Joint Effect of Investor Protection & Big 4 Audits on Earnings Quality (PDF)

The Joint Effect of Investor Protection and Big 4 Audits on Earnings Quality around the World* JERE R. FRANCIS, University of Missouri–Columbia DECHUN WANG, University of Nebraska–Lincoln 1. Introduction We examine whether earnings quality is jointly affected by the investor protection environment where a firm is located and the firm’s choice of a Big 4 versus non– Big 4 auditor. At issue is whether there are any differences in earnings quality in countries around the world due solely to differences in investor protection regimes, or if differential audit quality measured by the well-known Big 4 / non – Big 4 dichotomy plays a role in mediating how investor protection regimes affect earnings quality.1 The role of auditing is to enforce the application of proper accounting polices. Managers prefer discretion in the reporting process, and auditors may go along with earnings management behavior and the reporting of low quality earn- ings in order to avoid dismissal by clients. However, auditor incentives change as investor protection regimes become stricter, and there is a greater likelihood that client misreporting is detected and auditors are punished. Our conjecture is that Big 4 auditors are more sensitive to the cost of client misreporting and its effect on audi- tor reputation and are more likely to enforce higher earnings quality as investor protection regimes become stronger. In contrast, non–Big 4 auditors are less affected because they have less reputation capital at risk and therefore are less likely to risk client dismissal by enforcing a higher level of earnings quality. Three properties of earnings that have been widely used in prior earnings quality studies are investigated: the magnitude of signed abnormal accruals (Frankel, Johnson, and Nelson 2002); the likelihood of reporting a loss (Burgstahler and Dichev 1997); and earnings conservatism using the timely loss recognition framework of Basu 1997 and Ball, Kothari, and Robin 2000. All three earnings measures capture aspects of accounting conservatism in the sense that earnings are implicitly more conservative ceteris paribus if losses are reported, if signed abnor- mal accruals are income-decreasing, and if earnings are conservative using the Basu 1997 framework. For a large sample of firms from 42 countries over the period 1994–2004 we find that earnings quality is higher as the country’s investor protection * Accepted by Gordon Richardson. We thank the two reviewers and especially Gordon Richardson (editor) for their constructive suggestions. We also appreciate comments on earlier versions of the paper from workshop participants at Aston Business School (Birmingham), Hong Kong Polytechnic University, University of Houston, University of Melbourne, University of Missouri, University of Paris Dauphine, University of Toronto, Vanderbilt University, University of Wisconsin, European Auditing Research Network Symposium, and the annual meeting of the European Accounting Association. Jere Francis is an Honorary Professional Fellow at University of Melbourne, Australia. Contemporary Accounting Research Vol. 25 No. 1 (Spring 2008) pp. 157–91 © CAAA doi:10.1506/car.25.1.6 158 Contemporary Accounting Research regime becomes stronger, but only for firms with Big 4 auditors. Specifically, signed abnormal accruals are smaller (income-decreasing) and the likelihood of a loss is greater as the investor protection environment becomes stronger. In contrast, for clients of non–Big 4 auditors, abnormal accruals and the likelihood of report- ing a loss are unaffected by differences in investor protection regimes. Tests using the Basu 1997 framework are consistent with these results and show that earnings conservatism is increasing in the strictness of a country’s investor protection envi- ronment, but only for firms audited by Big 4 auditors. We conclude that the role of investor protection on earnings quality around the world is mediated by the incen- tives of Big 4 auditors to enforce higher earnings quality as investor protection regimes become stricter. Prior research documents greater financial transparency in countries with stronger investor protection regimes (Bhattacharya, Daouk, and Welker 2003; Bushman, Piotroski, and Smith 2004), and there is evidence that earnings are less managed and more value-relevant in these countries (for example, Ball et al. 2000; Hung 2000; Leuz, Nanda, and Wysocki 2003). Our study makes a distinctive con- tribution to the comparative accounting literature by showing that stronger investor protection regimes per se do not appear to affect the properties of accounting earn- ings without also considering the quality of enforcement by Big 4 and non–Big 4 auditors. The remainder of the paper is organized as follows. The role of auditing and investor protection on earnings quality is further developed in the next section. Investor protection variables are defined in section 3. The sample and models are presented in section 4. Primary results are reported in section 5, and sensitivity anal- yses and robustness tests are reported in section 6. The study concludes in section 7. 2. The role of auditing and investor protection on earnings quality There is evidence that earnings of U.S. companies with Big 4 auditors are of higher quality and that the stock market values earnings surprises of Big 4 clients more highly than earnings surprises of firms with non – Big 4 auditors (Teoh and Wong 1993; Krishnan 2003a). An explanation for this is that Big 4 clients have smaller abnormal accruals, which is consistent with Big 4 auditors constraining aggressive earnings management, thereby resulting in more credible earnings announcements (Becker, DeFond, Jiambalvo, and Subramanyam 1998; Francis, Maydew, and Sparks 1999; Krishnan 2003b). Another reason investors have greater confidence in the reported earnings of Big 4 clients is that Big 4 auditors are more likely to issue going-concern warnings than non–Big 4 auditors for the same set of client circumstances (Francis and Krishnan 1999, 2002). Why do earnings properties of U.S. companies differ systematically for com- panies with Big 4 and non–Big 4 auditors? The standard explanation is that Big 4 auditors in the United States impose a high level of earnings quality in order to protect their brand name reputation from legal exposure and reputation risk, which can arise from misleading financial reports by clients and, in particular, from overly optimistic earnings reports (DeAngelo 1981). If this explanation is correct then we should observe similar outcomes in other countries with strong investor CAR Vol. 25 No. 1 (Spring 2008) The Effect of Investor Protection and Big 4 Audits on Earnings Quality 159 protection (Newman, Patterson, and Smith 2005). Therefore the research question we investigate is whether the earnings properties of Big 4 clients observed in the United States also exist in other countries. There are three alternative scenarios regarding Big 4 behavior around the world. The first possibility is that Big 4 behavior in the United States is unique and stems from what the accounting profession has called the excessive litigation envi- ronment faced by U.S. auditors (Arthur Andersen, Coopers & Lybrand, Deloitte Touche, Ernst & Young, KPMG Peat Marwick, and Price Waterhouse 1992). The profession used these claims in the early 1990s to lobby successfully for litigation relief and the passage of the Private Securities Litigation Reform Act of 1995. Under this viewpoint, one would not expect to observe in other countries the kind of Big 4 behavior that occurs in the United States. In other words, the U.S. legal environment is an extreme outlier and the risk-management behavior observed in the United States leads to a unique level of earnings quality and accounting conser- vatism relative to other countries. A second viewpoint is that Big 4 accounting firms are international organiza- tions with global operations and, therefore, have incentives to develop and maintain uniform reputations around the world (Simunic and Stein 1987). This is achieved through standardized staff training and knowledge-sharing practices, and the global application of uniform audit methodologies. Under this perspective one would expect to