The Effects of Shared-opinion Audit Reports on Perceptions of Audit Quality

The Effects of Shared-opinion Audit Reports on Perceptions of Audit Quality Yan-Jie Yang, Yuan Ze University, College of Management, Taiwan. Email: yanie@saturn.yzu.edu.tw Qian Long Kweh, Universiti Tenaga Nasional, Department of Accounting, Malaysia. Email: qlkweh@gmail.com Abstract We explore how capital market investors perceive the shared-opinion audit report (SOAR). Using earnings response coefficients to proxy for investors perceptions of audit quality, we observe negative earnings surprise when the SOAR is announced as compared to unqualified audit opinions. The perceptions are more adverse if the SOAR involves a larger percentage of component firms that are not audited by the group engagement partner. Moreover, we find that the negative perceptions are more pronounced in the samples of Big 4 audit firms and controlling interests. These results are robust to estimation across various subsamples and using an alternative measure of earnings-returns regression model. Key words: Shared opinion; Audit quality; Earnings response coefficient; capital market perceptions JEL Classification: M41, M42, M48 1

1. Introduction An analysis of the impact of the SOAR on perceptions of audit quality could provide important insights into the usefulness of the SOAR, particularly whether investors discount SOAR announcements. In this study, we use audit data in Taiwan, where on average, 43.08% of the annual reports of Taiwanese listed firms received SOARs as compared to those in the United States of only 5.1% [1]. Unlike in the United States, SOARs are relatively common in Taiwanese listed firms. Although SOARs are prevalent in Taiwan, only Jiang and Yeh [2, written in Chinese] and Chang et al. [3, written in Chinese] have examined the association between SOARs and accruals quality and the impact of the SOAR on earnings management in Taiwan, respectively. Our paper differs from prior studies along two important dimensions. First, following prior studies [for example 4, 5-7], we apply a useful representation of the view of the capital market. Using the earnings response coefficient (ERC) as a proxy for perceptions of audit quality, we examine whether any difference exists between SOARs and unqualified opinions. Second, we hand collect 3,756 firm-year observations with respect to the percentage of component firms that are not audited by the group engagement partner. On the one hand, after controlling for other determinants of the ERC including systematic risk, growth, financial leverage, firm size, firm age, the deviation of control and cash flow rights, and diversification we find that the ERC of the SOAR sample is significantly lower than the ERC of the unqualified opinions sample. The ERC of the SOAR sample further decreases as the percentage of component firms that are not audited by the group engagement partner gets larger. On the other hand, our robustness tests show that the association between SOARs and ERC becomes weaker when it involves Big 4 auditors or it involves subsidiaries in the SOAR. On the whole, our results are generally consistent that investors may have doubts about the audit quality of SOARs. One implication of our empirical evidence is that capital market investor advocates might view SOARs as informative. Therefore, while this study is important for its own sake, it is also significant in the effect its insights may have on audit engagements for business groups as well as audit planning and issuance of audit opinions by the group engagement partner. The remainder of the paper is organized as follows. Section 2 reviews prior literature and develops hypotheses. Section 3 discusses the sample selection procedure and the research design. Section 4 presents the empirical results of the association between SOARs and ERC. Section 5 concludes the paper. 2. Literature Review If business groups engage the same audit firm to audit their financial reports, both the groups and audit firm are able to achieve economies of scale and audit efficiency. Using data from the Belgian capital market, Branson and Breesch [8] find that business groups auditor 2

