DETERMINANTS OF AUDIT FEE AT PUBLIC ACCOUNTING FIRMS IN EAST JAVA

This study examines the determinants of audit fee at Public Accounting Firms (KAP) in East Java, Indonesia. The determinants tested in this study consist of audit size, complexity, audit risk, non-audit services, and internal control. Respondents in this study are external auditors at KAP. Based on the convenience sampling method, the total sample in this study amounted to 60 respondents. This research uses Structural Equation Model (SEM) with SmartPLS analysis tool. The results indicate that audit size has a significant positive effect on the audit fee. Meanwhile, complexity, audit risk, non-audit services, and internal control have no significant effect on the audit fee.

The financial statements made by the company need to be audited by an independent and objective public accountant to ensure that the financial statements are free from material misstatements (Yuniastuti and Agoes, 2014). In this case, public accountants need to provide adequate audit quality. De Angelo (1981) states that audit quality is a combined probability in which the auditor can find and report fraud that occurs in the client's accounting system. Djamil (2000), mentions seven factors that affect audit quality, including 1) audit tenure, 2) some clients, 3) client's financial health, 4) the presence of third parties in reviewing audit reports, 5) auditor independence, 6) audit fee and 7) audit planning. This study examines the determinant one of audit quality's factors, namely the audit fee.
In Indonesia, Decree of the Chairperson of IAPI No. KEP. 024 / IAPI / VII / 2008 regulates guidelines related to the determination of service fees for public accountants. The regulation states that the determination of audit services benefits needs to consider the stages of work starting from audit planning, audit implementation, and audit reporting. Also, public accountants also need to consider other things, such as client's needs, duties and responsibilities, independence, level of expertise, complexity, length of time, and base of the agreed fee.
Since the research by Simunic (1980), the literature related to the determinants of audit fees is growing. Most are studies on private companies which show consistent results in nearly all factor ISSN: 1693-5241 99

Determinants Of Audit Fee At Public Accounting Firms In East Java
tested (Hay, et al., 2004). Simunic (1980) in Beattie, et al.(2000), developed a pricing model which shows that the audit fee is determined by the difference in loss exposure, the difference in loss-sharing ratio, the difference in the auditor's production functions, and auditor identity.
Xueand O'Sullivan (2012), concluded factors related to audit fees. First, the client's characteristics include size, complexity, risk, profitability, and leverage. Second, the characteristics of auditors, whether they are included as big-four or non-bigfour, audit tenure, auditor market share, and auditor base to conduct audits. Third, the number of engagements which include the audit timing, the length of the financial report and the signing of the audit report. Fourth, a joint provision for non-audit services that shows a positive relationship to audit fees.
The studies related to the determinants of audit fees are increasingly developing with researches thatexamine the effect of non-audit services variables and internal controls on audit fees. However, most studies are carried out from the client's perspective (auditee). This study examines the audit size, complexity, audit risk, non-audit services, and internal control variables from the perspective of external auditors at public accounting firms (KAP) in East Java.

Factors Affecting Audit Fee
According to Agoes (2012), the audit fee is a reward in the form of goods, money, or other forms given to or received from clients or other parties in an engagement. Based on the IAPI Chairman's Decree No. IAPI 024 / IAPI / VII / 2008, public accountants have to acknowledge the stages of audit work in determining audit fees. These stages, include: 1) Audit planning phase, which consists of preliminary planning, understanding the client's business, understanding the accounting process, understanding the control structure, conducting initial analysis, determining the level of materiality, creating an audit program, risk assessment, and discussing fraud with management. 2) The implementation phase, which consists of internal control testing, transaction substantive testing, analytical procedures, and transaction details testing.
3) The reporting phase consists of the review of contingent liabilities, review of events after the balance sheet date, testing of final evidence, evaluation and conclusions, communication with clients, issuance of audit reports, and capital commitment. Determination of the audit fee is commonly based on the contract between the auditor and the auditee in accordance with the time of the audit process, services, and the number of staffs needed for the audit process (Fierdha, et al., 2014). Arens and Loebbecke (2003),explain that the services provided by public accountants in the private sector can be in the form of operational audits, compliance audits, and financial statement audits. Most of the literature examining the determinants of audit fees shows that there are significant effects of factors derived from client's characteristics, such as size, complexity, risk, and non-audit services as well as factors. Factors derived from auditor's characteristics are size and location (Xueand O'Sullivan, 2012).