observe consistent Big 4 behavior in the treatment of their clients around the world with respect to earnings quality and accounting conservatism. While plausible, we believe the above two scenarios are less likely than a third perspective that draws on recent research documenting cross-country differences in legal institutions and investor protection, and their effects on accounting prac- tices (La Porta, Lopez-de-Silanes, Shleifer, and Vishney 1998, 2000; La Porta, Lopez-de-Silanes, and Shleifer 2006).2 Under this view, Big 4 behavior with respect to client earnings is neither uniform around the world, nor unique to the United States, but instead varies systematically with incentives in different institu- tional environments.3 We conjecture that Big 4 auditors impose higher earnings quality through greater accounting conservatism on clients’ financial reports as a rational response to stricter investor protection regimes, including the ability of investors to sue auditors for negligence and the power of regulators to sanction auditors for misconduct. In contrast, non–Big 4 auditors do not have the same repu- tation capital at risk as Big 4 firms and therefore do not have as strong an incentive to enforce higher earnings quality and risk dismissal by clients. To elaborate, we assume that clients prefer auditors who will allow some dis- cretion in the reporting of earnings and that auditors are willing to go along with such behavior to some degree in order to avoid dismissal. However, non – Big 4 auditors have less to lose than Big 4 auditors in accommodating clients and signing off on earnings that are of inherently lower quality, even in countries with stricter investor protection regimes where low earnings quality may be detected and pun- ished. In other words, the cost–benefit calculus is such that a non–Big 4 auditor has more to gain by appeasing clients on questionable accounting policies, while a Big 4 auditor has more reputation capital at risk and therefore is less likely to go CAR Vol. 25 No. 1 (Spring 2008) 160 Contemporary Accounting Research along with clients as the investor protection regimes becomes stricter. It follows that when investor protection is very low, the incentives for both groups of auditors are similar, in which case there may be no observable differences in earnings quality between Big 4 and non–Big 4 clienteles. Before proceeding, we contrast our study with a concurrent paper by Choi, Kim, Liu, and Simunic 2008. Choi et al. report evidence from 13 countries that Big 4 audit fees are higher than the fees of non–Big 4 auditors around the world, but the Big 4 premium decreases relative to non – Big 4 firms as legal regimes become stronger (stricter) and create greater litigation risk for auditors. Although their study only examines audit fees, higher fees imply higher quality audits, so at face value, the results suggest that Big 4 audits are clearly of higher quality relative to non–Big 4 audits in “weak” legal regimes; however, in “stronger” regimes there is less of a quality differentiation between the two groups of auditors. Thus the results in Choi et al. 2008 appear to be the opposite of our study with respect to Big 4 audit quality. Specifically, we find no differences in the quality of client earnings of Big 4 and non – Big 4 auditors in weak legal regimes, whereas the earnings of Big 4 clients are increasing in quality (more conservative) relative to non – Big 4 auditors as legal regimes become stronger. Thus our results imply that the “gap” in audit quality between Big 4 and non–Big 4 firms increases rather than decreases with the strictness of legal regimes. Even though the two studies seem to reach opposite conclusions, it is important to note that we examine client earnings characteristics, while Choi et al. (2008) examine audit fees, so the two studies are not directly comparable. Further, neither study directly observes audit effort or the auditor’s judgement and decision-making process. The contrasting implications of the two studies indicate the need for further investigation to better understand the effects of legal regimes and investor protection environments on auditing practices around the world.4 3. Investor protection variables Our study examines whether the quality of reported earnings improves as a country’s investor protection environment becomes stronger. Lower earnings quality is less likely to occur in countries with stronger investor protection because enforcement is better in such countries and there are greater consequences for auditors whose clients misreport in terms of civil and criminal liability and other punishment and sanctions imposed by regulatory agencies. We use multiple investor protection measures because there are multiple dimensions to the concept of investor protec- tion and because country-level metrics are likely to have measurement error. The testing of multiple variables is common in cross-country research and gives greater confidence when results are consistent across variables. La Porta et al. (2006) articulate a general framework in which investor protec- tion operates through a county’s legal tradition, corporate law, and securities law. A country’s underlying legal tradition is the foundation that defines basic legal rights, including the protection of property rights, and is also the lens through which cor- porate law and securities law are developed. Legal scholars classify legal traditions into two general families, common law and civil or code law (David and Brierly CAR Vol. 25 No. 1 (Spring 2008) The Effect of Investor Protection and Big 4 Audits on Earnings Quality 161 1985). England developed the common-law tradition that is characterized by rela- tively less reliance on statutes and a preference for contracts and private litigation to resolve disputes. In contrast, the civil- or code-law tradition is associated with France and other European countries and is characterized by greater reliance on explicit laws and procedural codes and a preference for state regulation over private litigation to resolve disputes. Prior research shows that the common-law legal tradition provides greater investor protection than does the civil- (code-) law tradi- tion because of its stronger orientation to private contracting and the protection of private property rights (La Porta et al. 1998). Therefore our first investor protection variable is LAW and is coded one for countries with a common-law legal tradition. Wingate (1997) reports anecdotal evi- dence based on insurers’ malpractice risk assessments that auditors have greater legal exposure in common-law countries than in code-law countries for breach of contract and the tort of negligence if they fail to detect misreporting by clients. Consistent with this view, our prediction is that litigation risk in common-law countries will have a greater effect on Big 4 auditors because of their reputation capital, and this creates an incentive for greater care in audits and the enforcement of higher earnings quality. A second level of investor protection comes explicitly from corporate law and, in particular, those mechanisms in corporate law that protect the rights of outside (minority) investors and attenuate agency problems between inside (controlling) owners and outside / minority owners. La Porta et al. (1998) develop what they term an antidirector rights’ index based on the presence / absence of six specific elements of investor protection in a country’s corporate law practices. The six- point index (ANTI_DIR ) measures how easily outside/minority stockholders can exercise their rights against opportunistic behavior by managers and dominant owners.5 When minority shareholders have greater legal recourse against opportun- istic behavior by majority owners, Big 4 auditors have incentives for a higher standard of care in order to avoid the misreporting of earnings by clients. The next three variables are based on a country’s securities laws such as the Securities Act of 1933 and the Securities and Exchange Act of 1934 in the United States. La Porta et al. (2006) worked with leading securities law attorneys around the world and distilled the protection provided by securities laws to three fundamen- tal factors: disclosure requirements, liability standards, and public enforcement. On the