International financial reporting : economic … · Economic Consequences of IFRS Adoption, Enforcement Reforms, and Interim Reporting ... 4 2 The German Accounting riadT

International Financial Reporting �

Economic Consequences of IFRS Adoption,

Enforcement Reforms, and

Interim Reporting Frequency

Inauguraldissertation

zur Erlangung der Würde

eines Doktors der Wirtschaftswissenschaft

der Fakultät für Wirtschaftswissenschaft

der Ruhr-Universität Bochum

vorgelegt von

Diplom-Kaufmann

Oliver Vogler

aus Gra�ng

2011

Dekan: Prof. Dr. Helmut KarlReferent: Prof. Dr. Jürgen ErnstbergerKorreferent: Prof. Dr. Bernhard PellensTag der mündlichen Prüfung: 11. Juli 2011

To Tanja � with love

Contents

List of Tables v

List of Figures vii

1 Introduction 1

1.1 Motivation and scope . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Summary and publication details . . . . . . . . . . . . . . . . . . . . 4

2 The German Accounting Triad � �Accounting Premium� for IAS/IFRS

and U.S. GAAP vis-à-vis German GAAP? 9

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2.2 Related literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.2.1 Information quality and cost of equity capital . . . . . . . . . 11

2.2.2 Determinants and impacts of the adoption of internationally

accepted accounting principles . . . . . . . . . . . . . . . . . . 13

2.2.3 Measurement of cost of equity capital . . . . . . . . . . . . . . 15

2.3 Institutional background . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.3.1 German �nancial reporting requirements . . . . . . . . . . . . 17

2.3.2 Accounting standards under investigation . . . . . . . . . . . . 18

2.3.3 German corporate governance and enforcement system . . . . 20

2.3.4 Germany's capital market . . . . . . . . . . . . . . . . . . . . 22

2.4 Hypotheses development . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.4.1 Overall impact of adopting IAAP . . . . . . . . . . . . . . . . 23

2.4.2 Impact of adopting IAAP in subperiods . . . . . . . . . . . . . 26

2.5 Research design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

2.5.1 Portfolio analyses . . . . . . . . . . . . . . . . . . . . . . . . . 27

2.5.2 Firm-level analyses . . . . . . . . . . . . . . . . . . . . . . . . 31

2.6 Sample selection and descriptive statistics . . . . . . . . . . . . . . . 33

i

Contents



2.7 Regression results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38



2.8 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

3 The Value and Accounting Premium for South African-listed Shares 53

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

3.2 Background and prior research . . . . . . . . . . . . . . . . . . . . . . 54

3.3 Data and portfolio construction . . . . . . . . . . . . . . . . . . . . . 55

3.4 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

3.4.1 Value premium . . . . . . . . . . . . . . . . . . . . . . . . . . 57

3.4.2 CAPM and multifactor model . . . . . . . . . . . . . . . . . . 58

3.4.3 Accounting premium . . . . . . . . . . . . . . . . . . . . . . . 59

3.4.4 The value premium and the accounting premium as risk factors 60

3.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

3.5.1 Value premium . . . . . . . . . . . . . . . . . . . . . . . . . . 61

3.5.2 CAPM and multifactor model . . . . . . . . . . . . . . . . . . 64

3.5.3 Accounting premium . . . . . . . . . . . . . . . . . . . . . . . 66

3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

4 Das Fama-French-Modell: Eine bewährte Alternative zum CAPM �

auch in Deutschland 71

4.1 Einleitung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

4.2 Historie: Vom CAPM zum Fama-French-Modell . . . . . . . . . . . . 72

4.2.1 CAPM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

4.2.2 Fama-French-Modell . . . . . . . . . . . . . . . . . . . . . . . 75

4.2.3 Multifaktorenmodelle: Weiterentwicklungen und Kritik . . . . 76

4.3 Das Fama-French-Modell in Deutschland . . . . . . . . . . . . . . . . 78

4.3.1 Bisherige empirische Erkenntnisse . . . . . . . . . . . . . . . . 78

4.3.2 Daten und Portfolio-Bildung . . . . . . . . . . . . . . . . . . . 79

4.3.3 Regressionsergebnisse . . . . . . . . . . . . . . . . . . . . . . . 81

4.4 Implementierung in der Praxis . . . . . . . . . . . . . . . . . . . . . . 83

4.4.1 Wann ist das Drei-Faktoren-Modell die richtige Modell-Wahl? 85

ii

Contents

4.4.2 Wie wirkt sich die Datenverfügbarkeit in der Praxis aus? . . . 86

4.5 Zusammenfassung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

5 Economic Consequences of Accounting Enforcement Reforms: The

Case of Germany 89

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

5.2 Institutional background . . . . . . . . . . . . . . . . . . . . . . . . . 92

5.3 Prior research and hypotheses . . . . . . . . . . . . . . . . . . . . . . 93

5.3.1 Enforcement reforms and the overall degree of enforcement . . 93

5.3.2 Earnings quality e�ects of the enforcement reforms . . . . . . 94

5.3.3 Capital market e�ects of the enforcement reforms . . . . . . . 96

5.3.4 Impact of enforcement through other mechanisms . . . . . . . 98

5.4 Research design, sample selection, and data . . . . . . . . . . . . . . 98

5.4.1 Regression approaches . . . . . . . . . . . . . . . . . . . . . . 98

5.4.2 Measurement variables . . . . . . . . . . . . . . . . . . . . . . 101

5.4.3 Sample selection and data sources . . . . . . . . . . . . . . . . 106

5.5 Empirical �ndings and sensitivity analyses . . . . . . . . . . . . . . . 107

5.5.1 Empirical �ndings . . . . . . . . . . . . . . . . . . . . . . . . . 107

5.5.2 Sensitivity analyses . . . . . . . . . . . . . . . . . . . . . . . . 112

5.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

6 The Real Business E�ects of Quarterly Reporting 121

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

6.2 Background on quarterly reporting in Europe . . . . . . . . . . . . . 125

6.2.1 EU Transparency Directive . . . . . . . . . . . . . . . . . . . . 125

6.2.2 Country-speci�c regulation . . . . . . . . . . . . . . . . . . . . 126

6.3 Related literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

6.3.1 Quarterly reporting . . . . . . . . . . . . . . . . . . . . . . . . 131

6.3.2 Real business e�ects . . . . . . . . . . . . . . . . . . . . . . . 132

6.4 Hypotheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

6.5 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

6.5.1 Real earnings management . . . . . . . . . . . . . . . . . . . . 139

6.5.2 Mandatory vs. voluntary quarterly reporting (MAND vs. VOL)142

6.5.3 �Suspect �rm-year� observations . . . . . . . . . . . . . . . . . 142

6.5.4 Controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

iii

Contents

6.5.5 Regressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

6.6 Sample and descriptive statistics . . . . . . . . . . . . . . . . . . . . . 147

6.6.1 Sample selection . . . . . . . . . . . . . . . . . . . . . . . . . 147

6.6.2 Descriptive statistics . . . . . . . . . . . . . . . . . . . . . . . 148

6.7 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

6.7.1 Main analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

6.7.2 Additional analyses on ACFO . . . . . . . . . . . . . . . . . . 155

6.7.3 Additional analyses on APROD and ADISC . . . . . . . . . . 157

6.8 Sensitivity analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

6.9 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167

Bibliography 169

iv

List of Tables

2.1 18 analyzed portfolios . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

2.2 Descriptive analysis for all 18 portfolios . . . . . . . . . . . . . . . . . 35

2.3 Descriptive statistics of yearly data . . . . . . . . . . . . . . . . . . . 37

2.4 Capital Asset Pricing Model . . . . . . . . . . . . . . . . . . . . . . . 39

2.5 Beta values comparison . . . . . . . . . . . . . . . . . . . . . . . . . . 40

2.6 Fama and French model . . . . . . . . . . . . . . . . . . . . . . . . . 41

2.7 Correlations of exogenous factors . . . . . . . . . . . . . . . . . . . . 42

2.8 GM model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

2.9 Improvements of adjusted R2 . . . . . . . . . . . . . . . . . . . . . . 44

2.10 Forecasts of cost of equity capital . . . . . . . . . . . . . . . . . . . . 44

2.11 Cost of equity capital comparison . . . . . . . . . . . . . . . . . . . . 45

2.12 First-stage regression . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

2.13 Second-stage regression (CAPM) . . . . . . . . . . . . . . . . . . . . 47

2.14 Second-stage regression (GM) . . . . . . . . . . . . . . . . . . . . . . 48

3.1 Summary statistics for the value premium . . . . . . . . . . . . . . . 61

3.2 CAPM and Multifactor model regression results . . . . . . . . . . . . 65

3.3 Summary statistics for the accounting premium . . . . . . . . . . . . 66

3.4 Estimation results for the multifactor model . . . . . . . . . . . . . . 68

4.1 Industrieklassi�zierung . . . . . . . . . . . . . . . . . . . . . . . . . . 80

4.2 Deskriptive Statistik der Variablen . . . . . . . . . . . . . . . . . . . 81

4.3 CAPM Schätzungen . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

4.4 Fama-French Schätzungen . . . . . . . . . . . . . . . . . . . . . . . . 84

5.1 De�nition of variables . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

5.2 Descriptive statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

5.3 Correlations for control variables . . . . . . . . . . . . . . . . . . . . 109

v

List of Tables

5.4 Regression results - non-matched sample . . . . . . . . . . . . . . . . 111

5.5 Impact of the level of enforcement through other mechanisms . . . . . 113

5.6 Regression results - matched sample . . . . . . . . . . . . . . . . . . . 115

6.1 Quarterly reporting environment in EU-15 countries . . . . . . . . . . 128

6.2 Simpli�ed model for expected REM activity . . . . . . . . . . . . . . 135

6.3 Sample overview by EU-15 countries . . . . . . . . . . . . . . . . . . 143

6.4 De�nition of variables . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

6.5 Descriptive statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

6.6 Mean equality tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

6.7 Regression of MAND, VOL and Suspect Year on ACFO . . . . . . . . 153

6.8 F -tests for hypotheses H2a, H2b, and H2c for ACFO . . . . . . . . . 156

6.9 F -tests for additional analyses . . . . . . . . . . . . . . . . . . . . . . 158

6.10 Regression of MAND, VOL and Suspect Year on APROD . . . . . . 160

6.11 Regression of MAND, VOL and Suspect Year on ADISC . . . . . . . 162

6.12 F -tests for hypotheses H2a, H2b, and H2c for APROD and ADISC . 166

vi

List of Figures

2.1 Excess returns from July 1997 to June 2005 . . . . . . . . . . . . . . 34

3.1 Distribution of the Value Premium among all �rms and among small

and big �rms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

3.2 Distribution of the Value Premium among small and big �rms during

the two subperiods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

3.3 Distribution of the Accounting Premium . . . . . . . . . . . . . . . . 67

vii

1 Introduction

1.1 Motivation and scope

Over the last decade, international �nancial reporting underwent several profound

changes in reporting harmonization � the most important being the rise of Interna-

tional Financial Reporting Standards (IFRS) as the globally predominant accounting

regime.

Indeed, as of April 2011, 118 countries either require or permit the use of IFRS

for listed companies in their domestic markets (IAS Plus, 2011). By the end of

2011, it is projected that the majority of Global Fortune 500 companies will employ

IFRS (Deloitte, 2010). All major economies that have not yet introduced IFRS have

established time lines to converge with or adopt IFRSs in the near future, which is

further supported by the Group of 20 Leaders (G20) (IFRS Foundation, 2011a). Of

particular importance is the convergence project between the International Account-

ing Standards Board (IASB) and the U.S. Financial Accounting Standards Board

(FASB). By June 2011, also the last remaining three major convergence projects

are targeted to be completed (FASB and IASB, 2009).1 Later in 2011, the U.S.

Securities and Exchange Commission (SEC) will make their decision on the use of

IFRS by U.S. domestic companies (Securities and Exchange Commission (SEC),

2010), after having accepted �nancial statements in accordance with IFRS without

reconciliation to U.S. GAAP from foreign private issuers since 2008 already.

In general, the objective of the IASB is to provide a �single set of high qual-

ity, understandable, enforceable and globally accepted� standards (IFRS Founda-

tion, 2011b). Participants supporting and promoting the worldwide implementation

of such standards expect bene�ts for companies and investors in several critical

1From the 10 major projects that were originally identi�ed as the joint programme in the 2006Memorandum of Understanding, the IASB and the FASB have only three major projects remain-ing to be completed � revenue recognition, leasing, and �nancial instruments.

1

1 Introduction

ways. For instance, when introducing the IAS regulation, the European Commis-

sion (2002) articulated the expectation that it would �help eliminate barriers to

cross-border trading by ensuring that company accounts are more reliable, trans-

parent, and comparable�. In turn, these accounting reforms, it was thought, would

increase market e�ciency and reduce the cost of raising capital for companies. Only

recently, IASB Chairman, David Tweedie, articulated that �global accounting stan-

dards will enhance the drive towards the free trade of capital internationally� and

that �all companies � large and small � are able to attract capital from a larger pool

of investors, driving down the cost of capital and facilitating cross-border mergers

and acquisitions activity and strategic investments� (Tweedie, IASB Chairman on

March 10, 2011).

With IFRS emerging as the globally dominant standard, however, also more

implementation-oriented questions are raised in the context of international �nancial

reporting.

One prominent question is which role enforcement plays in realizing the expected

economic bene�ts from introducing IFRS. Typically, national regulations regarding

the monitoring of compliance with the applicable reporting standards and providing

appropriate measures in case of infringements di�er quite substantially. In turn,

the variation in enforcement of international �nancial reporting is associated with

distinct economic consequences regarding, e.g., earnings management or market

valuation of a�ected companies. Therefore in Europe, the �IAS Regulation� (EC,

2002: No. 1606/2002) not only prescribes the application of IFRS for publicly traded

companies, but also requires all EU Member States to install e�ective mechanisms

for the enforcement of IFRS.

Another widely discussed topic in international �nancial reporting over the last

decade has been the level of mandatory disclosure that �rms should follow when

reporting under IFRS. E�ective in 2007, the European Commission adopted a Di-

rective to increase the transparency of �nancial reporting in Europe. Unlike other

directives, the �Transparency Directive� (2004/109/EC) is not a �maximum harmo-

nization� directive. Member states can set requirements above the minimum level

envisaged in the directive, e.g., with respect to the frequency of interim reporting.

This can be seen as a compromise after several countries (e.g., UK, Denmark, and

the Netherlands) had objected to further harmonized regulation regarding minimum

publication requirements for companies publicly-traded in the EU. As a result, for

2

1.1 Motivation and scope

instance, the Transparency Directive requires �rms only to disclose �Interim Man-

agement Statements� (IMS) � instead of full quarterly reports like in the U.S.

From an academic point of view, however, it remains an open question whether the

purported bene�ts touted by supporters of harmonizing international �nancial re-

porting have actually materialized � be it in terms of IFRS adoption e�ects, whether

the reforms in enforcement mechanisms have paid o�, or how, in turn, leaving cer-

tain choices in the implementation of disclosure regulation could actually level the

playing �eld at all.

The purpose of this dissertation is to analyze the economic implications of major

regulation changes in international �nancial reporting by examining three critical

aspects: �rst, the introduction of IFRS as the internationally common accounting

standard; second, the role of consistent enforcement mechanisms in realizing the

bene�ts of IFRS; and third, the implications of leaving national discretion regarding

interim reporting frequency. These three topics are analyzed based on 5 separate

papers that each individually contribute to the discussion and debate regarding the

bene�ts and costs of international �nancial reporting harmonization.

In the �rst part of this thesis, Papers I�III assess the general economic conse-

quences of IFRS adoption, comparing the cost of equity capital di�erences of IFRS

�rms with �rms that apply domestic or U.S. GAAP standards. The key paper on this

topic is The German Accounting Triad � �Accounting Premium� for IAS/IFRS and U.S.

GAAP vis-à-vis German GAAP? (Paper I), which compares the cost of equity capital

di�erences between IFRS, U.S. GAAP and German GAAP �rms. The paper makes

use of the unique setting in Germany before 2005, where all three di�erent types of

accounting regimes could be analyzed in the same homogeneous institutional setting.

Papers II and III represent more detailed analyses (or �robustness tests�) on speci�c

choices of how to assess the economic consequences of IFRS adoption in Paper I.

The paper The Value and Accounting Premium for South African-listed Shares (Paper

II) tests whether the �accounting premium� for IFRS also exists outside Germany,

looking at South Africa as another case of early IFRS adoption and documenting

cost of equity capital di�erences compared to domestic GAAP �rms. The paper

Das Fama-French-Modell: Eine bewährte Alternative zum CAPM � auch in Deutschland

(The Fama-French model: A proven alternative to the CAPM � in Germany as well)

(Paper III) provides more background on how to measure cost of equity capital

di�erences in general and speci�cally tests the di�erent performance of the CAPM

3

1 Introduction

and Fama-French model for a �naive� industry classi�cation unwinding the �sorting

portfolio� assumption, all in the institutional setting at hand, i.e. Germany.

The second part of this dissertation deals with the speci�c question of how en-

forcement regulation in�uences the economic consequences of IFRS (Economic Con-

sequences of Accounting Enforcement Reforms: The Case of Germany, Paper IV),

looking at recent enforcement reforms in Germany. The objective of these reforms

was to promote a consistent and faithful application of accounting standards, which

is associated with reducing earnings management as well as improving stock liquidity

and market valuation.

The third part discusses how the frequency (quarterly vs. semi-annualy) and type

(voluntary vs. mandatory) of disclosure leads to di�erent �business e�ects� within

international �nancial reporting (The Real Business E�ects of Quarterly Reporting,

Paper V), i.e., myopia-induced deviations from normal operational practices through

real activities manipulation. The focus lies on the resulting economic di�erences

of leaving discretion on the introduction of mandatory quarterly reporting to the

individual EU member states.

The remainder of this dissertation proceeds as follows. Section 1.2 gives a short

summary on the individual papers as well as publication details, before chapters 2

to 6 present the �ve research papers in all detail.

1.2 Summary and publication details

This dissertation is a cumulative work that consists of �ve separate papers in the

context of �International Financial Reporting � Economic Consequences of IFRS

Adoption, Enforcement Reforms, and Interim Reporting Frequency�. In this section,

each of the individual papers is brie�y summarized. Moreover, publication details �

including presentations on international conferences � are given.

Paper I: The German Accounting Triad � �Accounting Premium� for IAS/IFRS

and U.S. GAAP vis-à-vis German GAAP?

Co-author: Jürgen Ernstberger

Abstract: This paper critically examines the impact of voluntary adoption of Interna-

tionally Accepted Accounting Principles (IAAP, i.e. IAS/IFRS and U.S. GAAP) on

the cost of equity capital in Germany. We �nd that (1) overall cost of equity capital

4


estimates in the Capital Asset Pricing Model (CAPM) for companies applying IAAP

are signi�cantly lower compared to those applying German GAAP, (2) an enhanced

multifactor model which incorporates the accounting regime di�erences (called �GM

model�) absorbs the cost of equity capital di�erences, and (3) changes of the insti-

tutional background in Germany and of the accounting standards lead to di�erent

cost of equity capital e�ects for sub-periods of the 1998�2004 voluntary adoption

period, while particularly controlling for e�ects like self-selection, cross-listing, and

New Market (Neuer Markt) listing. The central thesis advanced in this paper is

that changes in the accounting standards and the institutional infrastructure can

in�uence the impact of applying IAAP. Therefore, we suggest to incorporating an

accounting factor into the cost of equity capital analysis.

Keywords: Accounting regime adoption, cost of equity capital, multifactor model,

IFRS, U.S. GAAP, Germany.

Publication details: Published in The International Journal of Accounting, 43 (De-

cember 2008), 339�386. A previous version of this paper was presented at the Illinois

International Accounting Symposium held at the University of Hawai'i at Manoa,

U.S. (June 2007), as well as at the 31st Annual Congress of the European Account-

ing Association in Rotterdam, NL (April 2008). The paper was granted the ERIM

Best Paper Award in �International Financial Accounting�.

Paper II: The Value and Accounting Premium for South African-listed Shares

Co-authors: Jürgen Ernstberger, Christian Heinze

Abstract: In the last decade, empirical research has found strong evidence that value

stocks provide higher returns than growth stocks (value premium). Firms with a

high ratio of the book value of equity to the market value of equity are regarded

as value stocks; a low ratio identi�es growth stocks. Most research is tailored to

the market in the United States of America. Only a few studies consider country-

speci�c distinctions. This research analyses the value premium for the South African

market and compares its magnitude to the �ndings for the U.S. market. Moreover,

the e�ects of the introduction of IFRS for companies listed at the JSE Limited are

examined. The adoption of IFRS is used to demonstrate that investors award an

accounting premium for voluntary compliance with this new accounting standard.

Keywords: Value premium, accounting premium, South Africa, asset pricing, Fama-

French model, multi-factor model.

5

1 Introduction

Publication details: Published in the Journal of Economic and Financial Sciences,

2 (October 2008), 187�202. A previous version of this paper was presented at the

Value 2008 conference held at Sun City, SA (May 2008).

Paper III: Das Fama-French-Modell: Eine bewährte Alternative zum CAPM

� auch in Deutschland (The Fama-French model: A proven alternative to the CAPM

� in Germany as well)

Abstract: In der Praxis ist das vorherrschende Modell zur Bestimmung der Eigenkap-

italkosten immer noch das Capital Asset Pricing Model (CAPM), obwohl vor 15

Jahren ein empirisch überlegenes Modell eingeführt wurde, nämlich das 3-Faktoren-

Modell von Eugene F. Fama und Kenneth R. French. In der Tat erweist sich das

Fama-French-Modell auch in Deutschland als empirisch überlegen, weil es einen

höheren Erklärungsgehalt besitzt als das CAPM und gezielt Anomalien berück-

sichtigen kann. Gegenstand dieses Beitrags ist es, die empirische Vorteilhaftigkeit

des Fama-French-Modells in Deutschlands sowie die praktische Anwendbarkeit und

jüngste Weiterentwicklungen der Multifaktorenmodelle aufzuzeigen.

Keywords: Bewertung, Fama-French-Modell, CAPM, Eigenkapitalkosten.

Publication details: Published in the FinanzBetrieb, 11 (Juli/August 2009), 382�388.

Paper IV: Economic Consequences of Accounting Enforcement Reforms: The

Case of Germany

Co-authors: Jürgen Ernstberger, Michael Stich

Abstract: This study investigates recent reforms of �nancial reporting enforcement

in Germany. The objective of these reforms was to promote a consistent and faithful

application of accounting standards. Conducting multivariate analyses, we �nd a

signi�cant decrease in earnings management, a signi�cant increase in stock liquidity,

as well as limited evidence for an increase in market valuation for the companies that

fall under the new enforcement regime. We document that companies that are char-

acterized by an overall low level of enforcement through other internal and external

mechanisms are particularly a�ected by these reforms. In our analyses, we carefully

control for e�ects of the mandatory IFRS adoption in Europe as well as for various

other e�ects arising from the application of di�erent accounting principles. Our

results hold for di�erent research designs, including a di�erence-in-di�erences ap-

proach and a matched sample approach. We conclude that the enforcement reforms

6


in Germany have leveled the playing �eld in the enforcement of �nancial reporting

and have enhanced the trust in �nancial reports of the a�ected companies.

Keywords: Enforcement, regulation, IFRS, earnings management, stock liquidity,

market valuation.

Publication details: Revise and resubmit (4th and �nal round) in the European Ac-

counting Review. Previous versions of this paper were presented at the American

Accounting Association (AAA) Annual Meeting in Anaheim, U.S. (August 2008),

at the AAA Annual Meeting in New York, U.S. (August 2009), at the 32nd Annual

Congress of the European Accounting Association (EAA) in Tampere, FI (May

2009), as well as at the 33rd Annual Congress of the EAA in Istanbul, TR (May

2010).

Paper V: The Real Business E�ects of Quarterly Reporting

Co-authors: Jürgen Ernstberger, Benedikt Link

Abstract: We examine the real business e�ects associated with di�erent interim

reporting frequency regimes. Speci�cally, we investigate the impact of voluntary

and mandatory quarterly reporting on real earnings management (REM), i.e., the

deviation from normal operational practices through real activities manipulation.

We hypothesize that mandatory (voluntary) quarterly reporting is associated with

higher (lower) REM compared to semi-annual reporting. Using a sample comprising

15 countries from the European Union (EU) that provide a unique regulatory setting

of both mandatory and voluntary quarterly reporting regimes, we �nd that manda-

tory quarterly reporters exhibit signi�cantly higher REM compared to semi-annual

reporters. This e�ect is particularly strong when company performance is below

industry average, accounting earnings management is high, analyst coverage is low,

and in countries with low minority shareholder protection (�suspect �rm-years�. In

contrast, voluntary quarterly reporting is associated with lower REM compared to

semi-annual reporters in �non-suspect �rm-years�, suggesting that primarily �good�

�rms voluntarily increase their disclosure frequency. In suspect �rm-years, however,

voluntary quarterly reporters exhibit disproportionally higher and thus comparable

REM levels to semi-annual reporters. This indicates that �good� �rms need to sig-

ni�cantly increase REM levels under adverse conditions suggesting that the higher

disclosure level becomes a burden. In line with expectations, mandatory quar-

terly reporters consistently exhibit higher REM compared to voluntary quarterly

7

1 Introduction

and semi-annual reporters across all settings underlining the evidence for reporting

frequency-induced REM in mandatory regimes.

Keywords: Interim reporting, quarterly reporting, real business e�ects, real earnings

management, international accounting, management myopia.

Publication details: Working Paper. This paper was presented at the AAA Mid-Year

Meeting of the International Accounting and Financial Accounting and Reporting

Sections in Tampa, U.S. (February 2011) and was afterwards discussed with 21

Accounting Professors at 10 U.S. Business Schools (among others, Chicago Booth

School of Business, Harvard Business School, Kellogg School of Management, MIT

Sloan School of Management, The Wharton School). The paper was granted the

AAA Best Paper Award in International Accounting.

8

2 The German Accounting Triad

� �Accounting Premium� for

IAS/IFRS and U.S. GAAP

vis-à-vis German GAAP?

Published in:

The International Journal of Accounting, 43 (December 2008), 339�386 (with Jürgen

Ernstberger)

2.1 Introduction

This paper critically examines the impact of the voluntary adoption of IAS/IFRS1

and U.S. GAAP (in the following referred to as �Internationally Accepted Accounting

Principles�, IAAP) by German companies. The results suggest that for companies

adopting IAAP an �accounting premium� is granted by investors, implying a lower

cost of equity capital. Our results speci�cally hold when controlling for e�ects like

self-selection, cross-listing, and New Market (Neuer Markt) listing. Based on these

accounting anomalies we can develop a novel multifactor model that captures the

�accounting premium� and leads to an improvement of the CAPM and Fama-French

model.

The paper contends that the adoption of IAAP could have direct and indirect

e�ects on the cost of equity capital. The indirect e�ects via improving earnings

1The International Financial Reporting Standards (IFRS) were initially called International Ac-counting Standards (IAS). In 2002, they changed name to International Financial ReportingStandards (IFRS). In general, we use a combination of both terms (�IAS/IFRS�). When we referto a speci�c time period, we use the term �IAS� for years before (including) 2002 and the term�IFRS� afterwards.

9


quality or disclosure levels as well as lowering information asymmetry have been

widely examined. However, the impacts and interrelations of these e�ects are di�-

cult to separate. Moreover, only studies on the entire link between the adoption of

IAAP and the cost of equity capital allow to capture the direct e�ects like additional

costs or impact on brand recognition. Thus, our study focuses on the entire link to

comprehensively test the indirect and direct e�ects.

For the empirical analysis, we use both a portfolio-based and a �rm-level analy-

sis. For the portfolio view on the one hand, we apply a Capital Asset Pricing Model

(CAPM) and an enhanced multifactor model to include the information about the

type of accounting regime applied as an additional factor. Even though Francis

et al. (2005a), Ecker et al. (2006), and Barth et al. (2006a) have worked with factor-

mimicking portfolios and new information risk related factors, to the authors' knowl-

edge so far no study has assessed the impact of the type of accounting regime applied

into an asset pricing model like the multifactor analysis based on the Fama-French

three factor model (Fama and French, 1993). Our combination of multifactor re-

gression and accounting regime factors leads to a novel approach of studying the

question whether di�erent accounting regimes justify empirically signi�cant di�er-

ences in excess returns in an asset pricing model. For the �rm-level analysis on the

other hand, we incorporate a two-stage estimation procedure in which we are able

to explicitly address the issue of self-selection. Moreover, we control for cross-listing

e�ects and the New Market (Neuer Markt) membership.

The motivation of this study is to test the major objectives of many German com-

panies for adopting an IAAP voluntarily. By substituting the domestic accounting

regime and therefore improving the transparency of �nancial reporting, German

companies expected to lower the cost of equity capital. This study tests this notion

by analyzing the cost of equity capital e�ect of adopting an IAAP.

This paper makes several contributions to the existing literature. First, we thor-

oughly examine the e�ects of adopting a new accounting regime theoretically and

for our speci�c setting. In doing so, we respond to the call of Holthausen (2003)

for research �determining the marginal e�ects of accounting standards, incentives,

ownership structure, institutional features of the capital markets and enforcement

on the quality of �nancial reporting� (p. 273).

Second, we develop a new method of comparing cost of equity capital for com-

panies applying IAAP. This method has the advantages of not being biased by the

10

2.2 Related literature

quality of analysts' forecasts like residual income models and of mitigating the prob-

lem of self-selection of companies adopting new accounting regimes. Consequently,

our sample is substantially bigger and less biased compared to other studies (e.g.,

Leuz and Verrecchia, 2000), since we do not lose (small) companies that are typically

not followed by analysts.

Third, we are the �rst to di�erentiate between sub-periods of introducing IAAP

in Germany. Our results support the expectation that the institutional changes over

time distinctly in�uence the e�ect on the cost of equity capital.

Fourth, although the classical Fama-French three-factor model has been success-

fully used in a variety of empirical studies (e.g., Fama and French, 1998; Liew and

Vassalou, 2000), the German capital market has been relatively overlooked in such

studies. What is more, previous studies have examined the Fama-French three-

factor model across di�erent countries, but to our knowledge this is the �rst study

to consider di�erent accounting regimes in a homogenous institutional setting. Our

�ndings suggest that the type of accounting regime applied is a priced risk factor in

the multifactor model.

Finally, we contribute to the literature comparing IAAP, like IAS/IFRS and U.S.

GAAP, to domestic GAAP. For our setting we �nd that the adoption of these ac-

counting regimes is associated with reduced cost of equity capital.

The remainder of this paper is organized as follows. In the following Section

2.2, we summarize related strands of literature. In Section 2.3, we describe the

institutional background of our study. Our hypotheses are developed in Section 2.4.

In Section 2.5 we develop the research design. In Section 2.6 we present the data

and descriptive statistics, followed by our econometric results in Section 2.7. Section

2.8 concludes with a summary of our results and discusses implications for future

research.


2.2.1 Information quality and cost of equity capital

The link between information quality and cost of equity capital is one of the most

fundamental and controversial subjects in recent accounting research. Theory sug-

gests a negative relationship between the quality of (accounting) information on the

11


one hand and the estimation risk and information asymmetry for investors, and

hence the cost of equity capital, on the other hand (Habib, 2006).

Also the level of disclosures is regarded to be essential for cost of equity capital.

Diamond and Verrecchia (1991) argue that voluntary disclosures reduce information

asymmetries among informed and uninformed investors and �nd that higher levels

of disclosure reduce estimation risk. Assuming estimation risk not being completely

diversi�able, investors will require a return premium as compensation for additional

risk components. This premium is interpreted as a higher cost of equity capital.

Botosan (2006) thoroughly reviews the link between disclosure and cost of equity,

assessing that �extent theory strongly supports the hypothesis that greater disclosure

reduces cost of equity capital� (p. 39). But she also admits that the underlying

assumption that public disclosure mitigates information asymmetry is not true for

all studies, suggesting additional research in this �eld of literature.

There are other critical voices regarding the assumption that public disclosure mit-

igates information asymmetry by displacing private information. Verrecchia (2001)

misses an underlying theory and attests no unambiguous empirical evidence for a

positive association between information quality and cost of equity capital. Kim

and Verrecchia (1994) argue that public disclosure might be processed into private

information again, also by informed investors. They state that more complex infor-

mation might improve the quality of private information of informed investors even

more than the information quality of public information for less informed investors

(Kim and Verrecchia, 1991).

As Hail and Leuz (2006) argue, the favorable e�ects of more disclosure are not

predictable as they might be relatively small or (to a large extent) captured by

traditional proxies of risk. Several theoretical studies argue that the size of the

economy examined might in�uence the magnitude of these e�ects (Clarkson and

Thompson, 1990; Coles et al., 1995; Clarkson et al., 1996; Easley and O'Hara, 2004;

Hughes et al., 2007).

Easley and O'Hara (2004) �nd that a higher proportion of private information

increases cost of equity capital, whereas cost of equity capital is decreased by higher

dispersion of private information and higher precision of private and public infor-

mation. They see private information as inducing a new form of systematic risk and

highlight that investors require compensation for that risk. They attest that �indi-

vidual �rms can in�uence cost of equity capital by choosing features like accounting

12


treatments� (Easley and O'Hara, 2004, p. 1554).

Hughes et al. (2007) work can be seen as an extension of Easley and O'Hara (2004).

In contrast to their predecessors, they �nd that in large economies idiosyncratic

risk is not priced. They call the fact that a large number of empirical studies

presume information asymmetry is priced, because of having to trade with privately

informed investors, �a commonly held misperception� (Hughes et al., 2007, p. 707).

This price-protection e�ect, also characterized in Easley and O'Hara (2004), is in

fact driven by under-diversi�cation and will disappear in large economies, in their

opinion. Nevertheless, while controlling for total information, they show that high

information asymmetry does lead to high cost of equity capital.

In our study, we investigate the impact of adoption of IAS/IFRS and U.S. GAAP

by German companies from 1998�2004. We argue that the speci�c institutional

setting in Germany and its changes over time give rise to question the positive

impact, and implicitly calls for a detailed analysis. Our results provide evidence for

a lower cost of equity capital of companies applying IAAP in Germany. However, the

cost of equity capital e�ects of applying IAAP is di�erent for the three sub-periods

examined. Moreover, we document that the type of accounting regime applied is a

priced risk factor in our sample.

2.2.2 Determinants and impacts of the adoption of

internationally accepted accounting principles

Various studies have addressed the determinants and impacts of the adoption of

internationally accepted accounting principles (IAAP). One stream of research iden-

ti�es attributes of companies which voluntarily change to IAAP (e.g., such analyses

are included in Leuz and Verrecchia, 2000; Gassen and Sellhorn, 2006).

The second stream of research investigates whether the adoption of IAS/IFRS and

U.S. GAAP causes signi�cant changes to �nancial statements. For German com-

panies, Moya and Oliveras (2006) on average �nd statistically signi�cant increases

in equity, but less obvious e�ects on net income. Several other studies corrobo-

rate these results (Küting et al., 2002; Burger et al., 2004, 2005, 2006; Küting and

Zwirner, 2007).

A third stream of literature examines di�erences in earnings attributes and accru-

als between the accounting regimes. Whereas many studies �nd a higher earnings

13


quality for IAS/IFRS companies compared to German GAAP companies in terms

of certain measures, e.g. timeliness, predictability, conservatism, earnings manage-

ment, value relevance, and analysts' forecast accuracy (Ashbaugh and Pincus, 2001;

Barth et al., 2008; Bartov et al., 2005), �ndings of other studies suggest similar

results for both accounting regimes (Van Tendeloo and Vanstraelen, 2005; Gon-

charov, 2005; Hung and Subramanyam, 2007). Some studies are inconclusive (Hung

and Subramanyam, 2007) or even provide evidence for a higher earnings quality of

German GAAP with reference to certain attributes (Gassen and Sellhorn, 2006).

Comparing the earnings attributes of IAS/IFRS and U.S. GAAP, studies �nd a

higher earnings quality for U.S. GAAP (Bartov et al., 2005; Barth et al., 2006b;

Goncharov and Zimmermann, 2006) or inconclusive results (Van der Meulen et al.,

2007). Reasons for the mixed results of these �accounting quality studies� might be

that the comparison of earnings attributes across accounting regimes could be biased

and that the self-selection of companies applying di�erent accounting regimes could

have confounded the results. Moreover, these studies only focus on certain summary

measures but neglect additional information included in �nancial statements, e.g.

the composition and presentation of assets or net income and the notes as well as

other information instruments like cash �ow statements. Finally, the implications

that can be drawn from certain measures are debatable (e.g., see Holthausen and

Watts, 2001, for value relevance studies).

A fourth stream of studies focuses on the capital market e�ects of the adoption

of international accounting regimes. To determine these e�ects, these studies rely

on various measures, like abnormal returns (Auer, 1996, 1998) stock price volatility

(Leuz and Verrecchia, 2000; Cuijpers and Buijink, 2005), bid-ask spreads (Leuz and

Verrecchia, 2000; Leuz, 2003b; Gassen and Sellhorn, 2006), percentage of trading

days (Gassen and Sellhorn, 2006) or analyst forecast based cost of equity capital

measures (Cuijpers and Buijink, 2005; Daske, 2006; Daske et al., 2007). No clear

conclusions have been drawn from these studies concerning the capital market im-

pact of the adoption of IAS/IFRS or U.S. GAAP.

Prior studies on the capital market e�ect of the adoption of IAAP in Germany

rest upon very speci�c samples (Leuz and Verrecchia, 2000) or on other time periods

(Daske, 2006). Moreover, all of these previous studies on the capital market e�ects

are conducted on a �rm basis, do not highlight the speci�c institutional background

of the country they examine and fail to explore the development of the impact

14


over time. Unlike these studies, we focus on a comprehensive sample of the entire

voluntary adoption period of IAAP in Germany between 1998 and 2004, shed light

on the possible e�ects of the institutional setting and of changes in this setting as

well as of accounting principles on the cost of equity capital impact and apply a new

methodology of measuring the impact of adoption of IAAP in Germany.