Audit Size
Client size is a scale that reflects the size of a company by looking at the total assets held (Suwito & Herawaty, 2005). Hay, et al. (2004) andSimunic (1980,argue that size is the most crucial determinant and has a significant positive effect on audit fees. In nearly all study, the proxy used to measure size is total assets. Measurement using size is typically conducted by running natural logarithm of the data to determine the linear effect between size and audit fee (Hay, et al., 2004 andChaney, et al., 2003). However, some studies such as Pong and Whinington (1994) in Chaney, et al. (2003), measure size using the proxy of assets turnover (the value of sales divided by total assets), with argument thatcompanies sales value and transaction volume are relative to its total assets, such asin service company. As a result, using only the total assets as measurement will not be sufficient. A study by Pop & Dana (2008), Cantoni, et al. (2011), Battaro, et al. (2015, andRusmanto &Waworuntu (2015), found a significant effect of audit size on audit fees. However, MARCH 2019

Audit Complexity
The complexity of the auditor's task is defined as the process of each to face the obstacles caused by the limitations of capability, memory and the ability to integrate problems (Widiarta, 2013). According to Wood (1988), complexity can be viewed in two aspects. First, it refers to the amount of information that needs to be processed and the stages of work performed on the audit assignment. Second, coordinative complexity which refers to the coordination level (between one part and another like a team) is needed in completing a task. Studies related to audit complexity by Ramzy (1988), Battaro et al. (2015), and Ulfasari andMarsono (2014), found a significant effect on the complexity of the audit on audit fees. On the other hand, studies byCantoni, et al. ( ), Ardiningsih (2013, and Rusmantro (2014), found that there was no significant influence on audit fees. H 2 : Complexity affects the audit fee.

Audit Risk
Audit risk is the risk faced by losses on third parties resulting from misstatement in a financial report. Tuanakotta (2014), mentions three components of audit risk, including the vulnerability of an assertion to material misstatement without taking related control into account (inherent risk), the vulnerability of an assertion to material misstatement that is not prevented, detected, or corrected in time by entity control (control risk), and the vulnerability in the procedure carried out by the auditor in suppressing the risk of material misstatement to a lower level that is acceptable (detection risk). Ulfasari and Marsono (2014), added that in the audit process, public accountants face litigation risk due to misstatements in the presentation of financial statements. Most studies that use proxy leverage, profitability, forms of ownership, and others (Hay, et al., 2004, Verbruggen, et al., 2011, andBattaro, et al., 2015 indicate a significant influence on audit risk on audit fees. However, Ulhaq (2015) and Chandra (2015), did not find any significant effect of audit risk on audit fees. H 3 : Audit risk affects the audit fee.

Non-Audit Services
Non-audit services are additional services (other than audit services) provided by KAP to clients. Examples of non-audit services at KAP include consultation, information and technology systems, legal, mergers and acquisitions, risk management, and others. Most studies describe consistent results that the use of non-audit services can increase the audit fee value (Ezzamel, et al., 1996). However, studies on health institution (NHS trusts) by Clatworthy, et al. (2002) and Basioudisand Ellwood (2005)

Internal Control
ISA 315 defines internal control as a process of designing, implementing, and managing companies by Those Charged with Governance (TCGW), management, and other employees to ensure the achievement of the entity's objectives. The internal control component based on the framework designed by COSO consists of the control environment, risk assessment, activity control, information and communication system control, and monitoring activities (Uwadiae, 2013). Goodwin-Stewart(2006), stated that high audit costs represent high audit quality. This is due to the audit committee's request for a high level of assurance that broadens the scope

Determinants Of Audit Fee At Public Accounting Firms In East Java
of audit checks (Goodwin-Stewart, 2006). On the contrary, research by Theodore (2015), found no significant influence on internal control on audit fees. H 5 : Internal control affects audit fees.