basis of these consultations, La Porta et al. (2006) develop indices that mea- sure a country’s disclosure level, liability standard, and public enforcement of securities law. By construction, the indices are continuous variables scaled from zero to one with larger values indicating countries with stronger investor protection regimes. The disclosure index ( DIS_REQ ) measures the extent to which there is required disclosure of information for firms issuing securities through a prospectus, including information on the compensation of executives, shareholder ownership structure, inside ownership, unusual contracts, and related-party transactions. More disclosure creates greater protection for investors by reducing information asymmetry. The liability index (LIT_STD ) measures the liability standard for CAR Vol. 25 No. 1 (Spring 2008) 162 Contemporary Accounting Research investors to recover damages from issuers of securities, company directors, invest- ment banks, and auditors when there has been misleading disclosures in the issu- ance of securities. La Porta et al. (2006) view the liability standard as a measure of the effectiveness of private enforcement through contract law and the burden of proof required to establish damages when there is malfeasance. The derivation of LIT_STD specifically incorporates the ease with which investors can sue auditors. In addition to “private enforcement”, investors also receive protection through the public enforcement of securities laws by regulatory agencies such as the Securities and Exchange Commission in the United States. La Porta et al. (2006) use the term “supervisor” to refer to the regulatory agency, and the public enforcement index (PUB_ENF ) is based on supervisor characteristics, rule-making powers, investi- gative powers, noncriminal sanctions, and criminal sanctions. Stronger investor protection exists when the supervisor has greater investigative authority and the ability to punish firms and auditors that violate securities laws. As with LIT_STD, the derivation of PUB_ENF specifically incorporates the extent to which auditors can be punished and sanctioned for failing to prevent client misreporting. In sum, as securities laws give greater protection to investors, Big 4 auditors are exposed to greater risks from the consequences of client misreporting and, therefore, are expected to enforce a higher level of earnings quality relative to non–Big 4 auditors. Table 1, panel A, reports values of the five investor protection variables for each of the 42 countries in the study. There are 15 common-law countries and 27 code-law countries in the sample. The United States has the highest level of investor protection for all five variables. Not surprisingly there is relatively high correlation among the five variables in Table 1, panel B. All pair-wise correlations are positive and significant at the 0.01 level or less, and Spearman correlations range in value from 0.379 to 0.647. The common-law/code distinction has been widely used to measure investor protection in prior research. Although viewed as a simplistic dichotomy, it is associated with other more specific measures of investor protection. In other words, countries with a common-law legal tradition also tend to have stronger investor protection through corporate law and securities law. 4. Sample and research design The sample and financial data are obtained from the COMPUSTAT Global Indus- trial and Commercial file for the period 1994–2004. Stock price and earnings per share data are retrieved from the COMPUSTAT Global Issue file for the same period. We exclude firm-year observations with nonfully consolidated financial statements, those not audited, and those with missing values for the dependent and independent variables. Next we keep only those observations in countries with investor legal protection measures for the 49 countries surveyed in La Porta 1998, 2006. We further delete observations from Japan, South Korea, India, and Pakistan because of potential miscoding of the auditor identification variable. These coun- tries report Big 4 auditor rates of close to zero due to the Big 4 practice of operating in these countries through the name of a local affiliate and therefore we have no precise way of determining which firms are audited by Big 4 auditors. Financial institutions (Standard Industrial Classification [SIC] 6000–6999) are also excluded. CAR Vol. 25 No. 1 (Spring 2008) TABLE 1 Investor protection in common-law and code-law countries Panel A: Measures of investor protection and Big 4 auditors’ market share for the 42 countries in the study No. of No. of No. of observations observations observations in abnormal in loss in earnings Big 4 accruals avoidance conservatism market Country analysis analysis analysis share LAW ANTI_DIR DIS_REQ LIT_STD PUB_ENF Argentina 83 216 161 74.5% 0 4 0.50 0.22 0.58 Australia 1,943 2,514 2,225 81.6% 1 4 0.75 0.66 0.90 Austria 405 563 409 49.2% 0 2 0.25 0.11 0.17 Belgium 516 748 503 66.2% 0 0 0.42 0.44 0.15 Brazil 569 935 664 88.3% 0 3 0.25 0.33 0.58 Canada 2,989 4,240 3,872 92.4% 1 5 0.92 1 0.80 Chile 492 610 520 88.0% 0 5 0.58 0.33 0.60 Colombia 59 94 70 50.0% 0 3 0.42 0.11 0.58 Denmark 830 1,079 797 92.0% 0 2 0.58 0.55 0.37 Egypt 10 25 27 40.0% 0 2 0.50 0.22 0.30 Finland 675 857 681 76.2% 0 3 0.50 0.66 0.32 France 3,220 4,400 3,544 50.1% 0 3 0.75 0.22 0.77 Germany 2,941 4,253 3,146 47.5% 0 1 0.42 0 0.22 Greece 321 425 290 36.0% 0 2 0.33 0.50 0.32 Hong Kong 776 987 784 87.5% 1 5 0.92 0.66 0.87 Indonesia 1,132 1,555 985 46.2% 0 2 0.50 0.66 0.62 Ireland 318 405 242 89.4% 1 4 0.67 0.44 0.37 Israel 152 283 224 51.9% 1 3 0.67 0.66 0.63 The Effect of Investor Protection and Big 4 Audits on Earnings Quality Italy 746 1,339 1,063 92.5% 0 1 0.67 0.22 0.48 Jordan 3 5 7 80.0% 0 1 0.67 0.22 0.60 (The table is continued on the next page.) CAR Vol. 25 No. 1 (Spring 2008) 163 TABLE 1 (Continued) 164 No. of No. of No. of observations observations observations in abnormal in loss in earnings Big 4 accruals avoidance conservatism market Country analysis analysis analysis share LAW ANTI_DIR DIS_REQ LIT_STD PUB_ENF Kenya 6 7 0 100.0% 1 3 0.50 0.44 0.70 Malaysia 3,766 4,829 3,704 65.8% 1 4 0.92 0.66 0.77 Mexico 389 561 375 76.1% 0 1 0.58 0.11 0.35 Netherlands 1,101 1,495 1,245 92.1% 0 2 0.50 0.89 0.47 CAR Vol. 25 No. 1 (Spring 2008) New Zealand 358 455 350 92.5% 1 4 0.67 0.44 0.33 Norway 644 948 719 92.4% 0 4 0.58 0.39 0.32 Peru 80 110 89 72.7% 0 3 0.33 0.66 0.78 Philippines 500 653 498 25.0% 0 3 0.83 1 0.83 Portugal 247 321 220 40.8% 0 3 0.42 0.66 0.58 Singapore 2,015 2,660 2,088 86.6% 1 4 1 0.66 0.87 Contemporary Accounting Research South Africa 454 683 621 86.1% 1 5 0.83 0.66 0.25 Spain 765 947 737 92.7% 0 4 0.50 0.66 0.33 Sri Lanka 13 17 21 88.2% 1 3 0.75 0.39 0.43 Sweden 1,428 1,890 1,413 84.4% 0 3 0.58 0.28 0.50 Switzerland 1,064 1,428 1,104 76.0% 0 2 0.66 0.44 0.33 Taiwan 928 1,199 1,022 77.7% 0 3 0.75 0.66 0.52 Thailand 1,385 2,088 1,336 40.1% 1 2 0.92 0.22 0.72 Turkey 116 212 163 64.6% 0 2 0.50 0.22 0.63 United Kingdom 7,287 9,215 7,399 81.6% 1 5 0.83 0.66 0.68 United States 17,184 29,852 24,791 93.1% 1 5 1 1 0.90 Venezuela 40 63 43 93.7% 0 1 0.17 0.22 0.55 Zimbabwe 16 27 15 96.3% 1 3 0.50 0.44 0.42 (The table is continued on the next page.) The Effect of Investor Protection and Big 4 Audits on Earnings Quality 165 TABLE 1 (Continued) Panel B: Spearman correlations of investor protection and Big 4 market share (n 42) Big 4 market ANTI_DIR DIS_REQ LIT_STD PUB_ENF share LAW 0.574 0.647 0.379 0.389 0.338 (0.01) (0.01) (0.01) (0.01) (0.03) ANTI_DIR 0.519 0.515 0.402 0.341 (0.01) (0.01) (0.01) (0.03) DIS_REQ 0.408 0.504 0.154 (0.01) (0.01) (0.33) LIT_STD 0.381 0.172 (0.01) (0.28) PUB_ENF 0.023 (0.88) Notes: p-values are in parentheses. LAW equals 1 if a country has a common-law tradition and 0 otherwise. ANTI_DIR is the antidirector rights’ index from La Porta et al. 1998. DIS_REQ is the disclosure requirements index from La Porta et al. (2006). LIT_STD is the liability standard index from La Porta et al. 2006. PUB_ENF is the public enforcement index from La Porta et al. 2006. Finally, we exclude observations that fall in the top and bottom 1 percent of abnor- mal accruals, those with the absolute value of studentized residuals greater than three in the abnormal accruals analysis, those