Ethical Awareness, Ethical Judgment and Whistleblowing: A Moderated Mediation Analysis

This study aims to examine the ethical decision-making (EDM) model proposed by Schwartz (J Bus Ethics, doi: 10.1007/s10551-015-2886-8, 2016), where we consider the factors of non-rationality and aspects that affect ethical judgments of auditors to make the decision to blow the whistle. In this paper, we argue that the intention of whistleblowing depends on ethical awareness (EAW) and ethical judgment (EJW) as well as there is a mediation–moderation due to emotion (EMT) and perceived moral intensity (PMI) of auditors. Data were collected using an online survey with 162 external auditors who worked on audit firms in Indonesia as well as 173 internal auditors working in the manufacturing and financial services. The result of multigroup analysis shows that emotion (EMT) can mediate the relationship between EAW and EJW. The nature of this relationship is more complex and then tested by adding moderating variables using consistent partial least squares approach. We found that EMT and PMI can improve the relationship between ethical judgments and whistleblowing intentions. These findings indicate that internal auditors are more likely to blow the whistle than external auditors; and reporting wrongdoing internally and anonymously are the preferred way of professional accountants to blow the whistle in Indonesia.


Introduction
Whistleblowing has gained the attention of the global community and the media in recent years, partly because of large awards offered by the Dodd-Frank Act of 2010 and partly due to a case of fraud involving Olympus corporation and Michael Woodford who was fired when he revealed payment irregularities Rao et al. 2011;MacGregor and Stuebs 2014). This indicates that a whistleblower does not only arise from inside the organization, but it can also come from outside, and referred to as an external whistleblower (Maroun and Atkins 2014b;Maroun and Gowar 2013).
An internal whistleblower can observe the various violations that occur within an organization such as discrimination, corruption, cronyism or other unethical behavior. Meanwhile, an external whistleblower can observe noncompliance with the fulfillment of corporate social responsibility and the environment (Culiberg and Mihelic 2016;Vandekerckhove and Lewis 2012). Thus, the important role of whistleblowers in detecting wrongdoing cannot be denied (Latan et al. 2016). However, being a whistleblower is not easy, because one must consider the positive and negative impacts caused, and also involves the complicated process of ethical decision making (EDM) (Ponemon 1994;Shawver et al. 2015;Webber and Archambeault 2015;O'Sullivan and Ngau 2014). EDM can be understood as deciding or judging whether the action or decision is ethical . Given that the internal control system is designed to minimize risks such as financial fraud, it will rely heavily on moral reasoning which is conducted by auditors (both internal and external). However, an auditor is often faced with ethical issues that pit ethics professional codes against ethical decisions. 1 The critical reviews conducted by Culiberg and Mihelic (2016) and Vandekerckhove and Lewis (2012) showed that there is still an empirical gap in this area that requires further testing. For example, most previous studies have focused too much on internal whistleblowers (such as employees, managers, internal auditors and management accountants) and ignored external whistleblowers (Latan et al. 2016;Alleyne et al. 2016;Miceli et al. 2014). 2 In this context, subjects, such as how to protect external whistleblowers (Maroun and Gowar 2013) and how they are perceived, need to be further addressed (Maroun and Atkins 2014a, b). At the same time, the body of literature currently offers little insight into how a person reacts to wrongdoings to make a decision to blow the whistle. This relates to the ethical decision-making (EDM) model proposed by Rest (1986), where there are four stages that must be passed, namely awareness, judgment, intent and actual behavior.
As stated by Culiberg and Mihelic (2016), most of the existing empirical research related to whistleblowing has examined the relationship between judgment and intent (Zhang et al. 2009;Chiu 2003Chiu , 2002Liyanarachchi and Newdick 2009) and supported it fully. However, there are no studies that extend this testing to other stages, such as considering the influence of ethical awareness on ethical judgment. In addition, some studies also show that there are other factors that can affect this process, such as moral intensity (Jones 1991) and emotion (Henik 2008(Henik , 2015Hollings 2013). Schwartz (2016) showed an EDM model of integration, combining the factors of rationality and nonrationality. This model assumes that ethical behavior depends on people who face ethical biases (related to mood or moral intensity) and the environmental situation at the time. Jones (1991) defines moral intensity as a measure of moral imperatives-related problems in certain situations.
Perceived moral intensity will help auditors when facing an ethical dilemma, while emotions are feelings that arise (such as anger or fear) when encountering wrongdoing, and influence the auditor's decision to blow the whistle (Jones 1991;Henik 2008Henik , 2015Latan et al. 2016). Both of these factors play an important role and are a key element in the EDM model of whistleblowing. Therefore, the purpose of this study was to extend the EDM model of testing for whistleblowing by considering the role of two whistleblower groups (internal and external) in the Indonesian context.
Indonesia provides the proper setting to test this model because it offers an interesting phenomenon to study. For example, according to a report from global fraud study conducted by the Association of Certified Fraud Examiners (ACFE) in 2016, Southeast Asia was in first position for cases of fraud, and Indonesia is one of five countries in the world experiencing higher levels of fraud after South Africa, India, Nigeria and China. This is an indication that auditors in Indonesia (internal and external) may be still reluctant to become whistleblowers (Latan et al. 2016). As stated by Jubb (2000), internal or external auditors are often faced with an ethical dilemma when wanting to reveal wrongdoings in the workplace, and therefore, they have conflicts of loyalty and professionalism. Hence, the decision to blow the whistle is complicated. However, research in Southeast Asia and Indonesia is rare, and there is still an empirical gap (Culiberg and Mihelic 2016;Latan et al. 2016). Thus, it is important to examine what factors are instrumental to the auditor's decision to blow the whistle.
Our study contributes to the current literature in several ways. First, this is the first study to extend the testing EDM model to whistleblowing, where there are many factors and relationships between variables that have not been tested in previous research. 3 Thus, this study answers the call research of Culiberg and Mihelic (2016) to extend the testing of these models in the context of accounting and ethics. Although some previous studies have discussed this model (Zhang et al. 2009;Chiu 2003;Arnold et al. 2013;Yu 2015), they can be developed further. Second, this is the first study to compare two groups of whistleblowers-internal and external auditors-which is helpful in explaining which group is more prone to blowing the whistle. Until now, no previous empirical studies have fully considered testing the two whistleblower groups together in a single model. Although Shawver et al. (2015) used professional accountants as samples (including internal and external auditors) in testing the EDM model for whistleblowing, they did not test samples separately. 4 Third, this study extends state-of-the-art research on whistleblowing by providing evidence from Indonesia. Based on our best knowledge, no study conducted in Indonesia has tested EDM models of decisions to blow the whistle. As there are no empirical results available from Indonesia on whistleblowing in the context of accounting, this study provides initial evidence of the importance of individual and non-rationality factors in favor of EDM model proposed by Schwartz (2016) which have been the focus of research lately. Finally, it is important to conduct this study with experienced professionals such as auditors, who experience real-life ethical dilemmas that maybe different from those outside professional organizations (e.g., employees, consultants, customers, shareholders). However, few studies use the auditor as a sample (Curtis and Taylor 2009;Latan et al. 2016;Culiberg and Mihelic 2016;Alleyne et al. 2013).
The remainder of the paper is organized as follows. The next section presents the development of the hypotheses, followed by the research methodology. Next, we discuss our results. Finally, we further analyze our results and provide important implications of our study as well as its limitations.

