chu et al - 2012 - does long tenure erode auditor independence

The average auditor tenure of the 500 100 largest companies in the U.S., based on market capitalization, is 21 28 years PCAOB 2011, 20.1Academic research has examined the association bet

Does Long Tenure Erode Auditor Independence?

Ling Chu School of Business and Economics Wilfrid Laurier University Waterloo, ON N2L 3C5 lchu@wlu.ca

Bryan K Church

800 West Peachtree Street College of Management Georgia Tech Atlanta, GA 30308-0520 404.894.3907 bryan.church@mgt.gatech.edu

Ping Zhang

Rotman School of Management University of Toronto Toronto, ON M5S 3E6 416-946-5655 pzhang@rotman.utoronto.ca

January 2012

Does Long Tenure Erode Auditor Independence?

Abstract

Regulators have shown a renewed interest in considering the merits of mandatory auditor

rotation A fundamental concern is that long tenure may undermine auditor independence We conduct a study to investigate the effects of long tenure on companies’ allowance for bad debts (ABD) We focus on ABD, as opposed to total accruals, because it allows us to hone in on the effect of auditor tenure on a specific accrual By examining ABD, we are able to conduct a more direct and powerful test and avoid some of the measurement error and noise associated with total accruals (e.g., discretionary or abnormal) We find evidence of a negative association between auditor tenure and estimated ABD The finding holds across a series of robustness tests Further analyses suggest that long tenure (around 15 years) is associated with downward bias in

estimated ABD With long tenure, the auditor endorses an estimate of ABD that is too

aggressive This result is consistent with the argument that long tenure leads to compromised independence Our findings suggest that a term limit of 10 years, which has been suggested elsewhere (e.g., PCAOB 2010; Chasan 2011), is sufficient to preserve auditor independence

JEL classification: M41

Does Long Tenure Erode Auditor Independence?

1 Introduction

Regulators are once again considering the merits of mandatory audit firm rotation Under Section 207 of the Sarbanes-Oxley Act of 2002 (SOX), the Comptroller General of the United States was directed to study the potential effects of requiring firm rotation The General

Accounting Office (GAO) completed a report in the fall of 2003 and concluded that, in light of other reforms enacted as a result of SOX, it was not cost beneficial to mandate audit firm rotation (GAO 2003) But, the GAO’s report states that further experience is necessary to evaluate the effect of other SOX reforms to gauge the implications for auditor independence and audit quality: that is, to determine whether other SOX reforms are sufficient

In the years since the GAO’s study, pundits (e.g., Sikka 2009; Rapoport 2010) have

claimed that auditors played a role in the global financial crisis, willingly turning a blind eye to clients’ questionable accounting tactics (leading to inflated asset values) The PCAOB also has gathered data through its inspection process, with inspection reports often suggesting that

auditors lacked sufficient professional skepticism (PCAOB 2008) Hence, regulators have a renewed interest in taking actions to strengthen auditor independence

On August 16, 2011, the Public Company Accounting Oversight Board (PCAOB) issued

term limits on auditor-client relations The Board is particularly interested in relationships that extend beyond ten years (PCAOB 2011, 20) The European Commission (EC) also is actively considering firm rotation and various proposals have suggested term limits of nine years and six

years (EC 2010; Barker and Hughes 2011; Journal of Accountancy 2011) With long tenure, the

auditor may face significant pressure to preserve client relationships, which may encourage the

auditor to acquiesce to client demands, undermining independence The practical implications of such concerns cannot be underestimated The average auditor tenure of the 500 (100) largest companies in the U.S., based on market capitalization, is 21 (28) years (PCAOB 2011, 20).1

Academic research has examined the association between auditor tenure and various proxies for audit/financial reporting quality Researchers have commonly used proxies based on unexpected accruals (signed and unsigned), with the findings being decidedly mixed, particularly with respect to long tenure Some studies find that quality is not affected by long tenure (e.g., Johnson et al 2002; Gul et al 2007); others find that quality improves with long tenure (e.g., Myers et al 2003; Srinidhi et al 2010); and still others find that quality diminishes with long tenure (e.g., Raghunathan 1994; Davis et al 2009) The mixed findings may be attributable, in part, to the measurement error inherent in estimating total accruals (see e.g., McNichols and Wilson 1988; Healy and Wahlen 1999) A fundamental concern with unexpected accruals is that, across a range of accounts, the potential for a noisy measure is magnified Furthermore, if

unexpected accruals change over time, measurement error may be compounded due to the

reversing nature of accruals (Chu et al 2011) In the current paper, we take a different approach

and focus on one specific accrual, companies’ allowance for bad debts (ABD) As compared to

total accruals, ABD is little affected by accounting estimates in prior periods: it is an offset to accounts receivable (a current asset), and, thus, valuation issues (the accuracy of the accrual) typically are resolved within a one-year time frame Our approach permits a more precise

measurement of the related accrual, though we readily acknowledge a tradeoff (i.e., financial reporting quality encompasses a broad array of accruals)

