brooks et al - 2011 - when does audit quality start to decline in firm audit tenure - an international analysis [mafr]

But first, regulators need to have some idea of the “turning point” of audit quality during audit firm tenure, and whether this turning point varies across countries in different legal l

When Does Audit Quality Start to Decline in Firm Audit Tenure?

– An International Analysis

Li (Lily) Z Brooks Department of Accounting E.J Ourso College of Business Louisiana State University zlily424@lsu.edu

C.S Agnes Cheng Department of Accounting E.J Ourso College of Business Louisiana State University acheng@lsu.edu

Joseph Johnston Department of Accounting College of Business City University of Hong Kong jjohnsto@cityu.edu.hk

Kenneth J Reichelt Department of Accounting E.J Ourso College of Business Louisiana State University reichelt@lsu.edu

September 25, 2011 Preliminary Draft

* We gratefully acknowledge the comments and suggestions of the workshop participants at

Penn State University

through a learning effect in early years but in later years it is likely to decrease due to a bonding

effect Using a quadratic model, we estimate the year when audit quality, measured by earnings

quality, starts to decline during an audit firm’s tenure We propose that the bonding effect should

be weaker in countries with stricter legal liability regimes, which implies that the year that audit quality begins to decline should be later in countries with stricter legal liability regimes We find that it takes 14 to 16 years for countries with stronger legal liability regimes while it only takes 4

to 10 years for countries with weaker legal liability regimes for audit quality to decline Our results are strong regardless of whether we measure legal liability from the perspective of legal origin (common or code law) or litigation risk Our findings have implications across the world for the current debate on the mandatory requirement of audit firm rotation across the world

1 Introduction

The purpose of our study is to examine whether the strength of the client’s country’s

legal liability regime affects the decline in audit quality during an audit firm’s tenure

Specifically, we examine whether a country’s strength on legal liability prolongs the point of

time when the audit quality starts to decline Regulators are concerned that bonding (or lack of

independence) between auditors and managers (hereafter, the ‘bonding effect’) will decrease

audit quality in later years; however, firms’ and auditors’ concerns of legal liability should

mitigate such a negative effect Accordingly, we propose that the legal liability regime should

affect the turning point when audit quality starts to decline The empirical evidence of the

‘point’ should help regulators to determine the necessity of mandatory audit firm rotation The

topic of mandatory audit firm rotation is currently being discussed by regulators such as the

European Commission and the Public Company Accounting Oversight Board (hereafter

PCAOB), as a potential solution to improve audit quality If regulators adopt mandatory auditor

rotation, they need to decide on the appropriate term Countries that already have audit firm

rotation in place all require audit firm rotation within 10 years.1 Recently, the US PCAOB is seeking comment on the proposal to limit audit firm tenure to 10 years However, it is unclear

whether a 10-year term limit should apply to all countries Our paper provides a plausible

explanation for varying the term limits across countries by showing that the legal liability regime

mitigates the ‘bonding effect’ and thus prolongs the turning point when audit quality starts to

deteriorate

The debate about mandatory audit firm rotation among regulators and professional bodies

has been long-standing and on-going (PricewaterhouseCoopers 2002; The Conference Board

1

Mandatory audit firm rotation exists in Brazil (5 years), India (4 years for banks, insurance companies, and the public sector), Italy (9 years), and South Korea (6 years)

2005; European Commission 2010; Public Company Accounting Oversight Board (PCAOB)

2011) Proponents argue that long audit firm tenure induces the economic and social bonding

between the auditor and the client, which can impair auditor independence and judgment

Accordingly, they support using mandatory audit firm rotation to mitigate the negative bonding

effect on audit quality Opponents argue that mandatory audit firm rotation is very costly to both

clients and investors In addition to the high out-of-pocket switching costs, newly hired audit

firms will be less effective at detecting errors due to their lack of client-specific knowledge If

audit quality does not improve after the initial stage, these costs are a deadweight loss to society

Many US studies tend to find evidence that audit quality is weaker when audit firm tenure is

shorter (e.g., Johnson et al 2002; Chung and Kallapur 2003; Myers et al 2003; Carcello and

Nagy 2004; Gul et al 2007; Stanley and DeZoort 2007) However, recent studies using US

firms (Davis et al 2009; Brooks 2011) find that long audit firm tenure decreases audit quality