Literature Review and Hypothesis Development
The Ethical Decision-Making Model (EDM) EDM is one of the issues that have attracted the attention of researchers in the field of business ethics, but also in other disciplines such as marketing, moral psychology, organizational behavior, philosophy and social economics. The extent of illegal and unethical behavior that occurs in organizations and society in general has motivated researchers to develop a EDM model on an ongoing basis. The main assumption among all bodies of knowledge in the literature on EDM is a rationality-based process. One of the most widely cited and tested EDM models was proposed by Rest (1986) which consists of four components, namely awareness, judgment, intent and actual behavior. Until now, there have been several theoretical models of EDM that have been proposed including a model of the contingency by Ferrell and Gresham (1985), a situational interactionist model by Trevino (1986), the general theory of ethics and its modifications Vitell 1986, 2006), modified Rest model by Jones (1991) and integrated EDM model by Schwartz (2016). The main purpose of building these models is to explain and predict the process, whereby a person makes ethical decisions and the factors underlying such decisions. Ferrell and Gresham (1985) adopted a framework for contingency aiming to explain the processes of EDM that influence ethical decisions of marketers. In this model, they propose three contingency factors: individual factors (e.g., knowledge, values, attitudes and intentions), organizational factors (e.g., organizational pressures and opportunities) and environmental factors (e.g., company policies and interactions between groups) that directly affect the ethical decisions of individuals. Trevino (1986) developed a situational interactionist model by combining individual factors (such as moral development) with the situational factors to explain and predict the EDM of individuals within an organization. More specifically, the model shows that the relationship between the individual cognitive moral development and ethical behavior will be moderated by the two factors. Individual factors include the strength of the ego, field dependence and locus of control, whereas situational factors include the immediate context of work, organizational culture and nature of work. 5 In addition, Trevino (1986) also adopted the six stages of cognitive moral development developed by Kohlberg which becomes operative in the EDM process. Hunt and Vitell (1986) proposed a general theory of ethics that is more comprehensive in explaining the process of EDM and widely accepted in the field of marketing. According to their theory, once a person is faced with an ethical dilemma, where there are alternatives and consequences (such as the influence of cultural, environmental professional, organizational, industrial and personal characteristics), they will make an evaluation (both deontological and teleological), before making ethical judgments. After that, the ethical judgment will directly affect the ethical intentions which in turn affect the actual behavior (Hunt and Vitell 2006). The Hunt-Vitell model also added feedback generated from the actual consequences of people behavior to make personal experience in the future.
Unlike the previous three competing models, Jones (1991) built an EDM model above Rest (1986) model. According to Jones (1991), the literature does not have a model which shows the characteristics of a moral problem itself which affects the EDM process, and he proposes an issues-contingent model of EDM. This combines the concept of moral intensity and organizational factors in the Rest's model, which is a new paradigm in EDM models. In addition, it considers that individuals who have a superior position in the organization, as a routine, more often faced ethical issues in decision making, and vice versa. Thus, the stronger the intensity of the ethical issues, the more likely the decision makers are to lean toward ethical behavior. Therefore, Jones (1991) makes the proposition that moral intensity and organizational factors play a role as predictor variables that directly and separately contribute to the EDM process.
Most recently, Schwartz (2016) conducted a synthesis of all existing EDM models and previous studies that have been conducted, to propose a new model called the ''integrated EDM model.'' This combines all theoretical and empirical models into a single comprehensive model. This study adopts the perspective of the framework proposed by Schwartz (2016), where we consider factors of non-rationality (such as emotions) as well as individual factors (such as moral intensity) as the mediation-moderation effects in the relationship between the variables that affect the decision-making process of auditor to blow the whistle (see Fig. 1). As stated by Schwartz (2016), EDM is a complex process that involves many variables that are interrelated (neurocognitive-affective processes) and influence each other. For example, in the EDM model described earlier, non-rationality factors were not fully discussed, and for this reason the rationalist approach seems to have limitations and shortcomings, especially in conditions that are unpredictable and dynamic. We chose the non-rationality factors to be tested for the reason that they are more dominant in the process of moral judgment, in which rationality plays a secondary role after ''a fact'' is clear. In other words, when someone finds wrongdoing, but it is outside of the organization's ethical code of conduct, for example, the non-rationality factor will dominate the EDM process. Conversely, when the wrongdoing is common and has been agreed upon, then the rationality factor will dominate. Given that the rights and duties of the auditor as a whistleblower have not been set out clearly with the law on protection, then the non-rationality of factors tends to be more important in the EDM process to blow the whistle. 6 The EDM model proposed by Schwartz (2016) also built on the model of Rest, but with additional modification factors of rational and non-rational as intermediaries as well as individual and situational factors as moderating variables. This model has not been widely tested in comparison with previous models, especially in decision making for whistleblowing. Butterfield et al. (2000) define ethical awareness as consciousness owned by an individual at a certain time point when faced with ethical dilemmas that require a decision or action that may affect the interests of themselves or others in a way that may conflict with one or more of moral standards. Classical theory of EDM found ethical awareness is a strong predictor of ethical judgment (Rest 1986;Jones 1991) and mediated by non-rationality factors (affective) such as emotions Henik 2008;Schwartz 2016). As proposed by Henik (2008) and developed further by Schwartz (2016), emotions (such as fear or anger) are also able to mediate the relationship between ethical awareness and ethical judgment for whistleblowing. Emotions can form prosocial or antisocial behavior which can affect a person's decision to reveal any wrongdoing. Previous research has found a significant relationship between ethical awareness and ethical judgment on marketing services (Singhapakdi et al. 1996), upper-division business students (Haines et al. 2008) and formal infrastructure (Rottig et al. 2011) and mediated by emotion (Connelly et al. 2004;Singh et al. 2016;Henik 2015). From the above discussion, the following hypothesis can be derived:

