Self-Reported Change in Antisocial Attitudes and Reoffending Among a Sample of 2,337 Males Convicted of Violent Offenses

Abstract

Antisocial attitudes are among the strongest predictors of reoffending; however, there is little evidence to show that treatment-induced changes in antisocial attitudes correspond to changes in individuals’ risk of recidivism. This study examined relationships between within-treatment change in antisocial attitudes derived from the Measures of Criminal Attitudes and Associates (MCAA) and reoffending among a large sample of males convicted of violent offenses (N = 2,337). Residual change scores (RCS) and categories of clinically significant change (CSC) were used as indices of within-treatment change. A number of MCAA factor scores significantly predicted general and violent reoffending when assessed before and after treatment. RCS calculations of within-treatment change on the Violence and Antisocial Intent factors were also significantly associated with general reoffending outcomes. There was no evidence that within-treatment change on any measure had predictive validity for violent reoffending.

Keywords

antisocial attitudes within-treatment change violence reoffending

Violent crime is a prevalent issue that has significant ramifications for the individual and broader economic and health costs (Klepfisz et al., 2014). Accordingly, research has devoted attention to developing and evaluating programs aimed at reducing violent offending. Research has shown interventions that adhere to principles of the Risk Need Responsivity (RNR) model can be effective in reducing reoffending among individuals convicted of general and violent offenses (Andrews, Zinger et al., 1990; Dowden & Andrews, 2000). The risk principle states that the intensity of treatment should match the risk level of the person; the responsivity principle informs how treatment should be delivered to accommodate individual learning styles, motivation, and cognitive abilities; and the need principle proposes that treatment should target criminogenic needs, or dynamic risk factors that have a causal relationship with likelihood of reoffending (Bonta & Andrews, 2017).

One dynamic risk factor commonly targeted in treatment is that of antisocial attitudes. Antisocial attitudes are cognitions, beliefs, or values that condone, justify, or minimize antisocial and criminal behaviors (Walters & DeLisi, 2013). Attitudes in general have been described as a “pre-condition” for behavior, in which the evaluative component of an attitude influences subsequent behavior (Allport, 1935). As such, it is not surprising that attitudes are highly predictive of behavior (Mills et al., 2002). Antisocial attitudes have been shown to be among the strongest predictors of general reoffending (Eisenberg et al., 2019; Gendreau et al., 1996) and have similarly shown strong associations with violent behavior (Harris et al., 1993; Stefanile et al., 2017; Stith et al., 2004). Antisocial attitudes have been identified as one of the “big four” criminogenic needs for offending alongside antisocial personality traits, history of antisocial behavior, and antisocial associates (Bonta & Andrews, 2017). The importance of antisocial attitudes in criminal behavior is aligned with the social learning theory foundations of the RNR model and the critical role of antisocial associates in developing and maintaining such attitudes (Kabiri et al., 2020; Mills et al., 2002).

Within-Treatment Change

In accordance with the need principle, a critical premise of most interventions is that change in the severity of a person’s criminogenic needs will result in changes in their likelihood of reoffending (Andrews, Bonta et al., 1990; Bonta & Andrews, 2017). A number of studies have sought to examine this by assessing differences in individuals’ dynamic risk factors between pretreatment and posttreatment (Klepfisz et al., 2014; Lewis et al., 2013). Several studies have demonstrated that people show improvement on measures of antisocial attitudes (for a recent review, see Banse et al., 2013) and other dynamic risk factors (Heffernan et al., 2019; Papalia et al., 2020; Serin et al., 2013) after treatment. In contrast, there is little research examining whether within-treatment change has a relationship with reoffending outcomes (Klepfisz et al., 2016; Serin et al., 2013). For example, a systematic review examining 24 studies concluded that there was no statistically robust evidence that interventions to address antisocial attitudes reduce reoffending (Banse et al., 2013). This conclusion was informed by findings that few studies have assessed whether within-treatment change in antisocial attitudes predict reoffending, and fewer have demonstrated significant associations between change and outcomes.

Evidence for the relationship between within-treatment change in antisocial attitudes and reoffending is particularly lacking for individuals convicted of violent offenses and evaluations of violence treatment programs (Serin et al., 2013). One study by Klepfisz and colleagues (2014) assessed change in self-reported procriminal attitudes before and after treatment among a sample of 42 adult males with histories of violent offending. They found change across treatment, as assessed by simple differences, reliable change, and clinically significant change, was not significantly associated with violent recidivism. Similarly, Woessner and Schwedler (2014) examined pre- to posttreatment changes in procriminal attitudes among 185 males with convictions for sexual and violent offending. They found that the magnitude of reductions in procriminal attitudes over the course of treatment was not significantly related to odds of committing further general, sexual, or violent offenses.

Most studies of within-treatment change in antisocial attitudes have used self-report psychometric measures that assess various latent constructs relating to content or process domains of these attitudes. One example, which is the focus of the current study, is the Measures of Criminal Attitudes and Associates (MCAA: Mills & Kroner, 2001). Part B of the MCAA is composed of four subscales that measure different thematic or content-oriented factors in antisocial attitudes. These include attitudes toward Violence (e.g., “It’s understandable to hit someone who insults you”), Entitlement (e.g., “Taking what is owed you is not really stealing”), Antisocial Intent (e.g., “For a good reason, I would commit a crime”) and Associates (e.g., “I have a lot in common with people who break the law”). Each of the factors can be aggregated to create a Total MCAA score, with higher scores indicating more severe general antisocial attitudes (Mills & Kroner, 2001).

A small number of studies have used the MCAA to examine associations between within-treatment change in antisocial attitudes and reoffending (Howard & van Doorn, 2018; Juarez & Howard, 2018; Kingston & Olver, 2018; Kroner & Yessine, 2013), with mixed results. For example, a study by Kroner and Yessine (2013) examined within-treatment change on the MCAA among a sample of 118 individuals with general offenses who had completed a treatment program targeting antisocial attitudes. The authors found that reliable change in scores on the Associates factor was significantly correlated with likelihood of any reoffending.

More recently, Howard and van Doorn (2018) examined the predictive validity of the MCAA in a large sample of adult males who completed behavior change programs while supervised in custody or the community (N = 1,858). Results indicated that the Total score in addition to the Entitlement, Antisocial Intent, and Associates factors predicted general reoffending at pretreatment. However, analyses of within-treatment change did not find any significant associations with odds of general reoffending. The authors suggested that responses on the MCAA may have been affected by increased response bias after treatment, whereby participants may have been more inclined to underreport the severity of their antisocial attitudes to demonstrate treatment gains or obtain benefits such as parole (see also Hanson & Wallace-Capretta, 2000; Soyer et al., 2017).

