The influence of the financial crisis on the relation between Big 4 auditor industry specialization and the asymmetric timeliness of earnings

(1)

Amsterdam Business School

The influence of the financial crisis on the relation between Big 4

auditor industry specialization and the asymmetric timeliness of

earnings

Name: T.S. Schavemaker Student number: 10408789

Thesis supervisor: dr. G. Georgakopoulos Date: 01-06-2016

Word count: 13658

MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam

(2)

Statement of Originality

This document is written by student Tim Schavemaker who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

This study conducts research on the effect of the recent financial crisis on the greater asymmetric timeliness of earnings for bad news relative to good news for clients audited by specialist Big 4 auditors compared to clients audited by non-specialist Big 4 auditors. Prior literature found this more conservative accounting in the earnings of specialist auditors' clients. A contribution to literature is established by extending this prior 1990s based research with studying the differences of timeliness of bad news in earnings for specialists' versus non-specialists' audited firms between both crisis and post-crisis periods, and pre-crisis period. The findings are consistent with prior literature. Contrary to the expectations, I find less timely incorporation of economic gains in earnings of specialist auditors' clients versus non-specialist auditors' clients for the recent financial crisis compared with the pre-crisis period. Furthermore, it can be suggested that the recent financial crisis had no significant impact on the difference between asymmetric timeliness of earnings for bad news relative to good news for clients audited by specialist auditors

(4)

1 Introduction

In a 2006 discussion paper on an improved conceptual framework, the IASB (International Accounting Standards Board) and the FASB (Financial Accounting Standards Board) state that conservatism and prudence are not desirable qualities of financial reporting information (Hellman, 2008). However, Barker (2015) observed in his recent study strong disagreement between the IASB and its stakeholders, which include both academics and practitioners. Specifically, the notion that financial reporting information should be conservative is supported broadly by research literature based on empirical evidence suggesting a market demand for conservatism and an economic theory explaining this demand. Accounting practitioners have long embodied a conservative approach of accounting. Therefore literature aligns with accounting practice in this regard.

The IASB and the FASB expressed a preference for reporting without bias, so called neutrality, in their discussion paper (Hellman, 2008). Contrary, literature broadly interpreters conservatism on the basis of the study of Basu (1997). This interpretation of conservatism leads to downward-biased earnings because bad news is more timely reflected in earnings than good news. It is often argued in literature that bias and noise in the financial statements is limited by conservatism. For example, Lafond and Watts (2008) mention the governance mechanism of conservative accounting which implies that manipulation and overstatement of the financial reports by managers is reduced by conservatism. The view of the standard setters that conservatism and prudence decrease accounting information relevance therefore contradicts with literature.

More quickly recognizing publicly available bad news about future cash flows than good news (i.e. conservatism), is a fundamental feature of accounting earnings. This asymmetric timeliness of earnings has extensively been studied and more timely recognition of losses appears to be enforceable by the auditing practice. Also, audit firms' expertise is dependent for an audit's effectiveness. Krishnan (2005) found that Big 6 auditors’ industry specialism is associated with more timely incorporation of bad news in earnings. He argues that Big 6 auditors who are specialist both have the expertise to detect losses and the incentives to persuade their clients to timely incorporation. This study elaborates on previous research into the relation between auditor expertise and accounting conservatism.

The recent financial crisis can be considered as the worst of his kind since the Great Depression in the 1930s (Vyas, 2011). In the financial crisis more firms find themselves in a situation of

(6)

financial distress (Francis et al., 2013). Auditors and their clients both have different incentives regarding conservative reporting during these financially distressed times. Because of the different level of expertise of specialist auditors compared to non-specialist auditors, the timeliness difference in incorporation of earnings by clients of both auditor groups can have changed because of the crisis. Therefore, this research focuses on the question whether the recent financial crisis influenced the greater asymmetric timeliness of earnings for bad news relative to good news for clients audited by specialist Big 4 auditors compared to clients audited by non-specialist Big 4 auditors.

Firms in financial distress try to manipulate earnings upward to influence debt-holders and investors (Vichitsarawong et al., 2010; Gul, Srinidhi & Shieh, 2002; Defond & Jiambalvo, 1994). However, for auditors it is more beneficial to be conservative in these circumstances. Because specialized auditors possess, more than non-specialized auditors, the resources to make this possible I expect the recent financial crisis to have an enhanced effect on the greater asymmetric timeliness of earnings for bad news relative to good news for clients audited by specialist auditors compared to clients audited by non-specialist auditors, in comparing the crisis with the period before the crisis. Furthermore, I expect the same for the post-crisis period relative to the pre-crisis period because conservatism can reduce the increased litigation risks and contracting conflicts and specialist auditors have the expertise to help clients become more conservative in their reporting.

My sample consists of 34,813 North-American (US Dollar based) firm-year observations from 2003 till 2014 retrieved from the Compustat database. The sample consists of Big 4 auditors' clients only to exclude differences due to audit firm name and to focus solely on industry specialism. I use the top three industry (two-digit SIC) portfolio shares per Big 4 auditor per year as the specialists observations, calculated by audit fees per industry as a percentage of the total audit fees received by the auditor in that year. Audit fees are retrieved from the AuditAnalytics database. To compare the three periods I divide my sample into the pre-crisis (2003-2006), crisis (2007-2008) and post-crisis (2009-2014) periods.

My results confirm that clients of industry specialized auditors have in general greater asymmetric timeliness of earnings for bad news compared with good news. Contrary to the expectations, I find less timely incorporation of economic gains in earnings of specialist auditors' clients versus non-specialist auditors' clients for the recent financial crisis compared with the pre-crisis period. Possible explanations can be benefits of conservative reporting for financially distressed firms (Francis et al., 2013; Balakrishnan et al., 2016) and intensified focus on auditors

(7)

(Stice, 1991), strengthened by the higher level of expertise of specialist auditors. My results are not sufficient in finding a greater difference in timely incorporation of economic losses in earnings of clients of specialist auditors for both the crisis and the post-crisis periods compared to the pre-crisis period. It can be suggested that the recent financial crisis had no significant impact on the difference between asymmetric timeliness of earnings for bad news relative to good news for clients audited by specialist auditors compared to clients audited by non-specialist auditors.

My study builds upon the study of Krishnan (2005). Next to the fact that this study has been published all the way in 2005 and uses data from 1989 till 1998, it does not include the recent financial crisis which occurred after the study of Krishnan. This study contributes by continuing research on the ongoing debate about the desirability of conservative reporting by providing more insight into factors influencing the use of conservative reporting, namely hiring an industry specialized auditor. Also, the study further researches audit quality by seeing auditor industry specialism, like Krishnan (2005), as a dimension of audit quality rather than implicitly seeing Big 4 auditors homogeneously as one group with regard to audit quality. Furthermore, this study contributes by extending literature about auditor industry specialization, specifically its effects during financially distressed times. Finally, the study also provides more research into the important topic about the influences the recent financial crisis has had on accounting.

