Do Auditors Use the Information Reflected in Book-Tax Differences ...

37 downloads 3326 Views 597KB Size Report
book-tax differences being associated with more modified audit opinions and a .... book-tax differences, the sample, variable measurement, and empirical tests.
Do Auditors Use the Information Reflected in Book-Tax Differences? Michelle Hanlon Assistant Professor Ross School of Business at the University of Michigan Ann Arbor, MI 48109 Phone: 734-647-4954 E-mail: [email protected] Gopal V. Krishnan Associate Professor of Accounting VSCPA Northern Chapter Professorship in Public Accounting School of Management, MSN 5F4 George Mason University Fairfax, VA 22030 Phone: 703-993-1966 E-mail: [email protected] First Version: September 20, 2004 Current Version: January 2, 2006 Comments Welcome ____________________________________________________________________ Abstract : Prior literature provides evidence consistent with book-tax differences providing information about financial accounting earnings quality. We investigate whether auditors use this information when auditing firms’ financial statements. The data are consistent with larger (absolute value) book-tax differences being associated with higher audit fees. We interpret this evidence as indicating that book-tax differences reflect information that represents a higher risk of earnings management which increases auditors’ efforts and time spent on the audit. We conduct supplemental tests on auditor opinions and auditor turnover. The data are consistent with larger book-tax differences being associated with more modified audit opinions and a greater incidence of auditor turnover. Similar to the audit fee results, we interpret this evidence as being consistent with the book-tax differences reflecting information about earnings management and with auditors examining and using this information in opining on the firms’ financial statements. This paper contributes to the literature that investigates the information reflected in book-tax differences and documents the relevance of book-tax differences on the audit process. _____________________________________________________________________

Hanlon appreciates funding from an Ernst & Young Faculty Fellowship. We thank Cristi Gleason, Sanjay Gupta, Gene Imhoff, Lil Mills, Tom Omer, and Terry Shevlin, and workshop participants at Arizona State University for helpful comments on earlier versions of this paper. We also thank the senior managers and partners from the audit firms who were willing to discuss our paper and questions with us.

Do Auditors Use the Information Reflected in Book-Tax Differences?

1.

Introduction This paper examines whether auditors utilize the information in book-tax differences.

Specifically, we investigate whether book-tax differences are associated with higher audit fees, more modified audit opinions, and greater auditor turnover. We measure book-tax differences as the absolute value of both total book-tax differences (temporary and permanent) and using a measure of temporary differences alone.1 The data are consistent with larger book-tax differences being associated with higher audit fees. We interpret this evidence as indicating that book-tax differences reflect information that represents a higher risk of earnings management which increases auditors’ efforts and time spent on the audit. Our results are also consistent with larger book-tax differences being associated with more modified audit opinions and a greater incidence of auditor turnover. Similar to the audit fee results, we interpret this evidence as being consistent with the book-tax differences reflecting information about earnings management and with auditors examining and using this information in opining on firms’ financial statements. The extant literature provides evidence consistent with large book-tax differences indicating lower ‘earnings quality.’ The underlying maintained hypothesis is that when a firm has large book-tax differences (i.e., where book and taxable incomes are very different) the company is manipulating one or both of the income measures opportunistically. While some research focuses on the information in book-tax differences for tax aggressiveness (e.g., Mills 1998, Manzon and Plesko 2002, Desai and Dharmapala 2005), others examine the information in book-tax differences about financial accounting earnings management. For example, Phillips et al. (2003) report that firms that report small positive earnings have a larger deferred tax expense 1

We define these concepts below.

2

consistent with these firms managing financial reporting income upward to meet the target but not reporting the additional income for tax purposes. Mills and Newberry (2001) report evidence consistent with the magnitude of book-tax differences being positively associated with financial reporting incentives such as prior earnings patterns, financial distress, and bonus thresholds while Joos, Pratt and Young (2000) conclude that firms’ earnings response coefficients are smaller in the presence of large book-tax differences. Further, Hanlon (2005) reports that firms with large book-tax differences have less persistent one-year-ahead earnings than firms with small book-tax differences and that investors interpret book income far in excess of taxable income as a ‘red flag’ about earnings quality. Lev and Nissim (2004) report that the ratio of taxable income to book income predicts subsequent five-year earnings changes. Weber (2005) extends Hanlon (2005) and Lev and Nissim (2004) to investigate sell-side analysts’ use of booktax difference information and finds that analysts generally misprice the information in total book-tax differences but seem to have lower forecast errors with respect to deferred tax expense. We extend this literature by asking whether auditors utilize the information in book-tax differences. We examine this question by testing for an association between book-tax differences and auditor fees, opinions, and turnover. We report evidence consistent with larger book-tax differences being associated with higher audit fees, more modified opinions, and a greater incidence of auditor turnover. We do not claim that auditors necessarily directly use the measure of book-tax differences as a distinct indicator of earnings quality. What we intend is to document that the book-tax differences, at a minimum, reflect the same information that the auditors use to assess earnings quality. Indeed,

3

in conversations with audit partners and senior managers from the Big Four accounting firms this indirect relation appears to be representative of audit procedures currently used in practice.2 A caveat of the paper is that the positive association between audit fees and book-tax differences could mean that larger book-tax differences indicate that the company is more complex, for example, more multinational, which requires more audits and thus more fees. We control for as many factors as we can in our regression analysis to eliminate the association being driven by complexity of the firm. In addition, we estimate our regression after excluding all observations with any amount of foreign source income and find similar results. However, to the extent some factors indicating complexity are still uncontrolled for our results could be affected. In addition, we conduct a wide array of sensitivity analyses including various controls, different methods of measuring control variables, using a changes specification for the audit fee test, and testing the relation of audit fees and lagged book-tax differences. We discuss all of these throughout the text, in footnotes, or in section 5 below. Overall, our results are consistent with our reported results – the absolute value of book-tax differences are positively associated with audit fees. The paper proceeds as follows. Section 2 develops our hypothesis. Section 3 describes book-tax differences, the sample, variable measurement, and empirical tests. Section 4 discusses the results, Section 5 describes our sensitivity analyses and Section 6 concludes.

2

One audit partner stated that if a firm is very tax aggressive this could indicate management’s mindset of compliance with rules and regulations – that is if they are willing to cheat on their taxes they might just be cheaters all around. All the practitioners we talked to said that book-tax differences are examined in relation to earnings quality but generally in an indirect sense. In further questioning they revealed that when the tax provision contains large items, they investigate those items in greater detail. For example, if a firm has a large write-off which is not tax deductible that write-off will be scrutinized in more detail. In addition, if deferred tax balance sheet balances are out of line with industry averages upon assessing their analytics, auditors will investigate further the potential reasons why this might be the case.

4

2.

Hypothesis Development Our tests investigate both whether book-tax differences are a proxy for low earnings

quality and whether auditors use proxies for earnings quality in conducting audits. Auditing standards suggest that auditors should respond to engagement risks by altering the nature, timing, and extent of audit procedures (e.g., SAS No. 47, AICPA 1983; SAS No. 82, AICPA 1997).4 Prior research is generally consistent with auditors being responsive to risks, although it is somewhat mixed. For example, using a small sample and private data, Bedard and Johnstone (2004) report that auditors plan increased effort and billing rates for clients with earnings manipulation risk (measured using engagement partner assessments of existing clients). Johnstone and Bedard (2001) using data from one audit firm over two years provide evidence consistent with risk having little effect on planned personnel hours but that the audit firm responds to fraud and error risk factors by applying engagement-planning strategies such as assigning more high-risk specialist personnel, assigning more industry expert personnel, applying more intensive testing, and/or performing additional review. In addition, Bell, Landsman, and Shackelford (2001) find that high business risk increases the number of audit hours, implying that audit firms can assess firm-level differences in business risk and then obtain compensation through billing additional hours (not by raising hourly fees, however).5

4

Recent standards such as SAS 99 Consideration of Fraud in a Financial Statement Audit and The Sarbanes-Oxley Act (SOX) impose additional duties on auditors with respect to audit procedures and audit documentation regarding earnings quality and management fraud. These standards would apply to one to two years of our sample period. 5 For other examples see also Davis et al. 1993; Mock and Turner 2001; Messier and Plumlee 1987; Maletta and Kida 1993). In addition, for studies that find that increased risks do not alter auditors planning or audit pricing see Bedard 1989; Mock and Wright, 1999; O’Keefe et al. 1994.

5

The prior literature on book-tax differences provides evidence consistent with the differences providing information about earnings quality. If auditors use the information reflected in book-tax differences in assessing earnings quality then large book-tax differences should indicate the necessity for auditors to exert more effort (more hours or including more specialists) which would increase the level of audit fees for these firms relative to firms with smaller book-tax differences, all else constant. We hypothesize that large book-tax differences in either direction can indicate earnings quality problems (i.e., earnings management--either upward or downward) based on reasoning and results in Joos et al. (2000) and Hanlon (2005) where large book-tax differences in both directions are associated with lower earnings response coefficients and less persistent earnings. We state our hypothesis as follows: H1: There is a positive association between audit fees and the absolute value of booktax differences.

