Effect of Audit Quality on Earnings Quality: Evidence From Indonesia Stock Exchange

This research was conducted to test the effect of audit quality on earnings quality. Its observed data consisted of 116 annual data of manufacturing companies listed in the Indonesia Stock Exchange within 2011-2014. The proxies of audit quality are auditor size, audit tenure and audit specialization. The earnings quality was formed from the attributes of accrual quality, persistence, predictability, and income smoothing. The analysis techniques for this research involved (1) an confirmatory factor analysis to form the earnings quality and (2) multiple regression analysis to test the effect of the auditor size, audit tenure and audit specialization on earnings quality. Analysis results showed that earnings quality is formed by the attributes of persistence and predictability. Research results showed that auditor size and audit tenure have effect on earning quality, while audit specialization do not.


Introduction
Earnings quality is one of the pieces of information that investors rely on to make decisions. Information of earnings given by the management through financial reports are also affected by management policies, which means that the management manages information that could be directed for certain purposes (Scott, 1997).
There are a lot of financial and accounting implementation cases of companies that directly or indirectly involve external auditors such as the Enron and WorldCom cases including several cases of public companies in Indonesia. External auditors are also considered to be less sensitive to the risks that occur in clients, less-independent, and lack of professional objectivity and responsibility which tend to affect audit quality. Based on this phenomenon, this study examines the effect of audit quality on earnings quality.
In this research, audit quality has the proxies of auditor size (the Big-4), audit tenure and audit specialization. Audit quality returned to public attention due to a number of accountancy and financial cases which indirectly involved public accountants, such as Waste Management, Enron, WorldCom, and several cases in Indonesia. Earnings quality in this research is a composite variable from the attributes of accounting earning qualities, made up of accrual quality, persistence, predictability, and income smoothing (Francis et al., 2004).

Theoretical Framework
Agency theory recognizes that there is a tug-of-war of interests between owners or principals and managers. The problem of information integrity appears, for which one reason is due to the information asymmetry. Investors possess limited access to sufficient information to make decisions. Beneficial behaviors become the tendency for managers who possess more information than other stakeholders (Jensen and Meckling, 1976). Information, including quality information on earnings, is very much needed by investors as material for making decisions. Auditors as external parties may be the ones to rely on in neutralizing conflicts of interests and reducing the information asymmetry, thereby maintaining the information integrity of financial reports (Almutairi et al., 2009;Varici, 2013). Companies may obtain economic and control benefits from auditing activities; one of these is that audits can increase information quality and earnings quality. Audits play an important role in decreasing the information asymmetry and moral hazards in order to provide certainty of information to stakeholders, so that the financial reports that are prepared by managers as agents can be relied upon .

Audit Quality and Earnings Quality
Audit quality has multiple meanings. In relation to earnings quality, audit quality is the accurateness of information given by an auditor to an investor to make more accurate estimations of company value (Titman and Trueman, 1986). Palmrose (1986) affirms that audit quality is indicated by financial reports that do not present the material misstatement. The measurement of audit quality in research is derived from, among others, audit firm size (Alzoubi, 2016;Francis and Wang, 2008;Wang et al., 2014) and audit tenure (Al-Thuneibat et al., 2011;Davis et al., 2000;Gul et al., 2009;Myers et al., 2003). Earnings quality shows accurate and unbiased earnings information which can explain company performance (Bissessur, 2008). The advantage of earning quality is that it can be used to help make decisions (Dechow et al., 2010). Further, this research considers the four indicators of accounting earnings quality, namely accrual quality, persistence, predictability, and income smoothing (Francis et al., 2004).

