THE EFFECT OF AUDITOR COMPETENCE, PROFESSIONAL SKEPTICISM, RED FLAG, AND INTERNAL CONTROL SYSTEM ON FRAUD DETECTION

Today more and more cases of fraud occur. In each of these cases, there is always an auditor's failure to detect fraud. It makes the fraud going on for a long time before being discovered. Based on this, we intend to research the factors influencing fraud detection. This study aims to determine the effect of auditor competence, professional skepticism, red flags, and internal control systems on fraud detection. The data collection technique in this study used primary data by distributing questionnaires to auditors working at the Public Accounting Firm (KAP) in the DKI Jakarta area using a purposive sampling method. The data analysis method uses a partial least square model using the SmartPLS 3.3 software. The results of this study indicate that auditor competence, red flag, and internal control system variables significantly affect fraud detection. In contrast, the professional skepticism variable has no significant effect on fraud detection.


INTRODUCTION
Quoted from the International Standard of Auditing (ISA, 2019: 240), fraud is an act that one or more individuals intentionally carry out among management, employees, responsible for governance, or third parties that involves fraud to obtain unfair or illegal benefits.Mangala and Kumari (2017) state that fraud is a serious threat to businesses and stakeholders, which is marked by an increase in economic crime and regulatory pitfalls.Therefore, further actions and efforts are needed to prevent fraud.
The Associate of Certified Fraud Examiners (ACFE) reported that in 2020, there were 2,504 cases of fraud from 125 countries around the world, with a total loss of $3.6 billion.The estimated loss for each organization due to fraud is 5% of revenue.ACFE also stated that corruption is the most common fraud scheme in every country.Misappropriation of assets is the most frequent fraud case with the smallest loss, while financial statement fraud is the least common fraud case with the greatest loss.Based on AppsFlyer's 2019 report entitled "Scams on the rise: How bots and malware harm APAC Apps", it is known that the fraud rate in Indonesia is the second highest in Southeast Asia after Vietnam at 58.2% (liputan6.com,accessed on 16 September 2021).Therefore, it can be said that fraud in Indonesia is still common.
It can be seen in Table 1.1,The most common and most detrimental fraud cases in Indonesia based on a survey conducted by the Association of Certified Fraud Examiners (ACFE) in 2019 sequentially, namely 167 cases of corruption (64.4%) with a total loss of Rp. 26 Billion.
Table 1 shows that the percentage of institutions that are most often harmed by fraud is the government at 48.5%, then BUMN at 31.8%, and private companies at 15.1%.

Table 1 Fraud Case Occurs in Indonesia
Source: Author Failure to detect fraud can occur due to several factors.These factors come from the internal side of the internal auditor and the external side.On the internal side, namely independence, skepticism, auditor experience, and the auditor's ability to capture signals conveyed by red flags.At the same time, the external side can be derived from the characteristics of fraud that are increasingly complex and well-organized by fraud perpetrators, namely the level of internal control.Sometimes the closed attitude of management makes it difficult for auditors to detect fraud (Indrasti & Karlina, 2019).
The first general standard (SA section 210 in SPAP, 2001) states that the audit must be carried out by someone who has sufficient technical expertise and training as an auditor.Meanwhile, the third general standard (SA section 230 in SPAP, 2001) states that in carrying out the audit in preparing the report, the auditor must use his professional skills carefully and thoroughly.Therefore, every auditor is required to have professional skills and expertise in carrying out their duties as auditors so that the audit process produces an accountable opinion.
The Public Accounting Professional Standards (SPAP) defines skepticism as an auditor's attitude that includes a mind that always questions and evaluates audit evidence critically, as well as being professional, careful, and thorough in carrying out auditing and preparing reports (SA section 230 in SPAP, 2011).Research conducted by the SEC (Securities and Exchange Commission) found that one of the causes of audit failure was an inadequate level of professional skepticism; where of the 40 audit cases studied, 24 cases (60%) of which occurred because the auditors did not apply an adequate level of professional skepticism (Waluyo, 2008).Professional skepticism is needed to improve audit quality because by being skeptical, the auditor will take the initiative to seek further information from management regarding accounting decisions taken and assess their own performance in exploring audit evidence that supports the decisions taken by the management.(Financial Reporting Council, 2010).
In carrying out their duties as auditors, it is important for auditors to pay attention to aspects that encourage them to be effective in the audit process.One of the most important is the red flag or commonly referred to as a signal that indicates fraud, which of course, must be detected by an auditor.Horne, Venter, and Lochner (2018) argue that it is important to identify red flags because this will be very helpful in detecting and preventing fraud, including in conducting the fraud investigation itself.Prasetyo (2015) states that the higher the level of red flags found, the easier it will be for auditors to detect fraud.Auditors can see signs of red flags in the corporate environment, such as management suddenly changing their lifestyle in a short period of time.
What is no less important that makes it easier for auditors to detect fraud is understanding the internal control system or internal control system.Internal control, according to Kulikova and Satdarova (2016) is a process that aims to increase the effectiveness of the company's operational activities, increase the reliability of financial reports, and realize the consistency of company activities with applicable policies and regulations.The implementation of internal control, it involves the board of directors, management, and all employees or staff.Abiola and Oyewole (2013) state that internal control is one of the most effective systems for detecting fraud because internal control operates in the same environment as fraud itself and serves as an effective and formidable enemy for fraud schemes.According to Abiola and Oyewole (2013), the internal control system is defined as a process, framework, or function, without touching the systematic concept.COSO (2011) states that the key to practices in detecting and preventing fraud in an entity, whether private or public, lies in the control measures implemented.