Healy and Wahlen (1999) suggest that a focus on specific accruals is a fruitful area for future research, as it may allow for more direct and powerful tests We examine ABD because

1

The Board notes that average tenure would be even longer if Andersen had not failed

(1) it directly affects the valuation of accounts receivable, which is a fundamental component of operating working capital; (2) it is subject to discretion and, thus, can be used to manage

earnings (e.g., McNichols and Wilson 1988; Teoh et al 1998); (3) it is associated with revenues (i.e., credit sales), which often are involved in fraudulent financial reporting (e.g., Beasley et al 1999; Beasley et al 2010);2 and (4) professional standards (AS No 12) require auditors to

consider revenues, specifically, in identifying and assessing the risks of material misstatement

(PCAOB 2010) We examine ABD to gauge its association with auditor tenure at any given point in time

We find evidence of a negative association between auditor tenure and estimated ABD The finding holds across a series of robustness tests Further analyses suggest that long tenure is associated with a downward bias in estimated ABD The downward bias appears with auditor tenure of around 15 years With long tenure, the auditor endorses an estimate of ABD that is too aggressive This result is consistent with the argument that long tenure leads to compromised independence Our findings suggest that a term limit of 10 years, which has been suggested elsewhere (e.g., PCAOB 2010; Chasan 2011), is sufficient to preserve auditor independence

The remainder of the paper is organized as follows Next, we provide a framework, including our research hypothesis Subsequently, we describe the research design and then present the empirical findings Lastly, we offer concluding remarks

accounting earnings The auditor learns a distribution, rather than an exact value, because

financial data depend on the application of accounting principles and significant estimates, both

of which require professional judgment The auditor’s best assessment of earnings is the mean

of the earnings distribution

The auditor opines on accounting earnings, which are announced and publicly disclosed

The auditor may incur various costs if accounting earnings are misstated, including increased

legal exposure and client frictions From the perspective of the auditing firm (the national office), the costs of legal exposure may be paramount The overall firm may prefer to report

conservatively in order to minimize the expected litigation cost, especially if the legal regime is strict As such, the auditing firm may prefer to endorse an accounting value that is biased

downward: that is, a value that falls below the mean of the earnings distribution

The engagement partner, on the other hand, may be more focused on the costs of client frictions, which arise when client relations are strained The partner’s performance evaluation, compensation, and professional status may be tied directly to his or her ability to cultivate and maintain client relations Thus, the partner may be incentivized to acquiesce to client demands (e.g., Francis 2004; Knechel et al 2011) The client, on average, may prefer to report

aggressively, with the aim being to create a more favorable financial picture (e.g., to secure better credit terms or to bolster stock prices) The engagement partner, in turn, may feel

compelled to support the client’s position Moore et al (2006) suggest that an unconscious, serving bias may make the partner susceptible to the client’s arguments and demands,

self-undermining objectivity Likewise, the PCAOB has expressed concern that audit partners “may have a bias toward accepting management's perspective, rather than developing an independent view or challenging management's conclusions” (PCAOB 2011, 7) Accordingly, the

engagement partner may be predisposed to endorse an accounting value that is biased upward: that is, a value that exceeds the mean of the earnings distribution

The engagement partner’s (audit team’s) bias may be magnified as auditor tenure

increases In the early years of an engagement, bias may subtly creep into auditors’ judgment, providing a ready means to foster client relations Bazerman et al (2002, 100) assert that the auditor may unknowingly adapt to “small imperfections in a client’s financial practices.” Over time, the auditor’s objectivity may deteriorate, ever so gradually, to the point that larger

imperfections are accepted, which may have significant implications for financial reporting quality (e.g., Bazerman et al 1997; Bazerman et al 2002; Moore et al 2006) In turn, the

potential negative consequences of rejecting the client’s position may increase as auditor tenure increases, with any backlash realized immediately In cases of long tenure, the auditor may come to interpret evidence from a partisan perspective (Moore et al 2006) Ongoing relations may cause the auditor to more closely identify with the client, making it more likely that the auditor will endorse the client’s position (e.g., Bamber and Iyer 2007) Simply put, long tenure may undercut objectivity