Ethical Awareness, Emotions and Ethical Judgment
H1a Ethical awareness has a positive direct effect on ethical judgment.
H1b Ethical awareness has a positive indirect effect on ethical judgment through emotions.

Moderating Effect of Perceived Moral Intensity on Ethical Awareness and Ethical Judgment
Jones (1991) defines moral intensity as a measure of moral imperative-related problems in certain situations. According to Jones (1991), EDM models should place emphasis on the characteristics of ethical issues themselves. Based on the issues-contingency perspective, Jones placed moral intensity as a predictor variable that affects every phase of the EDM process. Many previous studies have examined this variable in the context of business ethics Craft 2013;O'Fallon and Butterfield 2013) and provide result that can be developed further. We adopt this perspective that assumes individuals more easily identify ethical issues when they have high moral intensity. Moral intensity consists of six components (see Jones 1991); however, according to Curtis and Taylor (2009) only three factors are relevant in the context of the audit, which include the magnitude of consequences, probability of effect and proximity, and these three factors can affect the auditor's ethical judgment to blow the whistle (p. 198). 7 Magnitude of consequences is how much loss will result from the wrongdoings and affect the ethical judgment of the auditor. Probability of effect is the impact of that loss in the future (such as retaliation or job loss), and also how it will influence the ethical judgment of the auditor and the intention to blow the whistle. Finally, proximity is a direct influence caused by unethical behavior which harms one of the group members (such as co-workers or family members) and how it affects the ethical judgment of auditors to blow the whistle. In other words, if the impact of the one act does not directly affect the lives of people nearby, the auditor may be reluctant to disclose the error. Previous research has found a significant relationship between moral intensity and ethical judgments (Singer et al. 1998;Valentine and Hollingworth 2012;Yu 2015;McMahon and Harvey 2007;Leitsch 2004). Other studies of Beu et al. (2003) and Singh et al. (2016) showed that moral intensity moderates the relationship between several independent variables to ethical judgments. From the above discussion, the following hypothesis can be derived: H2 Moral intensity moderates the relationship between ethical awareness and ethical judgment.

