THE IMPACT OF CLIENT IMPORTANCE ON EARNINGS MANAGEMENT AND GOING-CONCERN OPINION: EMPIRICAL EVIDENCE FROM INDONESIA

This study investigates the impact of client importance on auditor independence measured by accrual earnings management and going-concern opinion. Our research sample consists of 1,080 firm-years of observation from listed companies in the Indonesia Stock Exchange (IDX) with eight years of observation (2004-2011). With pooled OLS we found that client importance has a negative impact on accrual earnings management. We also found evidence that client importance has a positive impact on going-concern opinion. Collectively, our evidence suggests that although audit firms have economic dependence toward their clients, those audit firms can maintain their audit quality by keeping their independency toward their clients, reflected by the lower accrual earnings management and higher tendency of issuing going-concern opinion. Our results are robust by considering the results of our sensitivity tests that support the main results.


INTRODUCTION
The audit quality of public accounting firms has been criticized in the last decade, since they could not protect the interests of investors (Coffee 2002;Levitt 1998). This audit quality has been defined by various definitions. DeAngelo (1981), for example, defines audit quality as the joint probability that the auditor will find a breach in the client's accounting system, and report that violation. In this definition, audit quality covers the dimensions of competence and independence. Fitriany (2011) suggests that a higher degree of compentence from auditors is not always be accompanied by a higher degree of independence. In terms of independence, audit quality can be measured from different aspects, i.e., economic dependence toward clients, the influence of client practices toward the propensity to meet earnings targets, corporate governance practices, the propensity to issue going-concern opinions, the conducted peer review, or other measure of independence (Bamber and Bamber 2009).
Due to the auditor-client relationship, auditors may have incentives to compromise with their clients depending on the auditors' economic interests on their clients (DeAngelo 1981;Reynolds and Francis 2001). The proposition that the auditors will report favorably to retain important clients is known as the economic bonding (DeAngelo 1981). The independence of auditors can be reflected in the extent to which auditors will tole rate opport unistic earnings management, and how likely the auditors will issue going-concern opinion (Herusetya 2012; Francis and Yu, 2009).
Most of prior studies found that economic dependence does not encourage larger clients to perform greater discretions in accrual-based earnings management (e.g., Reynold and Francis 2001;Francis and Yu 2009;Herusetya 2012). Some previous studies found evidence that auditor independence impaired when clients pay for non-audit services is relatively higher compared to the total fee (Frankel, Johnson and Nelson 2002). They found evidence of a positive association between client importance (i.e., the ratio of non-audit services to total fee) and the amount of discretionary accruals. While Chung and Kallapur (2003) found no evidence on the association between client importance and abnormal accruals.
Past research also found the association between client importance and audit opinion issued by public accounting firms (e.g., Reynolds and Francis 2001;Craswell, Stokes and Laughton 2002;Chen, Sun and Wu 2010). Reynolds and Francis (2001) found that Big 5 did not treat larger clients with more compro-mise than smaller clients based on the size of individual practice offices, i.e., by not reducing the propensity to issue going-concern opinions. Existing financial incentives were considered enough to motivate the auditors to be independent, despite the fact that the presence of economic dependence relationships was inherently in the auditor-client contract. Similarly, Craswell et al. (2002) found that fee dependence as a threat to auditor independence does not affect the likelihood of auditors to issue qualified audit opinions, both at national and local market levels.
The lack of evidence on the previous studies, which examined the associations between client importance, earnings management, and going-concern opinion are the main reasons of our study. T he lack of past research on auditor independence in Indonesia, as measured by the economic dependence of auditors toward their clients will make some contributions to this study. First, Marchesi (2000) for example, found that audit quality in ASEAN countries was so compromise. But as the climate of audit environment in the United States has started to change since the corporate failure in 2001, which also gave impact to the legal and audit environment around the world including Indonesia 1 (Fargler and Jiang 2008;Chen et al. 2010), we expect to provide evidence on these associations. Secondly, past researches on audit quality were more dominated by the measurement of the competence dimension, for example, using the size of accounting firms (Big 4, etc) (Fitriany 2011;Siregar 2006;Francis 2004). Audit quality of the Big 4 firms in Indonesia may differ from the Big 4 firms in the United States who are under the Sarbanes Oxley Acts (Gordon et al. 2013). Our study contributes to the measurement of independence of the audit firms in Indonesia using a proxy of client importance which includes a sample of the Big 4 firms. LITERATURE REVIEW AND HYPO-THESIS DEVELOPMENT Client Importance as a Measure of Auditor Independence It was long been a suspicion that the auditors' financial interest toward their clients can greatly affect the auditor independence (Mautz andSharaf 1961 in Chen et al. 2010). DeAngelo (1981) suggests that auditors' incentives to compromise their independence toward their client depends on the importance of a client to the auditor (client importance), i.e. the ratio of quasi rents specific to the client divided by the total number of quasi rents received by the auditor. DeAngelo (1981) observed that if the percentage of the entire fee depends on one client used as a measure of audit quality (i.e., perceived independence with respect to that client) then the audit quality will decrease. The proposition that auditors will report favorably to retain important clients is known as economic bonding (DeAngelo 1981; Francis and Yu 2009), fee dependence, or economic dependence (Reynolds and Francis, 2001;Chen et al. 2010). It is probable that the greater the size of the clients in auditor's portfolio, the greater the incentive for the auditor to retain his clients, in which audit quality is likely to be compromised (Chen et al. 2010).