For our purposes, long tenure increases the possibility that the auditor approves of more aggressive accounting estimates, effectively increasing earnings We are interested in the

auditor’s assessment of ABD, as it gives ample room for discretion Examples of companies’ that have understated ABD, under the auditor’s watchful eye, are plentiful including Friedman’s Jewelers (http://www.sec.gov/litigation/complaints/comp19477.pdf), Satyam (Kahn 2009), Advocat (Mulford and Comiskey 2002, 241), Miniscribe (Albrecht et al 2009, 144), and

NextCard (Knapp 2010, 78), just to name a few With long tenure, the auditor may come to embrace the client’s perspective, wanting to believe that the assessment is correct Hence, we

posit that as tenure increases, the auditor more readily accepts the client’s estimate of ABD,

which on average is too aggressive (too low) Formally stated, our research hypothesis is as follows

Hypothesis: Ceteris paribus, the allowance for bad debts (ABD) is a decreasing function

of auditor tenure

3 Research Design

We examine clients’ ABD over time to test for an association with auditor tenure As discussed above, the audit firm (national office) prefers to report conservatively, whereas the engagement partner (local office) prefers to report aggressively Because the engagement

partner makes the reporting decision, the partner’s preferences may prevail We conduct

empirical tests to examine this issue

We consider the following regression model

∆SALE change of sales from year t-1 to year t scaled by the beginning total assets

((data12t – data12t-1)/data6t-1);

CFO cash flow from operations scaled by the beginning total assets

(data308t/data6t-1);

SIZE log transformation of the year-end market value of equity (ln(data25t

*data199t));

ROA net income scaled by the beginning total assets (data172 t/data6t-1);

LOSS an indicator variable that equals to 1 if data172 t < 0 and 0 otherwise;

LEV total liabilities to total assets ratio (data181t/data6 t);

CR ratio of current assets to current liabilities (data4 t/data5 t);

FIRMGROW growth in sales of a firm (data12t/data12t-1);

BIG an indicator variable that equals 1 when the auditor is a Big 4 firm and 0

otherwise;

portion of total assets in an industry and 0 otherwise; and

Because accrual recognition varies across industries and time, we include two-digit SIC codes to capture industry fixed effects and indicators for years to capture time fixed effects

The control variables are similar to those in Myers et al (2003) and Gul et al (2009)

Firm age (AGE) is included to mitigate the possible confounding effects of firm maturity and to

control for differences associated with different firms’ life cycles (Anthony and Ramesh 1992) Other variables associated with accruals also are included We control for the change in

revenues (∆SALES), as it may directly impact ABD; cash flow (CFO) because prior studies show

an association between cash flow and accruals (e.g., Dechow 1994); and firm market value (SIZE,

SIZE 2 , and SIZE 3) because larger firms have more stable accruals (Dechow and Dichev 2002)

and the size effect is non-linear (Gul et al 2009) We also control for growth (INDGROW,

behavior (Gul et al 2009); leverage (LEV, CR) as there are documented associations between leverage and accruals (e.g., Butler et al 2004); performance (LOSS, ROA, and ROA_SD) because

it is associated with accruals; and finally auditor size and specialization (BIG, SPECIAL, and

TENURE_SPECIAL) because auditors’ specialization is associated with accruals (e.g., Gul et al 2009)

4 Data and Empirical Findings

4.1 Sample Composition and Descriptive Statistics

The initial sample consists of all firms for the years from 1988 to 2006 in the Compustat database We restrict our analyses to this time period because reported operating cash flows as

per SFAS No 95 (FASB 1987) are only available since 1988 We exclude firms in the financial

services industries (SIC codes between 6000-6999) The firm’s age and the auditor tenure

variables are calculated from all available data in Compustat

We delete firms with negative total assets, sales, debt, and market values of equity

because such observations introduce noise into the analyses To mitigate problems caused by extreme observations in cash flows and ABD, we exclude observations in the top and bottom 0.5 percentile To avoid the confounding effects of start-up firms, we drop observations of all firms

in their first five years of existence Firms that switch auditors in the first five years also are left out because such firms may differ systematically from other firms (Myers et al 2003)

