Do Demographic Factors Moderate How Well Criminal Thinking Predicts Recidivism?

Abstract

Is the relationship between criminal thinking and recidivism the same for criminal justice–involved individuals from varying demographic backgrounds? Relying on two independent samples of offenders and two measures of criminal thinking, the current studies examined whether four demographic factors—gender, race, age, and education—moderated the relationship between criminal thinking and recidivism. Study 1 consisted of 226 drug-involved probationers enrolled in a randomized clinical trial. Study 2 consisted of 346 jail inmates from a longitudinal study. Logistic regression models suggested that the strength of the relationship between criminal thinking and subsequent recidivism did not vary based on participant demographics, regardless of justice system setting or measure of criminal thinking. Criminal thinking predicts recidivism similarly for people who are male, female, Black, White, older, younger, and more or less educated.

Keywords

criminal thinking Criminogenic Cognitions Scale Criminal Thinking Scale recidivism

Criminal thinking encompasses “offense-supportive attitudes, cognitive processing during an offense sequence, as well as post-hoc neutralizations or excuses for offending” (Maruna & Mann, 2006, p. 155) and is often identified as a major risk factor in both the onset and maintenance of criminal behavior (Andrews & Bonta, 2010; Andrews, Bonta, & Wormith, 2006). Criminal thinking is a key component of several criminological theories, including subcultural (Anderson, 1999), differential association (Sutherland, 1947), and self-control (Gottfredson & Hirschi, 1990), as well as correctional rehabilitation frameworks (e.g., Risk-Need-Responsivity). It also serves as a primary target of several treatments for criminal justice–involved populations (e.g., Thinking for a Change; Bush, Glick, & Taymans, 1997). The relation between criminal thinking and recidivism, however, is modest on average (Walters, 2012, 2016; Walters & Lowenkamp, 2016) and essentially nonexistent in some studies (e.g., Mills & Kroner, 1997). The discrepancy in findings across studies motivates the question of why a theoretical construct so ubiquitous in correctional discourse proves only modestly, and in some cases not at all empirically, related to recidivism.

One possible explanation for these inconsistent findings regarding the predictive utility of criminal thinking is that the construct varies as a function of offender characteristics. Minimal empirical work has examined this notion in regard to demographic characteristics (Walters, 2014) or even more substantive moderators (e.g., substance use disorder symptoms [SUDS]; Caudy et al., 2015). The current study extends the existing literature by assessing whether demographic characteristics moderate the degree to which criminal thinking predicts recidivism within two distinct justice-involved populations, using two different assessments of criminal thinking.

Sometimes Criminal Thinking Predicts Recidivism

Empirical investigations of the criminal thinking–recidivism relationship, using a variety of criminal thinking assessments¹ in diverse samples, have yielded mixed results; a moderately robust link has been found in some studies (see Andrews & Bonta, 2010; Walters, 2012a, for meta-analytic support), but not in others (cf. Simourd & Van De Ven, 1999; Taxman, Rhodes, & Dumenci, 2011). Of note, some measures are less effective at predicting recidivism than others, while other measures are effective at predicting recidivism among some populations, but not others. For example, the Criminal Sentiments Scale–Modified (CSS-M; Simourd, 1997) predicts recidivism among violent offenders, but not nonviolent ones, whereas the reverse is true for the Pride in Delinquency Scale (Shields & Whitehall, 1991; Simourd & Van De Ven, 1999).

A recent meta-analysis synthesizing evidence from only one criminal thinking instrument, the Psychological Inventory of Criminal Thinking (PICTS; Walters, 1995), demonstrated a fairly consistent but modest link between criminal thinking and recidivism (Walters, 2012). Using six primary studies with seven unique inmate samples—four federal, one state, one non-U.S. prisoner, and one forensic patient—mean effect sizes (r) ranged from .14 to .27. The pooled mean effect size across studies was r = .20 (95% confidence interval [CI] = [0.15, 0.24]). In short, although widely discussed in the theoretical literature as a key criminogenic risk factor, empirical studies show only small to moderate effect sizes for criminal thinking when predicting recidivism. The small to moderate effect suggests the predictive utility of criminal thinking may depend on additional factors such as criminal history, age, gender, or other demographics. However, the meta-analysis did not conduct moderator analyses to explore this notion.

For Whom Does Criminal Thinking Predict Recidivism?

An unanswered question is whether criminal thinking is differentially predictive of recidivism for some people as opposed to others. It is important to investigate these types of questions to understand the boundaries of the effect. Does the modest relationship between criminal thinking and recidivism generalize across different groups of individuals? Or is it that it is a more (or less) salient risk factor for some individuals? Understanding the boundaries of the relationship can help individualize appropriate treatment and interventions.

One can imagine, for example, how measures developed primarily with males might be less meaningful and valid for females. Part of this may have to do with motivations for criminal behavior. According to Steffensmeier and Allan (1996), gender differences shape criminal motivation and styles of risk-taking. While males traditionally take risks to gain status or competitive advantage, females traditionally take risks to maintain relationships. While a sense of entitlement, power orientation, and other cognitions may be a strong risk factor for criminal behavior for males, this may be less true for females.