Moderating Effect of Emotions on Ethical Judgment and Whistleblowing Intentions
By recognizing that decisions can be divided into (a) rationalist based (i.e., reason) and (b) non-rationalist based (i.e., intuition and emotion) (Schwartz 2016), several previous studies have realized the importance of the role of emotions in influencing ethical decisions (Connelly et al. 2004;Curtis 2006). Emotions are feelings that arise (such as anger or fear) when encountering wrongdoing, and also influence the auditors' ethical judgment to arrive at the decision to blow the whistle (Henik 2008). Emotions can directly affect the ethical judgment and moral reasoning (Singh et al. 2016). For example, negative mood can be associated with lower intentions to report the unethical actions of others to a superior within the organization (Curtis, 2006). According to Schwartz (2016), emotions can also serve as a moderating variable on the relationship between ethical judgments and whistleblowing intentions. When the auditor is making ethical judgments on specific cases, for example, feelings like anger or fear will continue to be part of a subsequent decision, whether to reveal wrongdoing through internal routes (IWB), external (EWB) or anonymous (AWB) whistleblowing. If the auditor is quite afraid of revealing errors found, because it will affect personal and professional dimensions in the future, then the internal and anonymous route of whistleblowing is usually selected. Conversely, when the auditors ignore the risks, because wrongdoing affects the lives of many people (for example, Edward Snowden who leaked secret documents from the NSA), they will probably choose the route of external whistleblowing. Previous research has found a significant relationship between emotion and ethical judgments (Connelly et al. 2004;Curtis 2006) and the role of emotions as a moderator in the relationship between ethical judgments and whistleblowing intentions (Hollings 2013;Henik 2015;Schwartz 2016). From the above discussion, the following hypothesis can be derived: H3a Emotions moderates the relationship between ethical judgment and IWB.
H3b Emotions moderates the relationship between ethical judgment and EWB.
H3c Emotions moderates the relationship between ethical judgment and AWB.

Moderating Effect of Perceived Moral Intensity on Ethical Judgment and Whistleblowing Intentions
Recent research shows that high moral intensity can affect ethical judgments of auditors (Yu 2015) and will have a positive impact on the intention to blow the whistle (Alleyne et al. 2013). The model proposed by Jones (1991) placed moral intensity as a predictor variable in influencing every stage of the EDM process. We revise the role of the moral intensity variable by placing it as a moderating variable in line with the integrated EDM model proposed by Schwartz (2016). Ethical judgments made by individuals will be better when matched with high moral intensity and interaction, which in turn have a positive influence on the intention to report wrongdoings. In other words, the higher the perceived moral intensity of an issue, the more likely the person is to make ethical decisions, which in turn affects the intention to blow the whistle. Previous research has shown that ethical judgment has a positive influence on whistleblowing intentions (Zhang et al. 2009;Chiu 2003) and is moderated by moral intensity (Alleyne et al. 2013;Latan et al. 2016). From the above discussion, the following hypothesis can be derived: H4a Ethical judgment has a positive direct effect on IWB.
H4b Ethical judgment has a positive direct effect on EWB.
H4c Ethical judgment has a positive direct effect on AWB.
H5a Moral intensity moderates the relationship between EJW and IWB.
H5b Moral intensity moderates the relationship between EJW and EWB.
H5c Moral intensity moderates the relationship between EJW and AWB.

