Auditor search periods as signals of engagement risk: Effects on auditor choice and audit pricing

ADIAC-00339; No of Pages 15 Advances in Accounting xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

Advances in Accounting journal homepage: www.elsevier.com/locate/adiac

Auditor search periods as signals of engagement risk: Effects on auditor choice and audit pricing☆ Vivek Mande a, Myungsoo Son a,⁎, Hakjoon Song b a b

California State University, Fullerton, USA California State University, Dominguez Hills, USA

a r t i c l e

i n f o

Article history: Received 28 June 2016 Received in revised form 28 February 2017 Accepted 1 March 2017 Available online xxxx JEL classification: M41 M42

a b s t r a c t We examine the effect of auditor search periods (time taken from the dismissal/resignation of the old auditor to the appointment of the new auditor) on successor auditor choice and audit fees. Using a sample of auditor changes during the period 2002–2012, we find that clients associated with long search periods are less likely to be accepted by Big N auditors. Our results also show that successor auditors charge their clients higher initial audit fees following lengthier searches. Finally, we document that delays in appointing successor auditors following resignations are associated with a significantly negative stock market response. Our results suggest that investors, regulators and academics should be heedful of lengthy auditor search periods in their evaluations of audit quality and client risks. © 2017 Elsevier Ltd. All rights reserved.

Keywords: Auditor search period Audit fees Auditor choice Auditor resignation Auditor dismissal Market reaction

1. Introduction How auditors manage risks associated with auditing their clients is a topic of great interest to investors, regulators and academics (Johnstone, 2000, 2001). Studies show that auditors evaluate their own firms' business risks, and their clients' business and audit risks in client acceptance decisions (Ayers & Kaplan, 1998; Cohen & Hanno, 2000; Johnstone, 2000, 2001; Asare, Cohen, & Trompeter, 2005; Schroeder & Hogan, 2013).1 Once a client is accepted, audit firms actively manage their client portfolio risk, shedding risky clients when necessary (Johnstone & Bedard, 2003). They also implement other risk-management strategies, for example, assigning specialists and/or charging higher audit fees for their riskier clients (Johnstone & Bedard, 2004). ☆ Data Availability: Data are available from public sources identified in the paper. ⁎ Corresponding author at: 800 N State College Blvd., Fullerton, CA 92831, USA. E-mail addresses: [email protected] (V. Mande), [email protected] (M. Son), [email protected] (H. Song). 1 These risks are defined in Johnstone and Bedard (2004) as follows. Client business risk is “the risk that a potential client's economic condition will deteriorate in either the short term or long term”. Audit risk is “the risk that the auditor may unknowingly fail to appropriately modify his opinion on financial statements that are materially misstated”. Auditor business risk is “the risk that the audit firm will suffer a loss resulting from the engagement”.

Khalil, Cohen, and Schwartz (2011) examine whether a client's engagement risk is positively related to the time taken to find a new auditor following an auditor change (hereafter auditor search period (ASP)). In support of their hypothesis, they find that ASPs are positively associated with client-risk factors such as internal control weaknesses and illegal acts by top management. It is not clear from their study, however, whether long ASPs proxy for a “new” or unreported risk factor(s), since the authors only demonstrate a positive association of ASPs with “known” risk factors identified by prior research. Khalil et al. (2011) also do not examine whether lengthy ASPs in turn impact auditors' decisions. Unlike Khalil et al. (2011) we examine whether ASPs signal the presence of new or unreported risk factors that actually influence auditors' decisions. Specifically, we test whether lengthy ASPs have an incrementally significant impact over and above the known risk factors documented by prior work in influencing the choice of a new auditor and initial auditor fees. There has been no research that we are aware of, that has examined the impact of audit search periods on auditor choice and initial audit fees. Khalil et al. (2011) exclude auditor dismissals from their sample because they argue that lengthy ASPs have an association with risk factors only in the case of auditor resignations. However, because management reports whether an auditor was dismissed or resigned, some resignations may have been reported as dismissals

http://dx.doi.org/10.1016/j.adiac.2017.03.001 0882-6110/© 2017 Elsevier Ltd. All rights reserved.

Please cite this article as: Mande, V., et al., Auditor search periods as signals of engagement risk: Effects on auditor choice and audit pricing, Advances in Accounting (2017), http://dx.doi.org/10.1016/j.adiac.2017.03.001

2

V. Mande et al. / Advances in Accounting xxx (2017) xxx–xxx

to avoid a negative investor reaction (Lee, Mande, & Ortman, 2004; Griffin & Lont, 2010). Our results indicate that dismissals involving positive search periods are possibly de-facto resignations. Specifically, we find that dismissals accompanied by appointment delays are associated with the same high risk client-factors observed with resignations. Turner, Williams, and Weirich (2005) also report that auditor dismissals occur more often when there are concerns about internal control weaknesses and the reliability of financial reporting. As in Khalil et al. (2011), we measure ASPs by counting the number of days between the termination date of the predecessor auditor and the engagement date of the successor auditor. Although companies usually appoint new auditors on the old auditors' termination dates, there are at least two reasons for why high risk firms can have non-zero search periods (i.e., positive ASPs). First, auditors are known to take a longer time in their client acceptance decisions when they solicit or are solicited by high risk clients. A more complex and time-consuming process is involved before auditors can justify profitability when accepting a risky engagement. In some cases, a second review of a prospect by risk management partners is necessary (Johnstone & Bedard, 2003; Ayers & Kaplan, 1998) which in turn delays client acceptance.2 Second, a high risk client may have previously solicited and been rejected by another audit firm(s) which can lengthen the firm's search for an auditor. In general, we expect that the longer it takes for a client-firm to conclude its search for a new auditor, the more likely that there are risk factors present in that firm. Our sample consists of auditor switching firms in the years beginning after SOX and ending in 2012, as reported in Audit Analytics. Using both dismissal and resignation samples, we first investigate whether Big N audit firms are more or less likely to accept clients with long search periods than non-Big N audit firms. As Big N auditors have more to lose from audit failures (Jones & Raghunandan, 1998), we could expect that Big N auditors will be less likely to accept risky clients as proxied by long search periods. However, it is also possible that Big N auditors are more capable of accepting high-risk new clients because the auditors are able to diversify client-risk over a larger client portfolio (Simunic & Stein, 1990; Francis & Krishnan, 2002). Our empirical results, however, support a strategy of risk avoidance by Big N auditors. Specifically, we find that auditor search periods are negatively associated with the likelihood of subsequent acceptance by a Big N auditor (see also Shu, 2000; Catanach, Irving, Williams, & Walker, 2011). Second, we investigate whether a lengthy auditor search period is positively associated with the initial pricing of an audit following an auditor change. If lengthier search periods proxy for higher levels of engagement risk, successor auditors can be expected to exert more effort and/or charge a higher hourly rate for auditing their new risky clients. In support, we find that the length of the audit search period has an incrementally significant and positive association with audit fees after controlling for other known risk factors. This finding is noteworthy. Firms with long ASPs are not only less likely to be accepted by Big N auditors but they also face higher audit fees. In tests performed separately by Big N and nonBig N auditors, we find that even non-Big N auditors are able to charge a fee premium that increases with the duration of the search period. Finally, we examine whether the search periods are associated with the stock market reaction to auditor changes. After controlling for other determinants, we find that firms having positive search periods and whose auditors have resigned experience a significantly greater negative market reaction. Overall, therefore, the

2 Risk management partners are often more conservative than engagement partners (Ayers & Kaplan, 1998).

results suggest that long auditor search periods reflect risk factors that are considered by both markets and auditors in their decision making. Investors, policy makers and academics should be heedful of auditor search periods because their durations convey useful incremental information about client-firms' risks. Knowing whether an auditor resigned or was dismissed is a matter of great interest to investors, regulators and academics. Although auditors do not resign from their engagements very often, their resignations trigger a large negative reaction from market participants (Shu, 2000; Raghunandan & Rama, 1999; Krishnan & Krishnan, 1997). There is great interest in understanding what risk factors are being signaled to markets by an auditor resignation. Griffin and Lont (2010), for example, demonstrate that the stock market response to auditor resignations is largely driven by fundamental risk factors such as litigation and bankruptcy, while Ghosh and Tang (2015) find that resignations portend negative events for firms in the following three years such as internal control weaknesses, litigation and delistings. Our study contributes to this literature by showing that among the firms whose auditors have resigned, those experiencing delays in finding a new auditor are viewed by auditors and investors as being associated with significantly higher risk factors. We present supporting evidence indicating that appointment delays are associated with future adverse events such as delistings and litigation. Interestingly, we find that auditor dismissals are also associated with downward auditor changes and higher successor auditor fees, although the association is not statistically significant in sub-sample tests. The remainder of the paper is organized as follows. In the next section, we discuss prior literature and our hypotheses. This is followed by a discussion of the sample, empirical models and results. Our conclusion is in the final section.

2. Literature review and hypotheses development 2.1. Client-risk management by auditors Using proprietary pre-SOX data, Johnstone and Bedard (2004) examine client acceptance decisions of a large audit firm during 2000–2001 and provide evidence of active risk management that includes the shedding of riskier clients and acceptance of less risky clients. Pratt and Stice (1994) survey 243 auditors and report that the auditee's financial condition is of paramount importance in the decision to accept or continue with a client. Johnstone (2000) develops a model showing in an experimental setting that client-business risk, audit risk and auditor-business risk are all evaluated in client acceptance and continuance decisions. Using proprietary data for 1997–1998, Johnstone and Bedard (2003) provide evidence that auditors adopt a variety of strategies that include assigning specialist personnel and billing higher rates to manage risks in their client portfolios. Studies suggest that auditor changes reflect information about client risk factors more strongly post SOX than pre-SOX. Landsman, Nelson, and Rountree (2009), for example, demonstrate that in the periods immediately following SOX, the Big 4 auditors very aggressively managed their client portfolios by dropping risky clients (see also Griffin & Lont, 2010). They attribute these risk-based realignments to a reduced supply of auditor services (demise of Arthur Andersen) and a higher demand for those services (SOX requirements). Rama and Read (2006) also find that there were more Big 4 auditor resignations in 2003 than in 2001. Using post-Auditing Standard No. 5 (AS 5) data, Schroeder and Hogan (2013) document that the Big 4 continued to rebalance their client-portfolios by shedding high risk clients well beyond the early post-SOX years. This is interesting because resource capacity had increased at the Big 4 firms as there was a lower level

Please cite this article as: Mande, V., et al., Auditor search periods as signals of engagement risk: Effects on auditor choice and audit pricing, Advances in Accounting (2017), http://dx.doi.org/10.1016/j.adiac.2017.03.001

