Does Auditor Reputation Matter? The Case of KPMG Germany and ComROAD AG

Authors

  • JOSEPH WEBER,

    1. Sloan School of Management, Massachusetts Institute of Technology
    Search for more papers by this author
  • MICHAEL WILLENBORG,

    1. School of Business, University of Connecticut
    Search for more papers by this author
  • JIEYING ZHANG

    1. Leventhal School of Accounting, University of Southern California. We thank Liesbeth Bruynseels, Sung Gon Chung, Victoria Dickinson, Ron Guymon, Debra Jeter, Bruce Johnson, Bill Kinney, S. P. Kothari, Ling Lei, Roger Meuwissen, Reiner Quick, Doug Skinner, Ross Watts, David Weber, Jerry Zimmerman, two anonymous reviewers, and participants at the 2006 AAA Auditing Midyear Conference, Boston University, University of Chicago, University of Florida, Georgetown University, University of Iowa, Massachusetts Institute of Technology, Michigan State University, Penn State University, Tilberg University, University of Connecticut, University of Southern California, and University of Texas.
    Search for more papers by this author

ABSTRACT

We study the stock and audit market effects associated with a widely publicized accounting scandal involving a public company (ComROAD AG) and a large, reputable audit firm (KPMG) in a country (Germany) that has long provided auditors with substantial protection from shareholder legal liability. We use this event to study whether an auditor's reputation helps to ensure audit quality, a rationale for which recent literature and events provide scant support. Given the absence of a strong insurance rationale for audit quality, Germany permits a relatively clean test of whether auditor reputation matters. We find that KPMG's clients sustain negative abnormal returns of 3% at events pertaining to ComROAD, and that these returns are more negative for companies that are likely to have higher demands for audit quality. We also find an increase in the number of clients that drop KPMG in the year of the ComROAD scandal (mostly smaller, recently public companies that are similar to ComROAD). Overall, our results provide support for the reputation rationale for audit quality.

1. Introduction

We study stock and audit market effects associated with an accounting scandal involving a public company (ComROAD AG) and a large, reputable auditor (KPMG) in a low-litigation country (Germany). Per Klein and Leffler's [1981] model of endogenous quality, reputable firms provide high quality because they earn quasi-rents that they fear losing should they “cheat” by opportunistically providing low quality. DeAngelo[1981, p. 185] applies this logic to the audit market, arguing that larger audit firms have “more to lose” from supplying low audit quality. Our paper is essentially a test of DeAngelo's reputation rationale for audit quality.

A series of papers examining different contexts and international settings conclude that an insurance rationale for audit quality appears to dominate the reputation rationale (Lennox [1999], Willenborg [1999], Khurana and Raman [2004]). In addition, from a policy perspective, studying whether markets discipline auditors that supply low quality is important given today's climate where it is not obvious that, without a litigation incentive, auditors do high-quality work.1 While it stands to reason that “auditor reputation should matter,” recent literature and events suggest that its impact in terms of inducing audit quality is less than that of legal liability.

In Germany it is difficult for investors to sue auditors and there is a longstanding (since 1931) cap on auditor civil liability to shareholders, presently just €4 million per audit.2 These features of the German legal system suggest that auditors are unlikely to be a source of much insurance to investors. Germany therefore facilitates a clean test of the reputation rationale for audit quality, because it helps circumvent the insurance rationale that motivates wealthy auditors to provide quality in high-litigation countries. We use this context to study the response of investors and boards to events that, relatively unambiguously, damage an auditor's reputation.

The accounting scandal we study involves recognizing false revenues. ComROAD AG, a Munich-based company that develops and markets road transport telematics applications (e.g., navigation assistance, mobile Internet, emergency service), went public on Germany's Neuer Markt in November 1999.3 Soon thereafter, ComROAD began reporting fourfold growth in revenues, from DM 4.6 million to DM 20.0 million to DM 85.8 million for calendar years 1998, 1999, and 2000, respectively. However, by Summer 2001 ComROAD began issuing shareholder letters to defend against accusations of accounting manipulations, the most prominent of which was that they had “phantom partners in Asia.” On February 19, 2002, approximately one month before an analyst conference at which ComROAD was to provide its 2001 audited financial statements, KPMG declined its mandate as auditor, thereby effectively resigning. Per a spokesperson, KPMG stated that “there were justified doubts about the trustworthiness of ComROAD.” (Manager Magazin [2002]). ComROAD responded by announcing that they would hire a new auditor to fully investigate. In a press release on April 10, 2002, ComROAD stated that, according to their new auditor, Rödl & Partner, a major Hong Kong-based customer did not exist and, largely as a consequence, 97% of calendar-year 2000 revenues were fictitious. Later, on April 23, 2002, ComROAD announced, again per Rödl, that 86% and 63% of revenues for 1999 and 1998, respectively, were also fictitious. Arguably in “damage-control mode,” just one day later, on April 24, 2002 KPMG announced it would re-audit all of its Neuer Markt–traded clients.

We use the ComROAD scandal as a case study of whether auditor reputation matters. First, to study whether investors value auditor reputation, we analyze the stock market reaction of KPMG's clients to these events. Second, to examine the specifics of when auditor reputation/quality is more or less valuable to investors, we analyze characteristics associated with this market reaction. Third, to examine whether supervisory boards value auditor reputation, we analyze KPMG Germany's market share before, during, and after ComROAD.

We find that KPMG's clients sustain cumulative negative abnormal returns of about 3% around these events. The reaction is especially large at KPMG's resignation, and at ComROAD's announcement that the majority of their revenues were also bogus for 1999 and 1998, followed by KPMG's announcement that they would re-audit their Neuer Markt clients.

These results suggest that German investors value auditor reputation. Since German law makes it difficult for investors to sue auditors and, in most instances, caps auditor liability, the ComROAD scandal does not likely affect KPMG's viability as a source of insurance to investors. Therefore, in contrast to an insurance rationale, the market reaction more likely reflects KPMG's loss of reputation as a high-quality auditor. The reaction is consistent with investors assigning a higher probability that KPMG-audited financial statements contain material errors and, because of this, revising their beliefs downward regarding the quality of KPMG audits and upwards regarding whether KPMG clients incur costs to switch audit firms (DeAngelo [1981]).4

We find that abnormal returns are more negative for distressed companies, those that follow U.S. generally accepted accounting principles (GAAP) or International Accounting Standards (IAS), newer clients of KPMG, smaller companies, and companies with more subsidiaries. Taken together, we interpret these results as consistent with the view that when concerns regarding audit quality are more important investors consider damage to an auditor's reputation to be more costly.

We then expand our sample to encompass the German audit market and study KPMG's share before, during, and after ComROAD. Following the reputation literature, we focus on clients that switch from KPMG. We emphasize however, for several reasons, it is not obvious KPMG would suffer a loss of audit clients following ComROAD. For one, Germany neither permits auditors to advertise nor allows direct uninvited solicitation (Vanstraelen [2000]), activities that Chaney, Jeter, and Shaw [1997] find are positively associated with client/large auditor realignments. Such restrictions should limit the extent to which other auditors can poach KPMG's clients. For another, given other stakeholders, a German managers' objective function may not be to maximize shareholder value, and supervisory boards may be less likely to change from KPMG after ComROAD. Lastly, whereas KPMG's response was to re-audit their Neuer Markt clients, models of endogenous quality explicitly do not contemplate such a response.5

Despite these factors, we find that KPMG's rate of attrition (the fraction of their clients that change auditors the next year) doubles in calendar year 2002 (15.7% versus a three-year average of 7.7%). We conduct a multivariate analysis of auditor changes and confirm this increase in the number of clients that change from KPMG Germany for their 2002 audit.6 Partly because of this, for the 1999–2003 period, KPMG is the only major German auditor, other than Andersen, to suffer a net loss of clients. Moreover, the characteristics of companies that change from KPMG are consistent with concerns regarding audit quality. Taken together, our findings are consistent with the view that German supervisory boards chose not to retain KPMG for quality reasons.

The rest of our paper proceeds as follows. Section 2 surveys literature and discusses the German setting and the ComROAD event. Sections 3 and 4 provide our analyses of the stock and audit market effects associated with ComROAD, respectively. Section 5 concludes.

2. Background and Motivation

We first discuss theoretical literature on endogenous quality, which shows differential quality arises either because reputable firms have greater incentives not to perform a low-quality audit (reputation rationale) or because wealthier firms have a stronger bond to ensure a high-quality audit (insurance rationale). We then survey the empirical literature, which more strongly supports the insurance rationale. Last, we argue the ComROAD scandal provides a good setting to examine the reputation rationale for audit quality, which we test by studying whether an auditor's reputation for high quality matters to German investors and supervisory boards.

2.1 theoretical literature on endogenous (audit) quality

2.1.1. Reputation Models in Repeat-Purchase Settings. Klein and Leffler [1981] study a scenario in which competitive prices do not yield a level of profitability to induce a producer to consistently provide a high-quality product. In their model, price premiums arise “[t]o motivate competitive firms to honor high quality promises because the value of satisfied customers may exceed the cost savings of cheating them” (Klein and Leffler [1981, p. 623]). Because the existence of price premiums is incompatible with competitive markets, they characterize these “quasi-rents” as returns on nonsalvageable investments that customers perceive as a credible commitment to high quality. This perception translates into a willingness to pay a premium, thereby providing the company with a return on its investment. In essence, companies provide high quality because they earn a stream of profits that discourages them from the lure to cheat by lowering quality. Critical to this result is that “cheating” is caught by consumers, who punish the company by curtailing purchases.