Research Method Sample Selection and Data Collection
The respondents in our survey are professional accountants working for audit, manufacturing and financial services companies listed on the Indonesia Stock Exchange (BEI). We chose companies in manufacturing and financial services, as reported by ACFE 2016, because these sectors have the most cases of wrongdoing in Southeast Asia. We also ensure that external auditors who audited the companies were used as a sample and matched with the internal auditor of the companies. The data collection was done using an online platform which places the questionnaire used to collect data for this study. A web link to the questionnaire was later on emails sent to the firms. Email addresses from the audit firms were obtained from the directory of the Indonesian Institute of Certified Public Accountants (IAPI) for 2015. Email addresses of manufacturing and financial services companies were extracted from each company's website. Based on the directory and the information available, approximately 74 audit firms were contacted with 400 total respondents from external auditors. Furthermore, 223 manufacturing and financial services companies were contacted. These companies had, in total, 560 internal auditors. After sending a request to participate in the survey, we sent three subsequent emails as a reminder. To ensure data quality control, we checked the collected data, to verify whether there was missing data, straight line responses or similarity of answers. We found a few problematic cases that were removed from the data before further analysis. Finally, we made additional efforts to increase the response rate, by directly calling the target respondents. To convince the respondents, we conceal their identity (such as name and address of the company) and they remain anonymous. Furthermore, we determine the cutoff time for the return of the questionnaire, which is 3 months, for the purpose of testing nonresponse bias, as suggested by Dillman et al. (2014). Between July and October 2016, we obtained 179 questionnaire responses from external auditors and 194 questionnaires from internal auditors, of which 38 were incomplete, so the number of questionnaires that were valid and could be used in this study was 335 with a 34.89% response rate. Of the total questionnaires collected, 48.35% came from audit firms and the rest, respectively, 36.09% and 15.56%, came from manufacturing and financial services (see Table 1).
Results of the t test showed that there was no difference in statistical significance of responses (p \ 0.05) between public accountants who came from the Big 4 and non-Big 4 and also for the social desirability response bias problems (Randall and Fernandes 2013). This indicates that the size of the audit firm will not affect the results of analysis and there are no problems in social desirability response bias of the respondent's own reporting of whistleblowing intentions. 8 These results also indicate that there is no problem of selection bias that causes the auditor not to take part in the survey (Randall and Fernandes 2013). In addition, the statistical test results also showed that there was no significant difference between respondents who answered in the beginning of data collection, compared with respondents who answered at the end, which means there is no problem of non-response bias that occurs systematically (Dillman et al. 2014). To ensure there is no common method bias (Podsakoff et al. 2012), we use the full collinearity approach by (Kock 2015). The AVIF value obtained from analysis is less than 3.3, thus indicating that no common method bias problem in this study. Table 1 presents the profile of respondents in this study. The 335 completed questionnaires were divided into two subsamples: 162 external auditors and 173 internal auditors, 63.28% were male (while 36.72% are female), with an average age of 37.2 years. In terms of positions, 42.7% of the sample comprised senior audit staff and 57.3% comprised junior audit staff. As for qualifications, 61.8% held a bachelor's degree and 38.2% held a master's degree and doctorate, while 87.2% of the sample had professional qualifications, with 43.9% of the sample having completed a professional qualification CPA and 43.3% having completed the Qualified Internal Auditor (QIA) and Certified Internal Auditor (CIA) examinations.

The Survey Structure
The survey used to measure each of the variables in this study consists of three parts. The first section described the purpose and objectives of this research, by asking the respondent's willingness to participate in the survey. The second section asked for the respondents' demographic information such as gender, age, education level, occupation and qualifications. The third section presented scenarios and questions related to the variables to be studied. Given the difficulty in gaining access to the object in order to observe real unethical behavior, a scenario approach is commonly used in research in the field of accounting and ethics (for example, Alleyne et al. 2016;Arnold et al. 2013;Chan and Leung 2006;Curtis and Taylor 2009;Shawver et al. 2015). This approach illustrates a specific case, and the respondents are asked to respond and put themselves as an actor in such situations. The scenario used in this study was adopted from the scenario used by Bagdasarov et al. (2016), Clements and Shawver (2011), Curtis and Taylor (2009), Kaplan and Whitecotton (2001) and Schultz et al. (1993) with modifications, which highlights the numerous violations of professional ethics and wrongdoings in a company. 9 To create a scale able to measure the intentions to blow the whistle, we used a total of ten items of questions based on the internal, external and anonymous reporting routes adopted by Park et al. (2008). The survey respondents were asked about reporting routes that they use to select when they find wrongdoings that occur (hypothetical scenario). The variable ethical awareness was measured by three questions adopted from Arnold et al. (2013). Respondents were asked about whether an action in the case scenario is ethical or unethical behavior. The variable ethical 8 Social desirability response bias is broadly understood as the tendency of individuals to deny socially undesirable traits and behaviors and to admit to socially desirable ones. 9 The use of scenarios is more effective to give stimuli to the auditor in making ethical decisions when faced with certain situations.
Ethical Awareness, Ethical Judgment and Whistleblowing: A Moderated Mediation Analysis 295 judgment for whistleblowing was measured through four items inspired by Reidenbach and Robin (2013). Respondents were asked about whether an action in the scenario is moral or not morally right, just or unjust, acceptable or unacceptable and so on. Tables 2 and 3 show indicators and outcome measurement models for variables of ethical awareness, ethical judgment and intentions of whistleblowing. Furthermore, the moral intensity variable is measured by six questions adopted from Clements and Shawver (2011).
Respondents were asked to provide feedback on the scenarios to assess the intensity level of their morals. Finally, emotional variables measured four items of questions adopted from Connelly et al. (2004). Respondents were asked to provide feedback on the scenarios to assess the level of their emotions. The value of the loading factor, average variance extracted (AVE) and reliability derived from the analysis of the measurement model for all variables are: loading factor [0.60, composite reliability/ rho_A [ 0.70 and AVE [ 0.50, so it meets the recommended requirements (Hair et al. 2017;Henseler et al. 2018). However, there are some indicators of measurement models that were retained, with the value of the loading factor being [0.5. As stated by Hair et al. (2017, p. 114), the value of the loading factor shows the explained variance in a construct. So, if the value AVE is already more than 0.5, the indicator with low loading values can be kept to maintain the content validity. Table 4 shows the indicators and outcome measurement model for moral intensity and emotional variables.
In addition, we tested the discriminant validity or divergent validity for all latent variables in the model using  the heterotrait-monotrait (HTMT) ratio. As stated by Henseler et al. (2015), HTMT is a new procedure to test the discriminant validity and is more appropriate than the Fornell-Larcker criterion. The HTMT approach has reliable performance and overcomes bias in the estimation of parameters of the structural model. In Table 5, it is shown that the value of HTMT was smaller than 0.90, which means that it meets the recommended rule of thumb (Hair et al. 2017;Henseler et al. 2015).

