Validation of the Offending-Related Attitudes Questionnaire of CRIME-PICS II Scale (Chinese)

Abstract

This study examined the factor structure, reliability, and validity of the first part of the Chinese version of the CRIME-PICS II Scale, a self-administrated instrument assessing offending-related attitudes. Data were collected from three samples: male Hong Kong young offenders, female Mainland Chinese prisoners, and Hong Kong college students. Exploratory factor analysis and confirmatory factor analysis revealed a four-factor structure that differed from the one proposed by Frude, Honess, and Maguire. The Chinese version of the scale was found to have good internal consistency (α = .90) and good test–retest reliability (r = .86) and also to present evidence of construct, concurrent, discriminant, and predictive validity. Overall, the Offending-Related Attitude Questionnaire (Chinese) proves to have great potential utility in the context of the Chinese criminal justice systems but will benefit from further validation studies.

Keywords

reliability validity psychometric properties CRIME-PICS II offending attitudes offenders Chinese

Crime is believed to be the product of distorted cognitive patterns, such as distorted perceptions of behavior itself, its consequences and impact on others, or of lifestyles that are conducive to offending behaviors (Burgoyne, 2013; Lewis, Lobley, Raynor, & Smith, 2005; Lewis, Maguire, Raynor, Vanstone, & Vennard, 2003; McGuire & Hatcher, 2001; Meek, 2012). An accurate measurement of an offender’s attitudes, perceptions, and lifestyle could be used to predict recidivism and help criminal justice professionals identify the critical factors of attempts to desist from crime and provide corresponding services to overcome setbacks (Maguire & Raynor, 2006). The criminal justice field has witnessed a rapid development and application of related instruments in recent decades. Examples include, but are not limited to, the Psychological Inventory of Criminal Thinking Styles (Walters, 1995), the Criminal Attribution Inventory (Kroner & Mills, 2004), and the Level of Service Inventory (Andrews & Bonta, 1995). From within the wide range of choices, practitioners in the correctional system value not only the predictive power of instruments but also their ease of administration and interpretation, for pragmatic reasons. Given its capability in these areas, the CRIME-PICS II Scale has stood out from the other assessment tools.

A self-report instrument for measuring attitudes and life problems related to offending, the CRIME-PICS II Scale was developed in Britain through collaboration between criminology and psychology scholars with support from probation and prison service units in England (Frude, Honess, & Maguire, 2008; Clancy et al., 2006; Raynor, 1998). Since its development in 1994, the instrument has been widely used in program evaluation in probation and custodial settings and has demonstrated its effectiveness in England and elsewhere (e.g., Burgoyne, 2013; Chui & Chan, 2011a; CRG Research Ltd., 2003; Feasey & Williams, 2009; Maguire & Raynor, 2006; McGuire & Hatcher, 2001; Raynor & Kynch, 2000). The CRIME-PICS II instrument has been deployed as a standard assessment in both accredited and nonaccredited intervention programs throughout the National Offender Management Service of England and Wales as well as in social work settings in Scotland (Feasey & Williams, 2009; Raynor, 1998).

Despite the volume of studies conducted on the instrument and its wide use in Western custodial settings, CRIME-PICS II has neither been validated nor applied in the Chinese context. The primary objective of the current study is to validate a Chinese version of the Offending-Related Attitudes Questionnaire (Part 1) of CRIME-PICS II (hereafter the Attitudes Questionnaire) in order to introduce it to measure changes in offenders’ attitudes to offending in the Chinese criminal justice system.

CRIME-PICS II (Full Version)

In the original English version of CRIME-PICS II, five factors were found to identify the beliefs and needs of offenders in two parts: the Offending-Related Attitudes Questionnaire and the Problem Inventory (Frude et al., 2008). Part 1 includes an offender’s general attitude to offending, anticipation of reoffending, victim hurt denial, and evaluation of crime as worthwhile; Part 2 measures the respondent’s own perception of problems encountered in his or her current life (Frude et al., 2008).

Assessment tools that are able to help identify offenders’ criminogenic risks and needs and to predict the likelihood of recidivism are of great value to probation and custodial services (Andrews & Bonta, 1995; Raynor, 1998). Previous studies indicate that CRIME-PICS II is indeed able to discriminate between recidivists and nonrecidivists. Offenders who show positive changes in the Attitudes Questionnaire and a decrease in the Problem Inventory are less likely to reoffend. Simply put, a lower total score on CRIME-PICS II correlates with a lower risk of reoffending (Raynor, 1998; Raynor & Kynch, 2000).

Compared with other instruments, CRIME-PICS II expands the prediction of reoffending from past events to the ones; both attitudes and self-reported problems are susceptible to change—in other words, they comprise the dynamic risk factors of an offender (Kroner & Mills, 2004). Through repeated administration, pairs of scores can be obtained to register shifts in individuals’ attitudes and needs over time. When used in combination with tools that supplement information on other static and dynamic risk factors, this instrument could be part of a comprehensive risk and need assessment package. Service providers could be informed of offenders’ needs and risk levels, and supervision decisions could be evaluated and guided accordingly (Feasey & Williams, 2009; Raynor & Kynch, 2000).

Items contained in the CRIME-PICS II instrument do not have a situational-specific definition for “crime” and are not specific to any type of offending behaviors (Kroner & Mills, 2004; Kroner, Mills, Yessine, & Hemmati, 2004). This feature allows interpretation of criminogenic constructs even when an offender has denied an offense or negative self-attributes (Kroner & Mills, 2004). The general nature of the perception items also enables the instrument to be administered to noncriminal populations (Kroner & Mills, 2004).