Similarly, an assessment of “negative attitudes towards authority” (e.g., correctional officers, police officers; Tangney et al., 2012, p. 1343) might differ in meaning for a Black male, as compared with a White male living in the United States, given the attitudes of Whites and minorities to justice authority figures such as police (cf. Reisig & Parks, 2000). In regard to age, the behavior of younger people may be more strongly guided by peers than individual patterns of criminal thinking. Conversely, among older people, less influenced by peers, criminal thinking may be an especially potent predictor of recidivism. Or, maybe criminal thinking is a stronger risk factor for females or younger individuals. Whatever the reason, practitioners need to know if measures of criminal thinking are equally useful and meaningful for people of varying age, race, gender, and other factors.

Three recent studies assessed potential moderators of the predictive utility of criminal thinking in correctional populations (Caudy et al., 2015; Walters, 2014; Walters & Lowenkamp, 2016). Walters (2014) first examined whether criminal thinking differentially predicted recidivism as a function of education (<12 years as compared with ≥12 years) and race (White, Black, Hispanic) in a sample of 1,011 male inmates released from federal prison. Initial analyses revealed a significant interaction between the PICTS General Criminal Thinking (GCT; Walters, 1995) score and race, but not education. GCT was a stronger predictor of recidivism for Whites compared with minority participants. Secondary analyses demonstrated that the PICTS GCT score was significantly more accurate at predicting recidivism among more highly educated (≥12 years education) participants than among less-educated participants. As a higher proportion of White participants completed higher education, compared with Black or Hispanic participants, however, the Initial Race × GCT interaction was interpreted as an artifact of educational differences.

Replication of this finding with other measures of criminal thinking is necessary because this differential predictive validity based on education may be due to the ninth-grade reading level requirement for the PICTS (Disabato et al., 2016). Although it is common to exclude individuals high on the PICTS Confusion subscale from analyses as Walters (2014) did, the items in this subscale do not appear to be related to reading ability; they seem to reflect psychosis and distractability. For example, the item “I am afraid of losing my mind” does not capture one’s reading ability. Given that studies of adult correctional populations in the United States show that only about 50% read at or above a sixth-grade level (Bureau of Justice Statistics, 2003; Ryan, 1991), completing the PICTS may be too complicated for the average inmate, and particularly individuals who present with lower levels of education.

In a second study, Walters and Lowenkamp (2016) also relied on the PICTS to examine whether the relationship between criminal thinking and recidivism differed based on gender and ethnicity using receiver operating characteristic (ROC) analyses in a sample of 96,500 male and female offenders on federal probation or supervised release. They did not use moderation analyses for demographic moderators, but examined individual area under curve (AUC) values and 95% CIs for the total sample as well as the male, female, Black, White, Hispanic, and non-Hispanic subsamples; there were negligible differences between males and females and between Hispanics and non-Hispanics in how well the PICTS predicted recidivism. When comparing Blacks and Whites, the AUC values were approximately 5% lower in the subsample of Black individuals, but the effect sizes were of similar magnitude regardless of race.

In a third study, Caudy and colleagues (2015) examined SUDS as a moderator of the relation between criminal thinking and recidivism among the same two independent samples considered here—226 drug-involved probationers and 337 jail inmates. Criminal thinking, as measured by the Criminal Thinking Scale (CTS; Knight, Garner, Simpson, Morey, & Flynn, 2006), was not predictive of recidivism among the drug-involved probationers, and SUDS did not moderate the relation between criminal thinking and recidivism. Within the sample of jail inmates with varying levels of substance involvement, however, criminal thinking, as measured by the Criminogenic Cognitions Scale (CCS; Tangney, Meyer, Furukawa, & Cosby, 2002; Tangney et al., 2012), significantly predicted recidivism. Furthermore, SUDS moderated the relationship between criminal thinking and recidivism, whereby criminal thinking significantly predicted recidivism among offenders low in SUDS, but not among individuals high in SUDS.

In sum, the available research (Caudy et al., 2015; Walters, 2014; Walters & Lowenkamp, 2016) indicates a modest relationship between criminal thinking and subsequent recidivism with some variation between different subpopulations. However, despite the recent studies discussed here, relatively little research overall has been devoted to the study of potential individual differences in the strength of the criminal thinking and recidivism relationship.

A fourth study, a recent meta-analysis (Helmond, Overbeek, Brugman, & Gibbs, 2015), examined the relation between cognitive distortions and externalizing behavior, including whether demographic and study characteristics moderated this relation. Gender distribution, age, and ethnic composition of the sample did not moderate the relationship between cognitive distortions and externalizing behavior. The use of meta-analysis for examination of individual-level demographic moderators is inherently problematic, however, because the samples are heterogeneous. This results in problematic operationalization of individual-level moderators, such as classifying samples that are 40% male as “mixed” and comparing it with a 60% male sample classified as “Male” (Helmond et al., 2015). This is not a direct or effective test of whether demographic characteristics moderate the criminal thinking–recidivism relation, as is the aim of the current studies. Furthermore, although related to the current aims, both the independent and dependent variables differ from those considered in the current studies. The independent variable—cognitive distortions—is a broad construct including many different thinking patterns not all necessarily related to criminal thinking. The dependent variable—externalizing behavior—included a broad range of behavior, such as aggression and criminal behavior. Given the wide age range of samples included (e.g., childhood to adulthood), the behaviors captured in the outcome are likely qualitatively distinct and not directly comparable with adult criminal recidivism.