Effect of Competence of Auditor on Fraud Detection
According to attribution theory, there are two causes of behavioral attribution.First, it is internal or disposable.Second, external or environmental attribution.Internal attribution triggers a person's behavior by internal forces or tendencies.In external attribution, human behavior is triggered by environmental forces (Heider, 1958).Darwati (2015) is an amateur scientist who tries to understand the behavior of others by gathering and combining information until they reach a plausible explanation of the other person's particular behavior.Auditors must have audit experience, formal training, and work experience in their profession.Increased experience can be seen in the professional training of auditors, length of work, the number of customers, who are audited, and the length of time he has been auditor (Mulyadi, 2013).

Effect of Professional Skepticism on Fraud Detection
Confidence in the information or evidence obtained so that the examiner has a fairly high level of confidence in the information or evidence obtained from him in carrying out the examination process and also pays attention to its validity and adequacy (Anggriawan, 2014).The auditor's responsibilities in detecting fraud are the same, even though there are differences in the level of professional skepticism that the auditor has and the amount of experience the auditor has.
Research by Dasila and Hajering (2019), Helmiati (2021), Mokoagow et al. (2018), Ridwan et al. (2021), andSulistiyanti (2020) found that professional skepticism had a significant impact on fraud detection.Sanjaya (2017) found that professional skepticism had no significant effect on the auditor's ability to detect fraud.Some of the explanations put forward show the inconsistency of the relationship between professional skepticism and fraud detection.To detect fraud and material misstatements, the use of a skeptical mindset will lead to more conservative actions and develop additional information searches regarding the symptoms of fraud that occur.Therefore, the professional skepticism of an auditor is expected to assist the auditor in detecting fraud.H2: Professional Skepticism has a significant effect on fraud detection

Effect of Red Flag on Fraud Detection
Internal strength is closely related to the auditor's behavior and individual personality, including the auditor's ability to interpret red flags.A red flag indicates something is wrong, and the auditor needs to be aware of it.(Singleton and Singleton 2010:96), in their book, Fraud Auditing and Accounting Forensic, state that recognizing red flags is crucial for successful fraud detection and prevention.All of this is supported by several previous research results, namely Izevbigie (2020) and Laksana and Achmad (2020), which state that red flags have an effect on fraud detection.Based on this explanation, the hypotheses built are: H3: Red flag has a significant effect on fraud detection

Effect of Internal Control System on Fraud Detection
According to Laksana and Achmad (2020), auditors need to understand the company's management environment to detect fraud.It refers to policies, procedures, actions, and other activities that reflect the general attitude and behavior of the board, audit committee, and senior management regarding internal control.To detect fraud, the auditor must understand the company's management environment.According to Arens (2008from Dewi 2017), internal control consists of five components of the control environment, risk assessment, telecommunication control activities, and monitoring.These factors help the company achieve its goals.Based on previous research, namely from Yanti (2013), Izevbigie (2020), and Laksana and Achmad (2020) stated that the internal control system affects fraud detection.Based on this explanation, the hypotheses built are: H4: Internal control system has a significant effect on fraud detection Research framework of this paper is illustrated in figure 1 Figure 1 Research Framework