Factor Specificity in Antisocial Attitudes

Howard and van Doorn (2018) also proposed that assessing relationships between within-treatment change and recidivism may be affected by variability in sample characteristics and the relevance of measured risk factors to specific individuals and types of reoffending. The construct of antisocial attitudes is multifactorial (Mills et al., 2002, 2004), and different factors may not be consistently represented across people who are convicted of different offenses or have predictive validity for different types of offending, such as sexual or violent reoffending as compared with any reoffending. Considering that studies of within-treatment change on the MCAA have primarily examined any reoffending outcomes among heterogeneous samples of individuals with convictions of general offenses, this may increase measurement error in detecting relationships between change in a given domain of antisocial attitudes and likelihood of recidivism (Howard & van Doorn, 2018).

Differences in the presentation of risk factors across people convicted of different offenses are well established in the literature (e.g., Craig et al., 2006), and this extends to content domains of antisocial attitudes assessed by the MCAA. For example, attitudes toward entitlement, or what the person feels they are owed or deserve, are common features of offense supportive cognitions among individuals with convictions for sexual offenses and have been associated with risk of sexual recidivism (Hanson & Harris, 2000; Pemberton & Wakeling, 2009). Perhaps unsurprisingly, attitudes supportive of violence have a close theoretical relationship with perpetration of violent behavior and risk of future violence (Douglas & Skeem, 2005; Nunes et al., 2015). Consistent with this, previous research has indicated that the Violence factor is most predictive of verbal and physical aggression across MCAA measures (Van Hiel et al., 2007) and can have predictive validity for violent but not general reoffending (Mills et al., 2004). More recently, Kingston and Olver (2018) found that within-treatment change on the Antisocial Intent and Associates factors of the MCAA were significantly associated with general reoffending only, whereas change on the Violence factor was associated with violent reoffending only. While the evidence is mixed, the available literature provides some indication that factors on the MCAA, and change in those factors over treatment, can have predictive specificity for certain reoffending outcomes that do not generalize to other definitions of recidivism.

The Present Study

The aim of this study was to examine the relationship between within-treatment change in antisocial attitudes and reoffending by utilizing a large sample of adult males convicted of violent offenses who had completed MCAA assessments before and after participating in treatment programs. We hypothesized that by assessing within-treatment change in a comparatively homogeneous sample of violent males, and in reference to relevant specific outcomes (violent reoffending in addition to general reoffending), the sensitivity of MCAA factors and change in those factors for recidivism may be improved. In doing so, we also intended that the study would contribute to existing literature on how different content domains of antisocial attitudes correspond to risk of violent and other reoffending outcomes among individuals with convictions for violent offending. On the balance of existing research (e.g., Mills et al., 2004) we expected that scores on the MCAA Violence factor in particular, and change in that factor over treatment, would have greater predictive validity for violent reoffending compared with general reoffending.

Method

Participants

The sample consisted of 2,337 adult males who were under Corrective Services New South Wales (CSNSW) supervision in custody (n = 1,160; 49.6%) or community (n = 1,177; 50.4%) settings between January 2015, and May 2018. To be eligible for the study, individuals were required to have an index conviction for violent offenses; to have completed treatment in one of the EQUIPS suite of programs; and to have completed valid administrations of the MCAA prior to and after participating in programs. Individuals completing custodial orders also needed to have been released from custody to allow for time at risk of reoffending. The mean age of the sample was 32.92 years (SD = 9.12), with a range of 18 to 67 years. Thirty-three percent (n = 777) identified as having Aboriginal or Torres Strait Islander (ATSI) background. Ethics approval to conduct this research was approved by the Human Research Ethics Committee (HREC) of the University of New South Wales and the Ethics Committee of CSNSW.

EQUIPS is a set of four therapeutic programs delivered by CSNSW to address criminogenic needs among individuals supervised within custody- and community-based settings. Each program is delivered in group format and adheres to cognitive behavioral therapy (CBT) and RNR principles, with a focus on addressing attitudes and other cognitions that are relevant to their criminogenic needs. Each of the programs consists of 20 two-hour sessions totaling 40 hours of intervention. While each EQUIPS program is designed as a standalone intervention, they can be delivered in combinations of multiple programs according to the person’s individual needs and are intended to be integrated into other schedules of intervention as part of their routine case management in custody or the community.

EQUIPS Foundation is available for all offense types and targets dynamic risk factors associated with general offending. EQUIPS Addiction specifically targets motivation for abstinence and understanding of the relationship between the person’s substance use and offending behavior. EQUIPS Domestic Abuse is an offense-specific program for individuals with a history of intimate partner violence and focuses on increasing self-management and relationship skills, and responsibility for abuse. EQUIPS Aggression is available for individuals who have a violence-related offense and targets instrumental and expressive aggression by strengthening emotion regulation and self-management skills.

Individuals eligible for the EQUIPS programs included those assessed as medium or higher risk of general reoffending as measured by the Level of Service Inventory–Revised (LSI-R; Andrews & Bonta, 1995). Individuals were excluded from programs if they presented with active psychotic symptoms and/or alcohol or drug intoxication symptoms. The proportion of the study sample who completed each EQUIPS program are as follows: 25.9% completed EQUIPS Addiction, 18.5% completed EQUIPS Aggression, 27.9% completed EQUIPS Domestic Abuse, and 27.7% completed EQUIPS Foundation. On average, participants completed EQUIPS programs over a median period of 65 days, during which they completed one program corresponding to a total of 40 hr of intervention.

A small number of studies have evaluated the effectiveness of EQUIPS programs on recidivism outcomes. A study by Blatch and colleagues (2016) found that individuals supervised in the community who participated in an early version of EQUIPS Domestic Abuse were 15% less likely to reoffend than matched nonparticipants. Conversely, a subsequent reevaluation concluded that EQUIPS Domestic Abuse did not significantly affect recidivism outcomes for individuals with convictions for domestic violence offenses in the community (Rahman & Poynton, 2018). In a more comprehensive examination of EQUIPS treatment pathways for individuals with domestic violence convictions, a study by Zhang et al. (2019) found treatment effects on recidivism for the EQUIPS Domestic Abuse and Aggression programs but not for the Addiction and Foundation programs.

Measures

Measures of Criminal Attitudes and Associates (MCAA)

The MCAA is composed of two parts; part A is a measure of the respondent’s affiliation with criminal associates, and part B assesses the respondent’s antisocial attitudes (Mills & Kroner, 2001). Given the focus on antisocial attitudes, our study focused on responses to part B only. Part B is composed of 46 items which require respondents to give dichotomous responses (true /false) indicating their endorsement of attitudes across four factors, including Violence (12 items), Entitlement (12 items), Antisocial Intent (12 items), and Associates (10 items). The Violence factor assesses the respondent’s endorsement of, and willingness to use, violence as a method to attain a desired goal. The Entitlement factor measures beliefs about what the respondent deserves or is owed to them. The Antisocial Intent factor measures beliefs about the respondent’s likelihood of engaging in future antisocial behaviors, while the Associates factor assesses attitudes toward peers who are involved in antisocial or criminal behaviors.