Following Myers et al (2003) and Gul et al (2009), we eliminate firms that undergo merger and acquisitions, as Hribar and Collins (2002) show that estimating accruals for such firms is

problematic After imposing all the necessary requirements, we obtain a final sample of 15,226 firm-year observations

Figure 1 illustrates the relationship between estimated bad debts scaled by total accounts

receivable (ABD) and auditor tenure Panel A is a scatter illustration of the relationship Two features can be easily identified: the amount of ABD, as well as the range, is reduced as auditor

tenure increases Panel B illustrates the relationship between auditor tenure and various statistics

of ABD The mean of ABD is clearly a decreasing function of auditor tenure The 90th percentile

of ABD has a very strong negative relationship with tenure, while the 10th percentile does not show a systematic relationship

Insert Figure 1 about here

Table 1 reports the descriptive statistics for estimated bad debts, auditor tenure, and the

control variables for the sample period The average ABD is 6.1 percent, the average auditor

tenure is 10.34 years, and the average firm age is slightly over 27 years

Insert Table 1 about here

Table 2 presents pairwise correlations among the independent variables Notably, auditor tenure and firm age are highly correlated (with a correlation efficient of 0.616, p < 001) We further discuss this issue in our sensitivity analyses

Insert Table 2 about here

4.2 The Effect of Long Tenure on Allowance for Bad Debts

Table 3 presents the results of the main regression model The coefficient of TENURE is

negative and statistically significant at p < 0.01 The finding suggests that, on average, auditors who have longer relationships with their clients endorse financial statements with lower

allowance for bad debts This result is consistent with our research hypothesis

Insert Table 3 about here

4.3 Orthogonal Regression

As noted earlier (refer to Table 2), auditor tenure and firm age are highly correlated and the effect of each on ABD is similar The high correlation is inherent because longer tenure is conditioned on client existence (i.e., for auditor tenure to be long, the client must have a long life) Notwithstanding, the high correlation creates a multicollinearity problem and, in turn, it is not clear whether our findings are truly generated by auditor tenure or firm age To mitigate this concern, we rerun the regressions using an orthogonalized value of auditor tenure: that is,

orthogonal to firm age or to all control variables As shown in Table 4, inferences are unaffected

The coefficient of O_TENURE (orthogonal tenure) is negative and statistically significant at p <

0.01, which is consistent with our research hypothesis We point out that using the orthogonal

value of auditor tenure may understate its effect on ABD because AGE and TENURE are

positively associated

Insert Table 4 about here

4.4 The Effect of Long Tenure on Auditor Independence

We are interested in determining whether long tenure leads to impaired independence The negative association between auditor tenure and ABD, documented above, is consistent with

this explanation The auditor may face pressure to preserve client relations, particularly as tenure increases, which may prod the auditor to endorse an aggressive estimate of ABD (one that is biased downward) Over time the social bond with the client may strengthen, making it more likely that the auditor endorses an estimate of ABD that is too low Hence, the reported ABD, on average, decreases as auditor tenure increases

An alternative explanation, however, also applies In the initial years of an engagement, the auditor may endorse a conservative estimate of ABD (one that is biased upward), with the aim being to minimize expected litigation cost In this case, the engagement partner adopts the overall firm’s preference, with a focus on type II errors As tenure increases, the auditor

becomes more familiar with the client’s business, operations, and accounting policies and, in turn,

is better able to manage engagement risk (e.g., Geiger and Raghunandan 2002; Carcello and Nagy 2004) The improved knowledge allows the auditor to better gauge ABD Accordingly, over time the auditor may endorse an estimate of ABD that is less conservative, such that the bias in the estimate is reduced, but still positive (i.e., still conservative)

For our purposes, the competing explanations arise because ABD is an estimate and its

true (correct) value is not known a priori As such, it is difficult to determine whether the

auditor endorses an estimate of ABD that is aggressive or conservative To gain further insight into the negative association between auditor tenure and ABD, we identify a scenario in which the client is predisposed to report aggressively

Prior studies provide overwhelming evidence that companies making new equity

offerings have incentives to present a more favorable financial picture (Rangan1998; Teoh et al 1998; Shivakumar 2000) The objective is to increase the proceeds of the offering One means

to improve the financial picture is to make more aggressive estimates, including reducing ABD

Consequently, we examine the impact of a new equity offering on the reported ABD in the year immediately preceding the offering