2.2.3 Measurement of cost of equity capital

A �rm's cost of equity capital is usually de�ned as the expected return on a �rm's

stock (e.g., Lambert et al., 2007). In other words, cost of equity capital is the mini-

mum rate of return investors require to provide equity capital to the �rm (Botosan,

2006). Researchers have suggested and applied a variety of means to measure cost

of equity capital, each with speci�c advantages and drawbacks. Besides indirect

measures or proxies (e.g., stock return volatility), capital market researchers apply

direct measures of cost of equity capital like residual income and discounted cash

�ow models (e.g., Gebhardt et al., 2001), the Capital Asset Pricing Model (CAPM)

(e.g., Fama and MacBeth, 1973), or multifactor models (e.g., Barth et al., 2008).

Many studies assume that the CAPM is descriptive and use market beta to proxy

for non-diversi�able risk. If so, beta does include any estimation risk. However, by

using historical data to proxy for expected market risk premium the CAPM treats

estimated parameters as if they were true, ignoring estimation problems. Therefore,

the overriding conclusion in the literature is that the CAPM is not descriptive, and

theory suggests that market beta does not capture estimation risk. Investor's un-

certainty is not taken into account (Botosan, 2006). The fundamental debate about

estimation risk being diversi�able (not priced) or non-diversi�able (priced) is still

ongoing, though. For example, one possible counter-argument is that information

relevance declines with the degree of diversi�cation in large populations (e.g., Cready

and Gurun, 2007).

In discounted cash �ow models, cost of equity capital can be described as the

risk-adjusted discount rate that investors apply to the expected future cash �ows in

order to derive the current stock price. Implementations of these models are e.g. the

Botosan and Plumlee (2002) model, based on the short horizon form of the classic

dividend growth model, as well as the Easton (2004) price-earnings growth ratio

model, based on the abnormal growth in earnings. Daske (2006) directly estimates

the expected cost of equity capital e�ects through the implied rate of return of a

15


residual income model utilizing �nancial analysts' consensus earnings forecasts and

stock prices. General shortcomings of all the discounted cash �ow models are to

determine the forecast horizon and the terminal value (Easton, 2006).

Moreover, the use of analysts' forecasts has further disadvantages. Forecasts for

�rms that have changed from domestic principles to IAAP may tend to have a

di�erent degree of optimism than forecasts for �rms that have not changed. These

forecasts of the expected rate of return are generally likely to be higher than the

real cost of equity capital (Easton, 2006). Generally, in all these models analyst

forecasts serve as proxies for market beliefs. One common critic, however, is that this

implies measurement errors, since analysts cannot perfectly re�ect market beliefs.

Consequently, these type of models regularly perform unsatisfactorily in tests of

construct validity (e.g., Easton and Monahan, 2005). In addition, analysts typically

only follow companies with a high visibility or market capitalization which might

induce a selection bias (Francis et al., 2004).

In the meanwhile, many researches have proclaimed to extend the classical CAPM.

The revolutionary work for today's research practice was the three-factor model

suggested by Fama and French (1992, 1993), which outdated the classical CAPM.

Based on factor-mimicking portfolios, they showed that the CAPM beta was not

an e�ective or insightful model in their U.S. market studies and introduced new

regressors indicating a value premium which compensates the risk missed by the

CAPM: the �small-minus-big� factor (SMB), which represents the �rms' size in terms

of market capitalization, and the �high-minus-low� factor (HML) standing for the

ratio of book-to-market value. The factor-mimicking portfolio approach introduced

by Fama and French was also applied to accounting oriented research. Francis et al.

(2005a) create an accruals quality factor-mimicking portfolio (AQ factor) to estimate

asset-pricing regressions. Ecker et al. (2006) proposed their �e-loadings� concept as

returns-based representation of earnings quality. They see information uncertainty

as a non-diversi�able (priced) risk factor, gaining theoretical support from Easley

and O'Hara (2004), as well as Leuz and Verrecchia (2005).

In our study, we use both the CAPM and a factor-mimicking portfolio approach

where we incorporate the di�erences between the accounting regime applied as a new

factor in our model. This allows us to prevent measurement errors and selection bias

which might be present in analyst-based cost of equity capital estimates. To our

knowledge, no other study has applied such a model in this context before.

16

2.3 Institutional background


2.3.1 German �nancial reporting requirements

In Germany, accounting principles and rules are not released by a private standard

setter, but are enacted by the legislature and codi�ed in the German Commercial

Code (Handelsgesetzbuch, HGB). It is accompanied by standards and norms estab-

lished by court decisions or by the reporting practice. German GAAP encompasses

all codi�ed and non-codi�ed rules, standards, and norms a company has to observe

when preparing �nancial statements (Leuz and Wüstemann, 2004).

All German companies are required to provide individual �nancial statements ac-

cording to German GAAP for the legal entity, which are the basis to determine the

distributable income, for deriving the taxable income, and for other legal provisions

(Haller and Eierle, 2004). In addition, parent companies having one or more sub-

sidiaries are obliged to prepare consolidated �nancial statements. Basically, in the

consolidated �nancial statements also German GAAP had to be applied until 2005.

In the mid 90s, German multinational companies have started to apply IAS and

U.S. GAAP2 due to a cross-listing in the U.S. or due to a perceived need of a more

investor oriented reporting (Haller, 2002). Ultimately, this forced the German legis-

lator to enact the Capital Raising Facilitation Act (Kapitalaufnahmeerleichterungs-

gesetz, KapAEG) allowing publicly listed companies to report consolidated �nancial

statements according to IAAP and consequently substituting the provisions of Ger-

man GAAP (� 292a HGB). However, companies preparing consolidated �nancial

statements under U.S. GAAP in accordance with this option were generally not

obliged to comply with the disclosure requirements of the SEC (Wüstemann, 2001)

and were not subject to the enforcement of the SEC, unless they were cross-listed

in the U.S.

Since the enactment of the Capital Raising Facilitation Act the number of listed

companies in Germany exercising this option to adopt IAAP has increased. More-

over, the listing regulations of the New Market (Neuer Markt), a market segment

of the German Stock Exchange for growth �rms between 1997 and 2003 required

companies to apply IAAP (Glaum and Street, 2003).

2Thereby, companies adopted di�erent reporting strategies, e.g. a parallel reporting, providing twofull sets of �nancial statements or reconciliations of income and shareholders' equity (Leuz andVerrecchia, 2000).

17


Since 2005, all publicly traded European companies (including those in Germany)

are required to prepare consolidated accounts under IFRS according to the IAS Reg-

ulation EC No. 1606/2002 (with a few exceptions).3 Due to the so called �member

state options� of the regulations the German legislator has allowed companies to

provide additional individual accounts under IFRS (besides the individual accounts

under German GAAP) for publication purposes and has passed the option to apply

IFRS for consolidated accounts to all non-publicly traded companies.

2.3.2 Accounting standards under investigation

Fundamental di�erences exist between general properties of German GAAP and

IAAP. Firstly, IAAP are developed by a private standard-setting body within a

speci�ed due process, whereas in Germany the parliament owns the standard-setting

authority for accounting rules. Even though in 1998 the German Accounting Stan-

dards Board (GASB) was founded, the private-sector standard setting power of

this board is still restricted to developing recommendations for consolidated �nan-

cial statements and non-compliance with these recommendations is not sanctioned

(Sellhorn and Gornik-Tomaszewski, 2006).

Second, German GAAP is more strongly principles based and o�ers more explicit

choices (e.g., for the treatment of goodwill) than the IAAP. However, until recently

important areas (e.g., stock options) were not (su�ciently) covered by standards

and/or interpretations under IAS/IFRS. Furthermore, some provisions of IAS/IFRS

and U.S. GAAP are far more complex in comparison to German GAAP. For example,

the revenue recognition according to the percentage of completion method is more

complex than the completed contract method, the treatment of actuarial gains and

losses from pension obligations is more complex than the general rule to recognize

such adjustments at once, and the impairment test (especially for cash generating

units) is more complex than a simple write-down to the replacement costs. However,

it has to be taken into account that in certain areas like revenue recognition issues

German GAAP is likely to become complex as well when tax law, court decisions,

and particular standards (like GAS, which can be at least factually binding) have

3Companies publicly traded both in the European Union and on a regulated third-country marketand which are therefore applying another internationally accepted accounting system (especiallyU.S. GAAP) in their consolidated accounts are allowed to defer the application of IFRS until2007. This also holds for companies which only have publicly traded debt securities.

18


to be considered.

Third, German GAAP is � in contrast to the other accounting regimes investigated

� considerably in�uenced by tax considerations. Consolidated �nancial statements

following German GAAP are derived from individual �nancial statements, which

are closely tied to the tax accounts and serve as basis for determining dividend

restrictions. Due to the so-called �congruency principle� or �authoritativeness prin-

ciple� (Maÿgeblichkeitsprinzip) the determination of accounting income and taxable

income are directly interrelated (Pfa� and Schröer, 1996). This principle has pri-

marily an impact on the individual accounts of companies. Until 2002, it was also

possible to include tax-induced accounting practices into the consolidated accounts.

In 2002, however, the Transparency Act abolished this option.

Fourth, under IAAP several items of income or expenses are recognized directly

in equity (e.g., foreign currency translations SFAS 52.13/SFAS 52.20/SFAS 52.46;

IAS 21.32/21.37/21.39(c)/21.45; cash �ow hedges SFAS 133.18(c), IAS 39.95; reval-

uations of available-for-sale �nancial assets SFAS 115.13/115.15/115.16, IAS 39.51/

39.55/IAS 39.57) and thus two di�erent performance measures are de�ned (i.e. net

income and comprehensive income). Under German GAAP only one item of income

or expense (i.e. foreign currency translations) is recognized directly in equity leading

to di�erences in the adherence to the clean surplus relation.

Finally, U.S. GAAP and IAS/IFRS clearly focus on providing an undistorted

picture of the �nancial position of a company. Yet, German GAAP aim at investor

protection and are largely biased by the �principle of prudence�. This divergence in

the objectives leads to di�erent recognition, measurement and disclosure provisions.

Over time, several changes in the accounting regimes have occurred. However, it is

di�cult or rather impossible to determine the impact of these changes on the quality

of �nancial statements and on the cost of equity capital of companies individually.

By focusing on both the quantity and quality of the revision of standards or the

issuance of new standards we determine two crucial points of time where a major

change of at least one investigated accounting regime occurred. First, in 1998 under

U.S. GAAP a new standard for the disclosure of comprehensive income, under IAS

a revision of IAS 1 and under German GAAP new requirements for cash �ow state-

ments and segment reports became e�ective. Second, in 2000 new provisions for

derivatives and hedge accounting under U.S. GAAP and IAS 36�IAS 39 as well as

a revision of IAS 16 became e�ective. Third, the new standard for goodwill and six

19


other standards became e�ective under U.S. GAPP and under German GAAP the

option to include tax-induced accounting practices into the consolidated accounts

was abolished under German GAAP. Finally, in 2005 the so-called �Improvements

Project� changing 13 standards, two revisions of other standards and four new stan-

dards became e�ective under IFRS and the Accounting Reform Act brought several

revisions and new requirements for German GAAP.

2.3.3 German corporate governance and enforcement system

The corporate governance system in Germany is fundamentally di�erent to the

Anglo-Saxon system. These di�erences are due to di�erent legal systems or cul-

tural peculiarities. Essentially, Germany has a civil or code law system in contrast

to the common law system in the U.S. (e.g., Haller and Walton, 2003) and it is char-

acterized by a relatively high degree of uncertainty avoidance as well as collectivism

in comparison to other countries (Hofstede, 1984).

The corporate governance system of a German joint stock corporation, which

is the legal structure of nearly all listed companies,4 is often characterized as be-

ing insider-controlled and stakeholder-oriented (Schmidt, 2004). The joint stock

corporations have a two-tier system with a management board (Vorstand) for the

executive management of the company and a separate supervisory board (Aufsicht-

srat) for the overseeing of the management board.5 As several di�erent stakeholder

groups are represented in the supervisory board, the German governance system

is often characterized as stakeholder-orientated, where internal control mechanisms

play a central role (Franks and Mayer, 1997; Hackethal et al., 2005).

Until 2005, besides statutory auditors, no external enforcement mechanism for

overseeing the compliance of companies with accounting standards had been in

place. Auditors published a short-audit report to the public and provided a long-

audit report to the supervisory board. Based on this report, the supervisory board

assessed the compliance of the �nancial statements with the accounting rules and

4Some companies are partnerships limited by shares (Kommaditgesellschaft auf Aktien, KGaA)having at least one personally liable partner, e.g. Henkel KGaA or Merck KGaA. In 2004 theEuropean Company (Societas Europea, SE ) was introduced as a legal structure for German com-panies by the legislature. In 2006 and 2007 respectively, Allianz and Fresenius were the �rstcompanies to adopt this legal structure in Germany.

5The SE provides companies an option to establish a one-tier or two-tier board system (Noack andZetzsche, 2005).

20


the appropriateness of the accounting policies applied (Naumann, 2000).

During the last decade, several legal changes have a�ected these corporate gov-

ernance mechanisms. The monitoring of the management board by the supervi-

sory board has been improved by the Law for the Strengthening of Control and

Transparency (Kontroll- und Tansparenzgesetz, KonTraG) (Nietsch, 2005) in 1998

and the Transparency Act (Transparenz- und Publizitätsgesetz, TransPuG) in 2002.

The KonTraG included audit reforms changing the objective of the audit as well

as the reporting requirements and the legal liability for auditors. These reforms

increased the monitoring role of audits in Germany (Gassen and Ashbaugh-Skaife,

2009). Moreover, due to an amendment of the law for commercial stock companies

the compulsory establishment of risk management systems was required (� 91 II of

the Stock Corporation Act, Aktiengesetz, AktG).

A corporate governance code has been implemented providing standards of best

practice in 2002 (Nietsch, 2005; Noack and Zetzsche, 2005). In � 161 AktG an obli-

gation to �comply-or-explain� was included which should facilitate the acceptance of

these standards. However, some argue that these reforms have brought no structural

change for the governance mechanisms described above (Nietsch, 2005) or that can

rather be seen as a �marketing instrument� which should increase the attractiveness

of German shares to international investors (Noack and Zetzsche, 2005).

Two important reforms have had an impact on the enforcement system. First, the

Accounting Reform Law of 2004 (Bilanzrechtsreformgesetz, BilReG) implemented

certain measures strengthening the independence of statutory auditors and of mod-

ifying the audit report (�� 318�322 HGB) and the Auditor Oversight Act (Ab-

schlussprüferaufsichtsgesetz, APAG) established the Auditor Oversight Commission

(Abschlussprüferaufsichtskommission, APAK ) for overseeing the statutory auditors

(Haller et al., 2006). Second, an external enforcement system was established due

to the Accounting Law Control Act of 2004 (Bilanzkontrollgesetz, BilKoG) by in-

cluding �� 342b�342d HGB into the German Commercial Code. This two-step sys-

tem comprises a privately organized enforcement body called Financial Reporting

Enforcement Panel (Deutsche Prüfstelle für Rechnungslegung, DPR) and a state

authority called Supervisory Authority for Financial Services Institutions (Bunde-

sanstalt für Finanzdienstleistungsaufsicht, BaFin) (Delvaille et al., 2005; Noack and

Zetzsche, 2005).

21


2.3.4 Germany's capital market

Only a small proportion of German companies is publicly listed on a stock exchange.

While there exist about one million limited liability companies in Germany, only

15.000 stock corporations are registered, from which approximately 1.000 are listed

on regulated markets. Most companies in Germany, especially smaller companies,

are held privately (Noack and Zetzsche, 2005).

Traditionally, the German capital market is often seen as bank-based (Baetge

et al., 1995; Haller and Walton, 2003; Vitols, 2005). A major part of debt and also

equity �nancing is provided by few dominant universal banks, the so-called �house

banks� (Hausbanken) (Elsas and Krahnen, 2004). Besides being major creditors of

companies, banks also hold large stakes of the companies' equity, can increase their

in�uence by acting as proxies of their clients using depositary voting rights, and

play a key role in the internal corporate governance of companies (Fohlin, 2005).

Moreover, many cross-holdings exist between publicly listed companies (Schilling,

2001), leading to a high ownership concentration in comparison to other countries

(Hackethal et al., 2005; Enriques and Volpin, 2007). Being a typical bank-based

system, households asset are largely held as bank deposits and not as investments

in shares (Vitols, 2005).

Since the foundation of the New Market in 1997, the role of equity in the �nancing

of companies has become more important (e.g., the number of IPOs has increased)

and the number of stockholders has increased (Vitols, 2005). After the burst of the

capital market bubble in 2001, the public interest in stock investments has declined.

In particular, since a tax reform in 2000 the in�uence of banks on listed companies

by holding large equity stakes has decreased (Vitols, 2005; Hackethal et al., 2005).

However, to a certain extent, insurance companies have replaced banks in their role

as dominant shareholders of German companies (Vitols, 2005). Thus, raising equity

is still less important for the external �nancing of companies and shareholdings of

households are still signi�cantly lower than in other countries. Consequently, the

general capital market situation in the German capital market has not changed

structurally so far (Hackethal, 2004; Vitols, 2005)) and the �nancial system can still

be regarded as bank-based (Hackethal et al., 2005).

22

2.4 Hypotheses development


2.4.1 Overall impact of adopting IAAP

We investigate the impact of a voluntary adoption of IFRS or U.S. GAAP by German

companies in the entire period examined (1998�2004) as well as in several sub-

periods. The relation between applying a speci�c accounting regime and the cost of

equity capital is complex and in�uenced by several factors. Lambert et al. (2007)

argue that information provided to investors might have an indirect and a direct

e�ect on the cost of equity capital. Using a similar reasoning, the adoption of IAAP

could have an indirect impact on the cost of equity capital by improving the quality

and quantity of information, lowering information asymmetry (Gassen and Sellhorn,

2006), improving the liquidity of a company's shares by enlarging the investor base

(Merton, 1987; Covrig et al., 2007), and �nally lowering the compensation required

by uninformed investors in terms of returns, which means a lower cost of equity

capital (Easley and O'Hara, 2004).

In addition, direct cash �ow impacts of adopting IAAP could arise. On the one

hand, this could be negative impacts due to the costs of adoption as well as of the

application of the more complex IAAP. On the other hand positive impacts could

emerge because the adoption of IAAP could improve the brand recognition as well

as the recruiting of international employees, alleviate international co-operations

or acquisitions and foster the implementation of value-based management systems

(Weiÿenberger et al., 2004).

Three major problems impede evaluating these e�ects separately. First, in�uential

factors, like the incentives of managers and auditors (Ball et al., 2003; Ewert and

Wagenhofer, 2005; Gassen and Sellhorn, 2006), the expertise and capabilities of

managers, auditors and users of �nancial statements, other institutional settings

like corporate governance or enforcement and the importance or the integration of

the capital market (Hail and Leuz, 2006) could have diverse impacts on the (indirect

and direct) relations between standards and cost of equity capital for the di�erent

accounting regimes. Especially, the low rate of listed companies in Germany in

comparison to other countries and the minor percentage of people holding shares

might decrease the opportunities for diversi�cation and therefore in�uences pricing

of estimation risk. This is in particular true when a strong home bias towards

domestic stocks prevails. For Germany, such a bias is found in several studies (e.g.,

23


Tesar and Werner, 1995; Kilka and Weber, 2000). In contrast, the integration of

the stock markets in Europe and even worldwide could mitigate this e�ect (Harvey,

1991).

Second, these e�ects might interact with each other (Gietzmann and Trombetta,

2003) and with the factors explained, which hampers their exploration. E.g. the

interaction of accounting standards and of accounting practice is di�cult or rather

impossible to disentangle (Schipper, 2005; Sellhorn and Gornik-Tomaszewski, 2006).

Third, concepts like earnings quality, disclosure level, or information asymmetry

are unobservable and thus have to be measured by proxies (e.g., earnings quality by

persistence, predictability, conservatism, timeliness, discretionary accruals or value

relevance). Moreover, these concepts disregard the direct cash �ow impacts of adopt-

ing IAAP. This makes it di�cult to unambiguously investigate the overall impact

of adopting IAAP.

We assume, that the cost of equity capital is an important objective for adopting

IAAP. Several survey studies document this motive for adopting IAAP (Pellens and

Tomaszewski, 1999; Weiÿenberger et al., 2004). Also the EU and standard-setters

like the FASB aim at lowering the cost of equity capital when deciding about what

accounting standards should be applied. Moreover, only the cost of equity capital

is able to capture indirect and direct e�ects which pertain to the adoption of IAAP.

Thus we investigate the overall link between the adoption of IFRS or U.S. GAAP

and the cost of equity capital.

The impact of adopting IAAP in Germany is di�cult to predict. The application

of IAAP should re�ect the performance of a company more timely and with a lower

degree of conservatism. Moreover, the higher extent of value relevant items recog-

nized especially under IAS/IFRS (e.g., intangible assets) and the more timely and

less conservative measurement of items under IAS/IFRS and U.S. GAAP (e.g., at

fair value) should improve the ability of investors to predict future cash �ows and

thus lower estimation risk. The higher extent of explanatory notes and of additional

disclosures required by IAS/IFRS and U.S. GAAP in comparison to German GAAP

should all else equal lower the degree of information asymmetry. A company's de-

cision to switch to IAAP could be regarded as a strong commitment to increased

disclosure because it is very costly to reverse (Leuz and Verrecchia, 2000; Daske,

2006). As IAS/IFRS and U.S. GAAP are explicitly directed at investors, the adop-

tion of these accounting regimes in Germany should cause an exchange of private

24


information granted to certain stakeholders represented on the supervisory board

for public information (Daske, 2006). This should have a decreasing e�ect on the

degree of asymmetric information. However, (Francis et al., 2005b) argue that the

need for public information in bank-based �nancial systems like Germany is lower

than in market-based systems and that weak investor protection might impair the

credibility of the information provided by IAS/IFRS or U.S. GAAP. Therefore, these

e�ects are expected to be favorable for IAAP, but could turn out to be relatively

low in Germany.

In contrast, previous studies �nd that principles based accounting standards lead

to a higher earnings quality Webster and Thornton (2005) suggesting all else equal

a higher earnings quality of German GAAP and partly IAS/IFRS in comparison

to U.S. GAAP. The reduced comparability of �nancial statements under IAS/IFRS

and U.S. GAAP due to missing speci�cation of a reporting format for the income

statement and balance sheet could deteriorate the quality of information about Ger-

man companies available to investors. In addition, the adoption of IAS/IFRS or

U.S. GAAP might only be the use of a label but does not result in a material

change of the transparency of �nancial statements (Daske et al., 2007). This might

hold particularly in countries like Germany where individual shareholders have less

in�uence on the governance of the management and managers have more room for

pursuing their interests (Ball, 2006). Furthermore, the enforcement of IAS/IFRS or

U.S. GAAP �nancial statements might have been more di�cult for the statutory

auditors and the supervisory boards especially in the �rst years of adoption because

they lacked su�cient expertise in the new accounting regimes. Glaum and Street

(2003) provide empirical evidence on this notion. Following the arguments of Kim

and Verrecchia (1994), the adoption of the more complex accounting regimes IFRS

and U.S. GAAP could even have increased the information asymmetry because in-

formed investors are able to gain more insights than less-informed investors. This

argument might particularly hold for non-institutional investors in Germany, as they

rather neglect additional disclosures and focus only on the balance sheet and income

sheet (Deutsches Aktieninstitut (DAI), 2005). However, even analysts might have

di�culties in using IFRS or U.S. GAAP for earnings forecasts (Daske, 2006). A

further disadvantage of the adoption of IAAP could be that the previous domestic

standards have more e�ectively addressed the speci�c needs of �nancial statement

users or have accommodated the particular legal or economic system of a country

25


(Armstrong et al., 2010).

Based on this discussion, it ultimately remains an empirical question whether

over the examined period the adoption of IAS/IFRS and U.S. GAAP by German

companies has decreased their cost of equity capital. We therefore state the following

hypothesis:

H1: In Germany, the cost of equity capital is higher for companies ap-

plying German GAAP than for companies applying IAS/IFRS or U.S.

GAAP.

2.4.2 Impact of adopting IAAP in subperiods

In a second analysis, we examine whether and how the changes in the accounting

standards as well as in the German capital market, corporate governance system,

and enforcement system could have an impact on the cost of equity capital impact

of applying IAAP. As the description of the institutional background in the previous

section shows, several changes have taken place during the sample period. These

changes likely in�uence the di�erent e�ects of an adoption of IAS/IFRS or U.S.

GAAP. As explained above, in the years 1998, 2000, 2002 and 2005 major revisions of

accounting standards or new standards became e�ective. Moreover, the years 1998,

2002 and 2005 brought reforms for the corporate governance and the enforcement

system in Germany. Consequently, in the investigated period three relatively stable

sub-periods can be identi�ed: (1) 1998�1999, (2) 2000�2001 and (3) 2002�2004.

At the beginning of the �rst sub-period (1998�1999), several new disclosures and

presentation requirements under IAS and U.S. GAAP became e�ective which might

have a positive impact on the cost of equity capital e�ect of applying this two ac-

counting regimes. Moreover, the capital markets in Germany became more popular

especially to non-institutional investors which might have increased the demand for

more decision useful information for investors. This would ceteris paribus also imply

a positive impact of an adoption of IAAP, because German GAAP was regarded as

being less suitable for those purposes. In contrast, the IAS provided many options

and had no standards for several important issues in this time period. Moreover,

the compliance level of companies applying IAAP in Germany, especially in the �rst

years of application was low, since companies and auditors were not used to the new

accounting regimes. In addition, many analysts and investors in Germany might

26

2.5 Research design

not have been able to cope with the more complex provisions.

In 2000 which is the beginning of the second sub-period, several new standards as

well as revisions of key standards became e�ective under IAS, restricting options or

�lling important gaps of missing rules. Furthermore, companies, investors, �nancial

analysts, as well as auditors had become used to the provisions of IAAP and thus

were able to exploit the higher degree of transparency. In this time period vari-

ous IPOs took place and the interest in the investor-oriented accounting standards

further increased.

In the third sub-period (2002�2004), the Transparency Act became e�ective and

abolished the option under German GAAP to include tax-induced accounting prac-

tices into the consolidated accounts. Under U.S. GAAP the new provisions for

measuring goodwill are rather complex and provide a considerable degree of discre-

tion to managers. Concerning IFRS, a revised standard for employee bene�ts (IAS

19 (rev.) in 2002) became e�ective which might have an impact on the transparency

and thus on the cost of equity capital. The creditworthiness of accounting and es-

pecially of the IAAP was damaged due to some accounting scandals of companies

listed in Germany's New Market and applying IAS/IFRS or U.S. GAAP.

Concerning the cost of equity capital e�ects of applying IAAP in the three sub-

periods presented, we state the following hypothesis:

H2: The cost of equity capital impact of applying IAS/IFRS or U.S.

GAAP in comparison to that of applying German GAAP is di�erent in

the time periods 1998�1999, 2000�2001 and 2002�2004.

2.5 Research design

2.5.1 Portfolio analyses

The starting point for our analysis is the classical CAPM. This model describes how

the market return above risk free rate explains a stock or portfolio of stocks. The

CAPM is described by the following equation:

rit − rrft = αi + βi(rmt − rrf

t ) + εit (2.1)

where rit represents the individual stock (or portfolio) i at time t, rrft indicates the

risk-free interest rate and rmt the market return at time t. The Greek letters stand

27


for the intercept αi, the slope parameter βi, and the residuum εit of each individual

stock (or portfolio) i.

However, for voluntary changes in the disclosure level, the CAPM results could

be a�ected by a self-selection bias. When variables like company characteristics

explaining the decision of managers to change the disclosure level are omitted in

the analysis and are correlated to certain priced risk factors, the results are biased

(Hail, 2002). One way to mitigate this problem of self-selection is to include known

risk factors like market capitalization and market-to-book value into the analysis

(Berk, 1995). Therefore, we use an enhanced multifactor model gaining theoretical

support from Francis et al. (2004) as well as Francis et al. (2005a) and control for

these known priced risk factors. One assumption of the model is that there exists

an information factor in our sample which could not be diversi�ed away and thus is

priced. A second alternative for controlling a possible self-selection bias is a two-step

regression approach, which is applied in the �rm-level analyses in the next section.

The multifactor model we use in the following is based on the original Fama and

French model, with the following multifactor equation:

rit − rrft = αi + βi1(rmt − rrf

t ) + βi2SMB t + βi3HMLt + εit (2.2)

The �rst part of the Equation 2.2 is similar to the Equation 2.1, with the only

di�erence that the Greek slope parameters βij are now numbered from j = 1, 2, 3

for the three regressors (rmt − rrft ), SMB t, and HMLt. SMB t (�Small minus Big�)

represents the size of the companies. HMLt stands for a factor based on the book-

to-market ratio. Firms with high (low) book-to-market values are regarded as �value

stocks� (�growth stocks�) (Fama and French, 1993).

Similar to Fama and French (1993) we split our sample of companies into six

portfolios:

SH: �Small-High� (small size, high book-to-market ratio)

SM: �Small-Medium� (small size, medium book-to-market ratio)

SL: �Small-Low� (small size, low book-to-market ratio)

BH: �Big-High� (big size, high book-to-market ratio)

BM: �Big-Medium� (big size, medium book-to-market ratio)

28

2.5 Research design

BL: �Big-Low� (big size, low book-to-market ratio)

We rank our sample �rst according to the size in terms of market value. The

median discriminates between the �small� and �big� �rms. Secondly, we rank our

sample according to the book-to-market ratios and separate the highest 30% as

�high�, the lowest 30% as �low,� and the resulting 40% as �medium.�

The companies remain in one of the six portfolios for one year (starting in July,

ending in June of the next calendar year). Reference date for the re-alignment of

the portfolios is the end of June of the previous period for the size, and the end of

December of the previous period for the book-to-market ratio. The monthly returns

of the companies are averaged � weighted with their market value � in each of the

six portfolios. We obtain: rSH , rSM , rSL, rBH , rBM , rBL.

Eventually, we compute SMB t, and HMLt as follows:

SMB t = (rSH + rSM + rSL)/3− (rBH + rBM + rBL)/3 (2.3)

HMLt = (rSH + rBH)/2− (rSL + rBL+)/2 (2.4)

With these factors, the in�uence of size on book-to-market is reduced and vice-

versa.

Based on the Fama and French approach in Equation 2.2, we incorporate our

new model (that we call �GM model�), which is augmented with two new factors,

representing the accounting regime di�erences:

rit−rrft = αi +βi1(rmt−rrf

t )+βi2SMB t +βi3HMLt +βi4GMI t +βi3GMU t +εit (2.5)

The factors GMI t and GMU t explain the accounting regime impact. GMI t is

the return di�erence between the portfolios of companies using German GAAP and

IAS/IFRS (�German GAAP minus IAS/IFRS�) and GMU t is the return di�erence

between German GAAP and U.S. GAAP (�German GAAP minus U.S. GAAP�)

portfolios. For this analysis, we calculate the market value weighted means for each

of the three accounting regime groups and compute the di�erence for the �GM�-

factors as described for each month in the sample.

For our new �GM model�, we have to extend the number of portfolios according

29


Table 2.1: 18 analyzed portfolios

Book-to-Market

Low Medium High

German GAAPSize Small SLG SMG SHG

Big BLG BMG BHG

IAS/IFRSSize Small SLI SMI SHI

Big BLI BMI BHI

U.S. GAAPSize Small SLU SMU SHU

Big BLU BMU BHU

Notes: The 18 portfolios are built upon three criteria: (1) accounting regime applied (G: GermanGAAP; I: IAS/IFRS; U: U.S. GAAP), where companies are assigned into the July-to-June year bythe accounting regime they used at the reporting date within that period; (2) book-to-market value(L / M / H: Low / Medium / High), where the book value of equity is divided by the market valueof equity; we rank our sample according to the book-to-market ratios and separate the highest 30%as �high�, the lowest 30% as �low,� and the resulting 40% as �medium�; (3) size in terms of marketvalue of equity (S / B: Small / Big), where the median discriminates between the �small� and �big��rms. The companies remain in one of the 18 portfolios for one year (starting in July, ending inJune of the next calendar year). Reference date for the re-alignment of the portfolios is the end ofJune of the previous period for the size, and the end of December of the previous period for thebook-to-market ratio.

to the accounting regime and obtain the new 18 factor-mimicking portfolios, as

illustrated in Table 2.1.

The companies are assigned into the July-to-June year by the accounting regime

they used at the reporting date within that period. For example, if a company's

reporting date was December 31, 2000, the accounting regime they applied at that

point was assigned to the July 2000-to-June 2001 year.

Within these 18 portfolios, we again construct the market value weighted average

of the monthly returns. Ultimately, we computed our new factors as follows:

30

2.5 Research design

GMI t = (rSHG + rSMG + rSLG + rBHG + rBMG + rBLG)/6 (2.6)

−(rSHI + rSMI + rSLI + rBHI + rBMI + rBLI)/6

GMU t = (rSHG + rSMG + rSLG + rBHG + rBMG + rBLG)/6 (2.7)

−(rSHU + rSMU + rSLU + rBHU + rBMU + rBLU)/6

The idea of these factors is that, provided that these di�erences are signi�cant,

they help us to explain the performance of the returns of our 18 portfolios and

account for an information factor related to the di�erent disclosure levels among the

three accounting regimes examined. Knowing the accounting regime of a portfolio,

this should lead to greater explanatory power within the model.

We apply the seemingly unrelated regression (SUR) technique. Since it would be

unrealistic to expect that the equation errors are uncorrelated, this method explic-

itly allows analyzing a system of multiple equations with cross-equation parameter

restrictions and correlated error terms.6

2.5.2 Firm-level analyses

As already mentioned, dealing with voluntary adoption of IAAP can lead to a po-

tential self-selection bias between the di�erent accounting regime portfolios. One

way of tackling this issue is the concept of including an information quality factor in

the factor-mimicking portfolio analysis. But researchers have also successfully im-

plemented another means of addressing the self-selection problem at the �rm-level.

By applying the two-equation procedure, proposed by Heckman (1978), we can con-

trol for self-selection by incorporating the Inverse Mills Ratio (see, e.g., Leuz and

Verrecchia, 2000; Gassen and Sellhorn, 2006; Hung and Subramanyam, 2007).

In the �rst stage, we estimate a probit model to analyze the �rms' probability to

adopt IAS/IFRS or U.S. GAAP given a variety of explaining factors:

6As a robustness check, we also apply the ordinary least squares (OLS) method. The inference iseven stronger for OLS, but for the technical reasons mentioned we nevertheless incorporate theSUR method.

31


IAAP it = Probit(δ0 + δ1log(ME it) + δ2ROAit + δ3CAPINT it (2.8)

+δ4MANUF it + δ5NEWMARKET it + δ6USUK it + εit)

where for �rm i and time t IAAP it is a dummy variable for applying IAAP (i.e.

IAS/IFRS or U.S. GAAP), log(ME it) is the natural logarithm of the market equity,

ROAit is the return on assets, CAPINT it is the capital intensity (long term assets

divided by total assets), MANUF it is a dummy variable indicating if the company

is a manufacturing company (SIC < 4000), NEWMARKET it is a dummy for being

included in the New Market (Neuer Markt) segment of the Frankfurt stock exchange,

and USUK it is a dummy for indicating whether the company is cross-listed at the

U.S. or U.K. market.

Using this �rst stage probit estimation, we can compute the Inverse Mills Ratio

(λit) to account for self-selection in the second stage.

In the second stage, we analyze whether the adoption of IAAP (indicated by the

variable IAAP it) signi�cantly in�uences the cost of equity capital estimates which

we derived based upon our factor-mimicking models. The important di�erence here

is that we simultaneously control for self-selection bias and other e�ects included in

the �rst stage regression (e.g., cross-listing or New Market membership). Therefore,

we estimate in the second stage:

CoEC mit = ϕ0 + ϕ1IAAP it + ϕ2log(ME it) + ϕ3λit + εit (2.9)

where for �rm i and time t CoEC mit is the cost of equity capital calculated by method

m (CAPM or GM model), IAAP it is a dummy variable for applying IAS/IFRS or

U.S. GAAP, log(ME it) is the natural logarithm of the market equity, and λit is the

Inverse Mills Ratio. Even though we proclaim to have speci�ed the cost of equity

capital estimation model in the factor-mimicking section already, this procedure

can be meaningful as a ceteris paribus analysis for cost of equity capital versus the

adoption of IAAP.

32

2.6 Sample selection and descriptive statistics



We use data concerning the German stock market, sampled monthly from July 1997

to June 2005 (96 months). This ensures that only companies fully applying IFRS

or U.S. GAAP regulations are included as IAAP companies, whereas prior to that

period data quality was poor � e.g., reconciliations only. The source of the stock

returns (adjusted prices), number of common shares outstanding, and book values

is the Datastream Advance database. For the risk-free rate we apply the 3-month

interest rate of �rst the German Bundesbank (July 1997�December 1998), and after

that the EURIBOR 3-month rate as the money market reference rate for the Euro.

The type of accounting regime applied and the �scal year end data are hand collected

from the annual reports of the companies, as we found several missing or mistakable

entries in the Worldscope database (see Daske et al., 2007, for details). In accordance

with former studies (e.g., Ziegler et al., 2007) we exclude companies with a negative

book value. Also �nance and insurance companies are excluded from the sample

based on the Standard Industrial Classi�cation (SIC 6000 to 6999) since their book

values of equity are fundamentally di�erent to those of non-�nancial companies. No

company is allowed to have an interrupted time series. This leaves us with a �nal

sample of 548 companies that we assign into 18 portfolios. To minimize the bias

through outliers, we winsorize the return data at the 1% and 99% level respectively.

Table 2.2 shows the descriptive results of our analysis for the 18 portfolios, with

the mean excess returns, their standard deviations, the number of monthly observa-

tions, and the average number of �rms per month. Figure 2.1 illustrates the timely

development of the pooled portfolios into the three main classi�cations German

GAAP, IAS/IFRS and U.S. GAAP.

The �rst interesting �nding is that our data contain the Fama and French (1992)

value premium in the means. We �nd a positive impact of a high book-to-market

ratio on the (excess) returns. In our sample the mean of the 18 portfolios increases

with a rising book-to-market ratio, with only three exceptions (marked with † inTable 2.2), meaning that the value premium is empirically visible. Ziegler et al.

(2007) discover the same pattern for Germany as well.