The 46 items of MCAA part B can be summed to obtain a Total MCAA score ranging from 0 to 46. Higher scores indicate greater antisocial attitudes, and scores equal to or greater than 23 are considered elevated (Mills & Kroner, 2001). The MCAA has shown good reliability and support for the proposed factorial structure (Bäckström & Björklund, 2008; Mills et al., 2002; Mills & Kroner, 2001). In the current study, internal consistency statistics were α = .85 for the Violence factor; α = .69 for the Entitlement; α = .82 for Antisocial Intent; α = .75 for Associates; and α = .91 for the MCAA Total score.

Recidivism

Recidivism outcomes were measured using NSW criminal courts finalized conviction data obtained from the NSW Bureau of Crime Statistics and Research (BOCSAR). Separate outcome variables were generated for any (general) recidivism and violent recidivism. General recidivism was defined as any new proven offense with conviction during the survival period. Violent recidivism was defined as conviction for a new violent offense in the survival period, in accordance with Australian and New Zealand Standard Offense Classification (ANZSOC) codes. Violent offenses included homicide and related offenses (ANZSOC 01); assault-related offenses (ANZSOC 02) including sexual assault (ANZSOC 311-312); abduction, kidnapping, and threatening behavior (ANZSOC 05); and robbery (ANZSOC 611-612).

The survival period for individuals who completed treatment in custody was calculated as the number of days between release from custody to the date of their reoffense or, if no reoffense was recorded, to the recidivism data censoring date of May 31, 2018. Survival period for those supervised in the community was calculated as the number of days between completion of treatment to the date of reoffense or, if no reoffense was recorded, to the data censoring date. Survival period was calculated separately for each of the reoffending variables and adjusted for time spent in custody for reasons unrelated to the category of reoffending, to reflect the number of days the person was at active risk in the community.

Thirty percent (n = 689; 29.5%) of the sample committed a new general offense and 11.8% (n = 276) committed a new violent offense during the survival period. The mean survival period was 335.11 days (SD = 253.63) for general recidivism and 407.07 days (SD = 275.12) for violent recidivism. Violent reconvictions included assaults and other acts intended to cause injury (89.9%); abduction, harassment, and related offenses (12.3%); dangerous or negligent acts endangering others (3.6%); robbery, extortion, and related offenses (2.5%); and sexual assault (1.1%).

Analytic Plan

Within-Treatment Change

Two methods of assessing within-treatment change were employed for the purposes of this study. These included computation of residual change scores (RCS) and measures of clinically significant change (CSC). RCS were used as a method of assessing differences between pretreatment and posttreatment while adjusting for variability in individuals’ baseline scores prior to treatment. RCS were computed by first calculating simple difference scores for each of the MCAA Total and factor scores (posttreatment − pretreatment). For each of the measures a linear regression was then conducted, whereby simple difference scores were regressed onto pretreatment scores. The standardized residual from that regression was then used as the measure of change. RCS were used because the magnitude of possible within-treatment change is influenced by the pretreatment score (see Beggs & Grace, 2011); consistent with this, we found significant correlations between MCAA pretreatment scores and magnitude of simple difference scores (rs = .44–.55; ps <.001) in our sample. RCS were interpreted so that larger negative scores indicated greater reductions in antisocial attitudes over treatment and larger positive scores indicated increases in antisocial attitudes over treatment.

CSC analyses were also conducted to examine the statistical reliability and clinical relevance of within-treatment change (e.g., Nunes et al., 2011). CSC analyses involved two key steps: First, cutoff scores were used to define functional and dysfunctional ranges of scores on each of the MCAA measures; and second, indices of reliable change were computed to indicate whether change between pretreatment and posttreatment was statistically significant. Following other recent studies of within-treatment change, we applied an adapted version of cutoff B for establishing functional thresholds (Howard & van Doorn, 2018; Nunes et al., 2016; Wakeling et al., 2013). The functional threshold for each MCAA factor was defined as the mean derived from nonoffending community samples plus one standard deviation (Jacobson et al., 1984; Nunes et al., 2016). To derive functional means for the MCAA, we used normative data from a sample of 60 university students reported by Mills and Kroner (2001). For each of the MCAA factors, scores below the functional threshold were indicative of antisocial attitudes that were within functional ranges and scores above the functional threshold were indicative of dysfunctional or clinical elevations on the factor. A reliable change index (RCI) was then calculated for respondents’ scores on each of the MCAA factors (Jacobson & Truax, 1991). Reliable change is equal to the difference between a person’s posttreatment (x₂) and pretreatment scores (x₁) divided by the standard error of the difference (S_diff):

RC = \frac{x_{2} - x_{1}}{S_{diff}} .

The standard error of the difference was calculated using the following formula:

S_{diff} = \sqrt{2 {(S_{E})}^{2}} .

To calculate the standard error of the difference it was first necessary to compute the standard error of measurement ${(S}_{E})$ using the following formula:

S_{E} = s_{1} \sqrt{1 - r_{xx}} .

Reliability coefficients (r_xx) were obtained from Mills and Kroner (2001) who reported 4-week test–retest reliability coefficients for a sample of 41 individuals. The standard deviation $(s_{1})$ in this formula was also obtained from the normative sample of students reported by Mills and Kroner (2001). Jacobson and Truax (1991) suggested that change that exceeds an RCI of 1.96 (> 1.96 or < −1.96) was considered statistically significant.

Calculations of functional thresholds and RCI were then used to categorize individuals into one of four CSC categories for each of the MCAA measures. Individuals in the sample were categorized as recovered if their posttreatment scores were below the functional threshold and they showed reliable change between pretreatment and posttreatment. Individuals were categorized as improved if they showed reliable change although their posttreatment score remained above the functional threshold. Individuals were categorized as unchanged if they did not show reliable change between pretreatment and posttreatment. Finally, individuals were categorized as deteriorated if they showed reliable change although in a direction that indicated more severe antisocial attitudes after treatment. In accordance with the theoretical and statistical rationale for CSC (e.g., Jacobson et al., 1984; Jacobson & Truax, 1991), categories of change were only computed for those who had dysfunctional scores on a given MCAA factor prior to treatment.