In recent decades, there has been increasing concern over structured methods of assessment and evaluation for supervision and treatment programs in criminal justice agencies (CRG Research, Ltd., 2003). CRIME-PICS II can measure the impact of programs on offenders and estimate their success or otherwise (Raynor 1998; Raynor & Kynch, 2000). For instance, previous program evaluations using CRIME-PICS II have recorded a general improvement in offenders’ crime-related attitudes and reductions in self-reported problems resulting from supervision (Raynor, 1998). These results help us understand what targets and milestones can be pursued in services for offenders without a long follow-up period, as traditional reconviction studies have required (CRG Research, Ltd., 2003; Raynor, 1998).

Aims of the Study

Despite the extensive research already conducted on CRIME-PICS II, few studies have examined its psychometric properties (Frude et al., 2008). To fill this gap, the purpose of the present study is twofold: firstly, to identify whether the factor solution of Part 1 of CRIME-PICS II, developed in Britain, is applicable in the Chinese setting and secondly to test the reliability and validity of the Chinese version.

This study examined the use of CRIME-PICS II in four studies with three different samples: a criminal population from Hong Kong, another from Mainland China, and a college student sample. To validate the Chinese version of the CRIME-PICS II instrument, its factor structure across the two offender samples was assessed using exploratory factor analysis (EFA) and follow-up confirmatory factor analysis (CFA). Subsequently, its internal consistency and temporal reliability were examined. The data collected from the college sample were used to gauge the test–retest reliability of the Chinese version of the scale (Kroner & Mills, 2004; Kroner et al., 2004). Finally, the instrument’s correlation with criminal data and other measures of well-being were computed to calculate its construct, concurrent, discriminant, and predictive validity.

Attitudes and self-perceived problems are both criminogenic needs. CRIME-PICS II is intended to measure these needs, which are often closely associated with domains of life that make individuals prone to committing crime, and with protective factors that help individuals desist from crime (McGuire & Hatcher, 2001). The literature suggests that individuals with low self-esteem, greater substance consumption, poor self-management, and poor social skills are more likely to report problematic attitudes and issues (CRG Research, Ltd., 2003; McGuire & Hatcher, 2001). On the other hand, social attachment factors (also known as social bonds), such as strong social bonding with family, a strong attachment to conventional values, and finishing school, are all related to reduced involvement in crime and other antisocial behaviors, especially among young offenders (Chapple, McQuillan, & Berdahl, 2005; Chui & Chan, 2011a, 2011b, 2012; CRG Research, Ltd., 2003). It is thus hypothesized that crime-prone attitudes are negatively related to respondents’ self-esteem, social bonding, life satisfaction, and belief in the criminal justice system and positively related to theft, violent behaviors, and substance consumption, while variance in attitudes could account for the variance in current and previous offending behavior.

Study 1: EFA

Participants

The participants in Study 1 were 106 male young offenders. They were recruited through a nongovernmental organization that provides transitional residential services for young offenders in Hong Kong. Six cases were dropped due to missing responses on one or more CRIME-PICS II items. The age of this final sample of 100 respondents ranged between 14 and 29 years old with a mean age of 17.87. With regard to their current convictions, 26% had been convicted of violent crime, 25% of offenses against property, 11% of drug offenses, 8% of sex offenses, and the remainder had been convicted of other offenses. Noteworthy characteristics of the respondents included that many were under an 18-month probation order at the time (34%), most were living in a transitional house (78%), around half had been there for less than a month (53%), around half had past convictions (48%), and most had committed offending behaviors within the previous 12 months (78%). The majority were single (88%), had only completed junior secondary school (64%), and some were currently attending school (39%). We acknowledged a concern that the sample size (N = 100) and subject to item ratio (5:1 for the Attitudes Questionnaire) did not satisfy some stringent guidelines, such as the subject to item ratio of 10:1 or higher and a minimum sample size of 400 recommended by Osborne and Costello (2004). We also recognized that the impact of these shortcomings on the accuracy of parameter estimation should not be overlooked. However, due to constraints on time and resources for data collection, it was decided that data collection should be ended at this point of time. As such, exploratory rather than CFA was conducted on this sample of respondents to better reveal possible factor structures underlying the scale for Chinese respondents (Worthington & Whittaker, 2006).

Procedure

All respondents were recruited from a local transitional house for young offenders under probation orders. Eligible participants had to meet the inclusion criteria of (i) having been convicted of a criminal offense, (ii) having been sentenced to a probation order, and (iii) being willing to participate. Questionnaires were self-administered and participants were invited to seek clarification with a trained research assistant if necessary. The participants were briefed on the research purpose, their confidentiality was emphasized, and their written informed consent was obtained. They were then given 30 min to complete the questionnaire.

Measures

The CRIME-PICS II instrument (Frude et al., 2008) has been used to measure crime-related attitudes and risks. This measurement consists of 20 items on an Attitudes Questionnaire (scored on a 5-point Likert-type scale ranging from 1 = strongly agree to 5 = strongly disagree) and 15 items on a Problem Inventory (scored on a 4-point Likert-type scale ranging from 1 = big problem to 4 = no problem). Five domains could be obtained from the measurement: General Attitude to Offending (17 items), Anticipation of Reoffending (6 items), Victim Hurt Denial (3 items), Evaluation of Crime as Worthwhile (4 items), and Perception of Current Life Problems (15 items). A lower score in each domain represents higher offending-related attitudes and problems. Frude, Honess, and Maguire (2008) reported that the Cronbach’s α of these five domains ranged from .55 to .83. As the focus of this article is on the 20-item Attitudes Questionnaire, only analyses relevant to this part of the instrument are reported here.