Client Importance and Earnings Management
Since the corporate scandals involving auditors in the United States by the end of 2001 that gave birth to the Sarbanes-Oxley Act (SOX) of 2002, the change of audit environment is also felt through over the world, including Indonesia. The Indonesian Government issued various regulations concerning the audit services provided by audit firms. These regulations were expected to increase the competence and the independence of the auditors (Herusetya 2012).
Auditor independence can be reflected in the extent to which the auditor became more tolerant to the earnings management done by the clients. The higher the quality of the independence of the auditors, the lower the earnings management behavior was reflected. Previous research has examined the association between client importance and earnings management (e.g., Herusetya 2012;Reynolds and Francis 2001;Frankel et al. 2002;Chung and Kallapur 2003;Francis and Yu 2009). Reynolds and Francis (2001) examined the association between client importance and audit outcome, i.e. total accruals and total discretionary accruals as measures of earnings quality. They suspect that the client importance (INFLUENCE variable) at the level of individual practice office can reduce auditor independence, but they found that Big 5 auditors did not treat larger clients more compromisingly compared to smaller clients.
With a sample of 6,568 U.S. firmsyears observations for the period 2003 to 2005, Francis and Yu (2009) examined whether the larger offices of Big 4 auditors have higher audit quality. Using an INFLU-ENCE variable as a measure of the auditor's incentives with respect to a fee-dependence and used as a control variable, they found that larger Big 4 practice offices have a negative association with absolute abnormal accruals. Herusetya (2012) in his study used a composite measure of audit quality that covered the dimension of competence and independence, and client importance (CI) was one of the proxies used in the dimension of independence. By using a single proxy of CI, he found that CI has a negative association with the absolute discretionary accruals.
In summary these findings conclude that even though the auditors have economic dependence on their clients, but the auditors do not encourage greater discretion in accrualbased earnings management 2 . In other words, the auditors do not provide greater tolerance to earnings management behavior for their important clients 3 . Based on the arguments given above, the hypothesis to be tested is: H 1 : Client importance is negatively associated with earnings management.