There is also international evidence that long audit firm tenure decreases audit quality (Chi and

Huang 2005) Our study extends the literature by suggesting that the legal liability regime

should affect the adverse effects from bonding (or lack of independence) between auditors and

clients

Due to the recent financial meltdown and continuing poor audit quality, European and US

regulators have renewed their interest in mandatory audit firm rotation The European

Commission (EC) issued a Green paper (European Commission 2010) asking the European

Union to open up debate about mandatory audit firm rotation, among other issues, in order to

improve the audit quality The EC believes that improvements to audit quality will help stabilize

the financial system Their concerns stem from the fact that many banks in the United Kingdom

failed shortly after receiving an unqualified audit opinion (United Kingdom House of Commons

Treasury Committee 2009, p.76) The US PCAOB recently issued Concept Release No

2011-006 (2011) to solicit public opinion on means to improve audit quality, including mandatory

audit firm rotation The PCAOB’s concern for audit quality arises from the fact that even though

the PCAOB inspection process has been in place since 2004, they still find instances where

auditors fail to approach the audit with the required level of independence and professional

skepticism They note examples in US and foreign jurisdictions where auditors accepted

management’s perspective without challenging it or developing an independent opinion The

concept release notes that the audit firm tenure of the largest 100 companies averages 21 years,

and asks constituents to comment on a 10 year rotation requirement Should it be 5 years, 10

years or 15 years? If it likely that the adverse bonding effect starts at the 5th year (the turning point), then a requirement of 10 year rotation is too long and will not be able to protect investors

in time If the turning point is at the 15th year, then a requirement of 10 year rotation is too short and will generate a deadweight loss to society In choosing the number of years for mandatory

audit tenure rotation, regulators need to weigh these two types of costs But first, regulators need

to have some idea of the “turning point” of audit quality during audit firm tenure, and whether

this turning point varies across countries in different legal liability regimes

We propose that a stricter legal liability regime delays the deterioration of audit quality

during the audit firm’s tenure A stricter legal liability regime weakens the client-auditor bond,

through the threat of higher litigation costs, thus delaying the point in time when audit quality

begins to deteriorate We measure the strength of the legal liability regime by using legal origin

– common law vs code law (La Porta et al 1997, 1998), and by using litigation risk (Wingate

1997) We predict that audit quality will deteriorate later in common law countries than in code

law countries, and later in countries with higher litigation risk relative to countries with lower

litigation risk

We measure audit quality using earnings quality – the absolute value of abnormal

accruals Our main sample consists of 104,758 firm-year observations from 26 countries

covering the 14 year period of 1996 to 2009 We find that the turning point (when audit quality

starts to deteriorate) is longer for common law countries (16 years) than for code law countries (7

years) We also find similar results when we categorize countries into high and low litigation

risk To sum, we find that legal liability affects the turning point in a predicted fashion: legal

liability mitigates the bonding effect, hence, longer auditor tenure can be allowed Our results

are robust to an alternative measure of abnormal accruals (Francis and Wang 2008)

Our paper extends at least two streams of literature We contribute to the international

auditing literature by showing that the turning point in audit quality is affected by a stronger

legal liability regime Francis and Wang (2008) and Choi et al (2008) show that stronger legal

liability regimes are associated with higher audit quality; however, whether this audit quality

varies with audit firm tenure is not apparent in these studies We find that audit quality starts to

decline later in stronger legal liability regimes, suggesting that auditor-client bonding is less

severe in countries with stronger legal liability laws where litigation risk is higher We also

extend the one-country studies on the effect of audit tenure on audit quality, by Chi and Huang

(2005), Davis et al (2009) and Brooks (2011), to an international setting

Our paper provides evidence to regulators for their decision-making on the issue of

mandatory audit firm rotation Our evidence suggests that the extent of litigation risk creates

economic incentives for auditors to deliver high levels of audit quality for a longer period of

time For countries with a stronger (weaker) legal liability regime, the requirement of mandatory

rotation should be longer (shorter); this will avoid the deadweight loss to society while still

preserving investor protection

A limitation of our study is that we assume that audit quality is a quadratic function of

audit firm tenure We use this form for the sake of convenience in estimating the turning point

