감사인의 비감사 서비스 제공과 기업의 전기오류수정 사이의 관계

Regulators, academics, and investors have paid great attention on accounting error corrections recently, because the increasing number of firms announce the prior error corrections. Accordingly, Korea Accounting Standards Board announced its accounting standard No. 1 “Accounting Changes and Error Corrections” at year 2000. According to the statement, the accounting error occurred in prior periods can be corrected in three different methods: 1) financial statement restatement, 2) reporting in other revenue and expense section of the income statement, and 3) reporting as an adjustment to the retained earnings in the equity section of the balance sheet. In this study, we examine the relationship between the provision of the non-audit service by auditors and the prior accounting error corrections. If the provision of non-audit service impairs audit quality, as argued by regulators and media after the Enron scandal, it is possible that financial statements prepared by the auditor who provide non-audit service simultaneously with audit service could contain more unintentional error or intentional bias. It is because auditors are not able to preform quality audit which find and correct error or bias contained in the financial statements, compared with the firms that do not receive any non-audit service from auditors. These error or bias would be revealed in later which would result in more frequent error corrections by firms. Although there exist prior studies examining the relationship between financial statements restatements and the provision of the non-audit service by auditors, this is the first study that examine the relationship between the entire range of error corrections, including financial statements restatements, reporting them in income statement, and reporting them in balance sheet, and the provision of non-audit service. In this respect, we believe that this study can contribute academics, regulators, as well as practitioners and users of the financial statements on the much debated effect of the provision of non-audit service by auditors on audit quality. Although prior studies have investigated this issue frequently, the findings in the prior study is mixed and generally inconclusive. Most of the studies failed to find any significant relationship between the two. However, the most frequently used proxy for the audit quality in prior studies are either audit opinion or discretionary accruals. The two proxies have their own shortcomings. First, because almost all audit opinion is clean and thus non-clean opinions are seldom issued by auditors, it is difficult to have enough statistical power to get any meaningful results for a study to use the audit opinions. Second, the discretionary accrual measure is noisy (Dechow et al. 1995) and includes a lot of measurement errors, though the measure is not prone to the lack of statistical power problem that audit opinion has. In contrast, the audit opinion has no measurement error problems. The prior error corrections used in this study is a good measure that does not prone to the measurement error problems at all. In addition, this measure has statistical power problem in lesser degree compared with audit opinion measures, because relatively more firms correct prior errors compared with the number of firms that receive non-clean audit opinions. In this respect, we argue that prior error correction is a good measure to perform tests on audit quality. Using 2,330 firm-year observations, which include 192 observations that made prior error corrections collected over 2000-2004 period, we empirically test the relationship if firms correct prior errors receive more non-audit service in previous period when the firms commit error or bias. However, we are not able to find any significant relationship in a battery of tests. We measure the magnitude/degree of the provision of the non-audit service in various ways - including a dummy variable having a value of 1 if the non-audit service is simultaneously provide or 0 otherwise, ratios on the amount of non-audit service fees based on different deflators such as total fees, audit fees, total assets, and amount of non-audit service fees as well as different types of error corrections, including over-reporting errors versus under-reporting errors, and different reporting methods, including income statement report, statement of retained earnings report, and prior financial statements restatements, and types of non-audit service, including information system related, audit related, tax related, and other non-audit service. The dependent variables used in the analyses are both the existence of error corrections (a binary variable) and the magnitude of error corrections (both raw and absolute value of the error correction). However, we fail to find significant relationship in most of the tests. These findings are robust to several other sensitivity checks, including the use of abnormal non-audit fees. We use the abnormal non-audit service fees in the analyses because the normal level of non-audit service fees could be determined by various characteristics of the firms, such as firm size, complexity, and risks of the firms. Only the abnormally high level of the fees, which are the difference between actual and normal level of fees, could motivate auditors to compromise their independence so that audit quality could be impaired. In short, we do not find any evidence that audit quality is impaired or enhanced by the provision of non-audit service. These findings are consistent with those in several prior studies such as Ashbaugh et al.(2003), Chung and Kallapur(2002), and DeFond et al.(2002). We expect the findings in this study provide valuable insights into regulators on policy-making and academics on the measure of audit quality. In addition, we try various kind of methods and different measures of audit quality as well as the degree or existence of prior error corrections and non-audit service provision by auditors. The use of these various methods could provide a way for subsequent research to improve the credibility of their research findings.