METHODS
In this study, the research methodology used is explanatory research.Explanatory research is a text that contains an explanation of processes related to natural, social, scientific, cultural, and other phenomena.The purpose of the explanatory text is to explain the formation process or activities related to natural, social, scientific, and cultural phenomena (Kosasih, 2014, p. 177).The approach used in this study is a quantitative approach where the data used are in the form of numbers and use hypothesis testing.
The object of this research is an auditor who works at a Public Accounting Firm (KAP) in the Jakarta area, with a review of auditors conducting audits to detect fraud.The researcher chose Jakarta as the research location because this location is more accessible, has relatively the same socioeconomic conditions, and is expected to attract more respondents and be more generalized in this area of choice.
Sampling was carried out using a sampling technique dedicated to this method.The sampling method considers exclusive (Sugiyono, 2016, p. 126).This method is a sampling method that does not provide equal opportunities for each element or member of the population to be selected as a pattern (Sugiyono, 2016, p. 125).
Data analysis used to in this paper uses structural equation modeling partial least square (SEM PLS).We use Smart PLS Software to do data analysis.A partial least square model is chosen with consideration of providing a better measurement of model feasibility, validity, and reliability compared to ordinary regression analysis.In partial least square, we measure outer loading, average variance extracted, and discriminant validity which is not measured in ordinary regression analysis.
The method of determining the sample in this study is purposive sampling.The sampling criteria determined by the researchers were auditors who were active in the Jakarta Public Accounting Firm (KAP).In this study, researchers were sampled by using unknowns because the population was uncertain, and researchers could not obtain information about the number of accountants working at Public Accounting Firms (KAP).This formula is used to determine the selected sample.The exact number of the population is unknown (Wibisono, 2003, p. 53).
The following is the formula for the unknown population: Based on the illustration of the calculation from the equation, the number of samples produced in this study is 96.04, which is rounded up to 96.This means that 96 samples that work in Public Accounting Firms (KAP) in the Jakarta area will be respondents in this study.Below in table 1 is the operationalization of the variables that we use to measure the dependent and independent variables.

ANALYSIS
Respondents in this study were auditors who worked at the Public Accounting Firm (KAP) in the DKI Jakarta area.This questionnaire itself consists of 102 respondents.The following in Table 2 is a description of the identity of the research respondents that the researcher included in data collection:

Source: Author
Judging from the information we present in table 2, most of our respondents are female, aged between 21-30 years old, and have a bachelor's degree.

Outer Model Test
The measurement model in this study has the aim of evaluating the results of the validity test and construct reliability test, while the measurement model used to evaluate the relationship between latent variables and each attribute is the outer model.The purpose of this evaluation is to evaluate the relationship between latent variables and each statement item.
The convergent validity test in this study aims to test the validity of each latent variable and each item.The data can be said to be valid if the results of the evaluation of the measurement model in the Structural Equation Model (SEM) show the value of the loading factor > 0.7 (Sholihin, 2013).The following Table 3 and Figure 2 are the result of the outer loading value of convergent validity after elimination: The discriminant validity test in this study aims to test the construct of the latent variable in predicting the size of the variable construct value, which is larger than the size of the correlation value of the other variables.To ensure that the measurement of a diverse construct empirically represents the phenomenon, which in other equation models has not been tested for discriminant validity is used (Hair, 2014).The discriminant validity test in this study used the results of the cross-loading.The following is the cross-loading discriminant validity in this study.

Source: Author
Based on Table 4 above, where the value of the main cross-loading of the constructs of auditor competence, professional skepticism, red flag, internal control system, and fraud detection is greater than the cross-loading value of these variables compared to other variables, this is it can be concluded that all variables in this study have good discriminant validity.
In this study, reliability is seen based on the value of composite reliability, namely in evaluating composite reliability itself by comparing the Average Variance Extracted (AVE) value with the correlation value between constructs.The AVE value or cut-off value used in this study is 0.5, meaning the value of the AVE must be > 0.5.Furthermore, in looking at the reliability of the variable construct, namely by looking at the value of composite reliability and Cronbach's alpha.Composite reliability so that the value can be said to be reliable it must be greater than 0.7 (Hair, 2014).5 above, it can be concluded that the AVE value of each variable has a value > 0.5, so we can conclude that the evaluation of the measurement model in this study is valid.The composite reliability value of each variable also has a value above 0.7, so it can be concluded that the variable construct in this study has a good reliability value or all variables can be said to be reliable.Based on Table 5 above, it can also be concluded that Cronbach's alpha value of each variable has a value above 0.6, so it can be said that all instruments used in this study are reliable.