To provide further evidence on the question of whether auditors consider the information reflected in book-tax differences in auditing their clients, we also conduct supplemental tests that examine whether larger book-tax differences are associated with more modified audit opinions and greater auditor turnover. If auditors utilize the information reflected in book-tax differences to assess earnings quality and then communicate this to the market, we would expect to see a positive relation between the absolute value of book-tax differences and modified audit opinions. Further, if there are also more modified opinions and/or a higher incidence of auditor turnover for firms with large book-tax differences, the relation with fees is likely not simply the result of the firm being more complex or because more audit time is required on the tax expense line (i.e.,

6

the tax provision). Indeed, the auditors that we consulted stated that almost never was an opinion modified because of the tax expense itself.6 Similar to Bradshaw et al. (2001), a limitation of the audit opinion test is that auditors are not required to issue a modified audit opinion just because there is a higher probability of a violation of Generally Accepted Accounting Principles (GAAP) or an expected future decline in earnings.7 As Bradshaw et al. (2001) discuss, a qualified opinion is only required when there is a scope limitation or a departure from GAAP and generally the SEC will not accept qualified opinions.8 An adverse opinion and a disclaimer of opinion are also only very rare. However, auditors can modify their opinion by adding explanatory language in cases where they have going concern doubts, where there are material changes in accounting principles and their application, and in cases with other matters that the auditor may want to emphasize. Thus, if the information reflected in large book-tax differences causes the auditor to have concerns about earnings problems they can convey these concerns through a modified opinion. We also investigate whether audit turnover is associated with book-tax differences. If a client is perceived as being too risky for the auditor, the auditor can end the auditor-client relationship. For example, in 1997 the Wall Street Journal reported that the Big Six accounting firms were increasingly rejecting clients due to risks concerns and the article quoted Dan Guy, vice president of the American Institute of Certified Public Accountants as saying “More than ever, accounting firms don’t want to expose themselves to clients who will harm their reputations or generate costly litigation.” They also cited one case where Arthur Andersen 6

The auditors qualified this as being before Section 404 of the Sarbanes Oxley Act was effective. They all claimed that a preponderance of the material weaknesses in internal controls that were (are being) reported relate to deferred tax assets and liabilities not being reported accurately. Section 404 became effective after our sample period. 7 Although this is somewhat mitigated by the implementation of SAS 99 and SOX during the latter part of our sample period. The requirements for uncovering fraud are no different but the procedures and documentation requirements are much more rigorous. 8 See Article 4(a) or Regulation S-X, Securities Act of 1933. This was emphasized by the auditors we consulted with as well.

7

(ironically) dropped an account because they were uncomfortable with “management’s business philosophy” (MacDonald, 1997). Thus, if book-tax differences indicate higher audit risks and the auditors utilize this information then larger book-tax differences will be associated with a greater probability of auditor turnover.

3.

Book-Tax Differences, Sample, Variables and Empirical Tests

3.1

Book-Tax Differences Our main independent variable of interest is the measure of book-tax differences, or the

difference between what the firm reports as taxable income and what the firm reports as financial accounting income. There are in general two types of book-tax differences—temporary differences and permanent differences. Temporary differences are differences between book and taxable incomes in one period that will reverse out in a future period. Thus, the two income measures are different only because financial accounting reports the income or expense item in one period and the rules for the computation of taxable income require that same item of income or expense to be taken into account in a different period. An example of a temporary difference is the reserve for warranty expense. Under GAAP firms are required to estimate the warranty expense expected to be associated with current sales and record the expense in the current period. In contrast, for tax purposes estimates of expenses are not allowed and the firm must wait until the warranty costs are actually paid before a deduction can be taken. Under both income calculations the same amount of warranty expense should be recorded (assuming expectations are accurate) but the expensing will be earlier for financial accounting than for tax purposes. In contrast, permanent differences are income or expense items which are different between book

8

and tax and will never reverse. An example of a permanent difference is municipal bond interest which is included in financial accounting income but excluded from taxable income. Because taxable income is not publicly available, we use financial statement data to calculate the book-tax differences. We use the firm’s reported deferred tax expense as our estimate of the firm’s temporary book-tax differences (similar to Hanlon 2005, Phillips et al. 2003, and others). We estimate the firm’s total book-tax differences by first estimating taxable income following the methodology in Hanlon, Kelley and Shevlin (2005). We gross up the firm’s current tax expense by the statutory tax rate and then subtract from the result the change in the firm’s net operating loss. We then subtract the estimated taxable income from the firm’s reported income before taxes (data item #170). For both measures of book-tax differences, we use the absolute value of the book-tax differences because both large positive (book income in execess of taxable income) and large negative (book income less than taxable income) book-tax differences provide indications about lower earnings quality (Hanlon 2005 and Joos et al. 2000).9 We acknowledge that there are well known problems with using financial statement information to infer taxable income or total (especially permanent) book-tax differences (e.g., see Hanlon 2003). However, in our setting the use of financial statement data is appropriate because we are trying to interpret the auditor’s use of the information in the differences between book and taxable incomes, thus, we have to use the information available to the auditors at the time of the audit. For most publicly traded firms the tax return is not completed at the time of the audit. Tax returns are due two and a half months after the firm’s year end but can be and generally are extended for an additional six months. In contrast, the financial statements are required to be filed within 90 days of the firm’s fiscal year end. As a result, in most cases there

9

However, we report our results separately for positive and negative book-tax differences as well.

9

is no tax return that the auditors could look at when auditing the firm. What the auditors do have however is the tax provision which is included in the firm’s financial statements. In addition, we need to estimate the book-tax differences on world-wide income for the same entities included in the calculation of book income. The consolidation rules are different for financial accounting relative to tax purposes. For example, consolidation for financial accounting is required when ownership is 50% or greater but for tax return purposes consolidation is an election only when an entity is owned 80% or more by another entity (for more details see Hanlon 2003).10 While this issue creates no measurement error when using financial data to infer information about taxable income for the entities included in the financial statement consolidation, it does pose serious problems when trying to link tax return data to financial statement data (Hanlon 2003, Plesko 2002, and Mills and Plesko 2003). Thus, for these reasons we estimate taxable income and book-tax differences from financial statement data. Although we believe the use of financial statements to obtain our measure of book-tax differences is appropriate, we are cognizant of the measurement problems that are involved. One problem is that we actually capture more than book-tax differences in our calculation of total book-tax differences. For example, tax credits are essentially converted into a book-tax difference in this calculation which could affect our results because firms with foreign tax credits or more research and development credits could just be more complex firms requiring more time on the audit and thus more audit fees. In sensitivity analysis below, we estimate our fee regression by eliminating firms with any foreign source income and any research and development expense to investigate whether our results hold where book-tax differences can be

10

The ruled for consolidation for financial accounting purposes are technically based on the level of control of the entity. The percentage ownership is generally used as the indicator of such control.

10

estimated more accurately and the firms are less complex. We find very similar results which we discuss below. Our measure of total book-tax differences also captures amounts recorded for financial accounting that affect the tax expense line but are not really differences between pre-tax book and taxable incomes.11 For example, under Statement of Financial Accounting Standard 109 Accounting for Income Tax firms are required to record a reserve (valuation allowance) against a deferred tax asset if they are more likely than not to not realize the benefits of the deferred tax asset in the future. Another example is the tax contingency reserve (i.e., the tax cushion) recorded when firms have taken an aggressive tax position and need to record a reserve for the potential future costs associated with the position being overturned. Management can manage earnings through these accounts (i.e., the tax cushion, the valuation allowance, or in estimating the effective tax rate (Gleason and Mills 2002; Miller and Skinner 1998; Schrand and Wong 2004; and Dhaliwal, Gleason, and Mills 2004)). Thus, the use of total book-tax differences as measured from financial statements will incorporate earnings management using these methods as well as earnings management in pre-tax earnings and tax aggressiveness. Using our measure of total book-tax differences allows us to examine auditors’ use of all this information collectively. We separately test a measure of temporary book-tax differences, which is not subject to most of these concerns, to examine whether our audit fee results are solely attributable to the earnings management through the tax expense line or at least

11

These are, however, book-tax differences because they are items reported for financial accounting purposes but not taxable income purposes. They are accruals and reserves that affect the tax expense line itself rather than pre-tax earnings.

11

to some extent attributable to the information in temporary book-tax differences between pre-tax book income and taxable income. 12

3.2

Sample We start with all observations available on the 2004 Compustat database with nonmissing

asset data for the years 2000-2003 (N=35,516). We lose 26,302 observations missing any of the required variables for our tests. This leaves us with a final sample of 9,170 firm-years. (GOPAL WE SHOULD MAKE A TABLE 1 PANEL A FOR THE SAMPLE SELECTION SINCE WE LOSE SO MANY OBSERVATIONS I THINK PEOPLE WILL WANT TO SEE WHERE WE LOSE THE DATA. CAN YOU ADD TO T1 AND THEN I CAN WRITE UP?)

3.3

Empirical Model for H1and Variable Definitions

12

Note that under our calculations neither book income nor taxable income will include an expense (deduction) for stock option compensation. Thus, although there are tax benefits to stock option compensation and these create a book-tax difference we do not include this difference in our tests. We believe this is appropriate in this case because we are not interested in testing audit fees in relation to stock option compensation and the accounting that is associated with stock option compensation. Rather, we are interested in a firm’s level of reporting differently for tax and book for reasons other than stock option compensation. 17 We conduct tests using both the signed value of accruals, the absolute value, these values logged and these values scaled. We indicate the measure for the main tests in the respective tables and report the remaining results as sensitivity analysis in the text or footnotes.

12

In order to test H1, whether book-tax differences are associated with audit fees, we estimate the following model based on Simunic (1980) and Larcker and Richardson (2004), but include our measure of book-tax differences: Ln(AUDIT FEE)t = α + β1 Ln(ABS BTD)t + β2 ACCt + β3BIG5(4)t + β 4Ln(ASSET)t + β5BUS SEGt + β 6FGNt + β7INVt + β8RECt + β 9DEBTt + β10INCOMEt + β11LOSSt + β 12 AUD OPINt + β13-17 INDt +e (1)

We define the variables as follows: Ln(AUDIT FEE)t

=

Log of audit fee (obtained directly from Standard and Poor’s as disclosed by the firms in proxy statements);

Ln(ABS BTD)t

=

Log of the absolute value of the spread between pre-tax book income and taxable income (data #170 – ((data16-data50)/.35 – ∆data#52). Or, alternatively, the firm’s deferred tax expense for the year (data #50);

ACCt

=

The firm’s total accruals measured as the difference between earnings and cash flow from operations (data #18 – (data #308 + data #124), scaled by average total assets of the firm (data #6)17

BIG5(4)t

=

An indicator variable set equal to 1 for client-observations audited by a Big 5 auditor (or Big 4 auditor after the demise of Arthur Andersen), and 0 otherwise (item #149);

Ln (ASSET)t

=

Log of total assets (item #6);

BUS SEGt

=

Square root of the number of business segments owned by the client18;

FGNt

=

Ratio of foreign pre-tax income (item #273) to total pre-tax income (item #170);

INVt

=

Inventory to average total assets (item #3);

RECt

=

Receivables to average total assets (item #2);

DEBTt

=

Sum of short-term debt and long-term debt to average total assets (item #34 + item #9);

18

We obtain the number of business segments using the variable SEGNUM from Research Insight.