changes are controlled by the controlling firms. Besides, auditor changes in component firms are decided by the parent firm and auditor selection is associated with the business groups characteristics but not the component firms characteristics. Prior studies also find that nonexecutive directors draw on their networks of interlocking directorates to impact auditor choice [9, 10], which suggests the strengthened relationship between non-executive directors and auditors. One advantage of engaging the same auditor in a business group is that the auditor is able to apply uniform audit procedures, which would result in ease of monitoring by the group engagement partner and gaining of industry expertise through accumulated experience [8, 11]. Intrinsically, the group engagement partner is thus able to make appropriate udgment in issuing audit report. However, audit quality of business groups has not escaped scrutiny from capital market investors. Using modified Dechow and Dichev [12] model to estimate accruals quality, Jiang and Yeh [2] find that firms that are with high levels of global diversification and not audited by the group engagement partner have lower earnings quality when they receive SOARs. Meanwhile, Chang et al. [3] claim that earnings management in business groups is significantly greater than that in non-business groups when component firms of the business groups are audited by different audit firms. The authors further document that the earnings management phenomenon is particularly found in business groups whose parent firms are audited by Big 4 auditors. This scenario, however, is not rampant in business groups with component firms engaging the same auditor as the parent firm. Taken together, SOARs could manifest firm auditor selection and such firm characteristics as earnings management behavior. Furthermore, SOARs reflect audit quality, particularly the group engagement partner s degree of assurance, which might affect capital market investors evaluation on the financial statements of business groups. This situation in turn would have impacts on the share prices of business groups. Since the group engagement partner is allowed to not declare component auditors in the group audit report and investors are aware of SOARs, we expect that investors perceive the audit quality of business groups receiving SOARs as lower than those receiving unqualified audit opinions. If the group engagement partner declares the shared responsibility in the audit report, the percentage of component firms that are not audited by the group engagement partner shown in the report reveals the group engagement partner s unwillingness to assume component auditors responsibilities. Specifically, investors would perceive business groups receiving SOARs with higher percentage of shared responsibility as lower. These arguments lead us to develop the following hypotheses to test our expectation: Hypothesis 1: Capital market investors perceive the audit quality of SOARs as lower than that of unqualified audit opinions 3

Hypothesis 2: Capital market investors perceptions of the audit quality of SOARs become lower as the percentage of component firms that are not audited by the group engagement partner gets higher. 3. Methodology 3.1 Sample Selection The data used in this study are drawn from the Taiwan Economic Journal (TEJ) database for firms listed on the Taiwan Stock Exchange (TWSE) and the GreTai Securities Market (GTSM) for the period 2000 2010. We begin with all firms with audit reports and identify 11,881 firm-year observations, among which 3,997 cases (33.64 percent) are unqualified audit opinions and 7,838 cases (65.97 percent) are unqualified audit opinions with explanatory language. Of the 7,838 unqualified audit opinions with explanatory language, 4,677 cases are audit opinions whereby the group engagement partner engages other auditors or shares responsibilities with other auditors (SOARs); 2,515 cases are issued only because of shared opinions without any other explanatory paragraph (clean SOARs). In this study, we include only unqualified audit opinions and audit opinions whereby the group engagement partner engages other auditors or shares responsibilities with other auditors because we examine the impact of SOARs on perceptions of audit quality. After excluding firms in the financial services industry and firms that do not have data on the TEJ database, we obtain a sample of 6,895 firm-year observations, among which 3,139 cases are unqualified opinions and 3,756 are SOARs. As we trace and test SOARs, we only have 5,153 opinions for our sensitivity analysis (2,014 cases are issued only because of SOARs without any other explanatory paragraph). 3.2 Empirical Models Following Teoh and Wong [4], Ghosh and Moon [5], Lim and Tan [6], and Chi et al. [7], we use the ERC estimated from earnings-returns regressions as a market-based proxy for investors perceptions of audit quality. Specifically, we run the following OLS regression model to test whether investors perceive the audit quality of SOARs as lower than that of unqualified audit opinions. CAR UE Share UE Share CV 0 1 2 3 3 1 10 1 13 UE CV Year_and_Industry_Fixed_Effects 10 (1) Where: CAR = 12-month cumulative market-adusted abnormal returns (ending on the day of issuance of annual reports); UE = earnings surprise, measured by difference between continuing operations earnings per share (EPS) of year t and EPS of year t-1, scaled by stock price of outstanding ordinary shares at the beginning of the year. 4