This form is at best an approximation of what theory suggests It assumes that audit quality

increases first and then decreases in a similar degree However, the correct function could be

that audit quality increases slow (or fast) but it drops fast (or slow) in an earlier (or later) stage

Future research can address whether a more accurate functional form exists

The remainder of the paper is organized as follows The next section presents our

literature review and hypotheses The third section describes our empirical model The fourth

section describes our results and the final section concludes the paper

2 Literature Review and Hypothesis development

In this section, we first review the prior literature to learn how audit quality varies over

the audit firm’s tenure We next examine how the legal liability regime affects audit quality

Finally, we present a hypothesis about how the extent of legal liability regime affects how audit

quality varies over the audit firm’s tenure

2.1 Effect of Audit Tenure on Audit Quality

The prior literature, as well as the concerns of regulators and constituents, suggests that

two countervailing forces jointly determine the relation between auditor tenure and audit quality:

a learning effect and a bonding effect The learning effect occurs in the initial years of the

auditor’s tenure, where audit quality increases as the auditor gains a better understanding of the

client’s financial reporting system, business, industry and internal controls This effect is

consistent with the “learning curve” effect (Yelle 1979; Chen and Manes 1985) where new

information is acquired after each successive audit engagement at a diminishing rate, until a

maximum amount of information is acquired Thus, the auditor’s ability to detect material errors

in accrual estimates is sharpest in early years The learning effect results in a positive relation

between audit quality and audit tenure that gradually reaches a maximum level in later years

Two streams of audit firm tenure literature support the learning effect The first stream

examines audit failures and finds that audit failures are associated with shorter audit firm tenure

For instance, Geiger and Raghunandan (2002) find that firms entering bankruptcy are less likely

to have been issued a going concern audit opinion from audit firms with shorter tenure Carcello

and Nagy (2004) find that fraudulent financial reporting is more likely when audit firm tenure is

three years or less Stanley and DeZoort (2007) find that financial restatements are negatively

related to audit firm tenure A limitation with this literature stream is that audit failures are

infrequent; thus it is difficult to infer whether they are representative of the larger population

A second stream of literature that supports the learning effect finds that earnings quality

improves with longer audit firm tenure Chung and Kallapur (2003) and Myers et al (2003) find

that discretionary accruals are negatively related to auditor tenure Similarly, Johnson et al

(2002), and Gul et al (2009) find evidence of higher discretionary accruals in the early years of

the audit firm’s tenure A related study by Mansi et al (2004) find a negative relation between

cost of debt and audit firm tenure suggesting that perceived audit quality increases with audit

firm tenure

In addition to the learning effect, prior literature suggests that a second countervailing

force dominates in later years of the audit firm’s tenure We coin this effect – the bonding effect

We posit that the bonding effect consists of two complementary factors: an over-familiarity effect

and a lack of independence effect The over-familiarity effect reduces auditor skepticism and

diminishes audit quality as the auditor become more complacent and over-familiar with the

client For instance, Shockley (1981) states that “complacency, lack of innovation, less rigorous

audit procedures and a developed confidence in the client may arise after a long association with

the client” In turn, the auditor will have a tendency to anticipate results, such as the condition of

the client’s system and procedures (Hoyle 1978; Carey and Simnett 2006), and become less

vigilant to more subtle anomalies (Arrunada and Paz-Ares 1997) Consequently, the ability of

the auditor to detect either intentional or unintentional errors diminishes with the length of the

auditor’s tenure Therefore, the over-familiarity effect results in a negative relation between audit

quality and audit firm tenure

The second factor of the bonding effect is the lack of independence effect Independence

is the propensity of the auditor to correct material errors that they detected in the course of their

audit engagement A lack of independence occurs when the auditor will not correct a material

error Not correcting a material error may result when auditors develop an unwarranted trust of

management’s accounting judgments (King 2002) The audit partner acts as a ‘relationship

manager’ by taking steps to ensure that the client remains “happy”, in order to avoid being

replaced with a more accommodating partner (McCracken et al 2008) The auditor may also

socially bond with the client, and acquiesce to their client’s preferred position (Bamber and Iyer