Client Importance and Going-Concern Opinion
Auditor's independence can be reflected in the propensity to issue going-concern opinion. Auditor independence will decline if the auditor compromise not to report the financial condition of the clients related to going-concern problems, because the auditor has economic dependence to the client (DeAngelo 1981;Reynold and Francis 2001). Previous research found evidence on the association of client importance and audit opinion issued by public accounting firms (Reynolds and Francis 2001;Craswell et al. 2002;Chen et al. 2010) 4 . Reynolds and Francis (2001) found that Big 5 did not treat large clients with more compromise than the smaller clients. Reynolds and Francis (2001) found the size of individual practice offices of the Big 5 did not reduce the propensity to issue goingconcern opinion. Existing financial incentives were considered enough to motivate the auditor to be independent, despite the fact that the presence of economic dependence relationships was inherently in the auditor and client contract. Similarly, Craswell et al. (2002) found that the fee dependence as a threat to auditor independence does not affect the likelihood of auditors to issue a qualified audit opinion, both at the national and local market levels.
Chen et al. (2010) conducted a study in China, and found that the level of auditor independence increase in line with the improvement of legal and regulatory environment. Chen et al. (2010) found that after 2001, when the legal and regulatory environment began to grow well in China, the tendency for clients to receive modified audit opinion (MAO) is higher compared to the period before 2001 (i.e., 1995-2000) for client importance measured by individual auditors. Similarly, Li (2009) found that after the period of SOX in the United States, higher audit fees and total audit fee ratio have a positive associ-ation with the auditor's propensity to issue going-concern opinion. The findings of Li (2009) and Chen et al. (2010) provide additional evidence that in line with the improvement of the institutional and regulatory climate, the tendency of auditors to compromise in audit quality seems to decrease, given the risks of litigations and penalties for auditors are higher than the economic incentives they earned. Based on the above arguments, the hypothesis will be tested is: H 2 : Client importance is positively associated with the propensity to issue going-concern Opinion RESEARCH METHODOLOGY Population and Sample  Hypothesis H 1 of this research can be formulated in statistical form as follows: H 1 : 1 < 0; the expectation for each variable control is: 2 0, 3 < 0, 4 > 0, 5 < 0, 6 > 0, 7 > 0. All variables are defined as in Table 2.
The main variable in Model 1 is CI, and the coefficient 1 is predicted negative and significant toward the absolute discretionary accruals (ABSDAC), which indicates that client importance has a negative association with accrual earnings management. = Absolute discretionary accruals scaled by lagged total assets, using Kothari et al. (2005) estimated model, computed cross sectionally each year for each industry GCO = Dummy variable for going-concern opinion, equal to 1 if the company receive going-concern opinion in the current year; and 0 otherwise CI = Client importance, as the ratio of audit firm's economic dependence toward its client SIZE = Natural logarithm of total assets CFO = Cash flow from operation of client i in year t scale by lagged total assets LEV = Leverage ratio, defined as total liabilities divided by total assets at year end t LOSS = Dummy variable for loss firm, 1 if firm i at year t report net loss; 0 otherwise ABSTACC = Absolute value of total accruals of client i in year t SGRW = Sales growth, defined as (sales t -sales t-1)/sales t-1 PRIORGC = Dummy variable, 1 if firm i receives going-concern opinion in the last year financial statements; 0 otherwise CASH = Cash and cash equivalent LLOSS = Dummy variable for prior year loss, 1 if firm i report net loss in t-1; 0 otherwise Equation (1) To control other factors that affect absolute discretionary accruals, we include several variables in Model 1. Clients with larger size (SIZE) has less discretionary accruals, due to larger clients have more resources to maintain the sustainability of the entity (Becker, Defond, Jiambalvo and Subramanyam, 1998;Reynolds and Francis, 2001;Balsam et al. 2003). However, large companies have also higher market pressure to meet analysts' expectations (Barton and Simko 2002), thus having greater opportunities to perform earnings management (Tresnaningsih 2008). Therefore we do not predict the sign of SIZE coefficient. Higher level of sales would increase accruals but will reduce operating cash flows (Dechow, Sloan and Sweeney 1995), thus we expect a negative sign for CFO coefficient. Companies with the higher level of liabilities (LEV) have more incentives to deal with earnings management compared to the lower level of liabilities (Reynolds and Francis 2001;Balsam et al. 2003), therefore leverage (LEV) is predicted to be positively related to ABSDAC. Dechow and Schrand (2004), and Francis and Yu (2009) found a negative association of the company with net loss toward its accruals quality, therefore LOSS coefficient is predicted negative. ABSTACC, the absolute value of total accruals scaled by total asets t-1 is used to control the propensity to generate accruals, and is predicted to have positive relationship with ABSDAC (Balsam et al. 2003;Francis andYu 2009). McNichols (2000), and Menon and William (2004) found evidence that sales growth has a positive association with abnormal accruals, due to high growth companies have higher discretionary accruals adjustment. We predict SGRW has a positive association with ABSDAC.
The main variable in Model 2 is CI, and we estimate the coefficient 1 is positive and significant toward going-concern opinion (GCO), which indicates that client importance has a positive association with the propensity to issue going-concern opinion. Several control variables included in the Model 2 to control other factors that may affect audit firms to issue going-concern reports. Companies with larger size (SIZE) have greater resources and are more likely to survive compared to smaller companies (Francis and Yu 2009), therefore we predict variable SIZE has a negative association with GCO. Companies that received going-concern opinion on the previous year (PRIORGC) are more likely to receive goingconcern opinion in the current year (Reynolds and Francis, 2001). CASH is a liquidity measure that is the sum of cash and cash equivalent, scaled by total assets. Companies with more liquid assets have more resources to deal with financial difficulties, therefore we expect CASH coefficient is negative. Companies with high debt levels (LEV) and experience net loss in the current year (LOSS) are more likely to fail and more likely to receive going-concern reports, therefore, we predict LEV and LOSS coefficients have positive associations with GCO (Francis and Yu 2009). Operational Variables Absolute Discretionary Accruals (ABSDAC) We use accrual model from Kothari, Leone and Wasley (2005) to measure the discretionary accrual which is as follows:  (1) Discretionary accruals equal the value of residual errors ( ) or the difference between total accruals and the fitted value of normal accruals, defined as DA it = (TACC it ) -NDA it (Cohen, Dey and Lys 2008) 6 . The absolute value of discretionary accrual (ABSDAC) is used as a proxy of accrual earnings management by considering the possibility of positive and negative accrual earnings management (Myers, Myers and Omer 2003;Barton and Simko 2002;Cohen et al. 2008). All variables are defined as in Table 2.
Going-Concern Opinion (GCO) GCO is a dummy variable (1; 0), equal to 1 if the auditor assumes that the company cannot survive within 12 months after the balance sheet date, and receive going-concern opinion; and 0 if otherwise (Geiger et al. 2005).
Client Importance (CI) Client importance (CI) examines the tendency of auditors having economic dependence that may reduce the auditor independence (e.g., Frankel et al. 2002;Craswell et al. 2002;Chung and Kallapur 2003;Chen et al.  Where: CI it = client importance, as a measure of audit firm's economic dependence toward the client. SIZE it = natural logarithm of client's assets. = the sum of total assets (in natural logaritm) from n clients of certain audit firm.