in the top and bottom 1 percent of annual returns and earnings per share before extraordinary items, and those with the absolute value of studentized residuals greater than three in the accounting con- servatism analysis. After these screens, there are 57,966 observations for the period 1996 – 2004 in the abnormal accruals analysis, 85,193 observations for the period 1995 – 2004 in the loss avoidance analysis, and 68,167 observations for the period 1995–2004 in the earnings conservatism analysis.6 The sample selec- tion process is summarized in Table 2, and details of the three samples and variables used in each of the three tests are reported in Table 3. The number of firm-year observations for each of the 42 countries is reported in panel A of Table 1 for each of the three analyses in the study. Eight countries have less than 100 firm-year observations, 20 countries have from 100 to 1,000 firm-year observations, and 14 countries have more than 1,000 firm-year observa- tions. A sensitivity analysis shows that the tests are robust to the exclusion of smaller countries. Country-level Big 4 market shares are also reported (Table 1) and there is a wide range from 25 percent in the Philippines to 100 percent for Kenya (6 observations). Overall, the evidence is mixed with respect to the impact of investor protection on Big 4 market shares across countries. Table 1, panel B shows that Big 4 market share is positively related to legal tradition (LAW ) and to CAR Vol. 25 No. 1 (Spring 2008) TABLE 2 166 Sample selection Panel A: Abnormal accruals analysis No. of observations with no missing values on dependent and independent variables for 1996–2004 81,614 Less no. of observations from countries not on the list of the 49 countries in La Porta et al. 2006 (3,798) Less no. of observations from Japan, South Korean, India, and Pakistan (16,095) Less no. of financial institutions (SIC 6000–6999) (1,234) Less no. of top and bottom 1% of abnormal accruals (1,208) Less no. of observations with studentized residuals 3 (1,313) CAR Vol. 25 No. 1 (Spring 2008) Final no. of observations used in the abnormal accruals tests 57,966 Panel B: Loss avoidance analysis No. of observations with no missing values on dependent and independent variables for 1995–2004 117,825 Less no. of observations from countries not on the list of the 49 countries in La Porta et al. 2006 (5,466) Contemporary Accounting Research Less no. of observations from Japan, South Korean, India, and Pakistan (24,979) Less no. of financial institutions (SIC 6000–6999) (2,187) Final no. of observations used in the loss avoidance tests 85,193 Panel C: Earnings conservatism analysis No. of observations with no missing values on dependent and independent variables for 1995–2004 103,501 Less no. of observations from countries not on the list of the 49 countries in La Porta et al. 2006 (4,007) Less no. of observations from Japan, South Korean, India, and Pakistan (24,597) Less no. of financial institutions (SIC 6000–6999) (1,958) Less no. of top and bottom 1% of earnings per share before extraordinary items and annual returns (2,781) Less no. of observations with studentized residuals 3 (1,991) Final no. of observations used in the earnings conservatism tests 68,167 The Effect of Investor Protection and Big 4 Audits on Earnings Quality 167 antidirectors’ right index (ANTI_DIR), which means that we observe more Big 4 audits when investor protection is stronger as measured by these two variables. However, none of the three securities law variables ( DIS_REQ , LIT_STD , PUB_ENF ) is associated with Big 4 market share. Signed abnormal accruals analysis The first analysis tests whether signed abnormal accruals differ across countries as a function of investor protection regimes and whether there is a mediating Big 4 auditor effect. Larger abnormal (unexpected) accruals imply greater managerial opportunism and earnings of lower quality. We conjecture there are smaller abnormal TABLE 3 Descriptive statistics Panel A: Abnormal accruals tests (n 57,966) 25th 75th Variables Mean s.d. percentile Median percentile AB_ACCR 0.011 0.124 0.070 0.010 0.046 LSALES 5.357 2.010 4.067 5.367 6.668 CFO 0.067 0.166 0.012 0.082 0.149 LEV 0.541 0.252 0.369 0.540 0.689 GROWTH 0.126 0.386 0.050 0.069 0.217 PPE 0.132 0.356 0.013 0.071 0.195 LAG_LOSS 0.256 0.436 0 0 1 Panel B: Loss avoidance tests (n 85,193) 25th 75th Variables Mean s.d. percentile Median percentile LOSS 0.274 0.446 0 0 1 LSALES 5.355 2.106 3.990 5.376 6.751 LEV 0.551 0.266 0.374 0.548 0.699 GROWTH 0.201 0.636 0.043 0.081 0.244 Panel C: Earnings conservatism tests (n 68,167) 25th 75th Variables Mean s.d. percentile Median percentile EARN 0.046 0.160 0.000 0.052 0.103 R 0.101 0.583 0.252 0.020 0.324 DR 0.479 0.500 0 0 1 LMV 5.439 2.090 3.959 5.379 6.749 LEV 0.513 0.213 0.362 0.530 0.669 MB 3.104 5.844 0.983 1.681 2.967 (The table is continued on the next page.) CAR Vol. 25 No. 1 (Spring 2008) 168 Contemporary Accounting Research TABLE 3 (Continued) Notes: AB_ACCR is the signed abnormal accruals. LSALES is the natural log of client sales. CFO is the operating cash flows scaled by lagged total assets. LEV is the ratio of total liabilities to total assets. GROWTH is the sales growth rate, defined as the sales in current year minus sales in prior year and divided by sales in prior year. PPE is the growth rate of gross PPE (property, plant, and equipment), defined as the gross PPE in current year minus the gross PPE in prior year and divided by the gross PPE in prior year. LAG_LOSS equals 1 if net income before extraordinary items in the prior year is negative and 0 otherwise. LOSS equals 1 if net income before extraordinary items in the current year is negative and 0 otherwise. EARN is defined as earnings per share before extraordinary items, scaled by stock price at beginning of the period. R is the cumulative stock return for the fiscal year. DR is a dummy variable and equals 1 if R is negative and 0 otherwise. LMV is the natural log of market value of equity. MB is the market-to-book ratio. accruals in countries with stronger investor protection regimes because the conse- quences of misreporting earnings is greater for firms and their auditors in these countries, and Big 4 auditors are expected to be more sensitive to these conse- quences than non–Big 4 auditors. Signed abnormal accruals are used rather than absolute (unsigned) abnormal accruals for two reasons. First, we are interested pri- marily in the use of managerial discretion to increase reported earnings because this is the misreporting scenario most likely to damage an auditor’s reputation. Second, Hribar and Nichols (2007) report evidence that signed abnormal accruals are a better measure of earnings quality than the absolute or unsigned value of abnormal accruals. A cross-sectional Jones 1991 model is not practical for the calculation of abnormal accruals with international data. The reason is that the number of indus- try observations per country can be quite small, and this may explain, at least in part, why Jones-type abnormal accruals perform unreliably in international set- tings (Wysocki 2004; Meuwissen, Moers, Peek, and Vanstraelen 2005). We avoid this problem by using a linear expectation model adapted from DeFond and Park 2001 that uses a firm’s own prior year accruals in calculating the expectation benchmark. Specifically, expected accruals are based on a firm’s prior year ratio of current accruals to sales, and the prior year’s ratio of deprecation expense to gross property, plant, and equipment (PPE). Another benefit of this approach is that it implicitly controls for cross-country differences in accounting standards by using a firm as its own control to compute abnormal accruals. Therefore abnormal accruals are contextualized relative to the specific accounting standards of a particular country. To illustrate, assume that the ratio of current accruals to sales is 0.15 for a firm in year t 1, based on sales of $100 and current accruals of $15 in t 1. If sales in year t are $120, then predicted current accruals in year t will be $120 0.15 $18. The same procedure is used for predicted depreciation expense, which is based on the prior year’s ratio of depreciation expense to gross PPE. CAR Vol. 25 No. 1 (Spring 2008) The Effect of Investor Protection and Big 4 Audits on