Measuring criminal thinking among convicts imprisoned in Punjab prisons of Pakistan

Abstract

Purpose

This study aims to differentiate crime-related characteristics (such as the number of cases filed against current convictions and criminal history) based on the criminal thinking prevailing among convicts. However, because of the low reliability of subscales and poor structural validity of indigenous and translated versions of international instruments, a new instrument criminal attitude measure (CAM) was extracted to measure criminal thinking patterns among convicts incarcerated in central prisons of Punjab.

Design/methodology/approach

A cross-sectional research design was used. Data was collected from 1,949 male convicts (extracting mutually exclusive data from 649 respondents for EFA and 1,300 respondents for confirmatory factor analysis [CFA]). Both data samples were collected from convicts incarcerated in the nine (all) central jails of Punjab, Pakistan.

Findings

The results of this study showed poor model fit for both the indigenous criminal thinking scale and the translated version of criminogenic cognition scale. CAM was extracted through principal component analysis and proposed as a 15-item questionnaire with five factors extracted through varimax rotation. Those five factors are power orientation, mollification, entitlement, mistrust toward authorities and short-term orientation. The results of CFA for CAM confirmed the proposed five-factor structure for the construct. Findings based on MANOVA further found that CAM differentiates between the thinking patterns of recidivists, convicts with multiple charges filed against them in current convictions and convicts with a familial criminal record. The findings of this study showed that CAM is a practical, valid and reliable instrument for measuring criminal thinking among convicts.

Research limitations/implications

In this study, using the survey method was inevitable because of the restrictions imposed by the granted permission. However, this time duration was extended because of the courtesy of the Superintendent and Deputy Superintendent of each jail. This study is focused on a male sample only, and the findings cannot be generalized to females. The phenomena proposed (based on large data sets) in this study can further be elaborated using qualitative research designs and methods (using a small sample with an in-depth study). So, it is also suggested to test this new instrument on a comparative study between prisoners and non-prisoners to explore whether scale can differentiate between these two groups.

Practical implications

A short-scale and easy-to-administer instrument was developed for assessing major criminogenic needs among convicts for prison management, i.e. assigning barracks, allocating treatment and also detecting changes in attitude after imprisonment.

Originality/value

To the best of the authors’ knowledge, this study is the first study to explore and validate the construct of criminal attitudes among convicts using both the EFA and CFA. A small and valid instrument facilitates the measurement of criminogenic needs among prisoners. Data was collected from all central jails in Punjab. This study explored comparatively less researched crime characteristics in a relatively large sample.

Citation

Ishfaq, N. and Kamal, A. (2024), "Measuring criminal thinking among convicts imprisoned in Punjab prisons of Pakistan", Journal of Criminal Psychology, Vol. 14 No. 3, pp. 288-307. https://doi.org/10.1108/JCP-09-2023-0057

Introduction

Imprisonment aims to teach criminals that crime does not pay. The likelihood of reoffending is thus reduced by increasing the cost of imprisonment (through a harsh environment) rather than non-incarceration sanctions (Bonta and Andrews, 2017). According to a contrary perspective, criminologists typically maintain that being imprisoned is not simply a cost but also a social experience that fosters criminal behavior (Cullen et al., 2011). More than 10 million people detained internationally experience a higher burden of physical or psychological diseases and substance misuse problems than the general population. They often come from underserved or marginalized groups in the community (Stürup-Toft et al., 2018).

With more than 50% of the country’s population, Punjab is the most populous province of Pakistan (Adil et al., 2021), struggling with basic facilities and being a very high crime-reported area in Pakistan (Mahmood et al., 2019). In Pakistan, the crime/homicide rate is 3.88 per 100K population, and its trend has increased by 2.09% compared to 2019 (World Bank, 2021). Young people ranging in age from 25 to 29 years old reported high involvement in committing crimes such as theft, murder, abduction, robbery, drug-related offenses, illegal weapons and firing with criminal intent (Cheema et al., 2022). This is because of socioeconomic factors contributing to crime, such as unemployment, urbanization, illiteracy, early age and male gender (Kalsoom and Gul, 2020; Zaman, 2021). The increasing crime rate in Pakistan is affected by the significant long-term effects of human capital, corruption, quality of life, economic misery and the rule of law. Increased population density is essential in stimulating criminal activities in Punjab, Pakistan (Anwer et al., 2015).

Punjab has four kinds of prisons (central, district, special and sub-jails) depending upon their detention capacity, duration of imprisonment, nature of the crime and gender of the criminal (Nabi et al., 2021). Central jails have a maximum detention capacity of more than 1,000 prisoners, irrespective of the length of sentence (Punjab Prisons, 2023). The prison population comprises convicts, civil prisoners, on-remand criminals and any person ordered without trial under any law relating to the detention of such persons (Haq and Zafar, 2019). In Punjab alone, 29 out of the 41 prisons are overcrowded (Ministry of Human Rights, 2020). Jails have a myriad of problems, including overcrowding, torture, understaffing, underbudgeting, lack of basic prisoner needs, proper staff training, etc. In addition, obsolete prison rules rooted in the inefficient colonial criminal justice system and the State’s indifference toward prisons have greatly aggravated the problems (Gul, 2017). One of the root causes of major human rights violations in Pakistani prisons is massive overcrowding, where approximately 77,000 inmates are imprisoned instead of the authorized capacity of 56,634 inmates (Nabi et al., 2021).

Prison serves as a school for learning and reinforcing criminal behaviors, so some prisoners prefer to remain in prison as they feel safer and more connected there than life outside prison (Howerton et al., 2009). The imprisonment experience requires the deployment of the individual’s emotional and cognitive resources. Adjustment to imprisonment is described as the prisonization process, meaning the acceptance of behavioral rules dictated by the society of prisoners (Boduszek et al., 2021; Crewe et al., 2014). Prisoners with similar attitudes form groups within the prison and develop internal codes to rationalize their criminal behaviors by blaming external factors, i.e. distrustful authorities, discriminatory behavior, poverty, etc. (Žukov et al., 2009).

