrest, the severity of
rearrest, and the likelihood of probation revocation for offenders with a sentence of
incarceration, work release, house arrest, and traditional probation-or a combination of
these sentences. Consistent with previous research, prior record, gender, offense type,
and education were significantly related to the probability and severity of rearrest.
Furthermore, no matter which other sanction it was paired with, house arrest was asso-
ciated with reduced chances of rearrest and lower rearrest severity. The influence of one
of these sanction combinations was gender-specific: incarceration followed by house ar-
rest was associated with reduced odds of rearrest for women but not for men. However,
house arrest was associated with the considerably increased likelihood of probation re-
vocation. In addition, these effects remain after controlling for potential selection bias
stemming from the sentencing decision. This paper concludes by discussing the
theoretical and correctional implications of these findings and directions for future
research.
1971; Schur 1980; Hagan and Palloni 1990). For example, sanctions such as
imprisonment are said to stigmatize offenders, close off opportunities for
conventional activities and social ties, open up opportunities for participation in
deviant subcultures, socialization,and activities, and foster deviant self-definitions
(Ulmer 1994).
The 1990s has seen a recent theoretical convergence between labeling
theories, life-course oriented social control theories (Sampson and Laub 1993,
1997), and theories of differential social control (Matsueda 1992; Matsueda and
Heimer 1997). In particular, Sampson and Laub (1997, pp. 147-50) argue that
punitive, segregative sanctions like imprisonment “knife off” conventional social
bonds and opportunities, with repercussions for criminal behavior throughout the
life course. They also argue that criminal sanctions which entail structured ways
to preserve, mobilize, or foster the growth of social bonds to conventional adult
institutions such as work and family are far more likely to prevent future crimi-
nal offending in the short and long term. Furthermore, Braithwaite’s (1989) rein-
tegrative theory of criminal sanctioning synthesizes insights from labeling,
social control, opportunity, and differential association theories. Braithwaite
(1989) differentiates between reintegrative and stigmatizing sanctions. While
reintegrative sanctions can prevent future offending by restoring or preserving
conventional social bonds, opportunities, and socialization processes, stigmatiz-
ing sanctions can lead to continued or increased criminal offending by having the
reverse effects. This reintegrative theory of sanctioning has received some solid
empirical support (e.g., Makkai and Braithwaite 1991; Heimer and Staffen
1995). Also, Sherman (1993) argues that perceived procedural justice conditions
the effect of sanctions on future crime. Sanctions perceived by the offender to be
unjust tend to result in defiance, while those perceived as just tend to result in
shaming and reintegration.
Since the late 1980s, intermediate sanctions (sometimes also known as
alternative sanctions, alternative punishments, or community corrections), such
as house arrest with electronic monitoring, work release programs, and super-
vised labor in the community (usually known as “road crew” or “work crew”)
have become increasingly common. The primary policy motivation for such
programs is often thought to be a financial one-the conservation of scarce
corrections resources (Feeley and Simon 1992). However, these programs are
also sometimes touted as being potentially more effective at preventing recidi-
vism than traditional probation or incarceration (e.g., Morris and Tonry 1990;
Byrne 1990).
Byrne’s (1990) assessment of the possibilities and limitations of
intermediate sanctions provides an implicit linkage between the theoretical
directions outlined above and recent developments in corrections. Byrne (1990)
argues that intermediate sanctions will prevent recidivism to the extent that they
166 JEFFERY T. ULMER
another with no real benefit or cost saving). Of particular concern for this study
is the issue of “net strengthening.” If offenders are under the heightened surveil-
lance and control of intermediate sanctions like house arrest, there may be a
greater likelihood that any program rule violations or criminal behavior will be
caught, and thus a greater likelihood that offenders will be incarcerated for such
violations.
On the other hand, Jones (1990) argues that net widening and strengthening
are not inevitable by-products of intermediate sanctions. Furthermore, Blomberg
(1 995) argues that intermediate sanction programs become embedded in
complex organizational, correctional, community, and policy contexts, and this
can result in mixed or indeterminate outcomes for offenders and the programs
themselves. Blomberg (1995) calls for more empirical research on such poten-
tially mixed outcomes:
But what do we really know empirically about this particular penal reform movement [inter-
mediate sanctions]. Empirically speaking, little is known about the precise origins, operational
features, and potentially mixed outcomes of this particular reform [intermediate sanctions].
Multivariate analyses are surprisingly rare that directly compare the proba-
bilities of recidivism between intermediate sanctions vs. traditional sentences
using varied samples of offenders (e.g., samples that include a variety of
convicted offenses, as well as variation in offender characteristics such as gender
or age). Those that do exist provide mixed results. On one hand, some studies
fmd no significant difference in recidivism between offenders sentenced to
intermediate sanctions and offenders sentenced to incarceration or traditional
probation (Turner and Petersilia 1998; Smith and Akers 1993; Couhvright,
Berg, and Mutchnick 1997). On the other hand, Petersilia and Turner (1993)
found that in intermediate sanctions programs where offenders received both
supervision and rehabilitative treatment, recidivism was reduced by 20% to 30%
(see also Petersilia 1995). Geerken and Hayes (1993) found that offenders in
intermediate sanctions programs in Louisiana exhibited moderately lower overall
recidivism rates compared to those released from prison, while Glaser and Watts
(1992) found that drug offenders on house arrest in Los Angeles exhibited
significantly less probation revocations and rearrests than drug offenders on
traditional probation. In Washington state, Berliner, Schram, Miller, Milloy
(1995) found that while recidivism rates for sex offenders receiving intermediate
sanctions were not significantly different from sex offenders sentenced to incar-
ceration, other kinds of offenders receiving intermediate sanctions did exhibit
significantly less recidivism than their counterparts who were sentenced to
incarceration.