DeAngelo [1981] relates this logic to the audit market, and concludes that auditor size is a proxy for quality. She argues that for larger auditors the disincentives to “cheat” outweigh the incentives. For an incumbent audit firm, for whom the quasi-rents are client specific (i.e., the repeat purchase is by the same customer), the incentive to opportunistically provide a low-quality audit is to retain the present value of the client's quasi-rents. In this sense, cheating results in a benefit as the auditor continues to receive rents without incurring the additional costs of exerting high-quality effort. The disincentive to cheating relates to quasi-rents from all other clients, some of which will defect if they learn their auditor is providing low-quality audits. Therefore, as audit firms grow their clientele the costs of cheating exceed the benefits.

Later papers refine these insights. Shapiro [1983] extends Klein and Leffler [1981] to a multiperiod setting with free entry in which reputation is a cost of (not a barrier to) entry. In his model, companies earn a price premium either as an incentive to produce high quality or as a return on their investment in reputation. In a model relevant to professional service providers (where customers may not easily catch and punish cheaters), Rogerson [1983] formalizes DeAngelo's [1981] insight that high-quality companies are “larger,” in that they have more customers.

2.1.2. Insurance/Liability Models.  The insurance rationale for audit quality arises to the extent institutional details enable financial statement users to recover damages from auditors in the case of audit failures. A stream of literature studies the connection between audit quality and the insurance coverage that auditors offer to investors in the event of securities litigation. Simunic [1980] models audit fees as a linear combination of marginal cost plus expected losses from shareholder litigation in a manner that interrelates these components (i.e., additional effort increases the resource cost of the audit and decreases expected litigation losses). By expending more effort, auditors are more likely to detect material misstatements and satisfy professional standards, thereby mitigating each of the conditions necessary to bring successful litigation.

Dye [1993] models the audit as enhancing allocation of resources and providing investors with a claim on the auditor in the event of an audit failure, and demonstrates conditions under which an auditor's wealth serves as a bond for audit quality. Because larger firms tend to be wealthier (i.e., they have “deeper pockets”), depending on the institutional setting, larger/wealthier audit firms have greater incentives to provide high-quality audits.

2.2 empirical literature on audit quality

2.2.1. Event Studies.  Several studies focus on events where investors receive news that leads them to revise their beliefs regarding a company's auditor. For example, in November 1990, under a burden of litigation, Laventhol & Horwath (L&H), at the time the seventh-largest accounting firm in the United States, unexpectedly filed for chapter 11 bankruptcy. Two studies of the impact of L&H's bankruptcy on the prices of their clients' shares report results consistent with the notion that investors rely on auditors as a source of recovery for losses. Both Menon and Williams [1994] and Baber, Kumar, and Verghese [1995] find negative abnormal returns of about 2% to a portfolio of L&H clients at the time of L&H's bankruptcy filing, which could support either an insurance rationale or a reputation rationale for audit quality.

Chaney and Philipich [2002] study the stock market reaction to various Enron/Andersen events, focusing on Andersen's admission on January 10, 2002 that they destroyed a large number of Enron-related documents. They document significant negative returns to Andersen's clients of 2% at the time of the shredding disclosure, and 1% at the release of a report from Enron's Board and Andersen's hiring of Paul Volcker to head an independent oversight board. Chaney and Philipich[2002, p. 1243] interpret these negative returns as damage to Andersen's reputation for high quality, which “[paid] the ultimate market price for loss of reputation.”

However, because Enron so closely follows other big Andersen audit failures at Sunbeam and Waste Management, it is difficult to disentangle the reputation from insurance rationales for audit quality in the case of Enron/Andersen.7 In reacting to the shredding disclosure, within just months of Andersen's Sunbeam and Waste Management settlements, Assistant U.S. Attorney General Chertoff said Andersen was a “recidivist …[and that] the shredding at Enron was the worst case of corporate obstruction …[h]e had ever seen” (Alexander et al. [2002 p. 1]). Thus, it seems reasonable that a portion of the returns Chaney and Philipich [2002] document may reflect the stock market's reassessment of Andersen's prospects for survival, and the corresponding possibility that Andersen's insurance coverage to investors could be withdrawn.8

Whereas the L&H and Andersen incidents involve concerns regarding whether auditor resources would continue to be available to indemnify investor losses, the situation differs for KPMG and ComROAD. Because auditor liability to investors is limited under German law, despite the losses to ComROAD investors, the viability of KPMG Germany was arguably never in serious question. Moreover, even if KPMG was to suffer a similar fate as Andersen and L&H, investors in their German clients would likely retain the insurance coverage specified by German law (see section 2.3.1). Because of this, the ComROAD/KPMG incident enables us to abstract from the insurance rationale for audit quality and focus on the market reaction to events that are, more so than for the L&H and Andersen incidents, damaging to an auditor's reputation.

2.2.2. Other Studies.  Three recent empirical studies each conclude that the insurance rationale for audit quality appears to dominate the reputation rationale. Lennox [1999] finds that while larger U.K. auditors are more likely to be sued and criticized in the business press (e.g., for rendering a clean opinion to an ex post bankrupt company), such firms do not suffer a decrease in market share or an increase in client defections. Overall, he interprets this as consistent (inconsistent) with an insurance rationale (reputation rationale) for audit quality. In a study of small U.S. initial public offerings (IPOs), a context with substantial diversity in auditor type, Willenborg [1999] finds that the inverse relation between auditor size and underpricing is also evident for start-up IPOs, a setting wherein he argues the reputation rationale for audit quality is less important. He also finds audit fees are positively associated with IPO proceeds (the upper limit of the auditor's coverage to investors). He concludes that his “… results suggest the importance of an insurance-based demand for IPO audits” (Willenborg[1999, p. 237]). Khurana and Raman[2004, p. 475] find that the presence of a large auditor is inversely related to a measure of the ex ante cost of equity capital in the United States but not Australia, Canada, or the United Kingdom (i.e., lower investor litigation countries than the United States), and conclude “… litigation concerns, rather than reputation protection, drive the perceived higher quality and financial reporting credibility of Big 4 audits.”9

2.3 the german setting and the comroad ag/kpmg event

2.3.1. The German Setting.  Several characteristics of Germany are important to our study. To begin, it is difficult for clients and investors to sue auditors for damages associated with misstated financial statements. Clients may sue for breach of contract, but German law imposes a high bar by requiring they demonstrate the auditor acted intentionally or with reckless disregard for the truth. Investors may bring action under civil liability, but German law requires they show an implied contract or that the contract has “protective effects to third parties,” either of which are arguably quite difficult to demonstrate (European Commission [2001]).10

In addition to the difficulty in bringing investor suits against auditors, in most instances there is a limit on auditor civil liability. Since 1931, Germany's Commercial Code specifies a cap on the statutory auditor's exposure to legal damages to their client for negligence, and, since 1998, auditors are liable to third parties for negligence, though damages are also subject to cap. At the time of ComROAD, which follows the Act on Control and Transparency of Enterprises in April 1998, the maximum amount for negligence was €4 million per audit (i.e., $3.5 million on February 19, 2002, when KPMG quit as ComROAD's auditor). While third parties can sue under tort law, where damages are not capped, to successfully do so they must demonstrate immoral violation of professional duties by the auditor with the intent to damage. While information on successful litigation and damages is confidential, investor litigation probabilities and costs are arguably quite low for German auditors (Wingate [1997], LaPorta, Lopez-De-Silanes, and Shleifer [2006]).

In addition to these characteristics of their legal system (and in contrast to the American Institute of Certified Public Accountants Code of Ethics), Germany does not permit auditors to advertise or to engage in direct, uninvited solicitation of competitor's clients. As we discuss in section 2.1.1, a critical assumption of reputation models is that if a high-quality provider is caught cheating, consumers punish them by curtailing purchases. Germany's restrictions on auditor competition should bias against finding market-share evidence of a reputation rationale for auditing.

Lastly, because of the institutional details of board structure and stakeholder monitoring, independent audits of public companies are arguably less important in Germany. Public stock companies in Germany have two-tier governance structures in which banks often play an important monitoring role. The two-tier structure relates to both a management board and a supervisory board and, because German banks act as both commercial/investment banks and often hold substantial ownership positions, it is common for their representatives on supervisory boards to be active in corporate governance (Gietzmann and Quick [1998]). To the extent that auditing serves as a monitoring device to mitigate conflicts of interest (Watts and Zimmerman [1983]), in countries where the monitoring of management is already “high” the demand for audit quality should be lower. Following this, German investors may be less likely to place as much reliance on audit quality and German auditors are less likely to be subject to dismissal when their reputation becomes tarnished. This should bias against finding either stock market or market share evidence of a reputation rationale for audit quality.

To summarize, in Germany, the difficulty of successfully bringing suit against auditors coupled with a cap on client and investor damages for negligence reduces the insurance rationale for audit quality. As such, the German setting provides an opportunity to examine the reputation rationale for audit quality. However, because of certain institutional details, the German setting also imposes obstacles to finding evidence consistent with a reputation rationale for quality.