The Current Studies

Building on the minimal extant research on demographic moderators of the criminal thinking–recidivism relationship, the current studies tested whether the relation varied according to demographic factors (gender, race, age, education), as represented by significant Criminal Thinking × Demographic Characteristic interactions. Relying on the same samples as Caudy and colleagues (2015), we extend Walters’s (2014) recent study of federal prisoners by relying on two independent samples of justice-involved individuals and two different measures of criminal thinking, as well as by expanding the set of demographic moderators examined. Study 1 used a sample of drug-involved probationers who participated in a randomized clinical trial (RCT) designed to test the efficacy of a seamless model of substance abuse treatment. Participants were tracked over a 12-month period using self-report and official records. Study 2 used a sample of jail inmates housed in general population assessed while incarcerated and then tracked at 12 months postrelease. To measure criminal thinking, we also used two different measures: Study 1 used the Texas Christian University CTS (Knight et al., 2006) and Study 2 used the CCS (Tangney et al., 2002; Tangney et al., 2012).

The generality of criminal thinking in predicting recidivism has considerable implications for correctional intervention practices. Of note, if criminal thinking is not a significant predictor of recidivism for individuals from a certain demographic group (e.g., females, older adults), then interventions solely targeting criminal thinking are less likely to be effective in reducing recidivism for this particular group of people. A one-size-fits-all approach would thus squander limited resources. If there is no significant moderation, however, criminal thinking can be considered a relevant predictor of recidivism for a variety of individuals regardless of gender, race, age, and educational attainment; this could then expand its range of application.

Method—Study 1

Participants

Participants were 226 drug-involved probationers enrolled in a RCT of a correctional intervention to test the efficacy of a seamless model of substance abuse treatment for probationers. The treatment arm of the RCT involved the administration of manualized substance use treatment that was provided within the probation office during routine probation visits. The treatment was labeled as seamless because it occurred in the context of routine community correctional practices and included the treatment provider and probation officer jointly running sessions. The control condition involved making a referral to community-based substance abuse treatment. This was considered treatment as usual (TAU) in the study sites. The control group is the “standard care” for this organization.

Study recruitment began in March of 2007 by probation officer referral at three probation sites within the Maryland Division of Probation and Parole. Inclusion criteria were (a) at least 18 years of age, (b) English speaking, (c) sanctioned to probation supervision, and (d) substance use treatment as a condition of probation. Individuals were excluded if assigned to a specialized caseload (e.g., sexual offender). Probationers meeting the eligibility criteria were invited to enroll in the study; participation was voluntary. All participants were assessed at baseline and re-assessed at 12 months postrandomization. The baseline assessment generally occurred early in the probation sentence, but some participants had been on supervision for up to a year before study enrollment. Participants received a US$40 honorarium for each completed interview. All study procedures were approved by the affiliated institutional review board.

The Consolidated Statement on Reporting Trials (CONSORT) diagram for Study 1 is presented in Figure 1. A total of 415 probationers were assessed for eligibility; 164 of these did not participate in the study because they (a) did not meet inclusion criteria (n = 141), (b) declined to participate (n = 10), or (c) other reasons (n = 13). As such, 251 participants were randomized into the treatment (n = 126) and control (n = 125) groups. Of the 251 participants, recidivism data for both self-report and official records were available for 226 (90%). The official records of 25 participants were not available. There were no significant demographic differences between the participants who had official records available and participants who did not. The final sample for analysis included 226 participants with baseline and 12-month follow-up recidivism data. Descriptive statistics from baseline are shown in Table 1.

Figure 1.

CONSORT diagram for study 1.

Table 1.

Participant Demographic Characteristics.

Characteristics	Study 1 (n = 226)		Study 2 (n = 346)
Characteristics	M (SD)/%	Range	M (SD)/%	Range
Age (in years)	36.83 (11.54)	18-64	33.28 (10.15)	18-69
Male	73.5%	—	70.2%	—
Race/ethnicity
Black	68.1%	—	47.1%	—
White	30.1%	—	35.0%	—
Hispanic	0.4%	—	6.1%	—
Other	1.3%	—	11.8%	—
Marital status
Never married	68.1%	—	55.4%	—
Married	10.6%	—	15.1%	—
Separated	7.1%	—	9.9%	—
Divorced	12.8%	—	15.1%	—
Widowed	1.3%	—	0.3%	—
Other	—		4.4%	—
Education (in years)	11.41 (1.61)	6-17	11.91 (2.12)	5-18
Annual income (in US$)	8,391.10^a (10,647.69)	0-48,000	21,560.03 (23,998.50)	0-300,000
Employment status
Unemployed	43.4%	—	13.3%	—
Employed	56.6%	—	86.7%	—

Note. All demographic characteristics were reported at baseline.

Calculated based on self-reported income over the prior 30 days.

Baseline Measures

Demographics

Gender (0 = female, 1 = male), age in years, and years of education were assessed via self-report at baseline. Race was collected at this time, but due to the small number of individuals classified as Hispanic or Other (<2% of sample), we excluded those individuals from the analysis in which we use race (coded as 0 = White 1 = Black).