METHOD Sample
This study uses quantitative data obtained primarily through a research questionnaire on auditors in KAP East Java Region. Tomasoa (2013), states that mostly both small and large KAP are officed in East Java. Based on the IAPI directory data of 2016, there are 56 KAP in East Java, consisting of 43 offices in Surabaya, ten offices in Malang, 2 offices in Sidoarjo, and 1 office in Pasuruan.
The sample in this study consisted of 60 respondents. This number has passed the requirements of a study as stated by Roscue (1975), in Sekaran and Bougie (2013), that a decent sample size of a study is between 30 to 500. Roscue (1975), also added that for multivariate analysis research (i.e. correlation or multiple regression), the minimum number of sample is 10 timesthe number of variables tested.

Data Analysis
Data in this study were obtained using survey method with a seven points-likert scales questionnaire. The data obtained were then analyzed using the Variance Based Structural Equation Model (VBSEM) method with the SmartPLS analysis tool. According to Ghozali (2008), SEM is an evolution of multiple equation models developed from econometric principles and combined using the principles of psychological and sociological arrangements so that they are not based on many assumptions. Researchers used the PLS program because of the bootstrapping procedure. This method does not require a sample minimum limit. Testing in this study was carried out in two stages, including evaluation of the outer model (measurement model) and inner model (structural model).
Measurement model evaluation aims to test convergent validity and discriminant validity on the data. Yamin andKurniawan (2011), mention conver-gent validity testing is done by checking the indicators of validity and reliability, as well as Average Variance Extract (AVE) of a construct. On the other hand, discriminant validity is run by checking the cross loadings indicator and comparing the correlation of constructs with the root value of AVE.
The structural model evaluation aims to examine the level ofrelationship between independent and dependent variables by looking at the R-square value. Also, this test also shows the significance of the relationship of each independent variable to the dependent variable.

Testing of Measurement Models Convergent Validity Validity Indicators
This evaluation is completed by looking at loading factors for each indicator. The size of the reflection indicator is said to be high if it correlates to more than 0.7. However, Chin in Ghozali (2006), states that for the initial stage of developing a measurement scale loading values 0.5 to 0.6 are considered sufficient. This study uses a loading limit of 0.6 to test the validity indicator.
Based on the questionnaire which contains 30 indicators to test each audit size variable (5 indicators for audit size, 5 indicators for complexity, 3 indicators for audit risk, 5 indicators for non-audit services, 5 indicators for internal control, and 7 indicators for audit fees), a result is obtained by testing the validity indicator in Table 1.1 attached. The table shows that 5 out of 30 questionnaire indicators did not meet the limits of loading factors. Three out of the 5 statements (X1.4, X2.4, and Y1.7) that did not meet the loading factors were due to a negative statement. In this case, the respondent was unfocused in filling out the questionnaire which in result made the answer for the particular statement was invalid. Meanwhile, two statements (X5.2 and Y1.3) do not statistically have adequate loading factors. This can be caused by the fact that the researcher was not directly involved in the monitoring process of filling out the questionnaire, and therefore several respondents did not provide answers with adequate consideration. MARCH 2019

Nurkholis, Gede Krisnawan
The correlation between constructs and variables at the beginning has not met convergent validity because there are still indicators that have a value of loading factors below 0.6 to be retested by issuing the indicators that are valued under loading factors from the test. The value in Loading Factors Processed column is the result of testing the items that meet the loading limit in the Beginning Loading Factors column. The test results indicate that all indicators processed have met the loading limit of 0.6 so that none of the constructs for all variables is eliminated.