2.3.2. The ComROAD/KPMG Event.  On February 19, 2002 (Event1), KPMG resigned as ComROAD's auditor because of a “lack of clarity” about the existence of companies in Hong Kong and Spain. Per a February 22, 2002 press release, ComROAD announced their hiring of an independent auditor (Rödl & Partner) to investigate KPMG's concerns and that they were “endeavoring to swiftly and conclusively refute the continuing speculations, particularly in regards to some of our company's partnerships in Asia (Schnabel and Mehler [2002]).” Per an April 10, 2002 (Event2) press release, ComROAD stated that, according to Rödl & Partner, their major Hong Kong–based customer (VT Electronics) was a fictitious company and that invoices to them represented 97% of sales for 2000 (Mehler [2002]). Just two weeks later, on April 23, 2002 (Event3), ComROAD announced that, again per Rödl, 86% and 63% of revenues for 1999 and 1998, respectively, were also bogus billings to VT Electronics (Schwamm and Mehler [2002]).11 Then, just one day later, on April 24, 2002 (also Event3), KPMG announced that it would re-audit all its Neuer Markt clients (Bosch [2002]).

In October 2002, an investor lawsuit was filed against ComROAD's CEO, Bodo Schnabel, that was settled in May 2003 for €116,000 ($147,600) (General/Re Corporation [2004, 2005]). Schnabel was sentenced for seven years for fraud and market manipulation (Ryan [2003]). As of December 31, 2005, no lawsuit has been filed against KPMG nor were they the subject of professional or governmental investigation for ComROAD (Bollen et al. [2005]).

To summarize, reputable/large firms should provide high-quality audits for two reasons: (1) they have more to lose from providing low quality, in terms of being caught and punished by repeat purchasers; and (2) they have stronger incentives to protect their at-risk wealth in the event of litigation. Overall, empirical research concludes in favor of insurance over reputation. The ComROAD event provides an opportunity to test the reputation rationale for audit quality.

3. Stock Market Reaction

3.1 sample

We identify 128 German companies on Worldscope with KPMG as auditor for the 2001 financial statements, of which 92 have stock price information on Datastream for the 252 trading days from November 1, 2001 to October 31, 2002. In section 3.2, we examine the stock price reaction to the ComROAD scandal for these 92 clients of KPMG. Of these 92, 67 have financial statement, market, and other data for our cross-sectional analysis. In section 3.3, we examine the determinants of the stock price reaction to the ComROAD scandal for these 67 KPMG clients.12

3.2 overall stock market reaction

Average market-adjusted returns are significantly negative for all three of our event dates (panel A of table 1). In panel B of table 1, we present the results of Schipper and Thompson's [1983] multivariate regression model (MVRM), for which we estimate abnormal returns using a time series of daily returns to an equally weighted portfolio of KPMG clients. In our context, the MVRM conditions the return-generating process on the occurrence of the ComROAD event(s) by adding an indicator variable to the market model.13 The estimation accounts for possible contemporaneous correlation of residuals and cross-sectional heteroskedasticity. Because the MVRM assumes residuals are independent and identically distributed, if necessary, we adjust for time-series heteroskedasticity (White [1980]).

image(1)
Table 1. 
Market Reaction to ComROAD Events for KPMG Germany's Clients
Panel A: Descriptive statistics
 MeanMedian
Event 1 (−1, February 19, 2002, +1)
 Raw return(54 of 92 < 0)−0.016−0.008
(t-statistic −4.05)***(rank test −15)***
 MarketMDAX-adjusted return(52 of 92 < 0)−0.012−0.005
(t-statistic −3.22)***(rank test −5)
Event 2 (−1, April 10, 2002, +1)
 Raw return(40 of 92 < 0)−0.005 0.000
(t-statistic −1.21)(rank test 1)
 MarketMDAX-adjusted return(60 of 92 < 0)−0.011−0.006
(t-statistic −2.50)***(rank test −14)***
Event 3 (−1, April 23–24, 2002, +1)
 Raw return(52 of 92 < 0)−0.019−0.009
(t-statistic −3.57)***(rank test −10)**
 MarketMDAX-adjusted return(52 of 92 < 0)−0.018−0.008
(t-statistic −3.40)***(rank test −6)
Panel B:Schipper and Thompson [1983]regressions
Returnt01ReturnDAX,t2ReturnMDAX,t3ReturnSDAX,t+ΣθkEventk,tt (1)
VariableCoefficient (t-statistic)Coefficient (t-statistic)Coefficient (t-statistic)Coefficient (t-statistic)
  1. For panel A, observations are 92 KPMG-audited German companies (except ComROAD) for 2001 on Worldscope with stock returns on Datastream for the 252 trading days from November 1, 2001 to October 31, 2002. MarketMDAX-adjusted returns are raw returns minus the German Mid-Cap Index return over the event window. For panel B, observations are daily portfolios of these 92 companies. Estimation is OLS with t-statistics in parentheses.

  2. ** and *** indicate significance at the 5% and 1% levels, respectively (two-sided tests).

  3. Other variables are as follows:

  4. Event1= 1 for the three days around February 19, 2002 (i.e., when KPMG resigns as ComROAD's audit firm) and 0 otherwise.

  5. Event2= 1 for the three days around April 10, 2002 (i.e., when ComROAD admits their financial statements for 2000 contain material misstatements) and 0 otherwise.

  6. Event3= 1 for the four days around April 23–24, 2002 (i.e., when ComROAD announces the large majority of their revenues were bogus, not just for 2000, but for 1999 and 1998; and KPMG announces it will re-audit its Neuer-Markt traded clients, respectively) and 0 otherwise.

  7. Event1&2&3= 1 for the 10 days that comprise Event1, Event2, and Event3 and 0 otherwise.

  8. Returnt= The return to an equally weighted portfolio of 92 KPMG client companies (excluding ComROAD) with stock price information on Datastream for day t (t= 1, 2, …T), where T is the 252 daily return observations, from November 1, 2001 to October 31, 2002.

  9. ReturnDAX,t= The return on the DAX on day t. The DAX is the German Stock Index, comprised of 30 blue-chip stocks trading on the Frankfurt Stock Exchange.

  10. ReturnMDAX,t= The return on the MDAX on day t. The MDAX is the German Mid-Cap Index, comprised of 50 stocks that are smaller in market capitalization than those that comprise the DAX.

  11. ReturnSDAX,t= The return on the SDAX on day t. The SDAX is the German Small-Cap Index, comprised of 50 stocks that are smaller in market capitalization than those that comprise the MDAX.

Constant−0.00−0.00−0.00−0.00
(−3.01)***(−3.05)***(−3.01)***(−2.81)***
ReturnDAX,t0.160.160.150.15
(9.40)***(9.37)***(9.00)***(8.87)***
ReturnMDAX,t0.240.240.250.26
(5.70)***(5.64)***(5.85)***(6.07)***
ReturnSDAX,t0.420.420.420.41
(8.45)***(8.42)***(8.41)***(8.34)***
Event1−0.01 
(−2.20)** 
Event2 −0.00 
(−1.10) 
Event3 −0.01 
(−2.08)** 
Event1&2&3 −0.03
(−5.72)***
Observations252252252252
Adjusted R278.4%78.0%78.3%79.7%

Panel B of table 1 presents the results of estimating equation (1). Abnormal returns are negative for the first and third event windows, and a combination of all event windows.14,15,16 The coefficient estimates for Event1 and Event3 are similar to those in Chaney and Philipich [2002] for Andersen's clients of −2% and −1%, at the shredding disclosure and release of a report from Enron's Board and Andersen's hiring Paul Volcker, respectively. However, in contrast to Enron/Andersen, it is less plausible that the negative returns to KPMG's clients at the time of the ComROAD scandal relate to an insurance rationale for audit quality.

3.3 cross-sectional analysis of stock market reaction

In this section, we examine whether the negative stock market reaction to ComROAD is consistent with investors assigning higher probabilities that KPMG-audited financial statements contain material errors. As investors learn of KPMG's role in the ComROAD scandal, we expect them to downgrade their beliefs regarding the quality of KPMG's past and future audits (i.e., increase their expectations that other KPMG-audited financial statements may be materially misstated and, as a result, that some KPMG clients will defect).17 We estimate the association between abnormal returns to KPMG clients at the time of ComROAD and 11 characteristics: bankruptcy probability; the market-to-book ratio; if the company follows U.S. GAAP or IAS; whether the company recently changes to KPMG; size; the ratio of accounts receivable to assets; the number of subsidiaries and foreign subsidiaries; whether the client is a bank, insurer, or other financial company; and Neuer Markt and ADR listings.

We specify the probability of bankruptcy (Pr(Bankrupt)) per Zmijewksi [1984]. A negative association between returns and Pr(Bankrupt) supports a quality story (Baber, Kumar, and Verghese [1995]).18

Because auditors are arguably more important when information asymmetry between management and investors is greater (when investment opportunities comprise a larger portion of company value), we examine the association between returns and the market-to-book ratio (Market/Book). We expect the damaging effects of substandard auditing should increase with the market-to-book ratio.

We also consider whether a company follows U.S. GAAP or IAS (USGAAP/IAS). If German investors are less familiar with U.S. GAAP or IAS, then a decrease in audit quality is arguably more costly for companies that follow these accounting principles. Consistent with this, Leuz and Verrecchia[2000, p. 98] find German companies that switch to U.S. GAAP or IAS have lower bid–ask spreads and higher turnover (proxies for the information asymmetry component of the cost of capital) and conclude “a switch to international reporting represents a substantial increase in a firm's commitment to disclosure.” If investors bundle this commitment to disclosure with the auditor's reputation, the negative effects of revelations of faulty auditing should be more severe for companies that adopt international reporting.