Criminal thinking

The Texas Christian University CTS (Knight et al., 2006) is a 37-item instrument used to measure criminal thinking in six areas: (a) entitlement, (b) justification, (c) power orientation, (d) cold heartedness, (e) criminal rationalization, and (f) personal irresponsibility. Items were scored on a 5-point Likert-type scale ranging from 1 (disagree strongly) to 5 (agree strongly). A total CTS was computed by averaging scores across the six areas—higher scores represent higher levels of criminal thinking. The average score was 2.37 (SD = 0.41, range = 1.40-3.55) on the CTS, suggesting a moderate level of criminal thinking comparable with those reported in other samples of offenders (see Knight et al., 2006). The CTS demonstrated good test–retest reliability in justice-involved samples (Knight et al., 2006). In the present sample, Cronbach’s alpha was low, but acceptable (α = .63).

One-Year Follow-Up Measures

Recidivism

Recidivism during the 1-year follow-up period was assessed using self-report and official records. Participants were provided with a list of 25 offenses (e.g., robbery/attempted robbery/mugging, shoplifting/larceny/embezzlement, weapons offenses) and asked to self-report any arrest for each specific offense. Then, using the same list, participants reported any offenses they committed but were not caught for (undetected offenses). Official records of arrests were obtained from the State of Maryland’s Criminal Justice Information System. Using self-report and official indicators of recidivism, a dichotomous variable was coded such that 0 = did not recidivate, 1 = recidivated. Recidivists were any participants who self-reported offending, self-reported a new arrest, or showed a new official arrest during the 12-month follow-up period. To maintain consistency with a previously published paper using criminal thinking and recidivism from both these studies (Caudy et al., 2015), we created the recidivism variable without drug crimes for these analyses.² Probation violations were excluded from the recidivism variable. Overall, 46% (n = 104) of participants recidivated during the 12-month period.

Method—Study 2

Participants

Participants were 346 inmates in a county jail located in a suburb of the District of Columbia (see Table 1 for descriptive statistics at baseline). Participants were recruited for baseline assessment after assignment to general population. Participants were re-interviewed 1 year following release. Each participant received a US$15 to US$18 honorarium at baseline and US$50 upon completion of the 1-year follow-up interview. All study procedures were approved by the affiliated institutional review board.

Of the 628 inmates who consented and were enrolled in the study (74% of those approached), 508 completed full, valid baseline assessment (e.g., were not transferred or released to bond before the assessments could be completed) and were followed longitudinally. A CONSORT diagram for this sample can be seen in Figure 2. Of the 508 inmates enrolled in the longitudinal study, 484 had complete data for the criminal thinking variable at baseline, and 478 were eligible to be included in the 1-year postrelease recidivism data collection. Of the 478, one was detained by Immigration and Naturalization Service (INS)/deported and two were deceased or seriously disabled and therefore could not be interviewed. Of the remaining 475 eligible participants, 346 (73%) had baseline criminal thinking data and 1-year postrelease self-report and official record recidivism data. As such, the final sample for analysis included 346 participants. The retention rate compares favorably with other longitudinal studies of incarcerated individuals (Brown, Amand, & Zamble, 2009; Howerton, Burnett, Byng, & Campbell, 2009; Inciardi, Martin, & Butzin, 2004). Participants in the final analysis sample were significantly older, t(488) = −2.42, p = .016, than those with baseline criminal thinking data but no 1-year postrelease recidivism data; no other significant differences emerged.

Figure 2.

CONSORT diagram for study 2.

Baseline Measures

Demographics

Gender (0 = female, 1 = male), age in years, and years of education were assessed via self-report at baseline. As in Study 1, race was coded as 0 = White, 1 = Black; 18% of individuals identified as Hispanic or other and were excluded from the analysis of race.

Criminal thinking

The CCS (Tangney et al., 2002; Tangney et al., 2012) is a 25-item self-administered assessment that measures five dimensions of criminal thinking: (a) notions of entitlement, (b) failure to accept responsibility, (c) short-term orientation, (d) insensitivity to the impact of crime, and (e) negative attitudes toward authority. Items were rated on a 4-point scale ranging from 1 (strongly disagree) to 4 (strongly agree) and averaged to create a total criminal thinking score.

The average score was 2.24 (SD = 0.35, range = 1.28-3.28) on the CCS. Evidence for the reliability and validity of the CCS has been demonstrated with jail inmates drawn from the same longitudinal study examined here (Tangney et al., 2012). Cronbach’s alpha was acceptable in the present sample (α = .72).

One-Year Follow-Up Measures

Recidivism

As in Study 1, recidivism was assessed during the first-year postrelease using both self-report and official records. Participants reported whether they were arrested for any of 16 crimes (e.g., theft, assault, sex offenses, fraud) and whether they committed, but were not caught for committing, any of the 16 crimes. Official National Crime Information Center criminal records of arrests in the first year postrelease were also collected. Consistent with Study 1, a dichotomous recidivism variable was created (0 = did not recidivate, 1 = recidivated). As in Study 1, drug use crimes and technical violations of probation conditions were excluded from the recidivism variable. Participants who had an offense using any of the three indicators were considered recidivists. In the current sample, 63% of participants (n = 219) recidivated during the first year postrelease.