Share = a dummy variable equal to 1 for firms that receive SOARs, and equal to 0 otherwise. CV = one of the ten control variables discussed below, = 1, 2,, 9. UE ( ) or the ERC is the proxy for capital markets perceptions of audit quality. Our 1 variable of interest for the first hypothesis is UE Share ( ), which tests whether 3 earnings-returns relation differs between unqualified audit opinions and SOARs. If investors perceive the audit quality of SOARs as lower than that of unqualified audit opinions, is 3 expected to be significantly negative. In addition, the following cross-sectional time-series specification is used to test the hypothesized relation between SOARs and perceived audit quality described in H2: Investors perceptions of audit quality of SOARs becomes lower as the percentage of component firms that are not audited by the group engagement partner gets higher. 10 1 10 BS BS 0 1 2 3 3 1 CAR UE Share UE Share CV 13 UE CV Year_and_Industry_Fixed_Effects (2) where BS Share refers to the percentage of component firms that are not audited by the group engagement partner by using the proportion of the dollar amounts of audit shared, i.e. the ratio of the total assets of component firms audited by component auditors to the total assets of the group. Other variables are as defined previously. We predict that as the percentage of audit shared gets larger, perceived audit quality becomes lower; that is, the UE BS Share coefficient ( 3 ) is expected to be significantly negative. In addition to the variables measuring our hypothesized statements, our regression models include various control variables and their respective interactions with earnings surprise (UE ) to ensure comparability with prior studies [13-18]. They are as follows: Beta = the past 5-year market returns, calculated by using the market model; Growth = the ratio of market value of equity and book value of total debt to book value of total assets at the beginning of year t; Lev = the ratio of total debt to total assets at the beginning of year t; Loss = a dummy variable equal to 1 if earnings in year is negative, and equal to 0 otherwise; Size = the natural logarithm of market value of equity at the beginning of year t; Age = the number of years since the firm was established; Big4 = a dummy variable equal to 1 if the auditor is from a Big 4 audit firm, and equal to 0 otherwise; Tenure = the number of consecutive years that the firm has retained the audit firm; Dev_Vrcf = the deviation between voting rights and cash flows rights of the parent firm; Divf = the ratio of the absolute value of the difference between the operating income of the parent firm and that of the group to the operating income of the group. To control for the possibility of omitted time-specific effects, we allow the regression intercept to vary across years; and the predictability of earnings is industry-specific, we also include dummy variables for the industry, both of which are represented by Year_and_Industry_Fixed_Effects. Finally, because the error terms (ɛ) are likely to 5

exhibit cross-sectional correlation and autocorrelation, we compute the t-statistics in the regression models using Rogers s [19] robust standard errors, correcting for firm clusters. 4. Results and Discussion 4.1 Descriptive Statistics Table 1 reports descriptive statistics for variables in Models (1) and (2). 1 We divide the observations into two groups: unqualified audit opinions ( Share = 0, hereafter the Share0 group) and SOARs ( Share = 1, hereafter the Share1 group); and we test the statistical significance of differences between mean values of the Share0 group and the Share1 group using a paired t-test. The differences between the two groups are pronounced. The mean UE for the Share1 group is 0.027, whereas that for the Share0 group is 0.017. A two-tailed t-test suggests that the difference of 0.010 is significantly different from zero. Moreover, the average values of Beta, Lev, Size, Age, Tenure, Divf, and Loss of the Share1 group are significantly greater than those of the Share0 group. However, the mean values of Growth and Dev_Vrcf of the Share1 group are smaller than those of the Share0 group. On average, the result also shows that the percentage of component firms that are not audited by the group engagement partner ( Share BS ) is approximately 12.071 percent. Table 1: Descriptive statistics (n = 3,139) (n = 3,756) Variable Unqualified audit opinions Shared audit opinions Mean S.D. Mean S.D. Difference t-statistic CAR 0.005 0.406 0.004 0.399 0.009 0.88 Share 0-1 - - - Share BS 0-12.071 14.376 - - UE 0.017 0.216 0.027 0.234 0.010 1.70* Beta 0.768 0.309 0.866 0.273 0.098 13.88*** Growth 1.488 1.579 1.346 0.983 0.142 4.55*** Lev 0.409 0.167 0.452 0.163 0.043 10.72*** Loss 0.209-0.224-0.015 - Size 7.677 1.384 8.280 1.477 0.603 17.38*** Age 23.341 11.088 28.328 12.191 4.987 17.62*** Big4 0.838-0.831-0.007 - Tenure 10.505 5.173 12.062 6.091 1.557 11.31*** Dev_Vrcf 5.868 9.427 5.191 8.708 0.677 3.10*** Divf 0.121 0.208 0.229 0.233 0.108 20.20*** Note: *, **, *** represent significance levels of 10%, 5%, and 1%, respectively. An untabulated Pearson correlation matrix shows that Share and Share BS are not significantly correlated with CAR, suggesting that SOARs or the percentage of audit shared might not affect cumulative abnormal returns. The correlation matrix also reveals that Share and Share BS are significantly correlated with Dev_Vrcf and Divf, respectively, which underscores the need to include Dev_Vrcf and Divf in this study as control variables when 1 In this study, none of the variables appears to have significant outliers; therefore, we do not truncate or winsorize any observations from the analysis. 6