However, the second e�ect, discovered by Fama and French, of a positive impact

of a small size (i.e. small market values have higher returns) cannot be con�rmed

33


Figure 2.1: Excess returns of the German GAAP portfolio, the IAS/IFRSportfolio, and the U.S. GAAP portfolio from July 1997 to June 2005 (market value

weighted)

34


Table

2.2:Descriptive

analysisforall18

portfolios:returns(July1997�June2005,max.96

observations)

Mean

StdDev

Obs

NMean

StdDev

Obs

NMean

StdDev

Obs

N

GL

MH

SSLG

-0.0072

0.0672

9617.4

SMG

0.0045

0.0423

9636.4

SHG

0.0095

0.0510

9643.0

BBLG

0.0069

0.0604

9629.1

BMG

0.0147

0.0630

9643.1

BHG

0.0146†

0.0561

9622.6

IL

MH

SSLI

-0.0180

0.1275

7219.0

SMI

-0.0042

0.1204

8422.0

SHI

0.0136

0.1105

8421.7

BBLI

0.0101

0.0572

9621.5

BMI

0.0121

0.0635

9624.0

BHI

0.0028†

0.0773

8417.6

UL

MH

SSLU

-0.0205

0.1601

8412.3

SMU

0.0014

0.1184

7214.0

SHU

0.0260

0.0982

6010.0

BBLU

0.0122†

0.1227

9611.4

BMU

-0.0039

0.0684

9611.8

BHU

0.0119

0.0979

726.5

Notes:

Theportfolioconstructionisdescribed

inTable2.1.Table2.2

show

sthedescriptive

resultsofouranalysisforthe18portfolios.

Wehavemonthly

return

observationsfrom

July

1997untilJune2005.Mean:Themonthly

returnsoftheportfoliosare

averaged

byusing

market

valueofequityweightsofthecompanies.StdDev:standard

deviationofthemeanreturns;Obs:observations=number

ofmonths

data

are

availableforthespeci�cportfolio;N:averagenumber

of�rm

sincluded

forcomputingthemean/standard

deviationper

observed

month.†�valuepremium�exceptions:

thesevalues

donotre�ecttheFamaandFrench

�valuepremium�;FamaandFrench

foundin

1992

thatreturnsgrowwithhigher

book-to-m

arket

values.

35


in our data, more to the contrary. This is in accordance with Schrimpf et al. (2006)

and Ziegler et al. (2007). In our opinion this can be seen as a typical German

phenomenon. In Germany, the tendency toward public o�erings is substantially

smaller compared to the United States or the United Kingdom where considerably

more small companies are publicly traded.


For the two stage approach, we use yearly �nancial statement data gained from the

DAFNE database7 for the period from 1998 until 2004.8 To have these data available

for all �rms, however, we have to reduce the sample from the �nal 548 from above

to 494 companies. The cross-listing data are taken from a study of the �Deutsches

Aktieninstitut�9 (Glaum et al., 2006). The accounting regime data are, again, the

hand-collected data from the portfolio analysis. We estimate the monthly cost of

equity capital for each �rm based on the CAPM and the GM model using the risk

factors from above (rmt − rrft , SMB t, HMLt, GMI t, and GMU t). Afterwards, we

average the monthly data for every year and multiply them by 12 (see, e.g., Fama

and French, 1997) to obtain yearly �rm-level cost of equity capital estimates.10

Descriptive statistics of all data are provided in Table 2.3. Most important to note

is that the sample consists almost equally of IAAP (1542) and non-IAAP (1777)

companies.

7DAFNE is a database of detailed �nancial information for 140,000 German and Austrian compa-nies, hosted by Bureau van Dijk Electronic Publishing (BvDEP), one of Europe's leading electronicpublishers of business information.

8Before 1998, data quality regarding international accounting standards is very low, since only fewcompanies published IAS or U.S. GAAP �nancial statements. After 2004, the adoption of IFRSwas mandatory for most of the publicly listed companies.

9Deutsches Aktieninstitut e.V. (DAI) is the association of German exchange-listed stock corpora-tions and other companies and institutions with an interest in the capital market. The DAI is anindependent, non-pro�t institution.

10We also estimate the �rm-level models for monthly cost of equity capital data. For the inferencehowever, there is no di�erence between the monthly average and the monthly average multipliedby 12, since this is only linear transformation.

36


Table 2.3: Descriptive statistics of yearly data (1998�2004), 494 �rms

Panel A: Continuous variables

Obs Mean Median Max Min Std

log(ME) 2910 11.965 11.756 19.193 6.674 2.053

ROA 2924 0.016 0.047 0.839 -4.310 0.209

CAPINT 2883 0.346 0.334 0.951 0.000 0.195

CoEC (CAPM) 3394 0.060 0.062 2.997 -1.588 0.271

CoEC (DM Model) 3088 0.050 0.072 3.976 -1.903 0.406

Inverse Mills Ratio 2468 0.496 0.563 0.667 0.288 0.122

Panel B: Discrete variables

Obs Obs with Dep=0 Obs with Dep=1

IAAP 3319 1777 1542

MANUF 3952 2688 1264

NEWMARKET 3952 2736 1216

USUK 3952 3672 280

Notes: Variable de�nitions: log(ME) is the natural logarithm of the market value of equity, ROAis the return on assets, CAPINT is the capital intensity (long term assets divided by total assets),CoEC (CAPM) is the cost of equity capital calculated by the CAPM, CoEC (DM Model) isthe cost of equity capital calculated by the DM model, InvMillsRatio is the Inverse Mills Ratiocalculated based on the �rst stage regression in Table 2.10. IAAP is a dummy variable for applyinginternationally accepted accounting principles (IAS/IFRS or U.S. GAAP), MANUF is a dummyvariable indicating if the company is a manufacturing company (SIC < 4000), NEWMARKET isa dummy for being included in the New Market (Neuer Markt) segment of the Frankfurt stockexchange, and USUK is a dummy for indicating whether the company is cross-listed at the U.S.or U.K. market.

37


2.7 Regression results


Regarding the classical CAPM results in Table 2.4, we can testify a fairly high

explanatory power. All slope parameters are signi�cant on a level of 1%. The

goodness-of-�t is notably high, resulting in an adjusted R2 of 40.08% on average,

respectively an adjusted R2 of 54.78% applying the market value weighted average.

Looking at the parameter estimates, we �nd that 9 of 12 comparisons between

the German GAAP betas and IAS/IFRS and U.S. GAAP betas indicate di�erences

for market beta (the other three are indicated with † in Table 2.4).

Computing the average of the three accounting regime groups, we can also state

that German GAAP portfolios have lower betas. Their average is 0.5131, compared

to 0.8890 (IAS/IFRS) and 1.1030 (U.S. GAAP). Weighted with the number of ob-

servations in each portfolio, we see the same pattern: the German GAAP average is

0.5298 versus 0.8644 (IAS/IFRS) and 1.1380 (U.S. GAAP). Finally, we also regress

the return di�erences between IAS/IFRS and German GAAP, as well as between

U.S. GAAP and German GAAP, on the market excess return in the CAPM. Test-

ing the one-sided hypothesis that the resulting betas are smaller than zero can be

rejected at a 5%-level for the �IAS/IFRS minus German GAAP� regressand, and at

a 1%-level for �U.S. GAAP minus German GAAP� regressand. This suggests that

both the IAS/IFRS betas and the U.S. GAAP betas are signi�cantly higher than

the German GAAP betas. These results (Table 2.5) provide evidence for the infor-

mation risk e�ect of introducing a �new� accounting regime and leading to higher

betas in the CAPM model.

As expected, the extension of the CAPM with the Fama and French factors SMBt

and HMLt , applying the typical Fama and French approach as our second model,

leads to an increase of the goodness-of-�t, indicated by the results in Table 2.6.

The adjusted R2 rises to 57.20%. However, in terms of the market value weighted

average, it goes up less substantially to 59.93% only. Again, all βi are signi�cant

at the 1% level. Out of our 18 portfolios, seven (eight) SMBt (HMLt) parameters

do not achieve the 1%-level. This can be explained by the correlation of these two

factors in our sample, which leads to collinearity in the regressors and reduces the

power of the model (Table 2.7).

Applying our new model, with the �German GAAP minus IAS/IFRS�-factor

38


Table

2.4:CapitalAsset

Pricing

Model:r i

t−rr

ft

=α

i+β

i(r m

t−rr

ft

)+ε i

t

αi

βi

Adj.R

2α

iβ

iAdj.R

2α

iβ

iAdj.R

2

GL

MH

SSLG

-0.0127**

0,3065

***0,0951

SMG

-0,0002

0,2945

***0,2068

SHG

0,0033

0,4219

***0,2911

BBLG

-0,0027

0,7091

***0,5713

BMG

0,0042

0,8387

***0,7181

BHG

0,0077

*0,5079

***0,3340

IL

MH

SSLI

-0,0372***

1,3864

***0,5126

SMI

-0,0155

1,0132

***0,3167

SHI

0,0015

0,9368

***0,3078

BBLI

0,0022

0,6355†***0,5066

BMI

0,0022

0,7675†***0,5951

BHI

-0,0046

0,5948

***0,2488

UL

MH

SSLU

-0,0401***

1,6401

***0,4758

SMU

-0,0167*

1,2004

***0,4686

SHU

0,0132

0,5944

***0,1684

BBLU

-0,0067

1,4431

***0,5747

BMU

-0,0121**

0,6439†***0,3531

BHU

0,0014

1,0962

***0,4783

Notes:


inTable

2.1.TheCapitalAsset

PricingModel

(CAPM)describes

how

the

market

excess

return

(rm

t−rr

ft

)explainstheexcess

return

ofaportfolio

(rit−rr

ft

).Theexcess

return

meansthedi�erence

betweentheactualreturnsandtherisk-freemarket

interestrate.Weapplytheseem

inglyunrelatedregression(SUR)technique.

Abbreviations:α

i:intercept;β

i:estimatedparameter

formarket

excess

return;Adj.R

2:adjusted

R2(G

oodness-of-Fit).*,**,

and***meansthatthesevalues

are

signi�cantatthe10%,5%,and1%

level.†Exceptionsto

thethesisthattheCAPM

betas

oftheGermanGAAPportfoliosare

smaller

thantheirequivalentsin

theIAS/IFRSandU.S.GAAPportfolios.

39


Table 2.5: Beta values comparison between the German GAAP �rms, theIAS/IFRS �rms, and the U.S. GAAP �rms

Average Average CAPM CAPM with new

beta beta value regression endogenous variables

value weighted with of the three (with t-statistic):number of combined I�G

observations portfolios U�G

G 0.5131 0.5298 0.6232

0.1089** (1,8415)

I 0.8890 0.8644 0.7255

0.3750*** (4,5772)

U 1.1030 1.1380 0.9723

Notes: G: German GAAP; I: IAS/IFRS; U: U.S. GAAP; CAPM: Capital Asset Pricing Model;I�G: IAS/IFRS portfolio minus German GAAP portfolio, regressed in CAPM model; U�G: U.S.GAAP portfolio minus German GAAP portfolio, regressed in CAPM model; *, **, and *** meansthat these values are signi�cant at the 10%, 5%, and 1% level (here one-sided test).

(GMIt) and the �German GAAP minus U.S. GAAP�-factor (GMIt) results in the

best explaining model. Applying the di�erences between the accounting regimes al-

lows us to estimate our new model stated in Equation 2.5. The results are depicted

in Table 2.8.

As before, the highest validity is implied in the βi parameters, followed by the

SMBt and the HMLt parameters. Even though our new regressors GMIt and GMIt

are individually signi�cant only in 11 of 18 cases, the overall explanatory power

increases again. The new averaged adjusted R2 with 63.51% exceeds the CAPM by

23.43%-points and the classical Fama and French model by 6.31%-points. In terms

of market value weighted numbers, it outperforms the CAPM by 7.78%-points and

the Fama and French model by 2.63%-points.

To draw a �rst conclusion, including accounting regime information into the Fama

and French model leads to an improvement of the explanatory power.11 The returns

of the portfolios in our model can be described more exactly, compared with the

CAPM and the Fama and French approach. Table 2.9 illustrates the consolidated

comparison of all three models.

As a sensitivity check, we also compare the Schwarz criterion (SC) for the three

11Using the adjusted R2, takes already into account the loss of degrees of freedom. Including twoadditional regressors does not automatically increases the adjusted R2.

40

2.7 Regression resultsTable

2.6:FamaandFrenchmodel:r i

t−rr

ft

=α

i+β

i1(r

mt−rr

ft

)+β

i2SM

Bt+β

i3H

ML

t+ε i

t

αi

βi1

βi2

βi3

Adj.R

2α

iβ

i1β

i2β

i3Adj.R

2

GL

MSSLG

-0.0037

0.3707

***

0.6780

***-0.1170

0.3388

SMG-0.0017

0.4787

***

0.5920

***

0.3913

***

0.5393

BBLG

0.0012

0.6645

***

0.0164

-0.1895**

0.5900

BMG

0.0026

0.7786

***-0.3024***-0.0630

0.7528

IL

MSSLI

-0.0138**

1.2600

***1.1443

***-0.5012***

0.8111

SMI-0.0068

1.1703

***

1.1583

***0.0902

0.5061

BBLI

0.0020

0.6441

***0.0113

0.0168

0.4961

BMI-0.0046

0.8881

***0.0589

0.4270

***

0.6944

UL

MSSLU

-0.0170*

1.5782

***1.1327

***-0.6889***

0.7185

SMU-0.0015

1.2080

***1.0780

***-0.1569

0.6901

BBLU

0.0022

1.2869

***-0.0531

-0.5409***

0.6142

BMU-0.0157***0.6867

***-0.0372

0.1731

0.3601

GH

SSHG

0.0021

0.6426

***0.7294

***

0.4508

***

0.6289

BBHG

0.0005

0.6219

***

0.0101

0.4292

***

0.4677

IH

SSHI

0.0044

1.1808

***1.3118

***

0.4043

**0.5421

BBHI-0.0107

0.7871

***0.4139

***

0.5388

***

0.3535

UH

SSHU

0.0132

0.6859

***1.3744

***0.6034

***

0.5060

BBHU-0.0174**

1.3953

***

0.0442

0.8813

***

0.6865

Notes:


inTable2.1.TheFamaandFrench

modeldescribes

how

themarket

excess

return

(rm

t−rr

f),thesm

all-m

inus-bigfactor(S

MB),andthehigh-m

inus-lowfactor(H

ML)explaintheexcessreturn

ofaportfolio

(rit−rr

f).

Theexcess

return

meansthedi�erence

betweentheactualreturnsandtherisk-freemarket

interest

rate.

SMB

represents

thesize

ofthecompanies,

HM

Lstandsforafactorbasedonthebook-to-m

arket

ratio.Weapply

theseem

ingly

unrelatedregression(SUR)

technique.

Abbreviations:α

i:intercept;β

i1:estimatedparameter

i1formarket

excess

return

(rm

t−rr

f);β


i2forfactor

SMB;β


i3forfactor

HM

L;Adj.R

2:adjustedR

2(G

oodness-of-Fit).

*,**,and***meansthat

thesevalues

are

signi�cantatthe10%,5%,and1%

level.

41


Table 2.7: Correlations of exogenous factors

rmt − rrf SMB t HMLt GMI t GMU t Gt It

rmt − rrf 1.0000

SMB t -0.1553 1.0000

HMLt -0.3298 -0.3771 1.0000

GMI t -0.4365 -0.3470 0.3326 1.0000

GMU t -0.5044 -0.2321 0.4188 0.7484 1.0000

Gt 0.8911 -0.2844 -0.3030 -0.1921 -0.2757 1.0000

It 0.8293 -0.1810 -0.0947 -0.3737 -0.3047 0.7635 1.0000

Ut 0.8410 -0.0402 -0.3654 -0.5355 -0.6900 0.6923 0.6432

Notes: rmt−rrf : market excess return at time t ; SMB t: Small-minus-Big (market value) factor attime t ; HMLt: High-minus-Low (book-to-market value) factor at time t ; GMI t: German GAAP-minus-IAS/IFRS factor at time t ; GMU t: German GAAP-minus-U.S. GAAP factor at time t ; Gt:Average returns of the German GAAP companies at time t ; It: Average returns of the IAS/IFRScompanies at time t ; Ut: Average returns of the U.S. GAAP companies at time t.

di�erent models.12 Computing the average of the individual SC for the 18 portfolios

estimated with Least Squares, we �nd support fort he results obtained by the ad-

justed R2. The GM model has the lowest SC (-3.10) vis-à-vis the Fama and French

model (-2.99) and the CAPM (-2.77). Given a speci�c endogenous variable, the

regression model with the lowest SC is regarded to be the best explaining model.

Based on our three models, we can derive the cost of equity capital for the 18

portfolios as expected excess returns less the risk-free rate. The SUR system delivers

forecasts of cost of equity capital for both the CAPM and the GM (see Table 2.10).

They are applied to the 18 portfolios.

Most notably, di�erences between the three accounting regimes are quite obvious

for the CAPM based cost of equity capital (Table 2.11), as the signi�cant di�erences

of the bate values have already indicated. The German GAAP groups with a market

value weighted mean of 1.496% exceed the IAS/IFRS groups with 1.267% and the

U.S. GAAP groups with 0.832% in terms of expected cost of equity capital.

For the GM model calculation, however, the di�erences are not strong any more.

While the U.S. GAAP �rms with 0.730% still have lower cost of equity capital, the

German GAAP �rms (1.426%) and IFRS/IAS �rms (1.429%) show no signi�cant

di�erence any more.

12We prefer using the Schwarz criterion (SC) versus using the Akaike Info criterion, since the SC�punishes� the loss of degrees-of-freedoms even more.

42


Table 2.8: GM model:rit − rrf

t = αi + βi1(rmt − rrft ) + βi2SMB t + βi3HMLt + βi4GMI t + βi3GMU t + εit

αi βi1 βi2 βi3 βi4 βi5 Adj.R2

G LS SLG -0.0086 * 0.5973 *** 0.9540 *** -0.1425 0.4761 *** 0.2861 ** 0.4853B BLG 0.0004 0.7416 *** 0.0736 -0.2350 *** -0.0498 0.2744 *** 0.6167

I LS SLI -0.0103 * 0.8940 *** 0.8612 *** -0.5180 *** -0.5190 ** -0.1296 0.8181B BLI 0.0040 0.6329 *** -0.0901 -0.0560 -0.5479 *** 0.3844 *** 0.5348

U LS SLU -0.0116 0.8501 *** 0.6697 *** -0.5093 *** -0.3401 -1.1204 *** 0.8121B BLU 0.0024 0.9189 *** -0.2243 -0.2840 ** 0.3827 * -1.2483 *** 0.7410

G MS SMG -0.0035 0.5823 *** 0.6758 *** 0.3591 *** 0.0264 0.2455 *** 0.6068B BMG 0.0016 0.8222 *** -0.2244 *** -0.0368 0.2921 ** -0.1153 0.7604

I MS SMI -0.0018 0.6746 *** 0.5824 *** 0.0019 -1.9735 *** 0.6027 *** 0.6656B BMI -0.0026 0.8664 *** -0.0123 0.3787 *** -0.3461 *** 0.2303 ** 0.6981

U MS SMU 0.0067 0.4105 *** 0.7324 *** -0.1231 -0.1098 -1.0650 *** 0.7797B BMU -0.0154 *** 0.5913 *** -0.0907 0.2146 * 0.0337 -0.2182 0.3589

G HS SHG 0.0006 0.7481 *** 0.8251 *** 0.4142 *** 0.1023 0.2105 ** 0.6750B BHG -0.0008 0.7050 *** 0.1207 0.4326 *** 0.2476 * 0.0537 0.4907

I HS SHI 0.0082 0.6990 *** 0.8039 *** 0.3815 *** -1.4755 *** 0.2136 0.6755B BHI -0.0082 0.4841 *** 0.1867 0.5511 *** -0.4707 * -0.1128 0.3539

U HS SHU -0.0162 ** 1.2680 *** 0.1017 0.9047 *** 0.4246 -0.3934 ** 0.6960B BHU 0.0114 0.2493 1.3003 *** 0.7412 *** 0.9536 *** -1.3604 *** 0.6635

Notes: The portfolio construction is described in Table 2.1. The GM model describes howthe market excess return (rmt − rrf ), the small-minus-big factor (SMB), the high-minus-lowfactor (HML), the GMI factor, and the GMU factor explain the excess return of a portfolio(rit− rrf ). The excess return means the di�erence between the actual returns and the risk-freemarket interest rate. SMB represents the size of the companies, HML stands for a factor basedon the book-to-market ratio. GMI is the return di�erence between portfolios of companiesusing German GAAP and IAS/IFRS (�German GAAP minus IAS/IFRS�) and GMU is the re-turn di�erence between German GAAP and U.S. GAAP (�German GAAP minus U.S. GAAP�)portfolios. We apply the seemingly unrelated regression (SUR) technique. Abbreviations: αi:intercept; βi1: estimated parameter i1 for market excess return rmt − rrf ; βi2: estimatedparameter i2 for factor SMB ; βi3: estimated parameter i3 for factor HML; βi4: estimated pa-rameter i4 for factor GMI ; βi5: estimated parameter i5 for factor GMU ; Adj.R2: adjustedR2

(Goodness-of-Fit). *, **, and *** means that these values are signi�cant at the 10%, 5%, and1% level.

43


Table 2.9: Improvements of adjusted R2: Fama and French Model versus CAPMand GM model versus Fama and French model

FF vs. GM vs. FF vs. GM vs. FF vs. GM vs.

CAPM FF CAPM FF CAPM FF

G L M H

S SLG +0.2437 +0.1465 SMG +0.3325 +0.0675 SHG +0.3378 +0.0461

B BLG +0.0187 +0.0267 BMG +0.0347 +0.0077 BHG +0.1338 +0.0230

I L M H

S SLI +0.2985 +0.0070 SMI +0.1893 +0.1595 SHI +0.2344 +0.1334

B BLI -0.0105 +0.0387 BMI +0.0993 +0.0037 BHI +0.1048 +0.0004

U L M H

S SLU +0.2427 +0.0936 SMU +0.2291 +0.0896 SHU +0.3377 +0.1900

B BLU +0.0395 +0.1268 BMU +0.0070 -0.0012 BHU +0.2083 -0.0231

Notes: The portfolio construction is described in Table 2.1. The table shows absolute improve-ments (+) / deteriorations (-) of adjusted R2. Abbreviations: G: German GAAP; I: IAS/IFRS; U:U.S. GAAP; L / M / H: Low / Medium / High book-to-market value; S / B: Small / Big marketvalue (=size); CAPM: Capital Asset Pricing Model; FF: Fama and French Model; GM: �GermanGAAP-minus� Model.

Table 2.10: Forecasts of cost of equity capital (based on CAPM and GM model)

CAPM GM CAPM GM CAPM GM

G L M H

S SLG -0.377% -0.349% SMG 0.562% 0.651% SHG 1.090% 1.240%

B BLG 1.009% 0.993% BMG 1.759% 1.637% BHG 2.013% 1.996%

I L M H

S SLI -2.065% -1.804% SMI -0.389% -0.221% SHI 1.127% 1.230%

B BLI 1.335% 1.501% BMI 1.553% 1.722% BHI 0.451% 0.584%

U L M H

S SLU -2.228% -1.546% SMU -0.197% 0.470% SHU 1.880% 2.774%

B BLU 1.403% 1.152% BMU -0.045% -0.119% BHU 1.563% 1.812%

Notes: The portfolio construction is described in Table 2.1. The table shows the forecasts for costof equity capital, based on the CAPM and GM model. Cost of equity is the portfolio's excess-return plus the risk-free rate. Abbreviations: G: German GAAP; I: IAS/IFRS; U: U.S. GAAP; L/ M / H: Low / Medium / High book-to-market value; S / B: Small / Big market value (=size);CAPM: Capital Asset Pricing Model; GM: �German GAAP-minus� Model.

44


Table 2.11: Cost of equity capital comparison (monthly weighted averages)

Weighted by observation Weighted by market value

G CAPM 1.094% 1.496%

GM 1.115% 1.426%

I CAPM 0.506% 1.267%

GM 0.670% 1.429%

U CAPM 0.158% 0.832%

GM 0.472% 0.730%

Notes: The table shows a comparison between the forecasts for cost of equity capital, based on theCAPM and GM model, weighted (1) by observations and (2) by market value of equity. The sum ofweights in each accounting regime group (G / I / U) totals to 1. Abbreviations: G: German GAAP;I: IAS/IFRS; U: U.S. GAAP; CAPM: Capital Asset Pricing Model; GM: �German GAAP-minus�Model.

That this relationship is not applicable for the GM estimates is straightforward

given the speci�cation of our new model though. Since the GM model already

accounts for accounting regime di�erences, we do not expect signi�cant di�erences

between the cost of equity capital estimates.

Taken together, these �ndings support the �rst hypothesis of our paper. Compa-

nies applying German GAAP have to deal with higher CAPM-based cost of equity

capital expectations by the investors in comparison to �rms that have adopted IAAP.

We call this e�ect �accounting premium� for IAS/IFRS and U.S. GAAP �rms vis-

à-vis German GAAP �rms. The development of a novel multifactor model that

captures the �accounting premium� leads to an improvement of the CAPM and

Fama-French model.


As shown in Table 2.12, the �rst-stage regression of the IAAP dummy on the di�erent

explaining factors (see Equation 2.9) in a probit model delivers highly signi�cant

results.13 The McFadden R2 with 21.95% is quite substantial, the marginal e�ects

of all explanatory variables are signi�cant at least at the 10%-level. Especially

noticeable are the p-values of the New Market dummy and the cross-listing dummy,

13Since the mean of our dependent dummy variable IAAP is 0.4646, the probit model has to bepreferred to the logit model. For a robustness check however, we also estimate the logit model,�nding no substantial di�erences between the two approaches.

45


Table 2.12: First-stage regression: IAAP it = Probit(δ0 + δ1log(ME it) +δ2ROAit + δ3CAPINT it + δ4MANUF it + δ5NEWMARKET it + δ6USUK it + εit)

Variable Coe�cient Std. Error z -Statistic Prob.

Constant -1.045 0.188 -5.563 0.000

log(ME ) 0.072 0.015 4.654 0.000

ROA -0.487 0.151 -3.233 0.001

CAPINT -0.307 0.149 -2.058 0.040

MANUF -0.107 0.061 -1.746 0.081

NEWMARKET 1.670 0.077 21.631 0.000

USUK 0.635 0.111 5.721 0.000

Obs with Dep=0 1107 McFadden R-squared 0.219

Obs with Dep=1 1332 LR statistic (6 df) 737.6

Total obs 2439 Probability (LR stat) 0.000

Notes: In the �rst stage, we estimate a probit model to analyze the �rms' probability to adoptIAS/IFRS or U.S. GAAP given di�erent explaining variables. Variable de�nitions: IAAP is adummy variable for applying internationally accepted accounting principles (IAS/IFRS or U.S.GAAP), log(ME ) is the natural logarithm of the market value of equity, ROA is the return onassets, CAPINT is the capital intensity (long term assets divided by total assets), MANUFis a dummy variable indicating if the company is a manufacturing company (SIC < 4000),NEWMARKET is a dummy for being included in the New Market (Neuer Markt) segment ofthe Frankfurt stock exchange, and USUK is a dummy for indicating whether the company iscross-listed at the U.S. or U.K. market. Based on this regression, we can also calculate the InverseMills Ratio.

which are both smaller than 0.0001.

After computing the Inverse Mills Ratio from the �rst-stage regression, we can

analyze the e�ect of the IAAP dummy on the cost of equity capital, while simulta-

neously controlling for self-selection (Inverse Mills Ratio) and, implicitly, for cross-

listing and New Market membership in our second-stage.

The results are presented in Table 2.13 and Table 2.14. We distinguish between

the CAPM based (Table 2.13) and the GM model based (Table 2.14) cost of equity

capital estimates. For each panel we estimate four di�erent time periods: (1) the

whole sample (1998�2004), and three sub-samples, (2) 1998�1999, (3) 2000�2001,

and (4) 2002�2004.

Comparing the second-stage estimations, we �nd two major results. One, the

impact of the IAAP dummy is signi�cantly di�erent for the CAPM versus the GM

model. Comparing especially the whole-sample estimations (1998�2004), we �nd

that the IAAP dummy is highly signi�cant for CAPM (p-value 0.007) whereas IAAP

46


Table 2.13: Second-stage regression (CAPM):CoEC m

it = ϕ0 + ϕ1IAAP it + ϕ2log(ME it) + ϕ3λit + εit

Variable Coe�cient Std. Error z -Statistic Prob.

1) Time period: 1998�2004

Constant -0.172 0.046 -3.723 0.000

IAAP -0.036 0.013 -2.712 0.007

log(ME ) 0.019 0.003 7.019 0.000

InvMillsRatio (λ) 0.052 0.054 0.957 0.339

R-squared 0.030 F -statistic 21.39

Total observations 2108 Prob (F -statistic) 0.000


Constant 0.568 0.124 4.599 0.000

IAAP 0.100 0.036 2.733 0.007

log(ME ) 0.006 0.006 1.064 0.288

InvMillsRatio (λ) -0.835 0.142 -5.869 0.000




Constant -0.320 0.074 -4.323 0.000

IAAP -0.045 0.020 -2.314 0.021

log(ME ) 0.026 0.004 6.162 0.000

InvMillsRatio (λ) 0.162 0.081 2.002 0.046




Constant -0.161 0.062 -2.598 0.010

IAAP 0.009 0.020 0.455 0.650

log(ME ) 0.010 0.004 2.534 0.011

InvMillsRatio (λ) 0.142 0.076 1.878 0.061



Notes: In the second stage, we analyze whether the adoption of internationally accepted accountingprinciples (IAAP) signi�cantly in�uences the cost of equity capital estimates. Variable de�nitions:CoEC is the cost of equity capital calculated by the CAPM, IAAP is a dummy variable for applyingIAS/IFRS or U.S. GAAP, log(ME ) is the natural logarithm of the market value of equity, and (λ)is the Inverse Mills Ratio. We restrict the sample to four di�erent time periods: 1) 1998�2004, 2)1998�1999, 3) 2000�2001, and 4) 2002�2004.

47


Table 2.14: Second-stage regression (GM):CoEC m

it = ϕ0 + ϕ1IAAP it + ϕ2log(ME it) + ϕ3λit + εit

Variable Coe�cient Std. Error z -statistic Prob.


Constant -0.103 0.071 -1.458 0.145

IAAP 0.022 0.020 1.091 0.275

log(ME ) 0.005 0.004 1.186 0.236

InvMillsRatio (λ) 0.183 0.082 2.217 0.027




Constant -0.118 0.135 -0.878 0.381

IAAP 0.036 0.040 0.895 0.372

log(ME ) 0.029 0.006 4.584 0.000

InvMillsRatio (λ) -0.203 0.155 -1.310 0.191




Constant -0.633 0.092 -6.910 0.000

IAAP -0.044 0.024 -1.817 0.070

log(ME ) 0.043 0.005 8.276 0.000

InvMillsRatio (λ) 0.187 0.100 1.862 0.063




Constant 0.091 0.102 0.886 0.376

IAAP 0.040 0.033 1.225 0.221

log(ME ) -0.012 0.006 -1.934 0.053

InvMillsRatio (λ) 0.245 0.125 1.960 0.050



Notes: In the second stage, we analyze whether the adoption of internationally accepted accountingprinciples (IAAP) signi�cantly in�uences the cost of equity capital estimates. Variable de�nitions:CoEC is the cost of equity capital calculated by the GM model, IAAP is a dummy variablefor applying IAS/IFRS or U.S. GAAP, log(ME ) is the natural logarithm of the market value ofequity, and (λ) is the Inverse Mills Ratio. We restrict the sample to four di�erent time periods: 1)1998�2004, 2) 1998�1999, 3) 2000�2001, and 4) 2002�2004.

48


is not signi�cant for the GM model (p-value 0.275). Moreover, for the GM model

the F -statistic for the whole sample period has a p-value of 0.110, meaning that

all explaining variables together are not signi�cant at the 10%-level. Looking at

all of the panel estimations, while in the CAPM regressions IAAP is signi�cant in

three of four cases with two p-values below 1% and one below 5%, in the GM model

regressions IAAP is signi�cant only in one case (2000�2001), with a p-value below

10%.

The rational for these �ndings is that the cost of equity capital estimates in the

CAPM do not account for accounting regime di�erences, while the estimates in the

GM model do. Consequently, it is not surprising that on the one hand the IAAP

dummy has no signi�cant in�uence on the cost of equity capital estimates any more,

once we have already included this information for estimating the dependent vari-

able. On the other hand, documenting the di�erences for the CAPM regressions,

again, we �nd signi�cant di�erences between the regimes, which supports our hy-

pothesis H1 once more. It is important to note that this time we have also controlled

for self-selection, cross-listing and New Market membership e�ects.

The second result is that there are time period di�erences. Concentrating on the

CAPM method (Table 2.13), we �nd a negative sign for the IAAP dummy for the

whole period (1998�2004), which is highly signi�cant (p-value 0.007). This supports

our �rst hypothesis that applying a IAAP leads to lower cost of equity capital.

Looking at the sub-periods, however, delivers more detailed results. For the years

1998 and 1999, the e�ect was positive (p-value 0.007), only for the years 2000 and

2001 it was negative (p-value 0.021). Our rational for this observation is that in

the beginning uncertainty dominated the true substantial power of the IAAP, that

provisions for several important items were missing under IAS and that level of

compliance with IAAP was low. These disadvantages at least partly diminished in

the second sub-period when under IAS several new standards became e�ective and

companies as well as users of �nancial statements became used to the IAAP.

Yet for the period 2002�2004, we cannot draw any interference, since IAAP is

not signi�cant. Also for this phenomenon we have an explanation, namely that in

that period new revisions of accounting standards became e�ective which granted

leeway for management's discretion, the creditworthiness in the IAAP decreased

after several accounting scandals and that German GAAP abolished the option to

include tax-induced accounting practices into the consolidated accounts.

49


Consequently, this gives support to our hypothesis H2 that there are time-speci�c

di�erences in the e�ects of adopting IAAP on the cost of equity capital.

Regarding the goodness-of-�t, it is fair to mention that most of the resulting R2

are not considerably high, which can be seen as a drawback for this analysis. There

are two points to mention, however. First, this is economically comprehensible,

since we have stated that our cost of equity capital models (CAPM and GM model)

do already explain the cost of equity capital su�ciently enough. Consequently, we

would even expect that the explaining power of the second-stage regressions is fairly

low, meaning that our cost of equity capital methods were e�cient before. Secondly,

we rather see the second-stage as a ceteris-paribus analysis. In that context, low

R2s are in fact also econometrically expectable and acceptable and do not impair

the explanatory power, as long as the individual p-values are valid and the overall

F -test does not lead to refusing the signi�cance of all variables together (see, e.g.,

Wooldridge, 2002, p. 41).

In other words, we do not want to explain the calculation of the cost of equity

capital in the second-stage � we did this with our portfolio models already. Here, a

low R2 is rather a sign for a well-speci�ed cost of equity capital calculation, which

can also be seen as a robustness test for our hypothesis H1.

2.8 Conclusions

Our results suggest that the voluntary adoption of IAS/IFRS or U.S. GAAP by Ger-

man companies goes along with a decrease in their cost of equity capital. Notwith-

standing the speci�c institutional framework in Germany, our �ndings support the

general expectation that higher quality accounting standards lead to lower cost of

equity capital, using conventional models. We call this e�ect �accounting premium�

for IAAP. These results also hold when controlling for e�ects like self-selection,

cross-listing, and New Market (Neuer Markt) listing. Additionally, by developing a

novel multifactor model (that we call �GM model�), we can capture the �accounting

premium� e�ect and can hence improve the CAPM and Fama-French model.

Our study calls for more caution in future studies investigating the relationship

between the adoption of a certain accounting regime, information quality and the

cost of equity capital. Several additional e�ects and factors like accounting incen-

tives, enforcement, other institutional settings, and the properties of the capital

50

2.8 Conclusions

market as well as the interactions of these e�ects have to be considered.

For our analyses, we use the classical CAPM and a novel methodology including

the accounting regime information into a multifactor model, based on the Fama and

French model. This methodology has the advantages of not being biased or com-

promised by estimation errors of analysts' forecasts, of mitigating disclosure level

di�erences, and of being applicable to companies which are not followed by �nan-

cial analysts. The self-selection issue is addressed by using a two-stage estimation

procedure, explicitly controlling for self-selection and other e�ects, like cross-listing

and New Market membership

Similar to Barth et al. (2006a), we �nd that the inclusion of an information

quality or accounting factor improves the performance of the multifactor model.

Therefore, we provide further evidence that the traditional three factor model of

Fama and French seems to lack an additional factor proxying the level and quality

of information provided to investors.

We recognize, however, that our study is subject to caveats. First, we compare

the e�ects of voluntarily adopting IAS/IFRS or U.S. GAAP by German compa-

nies. The e�ects of mandatory adoption in periods from 2005 on might have been

di�erent as companies can no longer commit themselves to increased disclosure by

adopting IAS/IFRS or U.S. GAAP but are rather obliged to do so. Besides, the new

enforcement system in Germany and the new oversight body for statutory auditors,

established in 2005, might have impacted the accounting practice. Furthermore,

companies, auditors, and investors have gained more expertise in the provisions of

IFRS and U.S. GAAP and might thus cope better with the more complex infor-

mation provided by IAAP today. In summary, these changes might have lead to

a di�erent overall impact of the adoption of IFRS and U.S. GAAP by German

companies on the cost of equity capital after 2004.

Second, there is still a controversy in the literature about the assumption of diver-

si�ability of information risk. However, apart from our study several other papers

have already found results supporting the view that market beta does not su�ce

to explain cost of equity capital and see other information risk factors applicable

to improve the explanatory power (see, e.g., Francis et al., 2005a; Botosan, 2006).

Multifactor models are especially crucial for determining cost of equity capital on

the �rm-level for which the diversi�ability argument does not apply. We see our

research as another empirical contribution toward this discussion.

51


These two caveats provide opportunities for future research. Studies about the

impact of the mandatory adoption of IFRS, the changes in standards of IAS/IFRS

and U.S. GAAP, and of the new enforcement system as well as auditor oversight in

Germany on the information quality and the cost of equity capital might be valuable

tasks. Moreover, future research could examine the ability of our multifactor model

in evaluating the cost of equity capital impact of di�erent IAAP and could test our

model in other settings, e.g., investigate the impact of the adoption of IFRS in other

countries or institutional settings.