Ceteris paribus, a new equity offering in the subsequent year should not affect the

unbiased estimate of ABD in the current year But the scenario introduces tension because the auditor faces conflicting pressures On the one hand, the client prefers a smaller ABD, and the auditor may endorse a lower value to preserve client relations On the other hand, the stock offering increases the auditor’s exposure to potential liability, and the auditor only may endorse a higher value to minimize expected litigation cost We can determine whether the auditors with a particular tenure compromise independence by measuring the impact of the new equity offering

on ABD for firms by auditors with the given tenure

To measure the effect of new equity financing on ABD with given auditor tenure, we

need to add a variable FINANCE and an interactive variable TENURE_FINANCE to our

regression model The variable FINANCE is the amount of new equity financing in year t+1

scaled by total assets at the end of year t (data108t+1 /data6t) New financing increases the

auditor’s engagement risk, which suggests that the auditor may be conservative in assessing

ABD (and other accounting estimates) If so, the coefficient of FINANCE will be positive

The interactive variable TENURE_FINANCE is simply the product of auditor tenure and the amount of the new offering The sign of the coefficient may be negative because TENURE and FINANCE have opposite predicted effects on ABD The rationale is straightforward The

new equity offering increases the auditor’s engagement risk, introducing pressure to increase ABD But longer tenure has a counter effect, either creating pressure to compromise

independence or allowing the auditor to make a more precise estimate of ABD (and better

manage engagement risk) In either case, the effect is to decrease ABD As elaborated below,

we examine the sum of the coefficient of FINANCE and the product of TENURE and the

coefficient of TENURE_FINANCE to disentangle the competing explanations

The regression results are shown in Table 5 As before, the coefficient of TENURE is negative and statistically significant at p < 0.01 The coefficient of FINANCE is positive and statistically significant at p < 0.01 Further, the coefficient of TENURE_FINANCE is negative

and statistically significant at p < 0.01

Insert Table 5 about here

Because our objective is to determine whether auditors’ concern over preserving client relations leads to compromised independence, we need to isolate the total marginal impact of

new financing For a given value of TENURE, the total marginal impact is the sum of the

coefficient of FINANCE and the product of TENURE and the coefficient of TENURE_FINANCE

As we mentioned above, new financing in year t+1should not impact the unbiased estimate of ABD in year t If the total marginal effect is negative, it is indicative of impaired independence

A total negative effect suggests that the auditor complies with client pressure to endorse an

estimate of ABD that biased downward (aggressive) On the other hand, if the total marginal effect is positive, it suggests that the auditor maintain independence In this case, the auditor’s concern with potential liability prevails, and the auditor endorses an estimate of ABD that is biased upward (conservative)

The results of the total marginal impact of new equity financing with various years of auditor tenure are shown in Table 6 For each given value of auditor tenure (1-27), we compute the total marginal impact and test whether it is significantly different from zero For auditor tenure of one to eight years, the sum is positive and statistically significant at p < 0.05 As tenure

(p < 0.10) in the 18th consecutive engagement Moreover, the statistical significance improves over time (to 5 percent in the 19th year and 1 percent in the 26th year)

Insert Table 6 about here

The finding indicates that the impact of new financing on estimated ABD is positive in the earlier years of the auditor-client relationship, but that it becomes negative when the tenure is long In other words, auditors with short tenure endorse an estimate of ABD that is conservative When tenure becomes long, however, auditors endorse an estimate of ABD that is aggressive, which suggests that they compromise independence In our sample, about 25 percent of auditors have tenure longer than 16 years Therefore, the majority of auditors in our sample maintain independence

4.5 Additional Analyses

We consider the possibility that our results in Table 6 may be biased because we have assumed a linear relation between auditor tenure and the total marginal impact of new equity financing on estimated ABD Davis et al (2009) suggest that tenure may have a nonlinear effect

on the auditor’s willingness to approve aggressive accounting estimates If so, the estimated

coefficient for TENURE_FINANCE may not be the same for auditors with short and long tenure

To test this possibility, we divide our sample into two subsamples based on the length of auditor tenure: sample 1 includes companies with auditor tenure from 1-14 consecutive engagements and sample 2 to includes companies with auditor tenure with at least 15 consecutive engagements

We estimate the coefficients of FINANCE and TENURE_FINANCE for each sample (as

in Table 5) Then, using the sample-specific coefficients, we calculate the total marginal impact

of new financing on estimated ABD for a given auditor tenure The results, presented in Table 7, indicate that for sample 1 (auditor tenure of 1-14 years), the marginal effect is positive and