V. Mande et al. / Advances in Accounting xxx (2017) xxx–xxx

of audit effort involved in performing integrated audits under AS 5 (Doogar, Sivadasan, & Solomon, 2010; PCAOB, 2007). 2.2. Auditor search periods and risk factors There is only scant research on the impact of audit search periods on client acceptance decisions. Khalil et al. (2011) examine whether a client's engagement risk affects the search period for a new auditor following an incumbent auditor's resignation. Khalil et al. (2011) argue that for high risk clients, prospective auditors take additional time to gather and analyze client-information and more audit effort is required for risk assessments (see also Johnstone, 2000). Additionally, delays can occur when a second audit partner is asked to review the engagement partner's risk assessment (Ayers & Kaplan, 1998). The search can also become lengthy when prior to being accepted by a successor auditor, a high risk client has solicited and been rejected by other more risk-sensitive audit firms. A rejection can occur, for example, if the client was not compatible with the auditor's client portfolio risk management strategy or if the risk/return payoff was not acceptable (Krishnan & Krishnan, 1997; Shu, 2000). The rejected client-firm must then find another auditor who possibly has less stringent standards for accepting clients. We acknowledge that in some cases positive ASPs can occur due to non-risk related considerations. For example, appointment delays can result because the audit committee may have taken additional time for researching the successor auditor's expertise and independence, or because negotiations over audit fees did not progress in a timely fashion.3 However, if positive ASPs are mainly attributable to non-risk related factors, we would not expect to find support for our hypotheses which are predicated on risk-based responses from auditors. Khalil et al. (2011) only use a sample of client-firms whose auditors resign because they argue that client-firms dismissing their auditors are able to plan ahead for their auditor's termination and the successor auditor's appointment. They appear to suggest, therefore, that ASPs should be non-positive for auditor dismissal firms. In contrast to Khalil et al. (2011), we argue that auditor dismissals are also associated with audit risk factors. For example, auditees using aggressive accounting practices have incentives to dismiss their auditors for obtaining a favorable opinion from a successor auditor, and, cash-strapped clients may seek lower quality auditors who offer them deep cuts in audit fees. In support, Newton, Persellin, Wang, and Wilkins (2016) find that firms have incentives to dismiss their auditors for obtaining favorable internal control opinions from a new auditor. Similarly, Turner et al. (2005) report that many auditor dismissals are triggered by disagreements with auditors about matters concerning internal control weaknesses and the reliability of financial reporting. While Khalil et al. (2011) find that ASPs are correlated with risk factors that are reported in Forms 8-K such as internal control weaknesses, client-firms frequently fail to provide reasons for why their auditors resigned or were dismissed (Griffin & Lont, 2010). Indeed, a long standing complaint from investors has been that risk factors surrounding auditor changes are mostly not publicly disclosed. In our sample of 5524 auditor changes, for example, investors were not given an explanation for the turnover in about 73% of these instances.4 3 In the case of Herbalife, for instance, upon the resignation of KPMG, there were delays in appointing PwC as the new auditor because the latter firm had provided consulting work during the periods that had to be re-audited. Given this independence problem, Herbalife decided to seek permission from the SEC for appointing PwC as its auditor. Herbalife's audit committee also had to decide whether PwC's prior consulting work could impair the auditor's independence. This resulted in delays in confirming PwC as Herbalife's auditor. 4 Mandatory disclosures required in Forms 8-K such as, reportable events and/or auditor-client disagreements are not included in this count unless companies specifically cite those as reasons for the auditor-client realignment. Using Audit Analytics, we find that in companies that report reportable events in Forms 8-K, the disclosures concerned: internal control weakness (49%), auditor merger (20%), restatements (18%), disputes on audit opinions (17%), disagreements on accounting treatment (11%), fee disputes (9%), lack of independence (6%) and management representation not being reliable (3%).

3

We argue that ASPs could potentially proxy for the unobservable risk factors not required to be disclosed per SEC regulations on Forms 8-K such as contentious relationships with management or the board, and reputation of management.5 Second, they could proxy for reportable risk factors that managers choose not to disclose in the Forms 8-K filings. That is, some managers may have incentives to not disclose unfavorable information even when its reporting is mandated by SEC regulations. Ettrege, Johnstone, Stone, and Wang (2011), for example, provide evidence of noncompliance by companies with Forms 8-K and other SEC filing requirements.6 Third, we argue that when there is no auditor present (i.e., ASP N 0) the likelihood of restatements and other financial irregularities can increase, especially in companies having weak internal controls.7 We also provide preliminary evidence that positive ASP firms are more like to be delisted or sued in the future (see Table 2, Panel D). Therefore, to the extent that ASPs indicate risk factors that are not reported in Forms 8-K, we expect ASPs to have incremental content in explaining client acceptance decisions and audit fees. 2.3. Auditor search periods and auditor choice Large audit firms have more to lose from an audit failure due to litigation and loss of reputation than small audit firms (Feltham, Hughes, & Simunic, 1991; Jones & Raghunandan, 1998; Shu, 2000; Johnstone & Bedard, 2004). In support, studies document numerous risk-avoidance strategies used by large auditors for managing their client portfolio.8 On the other hand, large audit firms have a bigger client base over which high risks can be diversified (Francis & Krishnan, 2002; Feltham

5 There are maybe a number of risk related reasons for why an auditor will not accept an engagement: very high amounts of equity compensation paid to top management, significant deficiencies (not weaknesses) in internal controls, and the importance paid to the audit function by top management and the board. In a personal conversation with one of the authors, an audit partner stated that in some cases, their audit firm might walk away from an engagement because “things just don't look right”. The nature of these risk factors is such that it would be difficult to make rules mandating their disclosure in Forms 8-K. In that sense, ASPs are bit of a black-box because they could represent risk factors not required to be disclosed in Forms 8-K. While successor auditors may not actually use audit search periods in their fees and client choice decisions, the search periods represent an observable proxy that investors and others can use as signals of risks. We attempt to open the black box by relating ASPs to future adverse events (see Table 2, Panel D). 6 Managers are aware that there is a negative market reaction to the disclosure of unfavorable information (Ettrege et al., 2011; Whisenant, Sankaraguruswamy, & Raghunandan, 2003b). Ettrege et al. (2011) find that non-compliance increases in firms having: lower quality corporate governance, a reduced need for external financing, and bad news. Their study suggests that there is a need for increased monitoring by regulators. There is only anecdotal evidence of non-compliance with regard to reporting “reportable reasons” about auditor changes. A well publicized case is that of Computer Associates which in 1999 dismissed its auditor, E&Y, while stating at the same time that there was no disagreement about its accounting with its auditor. Investigations by the SEC later showed that Computer Associates has dismissed E&Y because there was an accounting disagreement (i.e., a reportable event) with the auditor regarding disclosures about compensation of top executives. 7 Generally, auditors like to review significant transactions around the time they occur so that the chances of misstatements or errors are reduced. When there is no auditor present, this arguably represents a risk, and, would require extra effort to audit the transactions later. A similar argument was made by the SEC in requiring quarterly auditor reviews of Forms 10-Q filings. The SEC argued that having an auditor present would facilitate the timely identification of accounting issues, reduce the likelihood of errors and year-end adjustments, and reduce pressure for earnings management during the interim periods. Conversely, when there is no auditor present, especially where lengthy searches are involved, there could result a reduction in the financial discipline over the reporting process, lower reliability of the financial statements and higher risks of misstatements, especially in firms having weak internal controls. 8 For example, Jones and Raghunandan (1998) find that Big 8 audit firms had fewer clients that were in financial distress and in high-technology industries during 1987–1994. Similarly, Choi, Doogar, and Ganguly (2004) document that client-firms audited by Big N auditors were less risky than those audited by non-Big N firms during 1990 to1994. Johnstone and Bedard (2004) find that newly accepted clients of a large audit firm were less risky than its continuing clients, which supports risk avoidance/mitigation.

Please cite this article as: Mande, V., et al., Auditor search periods as signals of engagement risk: Effects on auditor choice and audit pricing, Advances in Accounting (2017), http://dx.doi.org/10.1016/j.adiac.2017.03.001

4

V. Mande et al. / Advances in Accounting xxx (2017) xxx–xxx

Table 1 Sample selection and distribution. Panel A: Sample selection Sample selection

Observations for auditor choice and audit pricing

Observations for market reaction

Auditor switch firms on Audit Analytics (2002–2012) Less Firms without auditor search period Firms with extreme auditor search period Firms without audit fees on Audit Analytics Firms without restatements and internal controls on Audit Analytics Firms without financial data on Compustat Firms without market returns on CRSP Final sample

17,646 437 101 4145 1733 5706 N/A 5524

17,646 437 101 N/A N/A 502 12,294 4312

Panel B: Sample distribution by year Year

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Total

a

Sample for auditor choice and audit pricing

Sample for market reaction

# Firms

Percent

# Firms

Percent

1164 465 659 652 560 475 365 430 322 244 188 5524

21.07 8.42 11.93 11.8 10.14 8.6 6.61 7.78 5.83 4.42 3.4 100.00

945 359 483 523 436 346 263 322 278 182 175 4312

21.92 8.33 11.20 12.13 10.11 8.02 6.10 7.47 6.45 4.22 4.06 100.00

Panel C: Sample distribution by industry SIC

Industry

100–999 1000–1999 2000–2999 3000–3999 4000–4999 5000–5999 6000–6999 7000–9999 Total

Agricultural, mining Mining, construction Manufacturing—food, textiles, lumber, chemicals Manufacturing—rubber, metal, machinery, equipment Transportation, communication, utilities Wholesale, retail Finance, insurance, real estate Services

a

Sample for auditor choice and audit pricing

Sample for market reaction

# Firms

Percent

# Firms

Percent

16 432 786 1410 450 411 924 1095 5524

0.29 7.82 14.23 25.52 8.15 7.44 16.73 19.82 100.00

19 253 542 1035 356 306 1016 785 4312

0.45 5.87 12.56 24.01 8.25 7.10 23.56 18.20 100.00

As we only focus on the post-SOX period, the sample for year 2002 only includes firms with fiscal years ending in or after August 2002.

et al., 1991; Johnstone & Bedard, 2004).9 Industry expertise cultivated by big auditors also helps mitigate high audit risks to a greater extent than possible by small auditors (Cenker & Nagy, 2008). However, there is little empirical evidence to support the idea that large audit firms have a greater ability to accept and manage risky clients. In support of a strategy of risk avoidance in the post-SOX period, studies show that the Big 4 auditors dropped many of their risky clients due to capacity constraints caused by the demise of Andersen and SOX (e.g., Landsman et al., 2009). Schroeder and Hogan (2013) find that the Big 4 auditors also shed high risk clients after the passage of Auditing Standard No. 5 in 2007. Based on the arguments above, we expect that Big N auditors will be less willing to accept clients having long auditor search periods than non-Big N auditors, after controlling for other factors affecting auditor choice. H1. The length of the search period is positively associated with the likelihood that the successor auditor will be a non-Big N firm. 9 A recent study by Hackenbrack, Jenkins, and Pevzner (2014), however, questions whether the “portfolio” approach is used by audit firms to diversify risk. The authors suggest that decisions on client acceptance (and client continuation and audit fees) are largely made at the office and partner level rather than at the firm level. Their study supports the view that it is more difficult for a Big N auditor to diversify high client-risks.

2.4. Audit search periods and initial audit fees Prior studies indicate that subsequent to the acceptance of a high risk client, auditors increase the scope of the audit and/or increase audit effort to mitigate the high levels of engagement risk (Bell, Landsman, & Shackelford, 2001; Elliott, Ghosh, & Peltier, 2013). In addition to billing more audit hours, specialized partners and highly experienced staff are often deployed on high risk engagements (Venkataraman, Weber, & Willenborg, 2008; Elliott et al., 2013). Malpractice insurance premiums can also increase for auditors having a portfolio of risky clients as they are more likely to be targeted with lawsuits (Jones & Raghunandan, 1998, p 172). The additional costs incurred for auditing risky clients are generally passed on to the client-companies in the form of higher audit fees (Jones & Raghunandan, 1998).10

10 For example, in the case of Herbal Life Ltd. the firm's predecessor auditor, KPMG, resigned on April 8, 2013 because of illegal activity involving an audit partner. The company took 43 days to engage its new auditor, PwC. During 2011–2012, KPMG charged audit fees of $3,839,000 (2011) and $3,940,000 (2012) representing an annual increase of 2.63%. However, the new auditor, PwC, increased audit fees to $4,911,000 for 2013 (this amount excludes re-audit fees for prior financial statements), a 24.6% increase over the previous year.

Please cite this article as: Mande, V., et al., Auditor search periods as signals of engagement risk: Effects on auditor choice and audit pricing, Advances in Accounting (2017), http://dx.doi.org/10.1016/j.adiac.2017.03.001

V. Mande et al. / Advances in Accounting xxx (2017) xxx–xxx

Based on the discussion above, we hypothesize (H2) that auditor search periods will be positively associated with initial audit fees being charged by successor auditors, after controlling for other factors affecting audit fees. H2. The length of the search period is positively associated with the initial audit fees set by the successor auditor.