2007) Additionally, the next partner-in-charge will try to avoid the economic loss as well as the

stigma of losing a long-standing client (Doty 2011) Mautz and Sharaf (1961, p 231) state that

the auditor “must be aware of the various pressures, some obvious some subtle, which tend to

influence [their] attitude and thereby erode slowly but surely [their] independence.” In many

cases “the greatest threat to [their] independence is a slow, gradual, almost casual erosion of

[their] honest disinterestedness” (Mautz and Sharaf 1961, p 208) The close economic and

social bond that develops over time suggests that the auditor’s judgment will gradually impair

over time Thus, the lack of independence effect results in a negative relation between audit

quality and audit firm tenure

Two streams of empirical literature provide support for the bonding effect – the first

stream examines perceived audit quality, and the second stream examines actual earnings

quality The first stream suggests that auditors and investors perceive a bonding effect which

impairs audit quality from extended audit firm tenure Kealey et al (2007) find that the

successor auditor’s audit fee increases with the length of the prior auditor’s tenure, suggesting

that the successor auditor views longer tenure of the predecessor auditor as lower audit quality

Dao et al (2008) find that shareholders are more likely to vote against auditor reappointments

when the auditor’s tenure is longer, suggesting that shareholders view longer tenure as lower

audit quality Boone et al (2008) finds that cost of common equity capital decreases in early

years and increases in later years Their study implies that perceived audit quality increases in

early years due to a learning effect, and decreases in later years due to a bonding effect While

perceptions might provide insights into the appearance of audit quality, more germane to our

study is the examination of actual audit quality, such as earnings quality

Several recent studies (Chi and Huang 2005; Davis et al 2009; Brooks 2011) find that

earnings quality increases in the early years of audit firm tenure, and later deteriorates These

studies support the view that a learning effect dominates in early years and a bonding effect

dominates in later years Using Taiwanese data, Chi and Huang (2005) find that discretionary

accruals decrease in the first five years and later increase Using US data, Davis et al (2009)

finds that discretionary accruals to meet or beat analysts’ forecasts decrease in the first 15 years

but later increase, suggesting that earnings quality increases in early years but decreases in later

years However, they find their results only hold for the pre-Sarbanes Oxley (SOX) era (i.e.,

1988-2001), thus their results may not generalize to the post-SOX era Also using US data,

Brooks (2011) recently reports that accruals quality increases in early years of the audit firm’s

tenure and decreases in later years for both the pre-SOX and the post-SOX eras Brooks’ (2011)

study extends Davis et al.’s (2009) study from two perspectives: first, it uses accruals quality

(Dechow and Dichev 2002) as a measure of earnings quality which likely not only captures

management bias (as in Davis et al.) but also unintentional errors; second, it examines a larger

cross-section of firms, well beyond firms meeting or beating analysts’ forecasts

Based on the previous discussion, we expect that the learning effect and the bonding

effect (over-familiarity effect and lack of independence effect) act as countervailing forces

resulting in a non-monotonic relation between audit quality and audit firm tenure We express

this relation in equation (1) as follows:

where AQ is audit quality and T is auditor tenure Recall, that the learning effect results

in a positive relation between audit firm tenure and audit quality in early years, while the

bonding effect dominates in later years and results in a negative relation between audit firm

tenure and audit quality in later years We assume that their combined effect can be

approximated by equation (1) as a concave function, where α 1 > 0 and α 2 <0 (marginally

decreasing as T increases), and the turning point is equal to –α 1/ 2α 2.2 The effectiveness of this equation in approximating empirical data relies on the validity of our assumptions, i.e the

marginal increment in audit quality due to learning is higher (lower) than the marginal decrement

in audit quality due to bonding in earlier (later) years

While prior literature (Chi and Huang 2005; Davis et al 2009; Brooks 2011) supports the

non-monotonic relation between audit quality and audit firm tenure, our study aims to understand

how this relation varies across countries with differing legal liability regimes We now turn our

attention to a discussion of how the legal liability regime affects audit quality