Earnings Quality 169 Importantly, note that this model is not a random walk expectation model in which current accruals are simply expected to be the same dollar amount as last year’s accruals. Rather, accruals are assumed to have a constant linear relationship over time with sales (for current accruals) and gross PPE (for depreciation) that can be used to predict current period accruals for a given level of sales and gross PPE.7 Using data from COMPUSTAT Global Industrial and Commercial file, pre- dicted accruals are calculated as: Predicted accruals {[Sales (#1) in year t (current accruals in year t 1/sales in year t 1)] [gross PPE (#77) in year t (depreciation in year t 1 (#11 #13)/gross PPE in year t 1 (#77)]}/total assets (#89) in year t 1.8 Abnormal accruals are defined as the firm’s actual total accruals in year t, minus predicted total accruals for year t as defined above. Total accruals in year t are calculated as follows using data from the COMPUSTAT Global Industrial and Commercial file: Total accruals (Earnings before extraordinary items Operating cash flows)/ total assets (#89) in year t 1 where: Earnings before extraordinary items net income (#32) extraordinary items (#33); Operating cash flows9 Earnings before extraordinary items (as above) Depreciation and Amortization (#11) change of deferred income tax (#105) change of untaxed reserve (#108) change in other liabilities (#109) minority interest (#27) current accruals (as defined below). Current accruals change in non-cash working capital [total current assets (#75) cash and short term investments (#60) treasury stock shown as current assets (#73)] [total current liabilities (#104) total amount of debt in current liabilities (#94) proposed dividends (#102)]. Abnormal accruals are calculated each year for 57,966 firm-year observations using data from 1994–2004. Because data for three consecutive years are needed to calculate abnormal accruals, 1996 is the first observation year in the accruals analysis sample. Abnormal accruals are scaled by a firm’s lagged total assets, and CAR Vol. 25 No. 1 (Spring 2008) 170 Contemporary Accounting Research the mean (median) sample value is 0.011 (0.010). The 25th percentile value of abnormal accruals is 0.070, and the 75th percentile value is 0.046. A total of 44 percent of the sample has positive (income-increasing) abnormal accruals and 56 percent has negative (income-decreasing) abnormal accruals. The model in (1) below tests whether signed abnormal accruals differ around the world as a function of the country’s investor protection environment and the firm’s choice of a Big 4 versus non–Big 4 auditor, plus a set of controls for other factors that may affect accruals: AB_ACCRit 0 1BIG4it 2INVPRO 3BIG4it*INVPRO 4LSALESit 5CFOit 6LEVit 7GROWTHit 8PPEit 9LAG_LOSSit fixed effects eit (1), where: AB_ACCRit signed abnormal accruals scaled by lagged total assets for firm i in year t. BIG4it 1 if firm i is audited by a Big 4 auditor in year t, 0 otherwise. INVPRO proxies of investor protection, measured five ways: 1. LAW 1 for a common-law country and 0 otherwise, 2. ANTI_DIR antidirector rights’ index (La Porta et al. 1998), 3. DIS_REQ index of disclosure requirement (La Porta et al. 2006), 4. LIT_STD index of liability standard (La Porta et al. 2006), 5. PUB_ENF index of public enforcement (La Porta et al. 2006). LSALESit log of client sales in $ millions for firm i in year t. CFOit operating cash flows for firm i in year t scaled by lagged total assets. LEVit total liabilities/total assets for firm i in year t. GROWTHit sales growth rate, defined as the sales in year t minus sales in t 1 and scaled by sales in year t 1. PPEit growth rate of gross property, plant, and equipment (PPE), defined as PPE in year t minus PPE in t 1 and scaled by PPE in t 1. LAG_LOSSit dummy variable 1 if firm i reports negative income before extraordinary items in year t 1. fixed effects industry and year fixed effects. eit error term. CAR Vol. 25 No. 1 (Spring 2008) The Effect of Investor Protection and Big 4 Audits on Earnings Quality 171 We estimate (1) as a fixed effects model with year-specific dummy variables to con- trol for systematic time period effects and industry dummies based on two-digit SIC codes to provide additional controls for omitted variables that could affect firm-level accruals. For brevity, the year and industry dummies are not reported in the tables. Using the procedure in Rogers 1993, (1) is estimated to derive t-statistics and p-values that are robust with respect to heteroscedasticity. In addition, because country-level investor protection variables take on the same value for every firm within a country, it is possible that country effects are overstated due to repeated observations within countries. Therefore we also use the Rogers 1993 procedure to derive robust t-statistics and p-values that control for country clustering effects and the common variance among observations within a particular country (in addition to controlling for heteroscedasticity).10 Investor protection (INVPRO) is measured by the five ways discussed in sec- tion 3, and BIG4 denotes whether a firm is audited by a Big 4 auditor. The primary coefficients of interest are 2 and 3. The investor protection variable alone ( 2) captures the effect of investor protection regimes on firms with non – Big 4 audi- tors, while the coefficient on the interaction term 3 measures the incremental effect of investor protection regimes on Big 4 client accruals relative to the accru- als of non–Big 4 clients, for a given investor protection regime. If 3 is negative and significant, there is evidence that Big 4 clients have higher earnings quality (smaller abnormal accruals) than firms with non–Big 4 auditors as investor protec- tion regimes become stricter.11 The coefficient 1 tests whether accruals of firms with Big 4 auditors are different from non–Big 4 clients when investor protection is effectively zero (that is, extremely weak). We make no prediction about the coef- ficient 1 although it turns out that the variable is insignificant, which means there are no differences in the abnormal accruals of Big 4 and non – Big 4 clienteles when investor protection is extremely weak. The control variables in (1) are intended to control for other firm-specific fac- tors that can affect a firm’s accruals based on prior research (for example, Becker et al. 1998; Frankel et al. 2002). We control for company size, measured by natural log of total sales (LSALES). Prior studies document that large firms tend to have lower levels of accruals than smaller firms, even though accruals are scaled by firm size (lagged total assets). We control for operating cash flows (CFO) deflated by lagged total assets because there is a well-documented inverse relation between the operating cash flows and accruals. The variable for leverage (LEV) controls for the likelihood of bankruptcy, and a higher total debt to asset ratio indicates a higher possibility of debt covenant violation, which creates an incentive to increase reported earnings through accruals-based earnings management. A dummy vari- able is used for firms with prior-year losses (LAG_LOSS) as another proxy for financial distress and bankruptcy risk and therefore an incentive to increase reported earnings in the subsequent year. The final two variables control for firm growth, which could also affect yearly accruals if the relation between accruals and the accruals drivers (sales and gross PPE) is nonlinear. GROWTH is defined as the growth in sales relative to prior year sales, and PPE measures growth in gross PPE over the prior year. CAR Vol. 25 No. 1 (Spring 2008) 172 Contemporary Accounting Research Loss avoidance analysis The second analysis determines whether the likelihood of reporting a loss differs across countries as a function of investor protection and whether there is a mediating Big 4 auditor effect. There is evidence that firms systematically manage earnings to avoid reporting losses (Burgstahler and Dichev 1997; Degeorge, Patel, and Zeckhauser 1999; Brown and Caylor 2005). We conjecture that loss recognition is more likely to occur in countries with stronger investor protection regimes because the consequences of hiding or underreporting losses will be greater for auditors in these countries, and Big 4 auditors