Furthermore, Albonetti and Hepburn (1 997) compared the likelihood and
timing of probation revocation between offenders on traditional probation vs.
intensive probation (i.e., probation with more intensive probation officer super-
vision, more frequent drug/alcohol screenings, and more extensive drug/alcohol
treatment) among a sample of probationers in Phoenix. They found that those in
intensive probation were actually more likely to have their probation revoked,
lending support to the notion that intermediate sanctions can result in net
strengthening. Their explanation was that more intense supervision and
drug/alcohol testing meant that violations were more likely to be detected, and
did not necessarily mean that those on intensive probation actually committed
more violating behaviors than those on traditional probation. They also found
that probation revocation was more likely among males, those with more ex-
tensive and serious prior criminal histories, and those with low education.
Further, they found that the effects of offender’s gender, ethnicity, age, and su-
pervision level on probation revocation depended on offender’s education and
prior criminal history.
In sum, while prior research literature gives only modest and mixed
guidance on the comparative effect of intermediate sanctions on recidivism, most
prior studies agree that offenders’ gender, prior criminal record, type/severity of
170 JEFFERY T. ULMER
tative data from semi-structured interviews with each of the county judges who
handled criminal cases, the county district attorney and members of his stafT,
four members of the public defender’s office, the chief of adult probation, and
community corrections staff, as well as a varied subsample of offenders.
Although not the focus of this paper, these interview data inform and contexb-
alize the interpretation of the quantitative findings.
To select the sample, I obtained lists from county community corrections,
probation, jail, and department of corrections (DOC) agencies of adult felony
offenders released from house arrest or work release, supervised probation,
county jail, or state prison between 1991 and 1995. I then selected a stratified
random sample from these sampling frames. I was particularly interested in
obtaining adequate variation in convicted offenses, and obtaining adequate
numbers of women and black offenders for meaningful comparisons (the
county’s population was less than 9% black, and blacks were not highly overrep-
resented in the county’s offender population). Therefore, I grouped each
sampling frame according to offenders’ statutory felony severity level (A, B, C,
or D), gender, and race/ethnicity, randomly selected offenders within these
categories, and oversampled women and blacks. The sample consists of 528
offenders. Once these samples were selected, we collected and cross-checked
data from three sources: 1) county community corrections case files; 2) the
county courthouse’s computerized case database; 3) the state department of
corrections’ offender tracking data.
While the offenders in the sample were released between 1991 and August
of 1995, we tracked their rearrests and probation revocations until August of
1997. Offenders sentenced to jail andor prison were tracked from the time of
their release from incarceration.2 Offenders sentenced to house arrest, work
release, or traditional probation were tracked from the time of sentencing, since
their sentences allowed “on-street” time and thus immediate opportunity to be
rearrested. Thus, all offenders in the sample have a minimum of two years of
follow-up time within which to be rearrested or revoked. To address the disparate
lengths of on-street, non-incarcerated follow-up time for offenders in the sample,
I control for the number of months of “street time” (time since sentence or
incarceration release) in the later analyses.
A number of recent recidivism studies (DeJong 1997; Baumer 1997;
Albonetti and Hepburn 1997) have focused on the timing of various measures of
recidivism (e.g., self-reported reoffending, rearrest, probation revocation) using
time series or hazard models. The issue of the timing of recidivism, while
important, is conceptually distinct from the goal of this paper. My goal here is
simply to compare the relationships between different types and combinations of
sentence options and the overall probability and severity of rearrest and the
probability of revocation.
172 JEFFERY T. ULMER
Appendix A presents the variables used in the analysis and their coding.
Most of the variables and their coding are straightforward and self-explanatory.
However, three features of the data require further comment. First, data on two
key variables-marital status and education-were sometimes missing in the
three record sources above, and we were legally prohibited from accessing
probation department presentence report^.^ Rather than excluding these
important variables from the analyses completely, I utilized two common and
conservative techniques for dealing with missing data (Little and Schenker
1995). For education, I assigned the mean value of education (1 1.4 years) to
missing cases. For marital status, I created a dummy variable to indicate whether
marital status was missing or not, and included this variable in the models
presented below. This retains cases with missing marital status information
while also allowing a check on whether these missing cases are systematically
different from non-missing cases regarding the dependent variables (as seen in
the models below, they are not).
Second, since data on prior record and rearrests came from county comput-
erized courthouse records (supplemented by community corrections, adult
probation, and department of corrections records as appropriate), if for some
reason an arrest did not result in charges filed, (i.e., if the charge was dropped
prior to filing in court), then the data would not reflect such cases. In addition,
obviously neither the prior record nor rearrest data capture criminal behavior that
did not result in an arrest. The overall effect of these limitations, I believe, is that
the prior record and rearrest measures are conservativethat is, any errors or
omissions in these data almost surely underestimate the true amount of prior
criminal behavior and/or recidivism for the research subjects. On the other hand
it is arguably better from a data validity standpoint to underestimate these crucial
variables than overestimate them-conservative data are better than fictional
data. Thus, the prior record and rearrest data should be understood as represent-
ing clearly known and recorded prior convictions and rearrests.
Third, the prior criminal convictions scale (hereafter, prior record) reflects
both the number and severity of prior criminal convictions. As Appendix A
describes, different severities of offenses result in differing amounts of points on
the prior record scale. For example, a prior “A” felony would count four points
on the scale, while a prior misdemeanor would count only half a point4 In this
way, the prior record scale weights more severe prior convictions more heavily
than less severe ones. Finally, the rearrest severity scale ranges from zero to
eight, and simply corresponds to the statutory letter grading of misdemeanors
and felonies. For example a rearrest for a “C” misdemeanor counts “1” on the
scale, while a rearrest for an “A” felony counts “7.”
