1 Auditing Section Doctoral Consortium January 2006 Jere R. Francis Curators Professor KPMG Research Professor University of Missouri, USA [email protected].

1

Auditing SectionDoctoral Consortium

January 2006

Jere R. Francis Curators’ Professor

KPMG Research Professor

University of Missouri, [email protected]

2

Data Innovation in Empirical Audit Research

Do what you enjoy, and find the right data for the RQ Innovation through new data

Use new data to answer “old” questions better or to answer “new” questions that could not previously be investigated public data – new U.S. audit fee data or internal control reports

but risky because of competition private data avoids competition

but can be risky because it’s costly and the outcome is unknown Novel use of existing public data

Creatively use existing data in new ways link financial statement properties (e.g., abnormal accruals) with

auditor characteristics Becker et al. (CAR 1998) & Francis et al. (AJPT 1999)

office-level analysis of Big 4 accounting firms Reynolds &Francis (JAE 2000)

3

Audit Research:Intersection of practice, theory &

research design choices

Practice drives research questions (RQ) Some examples (next slides)

Theory informs/frames RQ Economics, agency & signaling theory, psychology

Research design – many choices Modeling, experimental, & archival

Inherent limits of RQs answerable by a single approach/method Value of triangulation/complementarity

4

Examples of Practice-Driven Empirical

Audit Research 1970s – Congressional investigations

Allegations of large-firm cartel pricing & predatory pricing Motivated first studies of audit pricing

1980s – effects of deregulation Allegations of low-balling and poor audit quality Motivated studies of initial engagement pricing

1990s – legal liability and audit litigation crisis (S&L, PSLRA) Motivated risk-screening studies, & effects of PSLRA

1990s – growth of industry specialization and non-audit services Motivated study of industry expertise and non-audit services

2000s – New regulations/institutions The effects of PCAOB and SOX on audit practice/quality

5

Triangulation - Low Balling Theory -modeling pricing incentives & consequences on

auditor independence DeAngelo JAE 1981; Dye JAE 1991

Experimental evidence Economics – induce LB in the lab & effects on audit quality

Schatzburg (TAR 1990); Schatzburg & Sevcik (CAR 1994); Dopuch & King (JAAF 1996)

Psychology – effects of “fee pressure” on the quality of auditor judgments Houston (AJPT 1999)

Empirical evidence Initial engagement discounts

Simon & Francis (TAR 1988); Craswell & Francis (TAR 1999) Consequences of LB on audit hours/audit quality

Deis and Giroux (JAPP 1996), sample of school district audits

6

Units of Analysis in Audit Research(“red” denotes prime empirical areas)

Audit testing & evidence-gathering procedures Auditor judgments about testing & evidence

Auditors apply/interpret testing procedures As individuals (individual judgments) In teams (group judgments)

Accounting firms Auditors work in firms

Observable “audit firm” outcomes are audit reports & audited financial statements (and who audits whom)

Audit industry and audit markets Audit firms operate & compete in audit markets

Regulatory institutions and public policy Auditors, firms and markets are regulated

7

Audit Testing Procedures(requires accounting firms)

Purpose Understanding, evaluating and improving the quality of audit

testing and evidence-gathering procedures Historical examples

Are accounts receivable confirmations effective/reliable? Caster (AJPT 1990)

Is classical sampling efficient/effective in auditing? Lead to development of dollar-unit sampling

Current examples Archival - Do control risk assessments affect other audit tests?

Mock and Wright (AJPT 1999), test of the “audit risk model” Experimental - Is the new “business risk” audit approach

effective? O’Donnell & Schultz (TAR 2005)

8

Auditor Judgment Research Purpose

To understand how auditors implement tests, interpret evidence Ultimate goal – to improve audit judgments in the field

Historical examples Descriptive models (1970s)

What information cues affect auditor decision-making? Heuristics and biases (1980s)

Are auditors different than other decision-makers? Memory and cognition (1990s)

What do auditors know and why?

Some current examples Do different “interventions” affect auditor-client negotiation outcomes? Do alternative forms of “review” affect auditor performance? Do “groups” outperform individuals?

Audit teams required to “brainstorm” for fraud risk Does the “business risk” model improve auditor judgments?

9

Accounting Firms(proprietary & largely a black box)

Purpose Understanding the organization and administrative structures of

accounting firms Field studies – e.g., Dirsmith & Covaleski (AOS 1985)

Examples of RQ How centralized/decentralized are firms?