The original scale was translated into Chinese by a research assistant fluent in both English and Chinese. Another research assistant who had no knowledge of the original scale then back translated the Chinese version. Items with an ambiguous translation were discussed with the first author and refined in order to develop the final version.

Results

Statistical analysis was conducted using IBM SPSS, Version 19. Prior to the EFA, the response distribution of the 20 items was analyzed. All items had a standard deviation within 1.5, and no excessive skewness or kurtosis was detected in any item. The centering of mean scale scores around the midpoint of the scale range demonstrated the ability to maximize the sensitivity of scale scores (Kroner & Mills, 2004). As analysis of the data was exploratory in nature, attempts were made to extract various numbers of factors and to apply different factor extraction and rotation methods. Steps of factor extraction, selection, and rotation were conducted iteratively to achieve a simple and interpretable factor structure (George & Mallery, 2012). After several rounds of data exploration, it was decided that a four-factor solution was adequate to fit the observed data of the 20-item Attitudes Questionnaire. These factor solutions were obtained by using principal component analysis for extraction with varimax rotation. We were fully aware of the exploratory nature of this study. The factor solutions obtained here should be treated as tentative and thus subject to subsequent confirmatory analysis with another data set from an independent sample.

Under Kaiser’s criterion (1960) and supported by Cattell’s scree plots (1966), the current sample yielded four factors with Eigenvalues greater than 1, and the four-factor solution for the Attitudes Questionnaire explained 55.50% of the total variance. The factorability of the current intercorrelation matrix had been tested by the Kaiser–Meyer–Olkin measure of sampling adequacy, which was .82, and Bartlett’s test of sphericity, which was significant at 602.01, df = 120, p < .001.

The factor analysis identified a new factor solution with four dimensions. Only items with a loading of at least .40 on their target factor and low cross-factor loading(s) on the nontarget factor(s) were retained (Hatcher, 1994). Following a review of the literature, moral disengagement theory was used to shed light on the factors identified by the factor analysis (Bandura, 1990). Six items, with loadings ranging from .49 to .83, constituted the first new factor, labeled Attitude toward the Victim. Another 7 items, the highest loading being .74 and the lowest .42, clustered together and formed Factor 2, named Attitude toward the Nature of the Behavior. Factor 3 (Attitude toward the Actor) and Factor 4 (Attitude toward the Consequences of the Behavior) consisted of 4 items and 3 items, respectively. The factor loadings of Factors 3 and 4 ranged from .45 to .78 and .58 to .75, respectively (see Table 1).

Table 1.
Results of Exploratory Factor Analysis for Attitudes Questionnaire of the CRIME-PICS II.

Items Factor Loadings M (SD) Corrected Item Total Correlation

1 2 3 4

13 .83 2.59 (1.20) .74

02 .81 2.70 (1.14) .55

18 .77 2.52 (1.24) .66

09 .66 2.59 (1.18) .52

17 .65 2.39 (1.24) .62

06 .49 3.26 (1.26) .40

04 .74 2.38 (1.41) .76

05 .71 2.49 (1.34) .69

03 .61 2.97 (1.14) .10

10 .59 2.32 (1.36) .80

11 .43 .52 2.48 (1.33) .79

14 .50 2.34 (1.29) .65

07 .42 2.71 (1.25) .60

15 .78 2.30 (1.12) .39

08 .66 2.08 (1.25) .13

16 .53 2.62 (1.17) .55

12 .42 .40 .45 2.69 (1.20) .64

19 .75 2.43 (1.48) .74

01 .69 2.35 (1.37) .69

20 .58 2.46 (1.32) .36

Percentage of variance 19.72 15.36 10.41 10.01

Study 2: CFA and Internal Consistency

Participants

A sample of 288 female offenders was recruited from a prison in a city in eastern China. This was the maximum number of prisoners allowed by the management of the data collection site. Noteworthy characteristics of this sample were as follows: A high proportion had been sentenced to less than 5 years (38.9%), were aged between 30 and 40 (33.7%), were married (44.4%), had a child (40%), and had not attended school beyond junior secondary (41.9%).

Procedure

After issuing invitations to 288 participants, 270 prisons agreed to participate, representing a response rate of 93.8%. The participants were briefed on the purpose of the research, and their written informed consent was obtained. Participants were then given 30 min to self-administer the questionnaire, and the prison officer was available throughout the process to answer any questions. Completed questionnaires with missing responses on 1 or more of the items were dropped from the subsequent analysis. After cleaning, an effective sample of 235 cases was retained.

The Hypothesized Model

A series of CFAs was conducted on the Attitudes Questionnaire using AMOS 19. A four-factor model was hypothesized based on moral disengagement theory. Six items served as indicators of Attitude toward the Victim, 7 items for Attitude toward the Nature of the Behavior, four for Attitude toward the Actor, and three for Attitude toward the Consequences of the Behavior. In Figure 1, circles represent latent variables, rectangles represent measured variables, and arrowed lines linking variables imply a hypothesized direct effect between variables.

Figure 1.
Hypothesized factor model for the Attitudes Questionnaire of the CRIME-PICS II.