Data Analysis
Before we analyze the overall model, we ensured that the adequacy of the sample size for estimation of the model has been fulfilled. Because the data analysis in this study uses the consistent partial least squares (PLSc) approach, then a sample needs to have at least 100 cases (Latan and Ghozali 2015). The main purpose of PLSc is to mimic the covariance-based SEM approach to test or confirm the theory (Dijkstra and Henseler 2015). By using PLSc, the estimator of the model will be consistent for the loading and the correlation between latent variables and allows us to access the goodness of fit (Dijkstra 2014). We chose PLSc with the consideration that it is more appropriate to test complex models, where the CB-SEM approach would be difficult to apply (Richter et al. 2016;Rigdon 2016). Previous research in this area already uses PLS-SEM as an analytical tool (Buchan 2005;Haines et al. 2008). In contrast to other SEM techniques, PLS-SEM does not rely on the assumption of normality (distribution free) because it is nonparametric. However, some assumptions such as multicollinearity and goodness of fit for the local models assessment need to be considered. Overall, the data analysis in this study will go through three stages. First, we analyze the measurement model to ensure an indicator constructs are valid and reliable using the full sample. Second, we examined multigroup analysis to compare the two subsamples for each path coefficient. Third, we examine the effect of mediation-moderation to determine the role of moral intensity and emotional variables.