Construct Reliability andAverage Variance Extract (AVE)
In this test, the criteria were declared reliable if the composite reliability value and Cronbach's alpha were more than 0.70. Also, the AVE value of each variable must be above 0.50 (Yamin and Kurniawan, 2011 Table 1shows that each latent variable in this study has a composite reliability value and Cronbach's alpha is above 0.70. Also, the AVE value for each latent variable is above 0.50, so the construct has good convergent validity.

Cross Loadings
In this test the construct is said to be valid if each indicator has loading factors that are greater than the correlation with indicators that form other latent variables. The following are the results:  Based on Table 2 above, it can be observed that all indicators forming latent variables have a value of loading factors greater than the correlation with indicators forming other latent variables.
The next test is run by comparing the correlation of constructs with the root value of AVE. This result can be seen through the Fornell-Larcker Criterion section on the results of testing using SmartPLS. The construct stated has good discriminant validity if the AVE root value is higher than the maximum construct correlation value. Here are the results: Table 3 shows that all constructs have a maximum correlation value that is not greater than the AVE root value, so the constructed model has good discriminant validity.

Testing of Structural Models R-Square
R-Square testing shows the influence of research variables consisting of audit size, complexity, audit risk, non-audit services, and internal control valued at 0.288. This can indicate that these variables only explain 29% of the determinants of audit fees. Meanwhile, the other 71% is explained by other variables out of the research scope.

Path Coefficient
In testing the path coefficient, an independent variable is stated significant to the dependent variable if the t-statistic result is greater than 1.64 with a significance of 5%. The following are the results of the t-statistical tests obtained in this study: MARCH 2019

Nurkholis, Gede Krisnawan
Table 4 above shows that the audit size variable (AUS) has an original sample value of 0.706 which means that the effect on the audit fee is positive. With a t-statistic value of 2.555, which is greater than 1.64, it is explained that audit size has a significant positive effect on audit fees. In this study, the H 1 hypothesis is supported, in other words, the larger the client size, the higher the audit fee set by KAP. Complexity (COM) has an original sample value of -0.161 which means that the effect on the audit fee is negative. The t-statistic value of 0.712 indicates that the H 2 hypothesis is not supported, which means that complexity has no significant effect on the audit fee.
Audit risk (AUR) shows the original sample value of -0.160, which means that the effect on the audit fee is negative. The t-statistic value of 0.711 indicates that the H 3 hypothesis is not supported, meaning that the audit risk does not affect the audit fee. Non-audit services (NAS) have an original sample value of -0.036, which means the effect is negative, but with a t-statistic value of 0.183, the H 4 hypothesis is not supported, meaning that non-audit services have no significant effect on the audit fee. Internal Control (INC) has an original sample of 0.080 which means its effect is positive. With a tstatistic value of 0.366, the H 5 hypothesis is not supported, meaning that internal control has no significant effect on the audit fee.

RESULT AND DISCUSSION Influence of Audit Size on Audit fee
The test results show that the audit size determinant has a significant effect on determining the audit fee with a t-statistic value of 2.555 indicating that H1 is supported. This result is aligned with most of the previous studies related to the effect of audit size on audit costs, such as the study by Simunic (1980), Ramzy (1988), Battaro et al. (2015), Amrita& Darya (2013),Ulfasari andMarsono (2014), and Ulhaq (2015). This relationship can be attributable to the condition that the higher total assets, transaction volume and sales, as well as the number of client subsidiary companies,the broader the scope of testing conducted by the auditor. This statement is supported by Ramzy (1988),who states that indicators in the form of inventory, work in progress, turnover, debtors, creditors, total assets, loans, share capital and reserves, current assets, sales profits, and profit before a tax can make the amount and complexity audit work. Increasing the volume of work would require more personnel, higher testing costs, and other expenses as the size of the client company being audited increased. Ulhaq (2015), mentions that with the increasing size of the company the use of labour and effort is also increasing in the audit work. Also, auditors tend to charge high audit fees to large clients because of adequate financial capacity (Battaro, et al., 2015).
This result is contrary to the research by Firth (1985) in Suharli and Nurlaelah(2008) and Rahayu (2017), who did not find a significant influence on the size of the company on audit fees. Firth examined the relationship between company size and audit fees in New Zealand in 1981-1983. Researchers use the company's total assets as a proxy for company size and divide the research sample by 96 companies into two groups based on the size of the company. Large companies are the ones with total as-