13

INCOMEt

=

The ratio of operating income after depreciation to average total assets (item #178);

LOSSt

=

An indicator variable set equal to 1 if income before extraordinary items and discontinued operations is negative in the current or two previous years, and 0 otherwise (item #18);

AUD OPINt

=

An indicator variable set equal to 1 if the firm receives a modified audit opinion, and 0 otherwise (item #149), a modified opinion is defined as anything other than a standard unqualified audit opinion coded as 1 by Compustat;

INDt

=

Industry indicator variables based on the two digit SIC codes.

Simunic (1980) chose these variables based on discussions with auditors from the then Big Eight firms and with representatives from firms that underwrote professional liability insurance coverage for accountants. These discussions produced five determinants of auditors’ loss exposure: 1) the size of the auditee, 2) the complexity of the auditee’s operations, 3) auditing problems associated with certain financial statement components, especially inventories and receivables, 4) the industry of the auditee, and 5) whether the auditee is a publicly or closely-held company. Thus we include Ln(ASSETS) to control for size. We include the number of business segments (BUS SEG) as a proxy for the complexity of the firm. We also include the percentage of total pre-tax book income that is from foreign sources as a proxy of the complexity of the firm.19 We include a ratio of the firm’s inventory to assets (INV) and receivables to assets (REC) to control for the loss exposure from the auditing difficulties in these areas, and we include industry indicator variables to control for the industry in which the firm operates. All of our companies are public companies and thus the final factor is not an issue in our study (nor was it in Simunic’s). Simunic includes proxies for the firm’s level of financial distress. Following

19

Simunic (1980) used survey data and included the percentage of foreign assets held by the firm. Compustat does not provide data on foreign versus domestic assets, thus we include the relative percentage of foreign sourced income as our proxy.

14

Simunic (1980) we include three such proxies: 1) the ratio of operating income to assets (INCOME) as a proxy for profitability, 2) an indicator variable set equal to one if income before extraordinary items and discontinued items is negative in the current or two prior fiscal years (LOSS), and 3) an indicator variable set equal to one if the firm received anything other than a clean, unqualified opinion in the current year (AUD OPIN). We also include a measure of the firm’s debt levels as an additional proxy for financial distress (DEBT). Finally, we include an indicator variable if the firm is audited by a Big Five (or Four after the demise of Arthur Andersen) accounting firm in the current year (BIG 5(4)) as an additional proxy for size and complexity of the firm. As a final control variable we include a measure of the firm’s total accruals. Some researchers have investigated a link between the firm’s accruals and the non-audit fees paid by a firm to test whether firms that pay high non-audit fees are allowed to record higher accruals (a proxy for lower earnings quality). The underlying hypothesis of these studies involves an economic bonding story in which the auditors are bound to firms that pay high fees and thus the auditor cannot conduct independent audits which results in low earnings quality. However, many of these studies fail to provide evidence in support of the economic bonding theory (see Larcker and Richardson (2004) for a discussion). Instead of an economic bonding story, we include the firm’s total accruals as a control variable that proxies for firm complexity and the increased risk associated with a firm reporting high accrual levels. Thus, our coefficient on the book-tax differences of the firm is the incremental effect of the book-tax differences beyond the effect of the risk associated with the firm’s accrual levels.20

20

There is some debate in the literature regarding whether the accruals measure should be signed or unsigned (see Larcker and Richardson (2004)). Because the accruals variable is a control variable in our study we include the variable measured in each of the various ways to ensure our results on the book-tax difference variable are robust to the various measures of accruals.

15

3.3

Supplemental Tests: Audit Opinions and Auditor Turnover We next investigate whether there is an association between audit opinions that are

modified in any way and book-tax differences. We estimate the following logistic regression based on Bradshaw et al. (2001) but with the addition of our test variable:

AUD OPINt = α + β 1(ABS BTD)dect + β 2 WCACCdec t+ β3 ln(ASSETS) + β 4CFOt + β5DEBTt + β6TIEt + ε

(2)

Where (ABS BTD)dec is the decile ranking (from 0 to 1) of the absolute value of the booktax difference measures (defined above), scaled by average total assets. Similarly, WCACCdec is the decile ranking (from 0 to 1) of the signed working capital accruals of the firm calculated as the changes in working capital accounts from the statement of cash flows (#302, #303, #304, #305 and #307), scaled by average total assets.21 The variable CFO is the firm’s cash flows from operations (data item # 308) scaled by average total assets and is included as a control for performance unrelated to accrual accounting. TIE is the firm’s operating earnings divided by interest expense (item # 178/#15) to control for the firm’s short term solvency and liquidity. The other variables are as defined above. We follow Bradshaw et al. (2001) (and Johnson and Lys (1990) and DeFond and Subramanyam (1998)) in estimating the following logistic regression to investigate whether firms with large book-tax differences have a higher likelihood of auditor turnover:

∆AUDITORt = α + β1 (ABS BTD)dec t + β2 WCACCdec t + β3 ABS(∆ASSETS)t + β4 ∆CFOt + β5 ∆FINt + β6 ∆TIE t + ε

(3)

21

We use signed working capital accruals to be consistent with Bradshaw et al. (2001) but also report results using total accruals and with no control for accruals at all.

16

Where the variable ∆AUDITOR is an indicator variable set equal to one if an auditor change occurs in year t and zero otherwise. The independent control variables in this regression are based on prior literature. Generally, audit pricing studies (e.g., Simunic, 1980; Francis and Simon, 1987) and audit differentiation studies (e.g., Danos and Eichenseher, 1982; Francis and Wilson, 1988 and others) were used to identify client characteristics that influence supplier costs. Jointly, the variables identified from those studies were classified into four broad categories by Johnson and Lys (1990): expansion, profitability, financing , and audit risk. In our study, we proxy for expansion using the change in firm size (assets). If large audit firms supply a higher quality audit (DeAngelo, 1981) and if client agency costs increase with client size, then increases in client size are likely to be associated with changes in auditors. Conversely, if clients undergo a contraction then they are predicted to realign with comparatively smaller audit firms (Johnson and Lys, 1990). We measure the change in firm size as the absolute value of the change in the firm’s assets in year t relative to the average of the last 3 previous years (ABS(∆ASSETS)). If three previous years are not available we use the number of years for which data can be obtained (abs(Assetst-Assetst-3)/Assetst-3) (data item # 6). We proxy for profitability using the change in the firm’s cash flow from operations. If poor cash flow performance is symptomatic of enterprise decline, this variable serves as a lead indicator for contraction. We measure the change in the firm’s cash flow from operations, ∆CFOt, as the change in the firm’s cash flow from operations relative to the average of the last 3 previous years. If three previous years are not available we use the number of years for which data can be obtained ((CFOt-CFOt-3)/Assetst-3) (data item # 308 and item #6). We follow Bradshaw et al. (2001) and proxy for financing using, ∆FINt,which is the change in the firm’s use of external financing measured as ( τ =1 EqIsst −τ + τ =1 DebtIsst −τ Assets t −3 ) (EqIss is data item #108, Sale 3

3

17

of common and preferred stocks, DebtIss is item #111, Long term debt issuance). Finally, our proxy for audit risk (as in Johnson and Lys, 1990 and Bradshaw et al., 2001) is, ∆TIEt, which is the change in times interest earned as defined above, TIEt-TIEt-3. As aidot risk increases (TIE decreases) the change of auditor is more likely. Other variables are defined above. Our main empirical prediction is that if larger book-tax differences lead to a greater likelihood of auditor turnover we expect β 1 to be positive.22

4.

Empirical Results

4.1

Descriptive Statistics In Table 1, panel B, we present the industry composition of our sample firms as well as

the industry composition of all Compustat firms during our sample period. We exclude financial services from our sample because of the regulatory rules surrounding those firms for both tax and financial reporting. In addition, deferred tax data are not available on Compustat for financial services firms. Relative to the Compustat population we have an overrepresentation of machinery and manufacturing (SIC 30-39) but the other industry classifications are fairly similar. We include industry fixed effects in our regressions to control for any effect of industry on audit fees. Table 2 presents descriptive statistics for our sample firms. We report the logged data because those variables are the ones included in the regression analyses. However, to get a better sense of the data, our median unscaled, unlogged audit fee for our sample of

22

A caveat to this test is that the auditor change could be a result of the auditor resigning or the client firing the auditor. Bradshaw et al. (2001) implicitly assume the changes are a result of auditor resignations. In contrast, Defond and Subramanyam (1998) implicitly assume the changes reflect client decisions. We follow Bradshaw et al. (2001) and assume that the changes represent auditors’ decisions to resign because we are examining the auditors’ use of the book-tax difference information. We believe this potentially adds noise to our tests but can think of no reason why this would induce our results.