Model Estimation

Maximum-likelihood estimation was used and a number of goodness-of-fit indices were computed. The χ² test of model fit tested whether the estimated parameters of the models were suitable for the obtained data, and a statistically nonsignificant result was desirable (Byrne, 2010). The comparative fit index (CFI) and root mean square error of approximation (RMSEA) were reported. The CFI tests “the fit of a user-specified solution in relation to a more restricted, nested baseline model” (Brown, 2006, p. 84). A CFI score above .95 and RMSEA below .05 would indicate a good fit of the hypothesized model to the observed data (Blunch, 2013). Before running the model, skewness and kurtosis were computed for each of the 20 items in the Attitudes Questionnaire. All 20 had skewness within ±1.70 (11 items had skewness within ±1.00). All but 3 items had kurtosis within ±2.00. Examination of the distribution of the scores for the 3 items with the largest kurtosis showed that responses were either clustered heavily on the agreeable or the disagreeable side. According to some practical guidelines suggested by scholars (e.g., Byrne, 2010), these values reveal that no item was substantially skewed and kurtotic.

Taking the theoretical underpinning of CRIME-PICS II as a base and using both the modification indices and standardized residual values as a guide, post hoc model modifications were performed in order to improve the fit. After several respecifications,¹ a final model was reached (see Figure 2). This final model included (1) a correlated error term within the factor Attitude toward the Nature of the Behavior (Item 4, “Crime has now become a way of life for me” and Item 5, “Crime can be a useful way of getting what you want”) and a correlated error term between an item from the factor Attitude toward the Victim (Item 9, “I don’t see myself as real criminal”) and another from the factor Attitude toward the Actor (Item 12, “Many so-called crimes are not really wrong”); (2) an item moved from the factor Attitude toward Victim to the factor Attitude toward the Consequences of the Behavior (Item 17, “When people have no money, they can’t be blamed for stealing”); and (3) deletion of 3 items that yielded statistically nonsignificant loadings on their target factor (Item 6, “I believe in living for now; the future will take care of itself” from Attitude toward the Victim; Item 3, “I will always get into trouble” from Attitude toward the Nature of the Behavior; and Item 8, “I definitely won’t get into trouble with the police in the next 6 months” from Attitude toward the Actor). The goodness-of-fit-indices for this final model were χ² = 340.114 (df = 161, p < .001), CFI = .913, and RMSEA = .066. Although the goodness-of-fit indices of this post hoc model indicated that the model fit was only satisfactory to good (Blunch, 2013), we decided to stop here to avoid overfitting.

Figure 2.
Final factor model for the Attitudes Questionnaire of the CRIME-PICS II.

Good internal consistency was achieved for the Attitudes Questionnaire (Cronbach’s α = .87). Cronbach’s αs ranging from .64 to .83 were obtained for its four factors. Information about each factor is presented in Table 2.

Table 2.
Reliabilities of Attitudes Questionnaire of the CRIME-PICS II and Its Factors.

Name of Scale and Factors M SD Cronbach’s α No. of Items

Attitudes Questionnaire 2.49 .84 .87 17

Victim 2.61 96 .82 4

Nature 2.45 .98 .83 6

Actor 2.53 .88 .64 3

Consequence 2.41 1.00 .80 4

The correlations of the full scale were examined. Table 3 shows that corrected item-total correlations for the Attitudes Questionnaire ranged from .36 to .80 (Mean r = .61), correlations of the Attitudes Questionnaire with its four factors ranged from .58 to .76, and correlations among the four factors of the Attitudes Questionnaire ranged from .26 to .51.

Table 3.
Correlations Among Attitudes Questionnaire and Its Factors.

Variables Attitude Questionnaire Victim Nature Actor

Attitudes Questionnaire — — — —

Victim .75 — — —

Nature .76 .28 — —

Actor .75 .51 .43 —

Consequence .58 .30 .33 .26

p < .01 (two tailed).

Study 3: Temporal Reliability

Participants

A sample was gathered by recruiting students from three undergraduate courses. One hundred and seventy completed questionnaires were collected from said sample, of which 164 were deemed valid for subsequent analyses. The age of the participants ranged from 18 to 48 (M = 21.71 years, SD = 3.94); 77 were male (46.9%) and 86 female (52.4%). One did not indicate gender. More than half were atheists (61%, n = 100), some were protestants (25%, n = 41), a few were Catholics (4.9%, n = 8) or held other religious beliefs (n = 3: 1 Muslim, 1 Taoist, and 1 Falungong), and the remainder did not indicate their religious belief (n = 12).

After a second session in which follow-up questionnaires were distributed, 126 questionnaires were collected, although only 54 could be matched back to questionnaires from the initial survey (due to reasons mentioned below). Of the participants among the successfully matched cases, gender was almost evenly distributed, 52% being male (n = 28) and 48% female (n = 26). They were aged from 18 to 33, with a mean age of 21 years and standard deviation of 2.72 years. Among this group, 25% had no religious background (n = 30), 13.3% were protestants (n = 16), 5% Catholics (n = 6), and the remainder did not indicate their religious beliefs.

Procedure

All prospective participants were briefed by their lecturer at the beginning of the class on the research’s purpose and invited to take part. As participation in this study was voluntary, the questionnaires were passed around the class and students who were willing to be involved took a copy for themselves. Written informed consent was obtained from all participants, and they were then given 15 min to complete the questionnaire. Participants were asked to provide some identification in the form of the last four digits of their student identity number, such that follow-up questionnaires could be matched.

