Investigating the students’ behavior towards the temptation to do academic misconduct in higher education: The moderation of religiosity

The purpose of this study was to investigate the factors that influence cheating intention (CI) and cheating behavior (CB) on accounting students with Islamic religiosity (IR) as a moderating factor. This study used the theory of planned behavior (TPB) concept in predicting academic cheating. The sample of this study used a cluster sampling approach. The data collecting technique used a survey method with a questionnaire. The population of this study was accounting students at an Islamic university in Yogyakarta, Indonesia with a total of 146 respondents. Regression analysis was employed to test this hypothesis by using a structural equation modeling (SEM) approach using smartPLS. The results of this study indicated that attitude toward cheating (ATC) and IR had a negative effect on CI, and perceived behavioral control (PBC) had a positive effect on CI and CB. However, it has no effect when moderated by religiosity. This finding also implied that religiosity was unable to moderate TPB to reduce students’ desire to commit academic misconduct. This paper contributes to the literature by providing empirical evidence of the moderating effect of religiosity on academic misconduct of students in higher education.


Investigating the students' behavior towards the temptation to do academic misconduct in higher education: The moderation of religiosity Introduction
Academic misconduct is a discreditable behavior and violates ethics Sarita & Dahiya, 2015) that may threaten academic integrity (Sarita & Dahiya, 2015). When unethical behavior such as cheating is conducted habitually, it is not impossible that the same thing will happen when the student enters the world of work (McCabe et al., 1996;Nonis & Swift, 2010;Sims, 2010;Weber & Gillespie, 1998), since Individual behavior is, to a large extent, determined by the experiences they have experienced (Greiner, 1998). Therefore, educational institutions must try to anticipate academic misconduct, apart from being providers of educational services, they also have moral responsibility in shaping the students' character that prioritize universal values.
There are certain studies in the case of academic misconduct, cheating, and (e.g. Crown & Spiller, 1998;Kisamore et al., 2007;McCabe et al., 2006;Stone et al., 2007;Stone et al., 2010) Those researchers generally use behavioral theory models based on demographic, situational, and personality variables (Stone et al., 2010). In predicting academic misconducts, another approach that can be used is based on the theory of planned behavior (TPB). The concept of TPB is considered the most appropriate to be used in predicting unethical behavior (Chang, 1998) and academic misconducts (Stone et al., 2010). Several researchers who have employed the TPB concept include Beck and Ajzen (1991), Passow et al. (2006), Harding et al. (2007), and Kisamore et al. (2007). With the exception of research by Beck and Ajzen (1991) and Stone et al. (2010), these studies modified the TPB concept (e.g. Passow et al., 2006;Stone et al., 2007), not based on the original concept (Harding et al., 2007;Passow et al., 2006), so that the relevance of TPB in predicting academic misconducts remains very difficult to be recognized.
In religious life, religiosity is often a subjective issue and to test this variable is demanding. Someone who has a religious personality does not reflect that the person is forever able to promote honest behavior in every aspect of life. Religiosity is believed to have an impact on behavior (Marquette, 2010), including cheating and other forms of academic misconduct. However, each individual has a different interpretation of religious life and this will have an impact on attitudes (Pavarala & Malik, 2010), behavior, and subjective norms (SN) (Pavarala & Malik, 2010;Weaver, 2002).
Based on the prevailing beliefs in Islam, in general, every follower must comply with the provisions of religious teachings in acting and behaving. Previous research that used the TPB concept in predicting academic misconduct only focused on psychological and behavioral aspects (Passow et al., 2006;Harding et al., 2007;Stone et al., 2007;Stone et al., 2010). This study use the basic model of TPB. Nonetheless, the researcher adds the variable Islamic religiosity (IR) personality (religious belief and religious behavior) as an antecedent factor in academic misconducts. The IR variable is considered very suitable for Indonesian academic life since practically every educational institution, both religiously affiliated and general, includes courses related to religion, especially Islam.
Based on the description, it can be perceived that previous studies have found inconsistent results related to the variables that affect cheating intention (CI) and cheating behavior (CB). Hence, further research is required, namely the factors that affect CI and CB in order to obtain information from the variables that have an effect on CI and CB more precisely compared to the previous studies. Therefore, this study scrutinizes the effect of TPB on CB and the role of IR as a moderating variable, as well as the correlation between CI and CB by students. It is expected that this research can contribute to the issue of students' academic misconduct behavior in Indonesia, so that they the appropriate curriculum can be applied to minimize the risk of academic misconduct in the future.