Inner Model
The structural model in this study aims to evaluate the quality relationship between latent variables.Furthermore, the structural model or inner model in this study is used to find out how much information can be explained by the relationship between variables.The inner model in this study itself uses the SmartPLS 3.3 software.The final result of the evaluation of the inner model in this study is seen based on the results of the R-Square (R2) value or the coefficient of determination and Predictive Relevance (Q2), where the R-Square itself is used to measure the dependent variable and Predictive Relevance (Q2) is used to see model validation.
The coefficient of determination test itself is used to measure and determine the measurement of the dependent variable.The criteria for the coefficient of determination test are if the R-Square value > 0.67 indicates a substantial effect, > 0.33 indicates a moderate effect, and> 0.19 indicates a weak effect.The following is the value of the coefficient of determination from the R-Square test in this study.
Table 6 Coefficient of Determination Variable Dependent R-Square Description Fraud Detection (Y) 0,691 Substantial

Source: Author
Based on table 6 above shows that the coefficient of determination of the dependent variable in this study, namely fraud detection, has an R Square value of 0.691.From this value, we can conclude that the fraud detection variable or the dependent variable in this study is substantial, which means the dependent variable, namely all independent variables, can influence fraud detection.The competence of auditors, professional skepticism, red flags, and internal control systems of 69.1% and the rest of 30.9% are not used in this study.
The predictive relevance test in this study was conducted to measure how well the model this study can be observed and predict model validation.According to Ghozali (2014), the criterion in the predictive relevance test is that it can be said to be predictive relevance if the Q2 value > 0, and vice versa.It cannot be said to be predictive relevance if the Q2 value is < 0. In addition, the Q2 value, which is getting closer to number 1, means it has a model that is getting better.Based on Table 7 above, we can see that the value of Q2 of the dependent variable in this study is 0.317.It shows that this study has predictive relevance because the value of Q2> 0, meaning that the value of Q2 meets the requirements.In addition, the table above shows that the diversity of data from the designed structural model can be explained by 31.7%, of which other factors outside the research model explain the remaining 68.3%.

Hypothesis Testing
The final stage of this research is to test the hypothesis to measure the effect of the independent variable on the dependent variable.Hypothesis testing was carried out to evaluate and see the conclusion of the relationship of all variables in this study.In this hypothesis test, the researcher uses the SmartPLS 3.3 software using the bootstrapping method and looking at the path coefficient.Hypothesis testing in this study was carried out by looking at the probability value (p-value) and the tstatistics value of each variable which was then compared with the t-table value.The criteria for testing the hypothesis in this study are using a significance level of 5% with a confidence level of 95% and t-table 1.96, meaning that the path coefficient value can be said to be significant if the t-statistical value > t-table (1.96) and the p-value is below 0.05.The following are the results of hypothesis testing in this study:

