Rehabilitative Services

Examining the Links Between Therapeutic Jurisprudence and Mental
Health Court Completion

Allison D. Redlich and Woojae Han
University at Albany, State University of New York

Research demonstrates that mental health courts (MHCs) lead to improved outcomes compared to
traditional criminal court processes. An underlying premise of MHCs is therapeutic jurisprudence (TJ).
However, no research, to our knowledge, has examined whether MHC outcomes are predicted by TJ
principles as theorized. In the present study, we examined whether principles measured at the onset of
MHC enrollment (knowledge, perceived voluntariness, and procedural justice) predicted MHC comple-
tion (graduation). Using structural equation modeling with MHC participants from four courts, a
significant, direct relationship between TJ and MHC completion was found, such that higher levels of TJ
were associated with higher rates of success. Although this direct effect became nonsignificant when
mediator variables were included, a significant indirect path remained, such that increased levels of initial
perceived voluntariness and procedural justice, and MHC knowledge, led to decreased rates of new
arrests, prison, MHC bench warrants, and increased court compliance, which, in turn, led to a higher
likelihood of MHC graduation.

Keywords: mental health courts, therapeutic jurisprudence, procedural justice, structural equation mod-
eling

The emergence of mental health courts (MHCs) in the last 15
years is an important new development in mental health law
(Petrila, 2003). MHCs are specialty criminal courts with dockets
usually exclusive to individuals with mental illnesses. MHCs,
developed as one among many strategies designed to stop the
revolving door of repeated cycling through the criminal justice
(CJ) system, are recognized as being much less formal than tradi-
tional adversarial courts (Casey & Rottman, 2003; Petrila, 2003).
In some ways, MHCs adhere to the risk–need–responsivity model
(see Andrews & Bonta, 2010) in that the courts typically identify
those most in need, develop individualized treatment needs, and
ultimately aim to decrease recidivism.

An underlying premise of MHCs is therapeutic jurisprudence
(TJ; Lurigio & Snowden, 2009; Wiener, Winick, Georges, &
Castro, 2010; Winick, 2002; Winick & Wexler, 2003). TJ is an
interdisciplinary approach that is concerned with how the law and
legal actors can be therapeutic change agents. Winick (2002) stated
that “problem solving courts often use principles of therapeutic
jurisprudence to enhance their functioning” (p. 1064), and John-
ston (2012) stated that “therapeutic jurisprudence is the most

popular justification for mental health courts” (p. 525). The theory
of TJ is specific in outlining that voluntary choice (in this case, into
the MHC) and appreciation of requirements are important to
promoting behavioral change. For example, in addressing how
problem solving court judges can be therapeutic change agents,
Winick (2002) emphasizes the TJ principles of avoiding paternal-
ism and respecting autonomy. He states, “If individuals who make
their own choices perceive them as noncoerced, they will function
more effectively and with greater satisfaction” (p. 1072). Addi-
tionally, according to the TJ model of problem solving courts
devised by Wiener et al. (2010), clients’ perceptions of TJ-relevant
concepts, such as procedural and distributive justice, and reinte-
grative shaming, serve to influence perceptions of the legitimacy
of law, which, in turn, leads to improved outcomes (case, mental
health, family and academic success, etc.).

Despite the important link between TJ and MHCs, no research,
to our knowledge, has examined whether improved court outcomes
are predicted by TJ principles. However, there is a smattering of
research on this link in other specialty courts. For example, Got-
tfredson, Kearley, Najaka, and Rocha (2007) examined the impact
of drug-court participation on perceptions of procedural justice,
and then on recidivism outcomes. They found that “participation in
the DTC [drug treatment court] increases the number of judicial
hearings attended, which directly reduces the variety of drugs used
and also reduces the variety of crimes committed by increasing
perceptions of procedural justice” (p. 26). Thus, Gottfredson et al.
found an indirect link such that participation in the drug court led
to increased feelings of procedural justice (via hearings attended),
which, in turn, led to fewer varieties of crimes committed. How-
ever, perceptions of procedural justice did not directly influence
the outcomes of drug variety and multiple drug use. Similarly,
Hepburn and Harvey (2007) did not find the threat of sanction in
drug court to influence success in the court. More specifically,

This article was published Online First June 17, 2013.
Allison D. Redlich, School of Criminal Justice, University at Albany,

State University of New York; Woojae Han, School of Social Welfare,
University at Albany, State University of New York.