Reoffending Analyses

A series of Cox proportional hazard regression models were conducted to examine associations between MCAA scores or change indices and reoffending outcomes, while accounting for individual differences in survival period. Reported hazard ratios represent the risk of general or violent recidivism occurring for every one unit increase in the MCAA measure of interest, adjusted for survival time. Analyses treated custody-based and community-based persons as a single sample to maximize statistical power and generalizability. While this approach was supported by the standardized format of EQUIPS and MCAA administration across contexts, interpretation of reoffending analyses may be complicated by differences in definition of survival period between those who were treated in custody (from release from custody) and those who were treated in the community (from completion of the program). To account for this, we calculated the interval between treatment end and the start of the survival period for persons that were custody-based (M = 114.70 days; SD = 148.76) and community-based (M = 0 days; SD = 0), which was included as a covariate in all models of reoffending outcomes. To account for the multiple planned comparisons and large sample size, a conservative alpha of .01 (p <.01) was used to determine significance of hazard ratios.

Results

Associations Between MCAA Scores and Reoffending

Table 1 gives the results of Cox regression analyses for associations between raw MCAA scores and reoffending outcomes at pre- and posttreatment. All MCAA measures were significantly associated with hazard of general reoffending at both pretreatment and posttreatment time points, with the exception of the Entitlement posttreatment measure. Associations were in the expected direction so that higher scores (indicating more severe antisocial attitudes) were predictive of higher odds of general reoffending.

Table 1:

Associations Between Pretreatment and Posttreatment MCAA Scores With General and Violent Reoffending

Administration	MCAA measure	n	M (SD)	Association with reoffending
				General		Violent
				Exp(B)	95% CI	Exp(B)	95% CI
Pretreatment	Total	2328	22.16 (9.54)	1.03**	[1.02, 1.03]	1.01	[1.00, 1.03]
	Violence	2330	4.82 (3.38)	1.05**	[1.03, 1.07]	1.02	[0.98, 1.06]
	Entitlement	2334	5.73 (2.69)	1.05**	[1.02, 1.08]	1.02	[0.97, 1.07]
	Antisocial Intent	2333	5.03 (3.34)	1.09**	[1.06, 1.11]	1.05*	[1.01, 1.09]
	Associates	2335	6.58 (2.49)	1.09**	[1.06, 1.13]	1.03	[0.98, 1.08]
Posttreatment	Total	2330	18.77 (9.05)	1.02**	[1.02, 1.03]	1.01	[1.00, 1.03]
	Violence	2332	3.62 (3.13)	1.06**	[1.03, 1.08]	1.04	[1.00, 1.08]
	Entitlement	2334	4.92 (2.69)	1.04	[1.01, 1.07]	1.02	[0.98, 1.07]
	Antisocial Intent	2334	3.95 (3.21)	1.08**	[1.05, 1.10]	1.06**	[1.02, 1.10]
	Associates	2334	6.26 (2.51)	1.08**	[1.04, 1.11]	1.02	[0.98, 1.08]

Note. MCAA = Measures of Criminal Attitudes and Associates; SD = standard deviation; Exp(B) = hazard ratio; CI = confidence interval.

p < .01. **p < .001.

Fewer MCAA measures were predictive of violent reoffending. The Antisocial Intent factor had a significant positive association with hazard of violent reoffending at pretreatment and again at posttreatment, so that each unit increase in score on this factor corresponded with a 5% increase and a 6% increase in hazard of violent reoffending, respectively. None of the other MCAA measures were significantly predictive of violent reoffending at pre- and posttreatment.

Associations Between Within-Treatment Change and Reoffending

Residual Change Scores

Table 2 gives Cohen’s d statistics indicating the magnitude of change in each of the MCAA scores between pretreatment and posttreatment. It can be seen that the measures showed change of low to moderate effect size on average, which was statistically significant in each case. The results of Cox regression models examining associations between RCS and reoffending are also given in Table 2. The Violence and Antisocial Intent factors had significant relationships with hazard of general recidivism, with increases in RCS (indicating growth in antisocial attitudes over treatment) on each of these factors associated with an 11% increase in the adjusted odds of general reoffending. RCS for each of the MCAA Total and factorial measures had nonsignificant associations with violent reoffending.

Table 2:

Magnitude of Change Between Pretreatment and Posttreatment Scores, and Associations Between Residual Change Scores and General and Violent Reoffending

MCAA measure	n	d	RCS analyses
			General reoffending		Violent reoffending
			Exp(B)	95% CI	Exp(B)	95% CI
Total	2321	.44**	1.09	[1.01, 1.18]	1.10	[0.97, 1.23]
Violence	2325	.39**	1.11*	[1.03, 1.20]	1.10	[0.98, 1.24]
Entitlement	2331	.30**	1.04	[0.96, 1.12]	1.04	[0.92, 1.17]
Antisocial Intent	2330	.37**	1.11*	[1.03, 1.20]	1.13	[1.01, 1.27]
Associates	2332	.15**	1.06	[0.98, 1.15]	1.01	[0.89, 1.14]

Note. MCAA = Measures of Criminal Attitudes and Associates; d = Cohen’s d; RCS = residual change score; Exp(B) = hazard ratio; CI = confidence interval.

p < .01. **p < .001.

Clinically Significant Change

The proportions of the sample classified in each CSC category for the MCAA measures are presented in Table 3. As previously mentioned, individuals were only assigned to a CSC category if they recorded scores in dysfunctional ranges at pretreatment. This included 39.4% (n = 893) of all individuals in the sample for the MCAA Total score; 24.1% (n = 562) for the Violence factor; 37.2% (n = 869) for the Entitlement factor; 34.7% (n = 810) for the Antisocial Intent factor; and 55.8% (n = 1304) for the Associates factor.

Table 3:

Distribution of Clinically Significant Change Categories Among Individuals Who Had Dysfunctional Scores at Pretreatment

MCAA measure	CSC categories n (%)
MCAA measure	Recovered	Improved	Unchanged	Deteriorated
Total	324 (36.3%)	58 (6.5%)	491 (55.0%)	20 (2.2%)
Violence	246 (43.8%)	—	316 (56.2%)	—
Entitlement	337 (38.8%)	43 (4.9%)	477 (54.9%)	12 (1.4%)
Antisocial Intent	297 (36.7%)	9 (1.1%)	489 (60.4%)	15 (1.9%)
Associates	123 (9.5%)	—	1181 (90.6%)	—

Note. MCAA = Measures of Criminal Attitudes and Associates; CSC = clinically significant change.

The majority of people who were classified as dysfunctional at pretreatment were assigned to the recovered (range = 9.5%–43.8%) or the unchanged (range = 54.9%–90.6%) categories of clinically significant change. Few were categorized as improved (range = 1.1%–6.5%) or deteriorated (range = 1.4%–2.2%) across the MCAA measures.