Literature Review TPB and Academic Misconduct
Although the use of TPB is able to predict behavior TPB (Ajzen, 1991) the application is still considered very rare in academic environment. Most researchers (e.g. Passow et al., 2006;Harding et al., 2007;Kisamore et al., 2007) modified the model so that the relevance of the original model became more difficult to understand. Passow et al. (2006) used TPB as an ex-post facto explanation, to examine the factors that drive behavior. Likewise, the study of Stone et al. (2007) who both used ex-post facto in the TPB component. These studies are an illustration of the application of the TPB concept in predicting academic misconduct. The next section of this paper will explain the concept of TPB from the point of view of academic integrity.
Researchers who directly applied the TPB concept in predicting academic misconducts were Beck and Ajzen (1991) which were applied to 146 psychology students to predict cheating and dishonesty in doing assignments and academic exams. To test the accuracy of the model, Beck and Ajzen (1991) applied two tests through the original TPB model and through a modified model by adding moral obligation variable which was considered to increase the prediction of academic misconducts. The results depicted that, of all the components of the TPB, namely attitudes, SN, perceived behavioral controls (PBC) it was found that behavioral controls were the differentiating variables in academic misconducts. Ajzen (1991; argued, when the resources and opportunities for cheating were absent in the will, then behavioral controls may be better at predicting misconduct than attitudes and norms. While the second test, using a modified TPB model, indicated that moral obligations only contributed 3% of the variance to academic misconducts, and this was further supported by Whitley (1998) who found the TPB to only have a 27.8% variance in CB.
In addition to Beck and Ajzen (1991), there are several other researchers who have modified the TPB in predicting academic cheating (e.g. Passow et al., 2006;Harding et al., 2007;Stone et al., 2007). Passow et al. (2006) made a modification by adding a moral obligations component, as a separate variable from the TPB. In general, the results have not been able to make a significant contribution to the accuracy of the model, and the research was carried out without testing CI. In contrast to its predecessors, by modifying the model, Harding et al. (2007) concluded that moral obligation is a driving factor for academic misconduct. Demographic variables (gender, discipline, fraternity membership) and previous CB, on the other hand, act as moderate variables between intention and CB. Model modification developed by Harding et al. (2007) formed a variant of cheating in doing homework by 27%. These findings indicate a strong modification model relationship compared to the findings of Passow et al. (2006).
Neither the research of Harding et al. (2007) nor Passow et al. (2006) concluded that the behavior control variable is the factor that significantly influences academic misconduct. The intention to engage in misconduct is possibly due to various factors. Stone et al. (2007) in his research on 217 students majoring in business found attitudes, SN, and PBC can determine a person's intention to cheat. Individuals who care about how their social group perceives them are more likely to change their attitudes toward certain behaviors (Sharma et al., 2021;Yu et al., 2021). The existence of external pressure encourages individuals to engage in socially desirable behavior (Sharma et al., 2021). This situation can be mediated by religiosity factors such as religious behavior and religious belief (Pavarala & Malik, 2010;Marquette, 2010;Michelson, et al., 2007;Rettinger & Jordan, 2005). Based on the above explanation, the research hypotheses are as follows: H1: TPB has a negative effect on CI H2: PBC have a negative effect on CB H3: CI has a positive effect on CB