Two weeks later, in the last session of the semester and the only occasion on which the same groups of students could meet the lecturer again, follow-up surveys were conducted: The lecturer again circulated the questionnaire in class and requested participants to complete it. After this follow-up session, 126 completed questionnaires were collected and 54 successfully matched back to participants’ first-round questionnaires. Failure to match occurred due to the following reasons: (i) participants did not provide the requested personal identity number either in the initial or follow-up questionnaire, (ii) they provided an invalid identification, or (iii) students who had taken part in the first round were not present at the follow-up session. As a 2-week interval is rather short for a test–retest reliability estimate (its value would be higher than that of a standard 4-week interval), interpretation of this estimated value should be cautious.²

Results

Based on the 54 successfully matched college participants’ responses, test–retest reliability was computed to examine the temporal stability of the measurement. Paired sample t-tests were conducted and no statistical differences were observed between Time 1 and Time 2 for the Attitudes Questionnaire (t = 1.40, p = .17). The Spearman–Brown coefficient for the Attitudes Questionnaire was .81, and the coefficients for the subscales ranged from .56 to .81 (see Table 4). These findings support the consistency of the measurement over time.

Table 4.
Means, Standard Deviations, Correlations of the Attitudes Questionnaire of the CRIME-PICS II With Other Measures in the Male Young Offender Sample.

Name of Scale and Factors r ^a M SD

Self-esteem .28 2.59 0.31

Social bond −.31 0.58 0.09

Theft −.34 0.49 0.21

Violent crime −.35 0.52 0.21

Belief in the legal system −.44 0.62 0.13

Peer attachment ns 0.47 0.16

Parental attachment ns 0.55 0.21

Life satisfaction ns 4.28 1.22

Substance abuse ns 2.12 0.64

Note. ns = nonsignificant.

^aCorrelation coefficients of the Attitude Questionnaire with other measures.

***p <* .01.

Study 4: Validity

Participants

The data of the young male offender sample recruited in Study 1 were utilized for analyses in Study 4, and discriminant validity was analyzed by comparing this sample with the college sample from Study 3.

Measures

Several relevant psychosocial variables were measured to be used in the construct validity analyses, in accordance with the findings of previous research. An indication of social attachment was measured using the Social Bond Scale (Chapple et al., 2005). This scale consists of 24 items covering eight factors: theft, violent crime, peer attachment, dependence on parents, parental attachment, commitment to education, involvement in organizations, and belief in the legal system. Higher scores indicate a stronger social bond.

Self-esteem was measured using the Rosenberg Self-Esteem Scale, the most widely used tool for assessing this construct (Rosenberg, 1965). The scale consists of 10 items on a 4-point Likert-type scale (1 = strongly disagree and 4 = strongly agree). A higher score indicates higher self-esteem. The scale has demonstrated good reliability in previous studies (Cronbach’s α = .84).

Life satisfaction was measured using the Satisfaction with Life Scale (Diener, Emmons, Larsen, & Griffin, 1985). This is a 5-item unidimensional scale developed to measure general satisfaction with life. Participants were asked to rate the items on a 7-point Likert-type scale (1 = strongly disagree to 7 = strongly agree); the higher the total score, the more satisfied an individual is with his or her life. The scale has been validated locally (Wang, Yuen, & Stanley, 2009) and shows good internal consistency with a Cronbach’s α of .88.

Substance abuse patterns were measured using the Substance Abuse Scale (Shek, 2005). The scale contains 9 items with on a 7-point Likert-type scale. A higher score indicates more frequent drug abuse. The scale was developed indigenously and thus includes the substances that are most commonly consumed by the local population, namely, “K (Ketamine)” and “Pills (e.g., Ecstasy or mandrax).” Adequate reliability has been demonstrated with Cronbach’s α ranging from .65 to .72.

Respondents’ criminal history was simply measured using a self-developed inventory detailing offending behaviors and previous convictions. A table with a list of criminal behaviors was presented and respondents were asked to indicate whether they had committed any of the types of offense shown therein within the previous 12 months, whether they were arrested or not, and the frequency of their engagement in said behaviors. Examples of offense types included “assault on police” and “possession of dangerous drugs.” Respondents were also asked to indicate the number of their previous convictions. It is worth noting that despite the subjectivity of such self-reports on offending behaviors and criminal history, the technique is considered suitable and the data obtained are considered to be valid, as the appropriateness of official reconviction records as measures of criminality have been questioned by scholars, especially for validation of attitudinal scales (Healy & O’Donnell, 2006). In order to encourage honest responses, respondents were assured of the confidentiality and anonymity of their reports (Healy & O’Donnell, 2006).

Results

Correlational analyses were conducted to examine the relationship between the Attitudes Questionnaire, the well-being of the offenders, and their criminal background. In general, the results agreed with the findings of past research, demonstrating relationships with the related constructs as predicted. To test whether the Attitudes Questionnaire of CRIME-PICS II could be associated with respondents’ well-being, its scores were correlated with the scores of measures of self-esteem, social attachment, life satisfaction, and substance abuse patterns.

The Attitudes Questionnaire correlated weakly to moderately with self-esteem (r = .28, p < .001) and social attachment (r = −.31, p < .001). The social bond subscale of Theft and Violent Crime specifically had slightly stronger associations with the Attitudes Questionnaire (r = −.34, p < .001; r = −.35, p < .001) than the previous two correlations. Specifically, individuals who reported less problematic attitudes tended to possess higher self-esteem, stronger social attachment, and were less involved in theft and violent behaviors.