We thank Dr. Henry J. Steadman and Policy Research Associates, as
well as Professor John Monahan and the John D. and Catherine T. Mac-
Arthur Foundation Network on Community Mandated Treatment, for gen-
erously supporting the research.

Correspondence concerning this article should be addressed to Allison
D. Redlich, School of Criminal Justice, University at Albany, SUNY, 135
Western Avenue, Albany, NY 12222. E-mail: [email protected]

T
hi

s
do

cu
m

en
t

is
co

py
ri

gh
te

d
by

th
e

A
m

er
ic

an
P

sy
ch

ol
og

ic
al

A
ss

oc
ia

ti
on

or
on

e
of

it
s

al
li

ed
pu

bl
is

he
rs

.
T

hi
s

ar
ti

cl
e

is
in

te
nd

ed
so

le
ly

fo
r

th
e

pe
rs

on
al

us
e

of
th

e
in

di
vi

du
al

us
er

an
d

is
no

t
to

be
di

ss
em

in
at

ed
br

oa
dl

y.

and Human Behavior © 2013 American Psychological Association
2014, Vol. 38, No. 2, 109 –118 0147-7307/14/$12.00 DOI: 10.1037/lhb0000041

109

mailto:[email protected]

http://dx.doi.org/10.1037/lhb0000041

Hepburn and Harvey studied two drug courts, one that threatened
a 120-day jail sentence for noncompliance and one that was
prohibited from imposing jail as a sanction. They found clients
from these two drug courts did not differ significantly on average
length of time in the court, the proportion who remained in the
court, or the proportion who successfully completed the program.
Thus, insofar as threat of jail as a sanction is an aspect of TJ,
Hepburn and Harvey did not find support for the theorized link
between TJ and later court outcomes.

Overall, little research has been conducted on the predictive
relationship between therapeutic constructs and future court suc-
cess. This paucity of research is surprising, given that (a) the
theory of TJ expects such a relationship, and (b) MHCs (as well as
other specialty courts) are premised on TJ (see Lurigio &
Snowden, 2009; Winick, 2002). The research on other problem-
solving courts, although scant and therefore preliminary, does not
show a readily recognizable relationship between TJ and success.

MHC Success

Success in MHC can be defined in a multitude of ways, includ-
ing not getting rearrested, reducing jail stays, becoming engaged
with community treatment, graduating, and generally complying
with court orders. The purpose of this article is to address whether
knowledge and perceived choice and procedural justice, principles
of TJ, measured at the beginning of MHC enrollment is predictive
of later MHC successful completion (graduation). There is consis-
tent, if not robust, evidence that MHCs lead to improved outcomes
in comparison with traditional criminal court processing. Most of
the outcome research on MHCs has focused on recidivism (for an
overview, see Sarteschi, Vaughn, & Kim, 2011); this corpus of
research has determined that MHC participants have reduced ar-
rests and jail days compared with offenders with mental illness not
diverted through such courts (Moore & Hiday, 2006; Steadman,
Redlich, Callahan, Robbins, & Vessilinov, 2011). However, within
MHCs, not all participants succeed, as some are terminated from
the court, some are arrested on new charges, and some abscond. In
a four-site study of MHCs, termination rates varied from a low of
17% to a high of 47% (Redlich, Steadman, et al., 2010). Thus,
although overall MHCs have shown to be superior compared with
treatment as usual, within MHCs, some defendants do not succeed.