18

firm-years is roughly $229,000. Frankel, Johnson and Nelson (2002) report a median audit fee amount of $191,000 for their sample which included amounts from SEC filings in the year 2001 only. Thus, our audit fee amount appears reasonable. The median unlogged absolute value of total book-tax differences for our sample is 4.7% of average total assets and the median absolute value of temporary differences scaled by average total assets is 0.45% of average total assets. Our book-tax difference measure is not identical to our accruals measure indicating that the difference between book and taxable incomes is not exactly the same as the difference between financial accounting income and cash flows (similar to Hanlon 2005). Our firms have a mean (median) value for size (log of assets) of 5.228 (5.194) similar to the overall Compustat population mean of 5.126 (5.261).

4.2

Results for H1: The Association between Audit Fees and Book-Tax Differences Table 3 presents the main results for H1 via the estimation of equation (1). As predicted

the data reveal that larger book-tax differences are positively associated with higher audit fees (β 1 = 0.060 significant at less than 0.01, one-tailed in the pooled model (column 6)). The coefficient is positive and significant in every year (columns 1-4) and with or without a control for accruals (columns 5 and 6). We interpret this as evidence consistent with larger book-tax differences reflecting information that represents a higher risk of earnings management causing auditors to spend more time on the audit. Examining the pooled model (column 6) for simplicity, all of the control variables are significant in the predicted direction, consistent with prior literature (e.g., Simunic, 1980 and Larcker and Richardson, 2004). For example, the number of business segments (BUS SEG), the percentage of income that is foreign (FGN), the

19

indicator variable for whether the firm is a Big Five (or Four) audit client (BIG5(4)), and size (Ln(ASSET)) are all significant and positive indicating that larger more complex firms will pay larger audit fees. Larger inventory (INV) and receivables (REC) accounts are also positively associated with larger audit fees consistent with these accounts requiring more time to audit and with prior literature (Simunic 1980). The proxies for level of distress also are significant in the predicted directions; higher debt levels (DEBT) are associated with higher audit fees (although the variable is not significant in every year), higher income levels (INCOME) are associated with lower audit fees, loss firms (LOSS) have generally higher audit fees as do firms with a modified audit opinion (AUD OPIN). With respect to the accruals variable we find that signed accruals scaled by average total assets are positively associated with higher audit fees (significant at 0.05, one-tailed in the pooled model (column 6)) but this result is not consistent across all four years (column 1-4). As stated above, there is some debate in the literature regarding whether the accruals variable should be signed or unsigned. To ensure that the results for our main variable of interest, book-tax differences, are not sensitive to the measure of accruals we also estimate equation (1) using the log of the absolute value of accruals (untabulated). In this specification we find that the coefficient on the book-tax difference variable, β 1, continues to be positive and significant (coefficient is 0.054 with a t-statistic of 12.20, in the pooled model). The accruals variable is also positive and significant (coefficient is 0.037 with a t-statistic of 7.11 in the pooled model) and consistently so across all four years. Seemingly both large positive and negative accruals will increase audit fees consistent with auditors being aware that firms can manage earnings either upward or downward, both of which require additional examination.23 23

In our main tests we do not scale audit fees or book-tax differences by a size scaler but instead include in the model size, BIG5(4), and other variables that should control for the effect of size to be consistent with prior

20

Table 4 presents the estimation of equation (1) over the sample partitioned based on the level of the book-tax differences in order to examine whether large positive book-tax differences (where book income is in excess of taxable income) or large negative book-tax differences (where book income is less than taxable income) are driving the results presented in Table 3. Within the groupings of both negative book-tax differences (groups 1 and 2) and the groupings of positive book-tax differences (groups 4 and 5) we find, consistent with Table 3, that larger (in absolute value) book-tax differences are positively associated with larger audit fees. In Table 5 we present the results of estimating equation (1) but rather than total book-tax differences, we include the variable for temporary book-tax differences only—the log of the absolute value of the deferred tax expense. Similar to the results for total book-tax differences, the data reveal that in the pooled sample (column 1) the larger the temporary book-tax differences the larger the firm’s audit fee (t-stat of 4.76 significant at 0.01, one-tailed). Again, the relation holds for both positive temporary differences (indicating book income is higher than taxable income (column 2)) and negative temporary differences (indicating book income is less than taxable income (column 3)). All of the control variables are also again significant in the predicted direction.24

4.3

Supplementary Tests

4.3.1 The Relation between Book-Tax Differences and Audit Opinions literature (Simunic 1980). To ensure that scaling does not alter our results we also estimate equation (1) after scaling audit fees, absolute book-tax differences, and absolute accruals by average total assets (the numerator is not logged) and find (in untabulated results) that the book-tax difference variable continues to be positive and significant (coefficient is 0.002 with a t-statistic of 12.90). In this specification the scaled absolute value of accruals variable is insignificant. 24 We lose 2,560 observations that have a zero deferred tax expense and when logged get dropped. To test the effect of these observations as well as testing the sensitivity of the results to scaling, we estimate the regression using scaled audit fees and scaled deferred tax expense (where the scaler is average total assets and the numerator is not logged) and find (in untabulated results) that the coefficient on the scaled deferred tax expense is 0.030 (t-statistic of 12.01).

21

Table 6 presents the results of our tests that investigate whether there is an association between modified audit opinions and book-tax differences. Column (1) reports the results for total book-tax differences. The data reveal that total book-tax differences are positively associated with the likelihood of a modified audit opinion (β 1 coefficient is 0.305, significant at 0.01 one-tailed). All of the control variables are significant in the predicted direction except for the times interest earned, which is significant but positive consistent with firms that have higher short term solvency and liquidity receiving more modified audit opinions.25 Although not our main variable of interest, the results for the accrual variable warrant additional discussion. In our sample period, we find results consistent with larger working capital accruals (signed) being negatively associated with an modified audit opinion (in column 1, the coefficient on the decile ranking of working capital accruals is -0.716, significant at 0.01, one-tailed). Thus, consistent with the Bradshaw et al. (2001) findings in an earlier time period-it appears that auditors do not issue more modified opinions to firms with higher accruals and, in fact, issue fewer. We use working capital accruals as the accruals control to be consistent with Bradshaw et al. (2001). However, in untabulated tests, we also estimate the model controlling for total accruals rather than working capital accruals and find that in that specification total book-tax differences are still significantly positively associated with modified audit opinions with a coefficient of 0.209 (chi-square statistic is 5.50 significant at 0.05. The coefficient on total

25

We have no explanation why the coefficient on this variable is significant in the opposite direction than what is predicted. When we estimate the regression by year we find that the coefficient on TIE is negative in only one year, 2003. We also estimate the regression after winsorizing TIE at the 90% and 10% levels of its distribution but obtain the same results. Bradshaw et al. (2001) do not provide descriptive statistics on this variable and thus we cannot compare to theirs. However, we note that when we compute the odds ratios of the variables we find that this variable while statistically significant offers almost no economic significance with only an odds ratio of 1 meaning that a change in the TIE variable does not change the likelihood of having a modified audit opinion. In contrast, changing the decile ranking of the absolute value of book-tax differences by one rank makes the likelihood of a modified audit opinion 1.4 times as likely.

22

accruals is significantly negative similar to the working capital accrual results above. We also estimate the regression without any control for accruals (results not tabulated) and find again that the coefficient on the total book-tax difference variable is positive and significant (coefficient is 0.361 and p-value is 0.0001). Thus, the inference with regard to our main variable of interest under both of these sensitivity analyses is consistent with reported results. Column (2) presents the results for the deferred tax expense (temporary book-tax differences only). The coefficient on the decile ranking of the absolute value of the deferred tax expense is -0.069 which is insignificant at conventional levels. Thus, it does not appear that temporary book-tax differences reveal information about earnings quality, beyond the information in working capital accruals, relevant in predicting whether the auditor will issue an modified audit opinion. Again, in untabulated sensitivity analysis we estimate this model controlling for total accruals rather than working capital accruals and find that in that specification the coefficient on the deferred tax expense variable is 0.156 (chi-square statistic of 3.15 significant at 0.10, onetailed). We also estimate this model without any control for accruals and find that the deferred tax expense variable becomes significant and positive (coefficient of 0.271 significant at less than 0.01, one-tailed). Thus, there is some evidence that the temporary book-tax differences reflect information useful to auditors in their decision to issue a modified audit opinion but it appears that this same information is reflected in working capital accruals.

4.3.2 The Relation between Book-Tax Differences and Auditor Turnover Table 7, column (1), presents results from our tests investigating the relation between book-tax differences and auditor turnover. The coefficient on total book-tax differences (decile

23

ranking of the scaled absolute value of book-tax differences) is positive and significant at less than 0.01, one-tailed. With regard to the control variables they are either insignificant or consistent with predictions. The accruals variable in Table 7 column (1), WCACCdec, is not significant. Bradshaw et al. (2001) report a negative coefficient on the working capital accruals variable consistent with higher working capital accrual firms having a lower incidence of auditor changes. In our sample period and firms we find no relation between the working capital accruals variable and the change in auditors.26 Similar to our tests for the modified audit opinion, we estimate several alternative specifications as sensitivity analyses (all of which are untabulated). First we estimate the model controlling for total accruals rather than working capital accruals and find that in that specification the coefficient on total book-tax differences is 0.243 (chi-square statistic is 3.02 significant at 0.02, one-tailed - untabulated). We also estimate the regression without any control for accruals and find again that the coefficient on the total book-tax difference variable is positive and significant (coefficient is 0.312 and the p-value is 0.01, one-tailed - untabulated). Thus the inference with regard to our main variable of interest under both of these sensitivity analyses is consistent with reported results. Table 7, column (2) presents the results using the temporary book-tax difference measure (i.e., the scaled, ranked absolute value of the deferred tax expense) as our book-tax difference variable. We find that there is a significantly negative relation between the deferred tax expense

26

We again estimate the odds ratios and find that for a change of one decile ranking in the book-tax difference measure the likelihood of an auditor change is 1.7 times greater. In contrast, a change in the working capital accruals by one ranking makes the likelihood of an auditor change only .936 times as likely. We also note that although the models used in Tables 6 and 7 for the modified audit opinion and auditor turnover have been used in prior research, the models do not provide very predictive tests. For example, we computed the area under the ROC curve as described in Hosmer and Lemeshow (2000) and find that the audit opinion model is poor (area = 0.6145) and the auditor change model fails (area = 0.5597) according to the guidelines published by Hosmer and Lemeshow (2000). Neither model is considered worthless (which would be an area under 50%), however, we recognize the weakness of the models.