52

3 The Value and Accounting

Premium for South

African-listed Shares

Published in:

Journal of Economic and Financial Sciences, 2 (October 2008), 187�202 (with Jürgen

Ernstberger and Christian Heinze)

3.1 Introduction

This research examines two patterns in stock returns for �rms listed at the JSE

Limited (previously the Johannesburg Securities Exchange): the value premium

and the accounting premium. Firms with a high ratio of the book value of equity to

the market value of equity are regarded as value stocks; a low ratio identi�es growth

stocks. The excess return of value stocks over growth stocks is referred to as a value

premium. An accounting premium is said to exist, if investors reward voluntary

compliance with International Financial Reporting Standards (IFRS) prior to its

becoming compulsory in 2005 with a lower expected return.

Research in empirical �nance, initiated by the pioneering work by (Fama and

French, 1992, 1993), has documented the existence of a value premium in a variety

of capital markets. In this strand of literature it is often claimed that the value

premium exists exclusively among small stocks. If this were the case, the majority

of investors' wealth would be una�ected. This critique is taken into account by

explicitly investigating the distribution of the value premium for di�erent company

sizes. Moreover, the introduction of mandatory IFRS accounting in 2005 for �rms

listed at the JSE Limited is used to investigate the existence of an accounting pre-

mium. This premium is examined within di�erent size classes as well as for value

53

3 The Value and Accounting Premium

stocks and growth stocks separately. Furthermore, the article analyzes at whether

Capital Asset Pricing Model (CAPM) captures the investigated patterns. Finally,

it is investigated, if the premiums qualify as risk factors in a multifactor model.

The �ndings suggest the existence of a value premium of about 1.5% to 2% both

in the pooled sample as well as in subsamples of small and large stocks. No indi-

cation of di�erences in the value premium between the two di�erent size classes is

found. Moreover, an accounting premium of similar size is identi�ed. A closer look

reveals, however, that the existence of the accounting premium is limited to small

stocks and value stocks, while large stocks and growth stocks do not share this pat-

tern. Furthermore, the CAPM does not capture the value premium. Both the value

property and the accounting standard in use qualify as priced risk factors. Thus,

the contribution to literature is twofold. First, most recent evidence on the value

premium is provided for a market that has been relatively disregarded so far, i.e.,

South Africa and that this e�ect does not depend on the size of the stocks. Second,

the existence of an accounting premium for voluntary compliance with an interna-

tionally standardized accounting standard is documented and it is shown that both

the value property and the accounting standard in use are priced risk factors in the

South African market.

The remainder of this paper proceeds as follows. Section 3.2 describes the back-

ground of the study and gives an overview of prior research with a similar focus.

After providing details about the sample and the portfolio construction (Section

3.3), the research methodology is described in Section 3.4. Section 3.5 discusses the

empirical results and Section 3.6 concludes with a summary and a �nal comment.

3.2 Background and prior research

Asset pricing is an important �eld of research in �nance. In addition to the CAPM

several multifactor models have been developed for explaining stock returns. Follow-

ing the seminal papers of Fama and French (1992, 1993) several other studies have

documented the failure of the CAPM to adequately explain asset returns. These

studies usually refer to anomalies, e.g., size e�ect or momentum e�ect, which are

not explained by the CAPM. However, most of these studies are conducted for the

U.S. market or other developed economies. Only a few studies investigate emerging

markets like South Africa which exhibits the largest market capitalization in Africa

54

3.3 Data and portfolio construction

(African Securities Exchanges Association, 2008) and the 19th largest worldwide

(World Federation of Exchanges, 2008). Trade was established in 1887; today the

JSE Limited lists more than 450 �rms on its main board (JSE Limited, 2008).

Since 2005 �rms listed at the JSE are required to comply with IFRS. Up to that

year �rms were free to choose between IFRS and South African Statements of Gen-

erally Accepted Accounting Practices (SA GAAP). Until 2005 the two accounting

regimes were harmonized (JSE Limited, 2004). Thus, only small di�erences remain

between them. This transition allows examination of whether investors reward the

higher degree of international comparability when applying IFRS with lower return

expectations.

Prior research on stock returns at the JSE includes Van Rensburg and Roberts

(2003). The research �nds that the cross-section of stock returns is captured by

similar variables as on the U.S. market: size, dividend yield, earnings-to-price, and

book-to-market. The results of this study are largely in line with the results of prior

papers by Page and Palmer (1991) and Van Rensburg (2001).

The contribution of this research lies in the speci�c consideration of the e�ect of

book-to-market (value premium) and the accounting standard in use (accounting

premium) on stock returns during the years following 2000.

3.3 Data and portfolio construction

The investigation is based on a panel of 185 �rms listed at the JSE during the time

from July 1998 to June 2007 (108 months). Finance and insurance �rms (SIC codes

6000 to 6999) are excluded for being fundamentally di�erent from non-�nancial �rms

in book value of equity. The remaining 159 �rms are investigated in the following

analyses. These �rms are used to form six portfolios on market value (market

capitalization, i.e. shares outstanding times share price) and book-to-market, that

is the ratio of book value of equity to market value of equity.

The portfolio design equals that of Fama and French (1993). Portfolio construction

proceeds as follows. In June 1998 �rms reporting a positive book value at the end of

1997 and whose returns are known for every month up to June 1999 are considered

for portfolio assignment. In line with Fama and French (1992) book value is de�ned

as book value of stockholders' equity minus preferred stock plus deferred tax. The

median market value in June 1998 and the 30% and 70% quantiles of book-to-market

55


in December 1997 act as breakpoints for portfolio construction. Each �rm is labeled

by its location relative to these breakpoints: a �rm is assigned either small (S) or big

(B) in size (market value) and either growth (G), neutral (N), or value (V) in book-

to-market. Portfolios pool stocks sharing the same labels and are named accordingly.

For example, a �rm with lower than median market capitalization and book-to-

market below the 30% quantile is assigned to the small-growth portfolio (SG). Each

June from 1999 to 2006 the stocks are reassigned in same fashion depending on the

book-to-market of previous December and the current market value.

During July 1998 and June 2000 the return series exhibit high volatility relative

to the rest of the sample period. To obtain a clearer picture on the properties of

the data generating process the sample is split into two subsamples. The earlier,

more volatile, period ranges from July 1998 to June 2000. The stable period begins

with July 2000. A di�erence in dispersion between the periods is con�rmed by two

nonparametric tests (Mood-test, Fligner-Killeen-test). These tests are most robust

against deviations from the Gaussian distribution (Conover et al., 1981).

In order to study accounting regime e�ects, the portfolio construction is extended

to cover three attributes. As IFRS became mandatory at the beginning of 2005 the

sample period is shortened for this analysis. July 2000 is chosen as a starting point

for two reasons. First, the IFRS portfolios are too sparsely populated in earlier

periods. Second, the higher volatility complicates the analysis of the underlying

mechanisms. Data are used up to June 2005 as once IFRS is mandatory the existence

of an accounting premium is ruled out by de�nition. The portfolio design equals

that of Ernstberger and Vogler (2008). Firms with missing data on monthly returns

or negative book equity are excluded. The size attribute (S vs. B) is assigned

dependent on the market value of a �rm in June relative to the median market

value in June. Its book-to-market ratio in previous December relative to the 30%

and 70% quantiles in previous December book-to-market governs the assignment of

the book-to-market attribute (G vs. N vs. V). An accounting attribute is assigned

depending on the accounting regime chosen the following December (in the sorting

period). �I� indicates IFRS and �G� South African GAAP. Portfolios gather �rms

sharing all three attributes. This leads to 12 portfolios. Each portfolio is named

according to the attributes it units, e.g., the portfolio including small �rms with

low book-to-market and IFRS accounting is termed �SGI�. Stocks remain in one

portfolio from July to June of the following year.

56

3.4 Methodology

3.4 Methodology

3.4.1 Value premium

First, the unconditional distribution of the value premium is considered using the

six portfolio sorting. Calculation of monthly average returns generates six monthly

return series. These realized returns are treated as proxies for expected returns.

This approach has been criticized as average returns are seen as noisy estimates.

However, less than half of the companies in the sample actually paid dividends,

which rules out estimation techniques based on dividend growth as recently applied

in Chen et al. (2008). The question of whether to use value weighted or equally

weighted averages is disputed in literature. Hence, all investigations are carried out

using both weighting schemes. In investigations of value weighted returns �rms are

excluded, if their market value is missing for one month or more during July and

June of the following year. As the results for both schemes are almost identical

the results on equally weighted portfolios are not reported. A portfolio is referred

to by its characterizing letters, e.g. SB, and its return by adding a time subscript

indicating the month of observation, e.g. SB t, t ∈ {1, . . . , 108}.

The value premium is de�ned as the return di�erence between value and growth

portfolios

VMG t = (SV t + BV t)/2− (SG t + BG t)/2 (3.1)

This construction results in the value premium factor originally proposed by Fama

and French (1992). Findings in Loughran (1997) negate the existence of a value pre-

mium for large stocks. Following Fama and French (2006) this critique is accounted

for by analyzing the value premium in two size classes separately. The size speci�c

factors are de�ned as

VMGS t = SV t − SG t

VMGB t = BV t − BG t (3.2)

Furthermore, the di�erences in the value premium between the small and large

stocks (VMGS t − VMGB t) is examined.

57


3.4.2 CAPM and multifactor model

Next, the research examines the conditional distribution of the value premium to

investigate whether the value premium is captured by the movement of the market

portfolio's excess return as predicted by CAPM. The excess return on a market

portfolio is de�ned as RM t − RF t. The market return (RM t) is calculated as the

average return across all stocks. This includes all stocks in the portfolios as well as

the stocks with negative book-value. A one month return on South African Treasury

bill with three month maturity (RF t) is subtracted.

Therefore the following equation is estimated by ordinary least square (OLS)

regressions:

RPt − RF t = a+ b (RM t − RF t) + εt (3.3)

with RPt ∈ {SV t, SN t, SG t,BV t,BN t,BG t,VMG t}

Heteroskedasticity is tested using the White test; autocorrelation is checked using

the Breusch-Pagan test. If evidence of heteroskedasticity in the error terms is found

heteroskedasticity consistent standard errors are used. When using robust standard

errors the reported t-value is marked with a star. Leverage points and outliers

are detected using indicators proposed in Belsley et al. (1980), and Cook's distance

measure (Cook, 1979). An outlier is de�ned as an observation showing a studentized

residual in excess of 3.5 times the series' interquartile range and at least two in

absolute value. A leverage point is understood as an observation of which the

diagonal entry in the hat matrix (the OLS projection matrix) exceeds 3.5 times

the hat values' interquartile range and at least three times the ratio of number of

covariates to number of observations. Leverage points or outliers are omitted in the

estimation when they result in a change of one standard error in at least one of the

slope estimates when omitted or when its Cook's distance exceeds the 50% quantile

of the respective F -distribution.

The often documented inadequacy of CAPM has led Fama and French (1992,

1993) to develop an alternative model. The CAPM is supplemented by two addi-

tional factors meant to indicate distress and hence higher exposure to non-diversi�able

risk (Berkowitz and Qiu, 2001; Fama and French, 1993). To test whether this model

dominates the CAPM in capturing stock return patterns, especially the value pre-

mium, the following models are estimated:

58

3.4 Methodology

RPt − RF t = a+ b (RM t − RF t) + c SMB t + dVMG t + ut (3.4)

where RPt ∈ {SV t, SN t, SG t,BV t,BN t,BG t}

SMB t is the size premium factor. It is de�ned as follows:

SMB t = (SV t + SN t + SG t)/3− (BV t + BN t + BG t)/3 (3.5)

Estimation techniques are identical to the CAPM case.

3.4.3 Accounting premium

The aforementioned twelve portfolio sort is used for the accounting premium analy-

ses. Value weighted and equally weighted returns are obtained for the twelve port-

folios. First, the unconditional distribution of the accounting premium is examined.

The accounting premium is de�ned as follows:

GMI t = (SGG t + SNG t + SVG t + BGG t + BNG t + BVG t)/6

−(SGI t + SNI t + SVI t + BGI t + BNI t + BVI t)/6 (3.6)

Second, the accounting premium is investigated within small and large stocks, i.e.

GMIS t = (SGG t + SNG t + SVG t)/3− (SGI t + SNI t + SVI t)/3

GMIB t = (BGG t + BNG t + BVG t)/3− (BGI t + BNI t + BVI t)/3 (3.7)

as well as for growth stocks and value stocks separately, i.e.

GMIG t = (SGG t + BGG t)/2− (SGI t + BGI t)/2

GMIV t = (SVG t + BVG t)/2− (SVI t + BVI t)/2 (3.8)

The di�erences in the accounting premium between size classes (GMIS t−GMIB t)

is examined as well as the accounting premium between growth and value stocks

(GMIV t −GMIG t).

59


3.4.4 The value premium and the accounting premium as risk

factors

Self selection might occur in the analyses because �rms with lower risk loadings

might use IFRS to try to signal this to potential investors (Hail, 2002). In this

case IFRS adoption and returns will show negative correlation. However, it will

not be linked to the accounting regime. Finding the accounting premium within

di�erent portfolios sorted by risk characteristics such as size and book-to-market is

an indication of accounting being a risk factor on its own. In addition, self selection

problems are addressed using regression analysis to control for known risk mimicking

portfolios, i.e. SMB and VMG (Berk, 1995).

For this analysis the size premium factor and the value premium have to be

adapted to the new portfolio structure:

SMB t = (SVI t + SVG t + SNI t + SNG t + SGI t + SGG t)/6

−(BVI t + BVG t + BNI t + BNG t + BGI t + BGG t)/6 (3.9)

VMG t = (SVI t + SVG t + BVI t + BVG t)/4

−(SGI t + SGG t + BGI t + BGG t)/4 (3.10)

The Fama and French multifactor model is estimated for the twelve portfolios

using the shorter time frame. This amounts to estimation of the following equations

RPt − RF t = a+ b (RM t − RF t) + c SMB t + dVMG t + εt (3.11)

where RPt ∈ {SV t, SN t, SG t,BV t,BN t,BG t,VMG t}

To test if the accounting premium re�ects a risk factor, the following equations

proposed by Ernstberger and Vogler (2008) are estimated:

RPt − RF t = a+ b (RM t − RF t) + c SMB t + dVMG t + eGMI t + νt (3.12)

where RPt ∈ {SV t, SN t, SG t,BV t,BN t,BG t}

The calculation of the market portfolio's excess return and the estimation is the

60

3.5 Results

Table 3.1: Summary statistics for the value premium

VMG VMGS VMGB VMGS-B

07/1998 - 06/2007

Mean 1.62* 2.20* 1.04* 1.16

(2.37) (2.83) (1.33) (0.56)

Median 2.08* 2.29* 1.71* 0.58

(<0.01) (<0.01) (0.03) (0.12)

07/2000 - 06/2007

Mean 1.92* 1.95* 1.89* 0.05

(3.80) (3.45) (2.67) (0.07)

Median 2.08* 1.85* 1.90* -0.05

(<0.01) (<0.04) (<0.01) (0.62)

Notes: Six portfolios are formed on size (small (S) and big (B)) and the book-to-market ratio(growth (G), neutral (N), and value (V)). These portfolios are realigned annually. Value weightedreturn series are calculated for each portfolio. VMG (value-minus-growth) represents the di�erencein returns between value and growth stocks; VMGS and VMGB embody the value premium forsmall and big �rms. VMGS-B gives the di�erence between the value premiums in both size classes.Distributions are investigated using three time frames: the full sample, the �rst two years whichexhibit excessive volatility (not tabulated), and the rest of the sample. The table shows means andt-ratios for the mean in parenthesis for the value premium series for the full sample period andthe second subsample period. In addition, the median and underneath the p-value of a Wilcoxonsigned-rank test with the null of a non positive median in parenthesis are reported for the seriesduring both time periods. The mean and the median are given as percentage points. * indicatessigni�cance at a 5% level.

same as for the CAPM.

3.5 Results

3.5.1 Value premium

The summary statistics for the value premium are shown in Table 3.1.

In the full sample period (top half of Table 3.1) estimates for the mean (1.62%,

2.20%, and 1.04%) and the median (2.08%, 2.29%, and 1.71%) are positive among

all stocks, small, and large stocks. Statistical signi�cance is indicated by both t-

statistic and Wilcoxon test in the pooled sample as well as for small stocks. For

big stock the t-test is inconclusive. Nevertheless, the Wilcoxon test shows signi�-

cance. Furthermore, the value premium di�erence between small and large stocks

61


Figure 3.1: Distribution of the Value Premium among all �rms and among smalland big �rms

is insigni�cant. Estimates of the value premium factor's distribution illustrated in

Figure 3.1 illustrate the positive location among all stocks, as well as among small

and big stocks.

To obtain further insight, the possibility of structural breaks is considered. Figure

3.2 shows that the volatility has decreased tremendously up to mid 2000. Test of

change in the variance reject a constant dispersion at 1% con�dence level (in case

of VMGB a 2% level is needed).

The estimates for the more volatile period June 1998�July 2000 (not tabulated) are

imprecise and allow little inference. The bottom part of Table 3.1 reports estimates

and test statistics for the period from July 2000 to June 2007. For all stocks,

small stocks, and large stocks the mean estimates (1.92%, 1.95%, and 1.89%) and

median estimates (2.08%, 1.85%, and 1.90%) show a value premium. T -statistics

62

3.5 Results

Figure 3.2: Distribution of the Value Premium among small and big �rms duringthe two subperiods

63


and Wilcoxon test show that these are signi�cant �ndings. Thus, more recently the

value premium for large stocks is clearly present. Di�erences in the value premium

between small and large stocks cannot be found. This is robust evidence for the

existence of a value premium for both stock sizes in the post-2000 period.

These �ndings bear resemblance to those of Fama and French (2006) for the U.S.

market. Fama and French investigate the value weighted return series for stock

portfolios ranging from July 1926 to June 1963. The portfolio design is identical to

the one in this study. They assess the value premium for the full sample as well as

for two subsamples ranging from July 1926 to June 1963 and from July 1963 to June

2004. As this research is conducted for very recent data using results on the latter

as benchmark is more adequate. Although being di�erent in quantitative terms,

the U.S. portfolio returns' means range from 0.62% to 1.22%, while here a range of

1.04% to 2.20% is observed, the qualitative implications coincide: Fama and French

�nd a value premium signi�cantly di�erent from zero (with t-ratios of 3.34 and 3.97)

when looking at the overall level and when looking speci�cally at small �rms. Their

evidence for a value premium among big �rms, a t-ratio of 1.87, is rather mixed.

Sorting by the earnings-price ratio instead of book-to-market however reveals clear

indication of its existence. Also when looking at a richer history of the full sample

period, they �nd a t-ratio of 2.23.

3.5.2 CAPM and multifactor model

Next the conditional distribution of the value premium is investigated to see whether

the premium is captured by the movements of the market as predicted by the CAPM.

The previous results have shown the later subsample (July 2000�June 2007) to

exhibit a di�erent distribution than the earlier subsample (July 1998�June 2000).

The more recent period gives a clearer picture on the value premium and exhibits

less volatility. The investigation is con�ned to this subsample. Results are given in

Table 3.2.

The theory surrounding the CAPM is free of an intercept. If the intercept is found

to be signi�cantly di�erent from zero, this amounts to a rejection of the theory. The

top part of Table 3.2 shows that the CAPM is rejected for all portfolios. Moreover,

the value is not captured: both value portfolios (SV and BV) exhibit lower betas

(0.458 and 0.784) than the respective growth portfolios (0.571 and 1.166). As a

result, the value premium shows a negative beta (-0.318). These �ndings are in line

64

3.5 Results

Table 3.2: CAPM and Multifactor model regression results, July 2000 to June2007 (84 month)

VMG SG SN SV BG BN BV

CAPM

a 0.044* 0.013* 0.026* 0.035* -0.011* 0.010* 0.015*

(4.14) (2.55) (5.82) (8.56†) (-3.70) (2.56) (2.85)

b -0.318 0.571* 0.425* 0.458* 1.166* 0.783* 0.784*

(-1.81) (6.88) (5.82) (5.05†) (24.25) (14.14) (8.94)

Multifactor model

a -0.001 0.001 -0.002 -0.003 0.002 -0.002

(-0.19) (0.08) (-0.74) (-0.97†) (0.45) (-0.30)

d -.0238* 0.300* 0.647* -0.373* 0.339* 0.743*

(-3.21) (3.97) (12.55) (-7.25†) (5.11) (8.51)

Obs. 84 84 84 84 84 84 84

Notes: Six portfolios are formed on size (small (S) and big (B)) and book-to-market (growth(G), neutral (N), and value (V)). The CAPM and a multifactor model are estimated for all sixportfolios. In addition, the CAPM is estimated separately for the value premium. The top part ofthe table provides the intercept and beta estimates for the CAPM and t-ratios in parenthesis. Thebottom part gives the intercept estimates as well as the slope estimates of the value premium formultifactor model, t-ratios, and the number of observations used. T -ratios are marked with a †,if the t-ratio is calculated using heteroscedasticity robust standard errors. * indicates signi�canceat a 5% level.

65


Table 3.3: Summary statistics for the accounting premium, July 2000 to June2005 (60 month)

GMI GMIV GMIG GMIV-G GMIS GMIB GMIS-B

Mean 1.83* 1.97 0.43 1.54 3.24* -0.19* 3.43*

(3.25) (1.65) (0.55) (1.12) (6.15) (-0.19) (2.15)

Median 1.78* 1.67* 0.85 0.82 3.57* 0.90 2.67*

(<0.01) (0.02) (0.21) (0.11) (<0.01) (0.44) (0.02)

Notes: Twelve portfolios are formed on size (small (S) and big(B)), book-to-market (growth (G),neutral (N), value (V)), and accounting regime used (IFRS (I) and SA GAAP (G)). Portfolios arerealigned each year. For each portfolio a value-weighted return series is calculated. GMI embodiesthe di�erence in the returns between SA GAAP stocks and the IFRS stocks. GMIV captures thisdi�erence for value stocks only; GMIG does the same for growth stocks. GMIS and GMIG give thisdi�erence within the class of small stocks and big stocks, respectively. GMIV-G is the di�erence inthe accounting premium series between value stocks and growth stocks. GMIS-B similarly capturesthe di�erence in the accounting premium between small and big stocks. The table supplies means,t-ratios for the means in parenthesis, as well as the median and underneath a p-value of a Wilcoxonsigned rank test against the null of a non positive location in parenthesis. Mean and median aregiven in percentage points. * indicates signi�cance at a 5% level.

with the results for the U.S. market of Fama and French (1993, 2006).

Results in the bottom part of Table 3.2 suggest that inclusion of additional risk

factors indeed mitigates these problems. Once the additional factors are controlled

for, all intercepts become insigni�cant. The multifactor model is more powerful in

explaining the cross-section of returns. The slopes in the multifactor model repre-

sent risk factor loadings. Evidence for a value premium was found earlier. Stocks

exhibiting a higher risk load should on average have larger returns. The results in

table 2 show that the book-to-market loadings are 0.6468 for value stocks versus

-0.2377 for growth stocks within the class of small stocks and 0.7427 versus -0.3726

when looking at large stocks. The factor loadings are signi�cantly di�erent from

zero without exception. Therefore, the multifactor model does better in explaining

the value premium. These �ndings are very similar to those of Fama and French

(1993) for the U.S. market.

3.5.3 Accounting premium

The summary statistics are found in Table 3.3 and illustrated in the left panel of

Figure 3.3.

66

3.5 Results

Figure 3.3: Distribution of the Accounting Premium among all �rms, withindi�erent size class, for value stocks and growth stocks

The estimated mean is 1.83% among all stocks. The estimated density of the

accounting premium for value weighted portfolio returns shows the strictly positive

location. The mean estimate is 3.25 standard errors from zero. The Wilcoxon signed-

rank test rejects the null of a non-positive median and supports the conclusion of

the existence of an accounting premium.

For the class of value stocks the �ndings are similar: an accounting premium of

slightly less than 2% exists. However, for growth stocks both t-ratio and Wilcoxon

test do not support its existence. Comparing the accounting premium for value

and growth stocks neither the Wilcoxon nor the t-ratio indicates a di�erence. The

accounting premium seems to be present in the class of small stocks. Large stocks

do not exhibit this pattern. Here both the t-ratio and the Wilcoxon test indicate a

di�erent premium. These �ndings are supported by the visual impression of the two

right panels in Figure 3.3. The value premium clearly di�ers between size classes,

67


Table 3.4: Estimation results for the multifactor model, July 2000 to June 2005(60 month)

GMI SGG SNG SNI SVG SVI BGG BGI BNG BNI BVG BVI

Multifactor model

a 0.022* 0.005 0.017* -0.062* 0.024* -0.029* -0.007 0.007 0.004 0.003 0.019* -0.031

(3.62) (0.45) (2.88) (-3.43) (4.71) (-2.38) (-1.04) (0.87) (0.86) (0.45) (2.24) (-0.17)

Enhanced multifactor model

a -0.003 0.007 -0.062* 0.013* -0.007 -0.027* 0.026* -0.008 0.003 -0.004 -0.013

(-0.29) (1.28) (-2.22) (2.99) (-0.78) (-4.62) (3.58) (-1.66) (0.32) (-0.52) (-0.89)

e 0.395* 0.474* 0.002 0.489* -1.274* 0.907* -0.848* 0.543* 0.025 1.066* -0.859*

(2.20) (3.74) (0.01) (4.78) (-5.31) (5.65) (-5.61) (5.60) (0.15) (6.85) (-3.22)

Obs. 60 59 60 12 60 24 60 60 60 60 60 59

Notes: Twelve portfolios are formed on size (small (S) and big (B)), book-to-market (growth (G),neutral (N), value (V)), and accounting regime used (IFRS (I) and SA GAAP (G)). Portfolios arerealigned annually. For each portfolio a value-weighted return series is calculated. The Fama andFrench multifactor model is estimated both for all twelve portfolios and the accounting premium.The upper part of the table supplies the intercept estimates of all series and t-ratios in parenthesis.The enhanced multifactor model is also estimated for all twelve series. The bottom part of thetable gives intercept estimates as well as slope estimates for the accounting premium. The SGIportfolio is omitted due to data unavailability. * indicates signi�cance at a 5% level.

comparing the value stocks to growth stocks yields less clear results. Nevertheless,

the evidence for the accounting premium in both classes separately is unambiguous.

Other risk factors are controlled for by estimation of an enhanced factor model.

Data limitations are an issue: estimation for the SGI portfolio is not possible, since

SGI features a zero population up to June 2005. When using rather sparsely pop-

ulated portfolios controlling outliers is important. Thus, the procedure described

above is used. Table 3.4 reports the estimation results on both the multifactor model

and the enhanced multifactor model.

The intercept of the regression of the accounting premium on the other risk factors

is signi�cantly di�erent from zero. Hence, other risk factors do not fully explain

this premium. The loading on accounting risk is found higher for �rms using SA

GAAP. Most of the loadings are signi�cantly di�erent from zero. The enhanced

multifactor model produces less signi�cant intercepts and hence outperforms the

classical multifactor model in explaining the cross-section of stock returns. Hence,

the results support the view of accounting standard in use being a priced risk factor.

68

3.6 Conclusion

3.6 Conclusion

The existence of two return premiums, the value premium and the accounting pre-

mium, is investigated in the South African stock market. A value premium exists, if

stock return expectations increase with the ratio of book value of equity to market

value of equity. An accounting premium exists if investors award a �rm's volun-

tary compliance to IFRS with a lower expected return. The sample of this study

consists of 159 stocks listed at the JSE Limited from July 1998 to June 2007. A

positive value premium of about 1.5% to 2% is identi�ed over the full period when

considering all stocks. This also holds within the class of small stocks. Taking into

account the high volatility prior to 2000 the results also extend to the class of big

stocks. The CAPM fails to explain the cross-section and more speci�cally the value

premium; a multifactor model dominates the CAPM. Evidence for the existence of

an accounting premium between 1.5% and 2% is found among all stocks. A closer

investigation shows that it is limited to the class of value stocks and small stocks.

Furthermore, it is shown that the accounting premium is robust against possible self-

selection. The �ndings on the value premium are qualitatively identical to recent

evidence for the U.S. market provided by Fama and French (2006). Further research

opportunities include a closer investigation of the existence and the properties of

both premiums in classes formed on di�erent risk related properties. Signi�cantly,

this would necessitate a richer data set.

69


70

4 Das Fama-French-Modell: Eine

bewährte Alternative zum

CAPM � auch in Deutschland

The Fama-French model: A proven alternative to the CAPM �

in Germany as well

Published in:

FinanzBetrieb, 11 (Juli/August 2009), 382�388

4.1 Einleitung

Das CAPMwird seit Jahren als nützliches Instrument zur Bestimmung der Eigenkap-

italkosten in der Praxis eingesetzt. In der Wissenschaft haben sich mittlerweile

jedoch alternative Modellansätze auf Basis einer groÿen Anzahl an empirischen Stu-

dien etabliert, die das CAPM systematisch zu Multifaktorenmodellen weiterentwick-

elt haben. Die neuen �Asset Pricing�-Modelle erlauben dabei einen höheren Grad an

Flexibilität im Modellansatz und unterlegen weniger restriktive Annahmen als beim

CAPM. Das heute wohl bekannteste ist das Drei-Faktoren-Modell von Eugene F.

Fama und Kenneth R. French, bei dem zwei zusätzliche Risikofaktoren, basierend

auf den Marktwertunterschieden bzw. dem Buch-zu-Marktwert-Verhältnis, höhere

Erklärungskraft zur Bestimmung der Überrenditen bzw. Eigenkapitalkosten liefern.1

In der Praxis hat sich das 1993 verö�entlichte Fama-French-Modell jedoch nicht

gegenüber dem CAPM durchgesetzt, obwohl es bei genauerer Betrachtung das em-

1Im Allgemeinen wird angenommen, dass die Eigenkapitalkosten der erwarteten Rendite einesrisikobehafteten Unternehmens entsprechen.

71

4 Das Fama-French-Modell: Eine bewährte Alternative zum CAPM

pirisch überlegene Modell zur Erklärung von Eigenkapitalkosten ist.2 Die Vorteile

des Drei-Faktoren-Modells liegen im besseren statistischen Erklärungsgehalt, ausge-

drückt durch das bereinigte Bestimmtheitsmaÿ, aber vor allem auch in der Elim-

inierung von statistisch relevanten, aber durch das CAPM unerklärbaren Rendite-

unterschieden, die in zahlreichen Studien nachgewiesen wurden � auch bekannt als

Anomalien.3

Dieser Beitrag hat zum Ziel, die Bekanntheit von Multifaktorenmodellen, ins-

besondere des Fama-French-Modells, zu steigern, sowie die Akzeptanz und Anwend-

barkeit dieser Modelle zu verbessern, und geht dabei in drei Schritten vor:

• CAPM vs. Fama-French: Zunächst werden Motivation und Grundlagen für

das Drei-Faktoren-Modell sowie die hauptsächlichen Unterschiede gegenüber

dem CAPM beschrieben. Es werden auch neuere Ansätze zur Erklärung von

Renditen in Mehrfaktorenmodellen vorgestellt und diese kritisch gewürdigt

(Abschnitt 4.2).

• Fama-French in Deutschland: Anschlieÿend wird das Fama-French-Modell auf

den deutschen Aktienmarkt für eine Einteilung von 548 börsennotierten Un-

ternehmen in Industriesegmente angewendet und aufgezeigt, dass das Modell

auch in Deutschland empirisch überlegen ist (Abschnitt 4.3).

• Implementierung in der Praxis: Schlieÿlich wird erörtert, wie das Fama-French-

Modell erfolgreich in der Praxis eingesetzt werden kann und wie sich die Daten-

verfügbarkeit auf die Anwendung des Modells auswirkt (Abschnitt 4.4).

4.2 Historie: Vom CAPM zum

Fama-French-Modell

4.2.1 CAPM

Das Capital Asset Pricing Model (CAPM) beruht auf den Arbeiten von Markowitz

zur Diversi�kation von Risiko (Markowitz, 1952, 1959) und wurde in der heute

2In einer Untersuchung gaben knapp 75% von 392 befragten CFOs an, das CAPM immer bzw.fast immer zur Bestimmung von Kapitalkosten zu benutzen, während nur ungefähr 10% ein Mul-tifaktorenmodell mit den Fama-French-Faktoren anwenden (Graham and Harvey, 2001).

3Eine Anomalie ist allgemein ein Phänomen, das von einem wissenschaftlichen Paradigma nichterklärt werden kann; mehr dazu in Frankfurter and McGoun (2001).

72

4.2 Historie: Vom CAPM zum Fama-French-Modell

bekannten Form von Sharpe (1964) und Lintner (1965) aufgestellt. Es ist das heute

am weitesten verbreitete Modell, um den Zusammenhang zwischen Rendite und

Risiko, genauer zwischen erwarteter Rendite und geschätztem Risiko, zu beschreiben:

E[Ri] = Rf + βi (E[RM ]−Rf ) (4.1)

Die erwartete Rendite einer Anlage i entspricht demnach einer risikofreien Rendite

Rf plus einem Risiko-Premium, das sich aus dem Produkt des geschätzten Beta-

Faktors der Anlage i, βi, und der Marktüberrendite, E[RM ]−Rf , ergibt. Der Beta-

Faktor beinhaltet dabei das systematische oder ideosynkratische Risiko der Anlage

i gegenüber der Marktrendite RM .

Das CAPM wird regelmäÿig zur Bestimmung der Eigenkapitalkosten verwendet,

die der erwarteten Rendite eines Eigenkapitalgebers an ein Unternehmen entsprechen.4

Auf Grund der hohen Bekanntheit des Modells sei an dieser Stelle nur auf zwei

in der Praxis relevante Probleme bei der Anwendung des CAPM verwiesen. Zum

einen muss eine geeignete Marktrendite ausgewählt werden, die exogen in das Mod-

ell aufgenommen wird. Gängig ist, entweder bei einer groÿ angelegten Studie die

Marktrendite als marktwertgewichtetes Mittel der Stichprobe selbst zu berechnen,

oder aber einen Index als Proxy zu verwenden. In der Literatur wird deshalb disku-

tiert, ob eine mögliche empirische Ablehnung des CAPM allein auf die Tatsache

zurückzuführen ist, dass die Marktrendite nicht beobachtbar ist, und damit das

Modell zwar theoretisch richtig, aber praktisch nicht überprüfbar sei (siehe z.B.

Roll, 1977). Zum anderen muss der Beta-Faktor für jede Anlage i durch lineare

Regression geschätzt werden. Der geschätzte Beta-Faktor ergibt sich als:

β̂i =cov(Ri, RM)

σ2(RM)(4.2)

Technisch kann man den Beta-Faktor interpretieren als �normiertes� Risiko für

Anlage i, das man erhält, indem man die geschätzte Kovarianz zwischen Ri und

RM , cov(Ri, RM), durch die geschätzte Varianz der Marktrendite, σ2(RM), dividiert.

Empirisch zeigt sich, dass die Schätzung von Beta auf aggregierter Portfolio-Ebene

präziser funktioniert als die Schätzung für jede einzelne Anleihe. Deshalb spielt in

allen Verö�entlichungen zu CAPM (und Fama-French-Modell) vor allem auch eine

4Siehe dazu im Rahmen dieses Journals z.B. Balz and Bordemann (2007); Aders and Wagner(2004); Timmreck (2002).

73


Rolle, wie die Portfolios der einzelnen Studien konkret zusammengesetzt werden

(siehe z.B. Ernstberger et al., 2011).

Die ersten Tests zum CAPM bestätigten grundsätzlich die Theorie, dass Anlagen

mit hohem Beta höhere Renditen aufweisen als Anlagen mit niedrigem Beta (siehe

z.B. Black et al., 1972). Jedoch wurde auch gezeigt, dass es einen signi�kanten

Renditeanteil gibt, der nicht durch das CAPM erklärbar ist. Jensen (1968) veröf-

fentlichte dazu die Idee, die Ursprungsversion des CAPM aus Gleichung 4.1 um einen

zusätzlichen additiven Parameter αi zu erweitern, durch das sog. Jensen-Alpha:

E[Ri]−Rf = αi + βi (E[RM ]−Rf ) (4.3)

Im klassischen CAPM müsste das Jensen-Alpha demnach Null sein. Empirische

Studien zeigen jedoch immer wieder, dass das geschätzte α̂i signi�kant von Null

verschieden ist, was zumindest für eine Erweiterung des CAPM um diesen Parameter

spricht.

Im Laufe der Zeit wurden jedoch eine Reihe weiterer Anomalien gegenüber dem

CAPM gefunden, die hier nur exemplarisch aufgeführt werden soll:

Size-E�ekt: Die durchschnittliche Rendite von Unternehmen mit geringer Mark-

tkapitalisierung ist höher als durch das CAPM vorhergesagt wird (Banz, 1981).

Price-Earnings-Ratio-E�ekt: Unternehmen mit einem hohen Kurs-Gewinn-Verhältnis

(KGV) haben im Durchschnitt höhere risikobereinigte Renditen als Unternehmen

mit niedrigem KGV (Basu, 1983).

Leverage-E�ekt: Die erwartete Rendite von Unternehmen steht in positivem Zusam-

menhang mit dem Verschuldungsgrad (Fremdkapital/Eigenkapital) (Bhandari,

1988).

Book-to-Market-E�ekt: Untersuchungen zeigen statistisch signi�kante positive Ko-

rrelationen zwischen den durchschnittlichen Renditen und dem Buchwert-zu-

Marktwert-Verhältnis von Unternehmen. Unternehmen mit hohem (niedrigem)

Buchwert-zu-Marktwert-Verhältnis werden auch als �Value stocks� (�Growth

stocks�) bezeichnet (Stattman, 1980).

Fama and French (1992) synthetisierten schlieÿlich diese Befunde und attestierten

einen hohen Erklärungsgehalt dieser Faktoren über das klassische CAPM hinaus.

74


4.2.2 Fama-French-Modell

1993 verö�entlichten Fama und French ihr Drei-Faktoren-Modell (Fama and French,

1993), mit dem sie das klassische CAPM um zwei innovative Faktoren erweiterten:

E[Ri] = Rf + βiM (E[RM ]−Rf ) + βiSMB E[SMB ] + βiHMLE[HML] (4.4)

Der vordere Teil entspricht der CAPM-Gleichung aus Gleichung 4.1. Der Faktor

SMB steht für �small minus big�, der Faktor HML für �high minus low�, die dazuge-

hörigen Betas, βiSMB bzw. βiHML, repräsentieren die entsprechenden Parameter im

geschätzten Regressionsmodell (auch als �factor loadings� bezeichnet). SMB stellt

dabei den Renditeunterschied zwischen Unternehmen mit groÿer und kleiner Mark-

tkapitalisierung dar und inkorporiert damit den �Size-E�ekt�. Demgegenüber steht

HML für den Renditeunterschiede zwischen Unternehmen mit hohem und niedrigem

Buchwert-zu-Marktwert-Verhältnis (�Book-to-Market-E�ekt�).