2.5. Auditor search periods and stock market reaction We also consider the effect of long ASPs on auditor changes from an investor's perspective. If zero ASPs are the norm, then positive ASPs can signal the presence of client-risk factors that are potentially priced by the stock market. Several prior studies have focused on examining stock returns surrounding auditor dismissals and resignations. Except for a few studies (e.g., Johnson & Lys, 1990; Klock, 1994), research generally views auditor changes as “negative” events that are accompanied by an adverse stock market reaction, with a more pronounced negative reaction observed for auditor resignations (Griffin & Lont, 2010; Knechel, Naiker, & Pacheco, 2007; DeFond & Zhang, 2014). If investors have a negative view of appointment delays, after controlling for other factors, the market reaction to a dismissal/resignation should be more negative when a successor auditor has not been found as of the dismissal/resignation date. It is worth noting, however, that investors have less information about auditor changes than do auditors, which in turn could affect the magnitude of the stock market reaction. In univariate tests, Smith (1988) finds that there are more negative returns to firms that fail to announce a successor auditor on the auditor-termination date when compared to other firms experiencing auditor changes. His results suggest that when there is a time lag between the previous auditor's termination and the successor auditor's engagement, it is viewed negatively by market participants. Although Smith (1988) suggests that ASP is a risk factor priced by the stock market, he only examines its usefulness to investors in the context of Accounting Series Release (ASR) No. 165.11 Additionally, in contrast to Smith's (1988) univariate analyses, we use a multivariate regression model. Our third hypothesis stated in a directional form is as follows: H3. The stock market reaction surrounding auditor changes is more negative for clients experiencing delays in finding successor auditors.

3. Research design 3.1. Sample selection Similar to recent studies on auditor changes, we test our hypotheses using post-SOX data (e.g., Khalil & Mazboudi, 2016; Huang & Scholz, 2012; Cenker & Nagy, 2008). Research shows that following SOX, there were increases in: auditors' litigation risk, audit fees, and the number of auditor switches (Huang, Raghunandan, & Rama, 2009; Landsman et al., 2009). Post-SOX

11 The SEC in ASR No. 165 changed the event that triggers Form 8-K reporting, from the new auditor's engagement date to the old auditor's termination date. Smith's (1988) results support the SEC's stance that requiring Form 8-K reporting on the termination date provides investors with more timely information. In the pre-ASR No. 165 period, investors could not discern whether companies changing auditors experienced a time lag in finding a new auditor.

5

studies show that the associations between client-risk factors and auditor fees/auditor changes have become stronger, as the SOX has increased auditors' legal exposure (Elder, Zhang, Zhou, & Zhou, 2009; Li, 2008; Rama & Read, 2006). Table 1, Panel A describes our sample selection procedures. We identify, using the Audit Analytics database, all firm-year observations representing auditor changes during the period 2002 to 2012.12 Our initial sample consists of 17,646 auditor changes. We then drop 437 firmyear observations that did not have the dates for auditor engagements and dismissals/resignations. Next, we remove 101 observations having “extreme” auditor search periods.13 Finally, we delete firm-years with missing data on either Audit Analytics, Compustat or CRSP. These procedures result in the following samples: 5524 observations for the auditor choice tests (H1) and the audit fee tests (H2), and 4312 observations for the stock market reaction tests (H3). Panel B presents the sample distribution by year. Consistent with prior research (e.g., Elder et al., 2009), there is a higher frequency of auditor changes in the years surrounding the passages of SOX Section 404 and Auditing Standard No. 2 (AS2) (PCAOB, 2004) in 2004. Panel C shows the sample distribution by industry using the Standard Industry Classification (SIC) codes. Firms in manufacturing, service and finance industries have a greater representation than firms in other industries. 3.2. Auditor choice model Drawing on prior work (e.g., Catanach et al., 2011; Huang & Scholz, 2012) we model auditor choice using the following logistic regression: Prob ðBIGNÞ ¼ β0 þ β1 LNASP þ β2 SIZE þ β3 FRGN þ β4 SEG þ β5 RECV þ β6 INVT þ β7 LEV þ β8 LOSS þ β9 CATA þ β10 ACCR þ β11 EBIT þ β12 QUICK þ β13 LITIND þ β14 IC þ β15 REST þ β16 BM þ β17 GC þ β18 DYE þ β19 PREAU þ β20 ISSUE þ β21 RESIGN þ year dummies þ industry dummies þ ε

ð1Þ

The dependent variable (BIGN) is coded 1 if the successor auditor is a Big N firm, and 0 otherwise. The test variable is the natural logarithm of the duration taken to find a new auditor (LNASP).14 In support of H1, we should expect to observe a negative coefficient on LNASP if a Big N auditor is less likely to accept a client with a long search period. Since the selection of an auditor and the determination of audit fees involve a similar consideration of risk factors, we use the same set of control variables in both models. We predict that Big N auditors are more likely to audit larger and more complex firms and firms having greater growth potential (Raghunandan & Rama, 1999; Catanach et al., 2011). As proxies we use the following: SIZE (natural logarithm of total assets), FRGN (ratio of foreign

12 Our post-SOX sample consists of observations having a fiscal-year ends after August 1, 2002. Similar to other studies on auditor changes (e.g., Griffin & Lont, 2010), our sample includes Arthur Andersen's clients that were forced to switch their auditors. We include these firms because lengthy search periods can signal the presence of risks even though the changes were involuntary. As a sensitivity test, we excluded these companies without changes to our conclusions. 13 We exclude firm-years with auditor search periods that are longer than 365 days. Khalil et al. (2011) report a maximum ASP of 262 days. 14 We use natural logarithm of ASP in our models because the distribution of ASP is highly skewed. We transform ASP into its logged form: LNASP = log (1 + ASP). As a check for robustness we also use the raw values of ASP. Using raw values provided us with higher statistically significant coefficients on the test variables.

Please cite this article as: Mande, V., et al., Auditor search periods as signals of engagement risk: Effects on auditor choice and audit pricing, Advances in Accounting (2017), http://dx.doi.org/10.1016/j.adiac.2017.03.001

6

V. Mande et al. / Advances in Accounting xxx (2017) xxx–xxx

sales relative to total sales), SEG (number of business segments), and BM (book to market ratio). Firms prone to litigation risk, measured by financial stress (LOSS) and firm leverage (LEV), are more likely to be audited by a non-Big N auditor because we argue that Big N auditors may view these firms as too risky to audit (Shu, 2000). Along the same lines, firms having a larger proportion of receivables and inventories (RECV and INVT) are less likely to be audited by a Big N auditor because these accounts are associated with litigation and SEC enforcement actions (Stice, 1991; Feroz, Park, & Pastena, 1991). Additionally, we include a control for litigation risk (LITIND) based on client-firms' membership in industries that are prone to litigation (e.g., Francis, Philbrick, & Schipper, 1994; Geiger, Raghunandan, & Rama, 2006; Venkataraman et al., 2008). However, since high litigation risk industries also experience high growth (Catanach et al., 2011), we acknowledge that the sign on this variable can be positive or negative. Also for risk avoidance reasons, we suggest that Big N auditors are more likely to accept audit clients that are more profitable and financially sound (EBIT, CATA and QUICK) and avoid clients that have internal control weaknesses (IC), restatements (REST), high levels of accruals (ACCR) and high-risk events (ISSUE) reported in their Forms 8-K (Whisenant et al., 2003b; Griffin & Lont, 2010). 15 Similarly, firms with going-concern opinions (GC) are also less likely to find a Big N auditor (Johnstone & Bedard, 2003); and Big N auditors are less likely to accept clients whose current auditors have resigned (RESIGN) (Raghunandan & Rama, 1999). Finally, we control for the client's fiscal year-end (DYE) and prior auditor type (PREAU) following Raghunandan and Rama (1999). 3.3. Audit fee model Audit fees are modeled as a function of audit complexity, litigation risk, and audit effort (Ferguson, Francis, & Stokes, 2003; Francis, Reichelt, & Wang, 2005; Francis & Wang, 2005; Hay, Knechel, & Wong, 2006; Whisenant, Sankaraguruswamy, & Raghunandan, 2003a). The same variables in the auditor choice model are used as controls for these attributes. LAF ¼ β0 þ β1 LNASP þ β2 SIZE þ β3 FRGN þ β4 SEG þ β5 RECV þ β6 INVT þ β7 LEV þ β8 LOSS þ β9 CATA þ β10 ACCR þ β11 EBIT þ β12 QUICK þ β13 LITIND þ β14 IC þ β15 REST þ β16 BM þ β17 GC þ β18 DYE þ β19 BIGN þ β20 ISSUE þ β21 RESIGN þ year dummies þ industry dummies þ ε

ð2Þ

The dependent variable (LAF) is the natural logarithm of initial audit fees paid to successor auditors. If a lengthy search period represents an incremental risk factor, then as stated in H2, we should observe a positive and statistically significant coefficient on LNASP. SIZE, along with the other proxies for audit complexity (FRGN, SEG, RECV and INVT), is expected to be positively associated with audit fees. Firm leverage (LEV), incidences of losses (LOSS), the ratio of current assets to total assets (CATA), earnings before interest and taxes scaled by total assets (EBIT), the quick ratio (QUICK), and membership in a high litigation industry (LITIND) are used as proxies for litigation risk. Additional controls for risk include the absolute value of total accruals (ACCR), internal control weaknesses (IC), the book to market ratio (BM), restatements (REST),

predecessor auditor's resignation (RESIGN), and reportable events (ISSUE). While EBIT, BM, and QUICK are expected to be inversely associated with audit fees, the remaining variables are expected to have a positive association (Hay et al., 2006; Francis & Wang, 2005). Attributes used to proxy for audit effort and quality (Whisenant et al., 2003a; Hay et al., 2006) are: going concern audit opinion (GC), busy season (DYE) and auditor type (BIGN). Firms having going concern opinions are expected to pay higher audit fees as more effort is needed for auditing these companies. DYE, a dummy variable taking a value of 1 when a firm's fiscal year-end is in December, is expected to be positively associated with audit fees because there are audit staff capacity constraints in the peak season. If a Big N premium exists, we should expect to find a positive relationship between audit fees and BIGN.

Table 2 Descriptive statistics. Panel A: Summary statistics for the ASP variable Sample

N

Mean

Std.

P25

Median

P75

Pooled Dismissal Resignation Positive ASP

5524 4282 1242 1452

6.709 2.168 22.366 27.472

25.428 15.818 41.205 41.057

0.000 0.000 0.000 4.000

0.000 0.000 2.000 12.000

1.000 0.000 33.000 38.000

Panel B: Summary statistics for the variables used in the auditor choice and audit fee tests Variable

Mean

Std.

P25

Median

P75

BIGN LNAF ASP LNASP SIZE FRGN SEG RECV INVT LEV LOSS CATA ACCR EBIT QUICK LITIND IC REST BM GC DYE PREAU ISSUE RESIGN

0.384 12.335 6.709 0.678 4.831 0.077 2.021 0.195 0.099 0.590 0.474 0.450 0.141 −0.084 2.542 0.283 0.127 0.054 0.482 0.129 0.703 0.591 0.252 0.225

0.486 1.430 25.428 1.310 2.299 0.341 1.973 0.198 0.139 0.433 0.499 0.310 0.212 0.349 3.721 0.451 0.333 0.227 1.918 0.335 0.457 0.492 0.629 0.418

0.000 11.348 0.000 0.000 3.120 0.000 1.000 0.051 0.000 0.301 0.000 0.167 0.029 −0.106 0.665 0.000 0.000 0.000 0.235 0.000 0.000 0.000 0.000 0.000

0.000 12.206 0.000 0.000 4.710 0.000 1.000 0.134 0.027 0.528 0.000 0.467 0.071 0.019 1.511 0.000 0.000 0.000 0.519 0.000 1.000 1.000 0.000 0.000

1.000 13.239 1.000 0.693 6.379 0.000 2.000 0.259 0.155 0.780 1.000 0.708 0.153 0.080 2.818 1.000 0.000 0.000 0.915 0.000 1.000 1.000 0.000 0.000

Panel C: Summary statistics for the variables used in market reaction tests Variable

Mean

Std.