Given the small cell sizes for the improved and deteriorated categories of clinically significant change, we examined associations between within-treatment change and reoffending by conducting pairwise comparisons for those in the recovered and unchanged categories. In the event that clinically significant change in MCAA scores is associated with reoffending, we expected that individuals in the recovered category would have significantly lower hazard of reoffending compared with those in the unchanged category.

The results of Cox regression models comparing the groups are given in Table 4. Results showed that when compared with the unchanged category (the reference group), individuals in the recovered category did not have significantly different hazard of general or violent reoffending on any of the MCAA measures.

Table 4:

Comparisons Between Individuals in the Recovered and Unchanged Categories of Clinically Significant Change on General and Violent Reoffending

MCAA measure	General reoffending		Violent reoffending
MCAA measure	Exp(B)	95% CI	Exp(B)	95% CI
Total	1.15	[0.90, 1.46]	1.18	[0.80, 1.75]
Violence	0.99	[0.75, 1.31]	0.85	[0.53, 1.36]
Entitlement	1.01	[0.79, 1.28]	0.97	[0.66, 1.44]
Antisocial Intent	1.20	[0.94, 1.52]	1.10	[0.75, 1.62]
Associates	0.99	[0.72, 1.37]	0.92	[0.55, 1.52]

Note. MCAA = Measures of Criminal Attitudes and Associates; Exp(B) = hazard ratio; CI = confidence interval.

Discussion

While antisocial attitudes have been established as an important risk factor for recidivism and are commonly targeted in interventions, there is little direct evidence to show that change in antisocial attitudes over treatment have an impact on peoples’ likelihood of reoffending. This study examined whether self-reported antisocial attitudes on the MCAA were associated with reoffending outcomes in a large sample of individuals convicted of violent offenses. In particular, we assessed the predictive validity of change between pretreatment and posttreatment scores for both general and violent reoffending.

Associations Between MCAA Raw Scores and Reoffending

Analyses of individual MCAA scores replicated previous studies (e.g., Howard & van Doorn, 2018; Mills et al., 2004) by showing that a number of measures had significant predictive validity for general reoffending when assessed prior to treatment. In contrast, only Antisocial Intent significantly predicted violent reoffending at pretreatment. The Antisocial Intent factor may have comparatively robust associations with multiple reoffending outcomes because behavioral intentions tend to be better predictors of future behavior than attitudes (e.g., Ajzen, 1988).

Notably, the Violence factor was associated with hazard of general but not violent reoffending at pretreatment. This is in contrast to previous findings that endorsement of violence on the MCAA can predict future violent behaviors (Mills et al., 2004; Nunes et al., 2015; Van Hiel et al., 2007). One interpretation is that while attitudes supportive of violence are an important risk factor for violent offending (e.g., Douglas & Skeem, 2005), in many cases their responses on the MCAA may be indicative of propensity toward antisocial behaviors in general, including but not limited to violence. People with a history of violence-related behavior and needs often exhibit generalization in future reoffending (Wan & Weatherburn, 2016). A related consideration is that violent reoffending was less prevalent than general reoffending (11.8% vs. 29.8%), which is likely to have impacted the power of analyses across outcomes.

In addition, a number of MCAA measures had significant associations with reoffending when assessed at posttreatment. Previous research has tended to find weaker relationships between self-reported dynamic risk factors and reoffending at posttreatment than at pretreatment (Barnett et al., 2012; Hanson & Wallace-Capretta, 2000; Howard & van Doorn, 2018). It is possible that individuals vary in their tendencies toward socially desirable responding before and after treatment, which may affect the validity of their self-reports at different points of measurement (e.g., Soyer et al., 2017). For example, Juarez and Howard (2018) found a “rebound effect” marked by endorsement of antisocial attitudes that decreased at the end of one program and increased at the start of a second program. Evidence for the predictive validity of posttreatment scores found in the current study is a promising indication that individuals’ self-reports may have been relatively unaffected by response bias. Potential moderators of response bias in this study will be explored in greater detail in the following section.

Within-Treatment Change and Reoffending

Primary analyses in this study focused on whether within-treatment change in scores on the MCAA was associated with reoffending outcomes. After adjusting for variance in pretreatment scores using RCS techniques, change on the Violence factor and on the Antisocial Intent factor was significantly associated with hazard of general reoffending. Associations were in the expected direction, so that greater reductions in antisocial attitudes corresponded with lower odds of reoffending. In contrast, RCS scores on all MCAA measures did not predict violent recidivism. The findings appear to support proposals that attitudes relating to violence and antisocial intent are important risk factors, and potentially viable treatment targets, for individuals with convictions for violent offenses; however, for many individuals these constructs may be more indicative of general antisociality as opposed to violence risk when assessed by the MCAA.

Findings for a significant association between within-treatment change and reoffending is uncommon in the literature (Banse et al., 2013; Serin et al., 2013) and warrants further exploration. In line with the aims of our study, it is possible that the sensitivity of within-treatment change analyses was improved by the relative homogeneity of the study sample. Individuals with violent convictions who are eligible for the same suite of intervention programs may have a more uniform range and severity of dynamic risk factors that are related to their offending behavior compared with cohorts of individuals with any convictions. As a result, it may be expected that elevations on a given construct of risk, and change in that risk over treatment, may have more consistent relationships with outcome as the homogeneity of the sample increases.

The degree of heterogeneity in samples may also have a relationship with tendencies toward response bias when completing self-report measures. Socially desirable responding has been found to have an inverse correlation with risk of reoffending (e.g., Mills & Kroner, 2005, 2006), suggesting that individuals who are at higher risk are more likely to give self-reports that are less affected by response bias. As a result, the predictive validity of self-reports may vary across samples in accordance with their risk profile (Howard & van Doorn, 2018). Meaningful change may have been detectable in the current study as a result of sampling priority individuals convicted of violent offenses who were required to meet risk-related eligibility requirements to participate in EQUIPS.

A number of previous studies have examined relationships between within-treatment change in antisocial attitudes and reoffending with homogeneous groups, typically with nonsignificant results (Banse et al., 2013). In this case, a number of additional methodological factors may also contribute to variability in findings, including the risk relevance of measures (e.g., Higgs et al., 2020; Nunes et al., 2016; Wakeling et al., 2013), and sample size limitations (e.g., Klepfisz et al., 2014). A large number of studies of self-reported within-treatment change have also utilized samples of individuals with histories of sexual offending, which may be prone to impacts associated with assessment of constructs that carry severe social stigma (Tierney & McCabe, 2001) as well as low base rates of reoffending.