Die beiden neuen Faktoren werden auf Basis sog. �factor-mimicking� Portfolios

berechnet. Dazu wird die Stichprobe in sechs Portfolios aufgeteilt:

SH �Small-High� (kleine Marktkapitalisierung, hoher Buchwert/Marktwert)

SM �Small-Medium� (kleine Marktkapitalisierung, mittlerer Buchwert/Marktwert)

SL �Small-Low� (kleine Marktkapitalisierung, niedriger Buchwert/Marktwert)

BH �Big-High� (groÿe Marktkapitalisierung, hoher Buchwert/Marktwert)

BM �Big-Medium� (groÿe Marktkapitalisierung, mittlerer Buchwert/Marktwert)

BL �Big-Low� (groÿe Marktkapitalisierung, niedriger Buchwert/Marktwert)

Während der Median typischer Weise zwischen �small� und �big� diskriminiert,

werden die gröÿten 30% der Buchwert/Marktwert-Verhältnisse als �high�, die niedrig-

sten als �low� und den restlichen als �medium� eingestuft. Die Unternehmen verbleiben

in den Portfolios für 12 Monate, bevor eine erneute Sortierung statt�ndet. Die bei-

den Fama-French-Faktoren werden schlieÿlich auf Monatsbasis wie folgt berechnet:

SMB = (RSH +RSM +RSL)/3− (RBH +RBM +RBL)/3 (4.5)

HML = (RSH +RBH)/2− (RSL +RBL)/2 (4.6)

75


Das Fama-French-Model bringt drei groÿe Vorteile mit sich:

• Die Erklärungskraft des Modells, ausgedrückt im bereinigten Bestimmtheits-

maÿ, steigt.

• Das Jensen-Alpha wird insigni�kant von Null verschieden, sprich es bleibt kein

unerklärbarer Rest zurück.

• Die angeführten Anomalien werden durch das Modell aufgefangen, auch der

Price-Earnings-Ratio-E�ekt sowie der Leverage-E�ekt.

Als Begründung für den Erfolg des Modells werden in der Literatur verschiedene

Ursachen diskutiert. Diese reichen z.B. von dem Argument, dass dies zugrun-

deliegende undiversi�zierbare Risiken sind, die zukünftige unsichere Ereignisse vor-

wegnehmen,5 über irrationales Verhalten der Investoren, bis hin zu dem Vorwurf,

dass die Ergebnisse abhängig von der gewählten Stichprobe seien. Abgesehen von

letzterem, ist es für die Bestimmung der Eigenkapitalkosten selbst jedoch nicht

entscheidend, ob es rationale oder irrationale Gründe sind, die letztendlich zum Er-

folg des Models führen, da beide als Bestandteil der Opportunitätskosten anzusehen

und damit einzupreisen sind (Stein, 1996).

4.2.3 Multifaktorenmodelle: Weiterentwicklungen und Kritik

Seit der Einführung des Drei-Faktoren-Modells wurde dieses in den unterschiedlich-

sten Studien und Ländern angewendet. Dabei gab es eine groÿe Anzahl von Auf-

sätzen, die das Modell untermauert, allen voran angeführt von den Forschungen von

Fama und French selbst (Z.B. in Fama and French, 1998). Das Modell gilt heute als

empirisch belegt und wird auch erfolgreich zur Bestimmung der Eigenkapitalkosten

eingesetzt (Siehe z.B. Fama and French, 1997).

Es wurden aber auch alternative Ansätze vorgestellt, die zum Ziel haben, das

Modell weiterzuentwickeln. An dieser Stelle soll exemplarisch auf zwei dieser Ansätze

eingegangen werden.

Zum einen gibt es eine bekannte Anomalie, die Fama und French nicht in ihr Mod-

ell aufgenommen haben, nämlich den von Jegadeesh and Titman (1993) nachgewiese-

nen �Momentum-E�ekt�. Dieser besagt, dass kurzfristig gesehen erfolgreiche Aktien

5Begründung der Autoren selbst.

76


oft weiter erfolgreich sind, während schwache Aktien oft weiter schlecht abschneiden.

Carhart (1997) hat darauf hin diesen E�ekt in sein Vier-Faktoren-Modell aufgenom-

men, zusätzlich zu der bekannten Fama-French-Spezi�kation. Der �Momentum-

Faktor� misst dabei den Renditeunterschied von kurzfristigen Gewinner- und Verlierer-

Portfolios. Fama und French erkannten zunächst auch an, dass ihr Modell an

dieser Stelle nicht in der Lage ist, diese Portfoliounterschiede zu erklären (Fama

and French, 1996a). Später zeigten andere jedoch auch (z.B. Ray et al., 2008),

dass Carharts Vier-Faktoren-Modell empirisch nicht prinzipiell dem Drei-Faktoren-

Modell von Fama und French überlegen ist, da keine signi�kanten Unterschiede im

Jensen-Alpha bescheinigt werden können.

Zum anderen gibt es auch Ansätze, Rechnungslegungsinformationen zur Erk-

lärung von Renditen aufzunehmen, wie beispielsweise Eigenschaften von bilanziellen

Unternehmenserfolgen (�earnings attributes�) � z.B. die Persistenz, Gleichmäÿigkeit

und Vorhersagbarkeit dieser Erfolge (Francis et al., 2004) � oder die Eigenschaften

des Unterschieds zwischen bilanziellen Unternehmenserfolgen und Cash�ows einer

Periode (�accruals quality�) (Francis et al., 2005a). Dabei werden die verschiede-

nen Rechnungslegungsgröÿen als zusätzliche Faktoren in das Fama-French-Modell

aufgenommen, wie etwa ein �Accruals Quality�-Faktor auf Basis entsprechender �fac-

tor mimicking�-Portfolios. Für den deutschen Aktienmarkt zeigen Ernstberger and

Vogler (2008), dass die Eigenkapitalkosten auf Basis des CAPM signi�kant niedriger

sind für Unternehmen, die nach internationalen Rechnungslegungsstandards, d.h.

IAS/IFRS bzw. U.S. GAAP, bilanzieren als für Unternehmen, die nach HGB ihre

Bücher führen. Diese Unterschiede verschwinden jedoch, wenn man ein neues Multi-

faktorenmodell anwendet, das mit zwei zusätzlichen Faktoren das �Accounting Pre-

mium� au�ängt. Weiterhin tragen jedoch auch die Fama-French-Faktoren zur Erk-

lärung von erwarteten Renditen, und damit den Eigenkapitalkosten, bei.

Trotz dieser Erfolge bleiben natürlich auch kritische Stimmen gegenüber Multi-

faktorenmodellen bestehen, die sowohl in der Theorie als auch in der praktischen

Anwendung verwurzelt sind.

In der Theorie fehlt allen Multifaktorenmodellen, so auch dem Fama-French-

Modell, die mikroökonomische Fundierung. Während das CAPM ein mit genauen

Annahmen hinterlegtes Gleichgewichtsmodell ist, wird dies in Multifaktorenmod-

ellen nicht weiter spezi�ziert. Im Gegenteil: Nach der �Arbitrage Pricing Theory

(APT)� von Ross (1976) steht die Begründung der ökonomischen Natur und Höhe

77


von Risikoprämien nicht im Vordergrund, sondern es wird sich auf die Identi�kation

von gemeinsamen Faktoren zur Erklärung des systematischen Risikos beschränkt.

Die Kritik an Multifaktorenmodellen ist deshalb, dass nicht unmittelbar de�niert

ist, ob die Faktorrisikoprämien mit einem spezi�schen ökonomischen Gleichgewicht

konsistent sind. Da die Testbarkeit des ökonomischen Gleichgewichts jedoch auch

im klassischen CAPM nicht gegeben ist (das e�ziente Marktportfolio muss immer

geschätzt werden), ist diese Kritik für die Praxis von untergeordneter Bedeutung.

Von höherer praktischer Relevanz sind jedoch empirische Studien, die kritisch

überprüfen, ob das Fama-French-Modell tatsächlich überlegen ist oder ob andere

Modelle vorzuziehen sind. Zwei Arbeiten sollen hier exemplarisch angeführt werden:

Mit am heftigsten kritisiert wurden die Arbeiten von Fama und French wohl von

Kothari et al. (1995). Sie argumentieren, dass der Ein�uss des Buch-zu-Marktwert-

Faktors schwächer und weniger konsistent ist sowie dass in den Daten von Fama und

French ein �selection bias� vorliegt. Darauf gehen Fama und French in einer eigens

dafür verfassten Verö�entlichung im Journal of Finance ein und stellen abschlieÿend

noch einmal klar: �The average-return anomalies of the CAPM are serious enough

to infer that the model is not a useful approximation� (Fama and French, 1996b, S.

1957).

Eng an das CAPM geknüpft bringen demgegenüber Jagannathan andWang (1996)

einen alternativen Modell-Vorschlag hervor, der zumindest in einem konditionalen

Sinne auf die CAPM-Theorie aufbaut (�conditional CAPM�). Sie entwickeln es von

einem statischen CAPM weiter hin zu einem dynamischen Modell. Schlieÿlich en-

thält aber auch ihr Modellvorschlag nicht mehr nur einen Beta-Faktor sondern drei,

wobei sowohl die Rendite auf Humankapital sowie die Zeitinstabilität des klassischen

Beta-Faktors berücksichtigt werden.

4.3 Das Fama-French-Modell in Deutschland

4.3.1 Bisherige empirische Erkenntnisse

Während die grundsätzlichen CAPM-Anomalien auch in Deutschland schon aus-

führlich getestet und nachgewiesen wurden (Beiker and Steiner, 1993; Sattler, 1994;

Wallmeier, 2000), spielt die empirische Überprüfung des Fama-French-Modells per

se bisher eine vernachlässigte Rolle. Eine Ausnahme ist die Studie von Ziegler et al.

78


(2007), in der der klassische Fama-French Ansatz aus dem Jahr 1993 für Deutsch-

land getestet wird. Das Ergebnis ist, dass das Fama-French-Modell eine höhere

Erklärungskraft besitzt und dass der Jensen-Alpha-Test besser ausfällt als für das

CAPM, wenn auch nur 6 von 16 Alphas im CAPM signi�kant sind.

Insgesamt beschränkt sich der Erfolg des Modells jedoch auf die von Fama und

French vorgenommene Einteilung der Aktien in �factor-mimicking� Portfolios. Das

heiÿt, der Erfolg des Modells beruht auf der Tatsache, dass die Unternehmen vorher

in Portfolios eingeordnet wurden, die Gröÿen- und Buch-zu-Marktwert-Unterschiede

nachbilden. Dies gilt grundsätzlich für die meisten Studien zum Mehrfaktorenmod-

ell, auch für die von Fama and French (1998) selbst durchgeführte internationale

Studie, in der auch der deutsche Markt behandelt wird. Nur in ihrer früheren Veröf-

fentlichung wenden Fama and French (1997) bewusst das Modell auf Industrieseg-

mente an, weisen zugegebener Maÿen jedoch auch hier den Erfolg des Modells nach.

Ziel dieses Beitrags ist es deshalb, den Erfolg des Fama-French-Modells in Deutsch-

land auch für einzelne Industriesegmente zu testen.

4.3.2 Daten und Portfolio-Bildung

Zur Überprüfung des Fama-French-Modells werden zunächst alle in Deutschland

gehandelten und in Datastream Advance verfügbaren Unternehmen im Zeitraum

von Juli 1995 bis Juni 2005 (120 Monate) betrachtet. Diese Stichprobe wird re-

duziert um Unternehmen mit negativem Buchwert sowie Finanzinstitutionen (SIC-

Code 6000�6999), da für diese besondere Rechnungslegungsvorschriften gelten. In-

sgesamt verbleiben damit 548 Unternehmen in der Stichprobe.

Die zu erklärenden Renditen werden daraufhin in 12 Industriesegmente nach der

Industrieklassi�zierung von French (2009) gruppiert (Tabelle 4.1).

Die erklärenden Variablen werden wie folgt bestimmt. Die risikolose Rendite, Rf ,

wird durch den dreimonatigen EURIBOR-Referenzzinssatz approximiert. Die Mark-

trendite, RM , ist das marktwertgewichtete Mittel aller verfügbaren Unternehmen

(inkl. SIC 6000�6999). Die Berechnung der beiden Fama-French Faktoren, SMB

und HML, erfolgt auf Basis der oben beschrieben sechs �factor mimicking�-Portfolios

sowie der Formeln 4.5 und 4.6. Nach 12 Monaten (jeden Juli) werden die Portfo-

lios neu zusammengestellt. Die deskriptive Statistik der Variablen ist in Tabelle 4.2

zusammengefasst. Es fällt vor allem auf, dass die Fama-French-Faktoren SMB und

HML weder untereinander noch mit der Marktrisikoprämie, RM − Rf , stark korre-

79


Table 4.1: Industrieklassi�zierung

Industrie- Beobach- Ø gesamter Markt- Ø gesamter

Segment Bezeichnung tungen wert (Mrd. EUR) Buch-/Marktwert

IND01 Kurzlebige Verbrauchsgüter 38 15.3 0.48

IND02 Gebrauchsgüter 29 115.9 0.55

IND03 Maschinen-/Anlagenbau 116 46.4 0.51

IND04 Energieerzeugung 1 3.4 0.63

IND05 Chemie 18 58.7 0.48

IND06 Computer 127 110.4 0.53

IND07 Telekommunikation 15 108.6 0.42

IND08 Versorger 15 90.6 0.45

IND09 Handel 72 47.2 0.52

IND10 Gesundheit 27 77.6 0.54

IND11 Finanzwesen - - -

IND12 Sonstige 90 37.0 0.54

Summe 548 711.1 -

Median - - 0.52

Notizen: Industrieklassi�zierung nach French (2009) auf Basis des SIC-Codes in zwölf Indus-triesegmente; Finanzinstitutionen (IND11, SIC-Code 6000�6999) werden nicht betrachtet.

80


Table 4.2: Deskriptive Statistik der Variablen

Korrelationskoe�zienten

Variable Mittelwert Std.abw. Rf RM RM −Rf SMB HML

Rf 0.0027 0.0007 1

RM 0.0146 0.0587 -0.1270 1

RM −Rf 0.0119 0.0587 -0.1384 0.9999 1

SMB -0.0113 0.0426 -0.3092 -0.1636 -0.1597 1

HML 0.0133 0.0489 0.0761 -0.3135 -0.3139 -0.3759 1

Notizen: Deskriptive Statistik der Variablen, bei RM inklusive Finanzinstitutionen (IND11, SIC-Code 6000�6999).

liert sind, was die Erwartung an die Erklärungskraft dieser Faktoren verbessert.

4.3.3 Regressionsergebnisse

Für die Marktwert-gewichteten Renditen der elf relevanten Industriesegmente des

deutschen Aktienmarkts werden sowohl das CAPM als auch das Fama-French-Modell

als Zeitreihenregression auf Basis der 120 Monate geschätzt.

Es ergibt sich, dass bereits das CAPM respektabel abschneidet (Tabelle 4.3). Mit

Ausnahme des Energieerzeugungssegments (IND04), für das nur ein beobachtetes

Unternehmen in der Stichprobe vorliegt, ist das bereinigte Bestimmtheitsmaÿ min-

destens 21,63% (IND08) bis maximal sogar 71,49% (IND12). Im Mittel liegt der

Erklärungsgehalt bei 45,3% (gleichgewichtet) bzw. 49,98% (Marktwert-gewichtet).

Abgesehen von IND04 sind alle Beta-Werte hochsigni�kant und sind auf ein ökonomisch

sinnvolles Intervall von 0,41 (IND08) bis 1,69 (IND06) verteilt. Das Jensen-Alpha

fällt nur für das Segment der kurzlebigen Verbrauchsgüter signi�kant von Null ver-

schieden aus. Alle anderen Alpha-Werte sind nicht signi�kant von Null verschieden.

Dies ist bemerkenswert, haben doch vor allem amerikanische Studien immer wieder

belegt, dass das Jensen-Alpha eigentlich zusätzliche Erklärungskraft besitzt. Allerd-

ings fällt auch bei Ziegler et al. (2007) der Jensen-Test für den deutschen Markt

relativ schwach aus.

Was die Signi�kanz der individuellen Parameter angeht, schneidet das CAPM für

die Industriesegmente damit noch besser ab als bei Ziegler et al. (2007), die �factor-

mimicking� Portfolios zugrunde legen. Der Erklärungsgehalt ist jedoch etwas kleiner

81


Table 4.3: CAPM Schätzungen: E[Ri] = Rf + βi (E[RM ]−Rf )

Industriesegment αi βi Ber.R2

IND01 0.0063 * 0.4376 *** 0.3397

IND02 -0.0031 1.0181 *** 0.5504

IND03 0.0001 0.7903 *** 0.6896

IND04 0.0034 -0.0590 -0.0038

IND05 0.0031 0.6334 *** 0.4291

IND06 -0.0033 1.6914 *** 0.7138

IND07 -0.0007 1.2153 *** 0.4451

IND08 0.0031 0.4072 *** 0.2163

IND09 0.0007 0.6702 *** 0.4224

IND10 0.0038 0.6856 *** 0.4653

IND12 -0.0023 0.9947 *** 0.7149

Mittel, gleichgewichtet 0.0010 0.7713 0.4530

Mittel, Marktwert-gewichtet 0.0000 0.9501 0.4998

Notizen: CAPM Schätzungen, Juli 1995�Juni 2005 (120 Monate), 548 Unternehmen aufgeteilt inelf Industriesegmente (IND01 bis IND12; IND11 wird nicht betrachtet, da es sich hierbei um Fi-nanzinstitutionen handelt); *, ** bzw. *** bedeutet, dass diese Werte auf einem Signi�kanzniveauvon 10%, 5% bzw. 1% von Null verschieden sind.

im Vergleich zu den durchschnittlichen 57,1% bei Ziegler et al. (2007), was auch an

deren längeren Betrachtungszeitraum (324 Monate) liegt.

Trotz der bereits hohen Modellgüte des CAPM bei der Erklärung der Renditeun-

terschiede in den Industriesegmenten, fällt das Ergebnis für das Fama-French-Modell

noch besser aus (Tabelle 4.4). Auch hier muss zwar das Energieerzeugungssegment

(IND04) wegen zu weniger Beobachtungen in der Stichprobe bei der Interpretation

ausgeklammert werden. Allerdings fallen die Werte des bereinigten Bestimmtheits-

maÿes ausnahmslos für alle Industriesegmente vorteilhafter aus als beim CAPM. Im

Schnitt verbessert sich der Erklärungsgehalt um ca. 5%-Punkte auf 50,42% (gle-

ichgewichtet) bzw. 55,73% (Marktwert-gewichtet).

Au�ällig ist vor allem, dass die hinzugefügten Fama-French-Faktoren zwar nicht

immer, aber meist signi�kant sind (in 15 von 22 Fällen) und Erklärungskraft liefern,

obwohl der Jensen-Alpha-Test beim CAPM ergeben hatte, dass zumindest als Kon-

stante kein Faktor �vergessen� wurde. Trotzdem bleiben alle Marktbeta-Werte

hochgradig signi�kant. Das Jensen-Alpha ist wie beim CAPM nur in einem Fall

auf 10%-Signi�kanzniveau von Null unterscheidbar (IND02).

82

4.4 Implementierung in der Praxis

Die Interpretation der Vorzeichen gestaltet sich bei dieser Analyse allerdings

schwierig. Auf den ersten Blick fällt auf, dass IND07 und IND08 bei SMBB sowie

IND06 und IND07 bei HML ein negatives Vorzeichen aufweisen. Während in den

nach Gröÿe und Buch-zu-Marktwert �sortierten� Portfolios, die man typischer Weise

für die Fama-French-Analyse verwendet, negative Vorzeichen bei SMBB bei groÿen

Unternehmen und negative Vorzeichen bei HML bei Unternehmen mit niedrigem

Buch-zu-Marktwert-Verhältnis auftreten, ist dieses Muster auf die �naive� Industrie-

Betrachtung schwer zu übertragen. Zwar sind IND07 und IND08 tendenziell (im

Durchschnitt) eher gröÿere und IND06 und IND07 eher Unternehmen mit niedrigem

Buch-zu-Marktwert-Verhältnis. Eine genauere Analyse des Vorzeichens lässt die Ag-

gregation der Daten in diesem Fall jedoch leider nicht zu.

Was den Erklärungsgehalt des Fama-French-Modells angeht bleibt dieser zwar

auch hier auf Grund des kleineren Beobachtungszeitraums etwas gegenüber dem

Ergebnis von Ziegler et al. (2007) zurück, die Signi�kanz der Beta-Werte ist aber

deutlich gröÿer, da bei Ziegler et al. (2007) nur 35 von 48 Parametern signi�kant

sind und gleichzeitig dort vor allem auch die Fama-French-Faktoren dem Marktbeta

Erklärungsgehalt nehmen (nur noch 10 statt vorher 13 sind signi�kant).

Insgesamt lässt sich konstatieren, dass das Fama-French-Modell für die Indus-

triesegmente in Deutschland nicht nur besser abschneidet als das CAPM, sondern

auch, dass es für Industriesegmente anwendbar ist und nicht nur für nach den

Fama-French-Faktoren sortierte Portfolios, wie z.B. bei Ziegler et al. (2007). Die

statistische Inferenz auf individueller Unternehmensebene, wie das Ergebnis des

Energieerzeugungssegments (IND04) im vorliegenden Fall gezeigt hat, bleibt aber

problematisch, ebenso wie die Interpretation der Vorzeichen auf Basis dieses Aggre-

gationsniveaus.


Bei der Bestimmung der Eigenkapitalkosten in der Praxis stellt sich nun die Frage,

wie und unter welchen Umständen das Fama-French-Modell implementiert werden

kann. Wie eingangs bereits erwähnt, geben nur ca. 10% der CFOs an, ein Multi-

faktorenmodell mit den Fama-French-Faktoren in der Praxis zur Bestimmung der

Eigenkapitalkosten anzuwenden (Graham and Harvey, 2001).

Neben der geringeren Bekanntheit gegenüber dem CAPM lassen sich hier vor

83


Table 4.4: Fama-French Schätzungen:E[Ri] = Rf + βiM (E[RM ]−Rf ) + βiSMB E[SMB ] + βiHMLE[HML]

Industriesegment αi βiM βiSMB βiHML Ber. R2

IND01 0.0049 0.5150 *** 0.2002 ** 0.2073 *** 0.3765

IND02 -0.0088 * 1.1318 *** 0.0838 0.3975 *** 0.5894

IND03 -0.0027 0.9245 *** 0.3253 *** 0.3689 *** 0.7725

IND04 0.0019 -0.0023 0.1196 0.1637 -0.0019

IND05 -0.0041 0.7180 *** -0.1112 0.3730 *** 0.5465

IND06 0.0072 1.5723 *** 0.1765 -0.5336 *** 0.7712

IND07 0.0064 0.9846 *** -0.4298 ** -0.6919 *** 0.5108

IND08 -0.0020 0.4369 *** -0.1849 * 0.1953 ** 0.2855

IND09 0.0010 0.7308 *** 0.2402 ** 0.1250 0.4357

IND10 -0.0010 0.7318 *** -0.1100 0.2257 ** 0.5082

IND12 -0.0028 1.0985 *** 0.3470 *** 0.2435 *** 0.7511

Mittel, gleichgewichtet -0.0000 0.8038 0.0597 0.0977 0.5042

Mittel, Marktwert-gewichtet -0.0002 0.9508 -0.0092 0.0068 0.5573

Notizen: Fama-French Schätzungen, Juli 1995�Juni 2005 (120 Monate), 548 Unternehmenaufgeteilt in elf Industriesegmente (IND01 bis IND12; IND11 wird nicht betrachtet, da es sichhierbei um Finanzinstitutionen handelt); *, ** bzw. *** bedeutet, dass diese Werte auf einemSigni�kanzniveau von 10%, 5% bzw. 1% von Null verschieden sind.

84


allem die mangelnde Überzeugung von dem Modell an sich sowie die mutmaÿlich

schlechtere Datenverfügbarkeit als Hinderungsgründe anführen. Beide Probleme

sollen an dieser Stelle adressiert werden.

4.4.1 Wann ist das Drei-Faktoren-Modell die richtige

Modell-Wahl?

Dass das CAPM empirische Schwächen aufweist (Jensen-Alpha, Anomalien), diese

allerdings mit dem Fama-French-Modell meist erfolgreich abgewendet werden kön-

nen, wurde oben bereits gezeigt. Unter den �Asset Pricing�-Modellen, wie beide

auch genannt werden, ist es also ratsam, das Drei-Faktoren-Modell zur Bestimmung

der Eigenkapitalkosten für �factor-mimicking�-Portfolios und Industriesegmente dem

CAPM vorzuziehen.

Dem gegenüber steht jedoch noch eine alternative Konzeption, die sich mit der

Bestimmung der Eigenkapitalkosten auf Basis zukunftsorientierter Schätzungen be-

fasst (auch �Implied Cost of Capital� genannt). Dabei werden zukünftige Über-

schusserwartungen (meist Dividenden) so diskontiert, dass sich der heutige Ak-

tienkurs des Unternehmens widerspiegelt. Die Eigenkapitalkosten entsprechen dabei

dem ermittelten Diskontierungssatz. Daske and Wiesenbach (2005) fassen die gängi-

gen Modell im Rahmen dieser Zeitschrift übersichtlich zusammen und stellen fest,

dass zukunftsorientierte Ableitungen der Eigenkapitalkosten �eine gute Alternative

zu den vergangenheitsorientierten Schätzmethoden darstellen�. Demgegenüber ver-

gleichen Easton and Monahan (2005) aber sieben verschiedene zukunftsorientierte

Schätzverfahren und �nden heraus, dass diese Verfahren im Allgemeinen �unzuver-

lässige Schätzungen� und nur mit Einschränkungen anwendbar sind.

Generell wird in der Literatur diskutiert, welche Konzeption die empirisch sta-

bilere zur Bestimmung der Eigenkapitalkosten ist. Während bei den �Asset Pricing�-

Verfahren die Verlässlichkeit historischer Daten in Frage gestellt wird, sind bei den

�Implied Cost of Capital�-Modellen vor allem die Qualität der Analystenschätzun-

gen per se aber auch die Festlegung des Prognosezeitraums sowie die Bestimmung

des Endwerts (�terminal value�) problematisch (Easton, 2006). Die Entscheidung

für oder gegen das Fama-French-Modell hängt somit also von der konzeptionellen

Wahl für oder gegen realisierte Renditen zur Bestimmung der Eigenkapitalkosten

ab.

85


4.4.2 Wie wirkt sich die Datenverfügbarkeit in der Praxis

aus?

Als zweiter Grund gegen den Einsatz des Fama-French-Modells soll hier die mut-

maÿliche mangelnde Datenverfügbarkeit angesprochen werden. Während im CAPM

nur die risikolose Rendite und die Marktrendite als exogene Faktoren approximiert

werden müssen, benötigt man für das Drei-Faktoren-Modell zusätzlich die beiden

Fama-French-Faktoren SMB und HML. Diese müssen prinzipiell erst auf Basis von

�factor-mimicking�-Portfolios geschätzt werden. Zwar bietet French (2009) für den

amerikanischen Markt diese Faktoren zum freien Download laufend aktualisiert im

Internet an. Allerdings wurde wiederholt gezeigt, dass landesspezi�sche Faktoren

besser erklären als globale Faktoren (z.B. Gri�n, 2002). Insofern wirkt sich die

Bestimmung der beiden zusätzlichen Faktoren als Zusatzaufwand im Vergleich zum

CAPM in der Praxis aus, was jedoch kein Problem der Datenverfügbarkeit an sich

ist, da die historischen Renditen ohnehin als zu erklärende Variablen und ggf. zur

Berechnung der Marktrendite vorliegen müssen.

Gegenüber dem Alternativkonzept, der zukunftsorientierten Schätzung von Eigenkap-

italkosten, haben CAPM und das Fama-French-Modell jedoch hinsichtlich der Daten-

verfügbarkeit einen tatsächlichen strukturellen Vorteil, da sie keine Analystenschätzun-

gen benötigen. Zum einen liegen Gewinnprognosen auf Basis von Analystenschätzun-

gen nur für weniger als die Hälfte der börsennotierten deutschen Unternehmen

überhaupt vor.6 Zum anderen ist die Verfügbarkeit von Analystenschätzungen mit

zusätzlichen Kosten verbunden, da sie häu�g nicht in den Standardversionen der

gängigen Finanzdatenbanken (z.B. Datastream Advance) hinterlegt sind. In der

Praxis scheitert die Anwendung der �Implied Cost of Capital�-Methode deshalb häu-

�g daran, dass keine oder zu wenige Analystenschätzungen vorliegen bzw. abrufbar

sind. Insgesamt spricht die Datenverfügbarkeit deshalb für die Anwendung von �As-

set Pricing�-Modellen. Innerhalb dieser ist das Fama-French-Modell das empirisch

überlegene Modell, wenn auch mit etwas Mehraufwand hinsichtlich der Datengener-

ierung gegenüber dem CAPM verbunden.

6Z.B. lagen in dem erwähnten Aufsatz von Daske and Wiesenbach (2005) nur für 303 von 717deutschen börsennotierten Unternehmen Gewinnprognosen in Bloomberg vor, wovon noch einmal97 weniger als drei verschiedene Analystenprognosen pro Unternehmen aufwiesen.

86

4.5 Zusammenfassung

4.5 Zusammenfassung

In diesem Beitrag wurde gezeigt, dass das Fama-French-Modell grundsätzlich em-

pirisch vorteilhafter für die Bestimmung der Eigenkapitalkosten ist als das CAPM,

weil es eine höhere Erklärungskraft besitzt und Anomalien au�ängt, wenn auch

weiter bestimmte Kritikpunkte, wie z.B. der Mangel an theoretischer Fundiertheit,

anzuzeigen sind. Die Vorteilhaftigkeit des Drei-Faktoren-Modells wurde auch für

Unternehmen des deutschen Aktienmarkts, die in Industriesegmente gruppiert wur-

den, nachgewiesen, ohne dabei auf eine Gruppierung in �factor mimicking�-Portfolios

zurückgreifen zu müssen. Allerdings zeigt sich, dass das CAPM in Deutschland rel-

ativ gesehen noch gut funktioniert, da der Jensen-Alpha Test schwächer ausfällt als

in internationalen Studien. Insgesamt emp�ehlt sich für die Praxis die Anwendung

des Fama-French-Modells auf Grund der einfacheren Datenverfügbarkeit gegenüber

zukunftsorientierten Konzepten (es werden keine Analystenschätzungen benötigt)

und der grundsätzlichen empirischen Überlegenheit gegenüber dem CAPM.

87


88

5 Economic Consequences of

Accounting Enforcement

Reforms: The Case of Germany

Revise and resubmit (4th and �nal round) in:

European Accounting Review (with Jürgen Ernstberger and Michael Stich)

5.1 Introduction

In the accounting literature, enforcement is often interpreted broadly as the pro-

cedures and mechanics to ensure observance of, or obedience to, security laws or

investor protection laws. This study deals with a speci�c area of enforcement, i.e.,

�nancial reporting enforcement. This is de�ned as �monitoring compliance of the

�nancial information with the applicable reporting framework; taking appropriate

measures in case of infringements discovered in the course of enforcement� (Com-

mittee of European Securities Regulators, 2003, Standard No. 1, Principle 2).

At the end of 2004, the German government enacted three new pieces of legisla-

tion that aimed to enhance the degree of �nancial-reporting enforcement for publicly

traded companies in Germany by, �rst, establishing a two-tier external enforcement

mechanism, second, enacting new independence rules for auditors, and, third, re-

structuring the auditor oversight. The goal of these reforms is to provide reliable

�nancial reports to investors and to restore credibility in the �nancial markets after

several accounting scandals (e.g., Comroad and Flowtex). These new regulations

are e�ective since 2005.

In this study, we explore the impact of these German reforms to the enforcement

of �nancial reporting on earnings management, stock liquidity, and the market val-

uation of a�ected companies. This research directly addresses an issue of great

89

5 Economic Consequences of Accounting Enforcement Reforms

importance to investors and regulators all over Europe. The so-called �IAS Regula-

tion� (EC, 2002: No. 1606/2002) not only prescribes the application of International

Financial Reporting Standards (IFRS) for publicly traded companies in Europe, but

also requires all EU Member States to install e�ective mechanisms for the enforce-

ment of IFRS. Thus, our research has practical implications for investors as well as

for regulators establishing new enforcement mechanisms because it documents how

the overall degree of enforcement a�ects reporting outcomes and capital market

properties.

The results of our study also emphasize the important role that enforcement

plays in shaping the quality of �nancial reports. In doing so, the paper responds

to a suggestion of Holthausen (2009) to analyze the impact of external enforcement

reforms on reporting outcomes and capital market properties. It sheds light on the

bene�ts of regulatory reforms on external �nancial reporting enforcement.

Other studies investigate the impact of broad measures of enforcement on ac-

counting quality as well as capital market properties (e.g., Ball et al., 2000; Hung,

2000; Ball et al., 2003; Bhattacharya et al., 2003; Leuz et al., 2003; Bushman et al.,

2004; Eleswarapu et al., 2004; Jain et al., 2006; DeFond et al., 2007; Ashbaugh-Skaife

et al., 2008). We contribute to this debate by examining the impact of mandatory

reforms that pertain speci�cally to the enforcement of �nancial reporting. Almost

all the previous studies on enforcement focus on common-law countries. Only three

recent empirical studies (Brown et al., 2008; Gassen and Ashbaugh-Skaife, 2009;

Ernstberger et al., 2010) , focus on Germany which can be seen as a typical code-

law country. Two of these examine the impact of internal control and audit reforms

enacted by the German government in the Gesetz zur Kontrolle und Transparenz im

Unternehmensbereich (KonTraG�Act of Control and Transparency of Enterprises)

in 1998 and �nd that the reforms have increased the frequency of quali�ed audit

opinions, the information content of �rst-time going-concern audit reports, the de-

mand for large-scale auditors, the number of auditor lawsuits, and earnings quality.

Ernstberger et al. (2010) focus on the e�ectiveness of the �name and shame� mech-

anism as established by the Bilanzkontrollgesetz (BilKoG�Accounting Enforcement

Act) of 2004 as a speci�c feature of the new external enforcement system. They ex-

amine the market reactions to the publication of error announcements by infringing

companies and �nd that the error announcements represent new, negative informa-

tion. They conclude that the �name and shame� mechanism of the new external

90

5.1 Introduction

enforcement system is e�ective in penalizing infringing companies.

We contribute to these studies by examining the overall market impact of the

new enforcement reforms on earnings quality, stock liquidity, and market valuation

of several a�ected companies. This allows us to provide evidence whether these re-

forms had positive e�ects on the quality and trustworthiness of �nancial reports and

which companies particularly bene�ted from these reforms. To assess the impact of

the enforcement reforms, we conduct multivariate analyses using di�erent segments

of the German stock market which are unequally a�ected by these reforms. We

explicitly control for the level and changes in other institutional factors, such as

auditors, cross-listings, as well as for di�erent accounting principles.

The results show that the enforcement reforms lead to a signi�cant decrease in

earnings management coming along with higher earnings quality. Further, we doc-

ument an increase in stock liquidity as well as in companies' market valuation. A

second set of tests investigates whether all companies are a�ected equally by the re-

forms, or whether companies characterized by a weaker overall enforcement through

other mechanisms are particularly a�ected. We assume that companies audited by

a �Big Four� auditor, companies that are cross-listed in the U.S. as well as larger

companies in general have a higher level of overall enforcement.

In our analyses, we �nd some evidence that companies showing a lower degree

of enforcement are a�ected more than their peers. We conclude that the reforms

under examination appear to level the playing �eld of the enforcement of �nan-

cial reporting, and have therefore enhanced the average earnings quality as well as

important capital market properties of the a�ected companies. Nevertheless, we rec-

ognize the di�culties of attributing changes in important company measures over

time to certain reforms. To address this concern, we use several speci�cations of our

multivariate analyses to rule out alternative interpretations of our �ndings.

The remainder of this paper proceeds as follows. In Section 5.2, we describe the

institutional background in Germany. Section 5.3 summarizes the earlier theoretical

as well as empirical work on this topic and develops the hypotheses. In Section 5.4,

we explain our research methodology and describe the data used for the analyses.

Section 5.5 discusses the results for our major analyses and for several sensitivity

analyses. Section 5.6 concludes.

91



Traditionally, the German enforcement system was based on the two-board sys-

tem, on statutory auditors, and on courts (especially tax courts). Based on a pro-

grammatic schedule published in 2003 by the German government, the German

enforcement mechanisms were reformed building on three steps. The �rst step was

the Bilanzkontrollgesetz (BilKoG�Accounting Enforcement Act) of December 15,

2004. This act created the legal basis for setting up an external �nancial reporting

enforcement in a two-tier structure. It comprises a private body called Deutsche

Prüfstelle für Rechnungslegung DPR e.V. (DPR�Financial Reporting Enforcement

Panel), which examines audited �nancial reports either in reaction to indications of

an infringement of �nancial reporting requirements or proactively based on random

sampling, and a public authority called Bundesanstalt für Finanzdienstleistungsauf-

sicht (BaFin�Federal Financial Supervisory Authority), which intervenes if a com-

pany is unwilling to cooperate voluntarily with the DPR. The DPR's investigations

are given authority through an adverse disclosure mechanism in the case of error

�ndings and the potential capital market reactions (�name and shame� mechanism).

The enforcement applies to all �nancial reports of companies that have securities

traded in a regulated segment of domestic stock exchange (DPR, 2009). The DPR

began its enforcement activities on July 1, 2005.

The second step taken by the German government to improve the enforcement was

the Abschlussprüferaufsichtsgesetz (APAG�Auditor Oversight Law) of December 27,

2004. The law established the Abschlussprüferaufsichtskommission (APAK�Auditor

Oversight Commission), being a private board that is responsible for disciplinary

oversight, for the adoption of international audit standards, and for the quality

assurance of auditors. The APAK has information and inspection rights for publicly

traded companies and co-operates with public oversight authorities within the EU

and other countries.