What predicts MHC success? This has been a driving question
of several research projects. As mentioned, most research has
examined recidivism as a main measure of success. For example,
via meta-analysis, Sarteschi et al. (2011) found a �0.54 effect size
in favor of MHCs reducing recidivism over that of comparison
samples. Although MHC participants have been found to have
lower recidivism rates and fewer jail days than similarly situated
offenders processed traditionally, criminogenic factors (such as
prior arrest history) tended to be more robust predictors of MHC
outcomes in comparison with demographic and clinical factors
(Steadman et al., 2011). However, though perhaps less robust,
noncriminogenic factors, such as severity of mental health prob-
lems (see Andrews & Bonta, 2010), have also been found to
influence the likelihood of recidivism and reincarceration among
MHC clients (e.g., Herinckx, Swart, Ama, Dolezal, & King, 2005;
Steadman et al., 2011).

Another factor found to predict recidivism is length of involve-
ment in the MHC, such that those who receive the full “dose” of

the MHC (i.e., graduate) are less likely to be arrested in compar-
ison with those not in the MHC and those who spent less time in
the court (Herinckx et al., 2005; Hiday & Ray, 2010; McNiel &
Binder, 2007). But what predicts completion status? In one study
of four MHCs, demographic and criminal factors, and data collec-
tion site, did not significantly predict MHC completion status
(Redlich, Steadman, et al., 2010). Rather, the only factor to sig-
nificantly predict completion status was perceived compliance
with judicial/court orders (as rated by court personnel), although
compliance itself was predicted by being White, having less seri-
ous arrest charges, and data collection site. Dirks-Linhorst, Kon-
drat, Linhorst, and Morani (2011) found that being male and
African American raised the likelihood of MHC termination.

Less research has been conducted on treatment-related out-
comes, but the research, in general, demonstrates that MHC par-
ticipants receive more treatment than comparison samples
(Boothroyd, Poythress, McGaha, & Petrila, 2003; Trupin & Rich-
ards, 2003). Luskin (in press) determined that MHC participants in
the Marion County, Indiana, court received more outpatient treat-
ment services than a treatment-as-usual matched sample. How-
ever, she also found that demographic (age, gender, and race) and
criminogenic (prior arrest history and current charge types) factors
were not significant predictors of treatment receipt 6 months after
enrolling in the court or being arrested; rather, having a diagnosis
of major depression (in comparison with diagnoses of schizophre-
nia or bipolar disorder) was influential.

Overall, the research demonstrates that MHC participants fare
better than those offenders in comparison samples who were
processed in the traditional CJ system. At the same time, within
MHC samples, some succeed and some do not. The factors found
to predict MHC “success” have not been always been consistent.
On the one hand, demographic factors of the MHC participants
themselves generally do not influence MHC success (but see
Dirks-Linhorst et al., 2011). On the other hand, other predictive
factors are found to be less consistent, often depending upon the
outcome measure of success being studied. For example, crimino-
genic factors, such as extent of prior arrests and severity of current
charges, have been found to be quite influential when the outcome
under examination is new arrests and perceived compliance in the
court, but have not been found to directly influence the outcomes
of MHC graduation or receipt of treatment in the community
(Dirks-Linhorst et al., 2011; Redlich, Steadman, et al., 2010).

MHC Comprehension

A major component of MHCs is that they are voluntary. Indeed,
if entry into the courts was not voluntary, arguably, the courts
would violate basic notions of equality under the law (Seltzer,
2005). And although there is no explicit requirement that the
decision to enroll in a MHC be knowledgeable per se, there is an
explicit requirement that legal decision making in general (includ-
ing the decision to plead guilty, which most MHCs require) be
knowing and intelligent (as well as voluntary). As such, in “The
Ten Essential Elements of Mental Health Courts,” Thompson,
Osher, and Tomasini-Joshi (2008) recommend that defendants
provide informed consent (see also Linhorst et al., 2010; Redlich,
2013). Specifically, Element Five states,