Auditor Size
Auditor size indicates the large scale income and organization of the public accountant firm, which is now called Big 4 (including its affiliation), comprising PricewaterhouseCoopers (PwC), Deloitte Touche Tohmatsu (Deloitte), Ernst and Young (EY), and Klynveld Peat Marwick Goerdeler KPMG). Previous research had documented the correlation between audit quality being measured by a measurement proxy of a Big 4 auditor (Al-Thuneibat et al., 2011;Alzoubi, 2016;Francis and Wang, 2008;Houqe et al., 2015;Kabir et al., 2011;Wang et al., 2014;Zakaria and Daud, 2013) and the earnings quality measured by a certain proxy such as earnings response coefficient (Alzoubi, 2016;Houqe et al., 2015;Teoh and Wong, 1993;Wang et al., 2014;Zakaria and Daud, 2013), discretionary accruals (Alzoubi, 2016), and abnormal accruals (Francis and Wang, 2008;Wang et al., 2014). The Big 4 auditors have a better audit quality than small or non-Big 4 auditors arguably because large audit firms have greater resources, knowledge, technical experience, capacity, and reputation compared to small audit firms.
Research by Zakaria and Daud (2013) also found that the Big 4 auditor have a significant and positive influence on earnings response coefficient (ERC). Results of research by Alzoubi (2016) showed that the level of earnings management is significantly lower for companies that use the Big 4 firms compared to those that use non-Big 4 firms. Houqe et al., (2015) found proof that companies that hire high-quality auditors in proxy with auditor size Big 4 have a negative influence on the earnings quality in proxy with discretionary accruals and income smoothing, which means that being audited by Big 4 auditors makes the earnings quality higher. The Big 4 auditors have greater audit quality compared to non-Big 4 auditors with the argument that the Big 4 auditors possess greater staff quality, knowledge, technical experience, capacity, and reputation compared to non-Big 4 auditors. Based on the previous research results and explanation, the hypothesis was formulated as H 1 : Auditor size has a positive affects on earnings quality.

Audit Tenure
Several studies have examined audit tenure with earnings quality, including (Al-Thuneibat et al., 2011;Davis et al., 2000;Ghosh and Moon, 2005;Gul et al., 2009;Johnson et al., 2002;Manry et al., 2008;Myers et al., 2003;Yazawa, 2014). There are two opinions that form the basis of the strength of the correlation between audit firm tenure and earnings quality. The first opinion is that audit firm tenure has a positive correlation with earnings quality, which is supported by the argument that a long audit firm tenure will result in the auditor having a better understanding of the client's condition. Research by Manry et al., (2008), Myers et al., (2003), and Gul et al., (2009) concluded that audit firm tenure has a negative correlation with discretionary accruals (better earnings quality). Audit firms that are continually assigned will have a sufficient understanding of the client's business, including understanding of the client's internal control structure. This is used by the auditor to design audit procedures, resulting in a higher-quality audit report. A longer audit tenure has the implication of a higher earnings quality in proxy with the earnings response coefficient (Ghosh and Moon, 2005). argument that a longer audit firm tenure will reduce the level of auditor's independence. Results of research by Johnson et al., (2002) support this opinion, that as the auditor-client relationship progresses, the report quality becomes lower, which indicated by a positive correlation between tenure and earnings persistence. A long audit firm tenure can threaten the independence of an auditor, caused by a client-auditor relationship that grows closer, which can result in the increase of the discretionary accrual proxy of audit quality (Al-Thuneibat et al., 2011;Davis et al., 2000). Davis et al., (2000) concluded that audit quality decreases as an audit tenure becomes longer, because as the audit tenure becomes further extended, the client possesses a greater financial reporting flexibility. This explanation indicates that there is a correlation between audit firm tenure and earnings quality. Based on the previous research results and explanation, the hypothesis was formulated as H 2 : Audit tenure has a positive affects on earnings quality.

Audit specialization
Krishnam (2003) and (Balsam et al., 2003) found that auditor with industry specialization had lower discretionary accruals than otherwise. Gul et al., (2009) examine whether auditors of industrial specialization affect the relationship between auditor tenure and earnings quality. The results show that the shorter audit tenure of auditor specialist and non-specialist industry who audit certain companies, the lower the earnings quality of the companies will be. This research supported research of Balsam et al., (2003).
Auditors who have a concentration on a certain industry and audit procedures are likely to to gain more and in-depth knowledge about the client's business and industry, so that auditors with industry specialization can work more effectively. The effectiveness of the auditor's work reflects the expertise in detecting accounting irregularities that have an impact on the quality of information. Based on the results of the research and explanation, the hypothesis to formulated as, H 3 : Audit specialization has a positive affects on earnings quality.