25th

Median

75th

CAR POSASP ISSUE LNMKT LNTENURE TIMING RESIGN

−0.001 0.250 0.530 5.010 7.482 0.179 0.191

0.078 0.433 1.096 1.983 1.222 0.383 0.393

−0.030 0.000 0.000 3.570 6.792 0.000 0.000

−0.002 0.000 0.000 4.884 7.545 0.000 0.000

0.026 1.000 1.000 6.242 8.276 0.000 0.000

Panel D: Summary statistics on adverse events occurring within a year of an auditor change for positive ASP and zero ASP firms 15

ISSUE is based on the following eight categories of reportable events in the Audit Analytics database: management representation not reliable assertion, illegal acts, SEC investigation, scope limitation, lack of independence, disagreement about audit opinion, disagreement about accounting treatments, and fee dispute. We assign a value of 1 for each of these events if present and sum the scores for each client-firm to obtain ISSUE.

Positive ASP firms Zero ASP firms Difference (t-test)

Mean class action lawsuits

Mean delistings

0.049 0.037 1.99**

0.061 0.036 3.94***

**, *** indicate statistical significance at the 10%, 5%, and 1%, respectively.

Please cite this article as: Mande, V., et al., Auditor search periods as signals of engagement risk: Effects on auditor choice and audit pricing, Advances in Accounting (2017), http://dx.doi.org/10.1016/j.adiac.2017.03.001

V. Mande et al. / Advances in Accounting xxx (2017) xxx–xxx Table 3 Auditor choice regression results. Variable

Expected sign

Pooled sample

Dismissal subsample

Resignation subsample

Intercept

±

LNASP

–

SIZE

+

FRGN

+

SEG

+

RECV

–

INVT

–

LEV

–

LOSS

–

CATA

+

ACCR

–

EBIT

+

QUICK

+

LITIND

±

IC

–

REST

–

BM

–

GC

–

DYE

±

PREAU

+

ISSUE

–

−4.551*** (−12.62) −0.068* (−1.52) 0.816*** (20.86) −0.053 (−0.45) −0.003 (−0.14) −1.810*** (−6.68) −1.594*** (−4.32) −0.927*** (−4.87) 0.216** (2.16) 1.497*** (6.35) −0.369* (−1.42) −0.287 (−1.07) −0.023 (−1.49) 0.301*** (2.87) 0.025 (0.16) 0.044 (0.19) −0.097*** (−2.98) −0.719*** (−3.67) 0.174* (1.87) 0.137 (1.27) 0.088 (1.09)

−4.207*** (5.55) −0.244*** (−3.59) 0.626*** (8.94) −0.156 (−0.47) 0.090** (1.86) −1.776*** (−2.89) −1.843** (−1.91) −0.425 (−1.12) 0.176 (0.76) 1.112** (2.20) −0.491 (−0.83) −0.660 (−1.33) 0.002 (0.07) −0.146 (−0.57) −0.253 (−0.88) 0.284 (1.21) −0.059 (−1.22) −0.776** (−1.79) 0.151 (0.68) 1.255*** (4.57) 0.092 (0.74)

RESIGN

–

−4.342*** (−6.67) −0.094*** (−2.75) 0.768*** (22.63) −0.056 (−0.50) 0.015 (0.71) −1.829*** (−7.50) −1.564*** (−4.71) −0.844*** (−4.87) 0.251*** (2.76) 1.330*** (6.24) −0.414** (−1.72) −0.334 (−1.12) −0.015 (−1.10) 0.224** (2.37) 0.053 (0.41) 0.007 (0.03) −0.091*** (−3.30) −0.734*** (−4.17) 0.152* (1.80) 0.271*** (2.82) 0.115 (1.33) −0.669*** (−5.70) 0.57 5524

0.57 4282

0.43 1242

Max-rescaled R2 N

This table presents results of logistic regressions that examine the effect of auditor search periods on auditor-choice. The dependent variable is BIGN, a dummy variable representing a choice of Big N auditor. The variable of interest is LNASP, natural log of auditor search periods. *, **, *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively (one-tailed if predicted, and two-tailed otherwise). All regressions are estimated using robust standard errors clustered by firm. See Appendix A for the variable definitions.

3.4. Market reaction model The sample used for tests of H3 consists of predecessor auditor changes having both positive and zero search periods. A dummy variable, POSASP, is used to indicate observations where a successor auditor was not engaged as of the predecessor auditor's dismissal or resignation (Model 3). CAR ¼ β0 þ β1 POSASP þ β2 ISSUE þ β3 LNMKT þ β4 LNTENURE þ β5 TIMING þ β6 RESIGN þ year dummies þ industry dummies þ ε

ð3Þ

The dependent variable, CAR, is cumulative abnormal stock returns over a three-day window (− 1, 1) surrounding the actual auditor

7

dismissal/resignation dates.16 We calculate abnormal stock returns by subtracting the CRSP equal-weighted market return from the firm's holding return (Griffin & Lont, 2010; Scholz, 2008).17 If investors view positive search periods negatively, then consistent with H3, we should observe a negative coefficient on POSASP. Following prior studies (Knechel et al., 2007; Griffin & Lont, 2010; Whisenant et al., 2003b; Hackenbrack & Hogan, 2002), we include as controls: reportable events (ISSUE), the logged values of market capitalization (LNMKT), the logged values of the predecessor auditor's tenure (LNTENURE), a dummy variable denoting late auditor switches (TIMING), and a dummy variable denoting auditor resignations (RESIGN). We expect negative coefficients on variables proxying for “adverse” events namely, ISSUE, LNTENURE, TIMING, and RESIGN. We do not have any directional prediction with regard to the coefficient on firm size (LNMKT). 4. Empirical results 4.1. Univariate correlations We examine univariate correlations between the variables used in the models (untabulated). Contrary to H2, LNASP is not statistically significantly associated with audit fees (LAF). However, supportive of H1, LNASP is negatively related to the likelihood that the successor auditor is a Big N firm (BIGN). POSASP is also negatively correlated with stock market returns (CAR) surrounding the departure dates of the old auditors, providing preliminary support for H3. Additionally, univariate correlations show that ASPs are significantly correlated with disclosed client-risk factors: LNASP is positively correlated with ISSUE, GC, LOSS, LEV, ACCR, IC, REST, LITIND, and RESIGN, and, negatively correlated with SIZE and BM. 4.2. Descriptive statistics In Panel A of Table 2, we report the distribution of the ASP variable across four different samples. In the pooled sample, the mean and median values of ASP are 6.709 and 0, respectively. The mean indicates that firms take, on average, about seven days to find a successor auditor. As predicted, following auditor resignations, client-firms take longer (mean ASP = 22.366 days) to find a successor auditor compared to dismissals (mean ASP = 2.168 days). Panel A also shows that the distribution of the ASP variable is highly skewed towards zero. Specifically, about 74% of the observations have zero values indicating that the departure of the predecessor auditor and the engagement of the successor auditor occur on the same day. When we exclude zero ASPs, we find that the mean ASP is about 27 days. Panel B of Table 2 presents descriptive statistics for the variables used in the regressions. It is noteworthy that the mean values of BIGN (denoting successor Big N auditor) and PREAU (denoting predecessor Big N auditor) are 0.384 and 0.591, respectively. This indicates that proportion of firms with Big N auditors decreased (from 59% to 38%) following an auditor switch. The mean and median reportable events (ISSUE) are 0.252 and 0, respectively, indicating that only a few firms disclosed reportable events, which is consistent with findings of Griffin and Lont (2010). About 13% of auditor switching firms received a going concern opinion (GC) from their departing auditors; 16 During our research period, the SEC (2004) reduced the due date for filing a Form 8-K from five to four business days, effective August 23, 2004. Since we use the event date returns (rather than the filing date returns), we do not expect any effect on our results due to changes in filing days. 17 For our main tests reported here we do not use the Forms 8-K filing dates, following Carter and Soo (1999) and Knechel et al. (2007) who argue that the effects of corporate events are better observed on the actual event dates. However, we obtain the similar results when we calculate abnormal returns using a market model estimated with CRSP value-weighted returns over days −220 to −20 relative to the filing date. As another test for sensitivity, we reperform our tests excluding all firms that filed Forms 8-K late, as in Griffin and Lont (2010).

Please cite this article as: Mande, V., et al., Auditor search periods as signals of engagement risk: Effects on auditor choice and audit pricing, Advances in Accounting (2017), http://dx.doi.org/10.1016/j.adiac.2017.03.001

8

V. Mande et al. / Advances in Accounting xxx (2017) xxx–xxx

Table 4 Audit fee regression results. Variable

Expected sign

Pooled sample

Dismissal subsample

Resignation subsample

Intercept

±

LNASP

+

SIZE

+

FRGN

+

SEG

+

RECV

+

INVT

+

LEV

+

LOSS

+

CATA

+

ACCR

+

EBIT

–

QUICK

–

LITIND

+

IC

+

REST

+

BM

–

GC

+

DYE

+

BIGN

+

ISSUE

+

9.072*** (70.92) 0.038*** (3.03) 0.532*** (58.40) 0.161*** (4.57) 0.054*** (7.38) 0.062 (0.57) 0.065 (0.51) 0.039 (0.95) 0.164*** (5.43) 0.494*** (6.76) 0.224*** (3.01) −0.205*** (−4.10) −0.031*** (−6.88) 0.085** (1.81) 0.193*** (4.39) 0.119* (1.47) −0.012* (1.47) 0.102** (2.16) −0.024 (−0.85) 0.129*** (4.09) 0.173*** (5.50)

9.339*** (41.94) 0.042*** (2.89) 0.520*** (30.81) 0.283*** (2.90) 0.048*** (2.59) −0.157 (−0.76) −0.024 (−0.12) 0.098** (1.73) 0.187*** (3.36) 0.501*** (4.68) 0.144 (1.23) −0.202*** (−2.70) −0.031*** (−4.82) 0.111* (1.34) 0.143** (1.97) 0.235** (1.84) −0.005 (−0.42) 0.016 (0.24) −0.062 (−1.21) 0.266*** (3.26) 0.145*** (3.19)

RESIGN

±

9.137*** (80.81) 0.037*** (4.07) 0.529*** (64.5) 0.180*** (5.39) 0.054*** (7.74) 0.030 (0.29) 0.041 (0.37) 0.062** (1.85) 0.168*** (6.36) 0.487*** (7.82) 0.218*** (3.47) −0.209*** (−5.05) −0.031*** (−8.04) 0.092** (2.22) 0.179*** (4.80) 0.162** (2.37) −0.009* (−1.32) 0.071** (1.83) −0.031 (−1.23) 0.137*** (4.70) 0.169*** (6.40) −0.094*** (−3.31) 0.70 5524

0.69 4282

0.68 1242

Adjusted R2 N

This table presents results of OLS regressions that examine the effect of auditor search periods on initial audit fees. The dependent variable is LNAF, natural log of audit fees. The variable of interest is LNASP, natural log of auditor search periods. *, **, *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively (one-tailed if predicted, and two-tailed otherwise). All regressions are estimated using robust standard errors clustered by firm. See Appendix A for the variable definitions.

and about 13% (5%) of firms reported prior internal control weaknesses (restatements). Table 2, Panel C presents descriptive statistics for the variables used in the market reaction tests. The mean and median cumulative abnormal returns (CAR) surrounding auditor changes are − 0.001 and −0.002. About 19% of the observations are auditor resignations. Firms do not generally disclose any reportable events (mean ISSUE = 0.530) which supports the argument that risk factors are largely undisclosed. Auditor changes in the fourth quarter (TIMING) constitute about 18% of the sample. Finally, for 25% of the observations there are delays in appointing a replacement auditor following an auditor change (mean POSASP = 0.25).18

18 The mean values of POSASP are 19% and 52% in the dismissal and resignation sub-samples, respectively.

In Table 2, Panel D we provide exploratory results on whether positive ASPs portend adverse events in the following fiscal year. As acknowledged earlier, ASPs constitute a bit of a “black box” as it is not clear what adverse events are being signaled by appointment delays. In an effort to open this black box, we examine whether firms having positive ASPs are associated with two types of future adverse events.19 Univariate evidence shows that in the fiscal year subsequent to an auditor switch, firms having positive ASPs are associated with more class action lawsuits and delistings, compared to zero ASP firms. Specifically, there is a statistically significant difference between positive and zero ASP firms in the 19 We follow Ghosh and Tang (2015) by coding a class-action lawsuit as 1 if a client is involved in a class-action lawsuit for issues related to accounting, financial reporting, tax, and SEC accounting, auditing, or enforcement releases within a year following an auditor switch (Audit Analytics category types 1, 2, 41, 43, 48, 54), and 0 otherwise. Delisting is coded 1 when a firm is delisted from a stock exchange for reasons other than merger/acquisition within one year following an auditor switch (CRSP delisting codes 170, 400–490, 535–587, and 589–591), and 0 otherwise.