Many criminological theories have focused on criminal thinking, e.g. subcultural, anomie, differential association, control, labeling and self-control theory. Differential association (Sutherland and Cressey, 1978) explained criminal thinking as a part of crime initiation. Others, such as the neutralization theory (Sykes and Matza, 1957), focused on the cognitive aspect of crime maintenance. Ward et al. (1997) presented a theoretical overview of cognitive distortions among sex offenders. They illustrated how attributional biases and perceptual distortions, such as misinterpreting cues from a woman, could initiate sexual offending or maintain attitudes, stereotypes and sex-offending behavior. Persistent offenders often hold distinct beliefs labeled as immoral cognitions that rationalize and perpetuate illegal activity (Tangney et al., 2007). Criminologists such as Sykes and Matza (1957) explained the techniques of neutralization that offenders use to reduce the dissonance between moral standards and immoral behavior – for example, minimizing harmful consequences by dehumanizing the victim. Maintenance of criminal behavior is because of cognitive distortions forming self-serving thinking bias, including blaming others for the offense committed along with a positive portrayal of the crime committed (Filkin et al., 2022). Individuals often rank themselves high in their opinions and tend to neutralize guilt or conscience. Rationalization of the criminal offense occurs to protect self-identity by attributing crime to external factors such as fate, genetic biology or provocative situations (Boduszek et al., 2021). This way, responsibility or accountability for crimes is intentionally shifted from the criminal to the victim or society. Antisocial cognitions and antisocial attitudes predict criminal behavior (Willmott and Ioannou, 2017).

Cognitions that initiate or maintain the violation of established laws promote criminal thinking among individuals. According to the lifestyle perspective, it is not until unlawful thought content and processes come together in time and space that the decision to act on a criminal thought then leads to crime (Walters, 2008). Walters's conceptualization of criminal thinking was based on Yochelson and Samenow's (1976) idea that antisocial behaviors are based on free choice. They developed an initial theoretical framework for conceptualizing criminals' thought processes. They noted that the criminal thinking process among offenders is persistent, interrelated, unaware and erroneous, yet there are distinct cognitive aspects of the criminal lifestyle. Overall, such thinking patterns result in inappropriate and illegal activity, and thus changing antisocial behavior requires a change in these erroneous thinking patterns (Walters, 2001). Walters (1996) developed eight cognitive patterns to measure criminal thinking: mollification, cutoff, entitlement, power orientation, sentimentality, super optimism, cognitive indolence and discontinuity. These patterns exhibit self-indulgent, subjective, hostile and rash decision-making or thinking styles contrary to societal standards (Mandracchia et al., 2007). Furthermore, according to Walters (2012), a positive association exists between a person's internalization of antisocial cognitions and the severity of their criminal conduct. He also distinguishes current criminal thinking as a measure of ongoing criminality and historical criminal thinking as an estimate of one's criminal past. Criminal thinking represents a more fluid (or state-dependent) process that can vary somewhat and show subsequent behavioral responses (Ellis and Ellis, 2011). Antisocial cognitions are a criminogenic dynamic risk factor amenable to intervention (Jackson et al., 2023).

The risk-need-responsivity model claims a pro-criminal attitude is a central criminogenic need to explain/predict recidivism, making it a significant part of prisoner risk assessment (Andrews and Bonta, 1998, 2010). Different instruments operationally defined criminal thinking through lengthy statements and reported low alpha reliabilities at the subscale level (Martínez and Andrés-Pueyo, 2015; Walters, 2002). A relatively short criminogenic cognitions scale (CCS) is based on the neutralization theory of Sykes and Matza (1957) and is comprised of 25 items (Tangney et al., 2002). Five subscales include notions of entitlement/demand for respect, a short-term orientation toward goals, negative attitudes toward authority, an external locus of control or failure to accept responsibility for one’s actions and an insensitivity to the impact of crime on victims and society. The alpha reliability of four out of five subscales was≥0.62, acceptable for research purposes and practical to administer on a large sample (Tangney et al., 2012). Major studies on criminal thinking were carried out in Western countries (e.g. the USA, UK, Spain, etc.), whereas crime and its interpretation depend on contextual culture. Countries (e.g. Canada, the USA and Australia) used the risk assessment model to reduce recidivism (Bonta and Andrews, 2017). Governments adapt the risk assessment model per their indigenous risk and need factors by developing or adapting tools and intervention plans (Giguere and Lussier, 2016; Giguère et al., 2021).

Risk assessment in Pakistan's prisons has been warranted for a long time (Gul et al., 2021; Khalid and Khan, 2013). Bhutta and Wormith (2015) conducted a study in Pakistan and found that criminal thinking predicted recidivism among offenders released on parole. From an indigenous perspective, the indigenous criminal thinking scale (ICTS) (Sana and Batool, 2017) was developed in Pakistan for offenders (including both males and females aged 18–60 years) in Punjab prisons. ICTS consists of 24 items covering five domains: criminal rationalization, power orientation and justification, personal irresponsibility, vindication and entitlement. In the present study, ICTS is used to cover the contextual conceptualization of the construct, and its validation through CCS is preferred because of the shorter length and easy-to-comprehend content of CCS. CCS was translated in the present study for this purpose. The current study aims to validate ICTS on a larger independent sample of convicts and establish its construct validity. Through identifying contemporary criminal thinking styles, one can explore, target and control the process of prisonization’s inclination toward recidivism and in-prison violence (Walters, 2012; Vrućinić, 2019). There is a strong need to explore criminal thinking for crime-related characteristics in the Pakistani sample of prisoners or offenders, especially those of convicts.