How much autonomy do local offices/partners have? A central issue in Andersen’s Enron audit

How do partners share profits? What are incentive effects? Large vs. small profit-sharing pools, and awarding of partnership units

Burrows & Black (AOS 1998); Trompeter (AJPT 1994); Liu & Simunic (TAR 2005)

Do firm reputations (audit quality) vary? Across offices? Across countries? If so, why? Begs the question of what is an accounting firm? More on this later

10

Observable Audit Outcomes Direct (who audits whom?)

Auditor resignations/client disagreements (8-Ks) Krishnan & Krishnan (TAR 1997); Shu (JAE 2000)

Audit reports Are reports informative? GC and New I/C reports

Indirect Financial statement/earnings quality

Jointly produced by companies/auditors Work to date is limited mainly to “accruals”

Secondary effects of differential audit quality Debt markets

Mansi et al. (JAR 2004) Equity markets

Khurana et al. (TAR 2004); Teoh & Wong (TAR 1993)

11

Examples of audit-outcome research questions

How do auditor characteristics affect audit outcomes? Firm size, brand name, industry expertise Locale/unit of analysis (global, country, office, partner)

How do engagement-specific characteristics affect audit outcomes? Proxies for independence

Fees (client influence), engagement tenure, auditor alumni,

How do client characteristics affect audit outcomes? Corporate governance (e.g., audit committees)

How do institutions affect audit outcomes? Regulatory agencies, litigation & investor protection

12

Audit Industry & Markets Industry Structure

Why is the industry dominated by large firms? Scale economies?

Yes – Banker et al. (JAE 2003) No – Zind & Zeghal (CAR 1991)

Positive spillovers in joint production of audit/other services? Economies of scope?

Simunic (JAR 1984); Whisenant et al. (JAR 2003) Audit Market Structure

Are audit markets competitive, or monopolistic? Oligopoly/monopolistic competition?

Is there demand for (supply of) differential audit quality? If so, why? How is it priced? How is audit quality affected?

How have mergers/consolidations affected audit markets?

13

Regulatory Institutions How do professional bodies & regulatory institutions

affect audits and accounting firms? Examples of RQ:

How do sanctions affect audits/firms? Wilson & Grimlund (APJT 1990) – SEC’s AAER Hilary and Lennox (JAE 2006) – POB Peer Reviews

How do legal liability regimes affect audits? Analytical/Experimental

Dye (JPE 1993; JAE 1997); Dopuch et al. (JAE 1997) Empirical

Francis & Krishnan (APJAE 2003); Lee & Mande (AJPT 2003); Current topics

How do current regulations (PCAOB/SOX) affect audits? Should nonaudit services (including taxes) be banned?

14

Empirical Research on Audit Quality

Audits are relatively cheap, and proven failures are infrequent. But this does not necessarily mean audits are always of high quality. What do we know about audit quality from empirical research?

15

Research on Audit Outcomes

What do know about audit quality from audit report research?

16

Inferring Audit Quality from Audit Reports

After 1989, only two primary audit reports Standard clean opinion (90%) Modified opinion for going concern (GC) (10%)

How do users respond to “modified” reports? Are modified reports perceived to convey

“negative” news? If so, then audit reports are informative and

auditing has value to users

17

Type 1 & 2 Error Rates(1995-2003 Compustat Population) Clean opinions (90%), GC (10%), n= 62,094

firm-year observations Overall Error Rate – 9.5%

Type 1 Errors (false negatives) – 9% (over-qualifying or auditor conservatism)

No bankruptcy within one year, but 9% had GC reports 5,467 GC reports for 785 bankruptcies (7 GC per

bankruptcy)

Type 2 Errors (false positives) – 55% (under-qualifying)

Bankruptcy within one year (785) but no GC 55% of time, i.e., 432 cases over an 8-year period

18

Do Modified Audit Reports Matter

(given type 1/2 errors?)

Informational value is difficult to assess Audit reports issued concurrently with financials Most GC reports are “repeat offenders”

Evidence that “surprise” first-time GC reports reduce share prices

Dodd et al. (JAE 1984), Loudder et al. (AJPT 1992) Predictive ability of modified audit reports for

material loss contingences (lawsuits) Raghundan (CAR 1993)

19

Audit Reports and the IPO Setting

Audit report has potentially more value Greater information uncertainty/asymmetry

Weber and Willenborg (JAR 2003) Microcap IPO’s < $10 million 23% have GC reports preceding IPO Pre-IPO audit reports predict delistings and future

stock returns But more so for large (Big 4 & national)

auditors

20

Who Audits Whom?

What do we know about auditor differences and audit quality?