Analytic Procedure

To test our hypotheses of whether demographic variables moderate the relation between criminal thinking and recidivism, logistic regression analyses were conducted. The main effect of each demographic factor and recidivism were entered first, with the interaction term entered in Step 2. Multicollinearity problems did not preclude conducting this series of analyses. Criminal thinking and age, the two continuous variables, were mean centered for analyses and for computation of the interaction terms. This analytic procedure was replicated across both data sources. Results from each study are reported separately and then considered together in the “Discussion” section.

Results

Study 1: Drug-Involved Probationers

Demographic differences in CTS (Knight et al., 2006) scores are presented in Table 2. No significant differences were found for gender or race. Older individuals endorsed significantly lower levels of criminal thinking, r(224) = −.14, p < .05, as did individuals with higher levels of education, r(226) = −.18, p < .01.

Table 2.

Intercorrelations Between Variables of Interest.

	1	2	3	4	5	6
Study 1 (n = 226)
Baseline
1. Criminal thinking	—	.10	.07	−.14*	−.18*	.08
2. Gender (male)		—	.09	−.01	−.00	.01
3. Race (Black)			—	.24***	−.13*	−.08
4. Age				—	.10	−.13
5. Education					—	.03
One-year follow-up
6. Recidivism						—
Study 2 (n = 346)
Baseline
1. Criminal thinking	—	.15**	.15*	−.21***	−.25***	.19***
2. Gender (male)		—	.11^†	−.18**	−.18**	.13*
3. Race (Black)			—	.04	−.04	.09
4. Age				—	.14**	−.11*
5. Education					—	−.09†
One-year postrelease
6. Recidivism						—

Note. Gender is coded 0 = female, 1 = male and race is coded 0 = White, 1 = Black. Age and education are in years.

†

p < .10. *p < .05. **p < .01. ***p < .001.

Demographics as a moderator of the criminal thinking-recidivism relation

The main and interactive effects of criminal thinking and each demographic on recidivism were then examined.³ Bivariate relations are presented in Table 2 and results of logistic regression analyses are presented in Table 3. Results for the main effects are presented from Step 1 and results from the interactions are presented from Step 2.

Table 3.

Logistic Regression Predicting Recidivism From Criminal Thinking and Demographic Factors.

	Study 1 (n = 226)		Study 2 (n = 346)
	B (SE)	OR [95% CI]	B (SE)	OR [95% CI]
Model 1: Gender
Criminal thinking	0.41 (.33)	1.50 [0.79, 2.86]	1.07 (.34)**	2.93 [1.51, 5.66]
Gender	0.02 (.31)	1.02 [0.56, 1.85]	0.49 (.25)*	1.64 [1.01, 2.65]
Criminal Thinking × Gender	−0.11 (.73)	0.90 [0.22, 3.77]	0.46 (.70)	1.58 [0.40, 6.26]
Model 2: Race^a
Criminal thinking	0.52 (.34)	1.69 [0.87, 3.26]	0.98 (.37)**	2.66 [1.28, 5.53]
Race	−0.38 (.30)	0.68 [0.38, 1.22]	0.29 (.25)	1.33 [0.81, 2.19]
Criminal Thinking × Race	0.43 (.76)	1.53 [0.35, 6.77]	−0.30 (.75)	0.75 [0.17, 3.23]
Model 3: Age
Criminal thinking	0.33 (.33)	1.39 [0.73, 2.66]	1.08 (.34)**	2.93 [1.51, 5.71]
Age	−0.02 (.01)	0.98 [0.96, 1.00]	−0.02 (.01)	0.99 [0.96, 1.01]
Criminal Thinking × Age	−0.02 (.03)	0.98 [0.92, 1.04]	−0.05 (.03)	0.95 [0.89, 1.01]
Model 4: Education^b
Criminal thinking	0.45 (.33)	1.57 [0.82, 3.02]	1.14 (.35)**	3.11 [1.58, 6.13]
Education	0.06 (.09)	1.06 [0.90, 1.26]	−0.04 (.06)	0.96 [0.86, 1.07]
Criminal Thinking × Education	−0.13 (.20)	0.88 [0.60, 1.29]	0.12 (.16)^b	1.12 [0.82, 1.54]

Note. Gender is coded 0 = female, 1 = male and race is coded 0 = White, 1 = Black. Education and age are in years. Main effects are from Step 1 and interactions are presented from Step 2. OR = odds ratio; CI = confidence interval.

Study 2: Race includes Black and White only (n = 284).

In Study 2, n = 344.

†

p < .10. *p < .05. **p < .01. ***p < .001.

In multivariate analysis, neither criminal thinking nor any of the demographic predictors were significantly related to recidivism. None of the interaction effects were significant. This suggests that results generalize across males and females, Blacks and Whites, and individuals of different ages and educational levels.

Study 2: General Population Jail Inmates

Demographic differences in CCS (Tangney et al., 2002; Tangney et al., 2012) scores are presented in Table 2. In contrast to the Study 1 sample, gender and race were significantly related to criminal thinking. Males endorsed higher levels of criminal thinking than females, t(344) = −2.71, p < .01, and Blacks scored higher than Whites, t(282) = −2.52, p < .05. Paralleling Study 1, older individuals endorsed significantly lower levels of criminal thinking, r(346) = −0.21, p < .001, as did individuals with higher levels of education, r(344) = −0.25, p < .001.