2.2 Legal Liability Regime and Audit Quality

Investor protection laws consist of well-functioning legal institutions and securities

regulations that serve to protect investors against exploitation by managers and controlling

shareholders Prior literature finds that cross-country differences exist in investor protection

laws (La Porta et al 1997, 1998) La Porta et al (2006) conjectures that investor protection laws

work to benefit investors by mandated disclosure rules (e.g., a prospectus with earnings and

ownership information), liability rules that facilitate private enforcement (e.g., liability standards

facing issuers, auditors and intermediaries when investors seek damages), and stricter

enforcement mechanisms (e.g., a Securities and Exchange Commission that can clarify legal

obligations and impose penalties to issuers and auditors) La Porta et al (2006) reports that

mandated disclosure rules and liability rules benefit investors in the form of greater stock market

development As well, Hail and Leuz (2006) report that investors benefit from all three

categories of investor protection laws, in the form of lower cost of equity The lower cost of

equity is partly contributed by higher quality financial reporting (Ball et al 2000; Hung 2000;

Leuz et al 2003; Bushman and Piotroski 2006), by serving to reduce information asymmetry

between investors and management

One contributing factor to higher quality financial reporting is higher audit quality

Auditing prevents investors from receiving misleading financial information from management

Studies that compare audit quality across different countries argue that the combination of strong

liability rules and strict enforcement mechanisms creates incentives for auditors to deliver high

audit quality from the greater risk of litigation (e.g., Choi et al 2008; Francis and Wang 2008)

We refer to this combined effect as strong legal liability

Strong legal liability means that auditors are more likely to be sued for negligence, public

enforcers are more powerful to punish them for misconduct, client misreporting is more likely to

be detected (Francis and Wang 2008), litigation costs are higher (Choi et al 2008), and

reputations seem to matter more (Cahan et al 2009) Litigation risk is so important that US

auditors, in the early 1990s, successfully lobbied to limit legal liability through the passage of the

Private Securities Reform Act of 1995 Litigation risk not only affects the reputation of the audit

firm, but more importantly, it reduces the personal wealth of the “deep-pocketed” audit firm.3 For instance, Lennox (1999) finds that large auditors are more prone to litigation, but they do

not suffer a reduction in demand from litigation, suggesting that “deep pockets” seem to matter

more than the audit firm’s loss of reputation The upshot is that strong legal liability creates an

incentive to deliver higher audit quality

Francis and Wang (2008) and Choi et al (2008) provide evidence that strong legal

liability regimes are associated with higher audit quality Francis and Wang (2008) find that

signed discretionary accruals are smaller for Big N auditors when the legal liability regime is

stronger Choi et al (2008) find that audit fees are higher when the legal liability regime is

stronger While these two studies find that audit quality is higher when legal liability regimes

3

Auditors have “deep pockets” and investors will frequently sue auditors to recover damages from auditor

misconduct For instance, an estimated 4,000 malpractice suits are filed annually against U.S accounting firms (Goldwasser et al 2008; Casterella et al 2010)

are stronger, they do not examine whether the legal liability regime affects audit quality over the

audit firm’s tenure

Several studies examine how audit firm tenure impacts audit quality within a specific

country, suggesting that cross-country differences exist Using US firms, Davis et al (2009) and

Brooks (2011) find that audit quality and audit firm tenure have an inverted U function – i.e.,

audit quality increases in early years of the audit firm’s tenure and later it deteriorates The

Brooks (2011) study finds that the turning point, when audit quality begins to deteriorate, is

longer in the post-SOX era (18 years) than in the pre-SOX era (14 years) Arguably the

Post-SOX era is a stronger legal liability regime because of more onerous auditing standards (e.g.,

PCAOB inspections of audit firms, and auditor reporting on the effectiveness of internal controls

over financial reporting) and Title XI of the SOX imposes further penalties and expands the

parties to whom auditors have a duty of care, suggesting that stronger legal liability regimes have

a longer turning point Using Taiwanese firms (1998-2001), Chi and Huang (2005) also report

an inverted U function of audit quality and audit firm tenure, and they report a shorter point in

time when audit quality starts to deteriorate (5 years), suggesting that weaker legal liability

regimes have a shorter turning point However, using privately-held Belgium firms, Knechel and

Vanstraelen (2007) fail to find evidence from going-concern opinions that audit quality

deteriorates over the audit firm’s tenure Their study may be limited by the power of their

sample which is limited to financially stressed firms Given that these studies provide some

evidence that the turning point, when audit quality begins to deteriorate, is longer in countries

with a stronger legal liability regime, we now turn our discussion to the hypothesis development