Data Collection, Sample and Analysis
Data samples were determined by purposive sampling from manufacturing companies listed in the Indonesia Stock Exchange within 2011-2014. The 116 observed data were collected from documentation. The data were obtained from annual reports published by the Indonesia Stock Exchange and other supporting data sources. This research used the confirmatory factor analysis to determine the composite variable from four earnings quality attributes, made up of accrual quality, persistence, predictability, and income smoothing (Fanani, 2011;Francis et al., 2004;Pagalung, 2006). The regression analysis was conducted to test the influence of the independent variable on the dependent variable after performing tests for normality, autocorrelation, multicollinearity, and heteroscedasticity (Santoso, 2015).

Dependent Variable
Dependent variable of this research is earnings quality. Earnings quality refers to the capability of reported earnings to reflect the actual earnings of the company, as well as the usefulness of reported earnings to predict future earnings (Bellovary et al., 2005;Francis et al., 2004). Earnings quality is determined based on the results of confirmatory factor analysis of the four attributes of earnings quality in accounting base, made up of accrual quality, persistence, predictability, and income smoothing (Fanani, 2011;Francis et al., 2004;Pagalung, 2006).

1) Accrual quality
Accrual quality refers to the amount of income which is recognized when business unit rights arises due to the delivery of goods to an outside party and costs which are recognized when the obligation arises due to the use of economic resources attached to the goods delivered (Francis et al., 2004). Accrual quality measurement using Francis et al., (2004) and Larcker and Richardson (2004)  Persistence is the condition where the current period profit is a reflection of the future period and the current period. This proxy for earnings quality is shown by the coefficient value of the annual earnings regression model with the formula (Francis et al., 2004). where: Earnings jt: earning before extraordinary accounts of company j in year t Earnings jt-1: earning before extraordinary accounts of company j last year Outstanding of share jt: Shares outstanding in company j in year t Outstanding of share jt-1: Shares outstanding in company j last year The measurement of persistence is done by looking at the beta value (β1) of the equation above. This variable is calculated based on the estimation from 2011-2014 using the fixed effect approach.

3) Predictability
Predictability is the ability of current earnings to predict future earnings. The predictibility is proxied by using the ARI model below, which is the root of the variance of squared errors from the model (Francis et al., 2004). where: Earnings jt: earning before extraordinary accounts of company j in year t Earnings jt-1: earning before extraordinary accounts of company j last year Outstanding of share jt: shares outstanding in company j in year t Outstanding of share jt-1: shares outstanding in company j last year Predictability is measured by looking at the error with the formula of the root of the error variable squared (√ (Ѷjt)) of the equation above. This variable is calculated by estimating during the year, 2011-2014 using the fixed effect approach.

4) Income smoothing
Income smoothing shows the difference between companies in terms of the variability of actual earnings according to cash flow. Proxy of financial reporting information quality is resulted from the calculation of the ratio between standard deviations of earnings before extraordinary accounts per total asset divided by standard deviation of operating cash flow per total asset (Francis et al., 2004).

1) Auditor Size
Audit firm size indicates the Big 4 group of audit firms and/or their firm affiliates in Indonesia. This variable is measured the same way as in a research by Alzoubi (2016), using a dummy variable of 1 if the company is audited by a Big 4 audit firm and 0 otherwise.

2) Audit tenure
Audit tenure indicates the number of years in a row a company is audited by the same audit firm, as utilized by Johnson et al., (2002) and Al-Thuneibat et al., (2011).

3) Audit specialization
An auditor is said to have industry specialization if the number of companies audited by an auditing firm is more than 15% of the number of companies audited by all auditing firms in a particular industry group, measured by a dummy variable i.e 1 for auditors who have industry specialization, and 0 otherwise (Solomon, et al., 1999;Zhou and Elder, 2004).