The analysis consists of multivariate models of rearrest, rearrest severity,
and the probability of probation revocation. Logistic regression is used for the
PROBABILITY AND SEVERITY OF RECIDIVISM 173
Mean (standard
Variable N (Percent) deviation)
Mean (standard
Variable N (Percent) deviation)
7 7 ( 1%)
8 or greater 7 ( 1YO)
Missing 11( 2%)
Basic sentence categories -
Incarceration 274 (52%)
Work release 75 (14%)
House arrest 295 (56%)
Traditional probation 500 (95%)
Sentence category combinations
Incarceration only 26( 5%)
Work release and probation 35( 7%)
Work release and house arrest 40 ( S o / )
Incarceration and house arrest 82 (16%)
Incarceration and probation 88 (17%)
House arrest and probation 171 (32%)
Traditional probation only 86 (15?'0)
Probation revoked (includes only
probation cases)
No 331 (66%)
Yes 169 (34%)
Rearrest -
No 295 (56%)
Yes 233 (44%)
Rearrest severity scale 1.7 (2.22)
None (0) 295 (56%)
D misdemeanor ( 1 ) 3 ( 0.5%)
C misdemeanor (2) 77 (14.7%)
B misdemeanor (3) 9 ( 1.7%)
A misdemeanor (4) 30 ( 5.8%)
D felony (5) 70 (13Yo)
C felony (6) 36 ( 6.8%)
B felony (7) 3 ( 0.5%)
A felony (8) 1 ( 0.2%)
Missing 4 ( 0.7%)
Table 2
Logistic Regression Models for Rearrest Odds (Probability)
Model 1 Model 2
a White is the reference for race/ethnicity. Single is the reference for marital
status. “Other” offenses are the reference for offense category combinations.
Probation is the reference for basic sentence categories and sentence com-
binations.
* Statistically significant at .05 or less.
** Statistically significant at .01 or less.
PROBABILITY AND SEVERITY OF RECIDIVISM 177
criminal record. Since the prior record scale is a continuous variable, its effect
on the odds of rearrest is cumulative-that is, each increase in the prior record
scale is associated with a 32% increase in the odds of rearrest (in Model 2). For
example, from both Model 1 and 2, offenders with a prior record score of “4”
would have about 2.86 times greater odds of being rearrested than an offender
with a prior record score of “0.”5 Thus, the more numerous and serious are
offenders’prior criminal records, the greater the likelihood of their rearrest. This
finding is no surprise, and coincides with the large majority of prior studies of
criminal careers and recidivism.
Offenders’ sex is also a strong predictor of rearrest. Females’ rearrest odds
are almost half those of males in both models, and also females have only a 35%
(Model 2) to 37% (Model 1) probability of rearrest compared to a roughly 60%
probability for males.6 This finding is also in line with previous criminological
literature, which finds that females are far less involved in crime, especially
serious crime, than males (Steffensmeier and Allan 1996), and less likely to
recidivate than males (e.g., Baumer 1997). The type of offense for which
offenders were convicted also influences the likelihood of rearrest in both
models. In particular, those convicted of property crimes have odds of rearrest
that are well over two-and-a-half times greater than other offenses, and have
about a 74% probability of rearrest. Those convicted of autohaffic felonies (i.e.,
repeat drunk driving or “habitual traffic violators”) are also significantly more
likely to be rearrested in Model 2. The lowest (but not statistically significant)
probabilities of rearrest (48%) are associated with non-rape sex offenses (i.e.,
child molesting, incest) followed by drug offenses (55%). Education is also
associated with moderately decreased odds of rearrest in both models. Each year
of education yields a roughly . I 8 decrease in rearrest odds.
Of the basic sentence categories in Model I, only house arrest is associated
with significantly reduced odds of rearrest compared to the reference category,
traditional probation. Those offenders for whom the court orders mandatory drug
or alcohol treatment are also moderately more likely to be rearrested in Model 1.
This finding may seem counterintuitive at first-a purpose of drug or alcohol
treatment, after all, is to prevent alcohol or drug related problems such as
criminal behavior in the future. In fact, a great deal of social science research
and criminal justice experience indicates that drugs and alcohol are in fact
strongly related to criminal activity (Blumstein 1993; Horney et al. 1995; Parker
1998). When a judge orders drug or alcohol treatment as a condition of proba-
tion, this indicates that an offender already has a drug or alcohol problem serious
enough to warrant treatment, and he/she who may already be at a modest
disadvantage in terms of recidivism. .
Model 2 depicts the effects of the various sentence combinations on
rearrest. First, each of the combinations result in decreased likelihood of rearrest
178 JEFFERY T.ULMER
"White is the reference for race/ethnicity. Single is the reference for marital
status. "Other" offenses are the reference for offense category combina-
tions. Probation is the reference for basic sentence categories and sentence
combinations.
* Statistically significant at .05 or less.
** Statistically significant at .01 or less.
180 JEFFREY T. ULMER
(Visher et al. 1991), though it may or may not reflect criminal behavior. On the
other hand, probation revocation is also a measure of criminal justice agencies’
surveillance and control activities (Albonetti and Hepburn 1997). In particular,
if intermediate sanctions such as work release or house arrest are associated
with comparatively high levels of revocation (and thus potential incarceration),
this may indicate the kind of “net strengthening” effect noted by Austin and
Krisberg (1 982).
The model in Table 4 examines the net effects on revocation of the sen-
tencing categories (incarceration, work release, community corrections) that can
be combined with probation. The reference category is traditional probation only.
This model includes only the 489 offenders who received probation as a part of
their sentence (minus 12 deleted for missing data). In terms of the non-
sentencing factors, age, sex, offense severity level, and offense type are all
significantly related to probation revocation. A decreased likelihood of probation
revocation is found among females, older offenders, and offenders convicted of
crimes with a more severe statutory grading. On the other hand, property
offenders are almost two-and-a-half times more likely to have their probation
revoked, and revocation is also moderately more likely for offenders who are
ordered into mandatory drug or alcohol treatment by the court. In contrast to the
analyses of rearrest and rearrest severity, prior record is not significantly associ-
ated with revocation.