Defendants fully understand the program requirements before agree-
ing to participate in the MHC. They are provided legal counsel to

T
hi

s
do

cu
m

en
t

is
co

py
ri

gh
te

d
by

th
e

A
m

er
ic

an
P

sy
ch

ol
og

ic
al

A
ss

oc
ia

ti
on

or
on

e
of

it
s

al
li

ed
pu

bl
is

he
rs

.
T

hi
s

ar
ti

cl
e

is
in

te
nd

ed
so

le
ly

fo
r

th
e

pe
rs

on
al

us
e

of
th

e
in

di
vi

du
al

us
er

an
d

is
no

t
to

be
di

ss
em

in
at

ed
br

oa
dl

y.

110 REDLICH AND HAN

inform this decision, and subsequent decisions about program in-
volvement. Procedures exist in the MHC to address, in a timely
fashion, concerns about a defendant’s competency whenever they
arise (p. 5).

Although recommended and legally required, research indicates
that significant minorities of MHC defendants are not making
informed and voluntary enrollment decisions (Boothroyd et al.,
2003; Poythress, Petrila, McGaha, & Boothroyd, 2002).

Recently, Redlich, Hoover, Summers, and Steadman (2010)
examined awareness of the voluntary nature (making a voluntary
decision, but also knowledge that the choice is voluntary) and the
level of understanding of court procedures and requirements
among newly enrolled MHC participants in two courts. Most
(66 % to 71%) said it was their choice to enroll in the court, but at
the same time, about 60% claimed not to have been told the
decision to enroll in the court was voluntary or told of the require-
ments prior to enrolling. Indeed, voluntary awareness and per-
ceived voluntariness to enroll were found to be related (see also
Poythress et al., 2002). Specifically, MHC clients claiming to be
unaware of the voluntary choice perceived more coercion than
those who were aware of the choice. In a multivariate analysis,
perceptions of voluntariness were not predicted by sociodemo-
graphic traits such as gender or criminal charges. Rather, MHC
voluntariness was predicted by awareness of the voluntary nature
(as mentioned), the number of times participants had been before
the MHC judge prior to enrollment, and insight into mental health
problems.

Redlich, Hoover, et al. (2010) also found generally poor levels
of MHC-related knowledge. Whereas most MHC clients were
aware of the basics of MHCs (e.g., having to go before the judge
periodically, taking medication), many were unaware of more
nuanced information. For example, 27% to 45% did not know that
pleading guilty was a condition of enrollment, and up to 30% did
not know they could stop being in the court if they chose. Redlich,
Hoover, et al. (2010) also examined adjudicative competence,
which is legally required and presumed present. They found that
one fifth to one third had impairments in their knowledge of legal
terms and concepts, and in their ability to reason pertinent to legal
decision making. MHC knowledge was most robustly predicted by
two adjudicative competence scores (understanding and reason-
ing). However, unlike the results regarding predictors of volun-
tariness, demographic/criminogenic factors significantly predicted
knowledge. Specifically, females, persons employed at the time of
MHC enrollment, and persons charged with less serious crimes
were more knowledgeable than their counterparts.

In addition to knowledge and voluntariness, a third important
principle relevant to TJ is perceptions of procedural justice (see,
e.g., Winick, 2002). Procedural justice refers to the degree to
which defendants feel respected, feel that they have a voice and are
heard, and are treated fairly (Lind & Tyler, 1988). Participants in
MHCs have been found to have higher perceptions of procedural
justice in comparison with those processed traditionally in the CJ
system (Poythress et al., 2002). For example, when asked the
degree to which they had sufficient opportunity to talk to the judge
about their personal and legal situations, MHC participants had an
average score of 5.4 versus participants in the comparison sample
in traditional court with an average score of 1.8 (1 � not at all; 7 �
very much so).