The third step taken by the German government to enhance con�dence in �nan-

cial information was the Bilanzrechtsreformgesetz (BilReG�Accounting Law Reform

Act) of December 4, 2004, which became e�ective for �scal years beginning after

January 1, 2005. This law modi�ed the previous regulations on auditor indepen-

dence. The BilReG prohibits auditors from carrying out a statutory audit when their

business, �nancial or personnel relationships cause a lack of objectivity. Even more

92

5.3 Prior research and hypotheses

restrictive rules apply to companies listed in regulated market segments. Moreover,

statutory auditors are subject to internal rotation every seven years.

Overall, these new enforcement mechanisms seek to establish incentives for com-

panies to comply with the accounting guidelines, thus, improving accounting quality.

They are also presumed to have a positive e�ect on capital market properties (i.e.,

stock liquidity and market valuation).


5.3.1 Enforcement reforms and the overall degree of

enforcement

In order to evaluate whether and how these reforms e�ective by 2005 have changed

the degree of �nancial reporting enforcement in Germany, we refer to our previously

stated de�nition of enforcement and argue that an enforcement regime can be seen

as stronger if it, ceteris paribus, increases the likelihood and/or the cost of getting

caught for erroneous �nancial reporting, in order to prevent future accounting er-

rors or frauds. The reforms are regarded as a fundamental change in the German

system of regulating �nancial reporting (Brown and Tarca, 2005). The new two-

tier enforcement system complements the enforcement by supervisory boards and

auditors. This increases the likelihood of companies (and auditors) being proven

to issue inaccurate �nancial reports. The sanction of the new mechanisms is the

publication of detected material errors which leads to adverse capital market reac-

tions. According to surveys, about 90% of the companies enforced by the DPR fear

negative consequences (especially, loss of reputation) resulting from such an error

announcement (Deutsches Aktieninstitut and PricewaterhouseCoopers AG, 2009).

Ernstberger et al. (2010) �nd that the capital market reactions are statistically

and economically signi�cant, however, much smaller than in the U.S. This illustrates

that error announcements are e�ective in penalizing infringing companies. Moreover,

the error �ndings increase the likelihood of litigation against fraudulent managers,

supervisory boards and auditors. The penalties were in place before the reforms, but

there was no external enforcement body to receive indications of accounting errors or

to conduct examinations on a random sampling basis. The auditor independence and

the auditor oversight reforms can both increase the power of auditors in negotiations

93


with management about certain accounting treatments (Nagy, 2005). Hence, the

enforcement reforms have the potential to reduce the risk of auditors colluding with

management and increase the prevention of �nancial reporting errors and frauds.

This leads us to synthesize that the German reforms have, ceteris paribus, in-

creased the overall degree of public enforcement for the �nancial reports of publicly

traded companies. The degree of enforcement for each company might still vary,

as the di�erences in other enforcement mechanisms such as the supervisory board

or the auditors still persist. However, the reforms set a certain minimum level of

enforcement for all companies listed in the regulated market.

5.3.2 Earnings quality e�ects of the enforcement reforms

Several prior studies show that levels of or changes in �nancial reporting enforce-

ment generally have the potential to in�uence accounting quality. In an analytical

study, Nagar and Petacchi (2005) show that the proportion of companies engaging

in earnings management varies between countries but that a temporary increase

in the enforcement budget can reduce this number. Goldman and Slezak (2006)

complement these �ndings by documenting that regulation � in terms of penalties

� can reduce earnings management if those penalties are su�ciently large. In turn,

the degree of earnings management can even rise with a new regulation, depending

on the incentives that might in�uence managers to exert more e�orts in window

dressing.

Empirically, several studies �nd that a higher degree of enforcement in a country

is associated with a lower degree of earnings management (Ball et al., 2000; Hung,

2000; Leuz et al., 2003; Lang et al., 2006). Hope (2003) uses a broad sample from 22

countries and provides evidence that stronger enforcement is associated with a higher

forecast accuracy of �nancial analysts. He also documents that enforcement is more

important when there is more choice in accounting methods. Any cross-country

study introduces problems of potentially confounding variables such as di�erent

accounting principles or other di�erences in institutions that are not controlled for

(Sloan, 2001; Holthausen, 2009).

Several cross-sectional empirical investigations have examined the impact of dif-

ferent properties of enforcement and other governance mechanisms on earnings man-

agement for di�erent companies in just one country and largely �nd that a greater

degree of enforcement and governance is associated with lower earnings manage-

94


ment (Dechow et al., 1996; Klein, 2002; Vafeas, 2005; Peasnell et al., 2005; Davidson

et al., 2005). Longitudinal studies about the e�ects of changes in the institutional

setting are bounded within a homogenous institutional and GAAP setting. As for

the cross-sectional studies, nearly all these studies focus on the United States, and

in particular on the impact of the Sarbanes-Oxley Act (SOX) on earnings manage-

ment (e.g., Bartov and Cohen, 2007; Ashbaugh-Skaife et al., 2008; Cohen et al.,

2008). Two exceptions are the studies of Brown et al. (2008) and of Gassen and

Ashbaugh-Skaife (2009). The latter document that the internal control reforms in

Germany mandated by the KonTraG have positively a�ected earnings management

and improved the monitoring role of audits.

The aforementioned enforcement reforms in Germany were expected to have var-

ious e�ects on earnings quality. First, the reforms were expected to curb man-

agement's opportunities and incentives to intentionally misstate or misrepresent

reported income. Given the potential sanctions of a material error �nding and its

publication, the expected net bene�t of (aggressive) earnings management falls as

the (possible) cost rises. Thus, the existence of the new enforcement laws can a�ect

companies' reporting choices and reduce earnings management.

Second, the enforcement reforms could increase the e�orts of other enforcement

institutions as the DPR states: �Accounting issues are being discussed much more

intensively by supervisory boards, and in particular by audit committees, as well as

by corporate governing bodies and auditors, and they are doing so frequently with

reference to the newly introduced enforcement procedure in Germany� (Deutsche

Prüfstelle für Rechnungslegung e.V., 2008). If the auditor does not carefully monitor

�nancial statements with high levels of earnings management, the client's risk of

being found to have material accounting errors increases and in the case of disclosure

of such an error, the auditor jeopardizes his/her reputation and risks litigation.

However, reputation is important in evaluating the future audit fees the auditor will

earn.

Third, the enforcement reforms can increase the bargaining power of auditors in

negotiations with the management about questionable accounting treatments (Nagy,

2005). Referring to the possible sanctions of error �ndings, they can convince the

management to use less aggressive earnings management. Based on these arguments,

the degree of earnings management should be lower after the German enforcement

reforms.

95


In contrast, other arguments would imply a less pronounced or even no impact of

the reforms. Many German companies are characterized by a high ownership con-

centration and major shareholders have access to information channels beyond the

�nancial statements. Accordingly, the demand for external monitoring might be lim-

ited for large shareholders (Ashbaugh and War�eld, 2003). Reforms that strengthen

external governance mechanisms might have a positive impact on earnings manage-

ment in countries with widely dispersed ownership such as the United States, but the

impact may be more muted in countries with a fundamentally di�erent governance

structure such as Germany (Co�ee, 2005).

To summarize, recent regulations have increased the degree of enforcement in

Germany but we note that the institutional infrastructure might in�uence the e�ects

of the reforms. If these reforms, however, have had their intended e�ects, then

we should observe a decrease in earnings management. This leads us to the �rst

hypothesis:

H1: Following recent enforcement reforms in Germany, the degree of

earnings management in �nancial statements of a�ected companies, ce-

teris paribus, is expected to decrease.

5.3.3 Capital market e�ects of the enforcement reforms

A study by (Daske et al., 2008) investigates the capital market e�ects of the manda-

tory IFRS adoption in 26 countries around the world. They study the impact on

market liquidity and on cost of capital, and provide evidence that the bene�ts of

IFRS adoption occur only in countries with a high degree of transparency and of

legal enforcement. Jackson and Roe (2009) document a signi�cantly positive asso-

ciation of strong enforcement with market capitalization, trading volume, and the

number of initial public o�erings.

The impact of enforcement and other governance reforms on the stock liquidity

of capital markets has been examined in several other studies. For example, the

Regulation Fair Disclosure has been shown to improve stock liquidity (Eleswarapu

et al., 2004). In addition, Jain et al. (2006) provide evidence that stock liquidity

measures for the U.S. capital market increased after the introduction of SOX.

Enforcement of �nancial reports also acts as a kind of �insurance for investors�. If

these reforms reduce the expected net bene�ts of earnings management and increase

96


the quality of the audit services provided to the company, they assure a faithful ap-

plication of standards and hence a higher quality of accounting outcomes. This

reduces information asymmetry between informed and uninformed market partic-

ipants (Diamond and Verrecchia, 1991). Lower information asymmetry is shown

to reduce stock return volatility and to enhance stock liquidity in terms of trading

volume and percentage of days traded (Welker, 1995; Leuz and Verrecchia, 2000;

Bushee and Leuz, 2005).

Furthermore, these reforms may have a direct impact on stock liquidity, because

they could be seen as measures to restore credibility in �nancial reports and therefore

increase stock liquidity. Yet, the speci�c institutional infrastructure in Germany

might in�uence the impact on stock liquidity. Based on the above discussion, our

second hypothesis is as follows:

H2: Following recent enforcement reforms in Germany, the stock liquid-

ity of a�ected companies, ceteris paribus, is expected to increase.

Several studies test the relationship between enforcement or governance and mar-

ket valuation. These studies document that better legal protection for investors is

related to higher valuation by the stock market (La Porta et al., 2002) and to higher

valuation of listed companies relative to their assets or changes in investments (Wur-

gler, 2000). By testing the predictions of their theoretical model, Durnev and Kim

(2005) provide evidence that companies with better governance enjoy higher valu-

ations and that this association is stronger in countries with weaker legal regimes.

Similarly, O'Hara (2003) argues that reducing the amount of hidden private infor-

mation can have a positive impact on asset prices by improving the price discovery

process and stock liquidity. Market valuation measures, such as Tobin's q, cap-

ture expected cash �ows, discount rates, and growth rates. In case the enforce-

ment reforms increase the transparency and credibility of �nancial reports and thus

stockholders' ability to monitor the management of a company, they have a positive

impact on the growth expectations of outsiders and therefore on the market valu-

ation of a company (Daske et al., 2008). Hence, the enforcement reforms should

increase the market valuation of the a�ected companies. In contrast, these reforms

might have only a limited impact on company valuation, because the level of overall

law enforcement in Germany was already relatively high prior to the reforms (Leuz

et al., 2003; Kaufmann et al., 2007). Additionally, a company's own valuation also

97


measures its cost of complying with the new enforcement mechanisms. Direct costs

are the (small) charges levied by the DPR. Further direct costs relate to increased

e�ort in preparing �nancial reports. Thus, the cost of compliance could mitigate

the expected increase in company valuation. Inferring a positive (net) impact of the

German reforms, we state the following third hypothesis:

H3: Following recent enforcement reforms in Germany, the internal val-

uation of a�ected companies, ceteris paribus, is expected to increase.

5.3.4 Impact of enforcement through other mechanisms

The purpose of the enforcement reforms is to ensure compliance with accounting

standards and to improve investors' con�dence in �nancial reporting. The exter-

nal enforcement bodies complement the other enforcement mechanisms and assure

a minimum level of overall enforcement. Investors of companies which are char-

acterized by a low level of enforcement through other mechanisms (e.g., non-Big-

Four auditors) should have a higher demand for an additional external enforcement

mechanism. Thus, we expect higher net bene�ts from the enforcement reforms for

investors of companies with an originally low level of enforcement and consequently

a more pronounced reaction to the enforcement reforms for those companies. Hence,

we state the fourth hypothesis:

H4: Following recent enforcement reforms in Germany, the positive im-

pacts are expected to be more pronounced for companies with, ceteris

paribus, a lower degree of enforcement through other mechanisms.

5.4 Research design, sample selection, and data

5.4.1 Regression approaches

The enforcement reforms in Germany provide a unique setting to evaluate the e�ects

of an increase in the degree of enforcement mandated by law, while controlling for

other important aspects, especially for the voluntary and mandatory adoption of

IFRS in Europe. We are able to isolate the impact of the reforms to the enforcement

from accounting principles adoption e�ects by exploiting the di�erent regulatory

requirements for the segments of the German stock market.

98


Regression for the total e�ect of the enforcement reforms

To test H1 to H3, we separately regress di�erent proxies for earnings management,

stock liquidity, or market valuation (MEASURE it) on a binary variable which in-

dicates if a company i in period t is a�ected by the enforcement reforms (ENF it).

Hence, we estimate the following regression model basically for all German listed

companies for the periods 2003 to 2006:

MEASURE it = α0 + α1 POST it + α2 ENF it + α3 SIZE it + α4 BIGFOURit

+α5 CROSSLit + α6 IFRS it + α7 IFRSFIRST it

+α8 IFRSMAND it + α9 USGAAP it + α10 CLOSEH it + α11 LEV it

+α12 MKTVOLAit + α13 ROAit + α14 VOLAit + uit (5.1)

where all variables are de�ned in Table 5.1 and the di�erent measurement concepts

are explained in Section 5.4.2. We estimate this equation for a balanced panel

using �rm-�xed e�ects and White period clustered standard errors. This regression

model is based on a di�erence-in-di�erences approach as we compare the changes in

the measurement variables for a treatment group (all companies in the Prime and

General Standard which are a�ected by the enforcement reforms) with the changes

in the measurement variables for a benchmark group (all companies in the Open

Market which are not a�ected by the enforcement reforms). POST it indicates if

an observation is related to a period after the enforcement reforms and captures

potentially biasing trends over time. In combination with the �rm-�xed e�ects

our variable of interest ENF it is conceptually comparable to an interaction term

of a binary variable indicating if a company belongs to the treatment group1 and

POST it. The coe�cient on ENF it shows whether the measurement variables have

signi�cantly changed for the treatment group as predicted by H1 to H3. This model

speci�cation also deals carefully with e�ects arising from di�erent accounting regimes

and from changes of the accounting principles towards IFRS. First, we include the

binary variables IFRS it and USGAAP it indicating the application of IFRS and U.S.

GAAP relative to German domestic GAAP in the respective period. Second, we

control for e�ects of the �rst-time application of IFRS (IFRSFIRST it). Besides the

1Such a binary variable cannot be implemented as we use �rm-�xed e�ects.

99


transition e�ects on the �nancial reporting �gures, this binary variable captures,

e.g., investors' enthusiasm concerning the adoption of IFRS. Third, we control for

the mandatory IFRS adoption since 2005 by companies which were not mandated to

apply IFRS before 2005, i.e., companies listed in the General Standard stock market

segment (IFRSMAND it).2

Regression for the impact of enforcement through other mechanisms

As stated in H4, we expect cross-sectional variations in the magnitude of the impact

of the German enforcement reforms on earnings management, on stock liquidity, as

well as on market valuation. To test H4, we expand the approach given in equation

(1) by an interaction term of ENF it and metrics for the enforcement quality through

other mechanisms. The DPR itself primarily attributes the detected erroneous �-

nancial reports to the �enormous scope and highly complex nature of IFRS�, which

�overwhelms� especially small companies' accountants and their auditors (Deutsche

Prüfstelle für Rechnungslegung e.V., 2009). Based on these assumptions we overall

expect a lower level of enforcement for smaller companies. As a second determinant

we use audit quality which we proxy by a binary variable indicating that �nancial

reports are audited by a �Big Four� audit company. This link is broadly documented

for the United Kingdom as well as for the United States (e.g., DeFond and Jiambalvo,

1991, 1993; Becker et al., 1998; Gore et al., 2001), whereas Maijoor and Vanstraelen

(2006) were not able to document such an e�ect for continental Europe. Further, we

assume that companies which are cross-listed in the United States are subject to a

higher degree of enforcement besides the German reforms under investigation (e.g.,

La Porta et al., 1997; Leuz, 2003a), since their �nancial reports are also enforced

by the Securities and Exchange Commission (SEC). Several studies document the

e�ectiveness of the SEC enforcement (e.g., Palmrose et al., 2004). Hence, we expect

a less pronounced e�ect of the reforms for companies cross-listed in the U.S. We

estimate the following equation:

2We note that our research design (inclusion of speci�c control variables as well as the estimationmethodology including �rm-�xed e�ects) is not adequate to detect general di�erences in ourmeasurement variables between accounting principles.

100


MEASURE it = β0 + β1 POST it + β2 ENF it + β3 SIZE it + β4 BIGFOURit

+β5 CROSSLit + β6 interaction it + β7 IFRS it + β8 IFRSFIRST it

+β9 IFRSMAND it + β10 USGAAP it + β11 CLOSEH it + β12 LEV it

+β13 MKTVOLAit + β14 ROAit + β15 VOLAit + vit (5.2)

where the independent variables are de�ned in Table 5.1 and the di�erent measures

are explained in Section 5.4.2. We estimate this equation for a balanced panel

using �rm �xed-e�ects and White period clustered standard errors. The variable

interaction it indicates if the impact of the enforcement reforms is in�uenced by

the level of enforcement through other mechanisms. Based on the aforementioned

argumentation, we expect the positive impact of the enforcement reforms to be

lower for larger companies, companies audited by �Big Four� auditors, a cross-listing

in the U.S. and with a higher market capitalization. Finally, we aggregate our

three proxies for the degree of enforcement through other institutions/mechanisms

(SIZE it, BIGFOURit, and CROSSLit) by a principal components analysis3 and use

the resulting factor score ENFOTHERit as an additional interacting variable.4

5.4.2 Measurement variables

Prior literature uses a variety of proxies to measure earnings quality (see, e.g., De-

chow et al., 2010b, for a recent overview), e.g., the assessment of the distribution

of earnings (e.g., Burgstahler and Dichev, 1998) and the investigation of changes in

accounting policies (e.g., Sweeney, 1994). In this study we focus on accrual quality

as a measure for earnings quality assuming that lower accrual quality indicates a

higher level of earnings management and thus a lower level of earnings quality. In

this context Jones et al. (2008) and Dechow et al. (2010a) provide consistent evi-

dence that accrual quality measures based on Dechow and Dichev (2002) perform

3We are aware that the aggregation of continuous and binary variables (as also used in otherstudies, e.g., by Armstrong et al., 2010) is potentially problematic as a principal componentsanalysis is conceptually designed for continuous variables only. We keep this aspect in mind forthe interpretation of our �ndings.

4To evaluate the impact of the factor score ENFOTHERit we adequately modify Equation 5.2 as wewithdraw the control variables SIZE it, BIGFOURit, and CROSSLit, and include ENFOTHERit

as well as the interaction ENF it · ENFOTHERit.

101


better in detecting earnings management in terms of the existence of fraud, of the

magnitude of fraud, and of non-fraud restatements than, e.g., conventional linear

discretionary accruals models introduced by Jones (1991). We rely on concepts an-

alyzing working capital accruals as it seem to be likely that working capital accruals

are subject of manipulations than other accruals (e.g., Becker et al., 1998; Teoh

et al., 1998; Young, 1999). The �rst measure is basically derived from the accrual

quality model developed by Dechow and Dichev (2002). In this model changes in

working capital are related to past, present, and future operating cash �ows. We

rely on the unexplained change in working capital as a measure for the accruals

quality.

Our second earnings management measure is the abnormal working capital ac-

crual metric proposed by DeFond and Park (2001). The underlying assumption of

this metric is that (normal) working capital of a company is a �xed proportion of

revenues of the respective company. We use the non-signed version of both earnings

management measures as we are interested in both, over- and understatements of ac-

cruals. Finally, we aggregate the non-signed versions of both measures by a principal

components analysis to derive a factor score representing the essential information

of the underlying measures concerning the earnings quality of the company. For

technical details about the earnings quality measures see Table 5.1.

The �rst stock liquidity measure is a bid-ask spread metric which is a commonly

used proxy for information asymmetries (e.g., Welker, 1995; Leuz and Verrecchia,

2000; Daske et al., 2008). It re�ects di�erences in buy-side and sell-side expectations

concerning companies' stock prices. Second, we use the proportion of trading days

without a stock price change to proxy for the stock liquidity. Prior research (e.g.,

Lesmond, 2005; Ashbough-Skaife et al., 2006; Bekaert et al., 2007) con�rms that

a zero-return ratio can capture the level of information inherent in stock prices

(see, e.g., Lesmond et al., 1999) for an argumentation concerning the occurrence

of zero-return trading days). Finally, we aggregate these measures by a principal

components analysis to derive a factor score representing the essential information

concerning companies' stock liquidity. For technical details about the stock liquidity

measures see Table 5.1.

As a last group of measures we use market valuation metrics, in particular, the

market-to-book ratio as well as Tobin's q (e.g., Servaes, 1991; Lang and Stulz, 1994;

Daske et al., 2008). Both metrics indicate di�erent expected future discount rates,

102


di�erent expected future cash �ows, and/or di�erent cash �ow growth expectations.

As we assume an improvement of the credibility of �nancial reports (e.g., a reduction

of uncertainties concerning the future performance of the company) arising from the

reforms to the enforcement and very limited cash �ow e�ects (e.g., Ernstberger

et al., 2010) we expect a net-increase in both measures. Finally, we aggregate these

measures by a principal components analysis to capture the essential information

concerning companies' market valuation. For technical details about the market

valuation measures see Table 5.1.

Table 5.1: De�nition of variables

Abbr. De�nition

Panel A: Measurement variables

DAC it Natural logarithm of the non-signed discretionary accruals scaled bylagged total assets of company i in period t calculated by a modi�edversion of the model proposed by Dechow and Dichev (2002). We com-pute scaled discretionary accruals as the residuals of the following model:

∆WC it = γ0 + γ1 CFO i,t−1 + γ2 CFO it + γ3 CFO i,t+1 + wit

where ∆WC it is the change in working capital scaled by lagged totalassets of company i in period t. CFO it stands for the cash �ow fromoperating activities scaled by lagged total assets of company i in period t(CFO i,t−1 and CFO i,t+1 respectively). We estimate this model over theinvestigation period (2003 to 2006) for each industry-accounting princi-ples combination (at least �ve observations are required per combination)separately. Higher values of DAC it indicate a higher level of earningsmanagement coming along with lower earnings quality.

AWCAit Natural logarithm of the non-signed abnormal working capital accrualsscaled by lagged total assets of company i in period t calculated by amodi�ed version of the model developed by DeFond and Park (2001).We compute abnormal working capital accruals of company i in periodt using the following equation:

AWCAit = WC it −WC i,t−1 · (SALES it/SALES i,t−1)

where WC it is the working capital of company i in period t (WC i,t−1

respectively). SALES it stands for revenues of company i in period t(SALES i,t−1 respectively). Higher values of AWCAit indicate a higherlevel of earnings management coming along with lower earnings quality.

Continued on next page

103


De�nition of variables � Continued

Abbr. De�nition

EM it Factor score for earnings management calculated by a principal compo-nents analysis of the non-signed values of DAC it and AWCAit. DAC it

and AWCAit are correlated by 0.055. EM it is correlated with DAC it

(AWCAit) by 0.792 (0.196) and represents 0.519 of the total informationof the underlying variables. Higher values of EM it indicate a higher levelof earnings management coming along with lower earnings quality.

BAS it Natural logarithm of the relative bid-ask spread, de�ned as the period-median of the daily di�erence between the closing ask-price and theclosing bid-price scaled by the daily average of these components forcompany i in the period t. Lower values of BAS it indicate a higher levelof stock liquidity.

ZRET it Natural logarithm of the proportion of zero-return days, de�ned as theproportion of trading days with no stock price change of company i inthe period t (relative to the company-speci�c total number of availableobservations). Lower values of ZRET it indicate a higher level of stockliquidity.

ILLIQU it Factor score for stock liquidity calculated by a principal componentsanalysis of BAS it and ZRET it. BAS it and ZRET it are correlated by0.436. ILLIQU it is correlated with BAS it (ZRET it) by 0.813 (0.875)and represents 0.762 of the total information of the underlying variables.Lower values of ILLIQU it indicate a higher level of stock liquidity.

MBRit Natural logarithm of the market-to-book ratio, de�ned as the ratio of theperiod-average market capitalization of company i in period t calculatedbased on closing-prices divided by the book-value of common equity atthe end of period t. Higher values of MBRit indicate a higher marketvaluation relative to its items reported in �nancial statements.

TOB it Natural logarithm of Tobin's q, de�ned as the book value of total assets,plus the market value of common equity at the end of period t, minus thebook value of common equity, scaled by the book value of total assetsof company i at the end of period t (e.g., Daske et al., 2008). Highervalues of TOB it indicate a higher market valuation of assets relative tothe assets reported in �nancial statements.

VALit Factor score for market valuation calculated by a principal componentsanalysis of the values of MBRit and TOB it. MBRit and TOB it arecorrelated by 0.740. VALit is correlated with MBRit (TOB it) by 0.925(0.928) and represents 0.867 of the total information of the underlyingvariables. Higher values of VALit indicate a higher market valuationrelative to its items reported in �nancial statements.


104



Abbr. De�nition

Panel B: Control variables

BIGFOURit Binary variable taking a value of 1 if company i is audited by a �Big Four�auditor (Deloitte, Ernst&Young, KPMG, or PriceWaterhouseCoopers)in period t, 0 otherwise.

CLOSEH it Ownership structure, de�ned as the average percentage of closely heldshares of company i over period t.

CROSSLit Binary variable taking a value of 1 if the company i in period t is listedin the United States (U.S.) and thus is subject to the SEC enforcementmechanisms, 0 otherwise.

ENF it Binary variable taking a value of 1 if the company i is subject of theDPR/BaFin enforcement mechanism in the respective period t, 0 other-wise.

IFRS it Binary variable taking a value of 1 if the �nancial statement of companyi in period t is prepared in accordance with IFRS, 0 otherwise.

IFRSFIRST it Binary variable taking a value of 1 if company i in period t for the �rsttime provides �nancial statements in accordance with IFRS, 0 otherwise.

IFRSMAND it Binary variable taking a value of 1 if the company i in period t is man-dated to apply IFRS since 2005 but was not permitted (e.g., by stockmarket requirement) to apply IFRS until 2004, 0 otherwise.

LEV it Financial leverage ratio, de�ned as the average book value of debt overthe period t divided by the average book value total assets over theperiod t of company i.

MKTVOLAit Natural logarithm of the volatility of the CDAX, de�ned as the standarddeviation of the daily CDAX returns over the period t.

SIZE it Size of the company, de�ned as the period-average market capitalizationof company i in period t calculated based on closing-prices.

USGAAP it Binary variable taking a value of 1 if the �nancial statement of companyi in period t is prepared in accordance with U.S. GAAP, 0 otherwise.

VOLAit Natural logarithm of the stock return volatility, de�ned as the standarddeviation of the daily stock returns (including dividends payments) ofcompany i over the period t.

Notes: This table shows de�nitions of measurement variables (Panel A) as well as of controlvariables (Panel B) which are used consistently within this study. Several continuous variables arewinsorized at the 2.5%- and 97.5%-percentiles. Factor scores are calculated based on normalizedbut not on winsorized data and are winsorized at the 2.5%- and 97.5%-percentiles afterwards.

105


5.4.3 Sample selection and data sources

Sample selection

Our basic sample contains observations of all German companies listed in a segment

of the German stock market (Prime Standard, General Standard, and Open Market)

for the periods 2003 to 2006.5 We require full data availability for each company

of the investigation period (balanced sample). This approach prevents biasing ef-

fects from changes in the sample distribution and mitigates potential endogeneity

concerns. Following previous studies, we eliminate companies from �nancial indus-

tries6 (Industry 11) from our sample in order to avoid distorting e�ects. We further

eliminate companies from the energy industry (Industry 4) from our sample, as we

require at least 30 observations per industry within our balanced sample. Next, to

simplify the research design and the interpretation of our �ndings we eliminate 8

observations which arise from companies which have changed the market segment

(from the General to the Prime Standard segment) over the investigation period.

Finally, our full sample of 1,312 observations contains 57.3% companies listed in

the Prime Standard, 37.5% companies listed in the General Standard, and 5.2%

companies listed in the Open Market segment. Within the subsample of companies

of the General Standard there is a ratio of �voluntary IFRS adopters� of 31.8%.

Within the subsample of Open Market companies the ratio of IFRS-applying com-

panies increases from 35.3% in period 2003 to 88.2% in period 2006. Within the

�nal sample of 1,312 observations 26.2% are from the business equipment and elec-

tronics industry (Industry 6), 20.1% are from the manufacturing industry (Industry

3), and 18.0% are classi�ed as �others� (Industry 12). Several other industries show

a proportion lower than 10.0%.

Data sources

We collect �nancial statement and stock market data from Datastream World-

scope. As the information for accounting principles are fragmentary within this

database and as previous studies found several unclear or even erroneous items

within databases in general (e.g., Daske et al., 2008), we double checked this in-

5As starting point for the sample selection we use the yearly lists of the Hoppenstedt Aktienführer.We also require su�cient data for 2002 where lagged data are necessary.

6We use the 12-industry classi�cation proposed by French (2009).

106

5.5 Empirical �ndings and sensitivity analyses

formation using the Hoppenstedt Aktienführer. In cases of contradicting entries we

take the information directly from the �nancial statements. Information about the

auditor as well as about the stock market segment is taken from the Hoppenstedt

Aktienführer. However, we also double checked the entries of this database con-

cerning the stock market segment using lists published by the BaFin (�companies

subject to the enforcement�). Information for the cross-listing variable is taken from

the New York Stock Exchange. For the stock market related variables we use data

from the main German stock exchange if the shares are traded on more than one

stock exchange. We de�ne the main stock exchange of a company as the one with

the highest trading volume (measured by market value) of the primary type of stock

(i.e., the one with the highest market capitalization over the investigation period).

Variables which are calculated based on daily data are only taken into account if

at least 25% of the observations per period are available. We note that these data

requirements as well as the balanced sample approach biases our sample towards

larger, mature, and surviving companies. To mitigate the in�uence of outliers and

potentially remaining erroneous items we winsorize continuous variables at the 2.5%-

and at the 97.5%-percentiles.


5.5.1 Empirical �ndings

Descriptive statistics and correlations

Table 5.2, Panel A presents descriptive statistics (non-logarithmized de�nitions) for

the measurement variables. Our non-signed accruals measures DAC it and AWCAit

show an average amount of 11.4% and 13.2%. Further, we �nd mean relative bid-ask

spreads (BAS it) of 3.6% and a proportion of 23.8% of trading days with no change

in stock prices (ZRET it). The average market-to-book ratio (MBRit) is 2.31 and

the average amount of Tobin's q (TOB it) is 1.48. Table 5.2, Panel B presents de-

scriptive statistics for the control variables. 72.0% (15.9%) of the observations relate

to IFRS it (USGAAP it) �nancial statements, the remaining observations relate to

German GAAP �nancial statements. 56.3% of the observations come from �nancial

statements audited by �Big Four� (BIGFOURit) auditors and 3.8% are cross-listed

in the U.S. The mean of CLOSEH it (43.9%) indicates a considerably high ownership

107


Table 5.2: Descriptive statistics

Mean [Prop.] Std. Dev. Lower quartile Median Upper quartile Obs.

Panel A: Descriptive statistics for measurement variables†DAC 0.114 0.109 0.043 0.092 0.153 1,248

AWCA 0.132 0.124 0.046 0.089 0.168 1,276

BAS 0.036 0.031 0.017 0.030 0.062 1,284

ZRET 0.238 0.247 0.093 0.216 0.478 1,312

MBR 2.308 2.093 0.969 1.648 2.294 1,312

TOB 1.483 0.736 1.002 1.459 2.089 1,312

Panel B: Descriptive statistics for control variables

SIZE† 685,951 1,448,201 20,943 64,636 209,480 1,312

BIGFOUR [0.563] 1,312

CROSSL [0.038] 1,312

IFRS [0.720] 1,312

USGAAP [0.159] 1,312

CLOSEH 0.439 0.272 0.340 0.490 0.880 1,312

LEV 0.211 0.184 0.065 0.183 0.346 1,312

MKTVOLA† 0.009 0.001 0.008 0.009 0.010 1,312

ROA 0.019 0.109 -0.029 0.038 0.089 1,312

VOLA† 0.028 0.014 0.018 0.024 0.045 1,312

Notes: This table shows descriptive statistics for measurement variables (Panel A) as well ascontrol variables (Panel B) for the largest balanced sample used within the multivariate regressionanalyses including 1,312 observations. Variables as de�ned in Table 5.1. † indicates non-logarithmde�nition.

concentration of the sample companies.

Table 5.3 shows correlations among all independent variables. The correlations

raise no serious concerns in terms of multi-collinearity. We keep in mind for the

interpretation of our coe�cients that SIZE it is moderately correlated with our other

proxies of the level of enforcement through other mechanisms (BIGFOURit and

CROSSLit).

Regression results for the total e�ect of the enforcement reforms

Table 5.4 presents regression results from estimating Equation 5.1 for the entire Ger-

man market including Prime Standard and General Standard companies which are

a�ected by the enforcement reforms (treatment group) and Open Market companies

108


Table 5.3: Correlations for control variables

(1) (2) (3) (4) (5) (6) (7) (8) (9)

ENF (1) -0.014 0.131 0.005 0.048 0.004 0.098 -0.061 -0.053

BIGFOUR (2) -0.003 0.134 0.287 -0.075 -0.039 0.133 -0.004 -0.087

CROSSL (3) 0.095 0.175 0.210 -0.143 0.046 0.038 0.006 -0.129

SIZE (4) 0.106 0.303 0.279 0.046 -0.074 0.353 0.015 -0.384

CLOSEH (5) 0.021 -0.060 -0.183 0.020 0.024 0.092 0.115 -0.058

LEV (6) -0.009 -0.024 -0.012 -0.061 0.010 -0.047 0.038 0.031

ROA (7) 0.112 0.068 0.072 0.361 0.080 -0.034 -0.077 -0.310

VOLA (8) -0.055 0.012 -0.015 -0.027 0.150 0.033 -0.096 0.118

MKTVOLA (9) -0.045 -0.156 -0.156 -0.474 -0.043 -0.001 -0.332 0.125

Notes: This table shows correlations for control variables for the largest balanced sample usedwithin the multivariate regression analyses including 1,312 observations. The table shows SpearmanCorrelations above the diagonal and Pearson Correlations below. Variables are de�ned in Table5.1.

which are not a�ected by the enforcement reforms (benchmark group). Column 1

through 3 of Table 5.4 show regression results for our earnings management mea-

sures (DAC it, AWCAit, and EM it). In line with H1 we expect a negative impact of

ENF it on our earnings management measures interpreted as a decrease in earnings

management associated with the enforcement reforms. For the discretionary ac-

crual model (DAC it) as well as for the earnings management factor score (EM it) we

document a signi�cantly negative impact. The impact of ENF it on the abnormal

working capital accrual measure (AWCAit) is also negative but not signi�cant at

conventional levels. We take these regression results as evidence in support of our

hypothesis H1.

Column 4 through 6 present regression results for our stock liquidity measures

(BAS it, ZRET it, and ILLIQU it). As stated in H2, we expect a negative impact

of ENF it on our stock liquidity measures interpreted as an increase in the stock

liquidity. Consistent with this prediction we �nd a signi�cant decrease in each of

our stock liquidity measure (at di�erent levels of signi�cance).

Column 7 through 9 show results for our market valuation measures (MBRit,

TOB it, and VALit). In line with H3 we expect a positive impact of ENF it on our

market valuation measures. For MBRit and TOB it we �nd a positive impact by

trend, however, on insigni�cant levels. The market valuation factor score (VALit)

provides little evidence for an increase in valuation by market participants.

109


The results for the variables controlling for accounting principle e�ects indicate

that the stock liquidity is higher for companies applying IFRS, especially for those

applying IFRS for the �rst time. The results for the earnings management measures

di�er, yet the coe�cients on the respective variables for the factor score regression

indicates a lower degree of earnings management for IFRS and U.S. GAAP observa-

tions. Interestingly, we �nd a lower market valuation for companies applying IFRS

and U.S. GAAP. We note, however, that the both measures (MBRit and TOB it)

could mechanically decrease due to a higher equity under IFRS or U.S. GAAP in

comparison to the equity under German domestic GAAP.

Overall, we �nd a signi�cant decrease in earnings management and increase in

stock liquidity for the companies that are a�ected by the enforcement reforms.

Moreover, we �nd limited evidence for an increase in market valuation of these

companies.

Regression results for the impact of enforcement through other mechanisms

To test hypothesis H4, we include an additional term in each of our regressions

which compares ENF it with variables being expected to proxy for a higher level

of enforcement through other mechanisms. In particular, we interact ENF it with

a binary variable indicating that a company's �nancial report is audited by a �Big

Four� auditor (BIGFOURit), with a binary variable indicating a U.S. cross-listing

(CROSSLit), with the size of the company (SIZE it), and with a factor score repre-

senting an overall higher level of �other enforcement� (ENFOTHERit). The inter-

action terms indicate if the impact of the enforcement reforms is in�uenced by the

level of �other enforcement�. As we hypothesize that the impact of the enforcement

reforms is higher for companies characterized by low level of enforcement through

other mechanisms we expect a coe�cient on the interaction terms having the oppo-

site sign in comparison to the sign of the general e�ect of the enforcement reforms

(coe�cient on ENF it).