are expected to have a relatively greater incen- tive to impose loss reporting on their clients relative to non – Big 4 auditors. Descriptive statistics in Table 3 indicate that 27.4 percent of firm-year observations reported a loss — that is, negative income before extraordinary items. The model in (2) tests whether loss reporting is affected by a country’s inves- tor protection environment and the firm’s choice of a Big 4 versus non – Big 4 auditor, plus a set of controls for size, leverage, and growth that may affect the likelihood of reporting losses, along with fixed effects for year and industry: P(LOSSit 1) 0 1BIG4it 2INVPRO 3BIG4it*INVPRO 4LSALESit 5LEVit 6GROWTHit fixed effects eit (2), where: LOSSit dummy variable 1 if firm i reports negative income before extraordinary items in year t, 0 otherwise. All other variables are as defined in (1), and the same coefficients ( 2 and 3) test the effects of investor protection and Big 4 auditor choice on the likelihood of reporting losses. As with (1), the model is estimated using procedures to derive asymptotic Z-statistics and p-values for coefficients that are robust with respect to heteroscedasticity and country clustering effects (Rogers 1993). Earnings conservatism test The third test uses the earnings conservatism framework of Basu 1997 to deter- mine whether there are differences in timely loss recognition across countries as a function of investor protection regimes, and whether there is a mediating Big 4 auditor effect. The sign of a firm’s annual stock returns is used to indicate whether the firm has experienced “good news” (positive returns) or “bad news” (negative returns) in the current fiscal year. Earnings conservatism exists if contemporaneous accounting earnings give recognition to bad news more quickly than good news. The premise of earnings conservatism is that losses are recognized immediately, while the recognition of gains (good news) is deferred until realized. Watts (2003) and LaFond and Watts (2008) argue that earnings conservatism is the defining fea- ture of high quality earnings and makes earnings more useful for contracting and for reducing information asymmetry between the firm and outside investors. CAR Vol. 25 No. 1 (Spring 2008) The Effect of Investor Protection and Big 4 Audits on Earnings Quality 173 Ball et al. (2000) use Basu’s 1997 framework and report that earnings are rela- tively more conservative in countries with common-law legal systems than in countries with code-law systems for a sample drawn from eight countries. We extend their study by using an expanded set of countries, and by testing whether investor protection effects are mediated by the firm’s choice of a Big 4 auditor. The following model in (3) builds on Basu 1997 as extended by Ball et al. 2000: EARNit 0 1DRit 2Rit 3R*DRit 4BIG4it 5BIG4it*DRit 6BIG4it*Rit 7BIG4it*Rit*DRit 8INVPRO 9INVPRO*DRit 10 INVPRO*Rit 11INVPRO*Rit*DRit 12INVPRO*BIG4it 13INVPRO*BIG4it*DRit 14INVPRO*BIG4it*Rit 15INVPRO*BIG4it*Rit*DRit 16LMVit 17LMVit*DRit 18LMVit*Rit 19LMVit*Rit*DRit 20 LEVit 21LEVit*DRit 22LEVit*Rit 23LEVit*Rit*DRit 24MBit 25MBit*DRit 26MBit*DRit 27MBit*Rit*DRit fixed effects eit (3), where: EARNit earnings per share before extraordinary items scaled by stock price at beginning of year t; Rit cumulative monthly stock return including dividend for the fiscal year t; DRit dummy variable, which equals 1 if Rit is negative and 0 otherwise; LMVit the natural log of market value of equity at the end of year t; LEVit total liabilities/total assets for firm i in year t; MBit market-to-book ratio at the end of year t. Following Ball et al. 2000, Rit is the buy-and-hold monthly stock returns including dividends for the current fiscal year t. Controls (and related interactions) are added for firm size (LMV), leverage (LEV), and growth (MB). Year and industry fixed effects are the same as specified in (1) and (2), and the model is estimated with robust t-statistics and p-values with respect to heteroscedasticity and country- clustering effects (Rogers 1993). The original Basu 1997 model is 0 through 3 in (3) and tests whether reported earnings are more strongly associated with negative contemporaneous stock returns ( 2) than with positive contemporaneous stock returns ( 3). A positive sign on 3 is consistent with accounting earnings giving recognition to bad news (timely loss recognition) more quickly than good news. Coefficients 4 through 15 introduce the effects of investor protection and Big 4 auditor choice, plus related interaction terms. The remaining terms in the model control for the effects of firm size (LMV), CAR Vol. 25 No. 1 (Spring 2008) 174 Contemporary Accounting Research leverage (LEV), growth (MB), and fixed effects for year and industry. The primary coefficients of interest are 11, which measures the effect of investor protection regimes on earnings conservatism of firms with non – Big 4 auditors, and 15 , which is the incremental effect of investor protection on earnings conservatism of firms with Big 4 auditors relative to non–Big 4 clients. As a caveat we note there are critiques of the Basu 1997 earnings conservatism model (for example, Dietrich, Muller, and Reidl 2007; Gigler and Hemmer 2001; Givoly, Hayn, and Natarajan 2007). Given the concerns raised in these critiques, we view the earnings conservatism analysis primarily as a triangulation with the other two tests. That is, if signed abnormal accruals are smaller (less income- increasing) and losses are more likely to occur in stronger investor protection regimes, then such earnings are implicitly consistent with the concept of account- ing conservatism in Basu 1997 wherein reported earnings are lower ceteris paribus due to timely loss recognition. 5. Results Signed abnormal accruals The signed abnormal accruals analysis is reported in Table 4. Five regression models are reported in which each investor protection variable is tested one at a time. All models are significant with adjusted R 2s of around 16 percent. Significance levels of individual coefficients are reported as two-tailed p-values. The investor protection variable by itself represents the effect on accruals of non–Big 4 clients as investor protection becomes stricter. The investor protection variable is insignificant at p 0.10 in all five model estimations. We conclude there is no evidence that abnormal accruals of firms with non–Big 4 auditors are affected by the strictness of a country’s investor protection regime. The interaction of investor protection with the Big 4 variable measures the incremental effect of Big 4 auditors relative to non – Big 4 auditors as investor protection becomes stronger. The interaction term has a negative coefficient and is significant at p 0.02 or less in all models except the model using ANTI_DIR, which is negative and significant at p 0.08. A negative sign indicates that abnormal accruals of Big 4 clients are consistently smaller (less income-increasing) relative to the accruals of non – Big 4 clients as a country’s investor protection regime becomes stricter. We also consider the Big 4 indicator variable by itself, which measures whether accruals of firms with Big 4 auditors are different from accruals of firms with non–Big 4 auditors when investor protection is effectively zero (i.e., extremely weak). The Big 4 variable by itself is insignificant (p 0.10) in all models except the model using DIS_REQ, which has a positive sign and is significant at p 0.07. On balance we conclude there is no consistent evidence that accruals are different for clients of Big 4 versus non–Big 4 auditors when investor protection is extremely weak. In sum, the evidence in Table 4 indicates that abnormal accruals are smaller (less income-increasing) as a country’s investor protection regime becomes stronger. CAR Vol. 25 No. 1 (Spring 2008) TABLE 4 Regression analysis of abnormal accruals (dependent variable is signed abnormal accruals: AB_ACCR) Investor Investor Investor Investor Investor protection protection protection protection protection ANTI_DIR, DIS_REQ, LIT_STD, PUB_ENF, LAW, estimate estimate estimate estimate estimate Independent variables (p-value) (p-value) (p-value) (p-value) (p-value) Intercept 0.011 0.013 0.001 0.011 0.003 (0.06) (0.23) (0.95) (0.21) (0.73) BIG4 0.003 0.002 0.012 0.001 0.007 (0.29) (0.77) (0.07) (0.89) (0.16) Investor Protection 0.005 0.000 0.016 0.005 0.015 (0.30) (0.92) (0.17) (0.62) (0.15) BIG4*Investor Protection 0.009 0.003 0.027 0.014 0.023 (0.01) (0.08) (0.01) (0.02) (0.01) LSALES 0.006 0.007 0.007 0.007 0.007 (0.01) (0.01) (0.01) (0.01) (0.01) CFO 0.322 0.322 0.322 0.322 0.322 (0.01) (0.01) (0.01) (0.01) (0.01) LEV 0.047 0.047 0.047 0.047 0.047 (0.01) (0.01) (0.01) (0.01) (0.01) GROWTH 0.004 0.004 0.004 0.004 0.004 (0.25) (0.24) (0.25) (0.24) (0.25) PPE 0.005 0.005 0.005 0.006 0.005 The Effect of Investor Protection and Big 4 Audits on Earnings Quality (0.38) (0.37) (0.37) (0.36) (0.37) (The table is continued on the next page.) CAR Vol. 25 No. 1 (Spring 2008) 175 TABLE 4 (Continued) 176 Investor Investor Investor Investor Investor protection protection protection protection protection ANTI_DIR, DIS_REQ, LIT_STD, PUB_ENF, LAW, estimate estimate estimate estimate estimate Independent variables (p-value) (p-value) (p-value) (p-value) (p-value) LAG_LOSS 0.014 0.013 0.014 0.013 0.014 (0.010) (0.010) (0.010) (0.010) (0.010) Adj. R 2 0.1623 0.1625 0.1624 0.1624 0.1623 CAR Vol. 25 No. 1 (Spring 2008) n 57,966 57,966 57,966 57,966 57,966 Notes: Coefficient p-values are two-tailed and robust to heteroscedasticity and country clustering effects using the method in Rogers 1993. Coefficients on year dummies and industry dummies based on two-digit SIC codes are not reported for brevity. BIG4 is a dummy variable and equals 1 if a Contemporary Accounting Research company is audited by a Big 4 auditor and 0 otherwise. CFO is the operating cash flows scaled by lagged total assets. All other variables are as defined in Tables 1 and 3. Statistics on test variables are shown in boldface. The Effect of Investor Protection and Big 4 Audits on Earnings Quality 177 However, this effect is mediated by the firm’s choice of auditor, and it turns out that abnormal accruals are smaller only when the auditor is a Big 4 auditor and when investor protection is stronger. For firms with non–Big 4 auditors there is no evidence that abnormal accruals vary across investor protection regimes, nor are there systematic differences in Big 4 and non–Big 4 clienteles when investor pro- tection is extremely weak. Loss avoidance The loss avoidance analysis is reported in Table 5 for five logistic regression models testing each investor protection variable one at a time. All models are significant with pseudo R 2s of around 18 percent. Significance levels of individual coefficients are based on two-tailed p-values for asymptotic z-statistics. The investor protection variable captures the effect of investor protection regimes on firms with non – Big 4 auditors and is insignificant (p 0.10) in all models in Table 5. On the basis of these results we conclude there is no evidence that firms with non – Big 4 auditors are more likely to report losses as country’s investor protection regime becomes stricter. The interaction of investor protection with the Big 4 indicator variable measures the incremental effect of investor protection for Big 4 clients relative to non–Big 4 clients as investor protection becomes stricter. The interaction term is positive and significant at p 0.01 in all models except the model using LAW, which is positive and significant at p 0.07. We conclude that the evidence consistently shows Big 4 clients are more likely to report losses than non – Big 4 clients, as investor protection regimes become stronger. The Big 4 variable by itself determines whether firms with Big 4 auditors are more likely to report losses than firms with non–Big 4 auditors when investor pro- tection is effectively zero (very weak). Overall there is no consistent evidence of differences in the likelihood of reporting losses for firms with Big 4 versus non– Big 4 auditors when investor protection is extremely weak. The model with LAW is significant and positive (p 0.04), while the model with PUB_ENF is negative and significant (p 0.05). The other three models are insignificant at p 0.10. We conclude that Big 4 clients are more likely to report losses than are clients of non–Big 4 auditors, but only as the country’s investor protection environment becomes stricter. There is no evidence that investor protection regimes affect loss reporting by non – Big 4 clients, or that there are systematic differences in Big 4 and non–Big 4 clienteles when investor protection is extremely weak. Earnings conservatism The earnings conservatism analysis is reported in Table 6 with five separate regres- sion models testing the investor protection variables one at a time. All models are significant with adjusted R 2s of around 22 percent, and significance levels of indi- vidual coefficients are reported as two-tailed p-values. The term R*DR is positive and replicates the basic finding in Basu 1997 that earnings are conservative in the sense of reflecting the timelier recognition of bad news (as proxied by negative stock returns) relative to good news (as proxied by positive stock returns). CAR Vol. 25 No. 1 (Spring 2008) TABLE 5 178 Logistic regression analysis of loss avoidance (dependent variable is the probability of reporting a loss: P(Loss 1)) Investor Investor Investor Investor Investor protection protection protection protection protection ANTI_DIR, DIS_REQ, LIT_STD, PUB_ENF, LAW, estimate estimate estimate estimate estimate Independent variables (p-value) (p-value) (p-value) (p-value) (p-value) Intercept 1.225 1.304 1.259 1.336 1.208 (0.01) (0.01) (0.01) (0.01) (0.01) CAR Vol. 25 No. 1 (Spring 2008) BIG4 0.219 0.191 0.355 0.032 0.301 (0.04) (0.12) (0.12) (0.83) (0.05) Investor Protection 0.147 0.037 0.143 0.262 0.097 (0.42) (0.61) (0.75) (0.46) (0.84) BIG4*Investor Protection 0.238 0.146 0.918 0.569 0.999 Contemporary Accounting Research (0.07) (0.01) (0.01) (0.01) (0.01) LSALES 0.470 0.480 0.474 0.485 0.476 (0.01) (0.01) (0.01) (0.01) (0.01) LEV 2.522 2.548 2.531 2.555 2.53 (0.01) (0.01) (0.01) (0.01) (0.01) GROWTH 0.016 0.012 0.013 0.009 0.014 (0.84) (0.88) (0.86) (0.90) (0.85) Pseudo-R 2 0.180 0.182 0.181 0.183 0.182 n 85,193 85,193 85,193 85,193 85,193 (The table is continued on the next page.) The Effect of Investor Protection and Big 4 Audits on Earnings Quality 179 TABLE 5 (Continued) Notes: Coefficient p-values are two-tailed and based on asymptotic z-statistics robust to heteroscedasticity and country clustering effects using the method in Rogers 1993. Coefficients on year dummies and industry dummies based on two-digit SIC codes are not reported for brevity. Variables are as defined in Tables 1 and 3. Statistics on test variables are shown in boldface. The three-way interaction term “Investor Protection*R*DR” tests the earnings conservatism of non–Big 4 clients across investor protection regimes. The coeffi- cients are negative and significant at p 0.05 in two models (DIS_REQ and PUB_ENF), and insignificant at the 0.10 level in the other three models. We conclude that there is no evidence earnings conservatism increases for firms with non–Big 4 auditors as investor protection becomes stricter. However, there is some evidence to suggest that the opposite occurs because the negative coefficients on DIS_REQ and PUB_ENF imply that firms with non – Big 4 auditors may actually have less earnings conservatism as investor protection becomes stronger, although we have no explanation for this result. The four-way interaction term “Investor Protection*R*BIG4*DR” tests the incremental earnings conservatism of Big 4 clients relative to non–Big 4 clients as investor protection regimes become stronger. The coefficients are positive and sig- nificant at p 0.01 or less in all models except the model with LAW, which is insignificant (p 0.23). Overall, the evidence indicates that firms with Big 4 auditors report relatively more conservative earnings than non – Big 4 clients as a country’s investor protection regime becomes stronger.12 In sum, we observe the same general pattern across all three tests. Earnings quality is consistently higher as investor protection becomes stronger, but only for those firms with Big 4 auditors. Signed abnormal accruals become smaller (less income-increasing) and the likelihood of reporting a loss increases for firms with Big 4 auditors relative to non – Big 4 clients as a country’s investor protection regime becomes stronger. Because