In summary, MHCs presume voluntary and informed decision
making. Research, however, shows that significant subsets of
MHC enrollees are not engaging in such decision making. Further,
criminogenic and court-specific (such as compliance and compe-
tence), and to a lesser degree, demographic, factors have been
found to influence MHC comprehension as well as MHC out-
comes. Redlich (2005) argued that “it stands to reason that those
who are unaware that entry into this [MHC] agreement is volun-
tary and thus do not understand this agreement may be less likely
to adhere to the agreement” (p. 613). This is an empirical ques-
tion— one that is addressed in the present study.

The Present Study

Based on the literature reviewed, it is hypothesized that MHC
participants who better understood court procedures and require-
ments, chose to enter the court, and felt more respected (as as-
sessed at the time of enrollment) would be more successful (i.e.,
more likely to graduate) in the courts than those with lower
appreciation of the courts and perceptions of choice and procedural
justice. Data collected as part of the MacArthur MHC Project (see
Redlich, Steadman, et al., 2010; Steadman et al., 2011) are ana-
lyzed. The MacArthur MHC study was a comprehensive under-
taking involving participants from four MHCs, comparison sam-
ples from each of the four sites, self-report interviews conducted at
baseline entry in the court/CJ system and 6 months later, and
extensive objective outcome data collection at 12 and 18 months
after entering the court. Structural equation modeling (SEM) is
used to test our research question. SEM is the ideal method to
handle measurement errors with multiples indicators. Further,
SEM is useful for mediation analysis (e.g., Cheung & Lau, 2008;
Frazier, Tix, & Barron, 2004; Jang & Chiriboga, 2010) and in
assessing the role of a mediating variable between the predictor
and the outcome. In the present study, CJ and MHC-related factors
are investigated as mediators that can potentially explain the
relationship between TJ and MHC outcome.

Method

Participants

Participants in the MacArthur MHC study came from four
courts: San Francisco County, California (n � 108), Santa Clara
County, California (n � 136), Hennepin County, Minnesota (n �
105), and Marion County, Indiana (n � 99). All of the courts
accepted persons charged with misdemeanors and felonies, and
persons with co-occurring substance use disorders. The courts,
which were established between 1997 and 2003, were in larger
urban jurisdictions and enrolled between approximately 100 (San
Francisco and Hennepin) and 250 (Santa Clara) MHC participants
annually. Information about demographics and mental illness was
provided by the MHCs. Aggregated descriptive statistics about
participants can be found in Table 1. Among the 448 individuals in
the study, mean age was 37.45 years and over half (58.2%) were
men. Approximately half of participants were White (49.3%).
Most (89.4%) individuals had serious mental health problems.
Symptomatology was measured at entry into the court using the
Colorado Symptoms Index (CSI; Conrad et al., 2001). The average
score in the present sample was 24.5 (SD � 13.31), with a range

T
hi

s
do

cu
m

en
t

is
co

py
ri

gh
te

d
by

th
e

A
m

er
ic

an
P

sy
ch

ol
og

ic
al

A
ss

oc
ia

ti
on

or
on

e
of

it
s

al
li

ed
pu

bl
is

he
rs

.
T

hi
s

ar
ti

cl
e

is
in

te
nd

ed
so

le
ly

fo
r

th
e

pe
rs

on
al

us
e

of
th

e
in

di
vi

du
al

us
er

an
d

is
no

t
to

be
di

ss
em

in
at

ed
br

oa
dl

y.

111THERAPEUTIC JURISPRUDENCE AND MHCS

of 60 points. (We also note here that CSI scores were unrelated to
the three TJ principles used here; rs range from �.02 to .04.)
Self-reported arrests varied quite a bit, ranging from 1 to 250;
participants had an average of 14.96 arrests since the age of 15
years.

Measures

Our dependent variable was completion versus noncompletion
in the MHC. Completion was defined as graduating within our
study time frame (n � 205). Noncompletion (total n � 228) was
defined as those who were removed from the court or had dropped
out (n � 127) and those still in the court (n � 101). The study time
frame was approximately 3.75 years. To answer our research
question, seven independent measures were utilized, which were
grouped into three categories with the following labels: (a) TJ; (b)
CJ outcomes; and (c) MHC outcomes.