The study is organized as follows. Section 2 contains the literature review in which the background of this study is discussed, divided into paragraphs about conservatism, industry specialization and the financial crisis respectively, and my two hypotheses are developed. Next, section 3 provides the methodology where the used measures and the sample selection are explained. In section 4 I present the results of my regressions followed by a discussion. Last, section 5 concludes this study.

(8)

2 Literature Review and Hypotheses

2.1 Conservatism

In its extreme form accounting conservatism was traditionally defined by anticipating no profit, but anticipating all losses (Watts, 2003). Later on, literature defines the concept of accounting conservatism as “capturing accountant’s tendency to require a higher degree of verification for recognizing good news than bad news in financial statements” (Basu, 1997, p. 4) or, more focussing on requirements, ‘‘the differential verifiability required for recognition of profits versus losses” (Watts, 2003, p. 208). Furthermore, Basu (1997) argues that accounting conservatism can also be reflected by: stronger concurrent earnings-to-return association than concurrent cash flow-to-return association for bad news relative to good news; and more persistent unexpected earnings growth with at the same time more temporary and volatile decline of unexpected earnings. According to Wang, Hogartaigh and Van Zijl (2008), the IASB sees conservatism as a degree of caution in exercising judgments for making the estimates required under uncertainties in such a way that assets or incomes are not overstated and liabilities or expenses are not understated. However, they argue that the construct of conservatism does not have a clearly articulated and commonly agreed upon interpretation.

These definitions indicate that cash flows should have a proper level of verifiability before recognizing them in accounts (Basu, 1997). Degrees of conservatism vary because of different verification requirements for gains and losses. A higher degree of conservatism originates from a greater difference between gains and losses; less timely recognition of unrealized gains, more timely recognition of unrealized losses. Conservative accounting incorporates firms recognizing unrealized losses, by writing assets down or liabilities up, when a loss in value occurs, but firms waiting to recognize gains until there is objective evidence for the realization.

Watts (2003) provides four explanations for the increased use of conservatism in accounting over the past 30 years; especially important are contracting conflicts and litigation costs, to a lesser extent taxation and accounting regulation. First, with respect to the agency theory, contracts align the interests of managers and firms. To be enforceable in court contracts need to contain verifiable measures. Contracts should also be conservative by displaying an asymmetric verification requirement. In debt contracts, conservative measures reduce the likelihood of

(9)

managers overstating earnings and assets, and making liquidating dividend payments to shareholders at debt holders' expense. Based on the study of Watts (2003), Zhang (2008) suggests and finds that conservative borrowers have a higher chance of violating debt covenants. This makes debt covenants more strict and attributes to debt holders' monitoring capacity. Therefore, Zhang (2008) finds that lenders have a higher willingness to offer lower interest rates to conservative accounting borrowers than to more aggressive borrowers. This is in line with the study of Ahmed et al. (2002) who find that more conservative accounting firms have more favorable debt ratings and incur on average lower debt costs.

In compensation contracts, conservative measures reduce the likelihood of managers overstating earnings and assets by trying to receive a high bonus, and embracing negative net present value projects. Conservatism can function as a corporate governance tool by signaling shareholders on a timely basis for possible negative net present value projects and bad performance. Lafond and Watts (2008) mention the governance mechanism of conservative accounting which implies that manipulation and overstatement of the financial reports by managers is reduced by conservatism. According to them, this governance mechanism generates a demand for conservatism as to control management. Ball et al. (2000) argue that conservative accounting facilitates monitoring of managers and (debt) contracts and that it is an important feature of corporate governance. They state that costs for managers are higher when they try to reverse bad investments than when they continue with the good ones. Informed monitoring forces managers to undertake the costs and therefore accountants use income and book values that incorporate economic losses more timely than economic gains.

Second, Watts (2003) states that the likelihood and costs of shareholder litigation are higher with an overstatement of earnings and net assets than with an understatement. So, an aggressive accounting style leads to more lawsuits than a conservative style. This is an incentive for both managers and auditors to be more conservative. The implementation of the Sarbanes-Oxley act in 2003 meant a significant increase of liability for managers (Lobo and Zhou, 2006). Consequently, the SOX-act imposed severe penalties for top-level managers. Now, also for managers - besides for shareholders, as described in the previous paragraph - there is a stronger incentive for conservative accounting. With respect to the threat of litigation, also auditors have the incentive to account more conservative.

Third, firms can have the incentive to act conservative, by deferring gains, to reduce taxable income and therefore reduce the present value of their tax payments. A firm's incentive for conservatism depends on the country it is in because a firm's expected tax burden is being

(10)

determined by its country's regime. A positive relationship between a firm's tax burden and degree of conservatism has been suggested by prior research, according to Bushman and Piotroski (2006). However, the authors themselves find mixed and inconclusive results with respect to the influence of tax regimes on conservatism. Thus, taxation can, to a lesser extent than contracting conflicts and litigation costs, be seen as an explanation for using conservatism. Last, Watts (2003) argues that regulators and standard setters are likely to face more criticism if there is an overstatement of firms' net assets compared to a situation in which there is an understatement of net assets. So, with the help of accounting conservatism political costs of regulators and standard setters are reduced. Noncontracting parties value the constrains of conservatism on opportunistic payments to, for example, managers. This asymmetry in political costs resembles the asymmetry in litigation costs.

The accounting regulation explanation of Watts (2003) clearly is in conflict to developments in the regulatory environment. Namely, in a 2006 discussion paper on an improved conceptual framework, the IASB (International Accounting Standards Board) and the FASB (Financial Accounting Standards Board) state that conservatism and prudence are not desirable qualities of financial reporting information (Hellman, 2008). The IASB and the FASB expressed a preference for reporting without bias, so called neutrality, in their discussion paper. There is a limited amount of research to be found discussing disadvantages of conservatism.

Lev et al. (2005) provide an argument regarding information inefficiencies following conservatism. Due to the conservatism, stock of these particular firms will be systematically undervalued. This can influence investors in such a way that resources are misallocated. Conservative firms will therefore be troubled with high cost of capital. Furthermore, Lev et al. (2005) study the treatment of R&D expenses. Following the conservatism principle would mean firms to expense their R&D costs immediately. The disadvantage with this is in the fact that assets will initially be undervalued and later on be overvalued when a new product is launched. All development costs are already expensed and therefore the project seems to be making profit only while this is not the case. Penman and Zhang (2002) argue that managers are able to increase their earnings because of conservatism. Conservatism in depreciation of investments means expecting shorter useful life estimates. Therefore, depreciation will be accelerated and earnings reduced. The actual value of the assets is higher than the books want you to believe, so unrecorded reserves exist. By lowering investments managers can release these reserves and increase their earnings.