Please cite this article as: Mande, V., et al., Auditor search periods as signals of engagement risk: Effects on auditor choice and audit pricing, Advances in Accounting (2017), http://dx.doi.org/10.1016/j.adiac.2017.03.001

V. Mande et al. / Advances in Accounting xxx (2017) xxx–xxx Table 5 Stock market reaction results around auditor dismissal/resignation dates. Panel A: Market reaction on dismissal/resignation dates Total (1)

Dismissal Resignation t-Test of difference (2) (3) between (2) and (3)

−0.0003 −0.0001 −0.0042 −0.0025 1.76** 0.64

Zero ASPs (A) Positive ASPs (B) t-Test of difference between (A) and (B)

−0.0006 −0.0065 2.28**

0.19 0.86

Panel B: Market reaction regression results Variable

Expected Pooled sign sample

Pooled sample with interaction

Dismissal Resignation subsample subsample

Intercept

±

POSASP

–

ISSUE

–

LNMKT

±

LNTENURE

–

TIMING

–

−0.003 (−0.38) −0.003 (−0.92) 0.001 (0.69) 0.000 (0.57) 0.000 (0.10) −0.002 (−0.39)

0.038* (1.66) −0.011 (−1.74)** −0.004** (−1.67) 0.000 (0.02) −0.004* (−1.46) 0.000 (0.02)

RESIGN

–

0.003 (0.41) −0.003 (−0.79) −0.001 (−0.46) 0.000 (0.54) −0.001 (−0.75) −0.001 (−0.31) 0.025 (0.60) −0.011* (−1.61) 0.01 4312

0.01 3487

0.01 825

0.004 (0.54) −0.006** (−1.97) −0.001 (−0.48) 0.000 (0.49) −0.001 (−0.80) −0.001 (−0.25) 0.003 (0.72)

RESIGN ∗ POSASP – Adjusted R2 N

0.01 4312

For these tests, we use firm-year observations consisting of all predecessor auditor changes having both positive and zero auditor search periods. *, **, *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively (one-tailed if predicted, and two-tailed otherwise). All regressions are estimated using robust standard errors clustered by firm. See Appendix A for the definition of variables.

9

accruals (ACCR), higher leverage (LEV), lower potential growth (BM) and auditors who have resigned (RESIGN), are less likely to be accepted by Big N auditors.21 On the other hand, firms that: face high litigation risk (LITIND), are larger (SIZE), have more current assets (CATA) and have a Big N firm as a predecessor auditor (PREAU), are more likely to find a Big N successor auditor. Interestingly, in contrast to the univariate association, firms having losses (LOSS) are positively associated with the likelihood of finding a Big N auditor.22 Overall, however, our results are consistent with predictions. Table 3 results also show that LNASP has a negative coefficient in both auditor dismissal and resignation subsamples that is statistically significant at least at p b 0.10. This suggests that while resignations are riskier events, auditor search periods proxy for client-risk factors even when predecessor auditors are dismissed. 4.3.1. Ordered logistic model As a test of sensitivity, we re-perform tests by further partitioning non-Big N auditors into large “second tier” auditors and “other” nonBig N auditors. Following Hogan and Martin (2009), we define large second tier auditors as consisting of four national firms—Grant Thornton, BDO, Seidman, McGladrey & Pullen, and Crowe, Chizek and Company. We replace the dependent variable BIGN in Eq. (1) with a variable called MULTIPLE which is coded 3, 2 and 1 for Big N, second tier and other auditors, respectively.23 Untabulated results for the ordered logistic model show that as auditor search periods lengthen, companies increasingly tend to move downwards in the audit quality hierarchy (the coefficient of LNASP is − 0.033 significant at p b 0.1). This result holds in both dismissal and resignation sub-samples: the coefficients on LNASP are − 0.028 and − 0.109 for the dismissal and resignation samples, respectively and are both significant at least at p b 0.1. Therefore, our results in Table 3 continue to hold using a finer classification of audit quality. 4.4. Audit fee results

average number of class-action lawsuits (0.049 versus 0.037 respectively) and delistings (0.061 versus 0.036 respectively). Comparing the means, we find that there are 32% more class-action lawsuits and 159% more delistings for the positive ASP firms. We also explore this issue using auditor switching firms not disclosing any 8-K reportable events and find the qualitatively similar results. These results must be interpreted with caution as they only provide preliminary univariate evidence that ASPs signal information about “hidden” risks not disclosed in Forms 8-K. We leave for future work the task of examining in more depth, the association of ASPs with subsequent adverse events. 4.3. Auditor choice results In Table 3, we present logistic regression results for the auditor choice model. All regressions use robust standard errors clustered by firm. In the pooled sample, the auditor search period (LNASP) is negatively associated with the likelihood of engaging a Big N successor auditor, which supports H1. The coefficient on LNASP is statistically significant at p b 0.01. The results suggest that non-Big N auditors are not as selective in accepting clients as Big N auditors (Bockus & Gigler, 1998). The economic impact of LNASP is also significant; a one standard deviation increase in LNASP (1.430 in Panel B of Table 2) leads to about a 14% decrease in the probability of hiring a Big N successor auditor.20 Regarding the control variables, client firms with: going-concern opinions (GC), relatively larger risky assets (RECV and INVT), higher

20 The economic significance of LNASP is calculated as follows: exp. (coefficient of LNASP ∗ standard deviation of LNASP) − 1 = exp (0.094 ∗ 1.43) − 1 = 0.144.

Table 4 presents results from estimating the audit fee model. All regressions use robust standard errors clustered by firm. 24 In the pooled sample, we find that the variable of interest, LNASP, has a positive coefficient that is statistically significant at p b 0.01. In support of H2, this suggests that successor auditors charge higher initial audit fees for clients with lengthier auditor search periods. Since the dependent and test variables are both log transformed, we interpret our results as follows. The coefficient on LNASP (0.037) indicates that a 1% increase in ASP (6.709 ∗ 0.01 or 0.0671) is associated with a 0.037% increase in audit fees ($227,522 ∗ 0.00037 or $84), where 6.709 and $227,522 are the mean values of ASP and audit fees, respectively. Put another way, a search delay of 6.709 days is associated with an increase in audit fees of $8418, or about 3.7% of mean audit fees. With the exception of RECV, INVT, and DYE, the coefficients on the control variables are statistically significant and are supportive of prior research findings. Interestingly, we find that firms whose auditors have resigned (RESIGN) pay lower audit fees. This appears to be inconsistent with expectations because we should expect riskier firms to pay higher fees. However, similar results have 21 When we replace ISSUE with eight separate indicator variables for each reportable event, we find that none of the indicator variables is individually statistically significant in the auditor choice model, while only one indicator variable (SEC investigation) is significant in the audit fee model. 22 Univariate correlations (untabulated) show that, as expected, LOSS is negatively correlated with BIGN at p b 0.01. 23 The breakdown of the sample observations is as follows: 2099 (38%), 939 (17%), and 2486 (45%) for Big N, second tier, and non-Big N, respectively. 24 Inspection of the variance inflation factors shows that all scores are under five for all the regression models. Thus, multicollinearity does not appear to be a significant concern.

Please cite this article as: Mande, V., et al., Auditor search periods as signals of engagement risk: Effects on auditor choice and audit pricing, Advances in Accounting (2017), http://dx.doi.org/10.1016/j.adiac.2017.03.001

10

V. Mande et al. / Advances in Accounting xxx (2017) xxx–xxx

Table 6 Pre- and post-section 404 regression results. Panel A: Auditor choice regression results Coefficient

Pre-404

Post-404

Pooled sample

LNASP

Dismissal sample

N LNASP

Resignation sample

N LNASP

−0.099 (−1.98)** 2288 −0.013 (−0.34) 1885 −0.372 (−3.54)*** 403

−0.082 (−1.80)** 3236 −0.045 (−0.74) 2397 −0.238 (−2.42)*** 839

N Panel B: Audit fee regression results Coefficient

Post-SOX, Pre-404

Post-404

Pooled sample

LNASP

Dismissal sample

N LNASP

Resignation sample

N LNASP

0.043 (2.06)** 2288 0.073 (2.57)*** 1885 0.047 (1.59)* 403

0.046 (3.96)*** 3236 0.057 (3.32)*** 2397 0.036 (2.01)** 839

N

Panel A presents results of logistic regressions that examine the effect of auditor search periods on auditor-choice, separately by pre- and post-SOX 404 periods. The dependent variable is BIGN, a dummy variable representing a choice of Big N auditor. The variable of interest is LNASP, natural log of auditor search periods. Panel B presents results of OLS regressions that examine the effect of auditor search periods on initial audit fees, separately by pre- and post-SOX 404 periods. The dependent variable is LNAF, natural log of audit fees. The variable of interest is LNASP, natural log of auditor search periods. We include all control variables but only report the results for variables of interest for the sake of brevity. *, **, *** indicate statistical significance at the 10%, 5%, and 1%, respectively (one-tailed if predicted, and two-tailed otherwise).

been documented by Elliott et al. (2013) in a sample consisting of non-Big N successor auditors.25 26 Lastly, our results show that LNASP has a positive coefficient in both auditor dismissal and resignation subsamples that is statistically significant at p b 0.01. This confirms earlier findings that that auditor search periods proxy as risk factors in all auditor change events, after controlling for known risk factors.

4.4.1. Big N versus non-Big N auditors Big N auditors are more sensitive to risk factors because of their “deep pockets” (Dye, 1993). Elliott et al. (2013) document that only Big N auditors are able to charge a premium for risky clients. Non-Big N auditors also have less bargaining power than Big N auditors in negotiating fees (Ghosh & Lustgarten, 2006). Interestingly, however, we find (untabulated) a statistically significant and positive association between LNASP and audit fees (p b 0.05) for both Big N and non-Big N auditor groups.27 Therefore,

25 Specifically, Elliott et al. (2013) find significantly lower (higher) audit fees for firms whose auditors have resigned in a sub-sample of non-Big N (Big N) successor auditors. They do not provide any explanation for their findings. Similar to Elliott et al., we estimate our regression models using two separate sub-samples: Big N and non-Big N successor auditors. We also find a significant negative coefficient on the RESIGN variable only for the non-Big N subsample. 26 We believe that there are two opposing forces at work here. On the one hand, as resignations are riskier events than dismissals, we expect clients to pay higher audit fees following resignations. On the other hand, client firms whose auditors have resigned are smaller in size and more likely to hire a non-Big N auditor, and, therefore, are expected to pay lower audit fees. Our findings support the idea that the net effect of these forces results in lower fees for clients hiring a non-Big N firm following resignations (see also Elliott et al., 2013). Our conjecture is that the non-Big N firms, relative to Big N firms, do not significantly increase fees (net) for resignation firms because they are less risk averse than the Big N firms who face greater litigation exposure. 27 We also find that ISSUE (i.e., reportable events in Forms 8-K) is positively and significantly related to audit fees at p b 0.01 in both Big N and non-Big N sub-samples. Our results showing significance on ASP suggest that ASP has valuable information that is not currently captured in the required disclosures.

our results do not support the idea that only Big N auditors are able to charge a premium for high audit risks.