Demographics as a moderator of criminal thinking- recidivism relation

Results of logistic regression analyses are presented in Table 3. Results for the main effects are presented from Step 1 and results from the interactions are presented from Step 2.

In multivariate analysis, criminal thinking and gender significantly predicted recidivism; no other demographic factors were significant predictors. Males recidivated at a higher rate than females; those higher in criminal thinking were more likely to recidivate than those lower in criminal thinking. None of the interaction effects were significant. This suggests, as in Study 1, effects generalize across males and females, Blacks and Whites, and individuals of different ages and educational levels.

Discussion

Extending Walters’s (2014) study of federal prisoners, the current studies relied on two independent samples of justice-involved individuals and two different measures to examine the link between recidivism and criminal thinking, a construct that is frequently touted as important for predicting whether criminal justice individuals will engage in future criminal behavior. Recent evidence suggests some individuals are more likely to engage in criminal thinking than others (e.g., younger offenders, less educated offenders, Black and Hispanic offenders; Mandracchia & Morgan, 2012), and therefore, the current studies examined if the relationship between criminal thinking and recidivism depends on demographics. The most important finding appears to be that the criminal thinking–recidivism relationship is generalizable across these demographic characteristics.

The criminal thinking and recidivism relationship revealed in the current studies clarifies the utility of this risk factor for risk assessment and management. Only one prior study has examined whether criminal thinking differentially predicts recidivism as a function of demographic characteristics—Walters’s (2014) study of 1,011 male inmates released from federal prison. In contrast to Walters’s (2014) evidence for stronger predictive validity of criminal thinking for more educated prisoners, educational attainment did not moderate the criminal thinking–recidivism relationship in either Study 1 or Study 2. It is possible Walters’s (2014) results are due to the PICTS assessment, which requires a ninth-grade reading level (Disabato et al., 2016), and may be too complicated for less educated individuals to accurately complete. It may also be that federal inmates are systematically different from probationers and jail inmates. Finally, it should also be noted that in contrast to Walters’s (2014) dichotomized measure of education (<12 years as compared with ≥12 years), education was captured in the present studies with a continuous variable reflecting years of education completed. Use of a continuous variable, compared with a dichotomous one, increases one’s ability to detect effects, so it is unlikely this difference explains the inconsistent findings. Furthermore, when analyses were conducted using a dichotomized education variable, education was still not a significant moderator of the criminal thinking-recidivism relationship.

More recently, Walters and Lowenkamp’s (2016) outcome study of the PICTS with a large sample of federal probationers showed there were negligible differences between males and females, Hispanics and non-Hispanics, and Blacks and Whites, in how well criminal thinking predicts recidivism. Results from the current studies seem to support the latter PICTS study (i.e., Walters & Lowenkamp, 2016), adding to the evidence that criminal thinking predicts recidivism, regardless of demographics. Specifically, findings from studies using different criminal thinking measures and diverse samples suggest there are minimal differences in the predictive validity of criminal thinking when considered by gender or race.

Practical Implications

The results of the current studies suggest criminal thinking can be used to predict recidivism for many different types of offenders with equal confidence. One important caveat, however, is there were relatively small but significant differences in mean levels of criminal thinking present in both samples. In the sample of drug-involved probationers, older and more educated individuals reported lower levels of criminal thinking. In the sample of jail inmates, females, Whites, older, and more educated individuals reported lower levels of criminal thinking. Furthermore, in a previous study, the CTS did not significantly predict recidivism among drug-involved probationers; level of substance dependence is an important variable to consider when using criminal thinking as a predictor in settings characterized by a high prevalence of substance use disorders (Caudy et al., 2015).

When interpreting scores on these criminal thinking assessments, it is critical to take into account the average for other inmates in that demographic group. For example, about two thirds of people fall within 1 standard deviation of the mean. Those scoring within 1 standard deviation of the mean would be considered average in criminal thinking; those scoring between 1 and 2 standard deviations above the mean would be considered high in criminal thinking; those scoring more than 2 standard deviations above the mean would be considered very high in criminal thinking—the same applies to standard deviations below the mean. Although not strict cutoff scores, these parameters may provide meaningful information regarding ones’ level of risk for recidivism based on criminal thinking score.

It is also important to remember, however, that criminal thinking is a modest predictor of recidivism—not everyone who scores high on criminal thinking assessments will recidivate. Other factors, such as those outlined in the risk-needs-responsivity model (Andrews & Bonta, 2010), need to be taken into account.

Limitations and Future Directions

The findings from the current studies indicate criminal thinking assessments have equal utility in predicting recidivism regardless of respondents’ gender, age, race, and years of education. Future research can build upon these findings by using a larger sample to increase power, exploring these relations in different criminal justice samples to increase generalizability, and examining more distal follow-up periods to determine whether there is evidence for moderation at later timepoints. Furthermore, it would be useful to employ other measures of criminal thinking to determine whether these effects generalize across instruments and examine additional moderators of the criminal thinking–recidivism relationship.