Control Variables
We use control variables which have been used in the prior research as follows: company size (SIZE), cash flow from operation (CFO), leverage (LEV), net loss (LOSS), absolute value of total accruals (ABSTACC), prior year going-concern opinion (PRIORGC), cash and cash equivalents (CASH), sales growth (SGRW), prior year loss (LLOSS). See Table  2 for variable definitions.

RESULTS AND DISCUSSION Descriptive Statistics and Correlations
The descriptive statistic of all operational variables can be seen on Table 3 7 . The mean of ABSDAC in Table 3, Panel A is 0.110 or 11.0% of the total assets, indicating the magnitude of accrual earnings management. The mean of GCO (Panel B) is 0.230 indicating that the average firm-years observation to receive going-concern opinion is 23.0%. The mean of CI is 0.239, indicating that the average proportion of audit firms' revenue per client is 23.9%. In Table 4, CI variable has a positive correlation with CGO at a significance level of 0.01, consistent with our prediction.     Table 2.

Test of Hypothesis 1 (H 1 )
Hypothesis H 1 predicts a negative association between client importance (CI) and accrualbased earnings management (ABSDAC). Empirical result of Model 1 in Table 5 has adjusted R-square of 85.68% with F-stat 923.766 (prob < 0.01) 8 . The coefficient of CI ( 1 ) is -0.011 (t-stat = -1.303), negative but not significant at 0.10 with two-tailed test. However, the coefficient of CI is negative and significant at 0.10 with one-tailed test (critical value of t-stat = 1.28), consistent with our prediction.  F-statistic) 0.000 n =1,080 ***, **, * significant at 0.01, 0.05, and 0.10, respectively in two-tailed tests; tstatistics are calculated using the Huber-White procedure to correct heteroscedasticity. All variables are defined as in Table 2.
This result implies that the bigger the proportion of the economic dependence of audit firms toward their clients, the smaller the magnitude level of accrual earnings management. Our result gives interpretation that audit firms in Indonesia that have larger clients and bigger economic dependence to their clients, are less likely to allow earnings management behavior. This result is also consistent with previous research that found larger offices of Big 4 provides higher audit quality and have clients with less earnings management behavior (Francis and Yu, 2009).
Four control variables are significant in the expected direction (CFO, LOSS, AB-STACC, SIZE), and the other control variable is not significant at the 0.10 (LEV), or is significant in the opposite direction (SGRW).