Determinants Of Audit Fee At Public Accounting Firms In East Java
sets of more or equal to NZ $ 21,000,000, while small companies are the ones with total assets below NZ $ 21,000,000. Based on that study's results, it did not find a significant influence on the size of the company on the determination of theaudit fees amount.

Influence of Complexity on Audit fee
The test results show that audit complexity has an insignificant relationship in determining the audit fee with a t-statistic value of 0.712, indicating that H2 is not supported. This result is in accordance with research conducted by Cantoni, et al. (2011), Ardiningsih (2013), and RusmantoandWaworuntu (2014. This can happen because companies in East Java have characteristics that tend to be homogeneous as compared to large cities, such as Jakarta and its surroundings so that the level of complexity faced by auditors in an audit assignment is a common thing. This statement is supported by the opinion fromArdiningsih (2013), who explained that although there are related party transactions in the client companies that may contribute to audit risk, these transactions are common, especially in large companies listed in the Indonesia Stock Exchange.The complexities which appear to be complex at first are then dismissed because it has become a common practice. Cantoni, et al. (2011), in the sector of charity organizations in the United Kingdom (UK) justify the insignificance of the influence of the complexity of the determination of audit fees, namely due to differences in the characteristics of the activities carried out by the organization concerned, for example in terms of organizational structure, dimensions and scope operation. This can be correlated with clients in the private sector, that the complexity of audit work is not only sourced from operational characteristics but also client organization.
The study by Suharli and Nurlaelah (2008), shows that there is no influence of some subsidiaries on audit fee. In the results of their research, Suharli andNurlaelah (2008), explained that companies going public with fewer subsidiaries incurred higher audit cost if audited by the Big 4 KAP. Meanwhile, companies that do not go public but with more subsidiaries incurred lower cost when audited by non-Big 4 KAP.
This finding is in contradiction to the results of Ramzy(1988), which found that audit complexity has a significant effect on determining audit fees. Ramzy (1988),rationalises the consequences of many client subsidiaries, decentralized financial controls, and minimum use of computer technology as the causes of increased audit fees.

Influence of Audit Risk on Audit fee
The test results show that audit risk has no significant effect in determining the audit fee with a tstatistic value of 0.711, indicating that H3 is not supported. This result is in line with research by (Battaro, et al., 2015,Ulhaq, 2015, and Chandra, 2017. This can occur because audit risk such as default risk which is a vulnerability on account balance or class of transactions against material misstatement cannot be controlled. Also, the existence of the risk of detection in an audit that comes from the vulnerability of the auditor in detecting material misstatements in the client's business is something that cannot be charged to the client even if the risk is related to the audit assignment to the client concerned. This is supported by research. The statement was supported by Pradika (2009), who found that detection risk did not have a significant effect because the risk was sourced from the auditor and not from the client. This is evidenced by the majority of studies which state that there is a significant influence on audit risk testing using proxy leverage, profitability, forms of ownership, and others (Hay, et al., 2004, Verbruggen, et al., 2011, and Battaro, et al., 2015. Also, study Ulhaq (2015), of external auditors in Pakistan supports the results of this study. Ulhaq (2005), indicates that most rules and laws in Pakistan are not as strict as developing countries. Countries with the application of strict audit rules tend to address the existence of risk in an audit assignment and the auditors always skeptical of fraud in the client company.
The results of this study are contrary to research by Fauziah (2010),Saputri, et al. (2017, and Rahayu (2017), who found that audit risk has a significant effect in determining audit fees.