Confirmatory Factor Analysis
The main objective of confirmatory factor analysis is to explain the structure of relationships among many variables in the form of factors or latent variables or composite variables. The criteria that must be fulfilled in the confirmatory factor analysis are a) a Kaiser-Meyer-Olkin (KMO) value ≥ 0.5 or the probability value of Bartlett's Test ˂ 0.05, b) a Measure of Sampling Adequacy (MSA) value that must be greater than 0.5, c) a Communalities value greater than 0.5, d) only one factor or component with eigenvalues greater than 1 (Hair et al., 1998), and e) examination of the loading factor or lambda value (λ). The loading factor shows the matching or unidimensionality of the indicators that form the variable. Indicators with loading factors ≥ 0.4 are included in the analysis (Ferdinand, 2002) because they constitute the variable, while the attributes with loading factors ≤ 0.4 or negative are not included in factor analysis.

Regression Analysis
The To show the effect of the independent variables on the independent variables t test is employed, which level of significant smaller than 5% (sig. value <0.05). The regression models are employed to test the alternative hypothesis (Ha) is Ho: β1 = 0; and Ha: β1 ≠ 0.

Results And Discussion
The summary of the confirmatory factor analysis results is shown in Table 1. In Table 1, shows that all values of KMO and Bartlett's Test for the variable of earnings quality are 0.500 with 0.000 significance. This indicates that factor analysis can be carried out. In the anti-image matrics, the attributes that has a low value (less than 0.5) are dropped and not included in the next stage of analysis. The values of communalities for the attributes of earnings quality are 0.538, 0.788, 0.786, and 0,492 for accrual quality, persistence, predictability, and income smoothing respectively. The value of each attribute is used to determine the number of representative factors from the original attributes. The total of all four communality values is 2.603. To achieve this value, there needs to be two factors or components that have eigenvalues greater than 1, which are single factors of 1.574 and 1.029, which is also the case for the next stages. The loading factor indicates the matching, agreement, or unidimensionality of the attributes forming the variable. Indicators with loading factors  0.4 or negative in value are not included in the model (Ferdinand, 2002). Empirically, the four attributes of earnings quality form a single composite variable supported by two attributes, which are persistence and predictability, each with a loading factor value of 0.887.
Results of normality testing using Normal P-Plot indicate that the data spread across a diagonal line and follow the direction of the diagonal line and supported by asymp. sig. value of one-sample Kolmogorov-Smirnov test is 0,089. Autocorrelation does not occur, as shown by a Durbin-Watson score of 2,066 that is between the value of 1,7504 and 2,2496 (between dU and 4-dU) and supported by asymp. sig. value of run test is 0,351 (≥ 0,05). Data multicollinearity did not occur, as shown by a tolerance value, less than 1, and VIF value less than 10. Heteroscedasticity testing used a scatterplot, indicates that the data show no clear spreading pattern and the points are spread out above and below 0 on the Y axis.
Descriptive statistics indicates that auditor size, with a minimum value of 0 and maximum value of 1 with a mean of 0.5259 indicates that more companies (53%) are audited by accounting firms that belong to major (Big 4) firms.
Auditor tenure, with a minimum value of 1 and maximum value of 6 with a mean of 3.4397, indicates that there are companies that are newly audited by certain firms, or reassignment of the old audit firm after changing to a new firm. There have also been many assignments to a certain firm in a row, up to 6 times. This condition is possible because of the requirement to change audit firms from old ones to new ones after six years of audit assignment, in accordance with regulation of Finance Minister, Republic of Indonesia No.17/Menkeu.01/2008. Audit specialization, with a minimum value of 0 and maximum value of 1 with a mean of 0.0862 indicates that around 8% of companies are audited by auditor with industry specialization. Earnings quality, with a minimum value of 0,02 and maximum value of 2,67 with a mean of 0.7308. Results of regression analysis (Table 2) for the auditor size variable on earnings quality indicates a probability value of 0.004, which means that auditor size has an influence on earnings quality, and thus hypothesis H 1 is accepted. The resulting coefficient value of -0,294 (negative direction) indicates a positive influence. It can be understood that continued audit by a Big 4 audit firm increases earnings quality progressively. Results of regression testing for the audit tenure on earnings quality indicates a probability value of 0.009, which means that a audit tenure has an influence on earnings quality, and thus hypothesis H 2 is accepted. The resulting coefficient value of -0,080 (negative direction) indicates a positive influence. It can be understood that a longer audit assignment by the accounting firm increases earnings quality progressively. Results of regression testing for the audit specialization on earnings quality indicates a probability value of 0.121, which means that a audit specialization has not an influence on earnings quality, and thus hypothesis H 3 is not accepted.
The F statistic value of 5.387 is significant to 0,002, which means that the independent variable has a simultaneous influence on the dependent variable. Meanwhile, the R 2 value is 0.126, which means that the contribution of the variable in forming the model is 12.6% and the remaining 87.4% is the contribution of other variables that may explain earnings quality.
The summary of regression analysis and hypothesis results is shown in Table 2.