TJ. TJ principles included three measures: perceived volun-
tariness to enroll in the court, perceived procedural justice, and
MHC knowledge. All were measured at the outset of court enroll-
ment.

Perceived voluntariness was measured with the MacArthur Per-
ceived Coercion Scale (MPCS), adapted for MHCs (see Poythress
et al., 2002). This measure includes eight statements concerning
the choice to enroll in MHCs (e.g., “I chose to be in the mental
health court”; “I had a lot of control over whether I went to the
mental health court”). A 5-point Likert scale was used, with 0 �
strongly agree to 4 � strongly disagree. Although the label of the
scale is perceived coercion, the inverse is perceived voluntariness.
The MPCS has been used with thousands of individuals and in
numerous countries and has been found to be sensitive and inter-
nally consistent (e.g., Cronbach’s alpha � .90; Gardner et al.,
1993) and to have obtainable interrater reliability (kappas range
from .73 to 1.00; Lidz et al., 1998). In this study, Cronbach’s alpha
was 0.76.

Perceived procedural justice was measured using responses to
five items employed by Poythress et al. (2002). Specifically,
participants were asked to rate the degree to which they had a

voice, were treated with respect and treated fairly by the MHC
judge, the judge seemed genuinely interested in them, and their
satisfaction with the judge (1 � not at all to 7 � definitely). In the
present study, Cronbach’s alpha was 0.89.

MHC knowledge was measured with a scale introduced by
Redlich, Hoover, et al. (2010). The measure included three series
of “true”/“false”/“I don’t know” statements with the following
opening stems: (a) “To be a mental health court participant, people
are agreeing to. . .” (14 statements); (b) “If people do not follow
the conditions of mental health court, they can. . .” (12 statements);
and (c) “If people follow the conditions of mental health court,
they can. . .” (7 statements). The statements consisted of actual
MHC procedures, requirements, and consequences (e.g., returning
for status review hearings, having to take prescribed medications
and go to treatment appointments, not using alcohol or drugs,
being sent to jail for noncompliance, and having their original
charges or convictions dropped), as well as false statements of
non-MHC procedures, requirements, and consequences (having to
take lie detector tests, having to do physical exercise, never being
arrested again, and being charged again for the same crime). To
create one MHC knowledge score, answers on the 33 statements
were scored as correct or incorrect (“don’t know” answers were
scored as incorrect) and then summed. Cronbach’s alpha for the
present study was 0.90.

CJ outcomes. CJ outcomes included number of new arrests
and whether the MHC client had been sent to prison (“yes” or
“no”), both measured 12 months after court enrollment. As de-
scribed in Steadman et al. (2011), arrest data were obtained from
the Federal Bureau of Investigation (the National Incident Based
Reporting System). Only new arrests were included; warrants and
violations were excluded. Prison data were obtained from the
Department of Corrections in the study states (California, Indiana,
and Minnesota).

MHC outcomes. MHC outcomes included a composite mea-
sure of compliance and number of bench warrants issued or stayed
by the court. Compliance was an average rating of compliance
with keeping treatment appointments in the community, judicial

Table 1
Descriptive Information of the Participants and Study Variables (N � 448)

Variable Range % or M (SD) Skewness (original) Kurtosis (original)

Demographics
Age 18–75 37.45 (10.69) .12 �.61
Gender (male) 0–1 58.2% .33 �1.90
Ethnicity (White) 0–1 49.3% .03 �2.01

Colorado Symptom Index 0–60 24.49 (13.31) .24 �.43
Lifetime arrest since 15-year-old 1–250 14.96 (27.92) 5.02 30.88
Therapeutic jurisprudence (baseline)