(11)

Barker (2015) observed in his recent study strong disagreement between the IASB and its stakeholders, which include both academics and practitioners. Specifically, the notion that financial reporting information should be conservative is supported broadly by research literature based on empirical evidence suggesting a market demand for conservatism and an economic theory explaining this demand. Accounting practitioners have long embodied a conservative approach of accounting. Therefore literature aligns with accounting practice in this regard. Literature broadly interpreters conservatism on the basis of the study of Basu (1997). It is often argued in literature that bias and noise in the financial statements is limited by conservatism. The view of the standard setters that conservatism and prudence decrease accounting information relevance therefore contradicts with literature.

2.2 Industry specialization

The factors influencing conservatism remain an interesting research topic because of the different views on the desirability of conservatism. Ball et al. (2000) see timely incorporation of losses as perhaps the single most important income reporting feature under common law. Identifying factors that enhance earnings timeliness, bad news in particular, has the interest of investors, analysts, the public and the earlier mentioned regulators (Krishnan, 2005, p. 209). Also, academics will be interested, indicated by the ongoing research on conservatism and because of the contrary views of regulators relative to academics. One factor influencing earnings timeliness, and therefore accounting conservatism, is empirically examined in this study, namely auditors' industry specialization.

The auditing practice plays an important role in enforcing timelier reflection of bad news about future cash flows relative to good news by earnings (Krishnan, 2005). This is a consequence of the fact that financial statements result from management's representations and auditor's assurance of validity to public. There are however several studies that suggest that variation in auditors' specialization is likely to change the effectiveness of auditing.

First of all, auditors with more experience in the manufacturing industry are better, relative to auditors with less experience, in identifying data-errors of clients in this specific industry (Bedard & Biggs, 1991). Krishnan (2005) argues in line with Bedard and Biggs (1991) basing it on the study of Maletta and Wright (1996) who find fundamental differences in methods of detection and error characteristics across industries suggesting an advantage in effectively detecting errors

(12)

for industry specialized auditors. Examples can be better assessments of loan loss provisions by banking industry specialized auditors and therefore improving reported earnings timeliness by appropriately recognizing economic losses, and better evaluation of appropriate provisions for warranties by retail industry specialists. The higher likelihood of detecting data-errors by industry specialists can also be found in the higher likeliness of specialists building databases with detailed industry specific best practices, risks and errors compared to the non-specialist auditors.

Second, specialization of auditors results in more auditor compliance with auditing standards if compared with not specialized auditors (O'Keefe et al., 1994). Furthermore, Johnson et al. (1991) find that an auditor's ability to detect fraud is enhanced by industry expertise. Also, specialized auditors are, compared to not specialized auditors, likely to invest more in information and audit technology, and staff recruitment and training in order to maintain their competitive advantage (Dopuch & Simunic, 1982 in Krishnan, 2005, p. 210). The study of Krishnan (2003) shows that absolute discretionary accruals are higher for clients of non-specialist auditors than for clients of specialist auditors, suggesting that auditor specialism moderates earnings management.

Specialized auditors can have particular incentives to find and report aggressive (non-conservative) accounting in addition to the above mentioned tools and expertise to actually find this aggressive practices. This is when auditing firms have a certain reputation following their brand-name. DeAngelo (1981) shows in her study that the big (in that time 8, now 4) accounting firms have a greater brand-name reputation. So, in an event of reputation loss, the big firms have more to lose than non-big firms due to the larger client base of big firms.

Combining the studies of Bedard & Biggs (1991), Krishnan (2005), Maletta and Wright (1996), O'Keefe et al. (1994), Johnson et al. (1991), Dopuch and Simunic (1982), Krishnan (2003) and DeAngelo (1981), it can be suggested that big firm auditors who are specialized possess both the resources and expertise to detect losses, and the incentives to encourage auditees to report them in a timely manner in their financial statements. Earnings' asymmetric timeliness is therefore likely to be affected by the industry specialism of the auditor. This study focuses on the question whether the tendency of firms to delay the recognition of economic bad news (in the form of losses) in earnings is mitigated by auditors' industry expertise.

Only big 6 auditors practice conservatism in accounting according to Francis and Krishnan (1999). Particularly, they issue modified audit reports for clients with a higher level of accruals. Francis, Maydew and Sparks (1999) find another difference between big 6 audit firms and non-big 6 audit firms relevant for this study, namely more constraining of accruals based earnings by big 6 auditors. From this it can be concluded that the big, higher quality auditors influence

(13)

asymmetric timeliness of earnings by persuading clients to recognize bad news more timely. To eliminate differences between big 4 and non-big 4 firms, and to isolate differences due to auditor industry specialism rather than differences due to brand name, this study will focus on firms audited by big 4 audit firms (Deloitte & Touche, EY, KPMG and PwC) only. This is in line with the study of Krishnan (2005)who also solely focuses on the big accounting firms, in that time the big 6.

2.3 Financial crisis

When taking the governance view of conservatism into account, Francis et al. (2013) expect and find that firms using more conservative accounting encounter smaller value losses compared to firms using less conservative accounting during the recent global financial crisis. They mention three main reasons for this finding which they extracted from prior literature.

First, they found a large amount of literature suggesting that more firms fall into a situation of financial distress in the case of a systematic crisis. Next to this, these firms will encounter increased information asymmetries and increased agency problems. As a consequence, firms will more aggressively manipulate their earnings leading to increased information-and agency risks on outside shareholders during periods of financial crisis. Inherent in accounting conservatism are some asymmetric verification requirements reducing earnings manipulation by managers. The requirements provide thus more reliable and transparent information to investors. By lessening information-and agency risks, conservative accounting mitigates firm value losses during crises. Furthermore, Francis et al. (2013) find evidence for a "flight-to-quality" during financial crises because a crisis could force investors to recognize the existing weaknesses in the quality of financial reporting. Reporting more conservative information signals that the quality of the accounting numbers is higher. So, it is expected that more conservative reporting firms experience a less reduced firm value during crises.

Third, firm value losses during a crisis can be prevented by conservatism's effect on real activities of firms, especially financing and investments. Behavior of excessive risk-taking is perceived as the biggest cause of the credit crisis and firms are generally short on credit and lack investment opportunities. Because of the positive impact of conservatism on the above, conservatism becomes more important in reducing firm value loss during a crisis. So, from the study of

(14)

Francis et al. (2013) it can be suggested that firms are intended to use more conservatism in their accounting during a financial crisis.

Balakrishnan, Watts and Zuo (2016) find a sharper decline of investment activity following the onset of the financial crisis for firms with less conservative reporting compared to firms with more conservative reporting. Furthermore, they find that more conservative reporting firms experience a lower decline in debt raising activity and in stock performance. The studies of Francis et al. (2013) and Balakrishnan et al. (2016) show more conservative financial reporting to be beneficial for firms during the financial crisis.

More conservative reporting during the financial crisis can also be desired by audit firms. When firms are in financial distress, auditors are more inclined to use a more conservative approach of accounting according to Stice (1991). This is because the situation of financial distress makes public focus on auditors increase. A going concern issue by an auditor will hurt a distressed firm badly and therefore brings the auditor-client relation in jeopardy. However, the market, society and regulators will punish an auditor when it fails to warn for bankruptcy.