Neither incarceration nor work release is significantly associated with the
likelihood of revocation, but house arrest is associated with considerably greater
revocation odds. Those offenders whose probation follows or runs concurrent
with a house arrest sentence have 2.25 times greater odds of probation revo-
cation, and a 69% probability of revocation. This finding can be explained by the
greater level of surveillance to which offenders in house arrest are subject,
compared to other offenders on probation. Given this more intensive level of
surveillance, it is likely that when community corrections offenders do violate
their conditions of probation, their violations are more likely to be detected and
dealt with compared to offenders under less surveillance-a pattern corroborated
in my qualitative data from interviews with house arrest staff and field observa-
tions. This is supported by the findings of Albonetti and Hepburn (1997), who
show that probationers under more intensive supervision are much more likely to
be caught for violations and have their probation revoked.
The analyses above control for a variety of factors that might influence
recidivism apart from the types or combinations of sanctions offenders receive,
and the correlations in Appendix B shows that the sentence types are far from
homogeneous in terms of the characteristics of the offenders who receive them.
Even so, it is important to rule out the possibility that the relationships between
the recidivism variables and the various sentence options described above are
Table 4
Logistic Regression Models for Probation Revocation (Probation Cases Only)
Odds (Probability)
-
"White is the reference for race/ethnicity. Single is the reference for marital
status. "Other" offenses are the reference for offense category combina-
tions. Probation is the reference for basic sentence categories and sentence
combinations.
* Statistically significant at .05 or less.
** Statistically significant at .01 or less.
182 JEFFERY T.ULMER
spurious due to selection bias. In other words, perhaps only those offenders least
likely to be rearrested are sentenced to house arrest in the first place. This would
render spurious the finding that those whose sentence includes a house arrest
component are least likely to be rearrested and tend to be rearrested for less
serious offenses. This possibility is unlikely for two reasons.
First, qualitative interview data from county judges, prosecutors, defense
attorneys, and house arrest staff indicate that offenders seen as less problematic
are sentenced to traditional probation, rather than house arrest, work release, or
especially incarceration. Instead, house arrest was reserved for either those
offenders whose offense severity or prior record precluded traditional probation,
or those offenders seen as unable to handle the comparative lack of structure and
supervision of traditional probation. Work release and incarceration, in turn,were
reserved for offenders who committed even more serious offenses, had more ex-
tensive prior records, and/or posed greater risks. Based on these descriptions of
the local sentencing decisions, one would expect results opposite those in Tables
2 and 3. If the results were due to selection bias from the sentencing decision, it
would favor offenders receiving only traditional probation (the reference cate-
gory in Tables 2 and 3), and not tKe other types and combinations of sentences.
Second, Table 5 shows the results of analyses that replicate the full models
in Tables 2-4, but also control for selection bias stemming from the sentencing
decision, using the two-step procedure commonly used to control for selection
bias (Berk 1983; Peterson and Hagan 1984).
Appendix C shows a logistic regression model of the likelihood of receiv-
ing house arrest, and one of the likelihood of being incarcerated. These models
were used to generate predicted values for the "hazard" of receiving one or the
other type of sentence. These predicted hazards, in turn, were included as pre-
dictors in models of rearrest, rearrest severity, and probation revocation that
replicated those in Tables 2 4 , but that controlled for the probability of receiving
a house arrest sentence, or being incarcerated, respectively. The results were
nearly identical between the recidivism models controlling for the probability of
receiving house arrest and the ones controlling for the probability of being in-
carcerated. Thus, Table 5 shows results for the sanction variables (other controls
included but not shown) when the probability or hazard of receiving a house ar-
rest sentence is controlled.
As the models in Table 5 show, the effects of the sentencing combinations
on rearrest and rearrest severity are almost identical to those in Tables 2 and 3.
In addition, the only sentence type significantly related to probation revocation is
house arrest, which-as in Table &more than doubles the odds of revocation.
Thus, the effects of house arrest-alone and in combination with other
sentences-on rearrest, rearrest severity, and probation revocation are apparently
robust with respect to selection bias stemming from sentencing decisions.
PROBABILITY AND SEVERITY OF RECIDIVISM 183
Table 5
Effects of Sentences on Rearrest, Rearrest Severity, and Probation Revocation:
Net of Non-Sentencing Control Variables, and Controlling for Sentencing
Selection Bias
Sentence combinations”
Incarceration only .68 (.40) -.05 (-.Ol)
Work release and probation .76 (.43) .09 ( .01)
Work release and house arrest .43 (.30) * -.40 (- .05)
Incarceration and house arrest .45 (.3 1) * -.54 (-.09)
Incarceration and probation .58 (.37) -.39 (-.07)
House arrest and probation .41 (.29) ** -.63 (-.13) *
Basic sentence typesa
Incarceration 1.37 (.58)
Work release 1.01 (.50)
House arrest 2.28 (.70) **
Selection bias correction factor 999.0 (-) .05 (.003) 188.1 (-)
Chi-squared 78.0 (p < .0001) 85.0 (p < .OOOl)
Gamma (predicted vs. observed) .44 .50
N 516 489
F 5.5 (p < .0001)
R-squared .16
“White is the reference for race/ethnicity. Single is the reference for marital status.
“Other” offenses are the reference for offense category combinations. Probation is
the reference for basic sentence categories and sentence combinations.
bThe logit coefficients for the selection bias correction factor in the models above are:
Rearrest model, 10.8; Revocation model, 5.2.
* Statistically significant at .05 or less.