21

Differential Audit Quality Research

Audit quality cannot be directly observed Except for “proven” audit failures (ex post) which are rare

Instead – differential audit quality is inferred Comparing audit outcomes between “classes” of auditors

All firms are assumed to meet minimum professional & legal requirements

But many different types of firms exist Suggests supply of differential auditing Implies differential demand (clienteles)

22

Audit Firm Size:Is Bigger Better?

The big firm/small firm dichotomy Big 8 (now Big 4) firms have brand name reputation and

incentives to protect their reputation Simunic and Stein (1987 CGA Monograph)

Auditor size proxies for quality due to less fee dependence (DeAngelo, JAE 1981) Client dependence = 1/n (where n=number of clients) Now we can use actual fee dependence

Doesn’t mean Big 4 audits are always better It just means that on average they are better Individual audit failures can and do occur

23

Evidence of Big 4 Audit Quality Differentiation

Audit pricing Big 4 audit fee premia of 20-50%

US, Australia, Hong Kong, UK Implies more quantity and/or better quality auditing

Higher fees imply voluntary demand for higher quality audits by some clients

Demand for quality Demand for Big 4 audits is increasing in agency costs

Francis & Wilson (TAR 1988), DeFond (AJPT 1992) IPOs with larger auditors have less IPO underpricing

Beatty (TAR 1989), Willenborg (JAR 1999)

24

Other Evidence Big accounting firms are sued less often

Litigation as quality proxy (Palmose, TAR, 1987) Big firms sanctioned less often by regulators

Feroz et al. (JAR 1991) Alternative explanation is that big firms are

more powerful and have resources to fight But evidence on audit “outcomes” is more

consistent with higher quality Let’s turn to that evidence

25

Evidence from Audit Outcomes Audit Reports of Large (Big 4) Auditors

More conservative reports (more modifications) Francis & Krishnan (CAR 1999), (APJAE 2003)

More informative reports (predictive power) IPO setting, Weber & Willenborg (JAR 2003)

More accurate reports in UK, Lennox (ABR 1999) Financial Statements Audited by Big 4 Auditors

Smaller abnormal accruals -- less managerial discretion Becker et al. (CAR 1998), Francis et al. (AJPT 1999)

Accruals/earnings surprises valued higher in the stock market Teoh & Wong (TAR 1993), Krishnan (AJPT 2003)

26

Conclusions on Auditor Size Evidence is supportive of higher Big 4 audit quality But an alternative explanation is endogeneity

i.e., “good” companies choose good (Big 4) auditors, in which case selection may explain outcomes, not audit quality

Heckman 2-stage approach is trendy But most studies using it support prior findings on audit quality

Weber & Willenborg (JAR 2003), Hogan (TAR 1997), Ireland & Lennox (JAAF 2002)

Chaney et al. (TAR 2004) is an exception Can we identify good “instruments” in accounting research?

Larcker & Rusticus (SSRN WP) are skeptical If there’s reason to think selection exists, an alternative is to

control through research design choices e.g., matched pairs, or samples limited to smaller clients

27

Moving Beyond the Big-Small Dichotomy

Differentiation within

the dominant Big 4 group

28

Within-Big 4 Differentiation

In past, Big 4 viewed as homogenous group B4 (90%) & NB4 (10%) U.S. market shares

Low power due to small variance in auditor type May be more variation within Big 4 firms

Some plausible within-Big 4 variations: Industry expertise Geographical variation Offices (cross-city differences) Countries (cross-country differences)

29

Industry Expertise Firms actively promote their industry expertise Industry experts have more experience and appear

to make better audit judgments Solomon et al. (JAR 1999); Low (TAR 2004)

Auditor clienteles proxy for industry expertise More clients create more opportunities to acquire deep

industry knowledge/expertise Industry shares are not evenly distributed

Using 2000-2001 U.S. fee data & 2-digit SIC codes #1 firm has 50% of industry fees, #2 firm only 22% Leadership among Big 5 firms for 63 industries

AA (14), DT (5), EY (16), KPMG (9), PWC (19)

30

Evidence from Audit Fees

Big 4 industry leaders have higher fees (10-30%) and higher fees imply higher quality audits

Financial statement outcomes evidence higher quality audits by Big 4 industry leaders Smaller abnormal accruals (less discretion)

Balsam et al. (APJT 2003), Krishnan (AH 2003) Higher valuation of earnings surprises

Balsam et al. (AJPT 2003) Less fraudulent reporting (AAERs)

Carcello & Nagy (2004 Managerial Auditing Journal) Less IPO underpricing and smaller accruals

Elder and Zhou (2003 WP)

31

What is an Accounting Firm?