4.4.2. Initial audit fees versus “normal” fees Next, we examine whether our results are possibly driven by differences in how auditors charge fees for their new clients versus their continuing clients. Specifically, our interest is in comparing initial audit fees for positive ASP clients with “normal” audit fees, as proxied by fees charged by auditors to the remainder of their continuing clients. Using the entire population of firms on Audit Analytics (58,500 observations) for which data on audit fee determinants are available, we first run a fee-regression that includes a dummy variable, SWITCH, taking a value of 1 when an auditor change occurs and 0 otherwise. In untabulated results, we find that the coefficient on SWITCH is −0.229 and statistically significant at p b 0.01. This supports the well documented practice of lowballing of fees (DeAngelo, 1981; Ghosh & Lustgarten, 2006; Hogan & Wilkins, 2008), with our results showing that successor auditors charge 20% lower fees for the first audit, after controlling for all other determinants.28 Using the same sample, we then estimate a fee-regression with two dummy variables, POSASP, taking a value of 1 for positive ASP firms and 0 otherwise, and, ZEROASP, taking a value of 1 for zero ASP firms and 0 otherwise. We find statistically significant coefficients of − 0.129 and − 0.247 on POSASP and ZEROASP respectively, suggesting that new clients belonging to these two groups receive fee discounts of 12% and 22% relative to continuing clients. The difference between the coefficients on POSASP and ZEROASP is statistically significant at p b 0.01 (Chi-square 20.18). Therefore, while successor auditors offer a fee discount to the POSASP group, the

28

This is calculated as follows: 1 − exp (−0.229) = 0.20.

Please cite this article as: Mande, V., et al., Auditor search periods as signals of engagement risk: Effects on auditor choice and audit pricing, Advances in Accounting (2017), http://dx.doi.org/10.1016/j.adiac.2017.03.001

V. Mande et al. / Advances in Accounting xxx (2017) xxx–xxx

11

Table 7 Logistic regression (first stage) results: probability of positive versus zero ASP. Variable

Expected sign

Pooled sample

Dismissal subsample

Resignation subsample

Intercept

±

FINDIST

+

LOSS

+

SALGR

+

SIZE

+

VAR

+

RECINV

+

INTCTR

+

INTEGR

+

YREND

+

NAF

–

SPECIAL

–

FORG

+

INCONS

+

DFILE

+

B4NB4

±

NB4NB4

±

NB4B4

±

TECH

+

QUALGC

+

Q1

+

Q4

–

−1.068 (−1.69)* −0.018 (−0.86) −0.049 (−0.46) −0.068 (−1.00) −0.091 (−2.93)*** 18.357 (1.21) −0.028 (−0.14)) 0.470 (3.69)*** 1.385 (2.80)*** 0.039 (0.32) −0.288 (−1.36)* 0.013 (0.12) 0.171 (1.66)** 0.401 (1.61)* 0.001 (1.70)** −0.062 (−0.49) −0.140 (−0.89) −0.045 (−0.26) 0.307 (2.65)*** 0.236 (1.19) 0.038 (0.34) −0.064 (−0.47)

−0.359 (−0.49) 0.085 (2.13)** −0.344 (−1.63) −0.159 (−1.59) −0.009 (−0.15) 26.876 (1.02) −0.524 (−1.27) 1.031 (4.05)*** 1.723 (2.48)*** −0.037 (−0.16) −1.768 (−4.14)*** 0.662 (2.06)** 0.368 (1.64)* 1.049 (2.06)** 0.002 (2.81)*** 1.189 (3.95)*** −0.287 (−1.05) −0.029 (−0.07) 0.565 (2.51)*** 0.181 (0.56) 0.419 (1.87)** −0.156 (−0.60)

RESIGN

+

−1.395 (−2.24)** 0.005 (0.29) −0.046 (−0.50) −0.099 (−1.74)* −0.064 (−2.41)** 21.933 (1.70)** −0.120 (−0.67) 0.624 (6.04)*** 1.409 (3.95)*** −0.008 (−0.08) −0.590 (−3.16)*** 0.053 (0.52) 0.202 (2.22)** 0.563 (2.98)*** 0.001 (2.86)*** 0.116 (1.07) −0.486 (−3.44)*** −0.007 (−0.04) 0.370 (3.71)*** 0.309 (1.94)** 0.124 (1.32)* −0.070 (−0.58) 1.928 (20.55)*** 0.40 4888

0.38 3967

0.41 921

Max-rescaled R2 N

*, **, *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively (one-tailed if predicted, and two-tailed otherwise). All regressions are estimated using robust standard errors clustered by firm. The first stage model uses the same variables used in Khalil et al. (2011) which are as follows: FINDIST: Zmijewski's (1984) financial condition index. LOSS: dummy variable that is equal to 1 when a firm reports a loss in the year preceding the auditor change, 0 otherwise. SALGR: growth in sales scaled by prior year sales. SIZE: natural log of the market value of equity. VAR: variance in daily abnormal returns during the year preceding an auditor change. RECINV: sum of receivables and inventory scaled by total assets. INTCTR: dummy variable that is equal to 1 when internal control weaknesses is reported by the incumbent auditor or the firm in Audit Analytics. INTEG: dummy variable that is equal to 1 when the incumbent auditor or the firm reports issues related to management representation and/or the presence of illegal acts by top management, 0 otherwise. YREND: dummy variable that is equal to 1 for firms not having a June 30 or December 31 year end, 0 otherwise. NAF: nonaudit fees scaled by total fees. SPECIAL: an audit firm is coded a specialist when it audits 25% or more of the clients in the two-digit SIC industry. FORG: dummy variable that is equal to 1 for firms reporting foreign income, 0 otherwise. INCONS: dummy variable that is equal to 1 for cases where the exhibit letter filed by the incumbent auditor following an auditor change disagrees with the Form 8-K filed by the firm, 0 otherwise. DFILE: the fiscal year end plus 60 days minus the auditor change date for accelerated filers and the fiscal year end plus 90 days for non-accelerated filers. B4B4: dummy variable that is equal to 1 when a firm switches from a Big 4 to a Big 4 auditor, 0 otherwise. B4NB4: dummy variable that is equal to 1 when a firm switches from a Big 4 to a non-Big 4 auditor, 0 otherwise. NB4NB4: dummy variable that is equal to 1 when a firm switches from a non-Big 4 to a non-Big 4 auditor, 0 otherwise. NB4B4: dummy variable that is equal to 1 when a firm switches from a non-Big 4 to a Big 4 auditor, 0 otherwise. TECH: variable indicating firms with the following SIC codes: 2833–2836, 3570–3577, 3600–3674, 7371–7379, and 8731–8734. QUALGC: dummy variable that is equal to 1 when the audit opinion is qualified for scope limitation or going concern reasons in the year preceding the auditor change, 0 otherwise. Q1 (Q4): dummy variable that is equal to 1 when an auditor switch occurs in the first (fourth) quarter, 0 otherwise. RESIGN: dummy variable that is equal to 1 when an audit firm resigns, 0 otherwise.

discount is smaller, which is consistent with the idea that the positive ASP firms are riskier than zero ASP firms. 4.5. Stock market reaction results Panel A of Table 5 shows the three-day market returns (CAR) surrounding predecessor auditors' departure dates. We observe a

significant difference in CAR of zero ASP (− 0.0003) and positive ASP (− 0.0042) groups. As predicted, the market appears to perceive auditor switches more negatively where a subsequent auditor is not chosen as of the predecessor auditor's departure date. We also find that positive ASP firms whose auditors have resigned experience more negative market returns (− 0.0065) compared to zero ASP firms (− 0.0006).

Please cite this article as: Mande, V., et al., Auditor search periods as signals of engagement risk: Effects on auditor choice and audit pricing, Advances in Accounting (2017), http://dx.doi.org/10.1016/j.adiac.2017.03.001

12

V. Mande et al. / Advances in Accounting xxx (2017) xxx–xxx

Table 8 Second stage regression results. Panel A: Auditor choice regression results using the matched pair method Variable

Expected sign

Pooled sample

Dismissal subsample

Resignation subsample

LNASP

–

SIZE

+

FRGN

+

SEG

+

RECV

–

INVT

–

LEV

–

LOSS

–

CATA

+

ACCR

–

EBIT

+

QUICK

+

LITIND

±

IC

–

REST

–

BM

–

GC

–

DYE

±

PREAU

+

ISSUE

–

0.114 (1.36) 1.297 (41.49)*** −0.247 (0.49) −0.023 (0.07) −2.179 (4.33)** −1.596 (0.82) −2.256 (6.51)** −0.040 (0.01) 1.398 (2.44) 0.398 (0.17) −0.952 (1.36) −0.021 (0.14) 0.173 (0.15) 0.347 (0.43) 0.396 (1.01) −0.754 (11.77)*** −0.398 (0.26) −0.622 (3.11)* −0.569 (2.03) −0.389 (1.34)

−0.249 (4.42)** 0.857 (10.97)*** −1.070 (0.95) −0.163 (1.09) −1.418 (0.87) 4.628 (3.48)* −0.034 (0.00) 0.742 (1.34) 1.014 (0.52) −5.406 (2.89)* 1.312 (0.78) −0.212 (2.25) 0.490 (0.53) −1.252 (2.18) 0.202 (0.14) −0.489 (2.05) 0.072 (0.01) 2.550 (10.03)*** 0.835 (1.86) −0.019 (0.00)

RESIGN

–

−0.123 (4.08)** 0.741 (47.40)*** 0.327 (1.15) −0.017 (0.07) −1.471 (4.04)** −0.940 (0.90) −0.210 (0.16) 0.151 (0.30) 0.882 (1.74) −0.629 (0.55) 0.615 (1.01) 0.035 (0.38) 0.185 (0.37) −0.199 (0.30) 0.231 (0.69) −0.462 (8.85)*** −1.127 (3.47)* 0.859 (9.64)*** −0.361 (1.33) −0.222 (1.06) −2.687 (10.33)*** 0.47 1940

0.55 1104

0.55 836

Max-rescaled R2 N Panel B: Audit fee regression results using the matched pair method Variable

Expected sign

Pooled sample

Dismissal subsample

Resignation subsample

Intercept

±

LNASP

+

SIZE

+

FRGN

+

SEG

+

RECV

+

INVT

+

LEV

+

LOSS

+

CATA

+

ACCR

+

EBIT

–

QUICK

–

10.06 (16.02)*** 0.020 (1.62)* 0.528 (27.31)*** 0.217 (3.20)*** 0.049 (3.73)*** −1.029 (−6.76)*** −0.338 (−1.72)* 0.141 (1.28) 0.180 (2.96)*** 0.992 (8.05)*** 0.261 (1.71)** 0.140 (1.16) −0.045 (−3.77)***

8.56 (12.76)*** 0.015 (0.73) 0.586 (20.13)*** 0.073 (0.87) 0.011 (0.61) 0.215 (0.73) 0.33 (0.94) −0.087 (−0.61) 0.068 (0.79) 0.551 (2.63)*** 0.238 (1.05) −0.336 (−1.93)** −0.032 (−2.29)**

9.945 (16.26)*** 0.037 (2.18)** 0.569 (19.01)*** 0.124 (1.01) 0.042 (1.79)** −1.155 (−5.25)*** −0.519 (−1.59) −0.043 (−0.30) 0.453 (4.74)*** 0.922 (4.52)*** 0.441 (1.94)** 0.635 (3.74)*** −0.053 (−2.78)***

Please cite this article as: Mande, V., et al., Auditor search periods as signals of engagement risk: Effects on auditor choice and audit pricing, Advances in Accounting (2017), http://dx.doi.org/10.1016/j.adiac.2017.03.001

V. Mande et al. / Advances in Accounting xxx (2017) xxx–xxx

13

Table 8 (continued) Panel B: Audit fee regression results using the matched pair method Variable

Expected sign

Pooled sample

Dismissal subsample

Resignation subsample

LITIND

+

IC

+

REST

+

BM

–

GC

+

DYE

+

BIGN

+

ISSUE

+

−0.072 (−1.14) 0.238 (2.89)*** −0.011 (0.09) −0.068 (−1.94)* 0.278 (2.73)*** −0.115 (−1.97)** 0.099 (1.48)* −0.026 (−0.58) −0.607 (−3.61)*** 0.70 1940

−0.061 (−0.47) 0.223 (1.82)** 0.284 (1.46) −0.136 (−2.83)*** −0.058 (−0.34) −0.111 (−1.44) 0.093 (1.05) −0.031 (−0.44)

0.028 (0.28) 0.043 (0.39) 0.069 (0.41) −0.154 (−2.90)*** 0.640 (5.01)*** −0.144 (−1.37) 0.295 (2.69)*** −0.110 (−2.10)**

0.71 1104

0.69 836

RESIGN

± 2

Adjusted R N

*, **, *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively (one-tailed if predicted, and two-tailed otherwise). All regressions are estimated using robust standard errors clustered by firm. See Appendix A for the variable definitions.