studying, SOARs. Taken together, although these univariate analyses are interesting and informative, we rely on the multivariate analyses that follow in the following sections. 4.2 Perceptions of Investors and SOARs Table 2 reports the results of the relationship between SOARs and perceived audit quality. The first-column result in Table 2 shows that earnings surprise (UE) is significantly positively associated with cumulative abnormal returns (CAR) (coefficient = 0.527, t-statistic = 9.72), in line with prior studies. In Column (II), the coefficient on UE is 0.642 (t-statistic = 7.17). More importantly, in the regression without the control variables, the coefficient on the interaction term of UE Share is negative and significant (coefficient = 0.196, t-statistic = 2.15). Consistent with the hypothesis 1, we find that firms that receive SOARs are associated with a lower ERC, suggesting that investors perceive audit quality to be lower for SOARs. In Column (III), the results of our OLS analysis also show that, after controlling for the effects of firms characteristics and economic conditions, the coefficient of the variable of interest remains significantly negative (coefficient = 0.157, t-statistic = 2.51). On average, the regression estimates highlight that investors discount SOARs by approximately 23.4 percent (= 0.157/0.672) as compared to unqualified audit opinions. This effect survives, albeit its magnitude is smaller, even after replacing Share with Share BS. In Columns (IV) and (V), the coefficients on UE Share BS are significantly negative, indicating that as the percentage of component firms that are not audited by the group engagement partner becomes larger, and investors perceive audit quality as being worse. On average, these results point out that investors discount SOARs by approximately 0.8 percent (= 0.006/0.730) as the percentage of component firms that are not audited by the group engagement partner increases by an additional percent. Table 2: Earnings response coefficients and investors perceptions of audit quality Dependent variable= CAR A B C D E UE 0.527 0.642 0.672 0.586 0.730 (9.72)*** (7.17)*** (3.33)*** (9.29)*** (3.73)*** Share 0.008 0.018 ( 0.97) (2.20)** UE Share 0.196 0.157 ( 2.15)** ( 2.51)** Share BS 0.000 0.001 ( 0.50) (1.42) UE Share BS 0.005 0.006 ( 3.26)*** ( 4.46)*** Beta 0.011 0.008 ( 0.56) ( 0.41) UE Beta 0.021 0.031 (0.17) (0.26) Growth 0.085 0.085 (8.96)*** (8.98)*** UE Growth 0.083 0.085 7

( 2.44)** ( 2.50)** Lev 0.131 0.130 ( 4.66)*** ( 4.60)*** UE Lev 0.626 0.570 ( 3.46)*** ( 3.75)*** Loss 0.159 0.160 ( 12.76)*** ( 12.82)*** UE Loss 0.488 0.523 ( 5.83)*** ( 7.00)*** Size 0.048 0.047 ( 10.17)*** ( 10.00)*** UE Size 0.101 0.092 (3.24)*** (3.13)*** Age 0.003 0.003 (6.13)*** (6.34)*** UE Age 0.006 0.007 ( 2.15)** ( 2.43)** Big4 0.017 0.018 (1.55) (1.61) UE Big4 0.031 0.038 (0.41) (0.54) Tenure 0.003 0.003 (3.97)*** (4.14)*** UE Tenure 0.009 0.014 ( 1.62) ( 2.51)** Dev_Vrcf 0.001 0.001 (2.81)*** (2.83)*** UE Dev_Vrcf 0.000 0.000 ( 0.03) ( 0.06) Divf 0.063 0.056 (3.13)*** (2.79)*** UE Divf 0.027 0.141 (0.22) (1.49) N 6,895 Year and industry Included Included Included Included Included fixed-effects Ad-R 2 0.1584 0.1613 0.2701 0.1620 0.2716 Note: *, **, *** represent significance levels of 10%, 5%, and 1%, respectively. Refer to Table 3 for definitions of the variables. All coefficients are adusted using one-way cluster method [19]. To conserve space, we do not report the intercepts and the individual coefficients on the year and industry fixed-effect variables. 5. Conclusions and Recommendations Business groups are prevalent in Taiwan and the number of Taiwanese firms receiving unqualified audit opinions with explanatory language reaches 65.97 percent, of which 43.08 percent are SOARs (the most common audit opinions issued in Taiwan) over the period 2000 2010. Using the unique audit data from Taiwan, we analyze the effect of SOARs on investors perceptions of audit quality. The focus on a market-based approach enables us to incorporate capital market perceptions of audit quality. Using earnings response coefficients from earnings-returns regressions to proxy for capital market perceptions of audit quality [5-7], we find evidence consistent with our prediction that investors perceive the audit quality of SOARs as lower than that of unqualified audit opinions. We also find that the negative 8