Conclusion
Analysis results indicate that the formation of the variable of earnings quality is supported by the two attributes of persistence and predictability. Test results indicate that auditor size has a positive significantly influence on earnings quality. The test results imply that continued audit by a Big 4 audit firm increases earnings quality progressively. The results of this research support the results of previous research (Alzoubi, 2016;Houqe et al., 2015;Teoh and Wong, 1993;Zakaria and Daud, 2013) that concludes that auditor size has a positive influence on earnings quality. The measurement of earnings quality in this research is different from the measurement of earnings quality in previous research. Measurement of earnings quality in this research is composed of the attributes of persistence and predictability, while in previous research, earnings quality is measured by accrual quality, including discretionary accruals (Alzoubi, 2016;Houqe et al., 2015) and earning response coefficient (Teoh and Wong, 1993;Zakaria and Daud, 2013). The results of this research support the explanation that Big 4 auditors possess better audit resources than non-Big 4 auditors. Large audit firms possess resource capacities, including staff number and quality, application of methodology and technology in carrying out audits, internal control of the audit firms, and audit procedures that are better compared to small auditor firms.
The results of this research indicate that audit tenure has a positive significantly on earnings quality. The test results imply that extended audit by the same audit firm increases the values of the earnings quality. The results of this research are in line with results of prior research (Ghosh and Moon, 2005;Gul et al., 2009;Manry et al., 2008;Myers et al., 2003), which concludes that assignment of audit firms has a positive influence on earnings quality. There is a difference in the measurement of earnings quality in this research compared to previous research. Earnings quality in previous research is measured by the decrease of discretionary accruals (Manry et al., 2008;Myers et al., 2003) and absolute discretionary accruals (Gul et al., 2009), and the increase in earnings response coefficient (Ghosh and Moon, 2005). This research does not agree with the findings of Al-Thuneibat et al., (2011) and (Yazawa, 2014) in which it was found that audit tenure negatively influenced earnings quality, as extended audit by the same audit firm decreased earnings quality. The results of this research affirm that extended or repeated audit firm assignments further gives insight into the condition of the client, allowing the accounting firm to have the chance to improve work procedures and quality, thereby increasing information quality.
The results of this research indicate that audit specialization has not influence on earnings quality. The results of this research are not support the results of previouse research that audit specialization has positive influence on earning quality that showed through the decreasing of discretionary accruals (Balsam et al., 2003;Gul et al., 2009), absolute discretionary accruals (Krishnan, 2003) and increase the earning respone coefficient (Balsam et al., 2003). Another conclusion is that independent variables have a significant and simultaneous effect on dependent variables. The strength or contribution of the variable in forming the model is still weak, which means there are still other variables that can explain earnings quality.

Limitations and Suggestions
The sample for this research was determined based on specific criteria purposive sampling for manufacturing companies. Consequently, the results cannot be generalized for companies other than manufacturing companies. This research only used three independent variables, i.e. auditor size, audit tenure and audit specialization, resulting in a relatively small determination coefficient value R 2 of 0.124, which means the model still call for further development. Based on this limitation, it is suggested that further research should add other variables that may affect earnings quality (persistence and predictability) such as treasury flow volatility, accrual size, sales volatility, level of debt, and operational cycles (Fanani, 2010).
The brand name of Big auditor and audit tenur become determinant factors for better information of earning quality. Which can be used by management to consider when they appoint auditor and by regulator when they make regulation concerning, particurlarly the condition of partner and accounting firm. However, the future research need to explore other measurement or proxies for auditor size in addition to auditor label, Big 4 and non-Big 4.