However, some controversy exists about the question whether financial crises increase or decrease conservatism. Financially distressed firms are likely to use more aggressive accounting to upward manipulate earnings according to multiple other studies (Vichitsarawong et al., 2010; Gul, Srinidhi & Shieh, 2002; Defond & Jiambalvo, 1994). The influence of a financial crisis on the level of conservative accounting has been studied by Vichitsarawong et al. (2010). They looked at the 1997 financial crisis in Asia and found that there was less conservative accounting for earnings in the crisis period compared to normal economic periods. This is because firms in financial distress are more likely to issue good news compared to bad news to rebuild market's trust in the firm and reduce panic. The findings of Vichitsarawong et al. (2010) suggest a higher level of conservatism in the pre-crisis period compared to the crisis period. Gul et al. (2002) also studied the Asian financial crisis and came to the same conclusion as Vichitsarawong et al. (2010). They show that the economic downturn of Hong Kong was associated with a decrease in accounting conservatism.

The Asian studies provide some empirical evidence as for the question what effect an economic downturn has on the overall use of accounting conservatism. These studies show a decreased level of conservatism in this particular situation and environment. The studies of Francis et al. (2013) and Balakrishnan et al. (2016) are about the recent financial crisis and therefore more applicable for this study respectively. However, these studies do not provide empirical evidence

(15)

about the overall use of accounting conservatism in the recent financial crisis. They merely emphasize the benefits of more conservative accounting for firms.

Focusing on the situation of auditors' industry specialization, being central in this study, it is interesting to know what the incentives of audited firms are with respect to conservatism in the financial crisis. In a financial crisis there are more firms in financial distress (close to violating debt covenants) (Francis et al., 2013). Despite the benefits firms can have when applying a more conservative way of accounting, as shown in the papers of Francis et al. (2013) and Balakrishnan et al. (2016), firms attempt to actively upward manipulate earnings when in financial distress. This non-conservative behavior has been found by Defond and Jiambalvo (1994) by studying 94 firms that reported debt covenants violations in their annual reports. Debt covenants are arranged to restrict managers from engaging in financial decisions and investments that reduce debt holder claims' value. When we take the debt covenants into account, there are two reasons why managers make accounting choices (upward manipulate earnings) that make default less likely according to Defond and Jiambalvo (1994); debt covenants are frequently written in accounting numbers and debt covenant violation is costly. When default cannot be avoided by manipulating earnings, managers will still manipulate earnings in a positive direction to improve their bargaining position in the event of renegotiation. Furthermore, upward manipulation of earnings signals investors that the firm is doing better instead of facing bankruptcy in the foreseeable future. If a firm is in financial distress it is less attractive for investors to invest in this firm because a positive return on their investment is less likely. Firms in financial distress therefore desire a more lenient attitude from the auditor towards the recognition of earnings. Vyas (2011) states that litigation costs have increased following the financial crisis because the crisis triggered a shareholder class action litigation against financial institutions. He suggests that this increase continues after the crisis. Another interesting study applicable to the post-crisis effects is the study of Fahlenbrach and Stulz (2011). They found that banks with better alignment between shareholders' interests and CEO incentives performed worse during the crisis. In the post-crisis period they found no evidence that these banks performed better. This indicates that contracting conflicts about compensation were severe after the financial crisis. Because conservatism has been argued to provide benefits for corporate governance, like reducing litigation costs, and for contracting conflicts, it may be expected that firms desire a higher degree of conservatism after the crisis period than before the crisis period. Table 1 provides an overview of, for this study the most important, seminal papers used as theoretical background in this study

(16)

Table 1. Literature Overview

Author(s) Year Research matter Important for this study

Bedard & Biggs 1991 The effect of auditors' domain-specific experience on evaluation of management representations

Auditors with more experience in certain industries are better in identifying data-errors in this industries

Maletta & Wright 1996 Industry error characteristics in

audit evidence Advantages in detecting errors for specialized auditors O'Keefe, King &

Gaver 1994 Auditor industry specialization and compliance with GAAS reporting standards

Specialization of auditors results in more compliance with auditing standards

Johnson, Jamal & Berryman

1991 Effects of framing on auditor decisions

Auditor's ability to detect fraud is enhanced by industry expertise Krishnan 2003 Big 6 auditor industry expertise

and earnings management

Discretionary accruals are higher for clients of non-specialist auditors

Francis &

Krishnan 1999 Accruals and auditor reporting conservatism Big 6 auditors only, modify report of firms with high level of accruals

Francis, Maydew

& Sparks 1999 Big 6 auditors and credible reporting of accruals More constraining of accruals based earnings by Big 6 auditors Francis, Hasan &

Wu 2013 Benefits accounting during the recent of conservative financial crisis

Firms using more conservative accounting encountered less value losses during the financial crisis

Balakrishnan, Watts & Zuo

2016 The effect of conservatism on corporate investments during the crisis

Firms with less conservative reporting experienced a sharper decline of investment activity

Stice 1991 Pre-engagement factors

associated with lawsuits against auditors

Auditors are more inclined to use a more conservative approach with financially distressed firms

Vichitsarawong,

Eng & Meek 2010 The impact of the Asian financial crisis on conservatism Less conservatism in crisis period vs. non-crisis period DeFond &

Jiambalvo 1994 Debt covenant violation and manipulation of accruals Firms in financial distress try to manipulate earnings upward Vyas 2011 The timeliness of accounting

write-downs due to the financial crisis

Litigation costs have increased due to the crisis

Fahlenbrach & Stulz

2011 Bank CEO incentives and the credit crisis

Indication of severe contracting conflicts in the after math of the crisis

Krishnan 2005 The association between auditor industry expertise and asymmetric timeliness of earnings

More timely reflection of clients' bad news with specialist auditors vs. non-specialist auditors

(17)

2.4 Hypotheses development

The studies of Bedard & Biggs (1991), O'Keefe et al. (1994), Johnson et al. (1991), Dopuch and Simunic (1982) and DeAngelo (1981) suggest that big firm auditors who are specialized possess both the resources and expertise to detect losses and the incentives to encourage auditees to report them in a timely manner in their financial statements. Earnings' asymmetric timeliness is therefore likely to be affected by the industry specialism of the auditor. Krishnan (2005) proves this suggestion to be right. Like aforementioned, he found more timely reflected bad news in the earnings of clients of specialized auditors than in the earnings of clients of non-specialized auditors

This relation between auditors' industry specialization and accounting conservatism can have been affected by the recent financial crisis. In the financial crisis more firms find themselves in a situation of financial distress (Francis et al., 2013). Firms in financial distress try to manipulate earnings upward to influence debtholders and investors positively (Defond & Jiambalvo, 1994). Manipulating earnings upwards is a non-conservative accounting style. The client desires a lenient attitude from its auditor. The auditor is therefore more likely to be influenced or perceive pressure (because of its client-auditor relation) to recognize losses in a less timely manner. However, compared to clients of non-specialist auditors, clients of industry specialized auditors are normally more inclined to report in a conservative manner. Furthermore, Stice (1991) suggests it to be beneficial for auditors to remain conservative when auditing financially distressed firms. The specialist auditors have the expertise and incentives to maintain conservative reporting on a higher level.