The current study has the following objectives:

• to establish the psychometric properties and convergent validity of the ICTS with the translated version of the criminogenic cognition scale (CCS);

• to differentiate criminal attitudes based on crime-related characteristics; and

• to validate the criminal attitude measure (CAM) on an independent sample of convicts.

The overall purpose of the current study is to explain the difference in crime-related characteristics based on criminal thinking among convicts through a valid and reliable instrument for Pakistani prisoners.

Method

Sample

The data was collected from the 1,949 male convicts incarcerated at nine Central Jails of Punjab located at Bahawalpur (n = 182), Sahiwal (n = 166), Multan (n = 245), Dera Ghazi Khan (n = 128), Mianwali (n = 181), Gujranwala (n = 218), Faisalabad (n = 303), Lahore (n = 265) and Rawalpindi (n = 261). Inclusion criteria were convicts who have committed any crime and are punished or sentenced by the court under that offense, incarcerated in central jail during the data collection period (July–October 2018). Convicts who gave their willful participation in the study and signed informed consent before participating were included. The sample was comprised of male convicts with ages ranging from 17 to 80 years (M = 36.14, standard deviation [SD] = 7.78) and monthly income ranging from 0 to 1000000 PKRs (M = 35837.73, SD = 58256.81). Imprisonment duration ranged from 15 days to 53 years (M = 6.59; SD = 4.64), where only four cases were above 25 years of imprisonment; excluding these cases makes an ignorable difference (M = 6.53, SD = 4.42). In this sample, 77.07% have committed one crime and 22.04% reported that they had been sentenced for multiple charges in their current conviction. A major offense committed was murder (n = 1331, 68.30%), narcotics (n = 267, 13.70%) and kidnapping (n = 166, 8.50%). There were 1,590 (81.58%) who reported a current conviction as their first-time offense. In addition, 337 (17.29%) respondents reported a criminal history. The frequency of previous crimes committed ranged from once (n = 141) to 10 times (n = 4). Age at the time of a previously committed crime ranged from 10 to 52 years (M = 23.16, SD = 7.73). There were 286 (14.67%) convicts who reported familial or acquaintance involvement in their crime. The age of the convict at the time of the relative crime ranged from before birth to 49 years (M = 16.29, SD = 9.99, missing = 28).

The sample was split for EFA (n = 649) and confirmatory factor analysis (CFA) (n = 1,300) through randomization. Thus, the sample characteristics are the same for both phases of the analysis, and significant differences between sample characteristics were absent.

Instrument

Indigenous criminal thinking scale

The 24 items of the ICTS measure criminal thinking with five factors: criminal rationalization, power orientation and justification, personal irresponsibility, vindication and entitlement (Sana and Batool, 2017). It is a five-point rating scale from disagree strongly (1), disagree (2), uncertain (3), agree (4) and agree strongly (5). The scores ranged from 24 to 120. The high score reflected the offender’s high criminal thinking and the low score indicated low criminal thinking. Cronbach’s alpha coefficient ranges from 0.72 to 0.58 for criminal rationalization to entitlement. For the composite score, the alpha reliability is 0.86.

Criminogenic cognitions scale

A 25-item measure assesses a distinct set of beliefs that serve to rationalize and perpetuate criminal activity (Tangney et al., 2002). It includes five subscales labeled as notions of entitlement/demand for respect, a short-term orientation toward goals, negative attitudes toward authority, an external locus of control or failure to accept responsibility for one’s actions and an insensitivity to the impact of crime on victims and society (Tangney et al., 2012). The total score reflects the average of the 25 items on a four-point scale. Items were rated on a four-point scale, with 1 = strongly disagree, 2 = disagree, 3 = agree and 4 = strongly agree. Items were averaged to create the five dimensions and a total criminogenic cognitions score.

Translation of criminogenic cognitions scale

Formal permission was obtained from the author to translate CCS into Urdu. The English version of CCS was given to five bilinguals for translation into Urdu. Bilinguals were briefed about the purpose and sample of the present study. Instructions for translation were provided to them, including consideration for cultural and educational equivalence concerning the sample. A committee approach was held to select the final Urdu statement, which was then given to five bilinguals for backward translation. The final version of the selected statements was sent to the author for feedback. After incorporating his feedback, the Urdu version was finalized.

In translating CCS, keeping the intensity of every sentence close to the original English version of the scale was considered important. In Item 9 the word “history” was retained in the back translation as “past,” because conceptually and sentence-structure-wise, it was more appropriate. In contrast, history is often not used in casual communication. The expression in Item 14 that “makes too big of a deal” was retained in the back translation as exaggerated. In item 21, “people in positions of authority” was reflected as people appointed/designated at higher posts in the back translation. Five sentences (1, 3, 5, 19 and 24) were suggested to be revised because of their initial discrepancy with the original version regarding the direction and connotation of sentences. For example, the expressions “want” and “deliver” were missing in Item 1, or translators interpreted them in a milder or more positive form. So, it was meant to be revised in the selected forward translation. Those revised items were then back-translated by five new translators, and because of their closeness to the original version, old translations were replaced by their revised version. The translation procedure used was based on the guidelines of Acquadro et al. (2012).