Regarding alternative assessments, the field would benefit from a thorough assessment of the predictive utility of criminal thinking measures in diverse samples. In the current studies, we used two different assessments thought to assess the broader criminal thinking construct. It is possible these measures tap into different aspects of this broader construct and this may contribute to the inconsistency in findings. For example, whereas the CCS assesses both proactive and reactive criminal thinking, the CTS more thoroughly assesses proactive criminal thinking, which is thought to be a weaker predictor of recidivism compared with reactive criminal thinking (Walters & Lowenkamp, 2016).

To assess recidivism, the current studies relied in part on self-reported involvement in criminal activity. With regard to the use of self-report data, research fairly consistently demonstrates there are minimal differences between criminal justice–involved individuals’ self-reported criminal behavior and their official record data (Maxfield, Weiler, & Widom, 2000; Weis, 1986). Furthermore, studies also suggest that the stability of self-reports are higher for interviewer-administered instruments than for self-administered assessments (Catania, Gibson, Chitwood, & Coates, 1990; Weinhardt et al., 1998); both Studies 1 and 2 used interviewer-administered interviews. Research has further suggested that self-reported criminal behavior may be a more accurate measure of behavior than official records due to factors such as police discretion and inaccuracies in official criminal records (Elliott, 1994; Marquis, 1981). Nonetheless, this measurement decision could affect the findings.

Although the current findings suggest demographic factors including gender, race, age, and education do not moderate the relationship between criminal thinking and recidivism, this does not mean the relationship is universal; we know one factor that attenuates this link is the severity of SUDS (Caudy et al., 2015). Other factors that may be influential include self-control, mental illness, characteristics of the neighborhood one returns to following release from incarceration, and different social connections (e.g., marriage, antisocial peers). These factors should be investigated in future research. It is likely that several different investigators have data on this question; a series of brief reports with a wide variety of samples (e.g., jail inmates, prison inmates, probationers, juveniles) and measures (e.g., PICTS, CSS-M) testing the generalizability of the criminal thinking-recidivism relationship would advance understanding regarding what attenuates or magnifies the effect, as well as its magnitude. Understanding for whom criminal thinking is a salient risk factor may facilitate directing interventions to those with the greatest need and for those who would benefit the most.

Footnotes

Acknowledgements

Many thanks to members of the Human Emotions Research Lab, researchers at the Center for Advancing Correctional Excellence, and individuals who participated in our studies.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by grant numbers R01 DA14694 (principal investigators [PIs]: June P. Tangney and Jeffrey Stuewig), R01 DA017729 (PI: Faye Taxman), and F31 DA039620 (PI: Johanna B. Folk) from the National Institute on Drug Abuse of the National Institutes of Health.

Notes

References

Anderson

(1999). Code of the street: Decency, violence, and the moral life of the inner city. New York, NY: W.W. Norton.

Andrews

D. A.

Bonta

(2010). The psychology of criminal conduct. New Providence, NJ: Routledge.

Andrews

D. A.

Bonta

Wormith

J. S.

(2006). The recent past and near future of risk and/or need assessment. Crime & Delinquency, 52, 7-27.

Brown

S. L.

Amand

M. D. S.

Zamble

(2009). The dynamic prediction of criminal recidivism: A three-wave prospective study. Law and Human Behavior, 33, 25-45.

Bureau of Justice Statistics. (2003). Education and correctional populations (NCJ Publication No. 195670). Washington, DC: U.S. Government Printing Office.

Bush

Glick

Taymans

(1997). Thinking for a change. Longmont, CO: The United States Department of Justice, National Institute of Corrections.

Catania

J. A.

Gibson

D. R.

Chitwood

D. D.

Coates

T. J.

(1990). Methodological problems in AIDS behavioral research: Influences on measurement error and participation bias in studies of sexual behavior. Psychological Bulletin, 108, 339-362.

Caudy

Folk

J. B.

Stuewig

J. B.

Wooditch

Martinez

Maass

. . .Taxman

F. S.

(2015). Does substance misuse moderate the relationship between criminal thinking and recidivism? Journal of Criminal Justice, 43, 12-19.

Disabato

D. J.

Folk

J. B.

Wilson

Barboza

Daylor

Tangney

J. P.

(2016). Psychometric validation of a simplified form of the PICTS in low-reading level populations. Journal of Psychopathology and Behavioral Assessment, 38, 456-464.

10.

Elliott

D. S.

(1994). Serious violent offenders: Onset, developmental course, and termination. Criminology, 32, 1-21.

11.

Gottfredson

M. R.

Hirschi

(1990). A general theory of crime. Palo Alto, CA: Stanford University Press.

12.

Helmond

Overbeek

Brugman

Gibbs

J. C.

(2015). A meta-analysis of cognitive distortions and externalizing problem behavior: Associations, moderators, and treatment effectiveness. Criminal Justice and Behavior, 42, 245-262.

13.

Howerton

Burnett

Byng

Campbell

(2009). The consolations of going back to prison: What “revolving door” prisoners think of their prospects. Journal of Offender Rehabilitation, 48, 439-461.

14.

Inciardi

J. A.

Martin

S. S.

Butzin

C. A.

(2004). Five-year outcomes of therapeutic community treatment of drug-involved offenders after release from prison. Crime & Delinquency, 50, 88-107.

15.