We specify an indicator variable (AuditorΔ) for whether a company changes to KPMG during any of the three years prior to 2002, the year of the ComROAD scandal. Several papers document that financial reporting quality is lower, and that audit failures are more likely, during the early years of an audit firm's tenure (e.g., Johnson, Khurana, and Reynolds [2002], Erickson, Mayhew, and Felix [2000]), consistent with auditors being less familiar with the client's business and more likely to make errors or perhaps be fooled by management. A negative coefficient for AuditorΔ is consistent with investors being more suspicious of KPMG's clients having financial misstatements during the first three years of KPMG's tenure as the company's auditor.

We specify company size (Ln(Assets)) because audit failures may be less likely to occur in larger companies, consistent with better corporate governance and less evidence of earnings management (e.g., Myers, Myers, and Omer [2003]). If KPMG is performing substandard audits, they may be more likely to result in audit failures among their smaller clients (e.g., ComROAD).

Given that the ComROAD scandal involved falsification of a large amount of revenues, investors may be suspicious of KPMG clients with large balances in accounts receivable. To consider this, we specify the ratio of accounts receivable to assets (AcctsRec/Assets).

Another factor that could impact the strength of the market reaction to ComROAD is the complexity of the audit. Just as shorter auditor tenure seems likely to increase the chances of an error, so might audit complexity. To consider this, we specify the number of subsidiaries (Ln(1 +#Subsidiaries)) and the number of foreign subsidiaries (Ln(1+#ForeignSubsidiaries)).

We consider whether the company is in banking, insurance, or other financial industries (Financial) and Neuer Markt listing (NeuerMarkt). We expect the negative effects of disclosures of faulty auditing to be more acute for clients in industries where KPMG has substantial market share (panel B, table 3) and the young, equity-oriented companies on this exchange, respectively.

Table 3. 
Market Share Analysis: Overall German Audit Market from 1998 to 2003
Panel A: Sample selection
 CompaniesCompany-years
German companies on Worldscope 1998–20031,0183,616
less: Companies that appear for nonconsecutive years  150  505
less: Companies that change their fiscal year-end   52  228
less: Companies with non-December 31 year-ends  118  378
less: Companies with missing financial statement data   29   98
 
Sample for German audit market share analysis (tables 4–6)  6692,407
Panel B: Company-years: by audit firm by industry
 BanksIndustrialsInsuranceFinancial OtherTransportationUtilitiesTotal
AA  0  (0.0)  131  (7.7)  0  (0.0)  6  (2.5) 4  (7.4)  2  (1.8)  143  (6.0)
D&T  2  (1.3)   42  (2.4)  4  (3.0)  5  (2.0) 1  (1.9)  1  (0.9)   55  (2.3)
E&Y 12  (7.7)  237 (13.9)  6  (4.5) 12  (4.9) 2  (3.7) 11 (10.1)  280 (11.6)
KPMG 55 (35.3)  367 (21.4) 73 (55.3) 41 (16.7) 1  (1.8)  5  (4.6)  542 (22.5)
PwC 47 (30.1)  293 (17.1) 27 (20.5) 39 (15.9)27 (50.0) 58 (53.2)  491 (20.4)
 Big 5/4116 (74.4)1,070 (62.5)110 (83.3)103 (42.0)35 (64.8) 77 (70.6)1,511 (62.8)
BDO 11  (7.0)   99  (5.8) 17 (12.9) 15  (6.1) 6 (11.1) 11 (10.1)  159  (6.6)
Other auditors 29 (18.6)  542 (31.7)  5  (3.8)127 (51.9)13 (24.1) 21 (19.3)  737 (30.6)
 Non–Big 5/4 40 (25.6)  641 (37.5) 22 (16.7)142 (58.0)19 (35.2) 32 (29.4)  896 (37.2)
Overall156(100.0)1,711(100.0)132(100.0)245(100.0)54(100.0)109(100.0)2,407(100.0)
Panel C: Company-years: by audit firm by year
 199819992000200120022003Total
  1. For exposition, table 3 accords retroactive treatment of the 1999 merger between Coopers & Lybrand and Price Waterhouse to 1998. Audit firm market share percentages are in parentheses.

AA 17  (4.8) 20  (5.8) 55 (10.8) 45  (9.6)  6  (1.5)  0  (0.0)  143  (6.0)
D&T  6  (1.7)  6  (1.7)  7  (1.4)  6  (1.3) 16  (3.9) 14  (4.3)   55  (2.3)
E&Y 27  (7.6) 20  (5.7) 41  (8.1) 50 (10.7) 75 (18.3) 67 (20.9)  280 (11.6)
KPMG 97 (27.3) 94 (26.9)105 (20.7)100 (21.5) 76 (18.6) 70 (21.8)  542 (22.5)
PwC 71 (20.0) 74 (21.2)102 (20.1) 93 (20.0) 83 (20.3) 68 (21.2)  491 (20.4)
 Big 5/4218 (61.4)214 (61.3)310 (61.1)294 (63.1)256 (62.6)219 (68.2)1,511 (62.8)
BDO 26  (7.3) 23  (6.6) 30  (5.9) 30  (6.4) 29  (7.1) 21  (6.6)  159  (6.6)
Other auditors111 (31.3)112 (32.1)167 (32.9)142 (30.5)124 (30.3) 81 (25.2)  737 (30.6)
 Non–Big 5/4137 (38.6)135 (38.7)197 (38.9)172 (36.9)153 (37.4)102 (31.8)  896 (37.2)
Overall355(100.0)349(100.0)507(100.0)466(100.0)409(100.0)321(100.0)2,407(100.0)

Lastly, we control for whether the company trades ADRs (ADR). While we argue that German investors are unlikely to sustain a loss in auditor insurance coverage as ComROAD unfolds, shares of companies that trade ADRs may reflect an insurance effect. Also, given that Leuz [2003] shows that foreign listing is the major determinant of whether a German company adopts U.S. GAAP or IAS, ADR helps control for selectivity associated with the USGAAP/IAS variable.

To study the association between these characteristics and abnormal returns, we follow Sefcik and Thompson's [1986] portfolio weighting procedure. The primary advantages of this procedure are that it accounts for cross correlation between characteristics and adjusts standard errors for cross correlation of residuals (Bernard [1987]). Of the 92 companies in the MVRM estimation, 67 have the necessary data for the Sefcik and Thompson [1986] estimation.

We take the following steps. We first form a matrix F=[1 X2Xj], which has a column of ones and J−1 columns of firm characteristics. In matrix F, Xj is an N×1 vector for the jth firm characteristic, where N denotes the number of firms in the analysis. We then create a portfolio weight matrix W and calculate J sets of weighted portfolio returns, as we define below.

image
W

J×N matrix of portfolio weights ( J= 11 firm characteristics and N= 67 firms)

F

N×J matrix defined in the first step

Wp

N×1 vector, the jth row of portfolio weight

inline image

return on portfolio j on day t

Rit

individual firm's return on day t

image(2)

Panel A of table 2 provides descriptive statistics, which we partition by whether a company's event-day raw returns are at or above the sample median. KPMG clients that sustain a below-median reaction are more likely to follow U.S. GAAP or IAS and have more subsidiaries. With regard to size, because both Deutsche Bank and Allianz (the two largest companies in our sample, by four orders of magnitude) are among the 33 with below-median returns, no descriptive difference is evident. If we exclude these companies, the mean (median) Assets for the remaining 31 decreases to 18,171 (1,114) million DM. Panel B presents the results of estimating equation (2). The coefficients on the event-date variables measure the association between each company/security characteristic and the abnormal returns to the three events. The adjusted R2 provides a measure of the extent to which the events of interest explain the variation in the weighted return matrix associated with each particular characteristic.

Table 2. 
Cross-sectional Analysis of Market Reaction to ComROAD Events for KPMG Germany's Clients
Panel A: Descriptive statistics
Variable Total (N= 67)Market Reaction at or above Median (N= 34)Market Reaction below Median (N= 33)Tests of Difference (Two-Tailed p-Value)
Pr(Bankrupt)Mean 0.22 0.25 0.190.38
Median 0.09 0.11 0.070.44
Market/BookMean 1.02 1.03 1.010.94
Median 0.69 0.84 0.620.15
USGAAP/IASMean 0.46 0.32 0.610.02
AuditorΔMean 0.13 0.09 0.180.27
AssetsMean75,84628,321124,8120.21
Median1,6121,5971,8240.87
AcctsRec/AssetsMean 0.17 0.18 0.160.48
Median 0.17 0.18 0.140.05
#SubsidiariesMean35.6023.0348.550.03
Median24.0021.5030.000.01
#ForeignSubsidiariesMean18.2412.5024.150.05
Median10.00 8.5010.000.01
FinancialMean 0.27 0.29 0.240.64
NeuerMarktMean 0.18 0.12 0.240.19
ADRMean 0.16 0.12 0.210.31
Panel B:Sefcik and Thompson [1986]regressions
Returnj,tjjReturnDAX,tjReturnMDAX,tjReturnSDAX,tkθj,kEventk,tj,t (2)
VariableE[sign]Coefficient (t-statistic)Adjusted R2
  1. Observations are 67 KPMG-audited German companies (except ComROAD) with stock returns on Data-stream for the 252 trading days from November 1, 2001 to October 31, 2002 and necessary data (these companies are a subset of the 92 companies in table 1). Tests of difference in means are two-sample t-tests and tests of difference in medians are Mann-Whitney U-tests. Estimation is portfolio weighted least squares with t-statistics in parentheses.