An additional source of variance across studies relates to how within-treatment change is assessed. In this study, the significant associations found in RCS analyses were not replicated in CSC analyses, suggesting that CSC was less sensitive to valid change. It has been established that measurement of within-treatment change is critically influenced by variation in respondents’ baseline scores (e.g., Beggs, 2010; Beggs & Grace, 2011). When compared with RCS methods, CSC calculations may underadjust for pretreatment variance if a large range of scores are categorized as dysfunctional. In this case, individuals with extremely severe pretreatment needs may be treated as equivalent to those with moderate needs. CSC analyses may then be biased toward detecting change among those extremely dysfunctional individuals because they are able to show a greater magnitude of change, including as a result of measurement error. Consistent with this, studies have found clinically significant improvement to be associated with increased likelihood of reoffending (Nunes et al., 2016; Wakeling et al., 2011). On the balance of these findings, we suggest that future studies of within-treatment change would benefit from employing methods that more systematically account for pretreatment variability, such as RCS.

Limitations

Some limitations of the study are noted. While the sample was drawn from participants of EQUIPS programs which have standardized models of delivery, and each aim to address antisocial attitudes, the four programs nonetheless have different treatment targets. Our analyses may have been affected by differences in the specificity of each EQUIPS program for constructs assessed by the MCAA, and how they relate to risk of general or violent reoffending (Higgs et al., 2020). In addition, each EQUIPS program is relatively low intensity, and there was a small degree of variance in dosage across participants that was not accounted for in analyses. It may be argued that within-treatment change analyses are designed to be robust to variation in the specificity and dosage of interventions; for example, limited or nonspecific intervention for a given risk factor should result in a measurably low magnitude of change on that factor. However, we acknowledge that higher intensity interventions are more likely to derive psychologically meaningful and long-term change, which could in turn be more readily distinguished from error by assessments of within-treatment change.

A related limitation is that antisocial attitudes were the only dynamic risk factor accounted for in this study. While factors on the MCAA were found to have significant predictive validity for general reoffending, it is possible that change in these factors alone was not sufficient to have a detectable relationship with likelihood of violent reoffending. Individuals with convictions for violent offenses have been found to exhibit different risk profiles compared with individuals with nonviolent offenses including elevations in aggression, impulsivity and hostility (Craig et al., 2006). In this case, assessment of within-treatment change in multiple domains of risk may be more sensitive to violent reoffending outcomes compared with change in single MCAA factors.

Our study employed a large sample to improve statistical power and generalizability across individuals convicted of violent offenses, including those with different profiles of criminogenic needs, as well as those receiving intervention in custody as well as the community. While we adjusted for statistical variation in survival time for persons in custody and the community, it is possible that these cohorts are subject to other differences that influence their self-reports on the MCAA or associations between within-treatment change and reoffending outcomes. A previous study found that the predictive validity of the MCAA varies between custody-based and community-based individuals; however, this may be attributable to sampling rather than contextual factors (Howard & van Doorn, 2018). We also note that the sample did not include females or those with a low overall risk of reoffending, and the results may not be representative of these individuals.

Finally, as with other studies in this field, the validity of CSC analyses of within-treatment change is contingent on the quality of normative data that are used to calculate functional thresholds and the reliable change index. In this study, CSC calculations were derived from a single sample of 60 university students who were assessed during initial validation of the MCAA (Mills & Kroner, 2001). Use of community samples to calculate functional norms is arguably less problematic than using individuals convicted of offenses as an estimate of dysfunctional norms on a given construct of dynamic risk (Nunes et al., 2016). However, we acknowledge that the small size of the Mills and Kroner (2001) sample may affect the reliability of normative data and other statistics derived from the study.

Conclusion

This study adds to the existing literature on within-treatment change by examining relationships between self-reported antisocial attitudes on the MCAA and reoffending among a large, statistically robust sample of males convicted of violent offenses. When considered in conjunction with previous findings, the results suggest that change in dynamic risk factors over treatment can be meaningfully associated with likelihood of reoffending. However, measurement of this association may be confounded by a range of methodological challenges, including changes in individuals’ response styles before and after treatment (see Juarez & Howard, 2018); differences in the extent to which assessed risk factors are relevant to different populations and types of reoffending; the analytical techniques used to assess change; and low statistical power. Use of a relatively homogeneous sample in this study may have ameliorated some of these challenges by promoting greater consistency in presenting risk factors and their relationships to reoffending behaviors, as well as the situational and individual characteristics that could influence bias in self-reports.

This study has practical implications by showing that various content domains of antisocial attitudes in the MCAA have predictive validity for reoffending among individuals with convictions for violent offenses. These findings give support for the importance of related risk factors for people with violent offenses and suggest the MCAA could have utility in assessing their intervention needs. However, it appears that MCAA scores may be a more consistent indicator of general antisocial outcomes for individuals with violent offenses than of violent reoffending in particular. In addition, given the inconsistent and modest associations between within-treatment change and reoffending, it is not possible to conclude that change in MCAA scores are a reliable indicator of how a person’s risk has been influenced by treatment.

Research into the dynamics and validity of within-treatment change in risk factors is a developing area (e.g., Serin et al., 2013), and further study is needed to improve methods for assessing change. Reliance on self-report in treatment settings appears to be a key challenge in this regard. One recommended solution is to adjust for self-report bias using measures of socially desirable responding (e.g., Heffernan et al., 2019). However, existing measures have been criticized as assessing constructs that are meaningfully related to reoffending risk (Uziel, 2010), with the result being that adjusting for such measures impairs the sensitivity of risk assessments (Mills & Kroner, 2005, 2006). Other promising avenues could involve further development of robust behavioral or clinician-rated indexes of change, or adopting assessment schedules that work to minimize demand effects. The results of this and other studies suggest that improving methods of measuring and analyzing change are a critical condition for progressing toward more fundamental insights about dynamic risk factors and how they change over treatment to influence reoffending outcomes.

Footnotes

Acknowledgements

The authors are thankful to the University of New South Wales, particularly Professor Richard Kemp, for overseeing this study. It is noted that the second author was employed by CSNSW at the time of completing this study.

ORCID iD

Tanyia Juarez

Tanyia Juarez is a psychologist currently practicing in a therapeutic role, working with individuals convicted of violent offenses. Her research interests include assessing violent recidivism risk and the effectiveness of treatment programs for individuals convicted of violent offenses.

Mark V. A. Howard is manager of research and evaluation at the Corrections Research Evaluation and Statistics (CRES) unit of Corrective Services NSW. He is a forensic psychologist and has previously worked in assessment and expert evidence matters for the NSW Criminal Courts. In collaboration with a team of researchers and external partners, he is currently engaged in a number of research projects, with interests in violence within prisons, sexual offending, mechanisms of change in treatment, and program evaluation.

References

Ajzen

(1988). Attitudes, personality and behavior. Dorsey Press.