Knight

Garner

B. R.

Simpson

D. D.

Morey

J. T.

Flynn

P. M.

(2006). An assessment for criminal thinking. Crime & Delinquency, 52, 159-177.

16.

Mandracchia

J. T.

Morgan

R. D.

(2012). Predicting offenders’ criminogenic cognitions with status variables. Criminal Justice and Behavior, 39, 5-25.

17.

Mandracchia

J. T.

Morgan

R. D.

Garos

Garland

J. T.

(2007). Inmate thinking patterns: An empirical investigation. Criminal Justice and Behavior, 34, 1029-1043.

18.

Marquis

K. H.

(1981). Quality of prisoner self-reports: Arrest and conviction response errors. Santa Monica, CA: RAND.

19.

Maruna

Mann

R. E.

(2006). A fundamental attribution error? Rethinking cognitive distortions. Legal and Criminological Psychology, 11, 155-177.

20.

Maxfield

M. G.

Weiler

B. L.

Widom

C. S.

(2000). Comparing self-reports and official records of arrests. Journal of Quantitative Criminology, 16, 87-110.

21.

Mills

J. F.

Kroner

D. G.

(1997). The Criminal Sentiments Scale: Predictive validity in a sample of violent and sex offenders. Journal of Clinical Psychology, 53, 399-404.

22.

Mills

J. F.

Kroner

D. G.

Forth

A. E.

(2002). Measures of Criminal Attitudes and Associates (MCAA): Development, factor structure, reliability, and validity. Assessment, 9, 240-253.

23.

Mitchell

Tafrate

R. C.

(2012). Conceptualization and measurement of criminal thinking: Initial validation of the criminogenic thinking profile. International Journal of Offender Therapy and Comparative Criminology, 56, 1080-1102.

24.

Reisig

M. D.

Parks

R. B.

(2000). Experience, quality of life, and neighborhood context: A hierarchical analysis of satisfaction with police. Justice Quarterly, 17, 607-630.

25.

Ryan

T. A.

(1991). Literacy training and reintegration of offenders. Forum, 3, 17-19.

26.

Shields

I. W.

Whitehall

(1991, December). The Pride in Delinquency Scale. Paper presented at the Eastern Ontario Correctional Psychologists Winter Conference, Burritts Rapids, Canada.

27.

Simourd

D. J.

(1997). The Criminal Sentiments Scale-modified and Pride in Delinquency Scale: Psychometric properties and construct validity of two measures of criminal attitudes. Criminal Justice and Behavior, 24, 52-70.

28.

Simourd

D. J.

Van De Ven

(1999). Assessment of criminal attitudes criterion-related validity of the Criminal Sentiments Scale-modified and Pride in Delinquency Scale. Criminal Justice and Behavior, 26, 90-106.

29.

Steffensmeier

Allan

(1996). Gender and crime: Toward a gendered theory of female offending. Annual Review of Sociology, 22, 459-487.

30.

Sutherland

E. H.

(1947). Principles of criminology (4th ed.). Chicago, IL: J.B. Lippincott.

31.

Tangney

J. P.

Meyer

Furukawa

Cosby

(2002). The Criminogenic Cognitions Scale. Fairfax, VA: George Mason University.

32.

Tangney

J. P.

Stuewig

Furukawa

Kopelovich

Meyer

P. J.

Cosby

(2012). Reliability, validity, and predictive utility of the 25-item Criminogenic Cognitions Scale (CCS). Criminal Justice and Behavior, 39, 1340-1360.

33.

Taxman

F. S.

Rhodes

A. G.

Dumenci

(2011). Construct and predictive validity of criminal thinking scales. Criminal Justice and Behavior, 38, 174-187.

34.

Walters

G. D.

(1995). The psychological inventory of criminal thinking styles Part I: Reliability and preliminary validity. Criminal Justice and Behavior, 22, 307-325.

35.

Walters

G. D.

(2012). Criminal thinking and recidivism: Meta-analytic evidence on the predictive and incremental validity of the Psychological Inventory of Criminal Thinking Styles (PICTS). Aggression and Violent Behavior, 17, 272-278.

36.

Walters

G. D.

(2014). Relationships among race, education, criminal thinking, and recidivism: Moderator and mediator effects. Assessment, 21, 82-91.

37.

Walters

G. D.

(2016). Predicting recidivism with the Criminal Sentiments Scale: A meta-analysis of a putative measure of criminal thought content. Criminal Justice and Behavior, 43, 1159-1172.

38.

Walters

G. D.

Lowenkamp

C. T.

(2016). Predicting recidivism with the Psychological Inventory of Criminal Thinking Styles (PICTS) in community-supervised male and female federal offenders. Psychological Assessment, 28, 652-659.

39.

Weinhardt

L. S.

Carey

M. P.

Maisto

S. A.

Carey

K. B.

Cohen

M. M.

Wickramasinghe

S. M.

(1998). Reliability of the timeline follow-back sexual behavior interview. Annals of Behavioral Medicine, 20, 25-30.

40.

Weis

J. G.

(1986). Issues in the measurement of criminal careers. In Blumstein

Cohen

Roth

J. A.

Visher

C. A.

(Eds.), Criminal careers and “career criminals” (pp. 2-51). Washington, DC: The National Academy Press.