Table 5.5, Panel A shows the results for SIZE it as an interacting variable. In line

with our predictions we document a signi�cantly positive coe�cient on the interac-

tion term ENF it · SIZE it for ILLIQU it. This �nding supports our hypothesis that

especially small companies bene�t from the enforcement reforms as they are charac-

terized by weaker other enforcement mechanisms. However, we show a signi�cantly

negative coe�cient on the interaction term ENF it · SIZE it for EM it. This �nding

110

5.5 Empirical �ndings and sensitivity analysesTable

5.4:Regressionresultsfortheearnings

managem

ent,stockliquidity,andmarketvaluationmeasures-

non-matched

sample

Earnings

Managem

ent

Stock

Liquidity

MarketValuation

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

DAC

AWCA

EM

BAS

ZRET

ILLIQ

UMBR

TOB

VAL

POST

0.073

0.235

-0.091

-0.349

***-0.165

*-0.254

***

0.005

0.028

-0.030

ENF

-0.277

**-0.109

-0.204

**-0.428

***-0.146

*-0.174

**0.070

0.021

0.174*

SIZE

0.040

0.137

0.053

-0.121

**-0.329

***-0.240

***

0.513***

0.246***

0.613***

BIG

FOUR

0.059

-0.152

0.040

0.104

0.093

0.100**

0.039

-0.019

0.062

CROSSL

-0.083

0.542

-0.144

0.675***

0.181*

0.333***-0.194

*-0.094

-0.301

IFRS

-0.064

0.152

-0.067

*-0.074

0.002

-0.100

*-0.211

***-0.109

***-0.196

**IFRSFIRST

0.049

0.307**

-0.122

*-0.035

-0.156

***-0.053

0.009

0.009

-0.041

IFRSMAND

-0.162

***-0.282

-0.021

-0.095

0.142**

-0.210

0.082

-0.021

-0.032

USGAAP

-0.204

0.484***-0.240

**0.057

-0.223

***-0.068

-0.237

***-0.111

**-0.281

CLOSEH

-1.730

***

0.193

-1.713

***-0.031

-0.181

***-0.123

*0.169***

0.035

0.154*

LEV

0.236

0.753

-0.383

-0.090

0.036

0.034

0.341

0.177

0.148

MKTVOLA

0.074

-0.156

0.150

0.944***

0.472***

0.675***-0.223

***-0.094

***-0.210

**ROA

0.287

0.524

1.766***

0.487**

-0.124

0.030

-0.501

**-0.188

-0.419

VOLA

-0.031

0.090

0.020

0.375***-0.113

0.118

0.034

0.024

0.047*

Adj.R-squared

0.758

0.240

0.522

0.703

0.855

0.810

0.923

0.820

0.779

Observations

1,248

1,276

1,248

1,284

1,312

1,284

1,312

1,312

1,312

Firm-�xed

e�ects

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Notes:

Thistableshow

smultivariate

regressionresultsforH1to

H3forourentire

sample.Variablesare

de�ned

inTable5.1.Weestimate

theregressionsusingpanel

structure

andWhiteperiodclustered

standard

errors.*,**,and***indicate

one-tailed

signi�cance

atthe

0.1,0.05,and0.01levelforourvariableofinterest

(EN

Fit)in

accordance

withtheexpectedimpact

asstatedin

H1to

H3andtwo-tailed

signi�cance

atthe0.1,0.05,and0.01levelfortheother

variables.

111


contradicts our hypothesis H4 as it suggests that the bene�ts of the enforcement

reforms in terms of earnings quality are more pronounced for larger companies than

for smaller ones. One potential explanation is that SIZE it is a very crude proxy

for the degree of enforcement. A second reason could be that larger companies are

more visible in the public and thus fear a possible error announcement by the new

enforcement bodies because it could damage their reputation.

Table 5.5, Panel B shows the results for BIGFOURit as an interacting variable.

Consistent with H4 several coe�cients suggest that companies audited by a �Big

Four� auditor are less a�ected by the enforcement reforms by trend but the impacts

are not signi�cant at conventional levels.

Table 5.5, Panel C shows the results for CROSSLit as an interacting variable.

Even through several coe�cients have the sign as predicted by H4 only the impact

on ILLIQU it is signi�cant and suggests that companies which are not cross-listed

in the U.S. bene�t more from the reforms than those listed in the U.S.

Table 5.5, Panel D shows the results for ENFOTHERit as an interacting variable.

Once more, we �nd that several coe�cients support H4 by trend. However, only

for ILLIQU it we are able to document a signi�cant impact of ENFOTHERit as an

interacting variable which provides evidence that the e�ect of enforcement reforms

are particularly pronounced for companies which are characterized by weak other

enforcement mechanisms.

Overall, the �ndings summarized in Table 5.5 provide some evidence that com-

panies with a low level of enforcement through other institutions are particularly

a�ected by the enforcement reforms. Thus, the reforms leveled the playing �eld

resulting in a more homogenous overall level of enforcement for German companies.

5.5.2 Sensitivity analyses

To insure the robustness of our results, we conduct several sensitivity analyses.

We provide further evidence to mitigate the concern that the enforcement reforms

e�ects could be biased by the voluntary or mandatory application of IFRS. We also

take error �ndings through the DPR/BaFin mechanism into account and examine

whether infringing companies are particularly a�ected by the enforcement reforms.

Further, we address some aspects of our research design as well as some technical

issues of our empirical analyses.

112


Table 5.5: Impact of the level of enforcement through other mechanisms

EM ILLIQU VAL

Panel A: Size of the company as a proxy for enforcement through other mechanismsPOST -0.027 * -0.253 *** -0.029SIZE 0.066 -0.261 *** 0.609 ***ENF 0.134 -0.536 *** 0.088ENF·SIZE -0.027 † 0.039 *** 0.007

Adjusted R-squared 0.524 0.815 0.779Included observations 1,248 1,284 1,312Firm-�xed e�ects Yes Yes YesOther control variables Yes Yes Yes

Panel B: �Big Four� auditor as a proxy for enforcement through other mechanismsPOST -0.095 -0.257 *** -0.029BIGFOUR -0.008 0.072 0.085ENF -0.268 ** -0.105 0.205ENF·BIGFOUR 0.107 0.063 -0.051


Panel C: U.S. Cross-listing as a proxy for enforcement through other mechanismsPOST -0.091 -0.260 *** -0.029CROSSL -0.148 0.110 * -0.264ENF -0.204 * -0.079 0.176ENF·CROSSL 0.005 0.246 *** -0.040


Panel D: Factor score as a proxy for enforcement through other mechanismsPOST -0.017 * -0.093 *** -0.014ENFOTHER 0.004 -0.013 ** 0.025 *ENF -0.036 -0.052 ** 0.062ENF·ENFOTHER 0.019 0.038 *** -0.017


Notes: This table shows excerpts of multivariate regression results including interactions to eval-uate H4 for our entire sample. For each model we include several control variables as stated inequation (2). Variables are de�ned in Table 5.1. We estimate the regressions using panel structureand White period clustered standard errors. *, **, and *** indicate one-tailed signi�cance at the0.1, 0.05, and 0.01 level for our variable of interest (ENF it and interactions) in accordance withthe expected impact as stated in H4 and two-tailed signi�cance at the 0.1, 0.05, and 0.01 level forseveral other variables. † indicates one-tailed signi�cance at the 0.05 level.

113


Confounding IFRS adoption e�ects

To control for possible confounding IFRS adoption e�ects, we �rst perform a matched

sample approach. We match the Open Market companies of our benchmark group

in the primary analyses with similar companies in the Prime Standard and General

Standard (treatment group). We require that the companies from the treatment

group exhibit the same �accounting principles pro�le� over the investigation period

(2003 to 2006) as the respective companies of the benchmark group. Afterwards, we

use the size of the companies (average market capitalization over the investigation

period) as a second matching criterion. We end up with a maximum sample of 104

observations (19.2% of the observations are from the Prime Standard, 30.8% are from

the General Standard, and 50.0% of the observations from companies listed in the

Open Market). Hence, the companies in the benchmark and in the treatment group

show the same �accounting principles pro�le� and similar size. A mean equality test

for size indicates no signi�cant di�erence in size (t = 0.307). Table 5.6 presents

regression results for the matched sample. As reported in columns 1 through 3, we

�nd a signi�cant decrease of DACit related to the enforcement reforms. However,

we are not able to document a signi�cant impact of the reforms on AWCAit and on

the factor score EM it. Thus, these �ndings provide only limited evidence that the

degree of earnings management decreased by the enforcement reforms. Columns 4

through 6 provide strong evidence for a decrease of BAS it as well as of ILLIQU it

but we do not �nd a signi�cant impact of ENF it on ZRET it. Columns 7 through 9

homogenously show an increase in market valuation which is signi�cant for MBRit

and VALit. Overall the �ndings of these three sensitivity analyses to control for

IFRS adoption e�ects corroborate the �ndings for our full sample.

Second, we re-perform our analyses based on a sample of Prime Standard com-

panies applying IFRS only. This research design di�ers from our original approach

as the sample does no longer contain a benchmark group. Thus, we assume that

all changes in our measurement variables � after controlling for various company-

speci�c and macro-economical e�ects � are related to the enforcement reforms under

investigation. For our earnings quality variables, we �nd a slightly signi�cant de-

crease in each measure on a slightly signi�cant level. Our stock liquidity measures

indicate qualitatively the same impact of the enforcement reforms on the stock liq-

uidity as in our original analyses. Further, we �nd a consistent but insigni�cant

positive impact of the enforcement reforms on the market valuation of companies.

114

5.5 Empirical �ndings and sensitivity analysesTable

5.6:Regressionresultsfortheearnings

managem

ent,stockliquidity,andmarketvaluationmeasures-matched

sample

Earnings

Managem

ent

Stock

Liquidity

MarketValuation

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

DAC

AWCA

EM

BAS

ZRET

ILLIQ

UMBR

TOB

VAL

POST

0.063

0.239

-0.066

-0.619

***

-0.345

***

-0.184

***

0.115

0.037

0.015

ENF

-0.194

*0.363

-0.376

-0.511

***

0.018

-0.568

***

0.182

*0.026

0.287

*

SIZE

0.060

0.152

-0.242

-0.264

**-0.473

***

-0.147

***

0.563

***

0.267

***

0.657

***

BIG

FOUR

0.031

-0.338

-0.096

0.191

0.199

-0.007

0.121

-0.031

0.061

CROSSL

IFRS

0.299

*0.861

-0.071

0.086

0.249

-0.209

**-0.517

**-0.221

*-0.480

IFRSFIRST

-0.145

0.542

0.224

0.204

-0.234

-0.041

0.278

*0.143

0.367

IFRSMAND

-0.175

-1.778

*-1.387

*0.355

0.246

-0.037

-0.586

**-0.138

-0.342

USGAAP

0.071

1.046

-0.226

0.181

-0.278

-0.062

*-0.531

***

-0.240

***

-0.699

**

CLOSEH

-1.703

***

-0.459

-0.169

0.104

0.016

0.090

-0.065

0.054

-0.074

LEV

0.566

2.890

0.173

-0.794

*-0.378

-0.066

0.074

-0.168

-0.695

MKTVOLA

-0.102

0.009

-0.050

1.109

***

0.331

0.369

***

-0.209

-0.156

*-0.155

ROA

-0.087

0.723

1.129

1.209

*-0.121

0.175

-1.487

*-0.712

***

-1.707

VOLA

0.066

-0.334

-0.023

-0.048

-0.321

-0.060

0.294

0.081

0.506

*

Adj.R-squared

0.798

0.457

0.481

0.687

0.819

0.916

0.919

0.853

0.781

Observations

8896

88104

104

104

104

104

104

Firm-�xed

e�ects

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Notes:

Thistableshow

smultivariate

regressionresultsforH1to

H3forasampleoftreatm

entandbenchmark

companiesmatched

bythe

accountingstandard

pro�leandsize.Variablesarede�ned

inTable5.1.WeestimatetheregressionsusingpanelstructureandWhiteperiod

clustered

standard

errors.*,**,and***indicate

one-tailed

signi�cance

atthe0.1,0.05,and0.01levelforourvariableofinterest(E

NF

it)

inaccordance

withtheexpectedimpact

asstatedin

H1to

H3andtwo-tailed

signi�cance

atthe0.1,0.05,and0.01levelforseveralother

variables.

115


We �nd impacts with the following levels of signi�cance on our primary variables of

interest (ENF it): t = -1.398 for EM it, t = -3.735 for ILLIQU it, and t = 0.593 for

VALit. The overall-�ts of the regression models are slightly lower: 0.492, 0.735, and

0.730.

Third, we re-perform our analyses using a sample of companies which apply IFRS

though the investigation period (2003 to 2006). Thus, there are no longer changes

in accounting principles but some companies apply IFRS on a voluntary and some

on a mandatory basis. We adequately modify our research design, keeping our

control variables IFRSFIRST it and IFRSMAND it. Although the number of ob-

servations considerably decreases we �nd qualitatively very similar results for our

earnings management and stock liquidity measures. Further, we document a signif-

icant positive impact of the enforcement reforms on every valuation measure. We

�nd impacts with the following levels of signi�cance on our primary variables of

interest (ENF it): t = -1.670 for EM it, t = -1.873 for ILLIQU it, and t = 1.307 for

VALit. The overall-�ts of the regression models are slightly lower: 0.487, 0.762, and

0.748.

Error �ndings by the DPR/BaFin mechanism as an interacting variable

A key element of the enforcement reforms was the establishment of the two-tier

external �nancial reporting oversight mechanism. Since 2005 the DPR and BaFin

conduct investigations of �nancial reports and mandate infringing companies to an-

nounce error �ndings. We take these error announcements by infringing companies

as an additional proxy for the degree of enforcement by other mechanisms. Compa-

nies that are required to announce an error release by the DPR/BaFin enforcement

are likely to have a low degree of enforcement through other institutions/mechanisms

and hence might bene�t more from the enforcement reforms. In line with this argu-

ments, we �nd that the binary variable ERRORit which indicates whether a company

had an error announcement, is signi�cantly negative correlated with ENFOTHERit

(ρ = -0.186). We estimate Equation 5.2 including ENF it · ERRORit as an �inter-

action term� and �nd that there is no e�ect on EM it (t = 0.283) but two close-

to-signi�cance impacts indicating that companies characterized by a low account-

ing quality are particularly a�ected by the enforcement reforms (t = 1.276 for the

116

5.6 Conclusion

ILLIQU it and t = 1.273 for the VALit regression model).7

Technical issues

In our main analyses we use �rm-�xed e�ects to ensure that the results are not

in�uenced by static unobservable cross-sectional di�erences between the companies

in our benchmark and treatment groups. We also conduct our regressions using

industry-�xed e�ects. While the overall �ts of our regressions decrease considerably

(for EM it, ILLIQU it, and VALit we document adjusted R2s of 0.518, 0.494, and

0.190) we observe slightly higher impacts of the enforcement reforms (ENF it) for

several measurement variables. We note that the remarkably high overall �ts of

our regressions are considerably driven by the inclusion of �rm-�xed e�ects. The

adjusted R2s of our regressions without �rm- or industry-�xed e�ects for EM it,

ILLIQU it, and VALit are 50.4%, 48.1%, and 16.9%.

To control for the impact of our outlier treatment (winsorizing), we re-run our

regressions with variables truncated at the 2.5%- and 97.5%-levels. Still requiring

a balanced sample, we note that the number of observations within our sub-sample

of companies listed in the Open Market segment decreases by 41.0%. However, the

�ndings concerning H1 to H3 are very similar. For H4, we �nd a signi�cantly positive

impact of the ENF it ·BIGFOURit interaction term in the EM it regression indicating

that the e�ect of the enforcement reforms is more pronounced for companies not

audited by a �Big Four� auditor.

5.6 Conclusion

Accounting outcomes are in�uenced not only by the type of accounting regime ap-

plied but also by the degree of enforcement. Enforcement refers to a system of

mechanisms to prevent errors in �nancial reports. This study investigates recent re-

forms of the enforcement of �nancial reports in Germany. These reforms established

7To examine whether our �ndings are restricted to infringed companies, we conduct several analysesbased on a sample which excludes several companies for which the DPR/BaFin-mechanism founderroneous �nancial reporting for the �scal years 2005 and 2006. For our largest sample of 1,312observations this leads to an elimination of 116 observations. However, �ndings are very close toour original analyses. We �nd impacts with the following levels of signi�cance on our primaryvariable of interest (ENF it): t = 1.997 for EM it, t = 2.072 for ILLIQU it, and t = 1.584 forVALit. The overall-�ts of the regression models are slightly lower: 0.509, 0.804, and 0.768.

117


an external two-tier enforcement mechanism, restructured the auditor oversight, and

introduced new independence rules for auditors. An in-depth theoretical analysis

of the impact of the reforms shows that they increase the likelihood of detecting

companies that issue inaccurate �nancial reports, and that they reinforce sanctions

for such companies and their auditors. Hence, the reforms increased the degree of

enforcement in Germany.

In line with the goal of the reforms, we provide empirical evidence that they had a

positive impact on the earnings quality, the stock liquidity, and the market valuation

of the a�ected companies. Particularly a�ected are small companies, companies that

do not appoint a �Big Four� auditor and companies that are not cross-listed and thus

could be said to have a low level of enforcement. Our results hold for various model

speci�cations and for several other sensitivity analyses.

Still, we acknowledge that our inference of the economic consequences of the Ger-

man enforcement reforms is subject to a general caveat. We assume that these

changes are due to the enforcement reforms but our research design could still be

a�ected by the mandatory adoption of IFRS in 2005, which was a confounding event

for our study. To exclude other possible explanations for changes in the economic

variables, we control for (changes in) company characteristics, for changes in the

accounting principles applied, and for changes in the stock market. Moreover, we

perform a di�erence-in-di�erences analysis to isolate the enforcement impact from

the impact of changes of the accounting standards or other factors. We also perform

a matching procedure and use several di�erent samples to disentangle accounting

standard e�ects from enforcement e�ects. The speci�c setting in Germany where

due to listing requirements basically all companies in a certain stock segment had to

use IFRS since 2003 gives us the opportunity to control even for a (quasi-) manda-

tory adoption e�ect. This is important as prior research (e.g., Daske et al., 2008)

suggests that voluntary adopters also bene�tted from mandatory adoption, espe-

cially in terms of stock liquidity and market valuation. Yet, our results also provide

evidence for a decrease in earnings management activities which should not primar-

ily be impacted by changes in accounting standards. In the end, however, the results

of this study should be interpreted with some caution with respect to the bene�ts

of the enforcement reforms, because the mandatory adoption of IFRS in Europe

in 2005 could have had an additional impact on the quasi-mandatory companies in

Germany by further increasing the comparability to peers in other European coun-

118

5.6 Conclusion

tries. Still, Germany represents a unique opportunity to disentangle enforcement

changes from the IFRS adoption e�ect due to its nonesuch setting of di�erent stock

exchange segments.

Overall, we conclude that the reforms in Germany have leveled the playing �eld

in the enforcement of �nancial reporting. After the reforms, the average earnings

quality and capital market properties of the a�ected companies improved, especially

for companies with a lower degree of enforcement through other mechanisms. How-

ever, more research on the impact of mandatory enforcement reforms is needed to

add to our understanding of the bene�ts of regulatory reforms in general and of the

degree of voluntary or mandatory enforcement in particular.

119


120

6 The Real Business E�ects of

Quarterly Reporting

Working paper (with Jürgen Ernstberger and Benedikt Link)

6.1 Introduction

Anecdotal evidence suggests that managers, when required to issue quarterly �nan-

cial reports, are forced to make short-sighted business decisions to meet earnings

targets, often at the expense of long-term value. The most prominent example for

opposition to mandatory quarterly reporting is probably car manufacturer Porsche.

In 2003, the company's former CEO Wendelin Wiedeking refused to publish quar-

terly reports claiming that it triggers short-sighted management decisions. This

resulted in a delisting of the company from the German stock market index for mid-

caps (M-DAX).1 In line with this argument, in a recent survey among executives

in the U.S., 80% of CFOs admit they would decrease discretionary spending (e.g.,

R&D) given the pressure to meet quarterly earnings targets. Almost 40% would

even provide incentives for customers to buy more products in the current period

by, e.g., increasing discount levels (Graham et al., 2005). The authors argue that

managers are willing to sacri�ce long term value in order not to fail meeting short

term expectations of analysts and investors.

In order to test empirically whether reporting frequency leads to a deviation from

normal operational practices, i.e. has �real business e�ects�, we compare quarterly

reporters with companies that report semi-annually. The European Union (EU) con-

stitutes a unique regulatory setup to test for di�erences between reporting frequency

regimes because quarterly reporting is mandatory in some countries and voluntary

1Other prominent examples of companies returning to semi-annual reporting include BAA (world'sleading airport operator) and BHP Billiton (global leader in resource industry).

121

6 The Real Business E�ects of Quarterly Reporting

in others. Each member state and stock market operator in the EU can decide upon

the required frequency of �nancial reporting individually. In our sample, 5 countries

and 3 stock market operators have adopted mandatory quarterly reporting, while 7

countries remain voluntary reporting regimes. Moreover, the EU has a harmonized

institutional and regulatory base (e.g., all companies have to report under IFRS),

which makes it easier to compare di�erent countries and isolate speci�c e�ects.2

Other settings, notably the U.S., do not o�er such a variation in the frequency of

reporting and/or have greater cross-country di�erences to control for.

There is surprisingly little research on the micro- and macroeconomic implications

of disclosure regimes (Leuz and Wysocki, 2008). In fact, previous research on the ef-

fects of interim reporting frequency choices is still underdeveloped. Previous studies

mainly examine the e�ects of quarterly reporting on capital markets such as price

volatility, analyst following (Rahman et al., 2007), and timeliness of earnings (Butler

et al., 2007). There is also a signi�cant amount of literature on the earnings prop-

erties in quarterly reporting (e.g., Degeorge et al., 1999; Brown and Caylor, 2005;

Das et al., 2009). To date, the implications of reporting frequency on operational

practices, i.e. the �real business e�ects�, has not been addressed in the literature.

Our objective is to �ll this void. Speci�cally, we examine whether interim reporting

frequency has an e�ect on the level of real earnings management (REM) de�ned

as a deviation from normal operational practices, e.g., for the purpose of avoiding

earnings surprises and other adverse e�ects such as loss in reputation. Following

prior research (Dechow et al., 1998; Roychowdhury, 2006), we measure REM using

abnormal cash �ow from operations (ACFO), abnormal production cost (APROD),

and abnormal discretionary expenses (ADISC), where ACFO consolidates the di�er-

ent manipulative e�ects such as sales manipulation, overproduction, and a reduction

in discretionary expenses.

We examine whether companies in mandatory and voluntary reporting regimes

exhibit di�erent levels of REM. Based on a simple manager decision model, we ex-

pect managers of mandatory quarterly reporters to use more REM compared to

semi-annual and voluntary reporters because they are forced to disclose more than

they would in a voluntary setting. Higher reporting frequency leads to additional

expected costs for managers, such as increased pressure to meet earnings targets

2In 2004, the European Commission adopted a Directive to increase the transparency of �nancialreporting in Europe leaving reporting frequency regulation to the member states and stock marketoperators (for details see Section 6.2).

122

6.1 Introduction

due to increased transparency. Therefore, they are more likely to use REM in order

to reduce these additional expected costs. In turn, we expect managers of volun-

tary quarterly reporters to use less REM compared to mandatory and semi-annual

reporters. In line with Leuz and Wysocki (2008), we assume that primarily �good�

�rms have an incentive to disclose more than the average �rm in order to convince

investors and analysts of a higher justi�ed share price. For example, managers that

consistently believe to outperform earnings forecasts should have an incentive to

report quarterly. These �rms should have a lower level of REM compared to semi-

annual reporters because they do not need to use REM to meet earnings forecasts.

They might even use negative REM to manage earnings downwards in order not to

reveal the full earnings potential.

We also investigate whether the e�ect is dependent on potential discriminating

factors, because we hypothesize that the magnitude and nature of the e�ect depends

on speci�c settings. We analyze di�erent settings/conditions suspected to increase

the overall level of REM and in�uencing the e�ect of interim reporting frequency on

REM (�suspect �rm-years�). Our suspect �rm-year de�nitions include (1) �rm-years

with below industry-year median company performance due to a higher likelihood

of missing earnings forecasts, (2) �rm-years in which a company uses above average

accounting earnings management (AEM) assuming that managers tend to use both

forms of earnings management, (3) �rm-years with below median analyst following

due to lower monitoring by analysts (Yu, 2008), and (4) �rm-years in countries with

below median anti-director rights due to a low investor protection environment, in

which insiders are likely to have greater freedom to proactively manage earnings in

their best interest (Leuz et al., 2003).

Using a sample of 16,305 �rm-year observations from EU-15 countries, we �nd that

mandatory quarterly reporters generally exhibit higher levels of REM compared to

semi-annual reporters. This e�ect is particularly strong in suspect �rm-years. In

line with expectations, we �nd that voluntary reporters exhibit signi�cantly lower

levels of REM in non-suspect �rm-years compared to mandatory and semi-annual

reporters suggesting that primarily �good� �rms choose to report voluntarily. This

e�ect disappears in suspect �rm-years, where voluntary reporters exhibit similar

REM levels than semi-annual reporters.

Our �ndings are particularly interesting as they contribute to previous research

in at least two until recently underdeveloped �elds (Leuz and Wysocki, 2008). First,

123


the �ndings contribute to the growing �real e�ects� literature in accounting by pre-

senting evidence on real business implications of regulations. Roychowdhury (2006)

shows that when companies are close to zero-earnings, they tend to deviate from

normal operational practices by, e.g., price discounts and a reduction in discretionary

expenses. Other studies have analyzed the real e�ects of accounting regulation re-

cently in a di�erent context (e.g., Zang, 2007; Cohen et al., 2008; Gunny, 2005). For

example, Cohen et al. (2008) show that REM has been used more extensively com-

pared to accounting earnings management after the introduction of Sarbanes-Oxley.

We contribute to this literature as the �rst study to present evidence on the �real�

e�ects of interim reporting frequency.

Second, we add to the literature on the institutional and regulatory e�ects of re-

porting and disclosure regimes. We show that in mandatory reporting regimes quar-

terly reporting leads to signi�cantly higher REM compared to voluntary regimes.

We also provide evidence that the direction (voluntary setting) and magnitude (vol-

untary and mandatory setting) of the e�ects depends on the institutional environ-

ment. Previous studies on the e�ects of reporting frequency have often restricted

themselves to a voluntary setting (e.g., Rahman et al., 2007), in which the e�ects

can only be analyzed on an aggregate level. Other studies comparing mandatory

and voluntary reporting are either based on the U.S., where the analyzed e�ects

date back to the 1970s (e.g., Butler et al., 2007), or include only a limited number

of countries (Mensah and Werner, 2008) with diverse institutional and regulatory

settings (e.g., di�erent accounting standards).3 We compare a large set of countries

that all apply IFRS and are committed to similar disclosure regimes with one of

the few degrees of freedom being the disclosure frequency. Thereby, we are able to

better isolate the e�ect of reporting frequency and better control for di�erences in

reporting regimes than previous international studies. Moreover, none of the previ-

ous studies has particularly focused on investigating the e�ect of interim reporting

frequency on REM.

Section 6.2 describes the regulatory background of our study. Section 6.3 discusses

the related literature. Section 6.4 develops our hypotheses. Section 6.5 discusses the

methodology used. In Section 6.6, we describe the sample selection and descriptive

statistics. Section 6.7 presents the empirical results of our main and additional anal-

yses. Section 6.8 discusses corresponding sensitivity analyses, Section 6.9 concludes.

3Mensah and Werner (2008) compare the US, the UK, Canada and Australia.

124

6.2 Background on quarterly reporting in Europe


The regulatory and institutional background in the EU is most suitable to analyze

di�erences between quarterly and semi-annual reporters and mandatory and vol-

untary reporting regimes. The underlying reason is that the EU currently has a

unique regulatory setup: While the level of harmonization of �nancial regulation is

very high,4 member states and stock market operators have the choice to either make

quarterly reporting mandatory or leave it to the individual company to decide if it

wants to issue quarterly �nancial reports. Due to the high level of harmonization

on the one hand and the di�erent reporting frequency regimes on the other hand,

it is much easier to isolate the e�ect of reporting frequency than in comparable in-

ternational studies. In our sample, we include the 15 member states that joined the

EU before 1995, called the EU-15.5

6.2.1 EU Transparency Directive

In 2004, the EU took a major step in order to further harmonize regulation of

its member states, when setting minimum publication requirements for companies

publicly-traded in the EU. The corresponding directive, the so-called �Transparency

Directive�, aims at enhancing the level of information to investors, setting mini-

mum standards to the publication requirements for publicly-traded companies in

the member states, and improves the dissemination of information on issuers.6

One of the main novelties of the Transparency Directive, compared to the pre-

vious regulatory framework, is the requirement to disclose �Interim Management

Statements� (IMS) after the �rst and third quarter of the �nancial year. The Eu-

ropean Commission originally wanted all publicly-traded companies in the EU to

4All publicly-traded companies have to report according to IFRS. The level of enforcement isreasonably similar. Capital markets are highly integrated and there is a common supervisorybody next to the national regulators.

5The EU was formally established with the Maastricht Treaty in 1993 and Austria, Sweden, andFinland joined in 1995. Since 2004, the EU was successively enlarged to 27 member states asof today. Many of the new member states are transitional economies, however, and thereforelack comparability with the regulatory, economic and institutional environment of the incumbentmember states. For better comparison, we therefore restrict ourselves to the EU-15, i.e., thecountries that joined until 1995.

6Unlike other directives, the Transparency Directive 2004/109/EC is not a �maximum harmo-nization� directive. Member states can set requirements above the minimum level envisaged inthe directive, especially with respect to reporting frequency. It was adopted in December 2004,therefore our sample comprises the years 2005�2009.

125


move to a U.S.-like system of quarterly reporting to increase transparency towards

investors and other stakeholders. However, several countries (e.g., UK, Denmark,

and the Netherlands) objected this proposal or raised concerns about a potential

increase in management and investor myopia at the expense of long term value.

The European �nance ministers �nally decided to have audited �nancial statements

twice a year and (qualitative) IMS in the �rst and third quarter. In fact, the EU's

minimum quarterly disclosure requirements only demand a qualitative statement,

not necessarily including �nancial �gures. As opposed to semi-annual reports, there

is no requirement to issue a complete or condensed set of quarterly �nancial state-

ments in accordance with International Accounting Standard 34, Interim Financial

Reporting (IAS 34).7 In other words, the Transparency Directive has set minimum

standards on interim �nancial disclosure, while giving its member states the option

to enact stricter rules on quarterly disclosure. In all member states publicly-traded

�rms can opt to voluntarily publish full quarterly reports in accordance with IAS

34 instead of IMS.

While the usefulness of IMS is currently being debated,8 the Transparency Di-

rective provides a reasonably similar regulatory base for all countries included in

our sample. The choice of IMS versus full quarterly reports represents a primary

disclosure choice available to �rms, member states or stock market operators, sug-

gesting the current EU setting is favorable to isolate the e�ect of �nancial reporting

frequency on operational practices.

6.2.2 Country-speci�c regulation

Table 6.1 gives an overview about the speci�c requirements in each EU-15 country,

which we compiled from an extensive review of documents from the EU, national

regulatory authorities, stock exchanges, and from interviews with �nancial analysts

7IAS 34 requires interim �nancial reports to contain either a complete set of �nancial statements(as described in IAS 1) or a set of condensed �nancial statements (IAS 34.8).

8The Regulator has provided little guidance on the nature and presentation of information thatthey would expect IMS to contain. Consequently, there are variations in the way companies pre-pare IMS and there is no clear market consensus of what IMS should include (Ernst & Young,2009; Mazars, 2010). While some predictability and comparability has been provided by CESRin October 2009, the European Commission is currently reviewing the requirements in the Trans-parency Directive, also with respect to IMS. It has asked member states to submit suggestionsfor improvements to the current set of rules. This has no e�ect on the results of the paper as itonly includes observations from 2005�2009.

126


in the respective countries. In Finland, Greece, Italy, Portugal, and Spain, the reg-

ulatory authorities require listed �rms to publish full quarterly reports. In Sweden,

the stock market operator on both regulated stock markets requires full quarterly

�nancial reports for all companies. In Austria and Germany, companies listed in

speci�c stock market segments (�Prime Market� in Austria and �Prime Standard�

in Germany) are also obliged to publish full quarterly reports. The remaining seven

countries (Belgium, Denmark, France, Great Britain, Ireland, Luxembourg, The

Netherlands) do not mandate full quarterly reports or quarterly earnings announce-

ments.

Following previous research, we de�ne �rm-year observations from countries or

stock exchange segments that mandate companies to at least publish quarterly net

earnings as mandatory reporters (Cuijpers and Peek, 2010).9 The remaining com-

panies are either classi�ed as semi-annual reporters or voluntary quarterly reporters

depending on the corresponding Datastream item WC05200 (�earnings reporting

frequency�).10

9Companies that need to issue quarterly reports solely due to stock exchange segment regulationcan arguably also be excluded from the sample because they theoretically have the choice to exitthe index listing. Our results are robust to exclusion of these observations.

10We perform sensitivity checks on the validity of the Datastream item. Changing the de�nitionof quarterly reporters to those companies for which quarterly net earnings are actually availablein Datastream does not a�ect our results.

127


Table

6.1:Quarterly

reporting

environm

entin

EU-15countries

Country

Quarterly

Reporting

Manda-

tory?

QuarterlyReportingRules

Regulator

Stock

Market

Austria(AT)

Yes

(A)

Austrian

stock

market(W

iener

Börse)re-

quires

�rm

slisted

in�PrimeMarket�

seg-

mentto

publish

fullQuarterlyReports

(in

acc.

withIA

S34)

Finanzm

arktaufsicht

(FMA)

Wiener

Börse

Belgium

(BE)

No

Listed�rm

sarenot

required

topublish

quar-

terlyearnings

Com

mission

Bancaire,Fi-

nancière

etdes

Assur-

ances(C

BFA)

EuronextBrussels

Denmark(D

K)No

Listed�rm

sarenot

required

topublish

quar-

terlyearnings

(How

ever,beforeCopenhagen

Stock

Exchange

was

takenover�rstbyOMX

in2005

andthen

laterbyNASDAQ

in2008,

�rm

swereincentivized

bythestock

exchange

topublish

quarterlyreports;thisleadsto

astillhighQRpracticein

Denmarktoday)

Finanstilsynet

(Danish

FSA)

NASDAQ

OMX

Copen-

hagen

Finland(FI)

Yes

(B)

TheFinish

Securities

MarketAct

stillre-

quires

listed

companiesto

presentinterim

re-

sultsforthe�rstthree,sixandninemonths

ofthe�nancialperiod(chapter2,section5);

thereareonly

very

limited

exceptions

Finanssivalvonta

(FIVA)

NASDAQ

OMXHelsinki

Continued

onnext

page

128


Quarterly

reporting

environm

entin

EU-15countries�Continued

Country

Quarterly

Reporting

Manda-

tory?


Regulator

Stock

Market

France

(FR)

No

Listed�rm

sarenot

required

topublish

quar-

terlyearnings

(How

ever,theFrench

Mone-

tary

andFinancialCoderequires

listed

�rm

sto

publish

quarterlynetsalesbysubsidiaries)

Autorité

des

Marchés

Fi-

nanciers(A

MF)

EuronextParis

Germany(D

E)Yes

(A)

German

stock

market

operator

Deutsche

Börse

requires

�rm

slisted

in�PrimeStan-

dard�segm

entto

publish

fullQuarterlyRe-

ports

(inacc.

withIA

S34)

Bundesanstalt

für

Finanzdienst-

leistungsaufsicht(BaF

in)

Frankfurter

Wertpapier-

börse

Great

Britain

(GB)

No

Listed�rm

sarenot

required

topublish

quar-

terlyearnings

Financial

Services

Au-

thority(FSA)

NYSEEuronextLondon

Greece(G

R)

Yes

(B)

Greek

regulation

authority(H

CMC)princi-

pallyrequires

alllisted

�rm

sto

publish

quar-

terlyreports

(withlimited

exceptions)

Hellenic

Capital

Market

Com

mission

(HCMC)

AthensExchange

Securi-

ties

Market

Ireland(IE)

No

Listed�rm

sarenot

required

topublish

quar-

terlyearnings

Irish

Financial

Services

Regulatory

Authority

(IFSRA)

IrishStock

Exchange

Italy(IT)

Yes

(B)

Italianregulation

authority(C

ONSOB)prin-

cipally

requires

alllisted

�rm

sto

publish

quarterlyreports

(withlimited

exceptions)

Com

missione

Nazionale

per

leSocietàela

Borsa

(CONSOB)

Borsa

Italiana

Luxem

bourg

(LU)

No

Listed�rm

sarenot

required

topublish

quar-

terlyearnings

Com

mission

de

Surveil-

lance

du

Secteur

Fi-

nancier

(CSSF)

BoursedeLuxem

bourg

Continued

onnext

page

129


Quarterly

reporting

environm

entin

EU-15countries�Continued

Country

Quarterly

Reporting

Manda-

tory?


Regulator

Stock

Market

The

Nether-

lands(N

L)

No

Listed�rm

sarenot

required

topublish

quar-

terlyearnings

Autoriteit

Financiële

Markten(A

FM)

EuronextAmsterdam

Portugal(PT)

Yes

(B)

Portuguese

regulation

authority

(CMVM)

principally

requires

alllisted

�rm

sto

either

publish

fullquarterlyreports

(inacc.

with

IAS34)or

speci�cquarterly�nancials(in-

cludingquarterlyearnings)

Com

issão

do

Mercado

de

Valores

Mobiliários

(CMVM)

EuronextLisbon

Spain(ES)

Yes

(B)

Spanishregulation

authority(C

NMV)prin-

cipallyrequiresalllisted

�rm

sto

publish

spe-

ci�cquarterly�nancials(includingquarterly

earnings)

Com

isión

Nacional

del

Mercado

de

Valores

(CNMV)

Bolsa

deMadrid

Sweden

(SE)

Yes

(C)

Under

thelistingrulesforthetworegulated

markets

inSweden,listed

�rm

sarerequired

topublish

fullquarterlyreports(inacc.

with

IAS34)

Finansinspektionen

(SwedishFSA)

NASDAQ

OMX

Stock-

holm

Notes:

Thistablespresents

thedi�erentquarterly

reportingregulationbycountryin

theEU-15countries.

Yes

(A)indicatesrequired

by

stock

exchangesegment,Yes

(B)indicatedrequired

byregulator,andYes

(C)indicatesrequired

bystock

exchange.

130



6.3.1 Quarterly reporting

Several studies investigate earnings properties in quarterly �nancial reports. Dege-

orge et al. (1999) propose a quarterly earnings threshold hierarchy, i.e. managers

seek to avoid quarterly losses or quarterly earnings decreases more than meeting or

beating �nancial analysts' quarterly earnings forecasts. Brown and Caylor (2005)

examine and �nd temporal changes in this hierarchy. Since the mid-1990s �rms

appear to focus more on avoiding quarterly earnings surprises due to the increased

importance of analysts' forecasts. Das et al. (2009) document that patterns in quar-

terly earnings changes re�ect accounting earnings management behavior. Companies

that perform poorly in interim quarters try to increase earnings in the last quarter

to reach earnings benchmarks. Butler et al. (2007) examine the timeliness e�ects

of voluntary and mandatory quarterly earnings disclosure in the U.S., using a his-

torical sample (1950�1973) that allows drawing conclusions on voluntary quarterly

adoption also for the U.S. (i.e., before 1970). They �nd that voluntary reporters

recognize bad news timelier, whereas �rms mandated by the SEC to report more

frequently experience no signi�cant e�ects. Another stream of research on quar-

terly �nancial reporting examines how it in�uences capital market properties. In

a mandatory disclosure setting, Mensah and Werner (2008) examine how interim

reporting frequency a�ects stock price volatility. They use an international sample

based on U.S. and Canadian �rms (quarterly reporting) as well as British and Aus-

tralian �rms (semi-annual reporting) and show that semi-annual reporting leads to

lesser price volatility after accounting for in�uential e�ects. In contrast, Rahman

et al. (2007) build upon a voluntary quarterly reporting environment (Singapore).