Panel B regression results using the whole sample show a negative and statistically significant coefficient on POSASP (p b 0.05).29 The coefficient on POSASP (− 0.006) suggests that the effect of this variable is economically significant, i.e., it is 600% of the mean CAR of − 0.001 (Panel C of Table 2) for the full sample. Additionally, we find that investor reaction to POSASP is more negative in the sample of resignations (p b 0.10), as evidenced by a negative coefficient on RESIGN ∗ POSASP. In resignation and dismissal sub-sample tests, however, we find a negative coefficient on POSASP only in the case of auditor resignations. For this sub-sample, we also find negative coefficients on ISSUE and LNTENURE implying that the stock market takes a negative view of reportable event disclosures and lengthy tenures of predecessor auditors. 4.6. Additional analyses 4.6.1. Positive auditor search periods To alleviate concerns that systematic differences in risk factors between firms having positive and zero ASPs might be driving our results, we test our hypotheses using a sample of only positive ASP firms. Mean values of ASP (in days) are: 27.47, 15.26, and 41.12 for pooled, dismissal, and resignation samples, respectively. Untabulated results show in both auditor selection and fee models that the coefficient on LNASP has the predicted sign and is statistically significant at least at p b 0.1 in the pooled, dismissal and resignation samples. The findings confirm that ASPs matter in auditor choice and audit fee determinations even in a within-group analysis of positive ASP firms. The results for the control variables are similar to those reported in the previous sections. 4.6.2. Pre- versus post-section 404 Section 404 significantly increased reporting requirements for firms which in turn placed additional demands on manpower resources of audit firms and resulted in higher auditor fees. In response to Section 404, many auditors (mostly Big N) also dropped their riskier audit clients (Rama & Read, 2006). To examine how SOX 404 changes impact the associations between ASPs and auditor choice/audit fees, we separate the data into pre- and post-SOX 404 subsamples using November 29 Compared to the auditor-related regression results, a weaker reaction from investors could be expected because unlike auditors who have private information about their clients, investors face higher information asymmetry about the circumstances around auditor changes. As the stock market sample is a subset of the sample used to test H1 and H2, we replicate our tests of H1 and H2 using the smaller sample. Our results are qualitatively similar to those reported.

15, 2004 as the cutoff date. Results in Table 6 show that in both periods LNASP is significantly related to auditor choice only in the sub-sample of auditor resignations (Panel A). The effect of LNASP on audit fees, however, is positive in both post-SOX periods and in both dismissal and resignation sub-samples (Panel B). The results for the control variables (not shown) are generally qualitatively similar to those reported in Table 3. Overall, our prior findings are supported, excepting for the auditor choice results using the dismissal subsample where we find, as expected, a negative coefficient, albeit statistically insignificant.

4.6.3. Propensity score matching Despite our efforts to control for known risk factors, we acknowledge that there may be risk variables that we may have omitted or not adequately controlled for in the tests. We mitigate this concern using a matched sample method with propensity scores. Using this method, however, results in the loss of more than a majority of the observations (65% of sample). We first estimate a logistic model that predicts the probability of having a positive versus a zero ASP. We then match each observation from the group reporting positive ASPs with an observation from the group reporting zero ASPs whose propensity score (i.e., the predicted values from the logistic model) is closest to the score of the matched positive ASP observation.30 Therefore, the paired observations, although differing significantly in their ASPs, are expected to have similar other risk characteristics. In the first stage model, we obtain propensity scores using all of the variables used by Khalil et al. (2011).31 The dependent variable is coded based on whether a firm has a positive or a zero ASP. Additionally, we include a control for auditor resignations (RESIGN) because unlike Khalil et al. we also include dismissals in our sample. We also include two variables to control for the timing of an auditor switch.32 Firms switching 30 We use a one-to-one firm matching with a caliper of 0.1, where a suggested range is [0.1 to 0.9] (Caliendo & Kopeinig, 2008). 31 Khalil et al. (2011) is the only study that we are aware of, examining the determinants of auditor search periods. Their sample only consists of 216 auditor resignations. In contrast, our sample has 5524 auditor changes including 1242 resignations. 32 Control variables used in the first stage model include variables for: (1) client business risk – client's financial condition, sales growth, size, and stock price volatility, (2) audit risk – accounts receivables and inventory as a percentage of total assets, weaknesses in internal controls, going concern opinions or restatements, and management integrity issues, (3) auditor business risk – timing of the engagement, non-audit fees, whether a firm has foreign operations, and industry belonging of client, and (4) other control variables – the time interval from the auditor's termination date to the annual 10-K filing date, type of auditor switch, whether the client is a high tech firm, and a dummy for auditor resignations.

Please cite this article as: Mande, V., et al., Auditor search periods as signals of engagement risk: Effects on auditor choice and audit pricing, Advances in Accounting (2017), http://dx.doi.org/10.1016/j.adiac.2017.03.001

14

V. Mande et al. / Advances in Accounting xxx (2017) xxx–xxx

auditors in the first (fourth) quarter are possibly under less (more) time-pressure to find a new auditor and, therefore, can be expected to have longer (shorter) ASPs.33 The results of estimating the first stage regressions are presented in Table 7 for the pooled, dismissal and resignation samples separately.34 Overall, the results show that the likelihood of positive ASPs increases for firms having: lower growth (SALGR), smaller market capitalization (SIZE), higher volatility of stock returns (VAR), smaller non-audit fees (NAF), larger foreign income (FORG), technology industry membership (TECH), reportable issues (INTCTR, INTEGR, INCONS, and QUALGC) and auditor resignations (RESIGN). We also find that the coefficient on Q1 is positive and statistically significant at least at p b 0.05 in pooled and resignation samples, suggesting that client firms are under less time pressure to find a new auditor in the first quarter.35 These results are generally consistent with the notion that firms having positive ASPs pose greater risk to auditors than those with zero ASPs. Panels A and B of Table 8 show estimation results using paired observations matched on propensity scores for the auditor choice and auditor fee models, respectively.36 Due to the matching procedure, the statistical significance of many control variables disappears. In both models, however, we find that the variable of interest, LNASP, has the predicted sign and is statistically significant using the pooled and resignation samples. However, once again this statistical significance disappears when the dismissal sample is used. 5. Conclusion We investigate the impact of auditor search periods on the choice of a successor auditor, the pricing of initial audits following an auditor change, and the stock market reaction to the termination of an old auditor. A longer internal due diligence process and rejections by other more risk-sensitive audit firms can elongate the search for a successor auditor. Once auditors accept risky clients (i.e., those with long ASP), the successor auditors must expend more effort and time to gather and analyze information about prospective high risk clients. Therefore, we argue that lengthy searches for a new auditor signal the presence of high engagement risk to investors, regulators, and academics. Our main results based on multivariate regression analyses are as follows. First, we find that clients with long auditor search periods are less likely to be accepted by Big-N successor auditors. Second, after controlling for previously identified correlates, we find that audit fees are higher for clients having long search periods. Lastly, as predicted, we find that there are significantly higher negative market returns for firms that experience delays in appointing their new auditor relative to firms that do not experience appointment delays. Additional analyses find, however, that investors only regard search periods negatively for companies whose predecessor auditors have resigned. A weaker market response to engagement delays is consistent with the idea that investors face more information asymmetry about auditor changes than do the companies' auditors. We conduct numerous additional tests using sub-samples to check for robustness of our results. These additional analyses show that there is weaker support for our hypotheses in the case of auditor dismissals. While a weaker association is to be expected with dismissals,

33 Of 4888 observations used for the regression, 978 auditor changes (20%) of the sample occur in the first quarter and 880 changes (18%) in the fourth quarter. 34 The areas under the receiver-operating characteristic (ROC) curves, which measure a predictive power of the model, are 0.73, 0.72, and 0.74, respectively, which suggest that the models provide a good fit. 35 For example, audit firms do not generally perform any internal controls testing in the first quarter. 36 Matching is done with replacement due to relatively small sample size. Following Cram, Karan, and Stuart (2009), we use conditional analysis methods since a matched sample is a nonrandom sample.

the smaller sample size used in these sub-sample tests can also be a factor affecting the results. It is noteworthy that with auditor resignations, however, there is consistent support for our hypotheses. This study contributes to the scant accounting literature on auditor search periods (Smith, 1988; Khalil et al., 2011) and complements emerging research (Elliott et al., 2013) investigating the influence of risk factors on successor auditor choice and initial audit fees (Jones & Raghunandan, 1998; Shu, 2000; Francis & Krishnan, 2002; Choi et al., 2004; Catanach et al., 2011). Future research is needed on whether our results are robust to the time-period examined and whether they also hold in other countries. A limitation of our study concerns the measurement of the auditor search period. Because companies can start their auditor search prior to the reported date of the auditor change, the auditor search period used in this study underreports the actual search period. We are unable to predict the impact of this measurement error on our test results. Acknowledgement We appreciate the helpful comments and suggestions from workshop participants at the 2015 AAA Annual Meeting and the 2015 AAA Auditing Section Meeting. Appendix A. Variable definitions

Variable

Definition

Auditor choice and audit fee model BIGN 1 if the successor auditor is a Big N firm, and 0 otherwise; LNAF Natural log of audit fees; ASP Number of days between the resignation or dismissal of the predecessor auditor and the appointment of the successor auditor; LNASP Natural log of auditor search period (ASP); SIZE Natural log of total assets; FRGN Ratio of sales of foreign subsidiaries to total sales; SEG Number of business segments; RECV Ratio of total receivables to total assets; INVT ratio of total inventory to total assets; LEV Ratio of total liabilities to total assets; LOSS 1 if earnings before extraordinary items are b0, and 0 otherwise; CATA Ratio of current assets to total assets; ACCR Ratio of the absolute value of total accruals to total assets; EBIT Ratio of earnings before interest and tax to total assets; QUICK Ratio of current assets (less inventory) to current liabilities; LITIND 1 if the firm is in a high-litigation industry, and 0 otherwise; High-litigation industries include biotechnology (SIC codes 2833–2836, 8731–8734), computer (3570–3577, 7370–7374), electronics (3600–3674) and retail (5200–5961). IC 1 if a firm reports internal control weaknesses in a Form 8-K auditor change filing, and 0 otherwise; REST 1 if a firm reports restatements in a Form 8-K auditor change filing, and 0 otherwise; BM Total book values divided by total market values; GC 1 if a firm receives a going-concern opinion prior to the auditor switch, and 0 otherwise; DYE 1 if the firm has a December fiscal year-end, and 0 otherwise; PREAU 1 if the predecessor auditor is a Big N firm, and 0 otherwise; ISSUE Number of reportable events disclosed in the Form 8-K auditor change filing; and RESIGN 1 if the auditor resigns, and 0 if the auditor is dismissed. Market reaction model CAR Cumulative abnormal stock return over a three-day window surrounding the predecessor auditor departure date; POSASP 1 if a successor auditor was not immediately engaged, and 0 otherwise; ISSUE Number of reportable events disclosed in the Form 8-K auditor change filing; LNMKT Natural logarithm of the market value of equity; LNTENURE Natural logarithm of the number of years that the firm had engaged its predecessor auditor. TIMING 1 if an auditor switch occurs in the fourth quarter, and 0 otherwise; and RESIGN 1 if the auditor resigns, and 0 if the auditor is dismissed.