  2. ** and *** indicate significance at the 5% and 1% levels, respectively (two-sided tests).

  3. Variables are as follows:

  4. Pr(Bankrupt)= Zmijewski's [1984]index

  5. Market/Book, = Market value of equity/book value of equity

  6. USGAAP/IAS, = 1 if company follows U.S. GAAP or IAS and 0 otherwise

  7. AuditorΔ, = 1 if company switches to KPMG during 1999–2001 and 0 otherwise

  8. Assets= Total assets (000,000 DM)

  9. AcctsRec/Assets= Accounts receivable/total assets

  10. #Subsidiaries= Number of subsidiaries

  11. #ForeignSubsidiaries= Number of foreign subsidiaries

  12. Financial= 1 if a bank, insurer, or other financial company and 0 otherwise

  13. NeuerMarkt= 1 if on the Neuer Markt and 0 otherwise

  14. ADR= 1 if an American Depository Receipt and 0 otherwise

  15. Returnj,t= Weighted portfolio return on day t (t= 1, 2, …T), where T is the 252 daily return observations, from November 1, 2001 to October 31, 2002. For each of the j= 12 regressions, daily returns are weighted by a 67 × 12 matrix of portfolio weights so that parameter estimates reflect the effect of each characteristic (plus a constant term) on the market reaction to the three ComROAD/KPMG events

  16. ReturnDAX,t= The return on the DAX (i.e., the German Stock Index, comprised of 30 blue-chip stocks trading on the Frankfurt Stock Exchange) on day t

  17. ReturnMDAX,t= The return on the MDAX (i.e., the German Mid-Cap Index, comprised of 50 stocks that are smaller in market capitalization than those that comprise the DAX) on day t

  18. ReturnSDAX,t= The return on the SDAX (i.e., the German Small-Cap Index, comprised of 50 stocks that are smaller in market capitalization than those that comprise the MDAX) on day t

  19. Eventk,t= An indicator variable for the kth event, equal to 1 during the three- or four-day event period and 0 otherwise. Event1 is February 19, 2002 when KPMG resigns as ComROAD's auditor; Event2 is April 10, 2002, when ComROAD admits their financial statements for 2000 contain material misstatements; and Event3 is April 23 and 24, 2002, when ComROAD announces the majority of their revenues are bogus, not just for 2000, but 1999 and 1998, and KPMG announces it will re-audit its Neuer-Markt traded clients, respectively. For this analysis, we combine these three events.

Constant −0.1121.2%
(−2.50)** 
Pr(Bankrupt)t−0.051.7%
(−2.17)** 
Market/Bookt−0.001.8%
(−0.36) 
USGAAP/IASt−0.0317.9%
(−2.65)*** 
AuditorΔt−τ−0.044.6%
(−2.52)** 
Ln(Assets)t+0.0135.5%
(3.45)*** 
AcctsRec/Assetst0.020.1%
(0.50) 
Ln(1+#Subsidiaries)t−0.0213.7%
(−2.11)** 
Ln(1+#ForeignSubsidiaries)t0.016.0%
(0.85) 
Financialt0.013.0%
(0.59) 
Panel B:Sefcik and Thompson [1986]regressions
Returnj,tjjReturnDAX,tjReturnMDAX,tjReturnSDAX,tkθj,kEventk,tj,t (2)
VariableE[sign]Coefficient (t-statistic)Adjusted R2
 
NeuerMarktt−0.0011.3%
(−0.10) 
ADRt0.0021.8%
(0.27) 

The negative stock reaction to KPMG clients at the time of ComROAD is more severe for stressed companies, those following U.S. GAAP or IAS, those recently switching to KPMG, smaller companies, and those with more subsidiaries. These results suggest that as the events at ComROAD unfold, investors assign higher probabilities that KPMG-audited financial statements contain material errors. We interpret this as support for the reputation rationale for audit quality.

4. KPMG's German Audit Market Share—before, during, and after ComROAD

Following DeAngelo [1981], a high-quality auditor caught “cheating” by providing low quality should be “punished” by their clients. In this section, we study KPMG's market share before, during, and after ComROAD. We begin by providing descriptive statistics for the German audit market (section 4.1) and for changes in KPMG's share (section 4.2.1). We then analyze the likelihood that a KPMG client changes auditors, and whether this likelihood increases at the time of ComROAD (section 4.2.2). Lastly, we examine the characteristics of KPMG clients that change auditors following ComROAD versus those that do not (section 4.3).

4.1 sample and descriptive statistics

Panel A of table 3 details the sample for our market share analysis. We identify 1,018 German companies (3,616 company-years) on Worldscope with financial statement information for any of the years from 1998 to 2003. Because of our focus on changes in KPMG's share of the audit market around the time of ComROAD, we exclude certain observations. We eliminate 150 companies (505 company-years) because they do not appear consecutively on Worldscope and, therefore, we cannot reliably discern whether they change their audit firm. Because of the timing of ComROAD (February–April 2002), we exclude 52 companies (228 company-years) that change their fiscal year-end and 118 companies (378 company-years) with non–calendar year-ends.19 We eliminate 29 companies (98 company-years) with missing financial statement information to compute the control variables for the equation (3) regression. The audit market sample for our descriptive analysis is 669 companies and comprises 2,407 company-years.

Panels B and C of table 3 provide a descriptive analysis of these 2,407 company-years by audit firm/industry and by audit firm/year, respectively. We break out the Big 5 firms and BDO, due to its substantial presence in the German audit market (Ashbaugh and Warfield [2003]). To ease comparison, the market share analyses accord retroactive treatment to the 1999 merger between Coopers & Lybrand and Price Waterhouse (PwC). For the 1998–2003 time period, KPMG is the market leader, auditing 22.5% of total observations, with PwC the runner-up, auditing 20.4% (panel B). KPMG's leadership dips sharply in 2000, when the number of companies in our sample increases 45% (panel C). By 2003, the sample shrinks to pre-1998 levels and the German market is a three-firm race between KPMG and PwC and Ernst & Young (E&Y). We focus in on the year-to-year changes in the KPMG row in panel C of table 3 (see table 4).

Table 4. 
Market Share Analysis: KPMG Germany from 1998 to 2003
Year199819992000200120022003
Total NKPMGTotal NKPMGTotal NKPMGTotal NKPMGTotal NKPMGTotal NKPMG
+N(%)+N(%)+N(%)+N(%)+N(%)+N(%)
  1. Observations are 669 German companies (2,407 company-years) appearing on Worldscope (see table 3). KPMG's audit market share (in%) is in parentheses.

1998–2003154 46(29.9)154+4−446(29.9)154+3−247(30.5)154+0−245(29.2)154+2−344(28.6)154+3−146(29.9)
1998–2002 36  9(25.0) 36+0−2 7(19.4) 36+0−1 6(16.7) 36+0−1 5(13.9)36+1−1 5(13.9) 
   
1998–2001 39  8(20.5) 39+0−0 8(20.5) 39+0−0 8(20.5) 39+0−1 7(17.9)190 49(25.8) 
   
1998–2000 32  6(18.8) 32+0−0 6(18.8) 32+0−2 4(12.5)229 57(24.9) 
   
1998–1999 48 17(35.4) 48+0−116(33.3)261 65(24.9) 
   
1998 only 46 11(23.9)309 83(26.9) 
  
355 97(27.3) 
1999–2003 19 5(26.3)19+1−06(31.6)19+1−25(26.3)19+0−23(15.8) 19+0−0 3(15.8)
1999–2002  3 0 (0.0) 3+0 00 (0.0) 3+0−00 (0.0) 3+0−00 (0.0) 
   
1999–2001  0 0 (0.0) 0+0−00 (0.0) 0+0−00 (0.0)22 3(13.6) 
   
1999–2000  6 0 (0.0) 6+0−00 (0.0)22 5(22.7) 
   
1999 only 12 6(50.0)28 6(21.4) 
  
40 11(27.5) 
2000–2003 125 18(14.4)125+2−119(15.2)125+1−416(12.8)125+2−018(14.4)
2000–2002  30  6(20.0) 30+2−1 7(23.3) 30+0−3 4(13.3) 
   
2000–2001  25  6(24.0) 25+0−0 6(24.0)155 20(12.9) 
   
2000 only  38  4(10.5)180 32(17.8) 
  
218 34(15.6) 
2001–2003 13 1 (7.7)13+1−02(15.4) 13+0−0 2(15.4)
2001–2002 10 1(10.0)10+0−01(10.0) 
   
2001 only 12 4(33.3)23 3(13.0) 
  
35 6(16.7) 
2002–2003  8 1(12.5)  8+0−0 1(12.5)
2002 only 11 0 (0.0) 
  
19 1 (5.3) 
2003 only    2  0 (0.0)
Total355 97(27.3)349 94(26.9)507 105(20.7)466 100(21.5)409 76(18.6)321 70(21.8)
KPMG clients continuing 86(of 97 in 1998)72(of 94 in 1999)97(of 105 in 2000)83(of 100 in 2001) 66(of 76 in 2002)
KPMG rate of attrition (%) 8.1%(i.e., 7 of 86)6.9%(i.e., 5 of 72)8.2%(i.e., 8 of 97)15.7%(i.e., 13 of 83)1.5%(i.e., 1 of 66)

4.2 changes in kpmg's share

4.2.1. Descriptive Statistics. Table 4 provides a descriptive analysis of changes in KPMG's share of the German audit market for the period 1998–2003. For KPMG, 1999 is a stable year; while 11 of its 1998 clients drop out of the sample, KPMG gains 11 of the 40 companies that are new to the sample. Of the 355 companies for 1998, 309 continue in the sample through (at least) 1999. Of these 309, KPMG picks up 4 and loses 7 clients. These 7 are out of KPMG's 86 (97 minus 11) clients that continue in the sample from 1998 to 1999, for an 8.1% rate of attrition. Overall, KPMG's share goes from 27.3% (97 of 355) for 1998 to 26.9% (94 of 349) for 1999.