** Statistically significant at .01 or less.
Finally, I tested interaction terms for each possible combination of race and
combined sanction type and gender and combined sanction type (e.g., race *
house arrest/probation, gender * incarceratiodprobation, etc.) on rearrest,
rearrest severity, and probation revocation. Only one of these kinds of interaction
effects was significant-gender * incarceration with house arrest (logit = -2.14,
odds = .12, p < .01). That is, the significant effect of incarceration followed by
184 JEFFERY T. ULMER
house arrest on the odds of rearrest found in Table 2 pertains only to women.
Women receiving a period of incarceration followed by house arrest are signifi-
cantly less likely to be rearrested, but not men (full results available on request).
Conclusions
To summarize, a stratified random sample of 528 adult felony offenders
from a medium-sized Indiana county was selected from offenders released from
house arrest, work release, supervised adult probation, and the Department of
Corrections between 1991 and 1995. The offenders in the sample varied in their
demographic characteristics, offense types and severity levels, prior criminal
records, sentences and other characteristics.
First, prior record, sex, offense type and education were consistently and
significantly related to the probability and severity of rearrest-findings gener-
ally consistent with previous research. Prior record was the strongest predictor of
rearrest and its severity-ffenders with more extensive prior records were more
likely to be rearrested, and tended to be rearrested for more severe offenses.
Women, on the other hand, were less likely to be rearrested, and if they were, it
tended to be for less serious offenses. Property offenders and autokaffic offend-
ers also tended to have a higher likelihood of rearrest, and tended to be rearrested
for more severe offenses. In addition, more educated offenders were less likely
to be rearrested, and when they were, it tended to be for less serious offenses.
The sentence combination associated with the least likelihood of rearrest
was house arredprobation. The combinations of work releasehouse arrest and
incarcerationlhouse arrest were also significantly associated with decreased
chances of rearrest compared to traditional probation. Of the sentence combina-
tions, only house arrest/probation was significantly related to rearrest. When
those sentenced to house arrestiprobation were rearrested, it tended to be for less
serious offenses. In fact, no matter which other sentence option it was paired
with, house arrest was associated with reduced chances of rearrest and less
rearrest severity. Furthermore, whenever any other sentence option was paired
with house arrest, that sentence combination was associated with reduced
chances and severity of rearrest (with the community corrections/probation
combination having the least odds and severity of rearrest). When sentences such
as incarceration, work release, and traditional probation were not paired with
house arrest in these data, those sentences exhibited no significant relationship to
rearrest or its severity. However, house arrest also considerably increased the
likelihood of probation revocation.
Some important cautions in interpreting these findings are in order. Rearrest
and its severity and probation revocation are not only measures of offender
behavior, but also of the activities of police, house arrest surveillance officers,
and probation officers. This is especially crucial to keep in mind for the probation
PROBABILITY AND SEVERITY OF RECIDIVISM 185
revocation analyses. The results for the analysis of house arrest violations and
probation revocations may reflect both the behavior of offenders and the amount
of attention and watchfulness accorded offenders under different types of correc-
tional supervision. These are factors that are impossible to fully disentangle in
these data. Relatedly, it is also important to remember that rearrests and proba-
tion revocations may represent very different offender behaviors. While rearrests
represent police definitions of criminal behavior, violations and revocations may
or may not involve criminal behavior. Instead, violations and revocations may in-
volve non-criminal behavior that violates program rules. Thus, the different re-
cidivism variables reflect potentially different levels of seriousness of behavior.
What is it about house arrest in this county that might explain the lower
odds and severity of rearrest? House arrest is the program in this county that puts
the offender in touch with the most opportunities and resources for rehabilitative
services (such as substance abuse counseling and workshops, sex offender coun-
seling, anger management classes, and other services) and also directly channels
offenders into work or education. To be eligible for and remain on house arrest,
an offender must be employed, actively seeking employment, or going to school.
Education is a particularly attractive option in this county’s program, since it can
fulfill the program’s employment requirement, and since there are a number of
local low-cost or publicly subsidized educational opportunities (e.g., county sub-
sidized G.E.D. programs, a local community college, a large state university).
None of the other types of sentences or programs (including work release) pro-
vides these resources and opportunities to such an extent.
Given these characteristics of this county’s house arrest program vis-a-vis
the other sentence options, the findings here support the contentions of Byrne
(1990), Andrews et al. (1990), and MacKenzie, Brame, McDowall, and Souryal
(1995) that intermediate sanctions of any type may potentially reduce recidivism,
but only to the extent that they include a rehabilitative emphasis. Furthermore,
these findings also support the theoretical arguments of Braithwaite (1989),
Sampson and Laub (1993, 1997), and Matsueda and Heimer (1 997) that crimi-
nal sanctions which entail structured ways to preserve, mobilize, or foster the
growth of social bonds to conventional adult roles and institutions such as work
and family are far more likely to prevent future criminal offending.
The different sanctions examined here have different rules about what con-
stitutes infractions and violations, and different levels of surveillance and formal
control. Like Albonetti and Hepburn’s (1997) findings regarding intensive pro-
bation, the data here suggest a process of “net strengthening” (Austin and
Krisberg 1982). Offenders under the heightened surveillance and control of
house arrest may or may not be more likely to violate rules. If offenders are
under the heightened surveillance and control of intermediate sanctions like
house arrest, there may be a greater likelihood that any program rule violations
186 JEFFERY T. ULMER
or criminal behavior will be caught. And, of course, offenders who violate pro-
gram rules almost always face the possibility of incarceration or reincarceration.