Combining the above, I do expect that during the recent financial crisis industry non-specialists clients' use of conservatism was reduced more. So, during the crisis the difference in the degree of conservative accounting used by clients of specialist relative to non-specialist auditors is expected to increase. I derive the following hypothesis for the crisis period relative to the pre-crisis period:

Hypothesis 1: During the recent financial crisis the greater asymmetric timeliness of earnings for bad news relative to good news for clients audited by specialist auditors compared to clients audited by non-specialist auditors was enhanced relative to the period before the crisis.

(18)

To see the consequences of the financial crisis for the period after the crisis, I compare the post-crisis period to the pre-post-crisis period. Vyas (2011) and Fahlenbrach and Stulz (2011) show that litigation risks and contracting conflicts increased for the post-crisis period compared to the pre-crisis period. In reducing these problems firms could use more accounting conservatism. The auditor is more likely to be influenced or perceive pressure (because of its client-auditor relation and for specialists to remain attractive compared to non-specialists) to recognize losses in a more timely manner. Now, firms and auditors have the incentives to report more conservative. As explained earlier, specialist auditors have the expertise to do so more than non-specialist auditors have. Therefore, I derive the following hypothesis for the post-crisis period relative to the pre-crisis period:

Hypothesis 2: After the recent financial crisis the greater asymmetric timeliness of earnings for bad news relative to good news for clients audited by specialist auditors compared to clients audited by non-specialist auditors was enhanced relative to the period before the crisis.

The difference between the two hypotheses is that in the first hypothesis the difference between the two groups of clients increases because clients of non-specialist auditors are suggested to become less conservative to a larger extent than clients of specialist auditors, while under the second hypothesis the difference between the two groups of clients increases because clients of specialist auditors are suggested to become more conservative to a larger extent than clients of non-specialist auditors.

(19)

3 Methodology

3.1 Measuring conservatism

To measure financial reporting conservatism I will use, like Krishnan (2005), the method of Basu (1997). According to the survey based study from Wang et al. (2008) the asymmetric timeliness measure of Basu is extensively used in literature. In their study, the Basu measure was the most frequently used measure; 36 of the 53 papers they reviewed used it. They state three strengths of the measure. First of all, the wide appliance and the fact that it was the only measure used in literature to operationalize the asymmetric timeliness of conservatism for almost nine years. Second, researchers' confidence in the measure has increased because of the asymmetric timeliness measure of Basu producing results that are consistent with their predictions. Last, manifested by the use of the measure in very large scale international comparative studies, the measure is well adapted to large-sample cross-sectional analysis. There are researchers suggesting the asymmetric timeliness measure is biased (Wang et al., 2008). However, controversy exists about the question whether it is biased upward or downward and to what extent.

For the operationalization of conservatism in accounting, Basu (1997) focused on the implication of more quickly recognized bad news than good news in earnings. He was the first to link asymmetric timeliness of earnings with accounting conservatism (Wang et al., 2008). According to his measure, greater asymmetric timeliness of earnings means a greater degree of accounting conservatism within a firm. Basu (1997) used annual stock return to proxy for good and bad news.

Basu (1997, p. 13) regression:

Xit/Pit-1 = α0 + α1DRit + β0Rit + β1Rit × DRit (1)

The variable definitions are as follows: i represents the firm. Xit is the earnings per share for the fiscal year t. Pt-1 is the price per share at the beginning of the fiscal year. R is the annual return on stock. DR is a dummy variable; =1 if R < 0, =0 if otherwise. An explanation of the dummy variable will be given below.

In essence, accounting earnings (X/P) are being regressed on stock returns (R) separately for good-news firm years and bad news firm years (Wang et al., 2008). I calculate R using begin and end of year share prices and with taking dividends per share into account; ((Pt - Pt-1 + Div per

(20)

share) / Pt-1) (Berk & DeMarzo, 2011, p. 253). Incorporating dividends in the stock return calculation has also been done by Ahmed et al. (2002) and Ball et al. (2000). A stock return that is positive or equal to zero indicates a good-news firm year (Rit ≥ 0), a negative stock return indicates a bad-news firm year (Rit < 0). The estimated slope coefficient measures the timeliness of recognizing the news embodied in the stock return in earnings, conditional on the nature of the news. Krishnan (2005) excluded the first three months of (and included the three months following) the fiscal year to show only that part of the stock return that includes news. However, Basu (1997) compared his model using the stock return Krishnan (2005) uses and a measure with fiscal year stock returns and found no differences.

A dummy variable (D) separates good-news from bad-news. This allows for a difference in the slope coefficients and intercepts between the two. D equals 0 under good-news (Rit ≥ 0), the timeliness coefficient is now β1. D equals 1 under bad-news (Rit < 0), the timeliness coefficient is now β0+β1. β1 “measures the difference in sensitivity of earnings to negative and positive returns” (Basu, 1997, p.13). A higher β1 indicates more conservatism. Economic gains are captured by β0 which indicates the sensitivity to gains. The sensitivity ratio: (β1 + β0)/β0, describes a comparison between the total sensitivity of negative and positive returns on earnings. A higher ratio means higher sensitivity of bad news on earnings.

3.2 Measuring auditor industry specialization

Industry auditor expertise is unobservable and therefore there must be relied upon proxies to estimate it. Audit firm industry expertise is being estimated by the proportion of an audit firm's audit fees earned from one industry relative to all the industries the specific firm audits. This portfolio shares measure has been developed by Yardley et al. (1992) and was used in the study of Krishnan (2005). However, during the study of Krishnan (2005), data from the years 1989-1998, audit fees were not available. Therefore, he used square root of total assets instead. Now, audit fees are available and will therefore be used. Following, Dunn and Mayhew (2004), all companies within each two-digit primary SIC (Standard Industry Classification) code in the Compustat database are defined as an industry. Like mentioned before, to eliminate differences between big 4 and non-big 4 firms, this study will focus on the big 4 audit firms (PwC, Deloitte, E&Y and KPMG) only. This is in line with the study of Krishnan (2005) who only studied the big accounting firms, in that time the big 6. Following Krishnan (2005), I calculate auditors'

(21)

specialism for each year that is included in this study since their specialism may change over time. An auditor has incentives to preserve its competitive advantage from the audited industries that generate the highest income for the auditor. The top three portfolio's of an auditor are the auditor's specialties because these are its top three sources of income. I calculated per Big 4 auditor its audit fees per two-digit industry per year and calculated the percentages of the auditor's total audit fees received in that year. I code the remaining industries as the auditor's non-specialties. These steps are repeated for each Big 4 auditor.

A variable for specialist or non-specialist auditors will be implemented in model 1. The variable SPEC equals 1for firms audited by a specialist firm and the variable equals 0 for firms audited by non-specialist firms.