The Attitudes Questionnaire also correlated moderately with the social bond subscale of Belief in the Legal System (r = −.44, p < .001), indicating that offenders who held less crime-prone attitudes also had a stronger belief in the criminal justice system.

To test the discriminant validity of the Attitudes Questionnaire of CRIME-PICS II (i.e., whether it could discriminate between offender and nonoffender groups), independent sample t-tests were conducted on this measuring scale and its subscales. The results showed a significant group difference in the scores of the Attitudes Questionnaire and those of all of its subscales (see Table 5). The effect sizes (measured by Cohen’s d) were quite strong for the Attitudes Questionnaire and two of its subscales (Attitude toward the Nature of the Behavior and Attitude toward the Consequences of the Behavior). As expected, the offender sample presented significantly more crime-prone attitudes than the college sample.

Table 5.
Tests of Mean Differences Between the Hong Kong Male Young Offender and College Sample on the Attitudes Questionnaire of the CRIME-PICS II.

Variable Male Young Offender, M (SD) College Students, M (SD) t-Value Cohen’s d

Attitudes Questionnaire 2.49 (0.84) 3.34 (0.38) 11.37 1.30

Victim 2.61 (0.96) 3.00 (0.53) 4.30 0.50

Nature 2.45 (0.98) 3.57 (0.64) 11.35 1.35

Actor 2.53 (0.88) 2.91 (0.67) 3.99 0.49

Consequence 2.41 (1.00) 3.65 (0.54) 13.23 1.54

p < .01.

To test whether the Attitudes Questionnaire of CRIME-PICS II was associated with criminal history and offending behavior, hierarchical regressions were computed for it against the number of respondents’ previous convictions and the number of offending behaviors they had committed. The Attitudes Questionnaire scores were entered as predictors. Age and gender were entered as control variables and were found to yield no significant effects in the following analyses.

Both previous convictions and current offending behaviors were significantly predicted by the Attitudes Questionnaire (β = −.29, t = −2.71, p < .01, R ² = .08; β = −.33, t = −3.13, p < .01, R ² = .11), indicating that more problematic attitudes were associated with a higher likelihood that the individual had engaged in offending behaviors in the past 12 months and had a criminal record, thus confirming the predictive power of the Attitudes Questionnaire of CRIME-PICS II in this regard. The effect sizes of these two predictions were in the small to medium range. A summary of results for the construct, concurrent, and discriminant validity are shown in Tables 4 and 5.

Discussion and Conclusion

The primary aim of the current study was to demonstrate the utility of the CRIME-PICS II instrument in the Chinese context through an examination of its factor structure and by providing evidence of its psychometric properties. The study has examined the factor structure of the Attitudes Questionnaire of the Chinese version of CRIME-PICS II with reference to past theories, provided evidence of the reliability and validity of its psychometric properties, and reviewed the instrument’s utility in practical contexts, fulfilling the criteria of a scale validation study (Macey & Schneider, 2008).

The factor structure of the Attitudes Questionnaire of CRIME-PICS (Chinese) was developed using both EFA and CFA. EFA identified a four-factor structure that differed from the findings of Frude and colleagues’ (2008) study. The results also suggest that most of the items loaded onto one single factor. The findings from a second sample using CFA confirm this factor structure. However, several items were found not to be statistically significant in the model and thus were removed; these were the items that had exceptionally low item-total correlations (<.40) in Study 1.

Moral disengagement theory (Bandura, 1990) was utilized to explain the factors of the Attitudes Questionnaire. The theory posits that there are four loci in our psychological processes which disengage moral feedback from injurious behaviors and prevent self-sanctioning, namely, the locus of the behavior, the agent of the action, the outcomes that flow from the behavior, and the recipient of the actions (Bandura, 1990; Osofsky, Bandura, & Zimbardo, 2005). The four loci provide a theoretical framework for comprehending respondents’ crime-prone attitudes and informed our labeling of the four factors of the Attitudes Questionnaire.

The internal consistency for the Attitudes Questionnaire was good (Cronbach’s α = .87) and was similarly satisfying for each of its factors (Cronbach’s α between .64 and .83). The Cronbach’s αs obtained in this research were higher than those of the original scale (between .55 and .83). All factors correlated weakly to moderately among themselves (r between .26 and .51) and with the Attitudes Questionnaire (r between .58 and .76). The scale also demonstrated adequate temporal stability (Spearman–Brown coefficient = .81 with a 2-week interval), proving that the measurement could provide information that remained stable over time.

Correlation results supported the construct validity of the Chinese version of the instrument and provided preliminary support for its external validity. In line with previous findings (i.e., Chapple et al., 2005; CRG Research, Ltd., 2003; McGuire & Hatcher, 2001; Kilcommins, 2007), crime-prone attitudes showed a robust overall link with the psychosocial variables of individuals’ self-esteem, social attachment, and behaviors related to theft and violent behavior—less problematic attitudes were observed among participants who reported higher self-esteem, stronger social attachment, and lower involvement in theft and violent behaviors. The identified effects of this study confirm our literature-derived hypotheses, which reveals that human and social capitals, such as self-esteem and social networks, are intervening factors that help solve problems and thus facilitate a lifestyle free of crime (Farrall, 2004; Maguire & Raynor, 2006).

Convergent validity was conveyed by the significant correlations between the Attitudes Questionnaire and belief in the criminal justice system. These findings present a convincing argument that the Chinese version of CRIME-PICS II is valid for use in Chinese samples, as it has been demonstrated to be in Western samples by previous studies. With regard to discriminant validity, the Attitudes Questionnaire also demonstrated the ability to distinguish a criminal sample from a sample of the general population.