Results
In this study, data analysis and hypotheses testing were conducted by using variance-based SEM. One of the techniques available today is PLS-SEM, which is the most fully developed and has become a vital tool for researchers to examine various issues of social science. PLS-SEM was developed with the main purpose of prediction and then extended to test the theory with consistent results for the factor models. We chose to use PLSc (on selection algorithms and bootstrapping) considering that it will provide similar results to CB-SEM. 10 We use the SmartPLS 3 program  to analyze these models by using PLSc. PLS-SEM analysis requires through two stages, namely the measurement model and the structural model. Assessment of the measurement model is intended to test the validity (convergent and discriminant) and reliability of each indicator forming latent constructs. After we make sure that all the indicators constructs are valid and reliable (see Fig. 2), we continue the analysis to the second stage of assessing the quality of the structural model and run multigroup analysis to test the hypothesis. The results of the quality assessment for the structural model are given in Table 6.
In Table 6 it is shown that the whistleblowing intention (IWB, AWB and EWB) can be explained by the predictor variables of 0.425-0.507. This value indicates that the ability of the predictor variables to explain the outcome variables was approaching substantial (Latan and Ghozali 2015). The resulting effect size value of each predictor variable in the model ranged from 0.01 to 0.520, which is included in the category of small to large. The value of variance inflation factor (VIF) generated for all the independent variables in the model is \3.3, which means that there was no collinearity problem between the predictor variables. The Q 2 predictive relevance value generated excellent endogenous variables, i.e., [0, which means that the model has predictive relevance. The value of goodness of fit is generated through the standardized root-meansquared residual (SRMR) that is equal to 0.049 \ 0.080 and the normed fix index (NFI) 0.837 [ 0.80, which means that our model fits the empirical data. We run multigroup analysis to compare the two subsamples of internal whistleblower (internal auditor) and external whistleblower (external auditor) for each path coefficients using the PLS-MGA approach. The purpose of the analysis of PLS-MGA was to compare two groups of samples to determine statistically significant differences in group-specific parameter estimates (Matthews 2018;Sarstedt et al. 2011) and in this case which group is more prone or unlikely to blow the whistle. Before running the PLS-MGA, we consider it to test the measurement invariance of composite models (MICOM) using a permutation procedure. 11 We test measurement invariance to ensure that the specific-group difference of the estimation model does not affect the results for latent variables in the whole group (Henseler et al. 2016;Solovida and Latan 2017). The analysis showed that there was no significant difference between variance and average values for both groups (see Table 7), which means no invariance problem that will affect the outcome.
Based on the analysis in Table 7, it can be seen that the ethical awareness (EAW) has no effect on ethical judgment (EJW) for both internal and external group auditors. From the analysis results obtained value of coefficient (b) to the relationship EAW ? EJW each for both groups was 0.070 and -0.057 with 95% bias-corrected and accelerated (BCa) [0.05. This means that the hypothesis 1a (H1a) was rejected. These results support previous studies (Chan and Leung 2006;Valentine and Fleischman 2004). EAW cannot be a direct predictor of the EJW, and this is consistent with the integrated EDM model by Schwartz (2016), where there is another factor that mediates both. EAW of professional accountants in this study also found variance in their ability to respond to a case scenario. Furthermore, the value of the coefficient (b) to the relationship EAW ? EMT is 0.670; 0.553 and EMT ? EJW is 0.482; 0.386 with 95% bias-corrected and accelerated (BCa) \ 0.01, respectively. This means that the hypothesis 1b (H1b) is supported. We also tested the indirect effect by using the method proposed by Cepeda et al. (2018) and obtained the same results. 12 These results support previous studies (Henik 2015;Connelly et al. 2004;Singh et al. 2016;Curtis 2006). This suggests that emotions may serve as indirect-only mediation or full mediation of the relationship between EAW and EJW. When someone finds wrongdoing, it will affect their emotions prior to making ethical judgments. From these findings, it can be concluded that the internal auditors have more intense EAW, EMT and EJW than the external auditors. Finally, from Table 7 it is shown that the value of the coefficient (b) to the relationship EJW ? IWB is 0.446; 0.317, EJW ? AWB is 0.400; 0.419 and EJW ? EWB is 0.315; 0.309 for each group of samples with 95% biascorrected and accelerated (BCa) \ 0.01, respectively. This means that the hypothesis 4 (H4a, H4b and H4c) is supported. These results support previous studies (Zhang et al. 2009;Chiu 2003;Arnold et al. 2013;Buchan 2005). As stated by Culiberg and Mihelic (2016), most of the research in this area has provided conclusive results for the relationship between EJW and whistleblowing intentions. A professional accountant who has made ethical judgments can report wrongdoing found through one of these three route options available: internal, external or anonymous. The results showed that the internal route is the most preferred by the internal auditor followed by an anonymous and external route. In contrast, for external auditors, the anonymous route is the most preferred, followed by internal and external. This indicates that professional accountants of both groups in the cases of Indonesia chose an external route to blow the whistle as the last option. They are more likely to disclose an error discovered through internal and anonymous routes. One reason that might affect their decisions is fear of retaliation and the various risks that arise when using an external route for whistleblowing.
These findings indicate that internal auditors have a higher (more likely) intention to report any act than external auditors; and blowing the whistle internally and anonymously can be more useful for professional accountants. Findings are aligned with the general statement that employees are not the only ones with privileged information about a company, and consequently outsiders may observe various wrongdoings (Culiberg and Mihelic 2016). However, the present study adds a more detailed suggestion that internal auditors more likely to report than external auditors. Although the literature has suggested that there is not, a priori, a profile of whistleblowers that organizations (Henik, 2015), our findings suggest that internal auditors are more likely to blow the whistle than external ones. While the literature recognizes that there are challenges in fully protecting external whistleblowers (Maroun and Gowar 2013), our findings suggest that discussing how to fully protect internal auditors should also be a priority.
However, as discussed by Atkins (2014a, 2014b) there is an upward trend of increasing the availability of information to stakeholders and enhancing the level of expectation that the public have on auditors, in terms of transparency and accountability and in terms of relevance of audit reports Atkins 2014a, 2014b). If this was reinforced in Indonesia, our results would be different. This scenario will need to further consider the challenges in fully protecting external whistleblowers (Maroun and Gowar 2013).

Importance-Performance Map Analysis (IPMA)
We tested importance-performance map analysis (IPMA). Ringle and Sarstedt (2016) stated that the IPMA gives researchers the opportunity to enrich their PLS-SEM analysis and, thereby, gain additional results and findings. Nevertheless, PLS-SEM has several key advantages over traditional IPMA that typically relies on multiple regression analysis. First of all, in determining the importance scores, PLS-SEM is a valuable analytical tool as it is capable of integrally assessing a complex network of relationships connecting drivers to a target construct of interest. Second, it can incorporate latent constructs (see Streukens et al. 2018). The IPMA analysis results are shown in Table 8.
From the above analysis (Table 8), it can be seen that the EMT has a relatively low performance that is equal to 64.45. If matched by other constructs, EMT's performance is slightly below average. On the other hand, with a total effect of 0.595, this construct's importance is enough high. Therefore, a one-unit increase in EMT's performance from 64.45 to 65.45 would increase the performance of IWB, AWB and EWB is 0.595, 0.590 and 0460 points, respectively. Therefore, the companies aim to improve the IWB, AWB and EWB of internal and external auditors, their first priority must be to improve the performance aspects of EMT. Furthermore, aspects related to EJW, PMI and EAW follow as a second, third and fourth priorities.