What’s the relevant unit of analysis? Individual offices, national practices, or international operations?

32

Arguments for an Office-Level Analysis

Accounting firms are decentralized networks of quasi-independent practice offices

Engagement partners in local offices contract and administer audits to clients in the same locale

Issue audit reports on office letterhead Francis et al. (ABACUS 1999); Reynolds & Francis (JAE

2000); Ferguson et al. (TAR 2003)

Client influence/fee dependence is stronger at the office level

Enron < 2% of Andersen fees nationally Over 30% of Houston office fees

33

First Office-Level StudyReynolds & Francis (JAE 2000)

Auditors are more conservative for larger clients in U.S. practice offices They issue more GC audit reports Clients have smaller abnormal accruals

Client size measured “relative” to office clienteles Client size is not significant using national clienteles

to measure relative client size Conclusion

Office-level data provides better understanding No results using aggregate data; but evidence of auditor

conservatism using office-level data

34

Office-Level Reputation for Industry Expertise

Is industry expertise firm-wide (national), office-specific (city), or combination of both? Office-specific arguments

Deep expertise is client-specific and cannot be fully captured by firms and distributed across offices

Gilson & Mnookin (SLR 1985), law firms Firm-wide arguments

Standardized training/audit practices and knowledge sharing is possible across offices

35

How is National Versus City Industry Leadership Priced?

4 possibilities (Francis et al., TAR 2005) Engagements where auditor is both the national

and the city-specific industry leader 18% of US sample

Engagements where auditor is a national industry leader (alone)

10% of US sample Engagements where auditor is a city-specific

industry leaders (alone) 23% of US sample

Nonleaders (neither city/national leaders) 49% of US sample

36

Evidence Joint national-city #1 has +28% premium City-alone #1 has +10% premium National-alone #1 has no fee premium relative to

nonleaders (default group is nonleaders) Implication - audit quality is office-specific

Industry premia only when a city leader Either alone or joint national-city leader

National market share is driven by city leadership National #1, 86% of fees where city leader

37

RQ - Does City Leadership Affect Earnings Quality?

Follow-up WP to Francis et al. (TAR 2005) Uses same B5 sample from 2000-01 first-time cycle of fee

disclosures to measure industry leaders

What do we find? Clients of city industry leaders have smaller abnormal

accruals, and smaller income-increasing accruals Magnitude is 6-12% of pre-tax earnings

Clients of joint national-city leaders are less likely to meet or beat by +1 cent analysts’ forecasts (40% 29%)

City-only leaders are also less likely to do so (40% 33%)

In both tests results for national leaders (alone) are n.s.

38

Legal Systems and Differential Auditing

(over time/countries) Institutions & legal regimes affect auditor incentives Two research approaches

A regime change within a single country (over time) The Private Securities Litigation Reform Act of 1995

Lee and Mande (AJPT 2003), Francis & Krishnan (APJAE 2003)

Cross-country studies of different legal regimes Seetharaman, Gul and Lyn (JAE 2002)

higher audit fees of UK firms cross-listed in US implies higher fees due to increased litigation risk exposure

Khurana et al. (TAR 2004) B4 clients have lower COC in US (but not Australia, UK, Canada) Implies US litigation risk drives B4 behavior (not reputations)

39

RQ – Do B4 incentives vary across legal regimes?

Do B4 audits vary across jurisdictions? Three Scenarios

Big 4 are uniform around the world Incentives to maintain brand name (standardization)

U.S. is an outlier due to extreme liability exposure More lawsuits in US than rest of world combined

Big 4 conservatism is increasing in country-level investor protection and auditor litigation risk

40

Big 4 Auditor Conservatism

Uniform Worldwide

U.S. Outlier

High

Low

Conservatism

Low High (U.S.)Legal Liability/Investor Protection

41

Preliminary Results Indicate that Legal Regime Matters

Big 4 conservatism is Increasing in investor protection/litigation risk Holds both with/without U.S. observations Implies less discretion to manage earnings

Magnitude Abnormal accruals are smaller by around 5% of earnings

in common law countries with stronger investor protection

Non-Big 4 firms are uniform across countries (and less conservative than Big 4) Less incentive for reputation protection/litigation

avoidance behavior (same as within US)

42

Other Examples of Research

on Audit Quality

Auditor tenure Non-audit services (NAS) Audit committees/corporate governance Accounting firm alumni

43

Auditor Tenure Tenure and mandatory auditor rotation

Long tenure is good because auditors know more about the client (deeper knowledge)

Long tenure is bad because of entrenchment and self-serving bias

The evidence to date: Audit quality is lower in first 3 years

Johnson et al. (CAR 2002) But longer tenure (5+ years) increases audit quality

Meyer et al. (TAR 2003) Studies of partner tenure in Australia and Taiwan

Office level analysis

44

Nonaudit Services (NAS) Do NAS impair auditor independence?