In contrast to the stability of 1999, KPMG's share drops to 20.7% in 2000 (105 of 507), mostly because it gains just 34 of the 218 (15.6%) new companies to the sample, many of which were Neuer Markt IPOs. As for companies that continue in the sample from 1999 to 2000, KPMG adds 4 and loses 5 clients. These 5 are out of KPMG's 72 (94 minus 16 minus 6) clients continuing from 1999 to 2000, for a 6.9% rate of attrition.

For 2001, the calendar year-end prior to when they resign the ComROAD audit, KPMG adds 5 and loses 8 clients. These 8 are out of KPMG's 97 (105 minus 4 minus 4) clients that continue on from 2000 to 2001, for an 8.2% rate of attrition. For the 1999–2001 period, KPMG loses an average of 6.67 clients per year, an attrition rate of 7.8% (20 of 255).

For 2002, the calendar year of the ComROAD scandal, KPMG audits just 5.3% (1 of 19) of the new companies to the sample. As for companies that continue in our sample from 2001 to 2002, KPMG picks up 5, 2 of which are former Andersen clients (SAP and SAP Systems Integration). KPMG's attrition rate for 2002 is double their three-year average, as they lose 13 clients in 2002, in contrast to a yearly average of 6.67 clients for 1999–2001. These 13 are out of KPMG's 83 (100 minus 7 minus 6 minus 4) clients that continue from 2001 to 2002, or a 15.7% rate of attrition. A test of the change in proportions from the 7.8% attrition rate for 1999–2001 to the 15.7% attrition rate for 2002 yields a Z-statistic of 2.09, significant at 5% (two-tailed). This increase in client attrition does not occur for KPMG in the United States.

For 2003, KPMG adds five clients and loses one. This 1 is out of KPMG's 66 (76 minus 5 minus 4 minus 1) clients that continue in the sample from 2002 to 2003, for a 1.5% rate of attrition. A test of the change in proportions from the 15.7% attrition rate for 2002 to the 1.5% attrition rate for 2003 yields a Z-statistic of 2.94, significant at 1% (two-tailed).

Table 5 provides a zero-sum descriptive analysis that frames KPMG's gains and losses of continuing companies from table 4 in the context of the overall German audit market. The bottom (far right) of the KPMG column (row) shows their client gains (losses) among the companies in the sample that continue on for the next year. That is, the bottom (far right) of the KPMG column (row) depicts the 4, 4, 5, 5, and 5 (7, 5, 8, 13, and 1) new clients (lost clients), for 1999, 2000, 2001, 2002, and 2003, respectively, that we discuss above. With regard to Andersen, table 5 shows that of the 40 clients they lose in the 2002–2003 period, E&Y gains 32 and KPMG gains just two. KPMG's loss of 13 clients in 2002 is noteworthy, both in terms of KPMG's time series and that of its rivals (e.g., it is double PwC's largest year of client attrition). Of note, per the bottom row of table 5, other than Andersen, KPMG is the only major German audit firm to sustain a net loss of continuing clients for this time period.

Table 5. 
Market Share Analysis: German Auditor Changes from 1999 to 2003
  1. Table 5 accords retroactive treatment of the 1999 Coopers & Lybrand and Price Waterhouse merger to 1998. The “other” category includes Wollert-Elmendorff, acquired in 2002 by D&T (five companies from Wollert to D&T).

inline image

4.2.2. Multivariate Analysis.  Of the 2,407 (669) company-years (companies) in tables 3–5, there are 121 observations representing just one year of data for a company. Because we cannot discern an auditor change, we eliminate these companies, leaving 2,286 observations for 548 companies. We also eliminate the first year for each company, leaving 1,738 observations for 548 companies. Of these 1,738, Worldscope shows 259 auditor changes20, of which we exclude 96 relating to mergers, acquisitions or dissolutions (i.e., 51 that relate to the PwC merger, 5 to Deloitte & Touche's acquisition of Wollert-Elmendorff, and 40 to Andersen).21 As such, 1,642 company-years comprise the sample for our multivariate analysis, of which 163 change auditors.

Using these 1,642 observations, we estimate equation (3) via probit. This regression examines the likelihood a KPMG client switches in the subsequent year. We control for size, leverage, changes therein (DeFond [1992], Francis and Wilson [1988]), profitability, financial health, and industry. Our variable of interest, KPMGt−1*Year2002, is equal to one if KPMG is the auditor for calendar 2001 and the year is 2002, the first audited financial statement after ComROAD. A positive coefficient for this variable (β8) is consistent with the notion that clients “punish” a high-quality auditor caught “cheating” by providing low quality.

image(3)
Where:
AuditorChange =

1 if a change in the company's audit firm and 0 otherwise

Ln(Assets) =

Natural logarithm of total assets

%ΔAssets =

Percentage change in Assets

Leverage =

Total liabilities/total assets

ΔLeverage =

Change in Leverage

ROA =

Net income/total assets

Loss =

1 if net income is negative and 0 otherwise

KPMG =

1 if KPMG is the (prior year) audit firm and 0 otherwise

KPMG *Year2002 =

1 if KPMG is the audit firm for 2001 and current year is 2002 and 0 otherwise

Table 6 presents the results of estimating equation (3). The pseudo-R2 (likelihood ratio index) is 3.7%. With respect to controls, our findings indicate that auditor changes are more common for smaller, unprofitable companies. Consistent with the descriptive statistics in tables 4 and 5, the coefficient on our variable of interest, KPMGt−1*Year2002, is positive and significant at 5% (two-tailed).22,23 The marginal effect of KPMGt−1*Year2002's coefficient is 0.074, implying a 7.4% increase in the likelihood of an auditor change in calendar 2002 for those cases where KPMG is the incumbent for the December 2001 audit.

Table 6. 
Auditor Change Probit Regressions: Changes Away from KPMG Germany
inline image
VariableCoefficient (t-statistic)
  1. Observations are 1,642 years for 548 calendar-year companies with two or more years of consecutive data on Worldscope from 1998 to 2003, of which 163 (1,479) change (do not change) auditors. These 1,642 are a subset of the 2,407 company years in tables 3 and 4, after excluding: companies that appear on Worldscope for just one year; the first year for each company; and companies that change auditors due to mergers, acquisitions, or dissolutions (51 for the 1999 merger of Coopers & Lybrand and Price Waterhouse, 5 for the acquisition of Wollert-Elmendorff by Deloitte & Touche, and 40 for Andersen). Estimation is maximum likelihood probit with t-statistics in parentheses.

  2. ** and *** indicate statistical significance at the 5% and 1% levels, respectively (two-sided tests).

  3. Variables are as follows:

  4. AuditorChange= 1 if a change in the company's audit firm and 0 otherwise

  5. Ln(Assets)= Natural logarithm of total assets

  6. Assets= Percentage change in Assets

  7. Leverage= Total liabilities/total assets

  8. ΔLeverage= Change in Leverage

  9. ROA= Net income/total assets

  10. Loss= 1 if net income is negative and 0 otherwise

  11. KPMG= 1 if KPMG is the (prior year) audit firm and 0 otherwise

  12. KPMG*Year2002= 1 if KPMG is the audit firm for 2001 and current year is 2002 and 0 otherwise

Constant−0.35
(−1.06)
Ln(Assets)t−0.08
(−2.96)***
%ΔAssets0.01
(0.31)
Leveraget−0.15
(−0.79)
ΔLeveraget0.19
(0.54)
ROAt0.03
(0.13)
Losst0.26
(2.48)**
KPMG t −1−0.18
(−1.46)
KPMGt−1*Year20020.45
(2.19)**
Industry indicator variablesIncluded
Observations1,642
Psuedo-R2 (likelihood ratio index)3.7%

Because German audit fees are confidential, we cannot directly quantify the losses that KPMG sustains due to abnormal client attrition following ComROAD. However, given that extant literature demonstrates that company size is (by far) the primary determinant of audit fees (e.g., Menon and Williams [2001]), we can use assets to proxy for audit fees. The 13 KPMG clients for 2001 that change auditors for 2002 (see tables 4 and 5) represent 3.4% of client total assets for the 83 KPMG clients for 2001 that continue in our sample for 2002. Moreover, the total assets of these 13 companies is about 10 times the total assets of the 20 clients that change from KPMG during the three-year period 1999–2001 (see also tables 4 and 5). In addition to this, as table 4 shows, during 2002 KPMG audits just 5.3% (1 of 19) of the companies new to Worldscope, versus 16.7% and 15.6% for 2001 and 2000, respectively.24

4.3 characteristics of companies that change from/stay with kpmg

We take a closer look at KPMG's 83 calendar year-end clients for 2001, 13 of which change auditors for their 2002 audit. While significant and consistent with our prediction, the attrition we find may not be as large as some may expect (e.g., given a failure so severe that it is associated with a 3% drop in investor wealth, some may expect an exodus from KPMG).25

One barrier to auditor–client realignment is that Germany bans direct solicitation and advertising by audit firms, making it difficult for rival audit firms to lure KPMG's clients after ComROAD. To the extent these restrictions impede client–auditor realignments, an unintended consequence is that audit markets may not adequately adjust to revelations of declines in quality.