perceptions of the audit quality of SOARs get larger when the percentage of component firms that are not audited by the group engagement partner becomes greater. In addition, we find that the negative perceptions are more pronounced in the samples of Big4 audit firms and controlling interests. These results are robust to examining SOAR-specific samples. Finally, as an additional robustness check, we use an alternative model suggested by Easton and Harris [20]. The results are very similar to our main tests. References Ali A, Chen T-Y, Radhakrishnan S. Corporate disclosures by family firms. Journal of Accounting and Economics. 2007;44:238-86. Benston GJ. The market for public accounting services: Demand, supply and regulation. Journal of Accounting and Public Policy. 1985;4:33-79. Branson J, Breesch D. Referral as a determining factor for changing auditors in the Belgian auditing market: An empirical study. The International Journal of Accounting. 2004;39:307-26. Butler M, Leone AJ, Willenborg M. An empirical analysis of auditor reporting and its association with abnormal accruals. Journal of Accounting and Economics. 2004;37:139-65. Chang R-D, Fang CJ, Shen WH. Earnings Management of the Affiliated Business Groups: The Effect of Whether Being Audited by the Same CPA Firm. Management Review. 2008;27:29-52 (Written in Chinese). Chi W, Huang H, Liao Y, Xie H. Mandatory audit-partner rotation, audit quality and market perception: Evidence from Taiwan. Contemporary Accounting Research. 2009;26:359-91. Collins DW, Kothari SP. An analysis of intertemporal and cross-sectional determinants of earnings response coefficients. Journal of Accounting and Economics. 1989;11:143-81. Davison AG, Stening BW, Wai WT. Auditor concentration and the impact of interlocking directorates. Journal of Accounting Research. 1984;22:313-7. Dechow PM, Dichev ID. The quality of accruals and earnings: The role of accrual estimation errors. The Accounting Review. 2002;77:35-59. Easton PD, Harris TS. Earnings as an explanatory variable for returns. Journal of Accounting Research. 1991;29:19-36. Easton PD, Zmiewski ME. Cross-sectional variation in the stock market response to accounting earnings announcements. Journal of Accounting and Economics. 1989;11:117-41. Ghosh A, Moon D. Auditor tenure and perceptions of audit quality. The Accounting Review. 2005;80:585-612. Gul FA, Tsui J, Dhaliwal DS. Non-audit services, auditor quality and the value relevance of earnings. Accounting & Finance. 2006;46:797-817. 9

Hayn C. The information content of losses. Journal of Accounting and Economics. 1995;20:125-53. Jiang J-X, Yeh H-M. Corporate Diversification, Shared Audit Opinion, and Accruals Quality. International Journal of Accounting Studies. 2008;46:67-105. Johansen TR, Pettersson K. The impact of board interlocks on auditor choice and audit fees. Corporate Governance: An International Review. 2013;21:287-310. Kormendi R, Lipe R. Earnings innovations, earnings persistence, and stock returns. The Journal of Business. 1987;60:323-45. Lim C-Y, Tan H-T. Non-audit service fees and audit quality: The impact of auditor specialization. Journal of Accounting Research. 2008;46:199-246. Rogers WH. Regression standard errors in clustered samples. Stata Technical Bulletin. 1993;13:19-23. Teoh SH, Wong TJ. Perceived auditor quality and the earnings response coefficient. The Accounting Review. 1993;68:346-66. 10