24

and the likelihood of an auditor change (coefficient of -2.79, significant at 0.05, one-tailed). The control variables are similar in direction and significance to the results reported in column (1). In untabulated sensitivity analysis, we again estimate the model using total accruals rather than working capital accruals and without any control for accruals. The data reveal that in both of these specifications the coefficient on the deferred tax expense variable becomes positive and significant (at 0.01, one-tailed in both specifications). Thus, there is some evidence that deferred tax expense also reflects information about earnings quality that affects the decisions of auditors to fire the client (and issue an modified opinion), however, these results are sensitive to the inclusion and definition of the control variable for accruals.

5.

Sensitivity Analyses We perform several additional sensitivity tests (all of which are untabulated). Because

tax credits can cause problems with the measurement of total book-tax differences, we estimate our audit fee regression by excluding firms with a positive research and development expense in the Compustat database (data item #46) as these firms likely utilize the research and development tax credit. The results are consistent with our main results presented in Table 3 of the paper -- the absolute value of total book-tax differences is positively associated with audit fees (coefficient is 0.058 and the t statistic is 8.98). We also estimate the regression separately for firms above and below the median level of research and development expense in our sample and find similar results for both sub-samples—the coefficient on total book-tax differences is positive and significant at less than 0.05 level. We also estimate the audit fee regression after excluding firms with foreign source income on the Compustat database in order to exclude firms that may have a book-tax difference

25

measure affected by the foreign tax credit. Excluding these firms also provides an additional test of whether the complexity of the firm (proxied by the extent of multinational operations) is driving our results. Again we find results that are consistent with our main results—the coefficient on the total book-tax differences variable is positive and significant (t-stat of 13.32). Thus, it does not appear that our results are driven by measurement error due to tax credits in the calculation of book-tax differences. In addition, to the extent that these variables also proxy for firm complexity our results continue to hold in the presence of these controls. To further investigate whether the results are driven by only one direction of book-tax differences we create a variable which is equal to the positive book-tax differences when the firm’s book-tax difference is positive and equal to zero otherwise. Similarly, we create a variable that is equal to the negative book-tax differences when the book-tax differences are negative and is equal to zero otherwise. We include both these variables in our audit fee regression in place of the absolute value of the book-tax differences. The results are consistent with those reported earlier in Table 4 for the quintiles of firms—the larger the book-tax differences for positive book-tax difference firms, the higher the audit fees and the smaller (more negative) book-tax differences for negative book-tax differences firms, the larger the audit fees. Thus, the larger in absolute value the book-tax differences the higher the audit fees. To provide another control for the complexity of the firm and especially the tax complexity of the firm (Mills, Erickson and Maydew 1998 and Omer, Bedard, and Falsetta 2005) we include the tax fees paid to the auditor as a control variable in our audit fee regression. The tax fee variable is positive and significant at 0.01 consistent with tax fees proxying for complexity of the firm and complex firms requiring more time and effort spent on the audit. Our main variable of interest, the total book-tax differences, remains positive and significant with a

26

coefficient of 0.05 (t-stat 8.26). We note that the sample size is reduced significantly by requiring this variable in our sample (N=4,418) and the tax fees paid to the auditor does not include all tax fees paid by the firm. For these reasons we do not include this variable in our main analysis. We also estimate our audit fee regression after adding controls for the presence of special items and merger and acquisition activity. These two items often generate book-tax differences while at the same time may proxy for firm complexity or simply indicate a one time event for the firm. In some sense, we want to test whether book-tax differences provide information about these items to auditors about the quality of earnings, we also want to be sure that our results are not entirely driven by the presence of special items or merger or acquisition activity. Thus, we estimate equation (1) after including 1) an indicator variable set equal to one for firm-years with non-zero, non-missing special items (data #17) and zero otherwise, and 2) an indicator variable set equal to 1 when the firm has non-zero, non-missing cash flows from mergers and acquisitions and zero otherwise.27 After including these variables the coefficient on ln (ABS BTD) continues to be positive and significant (coefficient of 0.060, t-stat of 13.91). The special item indicator variable is also positive and significant at the .05 level but the merger and acquisition indicator variable is insignificant. Thus, although the presence of special items appears to increase audit fees (i.e., likely proxy for low earnings quality and require more audit time and effort) the presence of special items is not driving our results for the book-tax differences. In our main tests, we use contemporaneous book-tax differences and audit fees under the assumption that although audit fees are somewhat set prior to the start of the audit, the audit firm can bill the additional time and effort expended based on current year facts and circumstances. If

27

Note we also proxy for M&A using data item #249 (sales attributable to mergers and acquisitions) to construct our indicator variable and the inferences are unchanged.

27

this is not the case, however, it may be better to examine the relation between lagged book-tax differences and current audit fees. Thus, we re-estimate our audit fee regression (equation (1)) using lagged book-tax differences rather than current book-tax differences. We find that lagged book-tax differences (measured similarly to our main tests—log of the absolute value) are positively associated with current audit fees consistent with lagged book-tax differences influencing current audit fees as well. Our final sensitivity test with regard to the audit fee model is to investigate whether changes in book-tax differences are associated with changes in audit fees. Although to our knowledge prior literature has not implemented the use of a changes model with respect to audit fees we examine this relation to provide further support for our conjecture that book-tax differences provide information about a firm’s earnings quality that should be useful to auditors. We alter our main equation (1) by regressing the log of the change in audit fees on the log of the change in the absolute value of book-tax differences. We also change many of the control variables (except BIG5(4), LOSS, and OPINION) to a change specification to be consistent with the dependent variable. We find results consistent with a large change in the absolute value of book-tax differences being positively associated with a large change in audit fees (coefficient on the change in book-tax difference variable is 0.223 and the t-stat is 10.68). We also perform several robustness checks on our supplemental tests. Because Arthur Andersen was indicted and subsequently lost its license to practice, many companies changed auditors in the year 2002 from Arthur Andersen to another audit firm. If these firms also had large book-tax differences then our results could be affected (and if they did not have large booktax differences these firms would add noise to our tests). To examine this we exclude observations in the year 2002, the year the clients switched, and re-estimate the model. Our

28

results are very similar to those reported in Table 7 – the coefficient on the total book-tax difference variable is positive and significant (coefficient of 0.791 and a chi-squared statistic of 28.30). We follow Bradshaw et al. (2001) in using the portfolio rank of working capital accruals and also book-tax differences in our tests of auditor changes with all other control variables measured as a change consistent with the dependent variable. In order to take into account the possibility that a change in book-tax differences or change in accruals is the determinant of the change in the auditor we compute the change in the absolute value of book-tax differences and the change in working capital accruals from year t-1 to year t (the same period over which the auditor change is taken). We then rank these change variables. We find that the coefficient on the rank of the change in book-tax differences remains positive and significant (coefficient is 0.4437 with a p-value of 0.0001) but the coefficient on the rank of the change in the working capital accruals becomes insignificant (p-value of 0.1790). Finally, because audit fee data are not required for the tests of audit opinions and auditor turnover we can use years earlier than the year 2000 to test the association of book-tax differences and audit opinions and auditor turnover. Because the accounting for taxes changed significantly in 1993 we obtain data starting in 1994 and re-estimate regression equations (2) and (3). We find almost identical results to those reported and thus our results with respect to audit opinions and auditor turnover do not appear to be time period specific.

6.

Conclusions This paper investigates whether auditors utilize the information in book-tax differences.

Specifically, we examine whether book-tax differences are associated with higher audit fees,

29

more modified audit opinions, and greater auditor turnover. We measure book-tax differences as both total book-tax differences (temporary and permanent) and using a measure of temporary differences alone. This paper contributes to the literature that investigates the information reflected in the differences between book and taxable incomes and whether market participants utilize this information. The data are consistent with larger book-tax differences being associated with higher audit fees. We interpret this evidence as indicating that the book-tax differences reflect information that represents a higher risk of earnings management which increases auditor’s efforts and time spent on the audit. Our data indicate that these results are robust to alternative definitions of book-tax differences (temporary and total) providing evidence that the results are not simply driven by the auditors testing for earnings management of the provision accounts (i.e., the valuation allowance and the tax cushion) but that the book-tax differences reflect information about pre-tax earnings quality as well. Our results are also robust to several controls for firm complexity, attempts to control for measurement error in the calculation of book-tax differences, and a control for the accrual level of the firm. Our results are also generally consistent with larger book-tax differences being associated with more modified audit opinions and a greater incidence of auditor turnover. We interpret this evidence as providing additional support that the book-tax differences reflect information about earnings quality and that our audit fee results are not simply due to larger book-tax differences indicating a more complex, time consuming audit. Overall, we interpret the sum of the evidence as being consistent with the book-tax differences reflecting information about possible earnings management and with auditors examining and using this information in auditing and opining on firms’ financial statements.