Influence of Non-Audit Services on Audit fee
The test results show that non-audit services have no significant effect on audit fees with a tstatistic value of 0.183, indicating that H4 is not supported. This owes to the fact that KAP in East Java treats non-audit services as a separate part of the audit to maintain auditor independence so that the use of non-audit services does not affect the number of audit fees. Decision by the Chairman of the Capital Market Supervisory Agency (BAPEPAM) Number 20 / PM / 2002 dated November 12, 2002 point 3 (d) states that a public accountant is declared not independent if during his professional assignment period provides non-audit services to clients, such as bookkeeping or other services relating to client accounting records, financial information system design and implementation, actuarial, internal audit, and so on.
This statement is confirmed by many studies previously conducted regarding the effect of audit fees on auditor independence (Supriyono, 1998, Widodo, 2002,Ardiani and Ricky, 2011, and AditamaandUtama, 2015. The results in this study indicate that KAP does not consider significant the existence of non-audit services in the determination of audit fees, because it can threaten the independence of public accountants in the particular KAP. This result is in contrast to research by Fauziah (2010) and Cantoni, et al. (2011), who found a positive relationship between non-audit services and audit fees. On the other hand, Clatworthy, et al. (2002) and Basioudisand Ellwood (2005), found a negative influence from the use of non-audit services on the number of audit fees set by KAP. The variety of results obtained in previous studies lead to thedifficulty in drawing definitive conclusions on the relationship between non-audit service and the audit fee.

Influence of Internal Control on Audit fee
The test results show that internal control has no significant effect on audit fees with a t-statistic value of 0.366. This result is aligned with the research conducted by Widiasari, et al. (2008), Chandra (2015), and Pohan (2017). This can be caused by the presence of a board of commission-ers or an audit committee that is responsible for internal control of the client so that KAP tends not to consider the internal control of the client in determining the audit fee. Based the letter issuance from BAPEPAM SE-03 / PM / 2000 dated May 5, 2000, and the decision issued by the Jakarta Stock Exchange Number KEP-339 / BEJ / 07 / 2001, it is required that companies listed on the Jakarta Stock Exchange to have an audit committee (Prastuti, 2013). This statement was confirmed by Hay, et al. (2008),after the audit committee became a requirement for companies listing on the New Zealand Stock Exchange, testing the audit committee variable does not have a significant effect on the audit fee. This supports the statement by Yatim, et al. (2006), who found no significant relationship between external audit feewith leadership, board size, the frequency of board meetings, and the existence of a risk management committee. Chandra (2015),established his research based on the IAPI Decree Number KEP.024 / IAPI / VII / 2008 which concluded that the number of audit fees was determined according to the agreement between the KAP and the company's audit committee. Therefore it is unrelated to the internal control carried out by the company.
The results of this study contradict the research by Prastika (2016), who found a significant influence of internal control on the determination of audit fees. Like Prastuti (2013), Prastika (2016),argues that the internal control component is very important in gaining an understanding toprepare the overall audit work program which affects the consideration in determining the audit fee. In addition to internal control Prastika (2016), also found that the influence of strong, good corporate governance tends to demand high audit quality, thus affecting the number of fees determined.

Conclusion and Implications
Based on the results in this study, audit size has a significant positive effect on audit fees, meaning that the larger the size of the company in an audit assignment the higher the audit feeset. Complexity does not affect audit fees which can be caused by the homogeneity of the characteristics of the client being audited so that complexity is not a significant thing to consider. Audit risk does not have any significant effect on audit fees because the audit risk component cannot be fully charged to the client to determine the audit fee. Non-audit services do not have a significant influence on audit fees. KAP treat of non-audit services as a separate part of audit servicesto maintain auditor independence. Internal control does not have any significant influence on audit fees which can be attributable to the existence of the board of commissioners and audit committee.
The implications of this study, in which the further research can examine in the same field, is to use bigger sample size and examine other determinants of audit fees, such as auditor psychology in determining the audit fee. Also, respondents included in the study should be chosen from a partner or manager level who are directly involved in determining the amount of the audit fee.