2.3 Hypothesis development

An implication of the foregoing discussion is that a strong legal liability regime weakens

the bonding effect between the client and the auditor, and the resulting audit quality improves In

a strong legal liability regime, audit failures impose high litigation costs and high reputation

damage To minimize these costs, the auditor will act with greater independence and objectivity

by employing more professional skepticism,4 and by employing practices to reduce

over-familiarity of the engagement (e.g., partner and staff rotation requirements) Consequently, the

auditor is better at detecting errors in management judgment, forming their own opinion of

management’s accrual estimates, and challenging management’s estimates

In an ideal world, a strong legal liability regime should completely eliminate the bonding

effect However, even in the US where litigation risk is high (e.g., Khurana and Raman 2004),

we still observe that audit quality declines in later years of the audit firm’s tenure (Davis et al

2009; Brooks 2011) Thus, from a practical point of view, a strong legal liability regime can

only weaken the bonding effect, but it cannot completely eliminate it

From our earlier discussion, we posited that the bonding effect results in a negative

relation between audit quality and audit firm tenure In contrast, the learning effect dictates a

positive relation between audit quality and audit firm tenure When the bonding effect prevails

over the learning effect in later years, audit quality will begin to deteriorate A stronger bonding

effect will deteriorate audit quality sooner, and the effect will also result in lower audit quality

Thus, in a strong legal liability regime where the bonding effect is weaker, audit quality will be

higher, and it should deteriorate later We formalize the hypothesis as follows:

4

For instance, auditors became less tolerant of accruals-based earnings management (Cohen et al 2008) in the SOX era

post-The point in time when audit quality starts to deteriorate during an audit firm’s tenure is longer in countries with a strong legal liability regime, than in countries with a weak legal liability regime.

3 Methodology

In this section, we discuss the methodology employed to test our hypothesis We use the

magnitude of abnormal accruals as a proxy for audit quality, consistent with prior research

quality (Johnson et al 2002; Ashbaugh et al 2003; Balsam et al 2003; Myers et al 2003) To

measure abnormal accruals, we use two different accrual estimation models The first model is

the performance-adjusted accruals model following Kothari et al (2005) and Jones (1991) The

second model is suggested by Francis and Wang (2008) for use with international data where

cross-sectional estimation may be limited

To measure performance-adjusted abnormal accruals, we first estimate the following

equation cross-sectionally by 2-digit SIC code5 within each country and within each year:

,

,    ,  ,,  ,,  ,,  2

Where:

!"## = Total accruals, measured as net income before extraordinary items minus

operating cash flows ($%& -('"(#) -*+,'# )6 for firm i in year t in country k;

-./0 = Change in revenues (1234 - 1234 ) for firm i from year t-1 to year t in

country k;

66/ = Gross value of property, plant, and equipment (77489 for firm i in year

t in country k;

.'" = Return on assets ($% ) for firm i in year t in country k;

"8"11491 = Average assets, measured as ((29 +29 )/2), for firm i in year t in

 = Error term

The residual from equation (2) is abnormal accruals We negate the absolute value of the

residual so that the measure (denoted as AQ1) is increasing in audit quality

Our second abnormal accruals estimation model, addresses the concern that international

data have a small number of industry observations which may render unreliable results in a

cross-sectional Jones (1991) model Following Francis and Wang (2008), we use a linear

expectations model to predict firm-level accruals, and to estimate abnormal accruals

Specifically, we model accruals as follows:

6:4;"&&  <234 =>#"<234 ? @ 66/ =66/,47 ? 3

"%"&&  6:4;"&&"11491@ !"## 4

Where:

<234 = Sales for firm i from year t-1 to year t in country k;

>#" = Change in non-cash working capital, defined as ∆(total current assets (act) –

cash and short term investments (che)] – ∆[total current liabilities (lct)-total amount of debt in current liabilities (dlt) – taxes payable (txp) – depreciation (dp)] for firm i in year t in country k;

66/ = Gross value of property, plant, and equipment (77489 for firm i at year t in

country k;

,47 = Depreciation expense (dp) for firm i in year t in country k;

"11491 = Total assets (at) for firm i at year t in country k

!"## = Total accruals, measured as net income before extraordinary items plus

depreciation and amortization minus operating cash flows ($%& ('"(#) -*+,'# ) for firm i in year t in country k; 7