30

REFERENCES American Institute of Certified Public Accountants (AICPA). 1983. Statement on Auditing Standards: No. 47 “Audit Risk and Materiality in Conducting and Audit.”, New York, NY: AICPA. American Institute of Certified Public Accountants (AICPA). 1997. Statement on Auditing Standards No. 82 “Consideration of Fraud in a Financial Statement Audit.” New York, NY: AICPA. American Institute of Certified Public Accountants (AICPA). 2003. Statement on Auditing Standards No. 99 “Consideration of Fraud in a Financial Statement Audit.” New York, NY: AICPA. Bedard, J. C. 1989. “An Archival Survey of Audit Program Planning.” Auditing: A Journal of Practice & Theory (Fall): 57-71. Bedard, J. and K. Johnstone. 2004. “Earnings Manipulation Risk, Corporate Governance Risk, and Auditors’ Planning and Pricing Decisions.” The Accounting Review 79 (April): 277304. Bell, T., W. Landsman, and D. Shackelford. 2001. “Auditors’ Perceived Business Risk and Audit Fees: Analysis and Evidence.” Journal of Accounting Research 39 (June): 35-43. Bradshaw, M., S. Richardson, and R. Sloan. 2001. “Do Analysts and Auditors use the Information in Accruals?” Journal of Accounting Research 39 (June): 45-74. Danos, P. and J. Eichenseher. 1982. “Audit Industry Dynamics: Factors Affecting Changes in Client Industry Market Shares.” Journal of Accounting Research 20 (Autumn): 604-616. Davis, L. R., D. N. Ricchuite, and G. Trompeter. 1993. “Audit Effort, Audit Fees, and the Provision of Nonaudit Services to Audit Clients.” The Accounting Review 68 (January): 135-150. DeAngelo, L. E., 1981. “Auditor Size and Audit Quality.” Journal of Accounting and Economics 3 (December): 183-199. DeFond, M. and K.R. Subramanyam. 1998. “Auditor Changes and Discretionary Accruals.” Journal of Accounting and Economics 25 (1998): 35-67. Desai, M. and D. Dharmapala. 2005. “Corporate Tax Avoidance and High Powered Incentives.” Forthcoming, Journal of Financial Economics. Dhaliwal, D., C. Gleason, and L. Mills. 2004. “Last Chance Earnings Management: Using the Tax Expense to Achieve Earnings Targets.” Contemporary Accounting Research 21 (Summer): 431-459.

31

Francis, J. and D. Simon. 1987. “A Test of Audit Pricing in the Small-Client Segment of the U.S. Audit Market.” The Accounting Review 62 (January): 145-157. Francis, J. and E.R. Wilson. 1988. “Auditor Changes: A Joint Test of Theories Relating to Agency Costs and Auditor Differentiation.” The Accounting Review 63 (October): 663682. Frankel, R. M., M. F. Johnson, and K. K. Nelson. 2002. “The Relation Between Auditor’s Fees for Non-Audit Services and Earnings Management.” The Accounting Review 77 (Supplement): 71-105. Gleason, C., and L. Mills. 2002 “Materiality and Contingent Tax Liability Reporting.” The Accounting Review 77 (April): 317-342. Hanlon, M. 2003. “What Can We Infer About a Firm’s Taxable Income from its Financial Statements?” National Tax Journal 56 (December): 831-863. Hanlon, M. 2005. “The Persistence and Pricing of Earnings, Accruals and Cash Flows when Firms Have Large Book-Tax Differences.” The Accounting Review 80 (January): 137166. Hanlon, M. S. Kelley, and T. Shevlin. 2005. “Evidence on the Possible Information Loss of Conforming Book Income and Taxable Income.” Forthcoming, Journal of Law and Economics. Hosmer, D. W. and S. Lemeshow. 2000. Applied Logistic Regression. John Wiley and Sons, Inc. New York, New York. Johnson, W. B. and T. Lys. 1990. “The Market for Audit Services: Evidence from Voluntary Auditor Changes.” Journal of Accounting and Economics 12 (January): 281-308. Johnstone, K. and J. Bedard. 2001. “Engagement Planning, Bid Pricing, and Client Response in the Market for Initial Attest Engagements.” The Accounting Review 76 (April): 199-220. Joos, P., J. Pratt, and D. Young. 2000. “Book-tax Differences and the Value Relevance of Earnings.” Working paper, Massachusetts Institute of Technology, Indiana University and Insead. Larcker, D. and S. Richardson. 2004. “Fees Paid to Audit Firms, Accrual Choices and Corporate Governance.” Journal of Accounting Research 42 (June): 625-658. Lev, B. and D. Nissim. 2004. “Taxable Income, Future Earnings, and Equity Values.” The Accounting Review 79 (October): 1039-1074.

32

MacDonald, Elizabeth. 1997. “More Accounting Firms are Dumping Risky Clients – Big Six Worry About Their Reputations and Chance of Costly Litigation.” The Wall Street Journal (April 25): A2. Maletta, M., and T. Kida. 1993. “The Effect of Risk Factors on Auditors’ Configural Information Processing.” The Accounting Review 68 (January): 681-691. Manzon, G., Jr., and G. Plesko. 2002. “The Relation Between Financial and Tax Reporting Measures of Income.” 55 Tax Law Review. Messier, W. and D. Plumlee. 1987. “The Effects of Anticipation and Frequency of Errors on Auditors’ Selection of Substantive Procedures.” Accounting and Business Research: 349358. Miller, G. and D. Skinner. 1998. “Determinants of the Valuation Allowance for Deferred Tax Assets Under SFAS 109.” The Accounting Review 73 (2): 213-233. Mills, L. 1998. “Book-Tax Differences and Internal Revenue Service Adjustments.” Journal of Accounting Research (Autumn): 343-356. Mills, L., M. Erickson, and E. Maydew. 1998. “Investments in Tax Planning.” The Journal of the American Taxation Association 20: 1-21. Mills, L., and K. Newberry. 2001. “The Influence of Tax and Non-Tax Costs on Book-Tax Reporting Differences: Public and Private Firms.” Journal of American Taxation Association 23 (Spring): 1-19. Mills, Lillian F., and George A. Plesko. 2003. “Bridging the Reporting Gap: A Proposal for More Informative Reconciling of Book and Tax Income.” National Tax Journal 56 No. 4 (December): 865-893. Mock, T.J. and J.L. Turner. 2001. “An Archival Study of Audit Fraud Risk Assessments Made Under SAS 82.” Auditing Standards Board of the AICPA Research Monograph. Mock, T. J. and A. Wright. 1993. “An Exploratory Study of Auditors’ Evidential Planning Judgments.” Auditing: A Journal of Practice & Theory 12 (Fall): 39-61. O’Keefe, T. B., D. A. Simunic, and M. T. Stein. 1994. “The Production of Audit Services: Evidence from a Major Public Accounting Firm.” Journal of Accounting Research 32 (Autumn): 241-261. Omer, T., J. Bedard, and D. Falsetta. 2005. “Auditor-Provided Tax Services: The Effects of a Changing Regulatory Environment.” Working paper, Texas A &M University. Phillips, J., M. Pincus, and S. Rego. 2003. “Earnings Management: New Evidence Based on the Deferred Tax Expense.” The Accounting Review 178 (April): 491-522.

33

Plesko, George A. 2002. “Reconciling Corporation Book and Tax Net Income, Tax Years 19961998.” SOI Bulletin (Spring): 1-16. Sarbanes-Oxley Act of 2002. P.L. No. 207-204. Washington D.C.: Government Printing Office. Schrand, C. and M.H.F. Wong. 2004. “Earnings Management Using the Valuation Allowance for Deferred Tax Assets Under FAS 109.” Contemporary Accounting Research 20 (Fall): 579-611. Simunic, D., 1980. “The Pricing of Audit Services: Theory and Evidence.” Journal of Accounting Research 18 (Spring): 161-190. Weber, D. 2005. “Book-Tax Differences, Analysts’ Forecast Errors, and Stock Returns.” Working paper, The University of Connecticut.

34

TABLE 1: Industry Distribution for Sample Firms

2digit SIC code 1-14 15-17 20-21 22-23 24-27 28-32 33-34 35-39 40-48 50-52 53-59 49, 70-79 80-99 60-67

Agriculture and mining Construction and plumbing Food and kindred products and cigarettes Textile mill products and apparel Lumber , furniture, paper, and printing Chemicals, petroleum, rubber, leather and stone Metal Machinery, electrical and computer equip., scientific instruments, miscellaneous manuf. Railroads, motor freight, transportation, communications Wholesale goods, building material, hardware retail Stores merchandise, auto dealers, apparel, home furniture stores, eating and drinking, misc. retail Lodging services, business services, other services

290 77 204 137 321 1,277 266 2,789

3.17 0.84 2.23 1.49 3.49 13.93 2.91 30.41

Compustat Number of Firm-Year Observations 5,224 1,020 1,704 1,260 2,436 6,772 2,080 15,920

512

5.59

5,408

6.06

369

4.02

3,520

3.94

712

7.76

5,012

5.62

1,688

18.41

15,292

17.15

Other Financial services Total

528 0 9,170

5.75 0 100%

4,572 18,996 89,216

5.12 21.29 100%

Industry

Number of Firm-Year Observations

%

Data are for the years 2000 through 2003. Total number of firm-year observations equals 1,506, 2,530, 2,682, and 2,452, respectively, for years 2000 through 2003.