Source: Author
The results of hypothesis testing in this study indicate that the auditor's competence variable has a significant influence on the fraud detection variable.It can be seen based on the p-value of 0.018 and the value of the t-statistic of 2.416, where the two values are respectively below the significance level of 0.05 and above the t-table 1.96.So it can be concluded that H1 is accepted.The results of this study are in accordance with the researcher's initial hypothesis, where the auditor's competence variable affects the fraud detection variable.
The results of this study support the results of previous research conducted by Wibowo & Nugroho (2020), which states that auditor competence influences fraud detection.Likewise, the results of research from Raihan (2020) were in the results of his research.He stated that auditor competence significantly affects auditors who will or are in the process of detecting fraud.Likewise, according to Sari & Komang (2020), auditor competence is one of the most significant influences.Dominant to fraud detection.It can be concluded from this study that the auditor's competence plays an active role and is important for an auditor who is or will do fraud detection.The higher the competence of an auditor, the higher the level of success of the auditor in detecting and fighting fraud.
The results of hypothesis testing in this study indicate that the professional skepticism variable does not have a significant effect on the fraud detection variable.This can be seen based on the p-value of 0.405 and the value of the t-statistic of 0.834, where both values are above the significance level of 0.05 and below the t-table 1.96.This concludes that H2 is rejected.The results of this study are not in accordance with the researcher's initial hypothesis, where the professional skepticism variable affects the fraud detection variable.
The results of this study support the results of previous research conducted by Sulistiyanti (2020), which states that professional skepticism does not affect fraud detection because there are still many factors that influence fraud detection more.However, this research is not in line with research conducted by Yanti (2013), Sanjaya (2017), Danila & Hajering (2019), and Ridwan (2021), who state that professional skepticism affects fraud detection.It can be concluded from this study that the amount of professional skepticism of an auditor does not affect the auditor in fraud detection.This states that in this study, the professional skepticism of an auditor cannot be used as the main factor for auditors who want to carry out fraud detection.An auditor's skepticism may only be able to indicate fraud but not in general because an auditor's professional skepticism may only help an auditor to have an untrustworthy attitude so as not to be influenced by either the internal or external auditors themselves.
The results of hypothesis testing in this study indicate that the red flag variable significantly influences the fraud detection variable.It can be seen based on the p-value of 0.038 and the value of the t-statistic 2.011, where both values are below the significance level of 0.05 and above the t-table 1.96.So it can be concluded that H3 is accepted.The results of this study are in accordance with the researcher's initial hypothesis, where the red flag variable affects the fraud detection variable.
Red flags positively affect auditors in detecting fraud (Muzdalifah & Syamsu, 2020).The results of this study support the results of previous research conducted by Adedokun & Oyewole (2013), which states that red flags make it easier for auditors to audit fraud detection.This research is also supported by Horne (2018) and Izevbigie (2019).They state that fraud detection will significantly benefit if the auditor can know and read the red flags in the fraud detection case so that the auditor can quickly and precisely conduct the fraud detection audit itself.It can be concluded that from this study, the number of red flags found by auditors influences the fraud detection process.The more red flags found, the more helpful the auditor's performance in the fraud detection process is because the auditor can be helped to find errors in the audit.
The results of hypothesis testing in this study indicate that the internal control system variable significantly influences the fraud detection variable.This can be seen based on the p-value of 0.000 and the t-statistic value of 3.975, where the two values are respectively below the significance level of 0.05 and above the t-table 1.96.So it can be concluded that H4 is accepted.The results of this study are in accordance with the researcher's initial hypothesis, where the internal control system variable affects the fraud detection variable.
The results of this study support the results of previous research conducted by Sanjaya (2017), which states that the internal control system is one of the mandatory components to measure fraud detection because the better the internal control system of a company or organization can make it easier for auditors to detect fraud.Previous studies that also support this research are Albert & Joseph (2015) and Kulikova & Satdrova (2016), where they state that the internal control system affects fraud detection.
It can be concluded that from this study, an auditor's internal control system plays an active role and is an essential thing for an auditor who is or will carry out fraud detection.It is stated in this study that an auditor's internal control system in an organization or company will greatly assist the auditor in doing so if it is functioning and organizes very well.So whether or not the weak internal control system can affect the fraud detection process.

CONCLUSION
The results of this study state that the competence of auditors has a significant influence on fraud detection.Professional Skepticism does not have a significant effect on fraud detection.Red flags have a significant influence on fraud detection.
Season and due to Covid-19 pandemic restrictions.The researcher realizes that this research could be better, so there are limitations during the process of compiling this research.The following are some of the limitations the researcher faces, including the time spent in data collection, which is relatively long because the researcher distributes questionnaires when the auditor or respondent is working in peak month.This research is only limited to the Public Accounting Firm (KAP) auditors in the DKI Jakarta area, so it cannot represent the condition.The limited sources of previous research are relevant to the title of this research.
The researcher has several suggestions for further research to produce much better research.Here are suggestions from the researcher: further research is expected to collect data from respondents outside the Peak Season to ensure a smoother data collection process.For further research, it is expected to determine the location not only in the DKI Jakarta area so that the research conducted can better describe the relationship related to Fraud Detection.For the academic community, that is being able to conduct research related to fraud cases and use more diverse independent variables.

Table 1
Operation of Variable Source: Author

Table 2
Description of Respondents

Table 3
Convergent Validity

Table 4
Discriminant Validity

Table 5
Reliability Test

Table 7
Predictive Relevant

Table 8
Hypothesis Testing