-"%"&& = Abnormal accruals for firm i in year t in country k;

7

When operating cash flow is not readily available, we calculate total accruals using the balance sheet method

This method of calculating abnormal accruals does not rely on estimation methods but

rather assumes that changes to the ratio of working capital to sales and the ratio of depreciation

expense to property, plant and equipment should be roughly constant across time for a firm As

with our first measure, high quality audits should yield smaller magnitudes of abnormal accruals

We therefore negate the absolute value of "%"&& as our second measure of audit quality

(AQ2)

To test our hypothesis that the turning point of audit quality is longer for firms in

countries with a stronger legal liability regime, we estimate the following equation with OLS:

"C  D D!4EF:4  D!4EF:4  D'#)  DG3H8I0 

"C = Audit quality (AQ1 or AQ2) for firm i in year t in country k;

!4EF:4 = the number of consecutive years that firm i has retained the auditor

since 1974 in year t in country k;

!4EF:4 = the square of !4EF:4 ;

'#) = cash flow from operations scaled by average total assets (H2E&\ /

29  29 /2 for firm i in year t in country k;

3H8I0 = the natural log of market value of equity, where market value of

equity is calculated as the closing price times the number of shares

outstanding at the end of the fiscal year for firm i in year t in country

k; I9HK = the market to book ratio, calculated as the market value of equity

divided by the book value of common equity (ceq) for firm i in year t

in country k;

"84 = the natural log of the number of years since firm i first appeared in

Compustat since 1950 in year t in country k;

+(,N = industry sales growth, calculated as ( ( ∑ 1234^ /

-1) by SIC-2 industry groups for firm i at year t in country k;

N:HP9Q = growth in gross domestic product adjusted for inflation, calculated as

(8;7:1 @ 8;7:1 /8;7:1 , for firm i in year t in country

k;

#HFE9:S,FTT$41 = 1 for country k, and 0 otherwise Twenty six countries are included in

the sample Please refer to Table 1 for details of the countries

+E;F19:S,FTT$41 = 1 for industry w, and 0 otherwise Nine industry (two digit SIC)

groups are formed as follows: 1-9 (Agriculture/Forest), 10-17 (Mining/Construction), 20-39 (Manufacturing), 40-47

(Transportation), 48 (Communication), 49 (Utilities), 50-51 (Wholesale), 52-59 (Retail), and 70-99 (Other)

X42: ,FTT$41 1 for year x, and 0 otherwise The sample includes fourteen years

Following prior literature, we control for firm size, operating cash flow, industry growth,

firm growth opportunities, firm age, and general economic condition We control for firm size

(3H8I0 since accruals quality increases with firm size because of greater stability and diversification of the portfolio of activities (Dechow and Dichev 2002) We control for

operating cash flow ('#) ) because firms with higher operating cash flow are more likely to perform better (Frankel et al 2002) Industry growth +(,N is included because industry growth is positively related to accruals (DeFond and Jiambalvo 1994) We control for market to

book ratio aI9HK b because growth firms have stronger incentives to manipulate earnings (Frankel et al 2002) Firm age ("84 ) is included because accruals differ with changes in the firm’s life cycle (Anthony and Ramesh 1992; Dechow et al 2001) We control for general

economic conditions (N:HP9Q ) because managers have stronger incentives to manipulate earnings during economic downturns.8

8 N:HP9Q only varies across years and country

As discussed previously, we expect that the coefficient on Tenure (β 1) is positive, and the

coefficient on Tenure 2 (β 2) is negative for both high and low legal liability regime countries,

indicating that audit quality increases in earlier years of audit firm tenure and decreases in later

years of audit firm tenure To test our main hypothesis, we calculate the turning point when AQ

reaches its maximum as follows:

Turning Point = - cd

ce (6)

Since the turning point is determined by the negative ratio of D to D, the smaller the

coefficient on Tenure 2 , the higher the turning point, ceteris paribus We predict that the turning

point in audit quality for countries with a strong legal liability regime is longer than the turning

point for countries with a weak legal liability regime To test our prediction, we estimate

equation (5) by legal regime and test for the difference in the turning point (equation 6)