Islamic Religiosity and Academic Misconduct
Since centuries ago, religious life has become a topic that is often discussed in psychology, especially in the western world. Religiosity is known to have dominated people's lives (Al Marri et al., 2013). In general, religiosity can be interpreted as a belief that is used as a way of reflecting on life based on individual/group values and attitudes (Fam et al., 2004). These values can shape the behavior and actions of a person or group (Eid & El-Gohary, 2015). Individuals with high level of religiosity tend to avoid deviant behavior (Mustapha et al., 2016;Baier & Wright, 2001) such as cheating.
In the view of Islam, a person is prohibited from taking other people's wealth without the consent of the owner (Moten, 2014) and all forms of deceit such as cheating are considered as behavior that deviates from religious teachings. Islam guides its followers to behave honestly and academic cheating is an act that is prohibited and must be avoided (Mustapha et al., 2016). Cheating and any other form of academic misconduct is an act that cannot be tolerated, and Islam firmly rejects it. In certain countries such as Indonesia, the role of religion has a very central role, especially the Muslim community (Azizah, 2006). The majority of ethical values originate and are determined based on religious teachings (Hadjar, 2017). Islam prioritizes fairness and honesty as characteristics that must be upheld by every follower (Hadjar, 2017;Aldulaimi, 2016) and all forms of cheating are behaviors that violate religious provisions.
Morally, cheating is an unacceptable behavior in society (Dodeen, 2012), although this is often and easily found in everyday life. Cheating, generally, occurs in the lives of the younger generation (Hadjar, 2017). Unfortunately, this behavior also often involves academics, teachers or lecturers. In addition to academic sanctions, regulations based on religion are expected to encourage honest and fair behavior because the academic process is an effort to build moral character and uphold integrity so that it can be universally accepted.
It cannot be denied that religion has an important and inseparable role in one's life. Obedience to such a religion will affect the formation of behavior. In connection with this, Weaver & Agle (2002) in their research concludes, all regulations based on religion will affect ethical behavior, and this can be mediated by religious identity and religious motivational orientation factors (Weaver & Agle, 2002). This opinion was later supported by Michelson, et al. (2007) by arguing that students who commit academic misconduct are less likely to engage in religious activities.
Research conducted by Michelson, et al. (2007) as described above further strengthen the research that has been done (Rettinger & Jordan, 2005). The two conclusions of the study agreed that the intensity of a person's involvement in religious activities will have an impact on the behavior in question. In other words, the higher a person's involvement in religious activities, the lower the likelihood of their involvement in academic misconduct. Obedience to religion is believed to have an impact on behavior (Marquette, 2010), including cheating and other forms of academic misconduct. Each individual has a different interpretation of religious life and this may affect attitudes (Pavarala & Malik, 2010) and behaviors (Pavarala & Malik, 2010;Weaver, 2002).
In religious life, religiosity is often a subjective issue. To test the religiosity variable is not an easy job. Someone who has a religious personality does not mean that the person is forever able to promote honest behavior in every aspect of life. Thus, IR in addition to attitudes and behavior is also an important factor that influences SN (Jahya, 2004). Indonesian society is known as a religious society that puts forward the provisions of life based on religion, especially Islam. Every action taken in general must be in accordance with the rules set by religious beliefs. Several previous studies that used the basic concept of TPB in predicting academic misconduct only focused on psychological and behavioral aspects (Passow et al., 2006;Harding et al., 2007;Stone et al., 2007;Stone et al., 2010). By adding the religiosity variable, this research becomes very relevant to be applied in academic life in Indonesia. The inclusion of the religiosity variable is also a differentiator of this study with previous studies.
Based on the above literature review, it is attained that there are limitations to the application of the TPB basic model in predicting academic misconduct. There are only two studies that have applied the basic concept of TPB (Beck & Ajzen, 1991;Stone et al., 2010). Therefore, to predict CI and CB, this study employed the basic TPB model by adding components of religious behavior and religious belief as moderating variables. Based on the explanation, the research hypotheses are as follows: H4: IR has a negative effect on CI. H5: TPB has a negative effect on CI with IR as the moderation