Thus, I found the kind of mixed outcomes for intermediate sanctions predicted
by Blomberg (1995). House arrest was associated with reduced chances of rear-
rest and its severity, but also much greater chances of probation revocation. One
very important direction of future research might be to determine whether house
arrest’s savings in correctional resources gained through preventing rearrests is
outweighed by greater levels of probation revocation, and potential subsequent
in~arceration.~ Some key empirical questions to answer are, what happens to
house arrest offenders whose probation is revoked-are they reincarcerated or
kept in the house arrest program but sanctioned some other way (e.g., given a pe-
riod on road crew, given an increased house arrest term, etc.), and how does this
vary according to the behavior in question, prior record, offenders’ social sta-
tuses, organizational program characteristics, or other factors?
We have a good deal more to learn about the comparative and conditionalout-
comes of traditional sanctions on one hand, and intermediate sanctions on the
other. The general recidivism literature contains several studies that examine both
the probability and timing of recidivism using sophisticated event history models,
but these studies typically do not include or compare a variety of intermediate
sanctions with more traditional sentences. The literature on recidivism and inter-
mediate sanctions seldom investigates the timing of recidivism, and some studies
do not even contain multivariate models of the probability of recidivism. Further,
both literatures often rely on offender samples that lack variation in key character-
istics. For example, many focus only on limited types of offenses, some offender
samples lack variation in criminal history, and many are limited only to males.
Future research needs to put all of these pieces together by investigating the prob-
ability and timing of a variety of recidivism measures, comparing a wide variety
of newer and more traditional sentence types, and using samples of offenders that
vary in their offenses, prior records, sex, race and ethnicity, and social resources.
As mentioned earlier, with its plentiful jobs, education opportunities, and its
comparatively well-funded, supported, and organized intermediate sanction pro-
grams like house arrest, this particular county would seem to have a social struc-
tural context which favors the success of community corrections. Thus, a crucial
direction for future research would be further examine the relationships between
intermediate sanctions, traditional sentencing options, and the probability and
timing of recidivism, to do so across a variety of jurisdictions with different con-
textual characteristics, and across intermediate sanction programs with different
organizational characteristics.
Further, research should not only continue to quantitatively analyzerecidivism
outcomes or series of outcomes, but should also qualitatively explore the processes
by which intermediate sanctions-like house arrest for example-may or may not
PROBABILITY AND SEVERITY OF RECIDIVISM 187
prevent recidivism (see also Blomberg 1995). Future research should focus on the
situational and interactionalprocesses by which differentkinds of sanctions affect-
the theoretical mechanisms discussed by Braithwaite (1989), Sampson and Laub
(1993,1997) and Matsueda and Heimer (19 9 7 F t h e balance of criminal and con-
ventional opportunities, the nurturing vs. or “knifing off” of conventional commit-
ments and social bonds, and criminal vs. conventional self-definitions.
APPENDIX A
Variables and Their Coding
Variable Coding
APPENDIX A (continued)
Variable Coding
.35** .OI -.04 -.II* -.II*' .01 .lo* -.W* .03 -.01 -.W* .23** .11** .04 -.IS** -.12** .07 .04 .OL .05 -.11' .02
.30" .01 -.01 -.13'* -.W' .02 .06 -.lo*' .04 .01 -.lo' 26'. .09* .04 -.IS'* -.II*' .11** .05 .04 .08 -.08' .01
- .M -.13** -.06 -.01 .08 -.14* -.03 .02 -.05 -.W* .13** .13** .01 -.13** -.13*' .06 .IS* .01 .07 .2l" .14**
- -.W -.05 -.05 .03 .I0 -.02 .02 -.01 -.01 -.06 -.01 -.03 -.04 .04 .02 .01 .02 -.W -.18** -.07
- .OL .17** .20'* .04 .01 .01 -.05 -.07 .01 -.24*' -.02 .02 .16*' .18** .01 .05 -.01 -.05 -.09*
- .01 .01 .04 .03 -.05 -.04 -.12** -.08 .07 -.IS** .21'* -.16** .01 .02 -.21** -.07 .03 .12**
- -.91** -.33*' -.02 -.05 -.04 -.03 .02 -.M .01 .II" .02 -.06 .03 .M .01 .12** -.12**
- -.35** -.05 -.03 -.04 .08* .09* -.01 .03 -.02 .11** -.07 .07 .09* -.01 .03 -.06
- .04 .02 -.01 -.20*' -.W -.M -.01 .II" -.11" .19* -.08 -.20*' -.06 -.27*' .IF*
- .01 -.OS .03 -.05 .01 .01 .07 -.05 -.01 .05 -.04 .02 .02 .I0
- -.05 -.02 .02-.01 .21** -,OS -.02 -.W* .02 .lo* .07 -.lI* .02
- -.01 -.02 -.03 .04 .01 -.04 .05 .03 -.03 .01 .03 -.02
- -.W* -.03 .04 .01 -.04 -40" -.01 .27** -.02 -.05 -.07
- .08* .Ol -.08 -.13* .06 -.01 .lo* .03 .08 -.04
- -.26** -.25*' -.17** -.37** -.20*' -.M -.04 -.05 -.04
- -.17*' -.12** -.2S** .01 .08 -.Ol -.I2** -.04
- -.11** -.24** .08* .lo* -.02 .05 .07
- -.17** .01 .12** -.07 -.04 -.IS**
- .17** -.lo* .11** .14** .II*
- .16** .13" .16** 29"
- .39" -.22** -.21"
- -.02 .09*
- .23**
APPENDIX C
Logistic Regression Models of Probability of Receiving House Arrest and Incarceration Sentences
Odds (Probability)
House Arrest Incarceration
aWhite is the reference for race/ethnicity. Single is the reference for marital status. “Other” offenses
are the reference for offense category combinations.
* Statistically significant at .05 or less.