We use two measures of the strength of legal regime First, we measure strong legal

liability regime countries as common law countries and weak legal liability regime countries as

code law countries, consistent with prior literature (e.g., Ball et al 2000; Francis et al 2003;

Francis and Wang 2008) Prior literature indicates that higher litigation risk in common law

countries is a key incentive for auditors to provide high quality audits (Francis and Wang 2008)

Our second measure of the strength of the legal liability regime uses the litigation risk rating

(Wingate 1997; Choi et al 2008; Francis and Wang 2008) Wingate (1997) develops a litigation

risk rating from an international insurance underwriter for one of the Big 6 audit firms The

rating (LITIGATE) measures the risk and cost of performing audit services in a particular

country The rating takes on values of one to 15, with higher values assigned to countries with

higher litigation risk High litigation risk countries (High-LIT) are assigned values of ten or

greater and low litigation risk countries (Low-LIT) are assigned values of less than ten

4 Empirical Results

4.1 Sample Selection and Descriptive Statistics

We begin our sample selection process by including all financial data obtained from

Compustat Global and Compustat North America for the period of 1996 to 2009, consisting of

379,977 firm-year observations We eliminate 150,487 observations with missing dependent and

independent variables, 33,960 observations with negative book values, 36,433 firm-year

observations in financial institutions (Standard Industrial Classification (SIC) 6000-6999), and

48,346 firm-year observations in the year of an auditor change We require a minimum of 50

firm-year observations for each country, which excludes 527 firm-year observations Finally, to

omit extreme outliers, we exclude 5,466 firm-year observations from the 1st and 99th percentile

of all continuous variables and firm-year observations with a studentized residual greater than 3

from the estimation of equation (5) using AQ1 After these screens, our final sample has 104,758

observations for the period of 1996 to 2009 Table 1 summarizes the sample selection process

[INSERT TABLE1 HERE]

Table 2 reports descriptive statistics by country, beginning with common law countries (a

total of 10 countries) and followed by code law countries (a total of 16 countries), sorted in order

of litigation risk (LITIGATE) The mean of LITIGATE for the ten common law countries is

7.000, and for the 16 code law countries is 4.938, consistent with prior literature that suggests

that litigation risk is higher in common law countries (Francis and Wang 2008) Comparing the

mean values of common law countries to code law countries for the other variables, we find the

following Mean accruals quality (AQ1) is lower for common law countries (-0.071) than for

code law countries (-0.044) The lower accruals quality of common law countries is likely due to

higher growth and risk characteristics of common law countries that increase the dispersion of

accruals Our sample descriptive statistics support this argument Specifically, common law

countries having lower mean operating cash flows (0.044) than code law countries (0.065), a

higher market to book ratio (2.464 vs 2.297), and higher industry growth (0.087 vs -0.002)

Furthermore, firms in common law countries tend to be smaller than those in code law countries

(4.378 vs 5.290) These findings support the need to control for these variables in our

multivariate analysis Mean audit firm tenure is slightly longer for common law countries

(6.944) than for code law countries (6.883) Mean GDP growth is lower for common law

countries (1.027) than for code law countries (1.053), suggesting that common law economies

have slower economic growth - likely due to the recent recession

[INSERT TABLE 2 HERE]

Table 3 reports the correlation coefficients among the variables used in our multivariate

analysis Pearson correlation coefficients are reported in the upper diagonal and Spearman

coefficients are reported in the lower diagonal For accruals quality and audit firm tenure, we

find that AQ1 and Tenure are positively correlated: the Pearson coefficient is 0.152 and the

Spearman coefficient is 0.127, consistent with prior literature We also find that AQ1 and

Tenure 2 are positively correlated: the Pearson coefficient is 0.152 and the Spearman coefficient

is 0.108, suggesting that the relation is non-linear For accruals quality and litigation risk, we

find that AQ1 and LITIGATE are positively correlated: the Pearson coefficient is 0.252 and the

Spearman coefficient is 0.255, suggesting that accruals quality is higher in countries with higher

litigation risk For accruals quality and legal origin, we find that AQ1 and Civil_com are

positively correlated: the Pearson coefficient is 0.290 and the Spearman coefficient is 0.257,

suggesting that accruals quality is lower in common law countries However, this relation may

change after we add control variables in our multivariate analysis