MHC knowledge 0–31 21.31 (6.90) �1.78 3.10
Procedural justice 5–35 28.51 (7.41) �1.38 1.23
Perceived voluntariness 0–31 17.19 (6.31) �.41 �.34

Criminal justice outcomes
Arrest (0, 1–2, 3�) 0–2 .31 (0.57) 1.67 (4.01) 1.77 (22.94)
Prison (Yes) 0–1 13.6% 2.13 2.54

MHC outcomes
Bench warrant (0, 1–2, 3�) 0–2 .63 (0.74) .70 (4.52) �.85 (38.05)
Average compliance rating 0–4 2.24 (1.15) �.16 �.86
MHC completion 0–1 47.3% .10 �1.99

Note. MHC � mental health court.

T
hi

s
do

cu
m

en
t

is
co

py
ri

gh
te

d
by

th
e

A
m

er
ic

an
P

sy
ch

ol
og

ic
al

A
ss

oc
ia

ti
on

or
on

e
of

it
s

al
li

ed
pu

bl
is

he
rs

.
T

hi
s

ar
ti

cl
e

is
in

te
nd

ed
so

le
ly

fo
r

th
e

pe
rs

on
al

us
e

of
th

e
in

di
vi

du
al

us
er

an
d

is
no

t
to

be
di

ss
em

in
at

ed
br

oa
dl

y.

112 REDLICH AND HAN

and court orders, and taking prescribed medications (rs � .87).
MHC coordinators rated compliance on these three aspects using a
scale of 1 � poor/not so good throughout to 5 � excellent/very
good throughout. Coordinators were instructed to consider the
entire year (or however long the person remained in the court) and
the frequency and severity of positive and negative events. Num-
ber of MHC bench warrants stayed or issued at the 12-month point
was obtained from MHC records.

Statistical Analyses

Variables with skewness above 2 and kurtosis above 7 were
considered abnormally distributed (West, Finch, & Curran, 1995).
Among the study variables, arrest and bench warrants were found
to be highly skewed (4.01 and 4.52, respectively) and have high
kurtosis (22.94 and 38.05, respectively). This is due to the large
portion of study participants with no new arrests (74%) and no
MHC bench warrants (52%). Arrests and MHC bench warrants
were categorized into three groups (0 [0], 1–2 [1], and 3� [2]; see
Table 1). These transformed variables were used in the descriptive
and SEM analyses. The SEM model was composed of three latent
variables: principles of TJ, CJ, and MHC (L for latent), and one
observed variable—MHC completion. TJ had three observed in-
dicators, all measured at the onset of MHC participation: MHC
knowledge, procedural justice, and perceived voluntariness.

A SEM is composed of measurement and structural parts. The
measurement part is validated from the exploratory factor analysis
(EFA): how appropriately each latent variable is measured by its
observed indicators (based on the model fit). The structural part is
validated from the confirmatory factor analysis (CFA): how well
relations among latent variables are hypothesized. Because each
data collection site did not satisfy the minimum sample size for
SEM (N �200; Garver & Mentzer, 1999; Harris & Schaubroeck,
1990; Hoelter, 1983), site variable was not included in the SEM
analyses. From the CFA, the number of factors to be included in
the model producing the best fit can be determined. Specifically,
we evaluated whether a four-factor model (the three latent con-
structs and the outcome) had a moderate model fit compared with
the direct or three-factor models. The direct model included the
outcome and the TJ latent variable (a combination of MHC knowl-
edge, procedural justice, and perceived voluntariness). The three-
factor model included the (a) outcome, (b) the TJ latent variable,
and (c) a CJ/MHC-L latent variable, which combined arrests,
prison, bench warrants, and compliance into one latent variable.
Finally, we analyzed a four-factor model with possible paths
among the three latent variables (TJ, CJ, and MHC-L) and one
observed variable (MHC completion). Mplus 6.0 (Muthén &
Muthén, 2010) was used for the SEM.

Although maximum likelihood is the most often-used estimator
to analyze data in a SEM, the weighted …