More fee dependence Inherent conflict of interest SEC attempted to ban NAS in 2000

Profession’s defense Increases auditor’s knowledge Clients demand and value NAS Incentives exist to protect reputations for

independence

45

Evidence Frankel, Johnson & Nelson (TAR 2002)

Compelling set of results against NAS Firms with higher NAS are associated with more earnings

management: Larger abnormal accruals (more discretion) More likely to meet analysts’ earnings targets

But also counter-evidence Ashbaugh et al. (TAR 2003), Chung & Kallapur (TAR

2003), Reynolds et al. (AJPT 2004), DeFond et al. (JAR 2003), Larcker & Richardson (JAR 2004)

46

Broader Implications Accounting journals generally don’t publish

replications and no-result studies But replications show fragility and sensitivity

to design choices e.g., Frankel et al. (TAR 2002) & Ashbaugh et al.

(TAR 2003) get different results on main tests, but with slightly different samples/models

47

Audit Quality and Corporate Boards/Governance

Audit quality is better when strong boards and audit committees exist (independence/outside directors) Auditors are more likely to issue GC reports

Carcello and Neal (TAR 2000) Less likely to be replaced following GC report

Carcello and Neal (TAR 2003) Auditors are more likely to detect fraud

Dechow et al. (CAR 1995) Accruals are smaller (less earnings management)

Klein (JAE 2002), Bedard et al. (AJPT 2004) Auditors have higher audit fees and are less likely to

perform non-audit services Abbot et al. (SSRN 2001), Abbott et al. (AJPT 2003)

48

Accounting Firm Alumni and Auditor Independence

Outplacement to clients impairs audit quality Clients know the auditor’s methods too well Auditor is cozy with former colleagues

Lennox (JAE 2005) Outplacement is less than perceived (10%) But GC reports are less frequent than predicted

(suggests client leniency) Menon & Williams (TAR 2004)

Firms with alumni have larger abnormal accruals

49

Policy Making

What’s the role of

academic auditing research

on audit quality?

50

The U.S. Experience

Academic research has little influence on regulation and policy-making

Two recent examples that ignored research SEC’s proposed ban on NAS in 2000

Arthur Levitt was convinced a ban was needed Quick adoption of Sarbanes-Oxley (SOX) in 2002

Political cover following Enron/Worldcom scandals

51

Why so little impact? Partly our fault

We rarely frame our research around normative regulatory issues or ‘sell’ its policy importance

We may not appear to be neutral There’s often an explicit anti-regulatory posture And we may (rightly) be viewed as apologists for the

accounting profession

Partly policy-makers’ fault No tradition of basing policy on research Policy-making is inherently political (not scientific) Overtures by the PCAOB are encouraging

52

Some ImportantUnanswered Questions

Does the U.S. evidence generalize to countries with different legal and regulatory institutions?

How much auditing is optimal? Audits are not very costly and proven failure rate is low Is ‘more’ auditing desirable, or cost-effective, e.g., SOX?

Is differential audit quality a good thing? If Big 4 or industry leaders are better, should they be

mandated? Or should there be choice?

Is government regulation (SEC/PCAOB) better than self-regulation with government oversight?

53

And Finally What level of “legal risk” achieves optimal audit

quality? U.S. is an outlier in legal-liability risk, although we observe B4

conservatism in other common law countries The risk of “audit firm failure” is real in the U.S. but nearly

impossible in rest of world Does “extreme” legal risk create the best incentive?

It didn’t prevent the Arthur Andersen collapse or frauds such as Worldcom, Adelphia, and HealthSouth

Another U.S. audit firm failure could end private-sector auditing around the world (global Big 4)

All Big 4 firms have serious legal problems Is a “Big n <= 3” viable for private-sector auditing

54

Follow-up Readings

J. Francis, “What Do We Know About Audit Quality?” The British Accounting Review (December 2004), Vo. 34, No. 4: 345-36

DeFond, M., and J. Francis, “Audit Research after Sarbanes-Oxley?” Auditing: A Journal of Practice and Theory (2005, Supplement): forthcoming.