Finally, respondents’ criminal history and offending behaviors were found to be correlated to their attitudes. Results from regression analyses thus proved the Chinese version of CRIME-PICS II to be predictively valid as is the original version.

Despite the merits of the present study, it has several limitations that warrant caution. The first is in the study’s cross-sectional design. Although multiple regression analyses were conducted to test the predictive validity of the instrument, they were based on past offending data and thus required self-report data from respondents’ memories. In order to truly register changes during interventions and to better evaluate the predictive validity of CRIME-PICS II as an evaluation tool, longitudinal data are required, which the authors are currently setting out to obtain.

Another limitation is in the sampling method. For the Hong Kong young offender sample, due to limited resource allocation in local probation settings, the study had to recruit respondents who had already been admitted to a transitional house within the previous 6 months. About half had already lived in the house for more than a month at the time of the questionnaire, which may have influenced our results as described earlier. In order to control this potential confounding factor, future research should attempt to recruit respondents who have just started their treatment program, or have only undergone a short period of treatment, such as a month.

In conclusion, this study presents a strong case for the validity of the Chinese version of the Offending-Related Attitudes Questionnaire of the CRIME-PICS II scale as a measure of attitude toward crime in the Hong Kong probation service within the Chinese custodial setting. It is hoped that future research will be conducted on the utility of the instrument, such that it will continue to develop as a useful assessment and evaluative tool for practitioners to improve their services and better assist their clients in breaking free from criminality.

Items	Factor Loadings	M (SD)	Corrected Item Total Correlation
13	.83				2.59 (1.20)	.74
02	.81				2.70 (1.14)	.55
18	.77				2.52 (1.24)	.66
09	.66				2.59 (1.18)	.52
17	.65				2.39 (1.24)	.62
06	.49				3.26 (1.26)	.40
04		.74			2.38 (1.41)	.76
05		.71			2.49 (1.34)	.69
03		.61			2.97 (1.14)	.10
10		.59			2.32 (1.36)	.80
11	.43	.52			2.48 (1.33)	.79
14		.50			2.34 (1.29)	.65
07		.42			2.71 (1.25)	.60
15			.78		2.30 (1.12)	.39
08			.66		2.08 (1.25)	.13
16			.53		2.62 (1.17)	.55
12	.42	.40	.45		2.69 (1.20)	.64
19				.75	2.43 (1.48)	.74
01				.69	2.35 (1.37)	.69
20				.58	2.46 (1.32)	.36
Percentage of variance	19.72	15.36	10.41	10.01

Name of Scale and Factors	M	SD	Cronbach’s α	No. of Items
Attitudes Questionnaire	2.49	.84	.87	17
Victim	2.61	96	.82	4
Nature	2.45	.98	.83	6
Actor	2.53	.88	.64	3
Consequence	2.41	1.00	.80	4

Variables	Attitude Questionnaire	Victim	Nature	Actor
Attitudes Questionnaire	—	—	—	—
Victim	.75**	—	—	—
Nature	.76**	.28**	—	—
Actor	.75**	.51**	.43**	—
Consequence	.58**	.30**	.33**	.26**

Name of Scale and Factors	r ^a	M	SD
Self-esteem	.28**	2.59	0.31
Social bond	−.31**	0.58	0.09
Theft	−.34**	0.49	0.21
Violent crime	−.35**	0.52	0.21
Belief in the legal system	−.44**	0.62	0.13
Peer attachment	ns	0.47	0.16
Parental attachment	ns	0.55	0.21
Life satisfaction	ns	4.28	1.22
Substance abuse	ns	2.12	0.64

Variable	Male Young Offender, M (SD)	College Students, M (SD)	t-Value	Cohen’s d
Attitudes Questionnaire	2.49 (0.84)	3.34 (0.38)	11.37**	1.30
Victim	2.61 (0.96)	3.00 (0.53)	4.30**	0.50
Nature	2.45 (0.98)	3.57 (0.64)	11.35**	1.35
Actor	2.53 (0.88)	2.91 (0.67)	3.99**	0.49
Consequence	2.41 (1.00)	3.65 (0.54)	13.23**	1.54

Footnotes

Declaration of Conflicting Interests

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

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Andrews

D. A.

Bonta

(1995). The level of service inventory-revised manual. Toronto, Canada: Multi-Health Systems.

Bandura

(1990). Mechanisms of moral disengagement. In Reich

(Ed.), Origins of terrorism: Psychologies, ideologies, theologies, states of mind (pp. 161–191). Cambridge: Cambridge University Press.

Blunch

N. J.

(2013). Introduction to structural equation modeling using IBM SPSS statistics and AMOS (2nd ed.). London, England: Sage.

Brown

T. A.

(2006). Confirmatory factor analysis for applied research. New York, NY: The Guilford Press.

Burgoyne

(2013). An evaluation of the ‘Think First’ program. Australian & New Zealand Journal of Criminology, 46, 88–100.

Byrne

B. M.

(2010). Structural equation modeling with AMOS: Basic concepts, applications, and programming (2nd ed.). New York, NY: Routledge.

Cattell

R. B.

(1966). The scree test for the number of factors. Multivariate Behavioral Research, 1, 629–637.

Chapple

C. L.

McQuillan

J. A.

Berdahl

T. A.

(2005). Gender, social bonds, and delinquency: A comparison of boys’ and girls’ models. Social Science Research, 34, 357–383.