Allport

G. W.

(1935). Attitudes. In Murchison

(Ed.), A handbook of social psychology (pp. 798–844). Clark University Press.

Andrews

D. A.

Bonta

(1995). Level of Supervision-Revised (LSI-R): An offender assessment system. User’s guide. Multi-Health Systems.

Andrews

D. A.

Bonta

Hoge

(1990). Classification for effective rehabilitation: Rediscovering psychology. Criminal Justice and Behavior, 17(1), 19–52. https://doi.org/10.1177/0093854890017001004

Andrews

D. A.

Zinger

Hoge

R. D.

Bonta

Gendreau

Cullen

F. T.

(1990). Does correctional treatment work? A clinically relevant and psychologically informed meta-analysis. Criminology, 28(3), 369–404. https://doi.org/10.1111/j.1745-9125.1990.tb01330.x

Bäckström

Björklund

(2008). The Measures of Criminal Attitudes and Associates (MCAA): Further testing of structural and criterion-related validity. Criminal Justice and Behavior, 35(11), 1398–1410. https://doi.org/10.1177/0093854808322239

Banse

Koppehele-Gossel

Kistemaker

L. M.

Werner

L. A.

Schmidt

A. F.

(2013). Pro-criminal attitudes, intervention, and recidivism. Aggression and Violent Behavior, 18(6), 673–685. https://doi.org/10.1016/j.avb.2013.07.024

Barnett

G. D.

Wakeling

Mandeville-Norden

Rakestrow

(2012). How useful are psychometric scores in predicting recidivism for treated sex offenders? International Journal of Offender Therapy and Comparative Criminology, 56, 420–446. https://doi.org/10.1177/0306624X11403125

Beggs

S. M.

(2010). Within-treatment outcome among sexual offenders: A review. Aggression and Violent Behavior, 15(5), 369–379. https://doi.org/10.1016/j.avb.2010.06.005

10.

Beggs

S. M.

Grace

R. C.

(2011). Treatment gain for sexual offenders against children predicts reduced recidivism: A comparative validity study. Journal of Consulting and Clinical Psychology, 79(2), 182–192. https://doi.org/10.1037/a0022900

11.

Blatch

O’Sullivan

Delaney

van Doorn

Sweller

(2016). Evaluation of an Australian domestic abuse program for offending males. Journal of Aggression, Conflict and Peace Research, 8, 4–20. https://doi.org/10.1108/JACPR-10-2015-0194

12.

Bonta

Andrews

D. A.

(2017). The psychology of criminal conduct (6th ed.). Taylor & Francis.

13.

Craig

L. A.

Browne

K. D.

Beech

Stringer

I. A. N.

(2006). Differences in personality and risk characteristics in sex, violent and general offenders. Criminal Behavior and Mental Health, 16(3), 183–194. https://doi.org/10.1002/cbm.618

14.

Douglas

K. S.

Skeem

J. L.

(2005). Violence risk assessment: Getting specific about being dynamic. Psychology, Public Policy, and Law, 11(3), 347–383. https://doi.org/10.1037/1076-8971.11.3.347

15.

Dowden

Andrews

D. A.

(2000). Effective correctional treatment and violent reoffending: A meta-analysis. Canadian Journal of Criminology, 42(4), 449–468. https://doi.org/10.3138/cjcrim.42.4.449

16.

Eisenberg

M. J.

van Horn

J. E.

Dekker

J. M.

Assink

van der Put

C. E.

Hendriks

Stams

G. J. J. M

. (2019). Static and dynamic predictors of general and violent criminal offense recidivism in the forensic outpatient population: A meta-analysis. Criminal Justice and Behavior, 46(5), 732–750. https://doi.org/10.1177/0093854819826109

17.

Gendreau

Little

Goggin

(1996). A meta-analysis of the predictors of adult offender recidivism: What works! Criminology, 34(4), 575–608. https://doi.org/10.1111/j.1745-9125.1996.tb01220.x

18.

Hanson

R. K.

Harris

A. J. R.

(2000). Where should we intervene? Dynamic predictors of sexual offense recidivism. Criminal Justice and Behavior, 27(1), 6–35. https://doi.org/10.1177/0093854800027001002

19.

Hanson

R. K.

Wallace-Capretta

(2000). Predicting recidivism among male batterers. Public Works and Government Services Canada.

20.

Harris

Rice

Quinsey

(1993). Violent recidivism of mentally disordered offenders: The development of a statistical prediction instrument. Criminal Justice and Behavior, 20(4), 315–335. https://doi.org/10.1177/0093854893020004001

21.

Heffernan

Wegerhoff

Ward

(2019). Dynamic risk factors: Conceptualization, measurement, and evidence. Aggression and Violent Behavior, 48, 6–16. https://doi.org/10.1016/j.avb.2019.06.004

22.

Higgs

Olver

M. E.

Nunes

Cortoni

(2020). Risk relevance of psychometric assessment and evaluator ratings of dynamic risk factors in high-risk violent offenders. Legal and Criminological Psychology, 25(2), 219–236. https://doi.org/10.1111/lcrp.12173

23.

Howard

M. V. A.

van Doorn

(2018). Within-treatment change in antisocial attitudes and reoffending in a large sample of custodial and community offenders. Law and Human Behavior, 42(4), 321–335. https://doi.org/10.1037/lhb0000290

24.

Jacobson

N. S.

Follette

W. C.

Revenstorf

Hahlweg

Baucom

D. H.

Margolin

(1984). Variability in outcome and clinical significance of behavioral marital therapy: A reanalysis of outcome data. Journal of Consulting and Clinical Psychology, 52, 497–504. https://doi.org/10.1037/0022-006X.52.4.497

25.

Jacobson

N. S.

Truax

(1991). Clinical significance: A statistical approach to defining meaningful change in psychotherapy research. Journal of Consulting and Clinical Psychology, 59(1), 12–19. https://doi.org/10.1037/0022-006X.59.1.12

26.

Juarez

Howard

(2018). Assessing offender change over treatment: The influence of treatment context on self-reported antisocial attitudes (Research Bulletin No. 37). Corrections Research Evaluation and Statistics, Corrective Services NSW.

27.

Kabiri

Shadmanfaat

S. M.

Smith

Cochran

(2020). A social learning model of antisocial coaching behavior. International Journal of Offender Therapy and Comparative Criminology, 64(8), 860–879. https://doi.org/10.1177/0306624x19899608

28.

Kingston

Olver

(2018). Psychometric examination of treatment change among mentally disordered offenders: A risk–needs analysis. Criminal Justice and Behavior, 45(2), 153–172. https://doi.org/10.1177/0093854817743539

29.