Data collection

Once the translated version was finalized, permission for data collection was obtained from Inspector General Prisons Punjab. The permission letter was shown to the Superintendents of all central jails in Punjab, along with the introduction and elaboration of the study purpose. Data was collected with the help of literacy teachers who were convicts but were paid by the Literacy Department (working under the Government of Pakistan) to teach other fellow prisoners. A positive aspect of assisted data collection was that rapport between teachers with their fellow prisoners had already been established. A team of volunteer teachers was formed, and the researcher then briefed them about the purpose of the research and instructed them on how to fill out their questionnaire and how they would distribute forms and collect data from either their students (convicted prisoners) or fellow inmates in their lockup cells. They distributed a fixed number of forms allotted to them concerning their willingness. They were instructed that if the respondent needed facilitation, they should only read the form and simply distribute the questionnaire and collect the data for literate prisoners. Informed consent was obtained through the signatures of respondents to ensure the anonymity of the respondent. Each form was later assigned a factual number, so the researcher did not know the names of the respondents, and the distributor remained unaware of the number assigned to the form. It was assured that the research process would not harm those who participated in the study, and the results would be kept confidential. The data were then analyzed using SPSS and AMOS version 25.

Data analysis

To measure the reliability and convergent validity of ICTS and CCS, the psychometric properties, internal consistency and convergent validity were established through descriptives, Cronbach’s alpha and bivariate correlation. To establish construct validity and determine the strength of each item in measuring the latent construct of criminal thinking, second-order CFA was carried out for ICTS and CCS. Because of the less missing data and the large sample size, the listwise deletion method was used to handle missing data. To evaluate the factor structure of instrument insignificant χ² is the most desirable index, but as it is greatly affected by sample size. Thus, if the sample size is greater, it is recommended to avoid decision-making based on significant χ² (Sharma et al., 2005). Criteria of good fit are specified as follows: the comparative fit index (CFI), incremental fit index (IFI) and goodness fit index (GFI) are greater than 0.95, whereas the root mean square error of approximation (RMSEA) and standardized root mean square residual (SRMR) for good fit have values less than 0.06 (Brown, 2015). Therefore, the χ² ratio should range from 2.0 – 5.0 to be considered acceptable (Hooper et al., 2008; Brown, 2015).

Poor factor structures showed that instruments lacked construct validity. Items are then pooled from these two measures scales to develop a new measure of criminal thinking labeled the CAM. Principal Component Analysis was carried out to extract components explaining the construct most appropriately. The varimax rotation was selected because of a low or insignificant inter-item correlation. Weak inter-scale correlations and the use of varimax rotation in different instruments measuring criminal thinking show these patterns are distinctive (Mitchell and Tafrate, 2012; Tangney et al., 2012). A total of 49 items were used as an initial pool (alpha = 0.90) while considering the theoretical framework of Walters (1996). Items with a factor loading less than 0.30 and a factor comprised of only one item were discarded from the scale (Field, 2013). In the end, 15 statements were selected, aiming to extract five factors based on the conceptual definition of the construct.

To replicate the factor structure of CAM, the extracted factor structure is then validated on a separate and independent sample through CFA. Second-order CFA was carried out on a new sample of convicts using AMOS software with a maximum likelihood estimation method and the same evaluation criteria (Ayubi et al., 2017). The proposed structural model was compared to a unidimensional model to further ensure the multidimensionality of the construct. The usefulness of the CAM is tested, where five criminal attitudes were taken as a dependent variable and group differences based on crime-related characteristics (i.e. number of cases filed and previous criminal history at the personal and familial level) were determined through MANOVA (Field, 2013). Post hoc findings of the associated univariate analysis were also reported, along with partial eta squared as an index to measure the effect size of the group difference.

Results

Validation of indigenous criminal thinking scale and extraction of criminal attitude measure

Table 1 shows the psychometric properties of the instruments, both translated and indigenously developed. A bivariate Pearson correlation was calculated to establish the convergent validity of ICTS with a translated version of CCS. Findings showed that the corrected item-total correlation of CCS ranged from 0.15 to 0.49. For ICTS, it ranged from 0.11 to 0.57. The correlation between CCS and ICTS was 0.58, p = 0.000. Moderate correlation between both instruments showed convergent validity (Field, 2013). However, both instruments lack internal consistency at the subscale level measured through Cronbach’s alpha (e.g. α = 0.37 for insensitivity to impact of crime from CCS and α = 0.47 for vindication and α = 0.44 for entitlement from ICTS).

CFA was carried out on both scales to further look into the construct validity of the instruments. Findings are presented in Figure 1 and Figure 2.

Figure 1 represents the factor loadings for CCS, where two items reported less than 0.3 loading (Field, 2013). For CCS, model fit indices were CFI = 0.65, Tucker–Lewis Index (TLI) = 0.58 and absolute index (RMSEA) = 0.07. The CFA of ICTS also showed poor model fit indices, with CFI = 0.62, TLI = 0.54 and RMSEA = 0.10. The values of indices were far less than conventional cut-offs (Shi et al., 2019). Factor loadings for ICTS shown in Figure 2 showed that four items reported low factor loadings as per criteria of 0.30 (Field, 2013). These model fit indices were too low to claim the validity and alpha reliabilities values (e.g. 0.3 and 0.4) for the subscales in instruments.

Thus, it was decided to develop a new short, valid and reliable instrument to measure criminal thinking using an item pool of both scales as an initial point. The items from both scales were considered as there was overlap in the statements, and both scales have their weak items for the current sample (e.g. items of vindication showed poor content validity as per subject matter experts in the present scenario). Therefore, the authors sought permission to use their respective scales as an item pool for developing a new instrument to measure criminal thinking.

Extracting criminal attitude measure

Value for Kaiser–Meyer–Olkin measure of sampling adequacy was 0.81. Bartlett’s test of sphericity measured approx. chi square (df) = 1846.95(105), p = 000 (Field, 2013).