Xit/Pit-1 = α0 + α1DRit + α2SPECit + α3DRit × SPECit + β0Rit + β1Rit × DRit + β2Rit × SPECit + β3Rit × DRit × SPECit (2)

Important in model 2 is the variable: Rit × DRit × SPEC. Greater asymmetric timeliness of earnings is associated with clients of specialist audit firms relative to clients of non-specialist audit firms when β3 > 0.

Previous studies also used an alternate measure for auditors' industry expertise, namely an auditor's market share per industry. I did not use this measure following the reasons of Krishnan (2003). He argues that despite the fact that portfolio shares and industry market shares are highly correlated, industry market shares may be a more noisy measure. Speciality industries under the portfolio shares measure are also speciality industries under the industry market shares measure but not the other way around. Furthermore, Krishnan's sample based on the industry market shares measure comprises of 51% of clients from speciality industries while under the portfolio shares measure it is only 12%. Therefore it appears that the industry market shares measure overstates an auditor's specialism.

3.3 Sample Selection

I use archival data to find answers to the two hypotheses. Data from the Compustat database and the AuditAnalytics database have been combined. Firm specific data has been obtained from the Compustat Capital IQ North American database. These annual fundamentals are obtained for the years 2002-2014. The year 2002 has been included to calculate the share price in the

(22)

beginning of 2003; observations in the year 2002 are removed later on. I restricted my sample to US dollars only; therefore Canadian dollars were excluded. "Earnings per share (basic) excluding extraordinary items" accounts for Xit. For Pt I used "share price close". For Pt-1 I used "share price close" from the previous year and for the dividends per share I used "Dividends per share Pay Date - Fiscal". Audit fee and auditor name are obtained from AuditAnalytics. In Stata, the sample is matched with Compustat firm financial data with the help of CIK company identifiers and fiscal years.

After removing not fully merged data, and in doing so removing data from 2002, the dataset consists of 102,268 observations. I removed 29,008 observations with missing data. I exclude financial institutions, insurance companies and real estate firms (SICs between 6000 and 6999), amounting to 15,046 observations. This is in line with the studies of Krishnan (2005) and Vichitsarawong et al. (2010) who argue that their high level of financial distress in the sample period makes them less comparable with firms operating in other segments. Furthermore, I removed non-Big 4 audited firms (21,968 firm-year observations). Following Basu (1997), Krishnan (2005) and Vichitsarawong (2010), I removed the top and bottom 1% of firms regarding earnings per share and stock returns. This is to minimize extreme observations' effects on regression results. This leaves a final sample of 34,813 observations. In creating the dummy variable for stock returns (DR) it was clear that 14,133 observations have a negative stock return; the remaining observations have a positive return or a return equal to 0. My dataset contains 9,834 firm-year observations for auditor industry specialism accounting for 28.25% of the total amount of observations. Table 2 presents firm-year observations per audit firm.

Table 2. Big 4 Auditor Observations

Auditor Name Freq. Percent Cum. Deloitte & Touche LLP 7,616 21.88 21.88

Ernst & Young LLP 10,898 31.30 53.18

KPMG LLP 7,355 21.13 74.31

PricewaterhouseCoopers LLP 8,944 25.69 25.69

Total 34,813 100.00

Number of firm-year observations per Big 4 auditor.

I use dummy variables to control for the effect of the financial crisis on this study's model. I will divide my sample into three sub-samples. Beltratti and Stulz (2009) and Fahlenbrach and Stulz (2011) argue the beginning of the crisis, the year 2007, to be uncontroversial. Both studies take

(23)

December 2008 as the end of the crisis due to poor performance of banks. Francis et al. (2013) show that the S&P 1500 index declined continuously between 2007 and 2009. Beltratti and Stulz (2009) and Fahlenbrach and Stulz (2011) admit that performance continued to be poorly in the first quarter of 2009 but argue resolution mechanisms and the uncertainty about nationalization to be the cause. Therefore, they think it be better to evaluate returns until the end of 2008. The continued poor performance of banks in the beginning of 2009 can be confirmed by Francis et al. (2013) who show the decline bottomed out and then experienced an upturn in March 2009. Furthermore, Vyas (2011) and Balakrishnan et al. (2016) use the period 2007-2008 as the crisis period. Kapan and Minoiu (2015) also use this timeframe as the crisis period, using October 2008 as the end of the crisis and the period thereafter as the "after" period.

Based on the above, I will use the period 2007-2008 as the crisis period. To measure the period before the crisis I will use the period 2003-2006. This is to exclude the shock of the 9/11 disaster, the scandals relating to Enron and Worldcom in 2001 and 2002, and pre-SOX act years respectively. I use the years 2009-2014 as the post-crisis period.

I develop model 3 to run the regression on the pooled samples of pre-crisis period and crisis period, together with a dummy variable Within. The dummy variable Within is used to compare the pre-crisis period with the crisis period. The variable Within equals 1 if the sample firm reports earnings in the crisis period and equals 0 if the firm reports earnings in the pre-crisis period. The coefficient of "Within × γ0" measures the difference in use of accounting conservatism between the crisis period and pre-crisis period. If "Within × γ0" is negatively significant the first hypothesis will be supported, meaning that there was less conservatism in the crisis period compared to the pre-crisis periods.

Xit/Pit-1 = α0 + α1DRit + α2SPECit + α3DRit × SPECit + β0Rit + β1Rit × DRit + β2Rit × SPECit + β3Rit × DRit × SPECit + Within × (γ0 + γ1DRit + γ2SPECit + γ3DRit × SPECit + δ0Rit + δ1Rit × DRit + δ2Rit × SPECit + δ3Rit × DRit × SPECit) (3)

The Within variable (2007-2008) contains of 5,443 firm-year observations, as can be seen in Table 3.

I develop model 4 to run the regression on the pooled samples of pre-crisis period and post-crisis period, together with a dummy variable Post. The dummy variable Post is used to compare the pre-crisis period with the post-crisis period. The variable Post equals 1 if the sample firm reports earnings in the post-crisis period and equals 0 if the firm reports earnings in the pre-crisis period. The coefficient of "Post × γ0" measures the difference in use of accounting conservatism between the post-crisis period and pre-crisis period. If "Post × γ0" is positively

(24)

significant the second hypothesis will be supported, meaning that there was more conservatism in the post-crisis period compared to the pre-crisis periods.

Table 3. Fiscal Year Observations

Fiscal Year Freq. Percent Cum.

2003 3,782 10.86 10.86 2004 3,627 10.42 21.28 2005 3,311 9.51 30.79 2006 3,111 8.94 39.73 2007* 2,884 8.28 48.01 2008* 2,559 7.35 55.36 2009 2,689 7.72 63.09 2010 2,617 7.52 70.61 2011 2,588 7.43 78.04 2012 2,551 7.33 85.37 2013 2,533 7.28 92.64 2014 2,561 7.36 100.00 Total 34,813 100.00

Number of firm-year observations per fiscal year. * Crisis period years, together 5,443 firm-year observations.