Research Methods
The population in this study were all active students of the Faculty of Business and Economics, Islamic University of Indonesia, Yogyakarta. Cluster sampling approach was employed as sample determining carried out where the sampling unit consists of one group. The units referred to in this study were students who were divided into ongoing semester intervals. The student units who were the respondents were students entering the second year. This determination was chosen because students in the second year were considered to have had learning experience and will still continue their academic activities. Thus, the total sample that fill out the questionnaire were 325 students.
Variable measurement of the attitude toward cheating (ATC) was measured in 7 question items. Furthermore, each of the PBC, intentions, CB variables had 3 statement items. Finally, the SN and IR variables were each measured by 4 question items. The indicator variable ATC, SN, PBC, CI, and CB were adopted from the research of Thomas et al. (2010). Meanwhile, the IR variable was adopted from Eid and El-Gohary (2015). This study used a Likert scale measurement of 1-6 from the answer choices of strongly disagree to strongly agree.

Validity and Reliability Testing
Convergent validity is the value of average extracted variance (AVE). The AVE value described the variance or diversity of the manifest variables owned by the latent construct. Thus, the greater the variance or diversity of the manifest variables contained in the latent variable, the greater the representation of the manifest variable on the latent construct that was carried out. This study also examined discriminant validity, namely validity that measured how different a measurement instrument was from other measurement instruments in measuring the same variable. The discriminant validity test was carried out using the Fornel and Larcke criteria, namely comparing the value of the square root of AVE with the correlation value between constructs. The measurement was declared valid if the AVE root value of each construct was greater than the correlation value between the constructs.
The reliability test of this study was conducted by looking at the Cronbach's Alpha value with the provision that the value above 0.7 which was a good measure of the reliability test although the value of 0.6 was still acceptable. From the results of the evaluation of the measurement model, the rule of thumb was summarized as shown in Table 1.

Data Analysis Methods
The research hypotheses testing was conducted through a Structural Equation Modeling (SEM) approach using smart-PLS. SEM was chosen because this multivariate technique could combine aspects of multiple regression and factor analysis to estimate a series of dependent relationships simultaneously. In addition, SEM was able to explain measurement errors in the estimation process that regression failed to explain.
The hypothesis was declared as supported if the t-statistic value in the output path coefficients of the smart-PLS application was greater than 1.96. The next thing to consider was the value of the original sample. The original sample value described the direction and the strength and weakness of the relationship. The original sample value was close to -1 or 1 indicating a strong relationship between variables. While the original sample value close to 0 indicated a weak relationship between variables.

Results and Discussion
Questionnaires that had been distributed totaled to 325 questionnaires. Then, there were 146 questionnaires returned in full and tested or totaled to 45%. Furthermore, the SmartPLS statistical test tool was carried out to assist in data processing. The results of data collection were displayed in Table 2.

Convergent Validity Results
The results of the convergent validity test of the instruments used in this study were comprehended from the value of the loading factor and AVE using factor analysis. In the first test of convergent validity using confirmatory factor analysis, it was found that some of the question items and dimensions did not meet the valid standard, namely having a loading factor value below 0.5. Thus, all variables and dimensions that did not meet the validity criteria were not included in the next test. Table 3 demonstrates the indicators and dimensions that were not included in further research, namely ATC2, PBC3, IR3, IR4, CB1, and CI1. 0,834 Note: ATC1-ATC7 is an indicator of attitude toward cheating. CB1-CB3 is an indicator of cheating behavior. CI1-CI3 is an indicator of cheating intentions which is an indicator of subjective norms. Furthermore, PBC1-PBC3 is an indicator of perceived behavioral control. IR1-IR4 is an indicator of Islamic religiosity The testing of the outer model was redone using valid instruments. The result was that all question items and dimensions included in the test had a loading factor value above 0.5. This indicated that all items that was included in the next test have met the convergent validity (see Table 4).