The findings of Table 2 showed that a factor structure emerged, explaining 59.20% of the variance in the construct. The corrected item-total correlation of items ranged from 0.27 to 0.49, showing that each item contributed to the measurement of the construct. Five subject matter experts doing a Ph.D. in psychology were approached to name the components by identifying their common underlying characteristics. Factors were then named as follows: Power Orientation measures self-centered use of power to handle others and was comprised of items 9, 3 and 6 (alpha = 0.75). Mollification measures social inequality, injustice and pressure as reasons for crime through item numbers 8, 5 and 15 (alpha = 0.68). Entitlement emphasized the urgency for respect attainment and was measured through items number 7, 14 and 2 (alpha = 0.65). Suspiciousness or mistrust toward authorities was measured through items 10, 12 and 11 (alpha = 0.55). Short-term orientation was measured through Items 1, 13 and 4 (alpha = 0.50) and represents the urgency to attain things on the spot without worrying about the future. Literature has shown that Cronbach’s alpha is related to the length of the instrument, i.e. if the scale is too short, there will be a decline in the value of alpha, and 0.5 is considered sufficient (Tavakol and Dennick, 2011). In a multidimensional scale, at the subscale level, it is common to have a reliability of 0.5 (Taber, 2018). The inclusion of redundant items makes data collection more complicated, especially for sensitive samples such as prisoners (Taber, 2018; Tavakol and Dennick, 2011; Ishfaq and Kamal, 2023). The overall alpha reliability of the scale is 0.78, where a high score represents a greater specified criminal attitude and a low score depicts the opposite.

Construct validity

The bivariate correlation between subscales and the total was calculated to explore the coherence of the scale.

Table 3 presents the findings that all subscales are positively correlated with CAM, but these factors are weakly interrelated. The correlation between subscales and the total score on CAM was medium to high, showing that each pattern of criminal attitude (i.e. power orientation, mollification, entitlement, mistrust toward authorities and short-term orientation) significantly contributes to overall criminal attitude. By indicating that each subscale is an index of the overall construct, this pattern also provides evidence for the construct validation of criminal attitude. Findings showed a low to medium correlation between subscales, highlighting that these patterns are not highly related to one another and, thus, are distinctive and diverse (Field, 2013).

Structural validation of criminal attitude measure

The construct validation of CAM was established on an independent sample of convicts imprisoned in the central jails of Punjab, Pakistan.

The findings in Table 4 validated the five-factor solution for CAM (see Figure 3). The unidimensional model showed a poor model fit as compared to Model 2 (five-factor model), where the reported indices in Model 2 fulfilled the criteria of CFI, IFI and GFI > 0.95 (Shi et al., 2019). For Model 2, the χ² ratio was 2.87; RMSEA and SRMR are depicted good fit with values less than 0.06 (Hooper et al., 2008; Brown, 2015). The unidimensional model (Model 1) failed to achieve model fit as per all model fit indices.

Usefulness of criminal attitude measure

Crime-related factors such as intensity/diversity of crime and previous criminal record, both personal and familial, were also included to test the usefulness of the CAM in discriminating against convicts involved in intense reoffending.

Table 5 shows the mean (M) and SD for group comparison on subscales of CAM. A one-way MANOVA was carried out to determine whether multiple cases filed in current convictions influenced the criminal attitudes of convicts. The overall analysis yields a significant main effect of the number of cases filed (see Table 5 for M and SD) on criminal attitudes (Wilk’s λ = 0.95, F(5,1279) = 14.10 with p < 0.001 and ε² = 0.05). The follow-up univariate analysis showed that scores on power orientation (F(1,1283) = 54.11 with p < 0.001 and ε² = 0.04) are higher for convicts with multiple cases (M = 7.61, SD = 3.59) than convicts sentenced for single cases (M = 6.06, SD = 2.97). A similar univariate trend was found on the scores for mollification (F(1,1283) = 20.60 with p < 0.001 and ε² = 0.02), which are higher for convicts with multiple cases (M = 11.83, SD = 2.98) than convicts sentenced for single cases (M = 10.80, SD = 3.47). For entitlement, univariate analysis (F(1,1283) = 16.34 with p < 0.001 and ε² = 0.01) showed convicts with multiple cases (M = 8.29, SD = 2.39) scored higher compared to convicts sentenced for single cases (M = 7.63, SD = 2.41). Mistrust toward authorities yields a significant univariate difference (F(1,1283) = 12.60 with p < 0.001 and ε² = 0.01), where convicts with multiple cases (M = 8.42, SD = 2.23) scored higher than convicts sentenced for single cases (M = 7.87, SD = 2.32). A similar univariate trend was found for short-term orientation (F(1,1283) = 20.40 with p < 0.001 and ε² = 0.02), where convicts with multiple cases (M = 7.47, SD = 2.21) scored higher than convicts sentenced for single cases (M = 6.78, SD = 2.27).

A one-way MANOVA was carried out to determine whether previous personal criminal records influenced the criminal attitudes of convicts. The overall analysis yields significant main effect (see Table 5 for M and SD) for criminal attitudes (Wilk’s λ = 0.98, F(5,1268) = 6.15 with p < 0.001 and ε² = 0.02). The follow-up univariate analysis showed that scores on power orientation (F(1,1272) = 22.64 with p < 0.001 and ε² = 0.02) are higher for recidivists (M = 7.33, SD = 3.51) than first-time offenders (M = 6.23, SD = 3.06). A similar univariate trend was found on the scores for mollification (F(1,1272) = 12.37 with p < 0.001 and ε² = 0.01) are higher for recidivists (M = 11.75, SD = 3.04) than first-time offenders (M = 10.89, SD = 3.43). For entitlement, univariate analysis (F(1,1272) = 6.54 with p < 0.01 and ε² = 0.01) showed recidivists (M = 8.15, SD = 2.35) scored higher compared to first time offenders (M = 7.69, SD = 2.41).