all clients are required to follow the applicable accounting standards within a particular country, systematic differences in accruals and loss reporting imply differences in enforcement by Big 4 and non – Big 4 auditors. Together these two tests imply greater accounting conservatism in the sense of reporting smaller earnings and/or losses and are consistent with the for- mal test of accounting conservatism using the Basu 1997 framework of timely loss recognition. Overall the evidence is compelling and consistently shows that investor pro- tection and auditing have a joint role in the production of higher quality earnings numbers. That is, the role of investor protection on earnings quality is mediated by auditing rather than being a direct effect in its own right. The evidence is consistent with Big 4 auditors having incentives to impose higher earnings quality on their clients as investor protection regimes become stricter. In contrast, the earnings CAR Vol. 25 No. 1 (Spring 2008) TABLE 6 180 Multivariate analysis of earnings conservatism of Big 4 clients (dependent variable is earnings per share before extraordinary items: EARN) Investor Investor Investor Investor Investor protection protection protection protection protection ANTI_DIR, DIS_REQ, LIT_STD, PUB_ENF, LAW, estimate estimate estimate estimate estimate Independent variables (p-value) (p-value) (p-value) (p-value) (p-value) Intercept 0.123 0.139 0.148 0.134 0.136 (0.01) (0.01) (0.01) (0.01) (0.01) CAR Vol. 25 No. 1 (Spring 2008) DR 0.031 0.009 0.013 0.023 0.006 (0.01) (0.63) (0.55) (0.15) (0.75) R 0.016 0.018 0.007 0.013 0.001 (0.31) (0.46) (0.76) (0.58) (0.96) R*DR 0.163 0.230 0.286 0.185 0.297 Contemporary Accounting Research (0.01) (0.01) (0.01) (0.01) (0.01) BIG4 0.011 0.008 0.002 0.010 0.006 (0.10) (0.62) (0.90) (0.32) (0.69) BIG4*DR 0.008 0.010 0.006 0.001 0.010 (0.32) (0.44) (0.74) (0.96) (0.44) BIG4*R 0.017 0.035 0.049 0.023 0.042 (0.11) (0.01) (0.01) (0.10) (0.02) BIG4*R*DR 0.019 0.078 0.105 0.025 0.112 (0.55) (0.04) (0.01) (0.49) (0.01) Investor Protection 0.024 0.009 0.053 0.038 0.045 (0.01) (0.03) (0.03) (0.02) (0.08) (The table is continued on the next page.) TABLE 6 (Continued) Investor Investor Investor Investor Investor protection protection protection protection protection ANTI_DIR, DIS_REQ, LIT_STD, PUB_ENF, LAW, estimate estimate estimate estimate estimate Independent variables (p-value) (p-value) (p-value) (p-value) (p-value) Investor Protection*DR 0.009 0.005 0.015 0.005 0.031 (0.52) (0.23) (0.59) (0.77) (0.13) Investor Protection*R 0.004 0.001 0.014 0.003 0.019 (0.76) (0.83) (0.55) (0.87) (0.38) Investor Protection*R*DR 0.009 0.021 0.168 0.043 0.203 (0.87) (0.10) (0.01) (0.52) (0.01) Investor Protection*BIG4 0.000 0.001 0.017 0.000 0.011 (0.97) (0.85) (0.44) (1.00) (0.62) Investor Protection*BIG4*DR 0.005 0.006 0.018 0.010 0.029 (0.69) (0.14) (0.49) (0.47) (0.17) Investor Protection*BIG4*R 0.033 0.01 0.065 0.041 0.066 (0.01) (0.01) (0.01) (0.02) (0.01) Investor Protection*BIG4*R*DR 0.049 0.037 0.209 0.120 0.254 (0.23) (0.01) (0.01) (0.01) (0.01) LMV 0.009 0.009 0.01 0.01 0.009 (0.01) (0.01) (0.01) (0.01) (0.01) LMV*DR 0.000 0.000 0.000 0.000 0.000 (0.96) (0.98) (0.92) (0.98) (0.99) LMV*R 0.000 0.000 0.001 0.000 0.000 The Effect of Investor Protection and Big 4 Audits on Earnings Quality (0.83) (0.95) (0.78) (0.86) (0.93) (The table is continued on the next page.) CAR Vol. 25 No. 1 (Spring 2008) 181 TABLE 6 (Continued) 182 Investor Investor Investor Investor Investor protection protection protection protection protection ANTI_DIR, DIS_REQ, LIT_STD, PUB_ENF, LAW, estimate estimate estimate estimate estimate Independent variables (p-value) (p-value) (p-value) (p-value) (p-value) LMV*R*DR 0.027 0.028 0.027 0.028 0.028 (0.01) (0.01) (0.01) (0.01) (0.01) LEV 0.048 0.048 0.048 0.047 0.047 (0.01) (0.01) (0.01) (0.01) (0.02) CAR Vol. 25 No. 1 (Spring 2008) LEV*DR 0.002 0.002 0.000 0.000 0.002 (0.81) (0.83) (0.98) (0.97) (0.89) LEV*R 0.079 0.078 0.08 0.079 0.081 (0.01) (0.01) (0.01) (0.01) (0.01) LEV*R*DR 0.061 0.058 0.052 0.064 0.052 Contemporary Accounting Research (0.12) (0.16) (0.22) (0.11) (0.22) MB 0.003 0.003 0.003 0.003 0.003 (0.01) (0.01) (0.01) (0.01) (0.01) MB*DR 0.000 0.000 0.000 0.000 0.000 (0.94) (0.92) (0.96) (0.92) (0.93) MB*R 0.002 0.002 0.002 0.002 0.002 (0.01) (0.01) (0.01) (0.01) (0.01) MB*R*DR 0.002 0.002 0.002 0.002 0.002 (0.37) (0.36) (0.33) (0.38) (0.34) Adj. R 2 0.215 0.217 0.216 0.216 0.215 n 68,167 68,167 68,167 68,167 68,167 (The table is continued on the next page.) The Effect of Investor Protection and Big 4 Audits on Earnings Quality 183 TABLE 6 (Continued) Notes: Coefficient p-values are two-tailed and robust to heteroscedasticity and country clustering effects using the method in Rogers 1993. Coefficients on year dummies and industry dummies based on two-digit SIC codes are not reported for brevity. Variables are as defined in Tables 1 and 3. Statistics on test variables are shown in boldface. quality of firms with non–Big 4 auditors does not vary systematically across inves- tor protection regimes, nor are there systematic Big 4/non–Big 4 differences when the investor protection regime is extremely weak. Separate Big 4 and non–Big 4 tests We are also interested in knowing whether earnings quality increases for firms with Big 4 auditors as investor protection regimes become stronger, irrespective of the relative comparison with non–Big 4 clients. To formally evaluate this, we limit the sample to just those firms audited by Big 4 auditors and reestimate the models in Tables 4 through 6. These untabulated results show that as a country’s investor protection environ- ment becomes stronger, Big 4 clients report smaller signed abnormal accruals and are more likely to report a loss. More specifically, for the abnormal accruals analy- sis, the investor protection variables are all negative and significant at p 0.10 except PUB_ENF, which is significant at p 0.10. Negative coefficients indicate smaller (less income-increasing) abnormal accruals. For the loss avoidance analy- sis, the investor protection variables are all positive and significant at p 0.05. A positive coefficient indicates a greater likelihood of reporting a loss as investor protection becomes stricter. These results are consistent with Tables 4 and 5 and show that earnings quality of Big 4 clients increases with stronger investor protection environments. However, the earnings conservatism analysis is less conclusive. Two test variables are positive and significant (ANTI_DIR and LIT_STD) at p 0.10, while the other three variables are insignificant. For completeness we also reestimate the models in Tables 4 through 6 just for firms with non – Big 4 auditors. This analysis finds no evidence that abnormal accruals or the likelihood of reporting a loss are affected by a country’s investor protection regime, which is consistent with the evidence for the full sample reported in Tables 4 and 5. Also consistent with Table 6, there is some evidence that earnings conservatism of non–Big 4 clients decreases as investor protection becomes stronger. Specifically, three of the investor protection variables are signif- icant (p 0.10) and have negative signs. 6. Robustness tests Deleting U.S. firms As discussed in section 2, an alternative scenario is that Big 4 auditors impose a unique level of earnings quality and accounting conservatism on their U.S. clients CAR Vol. 25 No. 1 (Spring 2008) 184 Contemporary Accounting Research due to the extreme litigation risk in the United States. If true, then we should not observe a significant interaction between investor protection and the Big 4 variable if the U.S. observations are dropped from the sample. Therefore we delete all U.S. firms and reestimate the models in Tables 4 through 6. These untabulated results are consistent with those reported in Tables 4 through 6. For the abnormal accruals analysis, the Big 4 interaction variable is significant at p 0.10 in all models (except ANTI_DIR, which is insignificant at p 0.24). For the loss avoidance analysis, the Big 4 interaction variable is significant in two of the five models at p 0.10, but is insignificant for

Joint Effect of Investor Protection & Big 4 Audits on Earnings Quality (PDF)

Document Details

Tags

Related

Summary

Full Transcript

Upgrade to continue