Discriminant Validity Results
Discriminant validity was employed to test the validity of a model. Discriminant validity was perceived through the cross loading value which demonstrated the magnitude of the correlation between constructs and their indicators and indicators from other constructs. The standard value used for cross loading must be greater than 7 or by comparing the square root of average variance extracted (AVE) value of each construct with the correlation between the construct and other constructs in the model. If the AVE root value of each construct is greater than the correlation value between the construct and other constructs in the model, then it is said to have a good discriminant validity value. The results can be seen in Table 5.  Table 5, it can be observed that all AVE values were higher than each of the correlations of the latent variables. Thus, it was concluded that this measurement meets the requirements of discriminant validity based on the Fornell-Larcker test. Subsequently, composite reliability measured the real value of the reliability of a construct. The rule of thumb or composite reliability must be greater than 0.7 even though a value of 0.6 was still acceptable. Testing the reliability of the variables in this study was carried out with the PLS application and the results can be seen in Table 6.  Table 6 displayed that the value of Cronbach's Alpha on the variables BC, IC, and IR was below 0.6. Meanwhile, all composite reliability values were greater than 0.7. Thus, all measurement items were declared reliable. It means, each indicator for each variable can consistently be trusted in measuring each research variable

Hypothesis Test Results
The hypothesis was accepted if the original sample value was close to 1 (positive relationship) or close to -1 (negative relationship). The next criterion was that the T-statistic value was greater than 1.96 or the p-value was less than 10%, 5%, and 1 %alpha.  Table 7 illustrated that the ATC had a negative effect on the CI variable with a value of -0.233 and a p-value of 0.002<0.01. Furthermore, the PBC had a positive effect on the CI variable with a value of 0.215 and a p-value of 0.015<0.05. However, the results of the relationship between the SN and CI were stated to have no effect because the p-value was 0.279> 0.05. Furthermore, the IR variable had a negative effect on CI with a value of -0.259 and a p-value of 0.008 <0.01. Then, PBC had a positive effect on the CB variable with a p-value of 0.093<0.1. While the CI variable had an effect on the CB variable with a p-value of 0.000.
On the other hand, the TPB variable (ATC, PBC, and SN) on the variable CI with IR moderation was declared to have no effect, this was because each p-value (0.928), (0.931), and (0.531) was greater than 0.05.

Theory of Planned Behavior and Cheating Intentions
Attitude is the amount of affection a person feels to accept or reject an object or behavior and is measured by a procedure that places the individual on a two-pole evaluative scale, for example good or bad, agree or reject, and others (Ajzen & Fishbein, 1980). ATC had a negative influence on CI. This proved the preceding research conducted by Kisamore et al. (2007) and Stone et al. (2010) that student attitudes affected a person to reduce the intention to cheat. According to Dewanti et al. (2020) Parents' motivation and social sanctions for students are known to commit misconduct, and they wanted to create a competitive learning environment that had an impact on students not to cheat. This phenomenon can occur because a student who considers academic cheating to be erroneous is not always followed by a decreased intention not to commit academic misconduct (Wijayanti & Putri, 2016).
By definition, SN were social factors that influence a person's attitude or behavior to take or not to take an action (Dewanti et al., 2020;Hagger & Chatzisarantis, 2006). The findings in this study indicated that SN had no effect on CI. This finding was supported by Dewanti et al. (2020) that social factors were able to influence a student to commit misconduct. An unsupportive environment would likely result in a disregard for SN.
The results of this study indicated that PBC had an effect on CI. When a student felt that academic cheating was easy and profitable to do both in doing assignments, it formed the intention of the student to cheat in doing assignments or taking exams (Dewanti et al., 2020;Kisamore et al., 2007;Wijayanti & Putri, 2016).