Another factor likely to hinder auditor switching is that the objective function of German managers may not be to maximize shareholder value. European companies in code-law countries such as Germany are subject to the demands of a broad set of stakeholders, including labor and banks. This means that there is likely to be cross-sectional variation in the demand for audit quality (e.g., some supervisory boards may choose to stay with KPMG because the costs to investors of retaining KPMG are offset by the benefits KPMG brings to other stakeholders). Following this, companies with less board representation by labor unions or banks and that recently raise public equity seem more likely to have stronger incentives to change from KPMG.

Lastly, auditor industry expertise and the severity of event-day returns could be factors that help explain why some KPMG clients change while others do not. Prior research finds evidence consistent with auditors who are industry leaders providing higher quality audits, where industry leadership is captured by client industry concentration (Francis, Reichelt, and Wang [2005]). This suggests that the propensity for a client to switch from KPMG may be a function of whether KPMG is the industry leader, although the direction of this impact is difficult to predict. For example, if KPMG is the industry leader, even a drop in the level of perceived audit quality in KPMG may still be higher than the quality of alternative auditors. This would suggest that there would be less switching away from KPMG when KPMG is the industry leader. On the other hand, clients with industry leaders may be particularly sensitive to drops in audit quality and be more likely to switch from KPMG as a result of a breach in providing the level of quality they expect. In addition, it seems likely that the larger the costs suffered by shareholders as a result of KPMG's audit failure, the stronger the incentives for managers to replace KPMG.

To investigate these rationales, we estimate equation (4a) and equation (4b) via probit, which examines the likelihood that a December 2001 KPMG client changes to a different auditor for their December 2002 audit. As with equation (3), we control for company size, leverage, and profitability. Because German companies with more than 500 employees must reserve 50% of the seats on their supervisory board for employee representation (Ashbaugh and Warfield [2003]), company size can proxy labor union representation on the board. We specify leverage to proxy for firms with debt financing, and hence creditor representation on the board. To proxy for companies that recently raise public equity, we include an indicator variable for whether a company is new to Worldscope since 1999, the year of ComROAD's IPO. As with equation (2), to proxy for industry specialization, we specify an indicator variable for whether the client is in the banking, insurance, or other financial industries. To consider the association between event-day returns and the likelihood a KPMG client changes auditors, we add the event-period returns in equation (4b). Because returns are not available for all KPMG's calendar 2001 clients, sample sizes for the equation (4a) and equation (4b) estimations are 83 and 64, respectively. Panel A of table 7 presents descriptive statistics and regression results for the sample of 83 KPMG calendar 2001 clients, whereas panel B presents the same for the sample of 64 clients.

image((4a))
image((4b))
Table 7. 
KPMG Germany's Calendar-Year 2001 Clients by Whether They Changed Auditor in 2002
Panel A: KPMG calendar-year 2001 clients (N= 83)
Variable Change from KPMG (N= 13)Stay with KPMG (N= 70)Tests of Differences (two-tailed p-values)
AssetsMean18,828100,8420.06
Median   527  1,7300.09
LeverageMean0.6630.6650.98
Median0.6590.6700.85
ROAMean−0.082−0.0280.40
Median−0.0020.0100.24
IPOMean0.6920.3430.03
FinancialMean0.3080.2570.73
inline image (4a)
VariableCoefficient (t-Statistic)
Constant0.60
(0.37)
Ln(Assets)−0.23
(−1.74)*
Leverage1.61
(1.53)
ROA0.41
(0.45)
IPO0.80
(1.91)*
Financial0.65
(1.27)
Observations83
Psuedo-R2 (likelihood ratio index)14.2%
Panel B: KPMG calendar-year 2001 clients with stock returns on Datastream (N= 64)
Variable Change from KPMG (N= 11)Stay with KPMG (N= 53)Tests of Differences (two-tailed p-values)
AssetsMean14,61591,6670.12
Median   527 1,8240.04
LeverageMean0.6360.6550.77
Median0.6590.6940.69
ROAMean−0.100−0.0350.39
Median−0.0380.0100.20
IPOMean0.7270.3580.03
FinancialMean0.2730.2830.95
Returns1 (Event 1)Mean−0.042−0.0130.13
Median−0.033−0.0070.21
Returns2 (Event 2)Mean−0.010−0.0070.73
Median−0.0080.0000.31
Returns3 (Event 3)Mean−0.025−0.0180.70
Median−0.008−0.0110.96
inline image (4b)
VariableCoef. (t-Stat.)
  1. For panel A, observations are 83 KPMG-audited, calendar-year German companies for 2001 on Worldscope for 2001 and 2002, of which 13 (70) change (do not change) auditors for their December 31, 2002 audit (i.e., the 100 KPMG clients in the “2001” column of table 4 minus the 17 companies that do not appear on Worldscope for 2002). For panel B, observations are the 64 of these 83 with stock returns on Datastream for the 252 trading days from November 3, 2001 to October 31, 2002. Tests of difference in means are two-sample t-tests and tests of difference in medians are Mann-Whitney U-tests. Estimation is maximum likelihood probit with t-statistics in parentheses.

  2. *indicates statistical significance at the 10% levels (two-sided tests).

  3. Variables are as follows:

  4. AuditorChange= 1 if a change in the company's audit firm (i.e., from KPMG) and 0 otherwise

  5. Ln(Assets)= Natural logarithm of total assets

  6. Leverage= Total liabilities/total assets

  7. ROA= Net income/total assets

  8. IPO= 1 if company is new to Worldscope in 1999 (i.e., the year of ComROAD's IPO), 2000, or 2001 and 0 otherwise

  9. Financial= 1 if a bank, insurer, or other financial company and 0 otherwise

  10. Returns1= Returns for event 1 (i.e., the three-day window around February 19, 2002)

  11. Returns2= Returns for event 2 (i.e., the three-day window around April 10, 2002)

  12. Returns3= Returns for event 3 (i.e., the four-day window around April 23–24, 2002)

Constant1.29
(0.63)
Ln(Assets)−0.27
(−1.73)*
Leverage1.36
(1.09)
ROA1.77
(1.51)
IPO0.84
(1.67)*
Financial0.49
(0.80)
Returns1−13.68
(−1.82)*
Returns24.24
(0.66)
Returns32.32
(0.51)
Observations64
Psuedo-R2 (likelihood ratio index)22.7%
Where:
AuditorChange

1 if a change in the company's audit firm (i.e., from KPMG) and 0 otherwise

Ln(Assets)

Natural logarithm of total assets

Leverage

Total liabilities/total assets

ROA

Net income/total assets

IPO

1 if company is new to Worldscope in 1999, 2000, or 2001 and 0 otherwise

Financial

1 if a bank, insurer, or other financial company and 0 otherwise

Returns1

Returns for the three-day window around February 19, 2002

Returns2

Returns for the three-day window around April 10, 2002

Returns3

Returns for the four-day window around April 23–24, 2002

Table 7 presents the results. Consistent with the descriptive statistics, the coefficients on Ln(Assets) and IPO are negative and positive at 10% (two-tailed), respectively, suggesting smaller, recently public companies are more likely to leave KPMG after ComROAD. As per panel B, the coefficient on Returns1 is negative and significant at 10% (two-tailed). This is consistent with the view that the larger the costs that shareholders sustain from ComROAD, the stronger the incentives for supervisory boards to replace KPMG. Among the companies that change from KPMG, those with the most negative Returns1 are smaller, recent IPOs (similar to ComROAD). This is consistent with the view that German supervisory boards more attuned to maximizing shareholder value are more likely to change from KPMG.

5. Conclusion

In a recent article, DeFond and Francis[2005, pp. 6, 13]“encourage researchers to address issues regarding the role and importance of litigation in maintaining high audit quality.” We examine the stock and audit market effects associated with an audit failure involving a public client of KPMG in Germany, a country that provides auditors with substantial protection from investor litigation. Our findings suggest that German investors and supervisory boards react to revelations of substandard audits by a firm with a reputation for high quality. From a stock market standpoint, investors appear to respond negatively to events relating to ComROAD, manifesting in cumulative negative returns of 3% for KPMG's other German clients. From an audit market standpoint, supervisory boards appear to respond by dropping KPMG, as KPMG's rate of attrition by their calendar-year clients doubles in the year of the ComROAD scandal.

Our findings are important because research has difficulty isolating the value of audit reputation to investors. We believe this difficulty is somewhat attributable to the research setting. By focusing on auditor reputation in relatively high shareholder litigation risk countries, it is difficult to separate reputation effects from insurance effects. In general, extant studies conclude the insurance effect dominates. By conducting our study in Germany, a major industrialized country with very low shareholder litigation risk, we abstract from the insurance rationale and are able to conduct a cleaner test of the reputation rationale for audit quality. We provide evidence that markets discipline auditors that deviate from providing high quality.