** Statistically significant at .01 or less.
ENDNOTES
An earlier version of this paper was presented at the 1998 annual meetings of the American
Society of Criminology, Washington, D.C. This research was funded by the Indiana Department of
Corrections, but the author is solely responsible for this article’s arguments. I thank David Kuebler,
Kristi Aufderheide, Amy Crook, Amy Earlywine, Janet Ahlfield, and Danielle Hsu for their efforts on
this research project.
‘These findings coincide with MacKenzie et a1.k (1995) research, which found that boot camps
exhibited decreased recidivism rates only to the extent that they included a rehabilitative or educa-
tional focus.
20f course, offenders could commit new criminal offenses while in jail or prison. I do not in-
clude such offenses. Incarcerated offenders are under much greater supervision surveillance than
those under other sanctions, and have much less opportunity to commit new criminal offenses than
PROBABILITY AND SEVERITY OF RECIDIVISM 191
those in the community. Thus, the time “at risk” for incarcerated offenders and those not incarcerated
is not strictly comparable.
3By state law, anyone who wants to view an offender’s presentence report must have that of-
fender’s written permission, and securing such permission from 528 offenders was impractical.
4This weighted prior record scale is very similar to those used by many states’ sentencing
guideline systems, such as Pennsylvania, Minnesota, and Washington (Kramer, Lubitz, and
Kempinen 1989).
5The cumulative effect on the odds for a given value of a continuous independent variable is
calculated by raising the odds ratio for that variable to a power equal to the value of interest. For ex-
ample, the cumulative odds of rearrest for a hypothetical offender with a prior record score of “4”
would be 1.34 = 2.86.
6The formula for converting odds ratios to probabilities is: odds / (odds + 1) (Lichter 1989).
’1 am grateful to Tom Bernard for suggesting this point.
REFERENCES
Agresti, Alan, and Barbara Finlay. 1986. Statistical Methods for Social Scientists. San Francisco:
Dellen.
Albonetti, Celesta, and John Hepburn. 1997. “Probation Revocation: A Proportional Hazards Model
of the Conditioning Effects of Social Disadvantage.” Social Problems 44(1):124-38.
Aldnch, John, and Forrest Nelson. 1984. Linear Probability, Logit, and Probit Models. Newbury
Park, C A Sage.
Andrews, D. A., Ivan Zinger, Robert Hoge, James Bonta, Paul Gendreau, and Francis Cullen. 1990.
“Does Correctional Treatment Work? A Clinically Relevant and Psychologically Informed
Meta-Analysis.” Criminology 28(3):369-404.
Austin, James, and Barry Krisberg. 1982. “The Unmet Promise of Alternatives to Incarceration.”
Crime and Delinquency 28:374409.
Baumer, Eric. 1997. “Levels and Predictors of Recidivism: The Malta Experience.” Criminology
35(4):601-28.
Becker, Howard. 1963. Outsiders. New York Macmillan.
Berk, Richard. 1983. “An Introduction to Sample Selection Bias in Sociological Data.” American
Sociological Review 48:38&98.
Berliner, Lucy, Donna Schram, Lisa Miller, and Cheryl Darling Milloy. 1995. “A Sentencing
Alternative for Sex Offenders.” Journal of Interpersonal Violence 10(4):487-502.
Blomberg, Thomas. 1995. “Beyond Metaphors: Penal Reform as Net Widening.” Pp. 4 5 4 1 in
Punishment and Social Control, edited by T. Blomberg and S. Cohen. New York Aldine.
Blumstein, Alfred. 1993. “Making Rationality Relevant.” Criminology 31(1):1-16.
Braithwaite, John. 1989. Crime, Shame, and Reintegration. Cambridge, UK: Cambridge University
Press.
Byme, James. 1990. “The Future of Intensive Probation Supervision and the New Intermediate
Sanctions.” Crime and Delinquency 36(1):M1.
Cole, George, and Christopher Smith. 1998. The American System of Criminal Justice. Monterey,
CA: Brooks/Cole.
Courtwright, Kevin, Bruce Berg, and Robert Mutchnick. 1997. “Effects of House Arrest with
Electronic Monitoring on DUI Offenders.” Journal of Offender Rehabilitation 24(3-4):35-5 1.
DeJong, Christina. 1997. ‘‘Survival Analysis and Specific Deterrence: Integrating Theoretical and
Empirical Models of Recidivism.” Criminology 35(4):561-76.
192 JEFFERY T. ULMER
Feeley, Malcolm, and Jonathan Simon. 1992. “The New Penology: Notes on the Emerging Strategy
of Corrections and Its Implications.” Criminology 30(4):449-74.
Geerken, Michael, and Hennessey Hayes. 1993. “Probation and Parole: Public Risk and the Future
of Incarceration Alternatives.” Criminology 3 1(4):54964.
Gendreau, Paul, Tracy Little, and Claire Goggin. 1996. “A Meta-Analysis of the Predictors of Adult
Offender Recidivism: What Works!” Criminology 34(4):575608.
Glaser, Daniel, and Ronald Watts. 1992. “Electronic Monitoring of Drug Offenders on Probation.”
Judicature 76(3):112-17.
Goffman, Erving. 1963. Stigma: Notes on the Management of Spoiled Zdentity. Englewood Cliffs, NJ:
Prentice-Hall.
Gottfiedson, Stephen, and Don Gottfiedson. 1994. “Behavioral Prediction and the Problem of
Incapacitation.” Criminology 32(3):441-74.
Hagan, John. 1992. “The Poverty of a Classless Criminology.” Criminology 30(1): 1-20.
Hagan, John, and Albert0 Palloni. 1990. “The Social Reproduction of a Criminal Class in Working
Class London: 1950-1 980.” American Journal of Sociology 92(2):265-99.
Heimer, Carol, and Lisa Staffen. 1995. “Interdependence and Reintegrative Social Control.”
American Sociological Review 60(5):635-54.