Perceived Behavioral Control dan Cheating Behavior
In this study, PBC was defined as the degree of ease or difficulty felt by a person with respect to performing ethical behavior. This study resulted in findings explaining that there was positive correlation between PBC and CB. Thus the hypothesis in this study is not supported. PBC is very important because it shows how effectively a person is able to resolve ethical conflicts that exist within itself (Bandura, 2001;Compeau et al., 1999;Wang et al., 2022). PBC has been found to be a significant predictor of behavioral intention. The findings in this study support previous research which found that PBC can lead to the intention to commit piracy (Harding et al., 2007;Hongwei et al., 2017;Passow et al., 2006;Sharma et al., 2021). Therefore, institutions need to help students develop a strong sense of self-control to better deal with academic cheating. Meanwhile, when someone's rationalization did not affect the intention to commit academic misconduct, then there were other factors that would influence someone to commit academic misconduct (Wandayu et al., 2019). This indicated that control over behavior could not be the main deterrent to students from cheating, because there were environmental influences that supported or form intentions to misconduct behavior (Dewanti et al., 2020). Students with lack of supervision or weak sanctions would tend to have the intention to commit academic misconduct (Salsabilla, 2020). Hence, it required the participation of the social and academic environment with integrity in reducing cheating (Salsabilla, 2020).

Cheating Intentions and Cheating Behavior
This study supported previous study conducted by (Salsabilla, 2020;Stone et al., 2010), stated that CI had a significant effect on CB. This finding demonstrated that the intention to cheat might lead to CB. If students already had the intention or plan to cheat, then this would be followed by committing actual academic misconduct (Salsabilla, 2020). Misconduct stigma might affect the relationship between intentions and behavior in a different way compared to favorable behavior (Stone et al., 2010). Furthermore, according to Stone et al. (2010), academic dishonesty was not only learned from observing peer behavior, but peer behavior provided a kind of normative support for CB. By getting used to a student cheating, CB was formed whose purpose was to make it easier for them to do assignments and pragmatically to get profits easily.

Islamic Religiosity and Cheating Intentions
Regarding the relationship between IR and attitude, the results displayed that IR had a significant negative correlation with CI to commit academic misconduct. According to Hongwei et al. (2017) that students who attended religious activities more often tended to cheat less often than students who rarely attended religious activities. If a person was very religious, then he would likely have a low level of tolerance for activities that were not Islamic. A more honest attitude in academic sessions and during exams was found in students who tended to be involved in religious activities such as religious assembly activities, discussion forum, and praying in mosques (Khan et al., 2019). Therefore, students who had a high level of religiosity which was observed through worshiping or following religious activities diligently would discourage them from committing academic misconduct. Thus, this study was supported by Hongwei et al. (2017), Khan et al. (2019), Niazi et al. (2019), and Salsabilla (2020).
Furthermore, this study found that there was no effect of TPB (ATC, PBC, and SN) on CI by moderating IR. Incidents of cheating might be more spontaneous than planned. Thus, the formation of intentions was limited by the reactionary nature of the behavior. This statement was consistent with the partially mediated model that had been previously tested by Stone et al. (2010), in attitudes, social norms, and perceptions of behavioral control acted not only on behavior through intentions, but also on behavior that could directly explain the actions that lead to more spontaneous behavior. A person can commit misconduct because of the opportunities and benefits that came from other sources, that is what is called opportunity. The greater the perceived opportunity, the higher the possibility of students committing misconduct (Fitriana & Baridwan, 2012).

Conclusion
This finding implied that TPB as measured by ATC had a negative effect on CI, meaning that a good student attitude indicated a low desire to commit academic misconduct. Then, PBC had a positive effect on CI and CB, so that in terms of controlling student behavior, it did not indicate a decrease in the desire to commit academic misconduct. However, the SN had no effect on the CI of accounting students. On the other hand, the IR variable had a negative effect on CI. This meant that students who had a high level of religiosity or a good understanding of religion were proven to have less desire to commit academic misconduct. Moreover, from this study, it was found that there was no effect of TPB (ATC, PBC, and SN) on the CI variable, moderated by the IR. It meant that IR was not able to moderate TPB to reduce students' desire to commit academic misconduct.
The limitations in the research data and the subject of the research were only accounting students, so that they could reflect university students in general. Measurement of religiosity variable for effectiveness in reducing the temptation to commit fraud could be done to detect it in different places or objects. For example, if it is done in a Sharia-based company and compared to a non-Sharia one. Then the findings will be interesting and become a reinforcement of the values that must be instilled by management, in this case the company's human resources (HR) division for its employees.