Xit/Pit-1 = α0 + α1DRit + α2SPECit + α3DRit × SPECit + β0Rit + β1Rit × DRit + β2Rit × SPECit + β3Rit × DRit × SPECit + Post × (γ0 + γ1DRit + γ2SPECit + γ3DRit × SPECit + δ0Rit + δ1Rit × DRit + δ2Rit × SPECit + δ3Rit × DRit × SPECit) (4)

For illustrative purpose I state Table 4 which shows the industry specialization (top three industries) and portfolio shares for each Big 4 auditor in 2014. So, 13.23% of Deloitte and Touche's total revenues seem to came from the Business Services industry. The top three portfolios of Deloitte and Touche (Business Services, Electric, Gas and Sanitary Services, and Chemicals and Allied Products) amount to 31.29% of its total revenues.

I compare the magnitude of the interaction variable coefficient, Rit × DRit, between non-specialist and non-specialist auditors' clients by separately estimating model (1) for both groups. Like Krishnan (2005), I perform cross-sectional regressions for models (1) and (2) for every year in the sample to mitigate cross-sectional correlation. Model (1), for both client groups, and model (2) are estimated for the three periods pre-crisis, within crisis and post-crisis to compare the different periods. Coefficients of these three periods are the mean coefficients for this particular period, calculated by adding up each year's coefficients based on a fixed number of average observations per year. The average number of observations is 2,081.6 per year for non-specialist

(25)

clients and 819.5 for specialist clients. I use the study of Fama and Macbeth (1973) to test the significance of parameter estimates with the help of t-statistics for the annual estimates' cross-temporal distributions. So, I report t-statistics for the mean coefficients to test significance. These are calculated by the ratio of my mean estimate of the coefficient to the standard error of the distribution of each year's coefficient, divided by the square root of the number of years in this period. Furthermore, I run pooled cross-sectional regressions for both model (3) and model (4) to measure the differences between pre-crisis and crisis periods and between pre-crisis and post-crisis periods.

Table 4. Big 4 Auditor Portfolio Shares for 2014

Auditor Industry (SIC code) Portfolio

shares in %

Deloitte & Touche Business Services (73) 13.23

Electric, Gas and Sanitary Services (49) 10.56 Chemicals and Allied Products (28) 7.50 Ernst & Young Electronic and other Electrical Equipment and

Components, except Computer Equipment (36) 9.32

Business Services (73) 9.16

Industrial and Commercial Machinery and

Computer Equipment (35) 8.41

KPMG Chemicals and Allied Products (28) 12.92

Communications (48) 8.05

PricewaterhouseCoopers Chemicals and Allied Products (28) 16.39 Electronic and other Electrical Equipment and

Components, except Computer Equipment (36) 9.38

Illustrative example - Top three portfolio shares per Big 4 auditor in 2014. Percentages per two-digit SIC code industry of total audit fees received per auditor. SIC codes retrieved from http://siccode.com/

(26)

4 Results

4.1 Descriptive Statistics

In Table 5 descriptive statistics for the sample firms are presented. About 28% of the total sample firm-year observations were audited by specialist auditors. Percentage of firms with a negative return is two times higher for firms in the crisis period compared to firms outside the crisis period for both clients of specialist auditors as for clients of non-specialist auditors. This indicates that the time-frames of the periods are valid; more firms fall into a situation of financial distress in the case of a systematic crisis (Francis et al., 2013). The annual returns' means and medians support this assumption by also presenting a downturn during the crisis years. Earnings seem to increase over the whole period indicated by the mean values and the median of firms audited by a specialist. Notable is the standard deviation for the post-crisis period of clients of non-specialist auditors, which is higher than the other standard deviations.

Table 5. Descriptive Statistics

Clients of Non-specialist

Auditors Clients of Specialist Auditors

Variable

Pre-crisis Crisis Post-crisis crisis Pre- Crisis Post-crisis Number of observations 9,812 3,900 11,267 4,019 1,543 4,272

% of loss firms 27.50% 29.62% 28.45% 38.12% 34.93% 32.09%

% of firms with negative return 34.53% 69.79% 35.55% 35.71% 71.03% 34.81% Earnings Mean -0.04 -0.01 -0.00 -0.11 -0.05 -0.05 Std. 1.37 0.28 5.78 1.23 0.55 1.55 P25 -0.01 -0.02 -0.02 -0.08 -0.05 -0.03 Median 0.04 0.04 0.04 0.02 0.03 0.04 P75 0.07 0.06 0.07 0.06 0.06 0.07 Annual Returns Mean 0.27 -0.14 0.23 0.33 -0.16 0.25 Std. 0.66 0.51 0.64 0.80 0.47 0.65 P25 -0.11 -0.48 -0.12 -0.13 -0.48 -0.11 Median 0.15 -0.22 0.13 0.14 -0.22 0.14 P75 0.46 0.07 0.41 0.49 0.04 0.43

This table provides descriptive statistics for variables Earnings and Annual Returns. Firm-year observations for 2003-2006 (pre-crisis), 2007-2008 (crisis) and 2009-2014 (post-crisis).

(27)

Table 6 presents the Pearson correlation matrix of the variables. The within-crisis period and the post-crisis period both significantly correlate to return and the dummy for negative returns. Not surprisingly it can be seen that the Within dummy is negatively correlated with return and positively with DR; for Post vice versa. This could indicate more negative returns during the crisis period and more positive returns during the period after the crisis.

Table 6. Correlation Matrix

Earnings Return DR SPEC Within Post Size

Earnings 1.0000 Return -0.0161* 1.0000 0.0027 DR -0.0084 -0.6088* 1.0000 0.1154 0.0000 SPEC -0.0072 0.0176* 0.0033 1.0000 0.1821 0.0010 0.5333 Within 0.0016 -0.2234* 0.2590* 0.0010 1.0000 0.7638 0.0000 0.0000 0.8580 Post 0.0048 0.0543* -0.0961* -0.0151* -0.3865* 1.0000 0.3721 0.0000 0.0000 0.0049 0.0000

This table provides Pearson correlations between selected variables. * Correlation coefficient is significant at the 0.01 level.

4.2 Regression Results

Table 7 presents pre-crisis annual regressions' mean coefficients and adjusted R-squares for model (1) and (2). Results for clients of non-specialist auditors and specialist auditors are presented separately in Panel A and Panel B respectively. Panel C represents the combined sample using model (2). As can be seen in Table 7 the adjusted R-squared in Panel A is very low. Suggesting that only 0.64% of the variation in earnings is explained by the independent variables. The adjusted R-squared for the regression of specialist auditors' clients is with 4.11% much higher. This suggests that returns of clients of specialist auditors explain variation in earnings more than returns of clients of non-specialist auditors do.

In line with Basu (1997) and Krishnan (2005), the intercept terms are positive, and highly significant (Panel A and C)/significant (Panel B). Basu (1997) states this to be consistent with currently recognized unrealized gains from previous periods that are uncorrelated with current news; these gains are already reflected in prior stock returns. So, this is expected if recognition of unrealized gains, and not unrealized losses, is postponed to later periods.