To evaluate the influence of a previous familial criminal record on the criminal attitudes of convicts, one-way MANOVA was carried out. The overall analysis yields a significant main effect (see Table 5 for M and SD) for criminal attitudes (Wilk’s λ = 0.98, F(5,1263) = 4.82 with p < 0.001 and ε² = 0.02). The follow-up univariate analysis showed that scores on power orientation (F(1,1267) = 6.68 with p < 0.01 and ε² = 0.005) are higher for convicts with a familial criminal record (M = 6.96, SD = 3.23) than convicts without a familial criminal record (M = 6.31, SD = 3.12). A similar univariate trend was found on the scores for mollification (F(1,1267) = 17.87 with p < 0.001 and ε² = 0.01), which are higher for convicts with a familial criminal record (M = 12, SD = 2.82) than convicts without a familial criminal record (M = 10.87, SD = 3.44). For entitlement, univariate analysis (F(1,1267) = 10.77 with p < 0.001 and ε² = 0.01) showed convicts with familial criminal record (M = 8.30, SD = 2.31) scored higher compared to convicts without familial criminal record (M = 7.68, SD = 2.41). Mistrust toward authorities yield a significant univariate difference (F(1,1267) = 6.15 with p < 0.01 and ε² = 0.005), where convicts with a familial criminal record (M = 8.37, SD = 2.18) scored higher than convicts without a familial criminal record (M = 7.92, SD = 2.31). A similar univariate trend was found for short-term orientation (F(1,1267) = 6.80 with p < 0.01 and ε² = 0.005), where convicts with a familial criminal record (M = 7.33, SD = 2.35) scored higher than convicts without a familial criminal record (M = 6.86, SD = 2.25).

Thus, findings showed that convicts who were involved in multiple crimes and reported familial criminal records scored high on all five criminal attitudes, whereas convicts with previous criminal records scored high on power orientation, mollification and entitlement.

Discussion and conclusion

In Punjab, criminal attitudes are linked with moral disengagement and psychological distress (Butt et al., 2019). Criminal attitude is an important factor to study among the prison population from the perspective of prison management as well as to control recidivism. Maintenance of pessimism is often carried out through selectively memorizing negative events and rumination (Syed et al., 2021). Revengeful and mistrustful attitudes prevailing among prisoners mark criminal attitudes as a baseline for future criminal behavior (Sheikh et al., 2022). Developed cities in Punjab, Pakistan (such as Faisalabad and Lahore) have their inherent socio-economic implications such as urbanization, societal disorganization, high population density and the free availability of weapons, which increase the vulnerability of homicide (Khalid et al., 2018). The current study explored crime-related characteristics and criminal thinking among convicts through a valid and reliable instrument. The findings showed that CAM is a short, valid, reliable and practically appropriate, i.e. a user-friendly, instrument to measure criminal attitudes among convicts. CAM also highlighted that criminal thinking is a multidimensional construct, where patterns that emerged in the current sample included power orientation, mollifications, mistrust toward authority, entitlement and short-term orientation.

Internal consistency

The current study showed that the alpha reliability of CAM is acceptable for research purposes (0.78), and for subscales, reliability of 0.50 is acceptable because of the overall low number of items in subscales, i.e. items of the short-orientation subscale, which represent urgency to attain things on the spot without worrying the about future, but it addresses three different aspects of short-term orientation. For example, the reluctance to plan, predict, or worry about the future represents different aspects of the same construct that can be the reason for low reliability. As per the expert opinion, the subscale has good face validity. Tangney et al. (2012) reported a 0.51 reliability of the short orientation subscale with five items. When it comes to criminal thinking scales, different studies have reported low reliabilities (<0.35) at the subscale level (Tonks and Stephenson, 2020; Walters, 2002a 2002b). All subscales showed a medium to strong correlation with the total score on CAM, making these factors a significant contributor to measuring criminal thinking among convicts. Low correlations between subscales reflect different thinking patterns associated with specific attributes and crimes (Tangney et al., 2012).

Structural validation of criminal attitude measure

The main objective was to establish the construct validity of the instrument and test the factor structure of the construct on a new sample of convicts (Ayubi et al., 2017; Sherretts and Willmott, 2016). The current study’s findings in Figure 1 reported an excellent model fit for CAM. Thus, CAM is a valid instrument to measure criminal thinking as it fulfills both criteria (EFA and CFA) suggested by researchers (Gibbons et al., 2021; Lace and Merz, 2020) to ensure the construct validity of the instrument. Furthermore, indices (e.g. GFI, CFI, RMSEA, SRMR and insignificant PCLOSE) presenting goodness of fit showed excellent measurement model fit (Brown, 2015) on a distinct and unique sample of convicts. For scoring, factor-level sums of scores are preferred, highlighting the multidimensionality of the construct.

Crime-related factors (utility of criminal attitude measure)

The current study showed that CAM could differentiate between recidivists and offenders, with multiple cases filed against them in recent convictions. This differentiation enhances the utility of CAM in law enforcement settings. The finding is consistent with the literature where criminal thinking instruments can differentiate between recidivist and first-time offenders, offenders committing multiple offenses in current conviction or charged for a single crime (Mandracchia and Morgan, 2011; Taxman et al., 2011). Many criminal behaviors coexist, which makes sentencing difficult (Sadiq et al., 2013; Pierce et al., 2015). The findings presented in Table 5 showed that convicts involved in more than one case in the current conviction scored high on power orientation, blaming society, entitlement, mistrust toward authority and short-term orientation. These findings are consistent with the claim of the Risk-Need-Responsivity Model, according to which criminal thinking is one of the central criminogenic needs to explain/predict criminal behavior (Andrews and Bonta, 1998, 2010).