35

%

5.85 1.15 1.91 1.41 2.73 7.59 2.33 17.85

TABLE 2: Descriptive Statistics and Correlations Panel A: Descriptive Statistics for Audit Fee Regression Variables Variable

Mean -1.354 2.191 2.556 0.847 5.228 1.366 0.087 0.126 0.156 0.211 -0.029 0.528 0.381

Ln(AUDIT FEE) Ln(ABS BTD) Ln(ABS ACC) BIG5(4) Ln(ASSET) BUS SEG FGN INV REC DEBT INCOME LOSS AUD OPIN

Standard Deviation 1.143 2.127 2.145 0.360 2.025 0.472 0.368 0.149 0.125 0.256 0.273 0.499 0.486

Minimum -5.846 -3.101 -6.908 0.000 -2.797 1.000 -1.247 0.000 0.000 0.000 -2.532 0.000 0.000

25% -2.133 0.797 1.164 1.000 3.845 1.000 0.000 0.005 0.062 0.006 -0.081 0.000 0.000

Median

75%

-1.473 2.186 2.557 1.000 5.194 1.000 0.000 0.078 0.135 0.143 0.047 1.000 0.000

Maximum

-0.686 3.604 3.953 1.000 6.545 1.732 0.000 0.193 0.217 0.331 0.114 1.000 1.000

4.382 10.732 10.851 1.000 13.381 3.162 2.031 1.708 1.007 4.323 0.847 1.000 1.000

Panel B: Descriptive Statistics for Audit Opinion and Auditor Change Regression Variables Variable

Mean

CFO TIE abs(∆ ASSET) ∆ CFO ∆ FIN ∆ TIE

-0.018 -11.981 1.075 -0.052 1.413 -4.053

Standard Deviation 0.263 60.212 3.703 0.791 3.004 73.174

Minimum -1.296 -200.000 -1.000 -5.709 0.000 -217.000

25% -0.063 -7.038 -0.191 -0.067 0.146 -7.043

Median

75%

0.049 1.249 0.172 0.022 0.505 -0.223

Maximum

0.116 5.717 0.845 0.129 1.317 5.586

0.416 100.000 28.302 1.943 22.480 170.000

Panel C: Correlations for Audit Fee Regression Variables ACC Ln(ABS BTD) Ln(ABS ACC) BIG5(4) Ln(ASSET) BUS SEG FGN INV REC DEBT INCOME LOSS AUD OPIN a, b,

0.709a 1.000

BIG5 (4) 0.362a 0.389a 1.000

Ln (ASSET) 0.693a 0.831a 0.456a 1.000

BUS SEG 0.204a 0.276a 0.084a 0.346a 1.000

FGN

INV

REC

DEBT

0.106a 0.135a 0.070a 0.183a -0.003 1.000

-0.238a -0.138a -0.096a -0.061a -0.008 -0.008 1.000

-0.218a -0.156a -0.107a -0.094a 0.112a 0.045a 0.131a 1.000

0.121a 0.160a -0.054a 0.152a 0.095a 0.016 0.041a -0.018c 1.000

INCOME 0.018c 0.207a 0.136a 0.438a 0.195a 0.097a 0.198a 0.225a 0.014 1.000

LOSS 0.019c -0.141a -0.098a -0.338a -0.156a -0.148a -0.203a -0.156a 0.014 -0.535a 1.000

and c indicate two-tailed significance at the 0.01, 0.05, and 0.10 levels respectively.

36

AUD OPIN 0.176a 0.176a 0.090a 0.161a 0.107a 0.043a -0.040a -0.043a 0.121a -0.011 0.027a 1.000

Table 2 (continued): Data are for the years 2000 through 2003. Total number of firm-year observations for Panel A equals 1,506, 2,530, 2,682, and 2,452, respectively, for years 2000 through 2003 for a total of 9,170 observations. The number of observations for the modified audit opinion model (Panel B first two rows above) is 8,918 and for the auditor change model (Panel B last four rows) is 7,155 because requiring the change in the variables and additional variables results in a loss of observations due to missing data. All variables for the modified audit opinion and audit change models are winsorized at the 1% and 99% of the distributions. Variables are defined as follows: Ln(AUDIT FEE) = Log of audit fee; Ln(ABS BTD) = Log of absolute value the total book-tax differences (i.e., spread between pre-tax book income and taxable) income (data #170 – ((data #16 – data #50)/.35 – ∆data#52). ACC t = Signed total accruals measured as the difference between earnings and cash flow from operations (data #18 – (data #308 + data#124) scaled by average total assets (data #6); BIG5(4) = An indicator variable set equal to 1 for client-observations audited by a Big 5 (or Big 4) auditor, and 0 otherwise (item # 149); Ln(ASSET) = Log of total assets (item # 6); BUS SEG = Square root of the number of business segments owned by the client; FGN = Ratio of foreign sourced pre-tax book income to total pre-tax book income (item # 273/# 170); INV = Inventory to average total assets (item # 3/item #6); REC = Receivables to average total assets (item # 3/item #6); DEBT = Sum of short-term debt and long-term debt to average total assets ((item #34 + item#9)/item # 6); INCOME = The ratio of operating income after depreciation to average total assets (item #78/item #6); LOSS = An indicator variable set equal to 1 if income before extraordinary items and discontinued operations is negative in the current or two previous years, and 0 otherwise (item #18); AUD OPIN = An indicator variable set equal to 1 if the firm receives a modified audit opinion, and 0 otherwise (item # 149); CFO = the firm’s cash flows from operations (data item # 308); TIE = the firm’s operating earnings divided by interest expense (item # 178/#15); ABS(∆ ASSET) = the absolute value of the change in total assets between t and t-3 divided by total assets at t-3; ∆ CFO = change in the firm’s cash flow from operations relative to the average of the last 3 previous years. If three previous years are not available we use the number of years for which data can be obtained. ((CFOt-CFOt-3)/Assetst-3) (data item # 308 and item #6); ∆FINt = change in the firm’s use of external financing measured as (

∆TIEt

=

τ

3

EqIsst −τ + τ =1 DebtIsst −τ Assets t −3 ) (EqIss is data item #108, Sale of =1 3

common and preferred stocks, DebtIss is item #111, Long term debt issuance); change in times interest earned as defined above, TIEt-TIEt-3.

37

TABLE 3: Regression of Log of Audit Fees on the Log of Book-Tax Differences and Controls Ln(AUDIT FEE) = α + β1 Ln(ABS BTD)t + β2 ACCt + β3BIG5(4)t + β 4Ln(ASSET)t + β5BUS SEGt + β 6FGNt + β7INVt + β8RECt + β 9DEBTt + β 10 INCOMEt + β11LOSSt + β12 AUD OPINt + β 13-17INDt +e (1) Independent Variables (Expected Sign) Intercept (?)

Ln(ABSBTD) (+) ACC (+) BIG5(4) (+) Ln(ASSET) (+) BUS SEG (+) FGN (+) INV (+) REC (+) DEBT (+) INCOME (-) LOSS (+) AUD OPIN (+) Adjusted R2 N

2000

2001

(1) -4.954 (-45.03)a 0.048 (4.44)a 0.039 (1.82)b -0.013 (-0.20) 0.481 (33.62)a 0.275 (8.87)a 0.235 (5.66)a 0.426 (3.39)a 1.069 (8.45)a -0.006 (-0.10) -0.338 (-4.85)a 0.093 (2.49)a 0.189 (4.68)a 0.746 1,506

(2) -4.591 (-64.45)a 0.058 (7.36)a -0.067 (-1.48)c 0.109 (3.26)a 0.422 (41.97)a 0.259 (10.67)a 0.286 (8.98)a 0.368 (4.16)a 1.014 (10.68)a 0.027 (0.65) -0.356 (-6.43)a 0.135 (5.05)a 0.146 (5.63)a 0.771 2,530

Coefficient (t-statistics) 2002 2003 Pooled Model (3) (4) (5) -4.445 -4.472 -4.368 (-60.92)a (-60.48)a (-110.86)a 0.071 0.054 0.060 (8.55)a (6.49)a (13.81)a −−− 0.007 -0.077 b c (1.88) (-1.33) 0.209 0.263 0.171 (6.34)a (7.82)a (9.36)a 0.416 0.448 0.433 (39.73)a (42.05)a (78.45)a 0.224 0.249 0.250 (8.74)a (9.65)a (18.88)a 0.178 0.189 0.217 (6.00)a (6.60)a (13.56)a 0.314 0.292 0.327 (3.33)a (2.89)a (6.56)a 0.955 0.938 0.986 (9.56)a (8.94)a (18.77)a 0.059 0.061 0.056 (1.23) (1.41)c (2.36)a -0.365 -0.421 -0.356 (-6.78)a (-6.68)a (-12.41)a 0.179 0.194 0.155 (6.74)a (7.11)a (10.82)a 0.130 0.096 0.133 a a (5.69) (4.01) (10.20)a 0.769 0.789 0.773 2,682 2,452 9,170

Pooled Model (6) -4.366 (-110.81)a 0.060 (13.84)a 0.007 (2.17)b 0.170 (9.31)a 0.433 (78.44)a 0.250 (18.87)a 0.217 (13.56)a 0.328 (6.57)a 0.985 (18.75)a 0.058 (2.47)a -0.358 (-12.48)a 0.155 (10.84)a 0.132 (10.17)a 0.773 9,170

a, b, c

indicates one-tailed significance at the 0.01, 0.05, 0.10 levels, respectively. See Table 2 for variable definitions. t-statistics are in parentheses. ACC here represents signed accruals scaled by average total assets. Both the yearly and the pooled models include twelve industry-dummy variables to represent the following two-digit SIC categories, respectively (similar to Ashbaugh et al 2003 TAR): SIC 0114, SIC 15-19, SIC 20-21, SIC 22-23, SIC 24-27, SIC 28-32, SIC 33-34, SIC 35-39, SIC 40-48, SIC 50-52, SIC 53-59, SIC 70-79. The pooled models also include three year-dummy variables to represent years 2000, 2001, and 2002.

38

TABLE 4: Regression of Log of Audit Fees on Log of Book-Tax Differences and Control Variables: Observations Partitioned by Signed Magnitude of the Book-Tax Differences Ln(AUDIT FEE) = α + β1 Ln(ABS BTD)t + β2 ACCt + β3BIG5(4)t + β 4ln(ASSET)t + β5BUS SEGt + β 6FGNt + β7INVt + β8RECt + β 9DEBTt + β 10 INCOMEt + β11 LOSSt + β12 AUD OPINt + β 13-17 INDt +e (1) Independent Variables (Expected Sign) Intercept (?)