Please cite this article as: Mande, V., et al., Auditor search periods as signals of engagement risk: Effects on auditor choice and audit pricing, Advances in Accounting (2017), http://dx.doi.org/10.1016/j.adiac.2017.03.001

V. Mande et al. / Advances in Accounting xxx (2017) xxx–xxx

References Asare, S., Cohen, J., & Trompeter, G. (2005). The effect of nonaudit services on client acceptance and staffing decisions. Journal of Accounting and Public Policy, 24(6), 489–520. Ayers, S., & Kaplan, S. E. (1998). Potential differences between engagement and risk review partners and their effect on client acceptance judgments. Accounting Horizons, 12(2), 139–153. Bell, T. B., Landsman, W. R., & Shackelford, D. A. (2001). Auditors' perceived business risk and audit fees: Analysis and evidence. Journal of Accounting Research, 39(1), 35–43. Bockus, K., & Gigler, F. (1998). A theory of auditor resignation. Journal of Accounting Research, 36(2), 191–208. Caliendo, M., & Kopeinig, S. (2008). Some practical guidance for the implementation of propensity score matching. Journal of Economic Surveys, 22, 31–72. Carter, M. E., & Soo, B. S. (1999). The relevance of form 8-K reports. Journal of Accounting Research, 37(1), 119–132. Catanach, A., Irving, J. H., Williams, S. P., & Walker, P. L. (2011). An ex post examination of auditor resignations. Accounting Horizons, 25(2), 267–283. Cenker, W. J., & Nagy, A. L. (2008). Auditor resignations and auditor industry specialization. Accounting Horizons, 22(3), 279–295. Choi, J., Doogar, R. K., & Ganguly, A. R. (2004). The riskiness of large audit firm client portfolios and changes in audit liability regimes: Evidence from the U.S. audit market. Contemporary Accounting Research, 21(4), 747–785. Cohen, J., & Hanno, D. (2000). Auditors' consideration of corporate governance and management control philosophy in preplanning and planning judgments. Auditing: A Journal of Practice & Theory, 19(2), 133–146. Cram, D. P., Karan, V., & Stuart, I. (2009). Three threats to validity of choice-based and matched-sample studies in accounting research. Contemporary Accounting Research, 26(2), 477–516. DeAngelo, L. (1981). Auditor independence, "lowballing" and disclosure regulation. Journal of Accounting and Economics, 3(2), 113–127. DeFond, M., & Zhang, J. (2014). A review of archival auditing research. Journal of Accounting and Economics, 58(2/3), 275–326. Doogar, R., Sivadasan, P., & Solomon, I. (2010). The regulation of public company auditing: Evidence from the transition to AS5. Journal of Accounting Research, 48(4), 795–814. Dye, R. A. (1993). Auditing standards, legal liability, and auditor wealth. Journal of Political Economy, 101(5), 887–914. Elder, R. J., Zhang, Y., Zhou, J., & Zhou, N. (2009). Internal control weaknesses and client risk management. Journal of Accounting, Auditing & Finance, 24(4), 543–579. Elliott, J. A., Ghosh, A., & Peltier, E. (2013). Pricing of risky initial audit engagements. Auditing: A Journal of Practice & Theory, 32(4), 25–43. Ettrege, M., Johnstone, K., Stone, M., & Wang, Q. (2011). The effects of firm size, corporate governance quality, and bad news on disclosure compliance. Review of Accounting Studies, 16, 866–889. Feltham, G. A., Hughes, J. S., & Simunic, D. A. (1991). Empirical assessment of the impact of auditor quality on the valuation of new issues. Journal of Accounting and Economics, 14(4), 375–399. Ferguson, A., Francis, J. R., & Stokes, D. J. (2003). The effects of firm-wide and office-level industry expertise on audit pricing. The Accounting Review, 78(2), 429–448. Feroz, E., Park, K., & Pastena, V. (1991). The financial and market effects of the SEC's Accounting and Auditing Enforcement Releases. Journal of Accounting Research, 29, 107–148 (Supplement). Francis, J. R., & Krishnan, J. (2002). Evidence on auditor risk management strategies before and after the Private Securities Litigation Reform Act of 1995. Asia-Pacific Journal of Accounting & Economics, 9(2), 135–157. Francis, J. R., & Wang, D. (2005). Impact of the SEC's public fee disclosure requirement on subsequent period fees and implications for market efficiency. Auditing: A Journal of Practice & Theory, 24, 145–160 (Supplement). Francis, J., Philbrick, D., & Schipper, K. (1994). Shareholder litigation and corporate disclosure. Journal of Accounting Research, 32(2), 137–164. Francis, J. R., Reichelt, K., & Wang, D. (2005). The pricing of national and city-specific reputations for industry expertise in the U.S. audit market. The Accounting Review, 80(1), 113–136. Geiger, M., Raghunandan, K., & Rama, D. (2006). Auditor decision-making in different litigation environments: The Private Securities Litigation Reform Act, audit reports, and audit firm size. Journal of Accounting and Public Policy, 25(3), 332–353. Ghosh, A., & Lustgarten, S. (2006). Pricing of initial audit engagements by large and small audit firms. Contemporary Accounting Research, 23(2), 333–368. Ghosh, A., & Tang, C. Y. (2015). Auditor resignation and risk factors. Accounting Horizons, 29(3), 529–549. Griffin, P. A., & Lont, D. H. (2010). Do investors care about auditor dismissals and resignations? What drives the response? Auditing: A Journal of Practice & Theory, 29(2), 189–214. Hackenbrack, K. E., & Hogan, C. E. (2002). Market response to earnings surprises conditional on reasons for an auditor change. Contemporary Accounting Research, 19(2), 195–223. Hackenbrack, K., Jenkins, N. T., & Pevzner, M. (2014). Relevant but delayed information in negotiated audit fees: Evidence from stock price crashes. Auditing: A Journal of Practice & Theory, 33(4), 95–117. Hay, D. C., Knechel, W. R., & Wong, N. (2006). Audit fees: A meta-analysis of the effect of supply and demand attributes. Contemporary Accounting Research, 23(1), 141–191. Hogan, C. E., & Martin, R. D. (2009). Risk shifts in the market for audits: An examination of changes in risk for second tier audit firms. Auditing: A Journal of Practice & Theory, 28(2), 93–118. Hogan, C. E., & Wilkins, M. S. (2008). Evidence on the audit risk model: Do auditors increase audit fees in the presence of internal control deficiencies? Contemporary Accounting Research, 25(1), 219–242.

15

Huang, Y., & Scholz, S. (2012). Evidence on the association between financial restatements and auditor resignations. Accounting Horizons, 26(3), 439–464. Huang, H., Raghunandan, K., & Rama, D. (2009). Audit fees for initial audit engagements before and after SOX. Auditing: A Journal of Practice & Theory, 28(1), 171–190. Johnson, W. B., & Lys, T. (1990). The market for audit services. Journal of Accounting and Economics, 12(January), 281–308. Johnstone, K. M. (2000). Client-acceptance decisions: Simultaneous effects of client business risk, audit risk, auditor business risk, and risk adaptation. Auditing: A Journal of Practice & Theory, 19(1), 1–25. Johnstone, K. M. (2001). Risk, experience and client acceptance decisions. National Public Accountant, 46(5), 27–30. Johnstone, K. M., & Bedard, J. C. (2003). Risk management in client acceptance decisions. The Accounting Review, 78(4), 1003–1025. Johnstone, K. M., & Bedard, J. C. (2004). Audit firm portfolio management decisions. Journal of Accounting Research, 42(4), 659–690. Jones, F. L., & Raghunandan, K. (1998). Client risk and recent changes in the market for audit services. Journal of Accounting and Public Policy, 17(2), 169–181. Khalil, S. K., & Mazboudi, M. (2016). Client acceptance and engagement pricing following auditor resignations in family firms. Auditing: A Journal of Practice & Theory, 35(4), 137–158. Khalil, S. K., Cohen, J. R., & Schwartz, K. B. (2011). Client engagement risks and the auditor search period. Accounting Horizons, 25(4), 685–702. Klock, M. (1994). The stock market reaction to a change in certifying accountant. Journal of Accounting, Auditing & Finance, 9, 339–347. Knechel, W. R., Naiker, V., & Pacheco, G. (2007). Does auditor industry specialization matter? Evidence from market reaction to auditor switches. Auditing: A Journal of Practice & Theory, 26(1), 19–45. Krishnan, J., & Krishnan, J. (1997). Litigation risk and auditor resignations. The Accounting Review, 72(4), 539–560. Landsman, W., Nelson, K., & Rountree, B. (2009). Auditor switches in the pre- and postEnron eras: Risk or realignment? The Accounting Review, 84(2), 531–558. Lee, H., Mande, V., & Ortman, R. (2004). The effect of audit committee and board of director independence on auditor resignation. Auditing: A Journal of Practice & Theory, 23(2), 131–146. Li, C. (2008). Does client importance affect auditor independence at the office level? Empirical evidence from going concern opinions. Contemporary Accounting Research, 26(1), 201–230. Newton, N. J., Persellin, J. S., Wang, D., & Wilkins, M. S. (2016). Internal control opinion shopping and audit market competition. The Accounting Review, 91(2), 603–623. Pratt, J., & Stice, J. D. (1994). The effects of client characteristics on auditor litigation risk judgments, required audit evidence, and recommended audit fees. The Accounting Review, 69(4), 639–656. Public Company Accounting Oversight Board (PCAOB) (2004). Auditing standard no. 2 (AS2): An audit of internal control over financial reporting performed in conjunction with an audit of financial statements. Washington, DC: PCAOB http://pcaobus.org/ Standards/Auditing/Pages/Auditing_Standard_2.aspx Public Company Accounting Oversight Board (PCAOB) (2007). Auditing standard no. 5 (AS5): An audit of internal control over reporting that is integrated with audit of financial statements and related independence rule and conforming amendments. Washington, DC: PCAOB http://pcaobus.org/Standards/Auditing/Pages/Auditing_Standard_5.aspx Raghunandan, K., & Rama, D. (1999). Auditor resignations and the market for audit services. Auditing: A Journal of Practice & Theory, 18(1), 124–134. Rama, D., & Read, W. (2006). Resignations by the big 4 and the market for audit services. Accounting Horizons, 20(2), 97–109. Scholz, S. (2008). The changing nature and consequences of public company financial restatements: 1997–2006. (Treasury Department Report). Schroeder, J. H., & Hogan, C. E. (2013). The impact of PCAOB AS5 and the economic recession on client portfolio characteristics of the Big 4 audit firms. Auditing: A Journal of Practice & Theory, 32(4), 95–127. Shu, S. Z. (2000). Auditor resignations: Clientele effects and legal liability. Journal of Accounting and Economics, 29(2), 173–205. Simunic, D. A., & Stein, M. T. (1990). Audit risk in a client portfolio context. Contemporary Accounting Research, 6(2), 329–343. Smith, D. B. (1988). An investigation of securities and exchange commission regulation of auditor change disclosures: The case of accounting series release no. 165. Journal of Accounting Research, 26(1), 134–145. Stice, J. (1991). Using financial and market information to identify pre-engagement factors associated with lawsuits against auditors. The Accounting Review, 66(3), 516–533. Turner, L., Williams, J., & Weirich, T. (2005). An inside look at auditor changes. CPA Journal, 12–21 (Supplement). Venkataraman, R., Weber, J. P., & Willenborg, M. (2008). Litigation risk, audit quality, and audit fees: Evidence from initial public offerings. The Accounting Review, 83(5), 1315–1345. Whisenant, J. S., Sankaraguruswamy, S., & Raghunandan, K. (2003a). Evidence on the joint determination of audit and non audit fees. Journal of Accounting Research, 41(4), 721–744. Whisenant, J. S., Sankaraguruswamy, S., & Raghunandan, K. (2003b). Market reactions to disclosure of reportable events. Auditing: A Journal of Practice & Theory, 22(1), 181–194. Zmijewski, M. E. (1984). Methodological issues related to the estimation of financial distress prediction models. Journal of Accounting Research, 22, 59–82 (Supplement).

Please cite this article as: Mande, V., et al., Auditor search periods as signals of engagement risk: Effects on auditor choice and audit pricing, Advances in Accounting (2017), http://dx.doi.org/10.1016/j.adiac.2017.03.001