In Punjab, limited and unpredictable resources led individuals to engage in risky explorations and select unconventional and novel ways to secure food and resources (Arif and Khan, 2019). Challenging circumstances such as natural disasters further reinforce the entitled attitude, especially in areas with strong ethnic backgrounds (Afzal and Nasreen, 2019). The situation gets worse as criminals believe that the law enforcement department in Punjab, especially Lahore, is highly influenced by power and politics (Siddiqi et al., 2014). Thus, having power and entitlement has functional value in Punjab, Pakistan. Strains common in certain Asian societies include teacher abuse, corruption, threats to traditional values and discrimination associated with rural-to-urban migration (Zaman, 2021). Parental strain, teacher strain and victimization were reported to be criminogenic (Lin and Mieczkowski, 2011).

Defense mechanisms used by prisoners include blaming, isolation, projection, acting out and displacement to project anger and frustration (Cheema et al., 2022). Blaming others (e.g. poor social skills) and desire for immediate gratification (e.g. impulsivity) are common among violent offenders (Shafqat et al., 2019). Khan et al. (2022) report restlessness, feeling irritated by their surrounding environment, low frustration tolerance and emotional exhaustion. A study conducted on the impact of movies on juvenile delinquents showed that they tend to watch sexual content in the movies, and their friends accompany them in watching movies and drug usage, showing their inclination toward immediate gratification (Sajid et al., 2022). Similarly, snatching hotspots were found in the densely populated areas of the city lacking streetlights, and the hotspots of robberies were observed in the posh areas of the city (Khalid et al., 2018).

Repetition of crime or reoffending behavior is one of the major concerns of criminologists (Luigi et al., 2020; Moore and Eikenberry, 2021). Recidivism, operationalized as previous arrest, lockup and incarceration in the present study, reports a significant group difference in different criminal thinking patterns (see Table 5), where convicts with a previous personal criminal record scored high on power orientation, entitlement and mollification as compared to convicts without a previous criminal record. Individuals tend to make excuses or blame external factors (stereotypes, discrimination, labeling, etc.) without accepting responsibility (Vrućinić, 2019). It is also essential to consider that convicts must have committed their previous crime at a relatively young age, so cognitive immaturity or a lack of interpersonal skills can also be contributing psychological factors (Sherretts et al., 2017). From the parental criminal record, Capaldi et al. (2021) reported that parental antisocial behavior predicts an individual’s arrests, and this relationship is significantly moderated by his late childhood cognitive function. Parental clashes, family disputes and interpersonal violence are associated with crime in Pakistan (Aslam et al., 2015). The findings are consistent with literature where the presence of familial criminal records differentiates convicts with high criminal attitudes. Prisoners in Pakistan have reported that their parents, other family members and caregivers often beat them for minor things and tend to punish them without any mistakes (Ali et al., 2023). Sometimes, the beating was so hard that it resulted in injuries such as broken bones or head injuries (Anjum and Bano, 2018). In addition, parents tend to ridicule, criticize, blame and taunt others, especially in the presence of others, generating feelings of insecurity and rejection as unwanted children (Ishfaq and Kamal, 2022). Early exposure to paternal violence increases the risk of becoming a violent person, and a higher conviction risk is associated with having a convicted parent (Van-de-Weijer et al., 2014).

Limitations

In the current study, using the survey method was inevitable because of the restriction imposed by the granted permission (i.e. three days for each jail). This was practically impossible as these prisons were in different cities, and usually, one day was required to build rapport with jail officials, staff, literacy teachers and convicts. However, this time duration was extended because of the courtesy of the Superintendent and Deputy Superintendent of each jail. The phenomena proposed (based on large data sets) in the present study can further be elaborated using qualitative research designs and methods (using a small sample with an in-depth study). So, it is also suggested to test this new instrument on a comparative study between prisoners and non-prisoners to explore whether the scale can differentiate between these two groups. The current study is focused on male convicts only, and the findings cannot be generalized to a female sample.

Implications for practice

For the risk assessment of criminals and prisoners, criminal thinking is the backbone of the risk assessment model. To address the cultural and legislature-driven differentials, indigenously developed or adapted instruments are essential to applying internationally accepted approaches of Western developed countries in an underdeveloped and non-western country such as Pakistan. The validity and reliability of an instrument are prerequisites for addressing the potential pitfalls of assumptions about applicability and utility.

CAM proved to be a valid, short and practically useful instrument to measure criminal attitudes among prisoners. Accurate assessment makes it easy to identify, target, manage and change risk factors such as criminal thinking among convicts to control recidivism, prisonization or in-prison violence by directing therapeutic interventions toward specific concerns.

Compliance with ethical standards

All procedures performed in studies involving human participants were per the ethical standards of the institutional research committee named the National Ethic Committee for Psychological Research (NECPR), which ensured that no potential harm could occur to respondents by looking into data protection strategies including anonymity of data, willful participation (informed consent) and the provision to withdraw from the data collection process whenever they wanted.