(28)

The interaction variable, Rit × DRit, coefficient measures the incremental sensitivity of earnings to negative over positive returns. The positive and highly significant results for the period 2003-2006 suggest more sensitive earnings for negative returns relative to positive returns. This is consistent with Basu's interpretation of accounting conservatism; more timely reporting of bad news by earnings about future cash flows than good news. In comparing Panel A and Panel B it can be seen that the interactive slope coefficient, β1, is greater for specialist auditors' clients than for non-specialist auditors' clients, which is in line with the results of Krishnan (2005). Clients of auditors with more industry expertise have a higher degree of sensitivity for earnings to bad news. More specifically, they are [(1.126-0.182)/(0.668-0.174)] 91% more sensitive to negative returns than to positive returns. Krishnan also found more sensitive specialist auditor clients, but his difference amounted to only 12%.

Table 7

Mean Coefficients and Adjusted R²s from Annual Cross-Sectional Regressions: Pre-Crisis Period 2003-2006.

Panel A: Model 1 (Clients of non-specialist auditors) N=9812

α0 α1 β0 β1 Adj. R² 0.063 (2.88)*** -0.051 (-1.17) -0.174 (-6.12)*** 0.668 (5.41)*** 0.64%

Panel B: Model 2 (Clients of specialist auditors) N=4019

α0 α1 β0 β1 Adj. R² 0.030 (2.09)** -0.002 (-0.06) -0.182 (-9.48)*** 1.126 (16.13)*** 4.11% Panel C: Model 2 N=13831 α0 α1 α2 α3 β0 β1 β2 β3 AdjR² 0.063 (2.88)*** -0.051 (-1.17) -0.033 (-0.88) 0.053 (0.72) -0.174 (-6.12)*** 0.668 (5.41)*** -0.082 (-1.80)* 0.458 (2.32)** 0.90%

The mean coefficients and adjusted R-squares are calculated by adding up coefficients of the years 2003-2006 based on a fixed number of average observations per year. Between parentheses are t-statistics for the annual estimates' cross-temporal distributions to test the significance of parameter estimates. These are calculated by the ratio of my mean estimate of the coefficient to the standard error of the distribution of each year's coefficient, divided by the square root of the number of years in this period. ***, **, * Significance at the 0.01, 0.05 and 0.10 level, respectively, for a one tailed test.

(29)

In the combined sample in Panel C, β2 is negative and marginally significant. This means, for clients of specialist auditors, a negative incremental sensitivity of earnings to good news. It can therefore be suggested that there is less timely incorporation of economic gains in earnings of specialist auditors' clients versus non-specialist auditors' clients. This can be explained by the greater concern for timely loss relative to gain recognition by auditors.

Furthermore, the significant β3 in Panel C tells us that there is a statistically significant different slope coefficient regarding bad news for clients of non-specialist auditors and clients of specialist auditors. The fact that β3 is positive tells us that the slope coefficient is higher for the specialist clients. Overall, the results show for all important variables a significant coefficient. Also, these coefficients are all in the same direction (positive or negative) as the coefficients in the results of Krishnan (2005). The coefficients from both models suggest that firms audited by a specialised auditor have a greater asymmetric timeliness of earnings for bad news than firms audited by non-specialists.

To compare the pre-crisis period with the crisis period, I now run the regressions for the years 2007-2008. The adjusted R-squares are larger in the crisis sample than in the pre-crisis sample and approach the adjusted R-squares in Krishnan's (2005) study more. This can be explained by more financially distressed firms during the crisis. Like explained earlier, earnings is expected to contain more timely information for firms with a negative return on stock. Therefore, the adjusted R-squared for the sample including more firms with bad news is expected to be higher. As can be seen in Table 8, the percentage of firms with negative stock returns was two times larger for firms during the crisis compared with years outside the crisis-period.

Like the pre-crisis period results, the crisis period results show positive and highly significant β1s. So, again earnings are more sensitive to negative returns than to positive returns. Also, the β1 for clients of specialist auditors (Panel B) is again larger compared with non-specialist clients (Panel A). For this study, it is interesting to compare the difference in sensitivity to negative returns and positive returns between clients of non-specialist and specialist auditors for the periods before and during the financial crisis. Clients of specialist auditors are, relative to clients of non-specialist auditors, [(0.909-0.536)/(0.356-0.152)] 83% more sensitive to negative returns than to positive returns during the crisis period. This is lower than the 91% in the pre-crisis period. So, the difference between the two types of clients has decreased during the crisis. This is not in line with the first hypothesis. The sensitivity of non-specialist auditor clients decreased with [1-(0.356-0.152)/(0.668-0.174)] 59% from the pre-crisis to the crisis period. Also, the

(30)

sensitivity of specialist auditor clients decreased with almost the same percentage; [1-(0.909-0.536)/(1.126-0.182)] 60%.

For the crisis period, β2 is negative and highly significant. The coefficient is more negative than in the previous period regression. Therefore, it can be suggested that there is even less timely incorporation of economic gains in earnings of specialist auditors' clients versus non-specialist auditors' clients during the crisis. Furthermore, β3, like during the pre-crisis period positive and highly significant, is larger for the crisis period. Both β2 and β3 suggest that the difference in asymmetric timeliness of earnings for bad news has become greater for clients of specialist auditors compared to clients of non-specialist auditors.

Table 8

Mean Coefficients and Adjusted R²s from Annual Cross-Sectional Regressions: Within-Crisis Period 2007-2008.

Panel A: Model 1 (Clients of non-specialist auditors) N=3900

α0 α1 β0 β1 Adj. R² 0.066 (6.57)*** -0.022 (-1.48) -0.152 (-10.37)*** 0.356 (11.54)*** 3.88%

Panel B: Model 2 (Clients of specialist auditors) N=1543

α0 α1 β0 β1 Adj. R² 0.102 (3.26)*** -0.017 (-0.41) -0.536 (-9.33)*** 0.909 (9.93)*** 7.89% Panel C: Model 2 N=5443 α0 α1 α2 α3 β0 β1 β2 β3 AdjR² 0.066 (6.57)*** -0.022 (-1.48) 0.036 (1.26) 0.005 -0.152 (-10.37)*** 0.356 (11.54)*** -0.384 (-7.92)*** 0.553 5.32% (0.13) (6.36)***

The mean coefficients and adjusted R-squares are calculated by adding up coefficients of the years 2007-2008 based on a fixed number of average observations per year. Between parentheses are t-statistics for the annual estimates' cross-temporal distributions to test the significance of parameter estimates. These are calculated by the ratio of my mean estimate of the coefficient to the standard error of the distribution of each year's coefficient, divided by the square root of the number of years in this period. ***, **, * Significance at the 0.01, 0.05 and 0.10 level, respectively, for a one tailed test.

The influence of the financial crisis on the relation between Big 4 auditor industry specialization and the asymmetric timeliness of earnings