Homey, Julie, D. Wayne Osgood, and Ineke Haen Marshall. 1995. “Variability in Crime and Local
Life Circumstances.” American Sociological Review 60(5):655-73.
Jones, Peter. 1990. “Community Corrections in Kansas: Extending Community Based Corrections or
Widening the Net?” Journal of Research in Crime and Delinquency 27(1):79-101.
Kramer, John H., Robin Lubitz, and Cynthia Kempinen. 1989. “Sentencing Guidelines: A
Comparison of Sentencing Policies in Minnesota, Pennsylvania, and Washington.” Justice
Quarterly 6565-87.
Land, Kenneth, Patricia McCall, and Karen Parker. 1994. “Logistic vs. Hazards Regression Analyses
in Evaluation Research.” Evaluation Review 18:411-37.
Lemert, Edwin. 1981. “Issues in the Study of Deviance.” The Sociological Quarterly 22(2):285-305.
. 1951. Social Pathology. New York McGraw-Hill.
Lichter, Daniel. 1989. “Race, Employment Hardship, and Inequality in the Nonmetropolitan South.”
American Sociological Review 54(3):43646.
Little, Roderick, and Nathaniel Schenker. 1995. “Missing Data.” Pp. 39-75 in Handbook of
Statistical Modeling for the Social and Behavioral Sciences, edited by G . Arminer, C. Clogg,
and M. Sobel. New York Plenum Press.
MacKenzie, Doris, Robert Brame, David McDowall, and Claire Souryal. 1995. “Boot Camp Prisons
and Recidivism in Eight States.” Criminology 33(3):327-58.
Makkai, Toni, and John Braithwaite. 1991. “CriminologicalTheories and Regulatory Compliance.”
Criminology 29(2): 191-220.
Matsueda, Ross. 1992. “Reflected Appraisals, Parental Labeling, and Delinquency: Specifjmg a
Symbolic Interactionist Theory.” American Journal of Sociology 97(6): 1577611.
Matsueda, Ross, and Karen Heimer. 1997. “Developmental Theories of Crime.” Pp. 17&86 in
Advances in Criminological Theov, edited by T. Thornberry. New Brunswick, N J Transaction.
Morns, W., and Michael Tonry. 1990. Between Prison and Probation: Intermediate Punishments in
a Rational Sentencing System. Oxford Oxford University Press.
Musheno, Michael, Dennis Palumbo, Steven Maynard-Moody, and James Levine. 1989.
“Community Corrections as an Organizational Innovation: What Works and Why.” Journal of
Research in Crime and Delinquency 26(2): 136-67.
Parker, Robert Nash. 1998. “Alcohol and Homicide in the United States, 1934-1995.” National
Institute of Justice Journal October: 14-5.
PROBABILITY AND SEVERITY OF RECIDIVISM 193
Petersilia, Joan. 1995. “A Crime Control Rationale for Reinvesting in Community Corrections.”
Prison Journal 75(December):479-96.
Petersilia, Joan, and Susan Turner. 1993. “Intensive Probation and Parole.” In Crime and Justice: An
Annual Review of Research, edited by M. Tonry. Chicago: University of Chicago Press.
. 1990. “Intensive Supervision for High-Risk Probationers: Findings from Three California
Experiments.” Santa Monica, CA: RAND.
Peterson, Ruth D., and John Hagan. 1984. “Changing Conceptions of Race: Towards an Account of
Anomalous Findings of Sentencing Research.” American Sociological Review 49:56-70.
Rhodes, William. 1986. “A Survival Model With Dependent Competing Events and Right Hand
Censoring: Probation and Parole as an Illustration.” Journal of Quantitative Criminology
2: 113-37.
Sampson, Robert, and John Laub. 1997. “A Life-Course Theory of Cumulative Disadvantage and the
Stability of Delinquency.” Pp. 133-61 in Developmental Theories of Crime and Delinquenq,
edited by T. Thornberry. New Brunswick, N J Transaction.
. 1993. Crime in the Making: Pathways and Turning Points Through Life. Cambridge, MA:
Harvard University Press.
Schur, Edwin. 1980. The Politics of Deviance: Stigma Contests and the Uses of Power: Englewood
Cliffs, NJ: Prentice-Hall.
Sherman, Lawrence. 1993. “Defiance, Deterrence, and Irrelevance: A Theory of the Criminal
Sanction.” Journal of Research in Crime and Delinquency 30(4):445-73.
Skolnick, Jerome. 1995. “What Not To Do About Crime.” Criminology 33(1):1-16.
Smith, Linda, and Ronald Akers. 1993. “A Comparison of Recidivism of Florida’s Community
Control and Prison: A Five-Year Survival Analysis.” Journal of Research in Crime and
Delinquency 30(3):267-92.
Spelman, William. 1995. “The Severity of Intermediate Sanctions.” Journal of Research in Crime
and Delinquency 32(2): 107-35.
Stebbins, Robert. 1971. Commitment to Deviance. Westport, CT: Greenwood.
Steffensmeier, Darrell, and Emilie Allan. 1996. “Criminal Behavior: Gender and Age.” In
Criminolofl, edited by J. Sheley. Belmont, CA: Wadsworth.
Tannenbaum, Frank. 1938. Crime and the Community New York Columbia University Press.
Turner, Susan, and Joan Petersilia. 1998. “Work Release: Recidivism and Corrections Costs in
Washington State.” NIJ Research in Brief Alternatives to Incarceration January/February:
10-11.
Ulmer, Jeffery. 1994. “Revisiting Stebbins: Labeling and Commitment to Deviance.” The
Sociological Quarterly 35(1):135-57.
Visher, Christy, Pamela Lattimore, and Richard Linster. 1991. “Predicting the Recidivism of Serious
Youthhl Offenders Using Survival Models.” Criminology 29(3):329-66.