Figures

Figure 1

Structural validation of criminogenic cognition scale among convicts

Figure 2

Structural validation of Indigenous criminal thinking scale among convicts

Figure 3

Structural validation of criminal attitude measure among convicts (N = 1300)

Table 1

Bivariate correlation and descriptives of the study variables (N = 643)

		1			2			3	4	5	6	7	8	9	10	11	12
1	Criminogenic cognitions scale		–		0.80**			0.73**	0.72**	0.64**	0.71**	0.58**	0.51**	0.44**	0.45**	0.33**	0.49**
2	Notions of entitlement				–			0.48**	0.51**	0.41**	0.45**	0.46**	0.40**	0.35**	0.37**	0.21**	0.41**
3	Short-term orientation							–	0.46**	0.28**	0.39**	0.47**	0.38**	0.38**	0.43**	0.25**	0.31**
4	Failure to accept responsibility								–	0.23**	0.40**	0.47**	0.43**	0.40**	0.38**	0.19**	0.36**
5	Negative attitudes toward authority									–	0.36**	0.35**	0.33**	0.17**	0.24**	0.31**	0.35**
6	Insensitivity to the impact of crime										–	0.33**	0.28**	0.29**	0.19**	0.22**	0.32**
7	Indigenous criminal thinking scale											–	0.82**	0.80**	0.84**	0.61**	0.72**
8	Criminal rationalization												–	0.55**	0.60**	0.39**	0.58**
9	Power orientation and justification													–	0.60**	0.37**	0.44**
10	Personal irresponsibility														–	0.37**	0.44**
11	Vindication															–	0.41**
12	Entitlement																–
	Number of items	25			5			5	5	5	5	24	5	6	6	3	4
	Cronbach’s alpha	0.81			0.55			0.64	0.50	0.58	0.37	0.86	0.57	0.58	0.77	0.47	0.44
	M	2.25			2.26			2.07	2.04	2.63	2.25	54.17	11.86	12.86	11.35	8.39	9.70
	SD	0.44			0.65			0.62	0.59	0.64	0.54	11.81	3.02	3.50	4.06	2.14	2.54
	Skew	−0.14			−0.01			0.25	0.26	−0.15	−0.12	0.03	−0.15	0.01	0.63	−0.53	0.02
	Kurt	0.29			−0.20			−0.41	−0.50	−0.20	0.01	−0.46	−0.10	−0.63	−0.25	−0.17	−0.32

Notes:

Skew = skewness; Kurt = kurtosis; **p < 0.01

Source: Created by the authors

Table 2

Five-factor solution using principal component analysis with varimax rotation (N = 643)

	Components
	1	2	3	4	5
CAM3	0.84	0.08	0.02	0.05	0.08
CAM9	0.82	−0.03	0.17	−0.01	0.02
CAM6	0.75	0.14	0.05	0.09	0.15
CAM8	0.04	0.79	0.11	0.09	0.07
CAM15	0.07	0.71	0.18	−0.02	0.01
CAM5	0.07	0.69	0.06	0.31	0.12
CAM7	0.12	0.14	0.73	0.00	0.17
CAM14	0.07	0.17	0.70	0.14	0.12
CAM2	0.05	0.11	0.69	0.34	0.05
CAM10	0.10	0.07	0.09	0.75	0.04
CAM12	0.04	0.02	0.13	0.73	0.24
CAM11	−0.03	0.30	0.19	0.54	−0.08
CAM1	−0.03	0.25	−0.03	−0.01	0.77
CAM13	0.14	−0.02	0.21	0.11	0.63
CAM4	0.23	−0.08	0.35	0.18	0.55

Note:

Italic values show strength of items that constitute a factor

Source: Created by the authors

Table 3

Correlation between CAM and its subscales (N = 643)

Sr. No.	Variables	1	2	3	4	5	6	M	SD	Skew	Kurt
1	Criminal attitude measure	–	0.56**	0.65**	0.74**	0.67**	0.65**	36.37	7.69	−0.24	−0.00
2	Power orientation		–	0.17**	0.25**	0.15**	0.28**	5.62	2.34	0.63	−0.49
3	Mollification			–	0.35**	0.34**	0.23**	8.73	2.43	−0.52	−0.49
4	Entitlement				–	0.42**	0.39**	7.70	2.45	−0.19	−0.62
5	Mistrust toward authorities					–	0.28**	7.71	2.31	−0.05	−0.53
6	Short-term orientation						–	6.62	2.20	0.30	−0.39

Notes:

Skew = skewness; Kurt = kurtosis

Source: Created by the authors

Table 4

Confirmatory factor analysis for CAM among convicts (N = 1,300)

		Goodness of fit
Facets	Model	χ2 (df)	χ2/df	GFI	IFI	CFI	RMSEA	SRMR	PCLOSE
0	1	1,208.54(90)***	13.43	0.70	0.69	0.68	0.10	0.11	0.000
5	2	253.80(85)***	2.87	0.97	0.96	0.95	0.04	0.04	1.000

Notes:

Model 1 = unidimensional model; Model 2 = five-factor model; GFI = goodness-of-fit index; IFI = incremental fit index; CFI = comparative fit index; RMSEA = root mean square error approximation; SRMR = standardized root mean square residual

Source: Created by the authors

Table 5

Mean scores (and standard deviations) on CAM subscales as a function of different crimes characteristics (N = 1,300)

	Crime characteristics
	Case filed in current conviction		Previous personal criminal record		Previous familial record
Criminal attitudes	Single (n = 1,003)	Multiple (n = 282)	Yes (n = 225)	No (n = 1,049)	Yes (n = 184)	No (n = 1,085)
Power orientation	6.06 (2.97)	7.61 (3.59)	7.33 (3.51)	6.23 (3.06)	6.96 (3.23)	6.31 (3.12)
Mollification	10.80 (3.47)	11.83 (2.98)	11.75 (3.04)	10.89 (3.43)	12 (2.82)	10.87 (3.44)
Entitlement	7.63 (2.41)	8.29 (2.39)	8.15 (2.35)	7.69 (2.41)	8.30 (2.31)	7.68 (2.41)
Mistrust toward authorities	7.87 (2.32)	8.42 (2.23)	8.26 (2.30)	7.93 (2.29)	8.37 (2.18)	7.92 (2.31)
Short-term orientation	6.78 (2.27)	7.47 (2.21)	7.15 (2.33)	6.89 (2.25)	7.33 (2.35)	6.86 (2.25)

Source: Created by the authors