Ln(ABS BTD) (+) ACC (+) BIG5(4) (+) Ln(ASSET) (+) BUS SEG (+) FGN (+) INV (+) REC (+) DEBT (+) INCOME (-) LOSS (+) AUD OPIN (+) Adjusted R2 N

Group 1

Coefficient (t-statistics) Group 2 Group 3 Group 4

-4.409 (-37.85)a 0.095 (6.18)a 0.040 (1.82)b 0.061 (0.89) 0.431 (30.64)a 0.323 (10.51)a 0.159 (5.22)a 0.342 (2.11)b 1.675 (10.85)a 0.056 (0.97) -0.469 (-6.57)a -0.009 (-0.21) 0.145 (4.79)a 0.729 1,834

-3.897 (-40.75)a 0.062 (3.36)a 0.013 (0.40) 0.202 (5.80)a 0.364 (30.01)a 0.219 (7.16)a 0.181 (5.05)a 0.208 (2.04)b 0.780 (7.31)a 0.044 (1.03) -0.213 (-4.12)a 0.084 (2.49)a 0.119 (4.38)a 0.618 1,834

-4.119 (-54.23)a 0.023 (1.97)b 0.007 (2.13)b 0.286 (9.99)a 0.397 (36.35)a 0.133 (4.39)a 0.228 (5.44)a 0.123 (1.48)c 0.575 (6.14)a 0.086 (1.88)b -0.250 (-4.14)a 0.127 (4.56)a 0.166 (5.91)a 0.656 1,834

-4.239 (-46.20)a 0.052 (2.42)a -0.221 (-2.04)b 0.261 (6.09)a 0.422 (32.35)a 0.137 (4.84)a 0.239 (5.88)a 0.322 (3.24)a 0.872 (8.06)a 0.134 (2.20)b -0.270 (-3.44)a 0.210 (6.53)a 0.087 (3.17)a 0.554 1,834

Group 5 -5.068 (-40.18)a 0.123 (7.39)a 0.097 (0.91) 0.229 (2.66)a 0.473 (32.24)a 0.256 (9.16)a 0.241 (7.03)a 0.732 (4.82)a 1.494 (10.58)a 0.041 (0.67) -0.318 (-2.71)a 0.197 (5.36)a 0.111 (3.65)a 0.730 1,834

a, b, c

indicates one-tailed significance at the 0.01, 0.05, 0.10 levels, respectively. See Tables 2 and 3 for variable definitions. Observations are partitioned into quintiles based on the book-tax difference variable (pre-tax book income less taxable income). Quintile 1 (5) represents observations with the highest negative (positive) book-tax differences. As in Table 3, three year-dummy variables and twelve industry-dummy variables are included as controls.

39

TABLE 5: Regression of Log of Audit Fees on Control Variables and Log of Absolute Deferred Tax Expense Ln(AUDIT FEE) = α + β1 Ln(ABS DEXP)t + β 2 ACCt + β 3BIG5(4)t + β 4Ln(ASSET)t + β5BUS SEGt + β 6FGNt + β7INVt + β8RECt + β 9DEBTt + β 10 INCOMEt + β11 LOSSt + β12 AUD OPINt + β 13-17 INDt +e (1)

Independent Variables (Expected Sign)

Intercept (?)

Ln(ABS DEXP)t (+) ACCt (+) BIG5(4) t(+) Ln(ASSET)t (+) BUS SEGt (+) FGNt (+) INVt (+) RECt (+) DEBTt (+) INCOMEt (-) LOSSt (+) AUD OPINt (+) Adjusted R2 N

Absolute Deferred Tax Expense (1) -4.552 (-92.26)a 0.023 (4.76)a 0.006 (1.87)b 0.146 (6.35)a 0.492 (76.60)a 0.230 (15.96)a 0.199 (12.06)a 0.239 (4.19)a 0.819 (13.48)a 0.057 (1.74)b -0.297 (-4.69)a 0.215 (12.64)a 0.137 (8.84)a 0.783 6,610

Coefficient (t-statistics) Positive Deferred Tax Expense (2) -4.634 (-69.55)a 0.018 (2.70)a 0.007 (1.87)b 0.160 (5.08)a 0.502 (57.77)a 0.208 (10.82)a 0.245 (9.37)a 0.307 (3.91)a 0.933 (10.94)a 0.051 (1.15) -0.368 (-3.89)a 0.253 (10.50)a 0.127 (6.19)a 0.783 3,732

Negative Deferred Tax Expense (3) -4.387 (-58.48)a 0.035 (4.87)a -0.055 (-0.95) 0.125 (3.72)a 0.475 (49.68)a 0.254 (11.72)a 0.155 (7.26)a 0.154 (1.82)b 0.670 (7.61)a 0.103 (2.05)b -0.206 (-2.33)b 0.138 (5.46)a 0.153 (6.50)a 0.787 2,878

a, b, c

indicates one-tailed significance at the 0.01, 0.05, 0.10 levels, respectively. Ln(ABS DEXP) = Log of absolute value of deferred income tax expense (data item #50). See Tables 2 and 3 for definitions of other variables. As in Table 3, three year-dummy variables and twelve industry-dummy variables are included as controls.

40

TABLE 6: Logit Regression of Audit Opinion on Book-Tax Difference Portfolios and Control Variables AUD OPINt = α + β1 (ABS BTD)dect + β2 WCACCdec t+ β3 ln(ASSET) + β4CFOt + β5DEBTt + β6TIEt + ε

Independent Variables (Expected Sign) Intercept (?)

(ABS BTD)dec t (+) (ABS DEXP)dect (+) WCACCdect (+) Ln(ASSET)t (?) CFOt (-) DEBTt (+) TIEt (-) # Clean # Modified N Likelihood χ 2 % concordant pairs (+ tied)

(2)

Coefficient (Wald Chi-square statistic) (1) (2) -1.164 -0.963 (202.62)a (195.56)a −− 0.305 a (13.41) −− -0.069 (0.76) -0.716 -0.752 (90.71)a (99.47)a 0.130 0.122 (121.21)a (108.84)a -1.566 -1.654 (183.74)a (199.62)a 0.045 0.046 (63.59)a (60.67)a 0.004 0.003 (56.03)a (49.45)a 5,806 5,749 3,112 3,093 8,918 8,842 a 403.906 394.151a 63.4%

63.6%

Wald Chi-square statistics are in parentheses. a, b, c indicates one-tailed significance at the 0.01, 0.05, 0.10 levels, respectively where a direction is predicted. If we do not have a predicted direction the test is a two-tailed test. AUD OPIN is an indicator variable set to one if the firm receives a modified audit opinion and zero otherwise. (ABS BTD)dec is the decile ranking (from 0 to 1) of the absolute value of the total book-tax differences as defined above, scaled by average total assets. (ABS DEXP)dec similarly, is the decile ranking (from 0 to 1) of the absolute value of the deferred tax expense (temporary book-tax differences only), scaled by average total assets. WCACCdec is the decile ranking (from 0 to 1) of the working capital accruals of the firm calculated as the changes in working capital accounting from the statement of cash flows (#302, #303, #304, #305 and #307), scaled by average total assets. The variable CFO is the firm’s cash flows from operations (data item # 308), scaled by average total assets, and TIE is the firm’s operating earnings divided by interest expense (item # 178/#15). The other variables are as defined above in Table 2.

41

TABLE 7: Logit Regression of Auditor Change on Book-Tax Differences and Control Variables

∆AUDITORt = α + β1 (ABS BTD)dec t + β2 WCACCdec t + β3 ABS(∆ASSET)t + β4 ∆CFOt + β5 ∆FINt + β6 ∆TIE t + ε Independent Variables (Expected Sign) Intercept (?)

(ABS BTD)dect (+) (ABS DEXP)dect (+) WCACCdect (+) ABS(∆ ASSET )t (?)

∆ CFOt (-) ∆ FΙΝt (+) ∆ TIEt (-) # No change # Change N Likelihood χ 2 % concordant pairs (+ tied)

(3)

Coefficient (Wald Chi-square statistic) (1) (2) -1.718 -2.069 (704.29)a (559.00)a −− 0.548 a (22.53) −− -0.279 (5.40)b -0.066 -0.131 (0.36) (1.40) -0.026 -0.022 c (2.42) (3.37)b 0.021 -0.009 (0.16) (0.03) 0.013 0.022 (0.61) (1.73)c -0.000 -0.000 (0.29) (0.48) 6,202 6,138 953 946 7,155 7,084 31.28a 13.409b 57.0%

55.8%

Wald Chi-square statistics are in parentheses. a, b, c indicates one-tailed significance at the 0.01, 0.05, 0.10 levels, respectively where a direction is predicted. If we do not have a predicted direction the test is a two-tailed test. ∆ Auditor is an indicator variable set to one if the firm experienced a change in auditors from year t-1 to year t, zero otherwise. (ABS BTD)dec is the decile ranking (from 0 to 1) of the absolute value of the total book-tax differences as defined above, scaled by average total assets. (ABS DEXP)dec similarly, is the decile ranking (from 0 to 1) of the absolute value of the deferred tax expense (temporary book-tax differences only), scaled by average total assets. WCACCdec is the decile ranking (from 0 to 1) of the signed working capital accruals of the firm calculated as the changes in working capital accounting from the statement of cash flows (#302, #303, #304, #305 and #307), scaled by average total assets. ABS(∆ASSETS) is the absolute value of the change in the firm’s assets in year t relative to the average of the last 3 previous years. If three previous years are not available we use the number of years for which data can be obtained (abs(Assetst-Assetst-3)/Assetst-3) (data item # 6). The variable CFO is the firm’s cash flows from operations (data item # 308) (scaled by average total assets), and TIE is the firm’s operating earnings divided by interest expense (item # 178/#15). The other variables are as defined above in Table 2.

42