Chui

W. H.

Chan

H. C.

(2011a). Baseline findings of a prospective study on pro-offending attitudes and self-reported problems among juvenile probationers. Hong Kong Journal of Social Work, 45, 13–26.

10.

Chui

W. H.

Chan

H. C.

(2011b). Social bonds and male delinquency while on probation: An exploratory test in Hong Kong. Children and Youth Services Review, 33, 2329–2334.

11.

Chui

W. H.

Chan

H. C.

(2012). An empirical investigation of social bonds and juvenile delinquency in Hong Kong. Child and Youth Care Forum, 41, 371–386.

12.

Clancy

Hudson

Maguire

Peake

Raynor

Vanstone

Kynch

(2006). Getting out and staying out: Results of the prisoner resettlement pathfinders. Bristol, MA: Policy Press.

13.

CRG Research, Ltd. (2003). An evaluation of summer plus: A cross departmental approach to preventing youth crime. Nottingham, England: Department for Education and Skills.

14.

Diener

Emmons

R. A.

Larsen

R. J.

Griffin

(1985). The satisfaction with life scale. Journal of Personality Assessment, 48, 71–75.

15.

Farrall

(2004). Social capital and offender re-integration: Making probation desistance focused. In Maruna

Immarigeon

(Eds.), After crime and punishment: Pathways to offender reintegration (pp. 57–84). Cullompton, England: Willan.

16.

Feasey

Williams

(2009). An evaluation of the Sycamore Tree Project based on an analysis of Crime Pics II data: Report. Sheffield, England: Sheffield Hallam University.

17.

Frude

Honess

Maguire

(2008). CRIME-PICS II manual (3rd ed.). Cardiff, England: Michael and Associates.

18.

George

Mallery

(2012). IBM SPSS statistics 19 step by step: A simple guide and reference (12th ed.). Boston, MA: Pearson.

19.

Hatcher

(1994). A step-by-step approach to using the SAS system for factor analysis and structural equation modelling. Cary, NC: SAS Institute.

20.

Healy

O’Donnell

. (2006). Criminal thinking on probation: A perspective from Ireland. Criminal Justice and Behavior, 33, 782–802.

21.

Kaiser

H. F.

(1960). The application of electronic computers to factor analysis. Educational and Psychological Measurement, 20, 141–151.

22.

Kilcommins

(2007). After crime and punishment: pathways to offender reintegration. British Journal of Criminology, 47, 844–846.

23.

Kroner

D. G.

Mills

J. F.

(2004). The criminal attribution inventory: A measure of offender perceptions. Journal of Offender Rehabilitation, 39, 15–29.

24.

Kroner

D. G.

Mills

J. F.

Yessine

A. K.

Hemmati

(2004). The generalized instructional set of the Criminal Attribution Inventory (CRAL): Socially desirable responding and faking. International Journal of Offender Therapy and Comparative Criminology, 48, 360–372.

25.

Lewis

Lobley

Raynor

Smith

(2005). The Irish on probation in England. Irish Probation Journal, 2, 4–19.

26.

Lewis

Maguire

Raynor

Vanstone

Vennard

(2003). The resettlement of short-term prisoners: An evaluation of seven pathfinder programmes. London, England: Home Office.

27.

Macey

W. H.

Schneider

(2008). The meaning of employee engagement. Industrial and Organizational Psychology, 1, 3–30.

28.

McGuire

Hatcher

(2001). Offence-focused problem solving: Preliminary evaluation of a cognitive skills program. Criminal Justice and Behavior, 28, 564–587.

29.

Maguire

Raynor

(2006). How the resettlement of prisoners promotes desistance from crime: Or does it? Criminology and Criminal Justice, 6, 19–38.

30.

Meek

(2012). The role of sport in promoting desistance from crime: An evaluation of the 2nd chance project rugby and football academies at Portland Young Offender Institution. Southampton, England: University of Southampton.

31.

Osborne

J. W.

Costello

A. B.

(2004). Sample size and subject to item ratio in principal components analysis. Practical Assessment, Research & Evaluation, 9. Retrieved March 28, 2016, from http://pareonline.net/getvn.asp?v=9&n=11

32.

Osofsky

M. J.

Bandura

Zimbardo

P. G.

(2005). The role of moral disengagement in the execution process. Law and Human Behavior, 29, 371–393.

33.

Raynor

(1998). Attitudes, social problems and reconvictions in the “STOP” probation experiment. Howard Journal of Criminal Justice, 37, 1–15.

34.

Raynor

Kynch

(2000). Risk and need assessment in probation services: An evaluation. London, England: Home Office.

35.

Rosenberg

(1965). Society and the adolescent self-image. Princeton, NJ: Princeton University Press.

36.

Shek

D. T. L.

(2005). Paternal and maternal influences on the psychological well-being, substance abuse, and delinquency of Chinese adolescents experiencing economic disadvantage. Journal of Clinical Psychology, 61, 219–234.

37.

Walters

(1995). The Psychological Inventory of Criminal Thinking Styles: Part 1: Reliability and preliminary validity. Criminal Justice and Behavior, 22, 307–325.

38.

Wang

K. T.

Yuen

Stanley

R. B.

(2009). Perfectionism, depression, loneliness, and life satisfaction: A study of high school students in Hong Kong. The Counseling Psychologist, 37, 249–274.

39.

Worthington

R. L.

Whittaker

T. A.

(2006). Scale development research: A content analysis and recommendations for best practices. The Counselling Psychologist, 24, 806–838.