Given present-day calls for auditor litigation reform (see Scannell [2007]), policymakers may find our results of some interest. Though we emphasize that a caveat for our paper is that our findings and inferences, because they pertain to a severe audit failure in a country where auditor liability to shareholders is quite low, may be subject to limitations on external validity.

Footnotes

  • 1

    For example: per Norris[2004, p. D-1], “[i]f [Big 4] auditors are doing a good job, they deserve to be protected from lawsuits that could put them out of business. But without the threat of such suits, will they do a good job?”

  • 2

    Overall, Germany has the lowest “liability standard index” of the 49 countries in LaPorta, Lopez-De-Silanes, and Shleifer [2006], suggesting the most difficulty recovering losses from issuers, directors, distributors, and accountants.

  • 3

    Germany's Neuer Markt began operations in March of 1997 as a stock exchange for small and mid-sized growth companies, comparable to NASDAQ. In September of 2002, following the sharp decline in share prices (a 95% decline from its March 2000 peak), the Deutsche Börse announced it was closing the Neuer Markt.

  • 4

    Because ComROAD is contemporaneous with Enron, other reasons for switching are that clients may expect sanctions or KPMG to shed clients and do not relate to the quality of KPMG's audits. However, ComROAD occurs two years after another scandal involving a client of KPMG Germany (FlowTex Technologie GmbH, a private pipe and cable installation company, leased nonexistent drilling equipment to related parties). As such, KPMG's clients may expect sanctions or KPMG to shed clients for reasons that do relate to the quality of KPMG's audits.

  • 5

    For example, per Tirole's[1992, p. 122] discussion of Klein and Leffler [1981] and Shapiro [1983], “[the high-quality price] must be such that the monopolist would not want to rebuild his reputation if he lost it. To do so, he could sell for one period at zero (or slightly negative) price and high quality.”

  • 6

    In addition, KPMG's merger with Andersen arguably broke down due to ComROAD. Instead of KPMG, Andersen merged with Ernst & Young, which took 32 of the 40 Andersen clients in our sample (versus just 2 of 40 for KPMG). “[KPMG's] talks to merge with the German unit of Arthur Andersen LP fell apart in the wake of the Comroad scandal … The deal would have boosted KPMG revenues in Germany to about $1.8 billion and cemented its dominance in the country. But less than a week after KPMG disavowed Comroad's accounting, Andersen backed out, choosing instead to partner with Ernst & Young Deutschland.” (Byrnes and Ewing [2002]).

  • 7

    The shareholder settlements and fines Andersen paid out are $110 million for Sunbeam in May 2001 and $107 million for Waste Management in June 2001 ($7 million of which is an SEC fine, the largest ever).

  • 8

    We do not suggest that investor litigation was the ultimate cause of Andersen's demise, which is the result of the U.S. Justice Department's decision to criminally indict Andersen (precipitated by violation of a consent decree).

  • 9

    Recent legal literature also questions the reputation rationale for audit quality (e.g., Prentice[2006, p. 786–7]).

  • 10

    Per Gietzmann and Quick[1998, p. 92], “[t]he scope for third parties to pursue actions against [German] auditors was (and still is) very much restricted … the auditor is liable to a third party when violating a protective law. Rules require that an intentional violation be established. The effect of the rule is to severely restrict the usage of actions by third parties since it is typically extremely difficult to prove intentional violation given the nature of audits.”

  • 11

    Per ComROAD's June 20, 2002 shareholder letter, “… Rödl & Partner has presented a comprehensive final report on the special audit of ComROAD AG for the years 1998–2000. The report confirms the preliminary findings that ComROAD already reported in its ad hoc press releases on the 10th and 23rd of April 2002. Large portions of revenues recorded in previous years' financial statements were from sales to VT Electronics—a company for which no evidence could be found of its existence. The existence of the revenues could not be substantiated.”

  • 12

    Industry classifications for the 92 companies are: 65 industrials, 11 insurance, 10 other financial, and 6 banks. Industry classifications for the 67 companies are: 49 industrials, 7 insurance, 8 other financial, and 3 banks.

  • 13

    Because German exchanges operate similarly to those in the United States, the Schipper and Thompson [1983] methodology seems appropriate in our context (see Campbell, Goldberg, and Rai [2003] for a comparable application).

  • 14

    As table 1 shows, each of the three market indices accounts for a significant portion of the variation in returns. Our results are qualitatively similar if we combine these three market indices into one equally weighted index.

  • 15

    Of the 92 KPMG-audited companies in our sample, 13 trade as American Depository Receipts (ADRs), for which non-German liability regimes are likely relevant. For Event1 and Event3, there is no significant difference between the negative returns for these cross listed companies and the remaining 79 (e.g., market-adjusted mean returns for Event1 are −0.014 and −0.012 for ADRs and non-ADRs, respectively; and for Event3 are −0.017 and −0.010 for ADRs and non-ADRs, respectively). For Event2, the 79 non-ADRs have significantly negative market-adjusted mean returns of −0.014, in contrast to 0.005 for the 13 ADRs. Table 1 results do not meaningfully change if we exclude these 13 ADRs and re-estimate equation (1) (the only change is the coefficient on Event2, which goes from −0.00 to −0.01; t-statistic from −1.10 to −1.61). These results suggest that a loss of insurance coverage relating to ADRs is not driving our findings.

  • 16

    We supplement this with an approach that specifies dummy variables for each day within a given event window. Abnormal returns are negative for the first two days in the first window (when KPMG resigns as ComROAD's auditor), though significant only on February 18th. Abnormal returns are negative for the last two days in the second window (when ComROAD admits their 2000 financial statements contain material misstatements), though no daily return is significant. For the four days in the third window (when ComROAD announces the majority of their prior revenues are bogus and KPMG announces it will re-audit its Neuer-Markt clients), abnormal returns are statistically negative for April 22nd, insignificantly positive for April 23rd and 24th, and insignificantly negative for April 25th.

  • 17

    Sanctions against German auditors are uncommon. For example, per Baker, Mikol, and Quick[2001, p. 778]“[e]xclusion from the profession is relatively rare (only seven instances since 1961).”

  • 18

    The Pr(Bankrupt) variable also proxies for audit risk. In contrast to Baber, Kumar, and Verghese [1995], we are unable to specify an audit opinion variable as a proxy for audit risk because all the companies in our sample have “clean” opinions.

  • 19

    The timing of ComROAD compromises the use of these companies because to include them in the analysis we would need to make assumptions about when during the course of a year such companies decide to change auditors. Of the 52 companies (228 company-years) we exclude because they change their fiscal year-end, one auditor change involves KPMG: Buderus switched from KPMG for the December 2001 audit to PwC for the December 2002 audit (i.e., around the time of ComROAD). Of the 118 companies (378 company-years) we exclude because they are non–calendar year, four auditor changes involve KPMG: two companies switched to KPMG and two companies switched away from KPMG. W.E.T. Automotive Systems switched from an “other” auditor for the June 2001 audit to KPMG for the June 2002 audit and Eurobike switched from PwC for the September 2001 audit to KPMG for the September 2002 audit (i.e., prior to ComROAD). Carl Zeiss Meditec switched from KPMG for the September 2002 audit to an “other” auditor for the September 2003 audit and Suedzucker switched from KPMG for the February 2003 audit to PwC for the February 2004 audit (i.e., around the time of ComROAD). As such, inclusion of these companies probably strengthens the result that ComROAD negatively affects KPMG's German audit market share.

  • 20

    Table 5 presents 208 of these 259 auditor changes (i.e., 259 minus the 51 that relate to the PwC merger).

  • 21

    We exclude auditor changes from acquisitions and dissolutions because, following DeAngelo [1981], we test for an increase in the number of clients that switch from KPMG following ComROAD. Including auditor changes for D&T's acquisition of Wollert-Elmendorff and Andersen's dissolution does not affect the results in table 6.

  • 22

    Table 6 results do not meaningfully change if we specify year indicator variables. In addition, we also estimate a version of equation (3) that includes a variable equal to one if KPMG is the auditor for the calendar year 2002 audit and the current year is 2003. When we do this, the coefficient on KPMGt−1*Year2002 remains positive and the coefficient for variable equal to one if KPMG is the auditor for the calendar year 2002 audit and the current year is 2003 is negative.

  • 23

    We also study contemporaneous changes from KPMG in the United States. We do this because ComROAD/KPMG follows on the heels of Enron/Andersen and the increase in client defections for KPMG Germany may stem from contagion effects and not from a decline in audit quality in Germany per se. To obtain a U.S. sample, we identify 32,303 company-years for 6,801 calendar-year companies that appear consecutively in the Compustat database for at least two years with necessary data (these compare to the 2,286 company-years for 548 German companies). We eliminate the first year for each company, leaving 25,502 company-years for 6,801 companies, and 767 auditor changes for Andersen. As such, 24,735 companies comprise the U.S. sample, of which 2,151 change auditors in a subsequent year. When we estimate equation (3) using this U.S. sample, the coefficient for KPMGt−1*Year2002 is insignificant.

  • 24

    Of course, these statistics do not capture other aspects of the cost of the ComROAD audit failure to KPMG (e.g., lower fees on clients that stay with KPMG, problems in attracting and retaining talented employees, an increase in regulatory scrutiny, and losing out on the opportunity to merge with Andersen Germany).

  • 25

    Alternatively, given Barton [2005] reports that 95% of Andersen's clients do not switch until after the firm's indictment on March 14, 2002, the extent of change we observe may be larger than some might expect.

Ancillary