Interaction Effect Analysis
We tested the interactions using the orthogonalizing approach. This approach was chosen because it produces an accurate estimate, has a high predictive accuracy and is able to minimize the collinearity problem. The results of the analysis of interactions are shown in Table 9. In Table 9 it is shown that H3a, H3b, H3c and H5a, H5B, H5c are fully supported where moral intensity and emotional may moderate the relationship between EJW and whistleblowing intentions. As for the relationship EAW 9 PMI ? EJW obtained insignificant results with coefficient (b) = 0.031 and 95% bias-corrected and accelerated (BCa) = 0.140 [ 0.05. This suggests that emotions or feelings of auditors themselves play an important role in improving the ethical assessment of auditors with the consequence that they have a higher whistleblowing intention to report any wrongdoing that occurs, reinforcing the discussion on non-rationalist-based decision making (Schwartz 2016). This finding can be understood by taking into account a broader discussion on how mood and emotions can influence whistleblowing (Curtis, 2006).
While the moral intensity that comes from the experience of auditors would assist in considering any magnitude of consequences, the possibility of future losses and the proximity to the organization influence actions to blow the whistle. Emotions felt would assist the auditor in considering the various risks arising from actions taken.
From the results of this analysis, we reached the same conclusion that the internal and anonymous route is a favorite choice for professional accountants in Indonesia to reporting wrongdoing. These results support previous studies (Hollings 2013;Henik 2015;Alleyne et al. 2016;Latan et al. 2016). Given the cultural and social norms' strength in Indonesia, the freedom to act and speak out becomes a supporting factor for professional accountants in improving the intention to report wrongdoing without fear of reprisal. Nevertheless, it is important to further develop institutional mechanisms capable of fully protecting whistleblowing (Maroun and Gowar 2013).

Conclusion
This study aims to examine the integrated EDM model proposed by Schwartz (2016), where we consider the factors of individual non-rationality that affect ethical judgments of the auditor to arrive at the decision to blow the whistle. We answered the call of Culiberg and Mihelic (2016) to extend the testing of EDM models in the whistleblowing context. In this paper, we argue that the intention of whistleblowing depends on EAW and EJW as well as on emotion and perceived moral intensity.
We support the hypothesis that EAW cannot directly affect the EJW, but must go through the non-rationality of factors such as emotion. We also found that internal and anonymous whistleblowing routes were used by professional accountants in the case of Indonesia. In terms of practical implications, these findings provide a deep understanding of how audit firms, manufacturing and financial services should be selective in choosing audit staff who uphold professional and ethical standards of behavior. In addition, companies need to make strong efforts to implement a comprehensive ethics program including training in ethics, codes of conduct, which provide guidance to staff auditors to resolve ethical conflicts and increase professional responsibility to report wrongdoing. Companies also need to apply the right strategy to enhance the auditor's whistleblowing intentions and reduce the fear of retaliation, for example, by providing a whistleblowing hotline or reporting of anonymity, which was a favorite choice for the Indonesian context.

Limitation and Future Research
There are several limitations to this study which need the attention of the reader. First, this study did not consider cultural factors that may affect the EDM process. Some cultural factors such as nationality, patriotism, religion and political system may affect the EAW and EJW of auditors. These findings may differ in other countries. Second, this study only considers the factors of non-rationality in the integrated EDM model proposed by Schwartz (2016), without examining the factors of rationality. Different results may be obtained when considering both. Third, this study only used two variables as mediation-moderation in the model. Lehnert et al. (2015) showed that there are still many relevant variables (moderation and mediation) more important to be considered and tested in the EDM model. Fourth, this study did not consider the effect of extraneous variables (such as age, gender, education or total tenure) and also unobserved heterogeneity that might interfere with the results. However, several previous studies showed inconsistency in the role of extraneous variables in the EDM model (Chan and Leung 2006;Cagle and Baucus 2006;Ebrahimi et al. 2005;Shafer et al. 2001;Marques and Azevedo-Pereira 2009). In addition, the selection bias could have been handled more carefully. Finally, this study only tested the whistleblowing intentions without testing actual behavior. Further research can follow up the testing of integrated EDM model by Schwartz (2015) for whistleblowing by considering factors of rationality and non-rationality as intuition, reason and confirmation. Cultural factors also need to be considered for further study. This is a call research to provide empirical evidence of the model. Furthermore, future research may use other moderating variables such as intrinsic religiosity, personal spirituality, moral obligation, retaliation, intelligence and others who have an important role in the EDM process (Liyanarachchi and Newdick 2009;Haines et al. 2008;Bloodgood et al. 2008). Replication studies on the other subject group (for example, consumers vs shareholder) and other organizations (e.g., government and public administration) will also allow access to generalize the findings of this study. Overall, the researchers feel that it is necessary to replicate this study by using qualitative approaches such as case studies or fuzzy-set qualitative comparative analysis (Ragin 2008), taking into account unobserved heterogeneity testing (Hair et al. 2012;Schlittgen et al. 2016), which might be fruitful for new avenues for future study, 13 as there are not many studies have used a qualitative approach to test the EDM model for whistleblowing.
Funding The author(s) received no financial support for the research, authorship and/or publication of this article.

Compliance with Ethical Standards
Conflict of interest The authors declare that they have no conflict of interest.
Ethical Approval This article does not contain any studies with human participants or animals performed by any of the authors.
Informed Consent We aware of the contents and consent to the use of our names as an author of manuscript entitled ''Ethical Awareness, Ethical Judgment and Whistleblowing: A Moderated Mediation Analysis.''