Klepfisz

Daffern

Day

(2016). Understanding dynamic risk factors for violence. Psychology, Crime & Law, 22(1–2), 124–137. https://doi.org/10.1080/1068316X.2015.1109091

30.

Klepfisz

O’Brien

Daffern

(2014). Violent offenders’ within-treatment change in anger, criminal attitudes, and violence risk: Associations with violent recidivism. International Journal of Forensic Mental Health, 13(4), 348–362. https://doi.org/10.1080/14999013.2014.951107

31.

Kroner

D. G.

Yessine

A. K.

(2013). Changing risk factors that impact recidivism: In search of mechanisms of change. Law and Human Behavior, 37(5), 321–336. https://doi.org/10.1037/lhb0000022

32.

Lewis

Olver

M. E.

Wong

S. C. P.

(2013). The Violence Risk Scale: Predictive validity and linking changes in risk with violent recidivism in a sample of high risk offenders with psychopathic traits. Assessment, 20(2), 150–164. https://doi.org/10.1177/1073191112441242

33.

Mills

J. F.

Kroner

D. G.

(2001). Measures of Criminal Attitudes and Associates (MCAA).

34.

Mills

J. F.

Kroner

D. G.

(2005). An investigation into the relationship between socially desirable responding and offender self-report. Psychological Services, 2, 70–80. https://doi.org/10.1037/1541-1559.2.1.70

35.

Mills

J. F.

Kroner

D. G.

(2006). Impression management and self-report among violent offenders. Journal of Interpersonal Violence, 21, 178–192. https://doi.org/10.1177/0886260505282288

36.

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(3), 240–253. https://doi.org/10.1177/1073191102009003003

37.

Mills

J. F.

Kroner

D. G.

Hemmati

(2004). The Measures of Criminal Attitudes and Associates (MCAA): The prediction of general and violent recidivism. Criminal Justice and Behavior, 31(6), 717–733. https://doi.org/10.1177/0093854804268755

38.

Nunes

K. L.

Babschishin

K. M.

Cortoni

(2011). Measuring treatment changes in sex offenders: Clinical and statistical significance. Criminal Justice and Behavior, 38(2), 157–173. https://doi.org/10.1177/0093854810391054

39.

Nunes

K. L.

Hermann

Maimone

Woods

(2015). Thinking clearly about violent cognitions: Attitudes may be distinct from other cognitions. Journal of Interpersonal Violence, 30(8), 1322–1347. https://doi.org/10.1177/0886260514540329

40.

Nunes

K. L.

Pettersen

Hermann

C. A.

Looman

Spape

(2016). Does change on the MOLEST and RAPE scales predict sexual recidivism? Sexual Abuse: A Journal of Research and Treatment, 28(5), 427–447. https://doi.org/10.1177/1079063214540725

41.

Papalia

Spivak

Daffern

Ogloff

J. R. P.

(2020). Are psychological treatments for adults with histories of violent offending associated with change in dynamic risk factors? A meta-analysis of intermediate treatment outcomes. Criminal Justice and Behavior, 47(12), 1585–1608. https://doi.org/10.1177/0093854820956377

42.

Pemberton

A. E.

Wakeling

H. C.

(2009). Entitled to sex: Attitudes of sexual offenders. Journal of Sexual Aggression, 15, 289–303. https://doi.org/10.1080/13552600903097212

43.

Rahman

Poynton

(2018). Evaluation of the EQUIPS domestic abuse program. Crime and Justice Bulletin, 211, 1–16.

44.

Serin

R. C.

Lloyd

C. D.

Helmus

Derkzen

D. M.

Luong

(2013). Does intra-individual change predict offender recidivism? Searching for the Holy Grail in assessing offender change. Aggression and Violent Behavior, 18(1), 32–53. https://doi.org/10.1016/j.avb.2012.09.002

45.

Soyer

Mcneeley

Zajac

Bucklen

K. B.

(2017). Measuring the criminal mind: The relationship between intelligence and CSS-M results among a sample of Pennsylvania prison inmates. Criminal Justice and Behavior, 44(11), 1444–1461.

46.

Stefanile

Matera

Nerini

Puddu

Raffagnino

(2017). Psychological predictors of aggressive behavior among men and women. Journal of Interpersonal Violence, 36(1–2), NP920–NP941. https://doi.org/10.177/0886260517737553

47.

Stith

S. M.

Smith

D. B.

Penn

C. E.

Ward

D. B.

Tritt

(2004). Intimate partner physical abuse perpetration and victimization risk factors: A meta-analytic review. Aggression and Violent Behavior, 10(1), 65–98. https://doi.org/10.1016/j.avb.2003.09.001

48.

Tierney

D. W.

McCabe

M. P.

(2001). An evaluation of self-report measures of cognitive distortions and empathy among Australian sex offenders. Archives of Sexual Behavior, 30, 495–519. https://doi.org/10.1023/a:1010239217517

49.

Uziel

(2010). Rethinking social desirability scales: From impression management to interpersonally oriented self-control. Perspectives on Psychological Science, 5, 243–262. https://doi.org/10.1177/1745691610369465

50.

Van Hiel

Hautman

Cornelis

De Clercq

. (2007). Football hooliganism: Comparing self-awareness and social identity theory explanations. Journal of Community and Applied Social Psychology, 17(3), 169–186. https://doi.org/10.1002/casp.902

51.

Wakeling

H. C.

Beech

A. R.

Freemantle

(2013). Investigating treatment changes and its relationship to recidivism in a sample of 3773 sex offenders in the UK. Psychology, Crime & Law, 19(3), 233–252. https://doi.org/10.1080/1068316X.2011.626413

52.

Wakeling

H. C.

Freemantle

Beech

A. R.

Elliott

I. A.

(2011). Identifying predictors of recidivism in a large sample of United Kingdom sexual offenders: A prognostic model. Psychological Services, 8(4), 307–318. https://doi.org/10.1037/a0025516

53.

Walters

DeLisi

(2013). Antisocial cognition and crime continuity: Cognitive mediation of the past crime-future crime relationship. Journal of Criminal Justice, 41(2), 135–140. https://doi.org/10.1016/j.jcrimjus.2012.12.004

54.

Wan

Weatherburn

(2016). Violent criminal careers: A retrospective longitudinal study (Crime and Justice Bulletin No. 198). NSW Bureau of Crime Statistics and Research.

55.

Woessner

Schwedler

(2014). Correctional treatment of sexual and violent offenders: Therapeutic change, prison climate, and recidivism. Criminal Justice and Behavior, 41(7), 862–879. https://doi.org/10.1177/0093854813520544

56.

Zhang

Wei

Howard

Galouzis

(2019). Evaluation of EQUIPS treatment pathways for domestic violence offenders in New South Wales. Corrections Research Evaluation and Statistics, Corrective Services NSW.