Test of Hypothesis 2 (H 2 )
Hypothesis H 2 predicts a positive association between client importance (CI) and going-concern opinion (GCO). Logistic regression in Model 2 (Table 6) has Cox and Snell R-square 53.70%, and Nagelkerke R-square 80.90% with the value of -2Log Likelihood 336.233 9 . The result of the test in Model 2 indicates that the coefficient of CI ( 1 ) is 1.218 (Wald test = 5.923), positive and significant at 0.05 with two-tailed test. This result implies that although the proportion of audit fee in audit firm's portfolio is bigger, audit firms can still maintain their independence, measured by the issuance of going-concern opinion. Our result is consistent with previous studies, which document that larger offices of public accounting firms are more likely to issue going-concern opinion, and give better prediction for financial distress in the upcoming period (e.g., Francis and Yu, 2009).
Five control variables are significant in the expected direction (SIZE, PRIORGC, LEV, LOSS, LLOSS) and the other control variable is not significant at the 0.10 (CASH). 80.9 n =1,053 ***, **, * significant at 0.01, 0.05, and 0.10, respectively in two-tailed tests. All variables are defined as in Table 2.

Sensitivity and Robustness Tests
We use alternate measure of client importance (CI), i.e., client's sales instead of total assets to test the robustness of our main results. Conversely to the hypothesis result of H 1 , we find no evidence of the association between client importance (CI) and accrual-based earnings management (ABSDAC) at 0.10 (not tabulated). The usage of client importance model using total sales as a proxy becomes sensitive and do es not support the main result.
Sensitivity test for hypothesis H 2 using the same measure of client importance (CI) finds similar result as in the main test ( Table  6) (not tabulated). Our result indicates that client importance (CI) has a positive association with going-concern opinion (GCO) at 0.05. The usage of the different measure of client importance (CI) in Model 2 is robust to all alternative test and support the main result. This result strengthens our main result and previous study (e.g., Francis and Yu 2009), which suggest that larger offices of audit firms are more likely to issue going-concern opinion rather than smaller offices.

CONCLUSION
Our study investigates the independence of auditors measured by the client importance, and reflected in the accrual-based earnings management and going-concern opinion. The main result of our study finds a negative association between client importance and accrualbased earnings management. Our evidence implies that audit firms with larger clients and with more economic dependence to their clients are less likely to allow earnings management behavior. This result is consistent with previous studies that suggest that larger audit firms, like Big 4 with more expertise and ability to detect earnings management behavior are less likely to allow earnings management rather than smaller audit firms (Francis and Yu 2009;Reynolds and Francis 2001).
Our second main test finds that client importance has a positive association with going-concern opinion. This evidence implies that although the proportion of client dependence is bigger, public accounting firms can maintain their independence measured by the issuance of going-concern opinion. Our result is consistent with the previous studies (e.g., Reynolds and Francis 2001) and robust to other sensitivity tests that support the main test result. The combine results indicate that client importance as a measure of auditor independence has a negative association with earnings management and has a positive association with going-concern opinion. This research has some limitations: (i) the conclusions drawn in this study should be conducted with caution, because we use only single measure of auditor independence, i.e., the client importance. There are some other measurements, such as the influence of client practices toward the propensity to meet earnings targets, and corporate governance practices (Bamber and Bamber 2009); (ii) accrualbased earnings management is the only measure of earnings management used in this study. Previous research documented several other earnings management techniques, such as real earnings management, and classification shifting (e.g., Cohen and Zarowin 2010;Fan et al. 2010).
We recommend for further study to use the composite measure of independence. Herusetya (2012) for example, the use composite measures of audit quality from the dimension of competency and independence. In order to reflect the different aspect of earnings management, we also recommend to examine the other tools of earnings management, such as real earnings management and classification shifting (Cohen and Zarowin 2010;.