They �nd that the key determinants of voluntary quarterly reporting are growth

perspectives, size and technology orientation. Moreover, quarterly reporting is as-

sociated with higher analyst following, but also with higher price volatility. Also in

a voluntary setting, Cuijpers and Peek (2010) analyze whether the choice between

quarterly and semi-annual reporting a�ects the precision of investors' information

and their private information acquisition activities. They document that voluntary

quarterly reporting leads to less precise pre-announcement information. They argue

that this is the consequence of reduced incentives to acquire private information.

This argument is supported by showing that quarterly reporters have lower infor-

131


mation asymmetry expressed by lower average bid-ask spreads and higher abnormal

share turnover.

Our study contributes to this literature by investigating the real business e�ects

of quarterly �nancial reporting. 80% of the managers responding to the Graham

et al. (2005) survey say they would decrease discretionary spending (e.g. R&D)

given the pressure to meet a quarterly earnings target. Almost 40% would even

provide incentives for customers to buy more products in the current quarter.

6.3.2 Real business e�ects

Prior research on earnings management has indicated that it occurs when managers

use judgment and discretion in �nancial reporting to mislead stakeholders about

the underlying economic performance (Healy and Wahlen, 1999). Managers have

at least two options to achieve this: They can either use accounting based earnings

management by accruals manipulation and classi�cation shifting or they can use real

earnings management de�ned as adapting their operational practices to in�uence

�nancial performance. A variety of studies have shown that managers in�uence

�nancial reporting through accounting discretion (typical �earnings management�)

and classi�cation shifting (e.g., Jones, 1991; Dechow and Dichev, 2002; Kothari et al.,

2005; McVay, 2006; Fan et al., 2010).

A growing stream of literature also investigates managers' willingness to depart

from normal operational practices ostensibly to achieve certain reporting goals �

often as a consequence of management myopia. We de�ne real business e�ects as

the �nancial e�ects of a deviation from normal operational practices for the purpose

of earnings manipulation. These practices include, for example, price discounts to

increase short term revenues, production increases to reduce average cost per unit,

and the reduction of discretionary expenses to boost short term pro�t. Previous

research on real earnings management focuses on opportunistic R&D expenditure

reduction to meet earnings forecasts or avoid share price dilution (e.g., Bens et al.,

2003; Dechow and Sloan, 1991). Other �ndings include the acceleration of sales and

the delay of R&D and other discretionary expenses (e.g., Healy and Wahlen, 1999;

Fudenberg and Tirole, 1995; Dichev and Skinner, 2002). In a more comprehensive

approach, Roychowdhury (2006) �nds evidence for abnormal levels of cash �ow from

operations, production cost and discretionary expenses triggered by managers trying

to avoid reporting losses. This is in line with evidence from surveys and experiments

132

6.4 Hypotheses

on myopic management behavior (e.g., Graham et al., 2005; Bhojraj and Libby,

2005). Using similar measures, Cohen et al. (2008) �nd that in the post-Sabanes

Oxley period, many �rms switched from accrual-based to real earnings management

methods. This is consistent with the model of Ewert and Wagenhofer (2005) who

suggest that tighter accounting standards might increase the expected total cost of

earnings management due to a switch from accounting to more costly real earnings

management.

We contribute to this literature by investigating the e�ect of di�erent reporting

regimes on real earnings management.

6.4 Hypotheses

We examine whether increased interim reporting frequency is associated with higher

levels of real earnings management (REM). We compare mandatory quarterly re-

porters, i.e., �rms that are mandated by country or stock market regulation to report

quarterly, voluntary quarterly reporters, i.e., �rms that report quarterly although

they would be allowed to report semi-annually, and semi-annual reporters. Based

on previous research, we develop a simpli�ed decision model for managers to derive

our hypotheses.

Quarterly reporting has bene�ts and costs for managers (e.g., Butler et al., 2007;

Graham et al., 2005; Leuz and Wysocki, 2008; Rahman et al., 2007).11 Speci�cally,

bene�ts of quarterly reporting include, for example, increased credibility towards

investors, better access to �nancing and the admittance to speci�c stock market in-

dices. This can in turn lead to a higher trading volume as well as an increase in the

share price and thereby result in an increase of performance-based part of a man-

ager's compensation. More frequent disclosure also improves the ability of analysts

and investors to determine the �real� value of a �rm. It can therefore be bene�cial

for �good� �rms/managers, e.g., managers that expect to outperform investor ex-

pectations, to disclose more than the average �rm in order to convince shareholders

of a higher justi�ed �rm value (Leuz and Wysocki, 2008). The associated costs for

11Previous research makes no clear distinction between bene�ts and costs of reporting frequencyfor managers and for companies. In many cases these will be similar. In our study, we focus onmanagers because we assume that managers can choose the reporting frequency of their companyindependently of other stakeholders. Consequently, manager's expected costs and bene�ts ofreporting frequency determine its reporting frequency choice.

133


managers of a higher reporting frequency include increased transparency towards

investors and analysts, a higher tendency to deviate from long-term value-optimal

operating practices due to induced myopia, as well as loss in economic rents due to

publicizing proprietary information (e.g., Harris, 1998; Butler et al., 2007; Rahman

et al., 2007). If managers have the choice, they will adopt quarterly reporting only if

the average expected bene�ts exceed the costs. This trade-o� decision is illustrated

in Table 6.2.

As outlined in Section 6.3, several studies investigate accounting earnings manage-

ment properties depending on di�erent reporting frequency regimes (e.g., Degeorge

et al., 1999; Balsam et al., 2002; Das et al., 2009). Our study focuses on the re-

porting frequency e�ect on REM, which has not been investigated so far. Although

we do not explicitly analyze the trade-o� between AEM and REM, we argue that

reporting frequency should a�ect both forms of earnings management. In line with

the ongoing discussion about reporting frequency-induced management myopia re-

sulting in short-sighted business decisions, we focus primarily on REM and discuss

the relation with AEM in one of the suspect �rm-year analyses. Moreover, Zang

(2007) demonstrates that the REM e�ect is especially important under high scrutiny

conditions.

In a mandatory quarterly reporting environment, all �rms have to issue quar-

terly reports independent of the manager's preferences. Consequently, in line with

our manager decision model, many managers are forced to report quarterly that

otherwise would only report semi-annually. In other words, many managers have

to report quarterly although the costs of additional disclosure exceed the bene�ts

(Table 6.2). Assuming a similar distribution of type A and B managers in volun-

tary and mandatory settings, there are relatively more managers with higher costs

then bene�ts from a quarterly reporting in a mandatory environment. This implies

that additional net costs are imposed to managers due to the mandatory report-

ing frequency. For example, quarterly reporting increases the pressure to meet short

term results, which in turn leads to myopic management decisions re�ected in higher

REM. Also, managers are likely to increasingly use REM to smooth earnings or avoid

missing earnings targets by managing earnings upwards through real activities ma-

nipulation. In line with the recent discussion in the EU, we argue that a higher

reporting frequency increases the short term pressure to deliver results, which might

force managers to adopt operational practices disconnected from the business cycle,

134

6.4 Hypotheses

Table

6.2:Simpli�ed

modelforexpectedREM

activity

depending

onmanager

typeandquarterlyreporting

regime

Quarterly

Reporting

Regim

e

Manager

Type

Description

Bene�ts

(b)

and

Cost(c)of

QR

Resulting

Reporting

Choice

Realized

Net-Bene�t

from

QR

Expected

REM

activ-

ity

Voluntary

AManagersthat

incur

higher

cost

than

ben-

e�ts

from

QR

b<c

Sem

i-annual

0n/a

(control

group)

BManagersthat

incur

higher

bene�ts

than

costsfrom

QR

b>c

Quarterly

b−c>

0Low

erREM

Mandatory

A/B

Mix

oftype

Aand

type

Bmanagers

(typeAmanagersex-

pectedto

dom

inate)*

If]A

>]B

:b<c(If

]B>]A

:b>c)*

Quarterly

(nochoice)

If]A

>]B

:b−c<

0(If]B

>]A

:b−c>

0)*

Higher

REM

(Low

erREM)*

Notes:

This�gure

illustratestheexpectedREM

activitylevelbymanager

typeanddi�erentquarterly

reportingregimes.]A

and]B

meansnumber

oftypeAmanagersandnumber

oftypeBmanagers,respectively.*Share

oftypeAmanagersfoundto

besigni�cantly

larger

involuntary

regimes;similardistributionassumed

inmandatory

regime.

135


such as excessive price discounts or more lenient credit terms. Consequently, we

expect to �nd higher REM levels in mandatory reporting regimes, when compared

to semi-annual reporters:

H1a: Mandatory quarterly reporters exhibit higher levels of REM com-

pared to semi-annual reporters.

In a voluntary reporting regime, managers can choose to either issue quarterly

or semi-annual reports. Managers will opt for quarterly reporting if the expected

bene�ts exceed the costs (Table 6.2). One of the major bene�ts of increased dis-

closure and transparency is the potential increase in stock price if the manager

expects to either systematically outperform analyst and investor expectations or

that the implied risk premium is excessively high (similar to Leuz and Wysocki,

2008). Revealing additional information could therefore help to convince analysts

and investors that company performance is higher than expected or that the implied

risk premium is too high. Both levers potentially increase the share price, which

would also bene�t the managers' performance-based part of compensation. Conse-

quently, managers that expect to on average outperform earnings forecasts are more

likely to opt for voluntary reporting.12 However, it is unlikely that these managers

reveal the full potential to the capital markets because managers have a tendency to

manage earnings downwards in order not to raise expectations to an unsustainable

level. Therefore, we expect managers to only reveal part of their performance poten-

tial adopting either negative REM to manage earnings downwards (e.g., less price

discounts, discretionary increase of expenses) or at least adopt a very low level of

REM. Previous studies have already found managers to manage earnings downwards

using REM (e.g., Edelstein et al., 2008). In any case, we expect voluntary quarterly

reporters to exhibit signi�cantly lower REM compared to semi-annual reporters.

H1b: Voluntary quarterly reporters exhibit lower levels of REM com-

pared to semi-annual reporters.

The e�ect of reporting frequency on REM depends on various factors including

manager and company characteristics as well as external factors. It is therefore

12The choice of reporting frequency cannot easily be changed. Therefore, only managers whobelieve that the bene�ts of increased reporting frequency consistently exceed the costs will optfor quarterly reporting.

136

6.4 Hypotheses

probable that the strength and robustness of our results depend upon several other

factors. While such factors might have an overall level e�ect on REM which holds

for all reporting frequency regimes, the incremental e�ect of reporting frequency on

REM is likely to depend on these factors setting. We identify at least four such

factors suspected to have an in�uence on the investigated REM di�erences. We

create four dummy variables and classify each �rm-year observation that satis�es

the respective condition as �suspect �rm-year�.

First, company performance is likely to have a strong e�ect on the manager's costs

associated with quarterly reporting. Deteriorations in performance need to be faster

communicated to the capital markets and the higher disclosure frequency leaves the

manager less time to counteract. The result is an increase in the probability to

miss analysts' and investor expectations. We hypothesize that managers are using

REM more excessively when the company performs relatively poorly.13 Second, the

e�ect of reporting frequency on REM is likely to increase in the level of accounting

earnings management (AEM). For example, if a manager has a higher tendency to

use AEM in general, or the economic situation results in an overall increase in use

of AEM (e.g., in times of high volatility), managers are probably also willing to use

�real� manipulation to reach their anticipated results. While some evidence suggests

that in case of strong regulatory changes AEM and REM are substitutes (e.g., Ewert

and Wagenhofer, 2005), we argue that, ceteris paribus, if the level of AEM is high,

the absolute level of REM is also higher compared to a low AEM setting. Third,

companies with lower analyst following tend to use more REM because of the lower

level of monitoring. Yu (2008) shows that analysts serve more as external monitors

to managers rather than putting excessive pressure on managers concerning earnings

targets. We therefore conjecture that the impact of quarterly reporting on REM is

more pronounced in an environment of lower monitoring, i.e. lower analyst coverage.

Fourth, in countries with low investor protection insiders conceal �rm performance

by managing earnings in order to protect private control bene�ts from outsiders,

while a strong investor protection mitigates this e�ect (Leuz et al., 2003). Thus, we

predict that countries with lower investor protection represent an environment in

which REM is higher. In line with the above reasoning, we de�ne �suspect �rm-years�

13It is important to de�ne performance relative to competitors as one could otherwise argue thatinvestors adapt their expectations if the economic environment causes the company's situation todeteriorate. Investors are likely to expect companies to perform at least as good as the industryaverage in the long run.

137


as �rm-years with below median company performance or above median accounting

earnings management or below median analyst coverage or �rm-years from countries

with below median anti-director rights score.

In order to test whether we have chosen relevant suspect �rm-years, we test each

of the factors separately by comparing the respective suspect �rm-year observations

to the rest of the sample. For this test we predict:

H2a: REM levels across all interim reporting regimes are higher in sus-

pect �rm-years.

It is important to discuss the e�ect of suspect �rm-years on voluntary and manda-

tory quarterly reporters separately. Following our above reasoning, we expect that

voluntary quarterly reporters exhibit less REM compared to semi-annual reporters as

a consequence of the tendency of �good� �rms to voluntarily disclose more frequently.

The suspect �rm-year conditions, however, are likely to reduce the willingness of

managers of �good� �rms to voluntarily disclose more as they are characterized by

below industry-average performance, an elevated level of AEM, lower monitoring

and low investor protection. In such an environment, managers of �good� �rms will

not materialize the full bene�ts from increased disclosure. In contrast, the higher

disclosure frequency is rather burdensome in suspect �rm-years as they reduce the

likelihood of a positive e�ect of increased disclosure on the share price. We there-

fore expect that voluntary quarterly reporters exhibit lower levels of REM than

semi-annual reporters in non-suspect �rm-years, especially.

In contrast, the expected reporting frequency-induced REM for mandatory quar-

terly reporters should be particularly strong in suspect �rm-years. Managers that

have to report quarterly, even though they would rather report semi-annually, are

likely to increase REM especially in suspect �rm-years, e.g., because of poor perfor-

mance or because of low investor protection.

H2b: Mandatory quarterly reporters (voluntary quarterly reporters) ex-

hibit higher (lower) REM levels in suspect �rm-years (non-suspect �rm-

years)

If the discriminatory nature of suspect �rm-years holds as expected and report-

ing frequency indeed has an e�ect on REM, we expect that the di�erence in REM

138

6.5 Methodology

between suspect and non-suspect �rm-years is signi�cantly higher for quarterly re-

porters (voluntary and mandatory) compared to the semi-annual reporter. In par-

ticular, we expect that also voluntary reporters adopt signi�cant REM in suspect

�rm-years and disproportionally increase their REM levels compared to semi-annual

reporters. We argue that the increased disclosure level becomes a burden in sus-

pect �rm-years and voluntary reporters need to use REM to counteract. Similarly,

in line with the above reasoning, we should be able to �nd a disproportional in-

crease in REM for mandatory quarterly reporters in suspect �rm-years compared to

non-suspect �rm years.

H2c: Mandatory and voluntary quarterly reporters exhibit a dispropor-

tionally high REM di�erence between suspect and non-suspect �rm-years

compared to semi-annual reporters.

6.5 Methodology

6.5.1 Real earnings management

To test for real business e�ects of interim reporting frequency, we investigate pat-

terns in real earnings management, speci�cally abnormal cash �ow from operations

(ACFO), abnormal production costs (APROD) and abnormal discretionary expenses

(ADISC). We use the model from Dechow et al. (1998), further re�ned by Roychowd-

hury (2006), to derive normal and abnormal levels of the corresponding measures.

Subsequent studies such as Zang (2007), Cohen et al. (2008), and Gunny (2005)

and a substantial amount of recent working papers validate the proxies. ACFO

consolidates the e�ect of di�erent earnings manipulation activities like sales manip-

ulation, overproduction and reduction of discretionary expenses. Correspondingly,

an increase in production cost (decrease in discretionary expenses) has a negative

(positive) e�ect on ACFO. Consequently, we focus on ACFO when discussing the

consolidated REM e�ect. APROD and ADISC are used as supporting evidence for

REM activity and as factors (partly) explaining ACFO.

Following Roychowdhury (2006), we de�ne sales manipulation as the acceleration

of current period sales through price discounts and more lenient credit terms. Price

discounts are often used to pull part of next-period sales into the current period to

temporarily boost (absolute) earnings. As a percentage of sales, however, ACFO

139


will decline due to declining margins as a consequence of price discounts. A second

option to boost current period sales is o�ering more lenient credit terms. Many

manufacturing companies (e.g., automotive manufacturers) o�er lower interest rates

towards the end of the �scal year for �nancing their products. This is similar to

price discounts and leads to reduced cash �ow at given sales levels. As a result of

sales manipulation, we expect current period ACFO to be negative.

We de�ne overproduction as managers' tendency to increase production above the

necessary level to reduce cost of goods sold (COGS) per unit by spreading �xed cost

over a larger amount of units (under the assumption that the increase in marginal

cost does not o�set this e�ect). However, the �rm incurs additional production and

holding costs that are not matched with corresponding sales.14 This results in higher

production costs and we expect abnormally high production cost for a given level of

sales. This has a negative e�ect on cash �ow from operations at given sales levels.

We de�ne discretionary expenses as the sum of R&D, advertising, and SG&A

expenses. In order to increase earnings, managers can reduce discretionary expenses

because a reduction does not immediately a�ect sales. If managers decrease dis-

cretionary expenses to meet earnings targets, they will experience abnormally high

earnings in the respective period. We therefore expect abnormally low discretionary

expenses for a given level of sales. This, in turn, increases cash �ow from operations.

As described before, sales manipulation, overproduction and reduction in discre-

tionary expenses have adverse e�ects on cash �ow from operations: While sales ma-

nipulation and production cost e�ects reduce cash �ow from operations, a reduction

in discretionary expenses has the opposite e�ect. Consequently, ACFO could either

be positive or negative. However, following, e.g., Roychowdhury (2006) and Cohen

et al. (2008), empirical evidence indicates that ACFO is indeed negative because

a reduction in discretionary expenses is overcompensated by the other factors (like

sales manipulation and overproduction). We therefore hypothesize to �nd negative

ACFO dependent on reporting frequency and reporting regime.

To estimate the �normal levels� of the dependent variables, we run the following

cross-sectional regressions for every industry and year. In line with Roychowdhury

(2006), we require at least 15 �rm-year observations per 2-digit SIC industry classi-

�cation group for each measure:15

14As stated by Roychowdhury (2006), managers only raise production if the bene�ts from decreas-ing average COGS are higher than the additional holding costs incurred.

15We also perform a sensitivity test using 30 �rm-year observations per industry and a one-digit

140

6.5 Methodology

CFO it

TAi,t−1

= α0 + α11

TAi,t−1

+ α2SALES it

TAi,t−1

+ α3∆SALES it

TAi,t−1

+ εit (6.1)

where all variables are de�ned in Table 6.4.16 ACFO is de�ned as the actual cash �ow

from operations minus the �normal� cash �ow from operations calculated using the

estimated coe�cients above. The residuals of Equation 6.1 are used as dependent

variable in the regression models discussed in this section to test the in�uence of

quarterly reporting. In line with the aforementioned reasoning, we expect ACFO to

be more negative for quarterly reporters.

Production costs are de�ned as the sum of cost of goods sold (COGS) and change

in inventory in the respective year. Following Roychowdhury (2006) and Cohen

et al. (2008), we calculate normal production cost using the following regression:

PROD it

TAi,t−1

= α0 + α11

TAi,t−1

+ α2SALES it

TAi,t−1

+ α3∆SALES it

TAi,t−1

+ α4∆SALES i,t−1

TAi,t−1

+ εit

(6.2)

where all variables are de�ned in Table 6.4. As outlined above, we expect production

costs to be abnormally high in the current period with in turn reduces cash �ow

from operations as a function of sales.

Discretionary expenses include SG&A, R&D and advertising expenses. As adver-

tising expenses are not available for European companies on Datatastream/Worldscope,

we calculate the measure using SG&A and R&D �gures only. Under IFRS, com-

panies capitalize the development part of R&D under certain circumstances which

are to a considerable degree within management's discretion. This level of dis-

cretion reduces the comparability of the o�cial R&D �gures available on Datas-

tream/Worldscope. We therefore use pro forma adjusted R&D �gures published

by the Joint Research Centre and Directorate General Research of the European

Commission, where no R&D is capitalized or impaired (similar to U.S. GAAP). We

model the normal level of discretionary expenses as follows:

DISC it

TAi,t−1

= α0 + α11

TAi,t−1

+ α2SALES i,t−1

TAi,t−1

+ εit (6.3)

SIC code. Our results are robust for this speci�cation.16In accordance with Roychowdhury (2006), we employ not only a scaled intercept when estimat-ing non-discretionary accruals (as is general convention in the literature), but also an unscaledintercept to ensure that the mean abnormal value for every industry year is zero.

141


where all variables are de�ned in Table 6.4. We expect discretionary expenses to be

abnormally low, which in turn reduces cash �ow from operations as a function of

sales.

6.5.2 Mandatory vs. voluntary quarterly reporting (MAND

vs. VOL)

In line with previous research (e.g., Cuijpers and Peek, 2010), we de�ne quarterly

reporters as �rms that publish quarterly earnings, i.e., quarterly net income. Con-

sequently, we use Datastream/Worldscope item WC05200 (�earnings reporting fre-

quency�) to identify quarterly reporters.17

Table 6.1 illustrates the speci�c country regulations on interim reporting. We

classify �rm-year observations of quarterly reporters into �mandatory (MAND)� and

�voluntary (VOL)� quarterly reporters according to the collected regulatory informa-

tion by country. Quarterly reporters in a voluntary reporting regime are named as

VOL and quarterly reporters in a mandatory environment are classi�ed as MAND.

Table 6.3 illustrates the corresponding distribution of �rm-year observations. Note

that there can be MAND observations in countries with voluntary quarterly re-

porting regime due to cross-listing in a mandatory reporting regime, i.e., a stock

exchange segment (index) that requires quarterly reporting. Most of these �rms are

listed in the Prime Market in Austria or the Prime Standard in Germany. As these

companies have to report quarterly due to their cross-listing, we classify them as

mandatory reporters. The distribution of MAND and VOL in the EU-15 countries

is shown in Table 6.3.

6.5.3 �Suspect �rm-year� observations

We de�ne suspect �rm-years as �rm-years suspected to have an e�ect on reporting

frequency-induced REM. As discussed in Section 6.4, managers are likely to adapt

the level of real activities management to the situation of the company (e.g., Roy-

chowdhury, 2006). We hypothesize that the e�ect of interim reporting frequency on

REM depends on di�erent discriminating factors. We identify four factors that po-

17To test the quality of the Datastream item WC05200 provided by Datastream, we alternativelytest a dummy variable indicating if �rms have di�ering net income data for all 4 quarters available(based on item DWNP) for each respective year. Our results are not a�ected.

142

6.5 Methodology

Table 6.3: Sample overview by EU-15 countries

Country Total Obs. QRs t/o Mand. QRs t/o Vol. QRs

Austria (AT) 233 230 160 7099% 69% 30%

Belgium (BE) 435 94 0 9422% 0% 22%

Denmark (DK) 439 346 0 34679% 0% 79%

Finland (FI) 528 528 528 0100% 100% 0%

France (FR) 2,371 158 5 1537% 0% 6%

Germany (DE) 2,223 1,578 1,242 33671% 56% 15%

Great Britain (GB) 5,576 167 7 1603% 0% 3%

Greece (GR) 592 592 592 0100% 100% 0%

Ireland (IE) 261 19 0 197% 0% 7%

Italy (IT) 882 882 882 0100% 100% 0%

Luxembourg (LU) 63 32 8 2451% 13% 38%

The Netherlands (NL) 551 227 35 19241% 6% 35%

Portugal (PT) 155 155 155 0100% 100% 0%

Spain (ES) 479 479 479 0100% 100% 0%

Sweden (SE) 1517 1517 1517 0100% 100% 0%

Total 16,305 7,004 5,610 1,394100% 43% 34% 9%

Notes: This table presents a sample overview along the EU-15 countries and split into the fullsample and the suspect years sample. Quarterly reporters are de�ned as classi�ed by Datastream(WC05200, �earnings reporting frequency�). For mandatory reporting, some companies are requiredto publish quarterly reports due to listings in other countries. Percentages are relative to totaloberservations per country.

143


tentially in�uence the relation between reporting frequency and REM: (1) company

performance, (2) the level of accounting earnings management, (3) analyst following

and (4) minority shareholder protection. The rationale behind the choice of the

suspect �rm-years is explained in Section 6.4.

Company performance is measured as operating income of the respective �rm per

year. We classify �rm-year observations with below industry-year median operating

income as suspect �rm-years (Perf. Low = 1, otherwise 0). We use a relative suspect

year de�nition in order to control for other relevant external factors that in�uence

company results such as the overall economic environment.

Accounting earnings management is measured using the modi�ed cross-sectional

Jones model (Jones, 1991) as described in (Dechow et al., 1998). We estimate the

following equation for accruals quality for each industry-year:

AQ it

TAi,t−1

= α0 + α11

TAi,t−1

+ α2∆SALES it

TAi,t−1

+ α3PPE it

TAi,t−1

+ α4 ROAi,t−1 + εit (6.4)

where all variables are de�ned in Table 6.4. We de�ne �rm-year observations with

above industry-year median earnings management levels as suspect �rm-years (AEM

high = 1, otherwise 0).

Analyst following is taken from the I/B/E/S database and measures the number

of analysts for which earnings forecasts are available in the database. We de�ne

�rm-year observations with analyst following below median as suspect �rm-years

(An. Follow. Low = 1; otherwise 0).

Anti-director rights is a measure for minority shareholder protection (Djankov

et al., 2008). We classify �rm-year observations from countries with below median

anti-director rights as suspect �rm-years (Anti-DR Low = 1, otherwise 0).

The distribution of suspect �rm-years is shown in Table 6.5, Panel A.

6.5.4 Controls

In line with Roychowdhury (2006), we employ size (SIZE), book-to-market ratio

(BTM), and net income (NI) as control variables in our regressions. We argue, how-

ever, that in order to make the regression more robust and less prone to omitted

variable bias and endogeneity, we need to include further controls for �rm perfor-

mance and �rm characteristics such as leverage (LEV), tangibility (TANG, ratio of

144

6.5 Methodology

tangible assets to total assets), and �nancial slack (SLACK, ratio of cash to PPE).

We expect �rms with higher leverage to exhibit a higher degree of real earnings

management because �rms must present speci�c results not only to investors but

also creditors (e.g., in the case of covenants). The other controls can in�uence the

level of real earnings management because they re�ect in part the level of economic

pressure on the company. De�nitions of all variables are provided in Table 6.4.

6.5.5 Regressions

We use a set of panel regressions with adjusted standard errors for heteroskedasticity,

serial-, and cross-sectional correlation using a two-dimensional cluster at the �rm

and year level (as proposed by Petersen, 2009).

In order to test hypothesis H1a and H1b, we regress ACFO, APROD, and ADISC

on our key experimental variables for mandatory and voluntary quarterly reporting,

while controlling for other known factors for in�uencing real earnings management

as described above. A description of all variables can be found in Table 6.4.

We estimate the following regression model:

AbnormalMeasure it = α0+α1 MAND it+α2 VOLit+∑

αj CONTROLs it+εit (6.5)

In order to test hypothesis H2a and H2b and to separate the e�ect of mandatory

and voluntary quarterly reporting under suspect conditions (�suspect �rm-years�),

we include 4 di�erent discriminating factors (SUSPECT) and interaction terms with

VOL and MAND (VOL·SUSPECT and MAND·SUSPECT, respectively). The 4

di�erent types of suspect �rm-years are described above.

We apply the following model:

AbnormalMeasure it = α0 + α1 MAND it + α2 VOLit + α3 SUSPECT it

+α4 MAND it SUSPECT it + α5 VOLit SUSPECT it

+∑

αj CONTROLs it + εit (6.6)

We test the di�erent hypotheses using t-tests (single coe�cient) and F -tests (mul-

tiple coe�cients).

145


Table 6.4: De�nition of variables

Variable De�nition (with Datastream/Worldscope data items)

Dependent variables

ACFO Abnormal cash �ow from operations, calculated based on Equation6.1

APROD Abnormal production cost, calculated based on Equation 6.2

ADISC Abnormal discretionary expenses, calculated based on Equation 6.3

Experimental variables

MAND Mandatory quarterly reporting (Y/N), compiled based on informa-tion presented in Table 6.1

VOL Voluntary quarterly reporting (Y/N), compiled based on informa-tion presented in Table 6.1

Perf. Low Low operational performance (Y/N), with value of 1 for OPINCbelow median (by industry year group), and 0 else

Acc. EM High High accounting earnings management (Y/N), with value of 1 forAQ above median, and 0 else

An. Follow. Low Low analyst following (Y/N), with value of 1 for number of analystsfollowing below median, and 0 else

Anti-DR Low Low anti-director rights (Y/N), with value of 1 for anit-directorrights below median (by country), and 0 else

Control variables

SIZE Size, de�ned as natural logarithm of total assets (WC02999)

BTM Book-to-market value of equity, calculated as common equity(WC03501), divided by market value of equity (MVE)

NI Standardized net income, de�ned as net income available to commonequity (WC01751), divided by lagged total assets (WC02999)

LEV Accounting leverage, calculated as total current liabilities(WC03101), divided by total assets (WC02999)

TANG Asset tangibility, de�ned as PP&E (WC02501), divided by laggedtotal assets (WC02999)

SLACK Financial slack, de�ned as cash and cash equivalents (WC02001),divided by lagged total assets (WC02999)

Other variables

AQ Accruals quality measure to detect accounting earnings manage-ment, calculated as discretionary total accruals, using the modi�edcross-sectional Jones model (Jones, 1991) as described in Dechowet al. (1995)


146

6.6 Sample and descriptive statistics


Variable De�nition (with Datastream/Worldscope data items)

CFO Standardized cash �ow from operations (CFO), de�ned as CFO(WC04860), divided by lagged total assets (WC02999)

MVE Market value of equity, calculated as number of shares outstanding(nosh) times share price at �scal year end (WC05001)

OPINC Operating income in percent, de�ned as operating income(WC01250), divided by net sales (WC01001)

SALES Standardized sales, de�ned as total sales (WC01001), divided bylagged total assets (WC02999)

TA Total assets (WC02999)

Notes: This table shows de�nitions of dependent variables, experimental variables, controland other variables.

6.6 Sample and descriptive statistics

6.6.1 Sample selection

We collect a sample of all shares covered by Datastream/Worldscope and I/B/E/S

in EU-15 countries between January 2005 and December 2009, also including shares

that were de-listed during the period in order to avoid any survivorship bias. To

ensure a common �nancial reporting and disclosure setup, we start our sample col-

lection in January 2005 after the Transparency Directive was o�cially adopted in

December 2004. This has the advantage that all companies in our sample are sub-

ject to the same transparency requirements. This gives us a starting sample of 7,445

�rms.

We use Datastream/Worldscope to collect all necessary company data.18 We

require that �rms have data on total assets for at least two consecutive years from

2004 until 2009, distinct �scal year end information and unique ISIN identi�ers.

This results in 22,194 �rm-year observations from 2005 to 2009. Next, we exclude

�nancial �rms (industry 12 according to Fama-French industry classi�cation; 4,927

18Datastream/Worldscope is widely used in international studies. In their paper on the �e�ectsof database choice on international accounting research�, Lara et al. (2006) list 33 key researchpapers for international accounting, of which 15 use Datastream or Worldscope.

147


�rm-years) and �rm-years in �nancial distress (962 �rm-years). This leaves us with

16,305 �rm-year observations for our study (�full sample�). In order to mitigate the

in�uence of outliers all continuous variables are winsorized at the 1% and 99% level.

6.6.2 Descriptive statistics

Table 6.5 provides the descriptive statistics as well as correlation coe�cients. Panel

A shows the number of observations for the reporting regime and suspect �rm-

year dummy variables. Only Analyst Following Low (Y/N) has less observations

than the full sample due to missing coverage by I/B/E/S. Panel B summarizes the

continuous variables. By de�nition, the average of ACFO, APROD, and ADISC is

practically zero. ADISC is only available for a total of 1,970 �rm-year observations

given relatively low number of overlapping R&D data by the European Commission

and SG&A data from Datastream/Worldscope. Panel C presents correlations for the

measure and control variables used in the regression models. Most control variables

are highly correlated (at a 1% signi�cance level) with the REM measures (ACFO,

APROD, ADISC). This indicates the potential relevance of the controls SIZE, BTM

and NI used by Roychowdhury (2006). In addition, however, it supports us in

including further variables for �rm characteristics and performance (such as LEV,

TANG, SLACK). There is no high correlation between the explanatory variables

limiting the risk of multicolinearity in our sample. The highest correlation is between

SIZE and NI (0.31).

6.7 Results

6.7.1 Main analyses

To test whether the mandatory and voluntary quarterly reporting as well as suspect

�rm-year dummies exhibit a signi�cantly di�erent level of REM, we perform mean

equality tests on ACFO, APROD, and ADISC. Table 6.6 shows the results (ACFO

in Panel A, APROD in Panel B, ADISC in Panel C). 15 out of 18 tests performed

exceed the required signi�cance level. The tests indicate that mandatory quarterly

reporting as well as the di�erent suspect �rm-years are associated with higher REM

levels, while voluntary quarterly reporting exhibits lower REM levels across ACFO,

APROD and ADISC.

148

6.7 Results

Table 6.5: Descriptive statistics

Panel A: Descriptive Statistics of Dummy Variables

Variable Obs. Obs (0) Obs (1)

Mandatory QR (Y/N) 16,305 10,695 5,610

Voluntary QR (Y/N) 16,305 14,911 1,394

Performance Low (Y/N) 16,305 8,443 7,862

Accounting EM High (Y/N) 16,305 6,928 9,377

Analyst Follow. Low (Y/N) 8,591 5,140 3,451

Anti-Dir. Rights Low (Y/N) 16,305 6,316 9,989

Panel B: Descriptive Statistics of Continuous Variables

Variable Obs. Mean Std. Dev. 25% Median 75%

ACFO 14,282 0.000 0.135 -0.053 0.008 0.066

APROD 12,388 0.000 0.232 -0.108 0.013 0.131

ADISC 1,970 -0.001 0.195 -0.122 -0.021 0.094

SIZE 16,305 12.039 2.332 10.388 11.838 13.556

BTM 15,513 0.932 1.096 0.339 0.619 1.099

NI 14,431 0.004 0.183 -0.020 0.037 0.082

LEV 16,123 0.338 0.178 0.206 0.320 0.451

TANG 14,379 0.253 0.254 0.052 0.171 0.373

SLACK 14,438 0.190 0.251 0.044 0.105 0.232

Panel C: Pearson Correlations Matrix

Variable ACFO APROD ADISC SIZE BTM NI LEV TANG

ACFO 1

APROD -0.38*** 1

ADISC -0.10*** -0.59*** 1

SIZE 0.11*** -0.01 -0.08*** 1

BTM -0.06*** 0.10*** -0.07*** 0.03*** 1

NI 0.56*** -0.23*** -0.12*** 0.31*** -0.07*** 1

LEV -0.11*** 0.11*** 0.02 0.02** -0.09*** 0.03*** 1

TANG 0.14*** -0.04*** -0.13*** 0.27*** 0.08*** 0.11*** -0.24*** 1

SLACK 0.00 -0.05*** 0.25*** -0.25*** -0.12*** -0.13*** -0.20*** -0.16***

Notes: This table presents descriptive statistics of dummy variables (Panel A) and continuousvariables (Panel B), as well as Pearson correlation coe�cients (Panel C) for continuous variables.Variable de�nitions are in Table 6.4. All continuous variables are winsorized at 1st and 99thpercentile. *, **, *** indicates signi�cant at the 0.10, 0.05 and 0.01 levels, respectively, using atwo-tailed test.

149


We interpret these �ndings as �rst supporting evidence of hypotheses H1a, i.e.,

mandatory quarterly reporters exhibit higher REM, H1b, i.e., voluntary quarterly

reporting with lower REM, and H2a, i.e., suspect �rm-years are associated with

higher REM compared to semi-annual reporters.

In order test our hypotheses while also controlling for other e�ects, we perform

multiple regression analyses. Table 6.7 presents the results of estimating equations

(4) and (5) for ACFO. As discussed in Section 6.5, ACFO is the most important mea-

sure because it consolidates the e�ect of the most important forms of REM, i.e., sales

manipulation, overproduction and a reduction in discretionary expenses. We there-

fore focus on the interpretation of ACFO in determining REM in the main analysis

section. APROD and ADISC are discussed as supporting evidence in the additional

analyses section. We subsequently regress mandatory and voluntary quarterly re-

porting (�basic regression�, Equation 6.5) as well as suspect �rm-year interactions

(�suspect �rm-year regressions�, Equation 6.6) on ACFO. For the basic regression, we

�nd that all controls are highly signi�cant and the overall goodness-of-�t is 34.8%.

In line with H1a, we �nd that mandatory quarterly reporting exhibits signi�cantly

higher REM compared to semi-annual reporters (negative sign, t-statistics of -3.73).

For testing H1b, that voluntary quarterly reporting is associated with lower REM,

we �nd an expected positive coe�cient, which is insigni�cant, however (t-statistics

of 0.17). Given these results, we �nd empirical evidence for H1a, but not for H1b.

In order to analyze if a potential REM e�ect of quarterly reporting depends on

speci�c settings, we de�ne suspect �rm-years (suspect �rm-year regressions). In

columns two to �ve of Table 6.7, we subsequently analyze the e�ect of the di�erent

suspect �rm-years on our results. For the inference on our hypotheses H2a, H2b,

and H2c, we perform several F -tests based on the suspect �rm-year regressions in

Table 6.7, which are summarized in Table 6.8.

In order to test the validity of the suspect �rm-years chosen, we expect to �nd

a level e�ect of REM for each of the di�erent suspect �rm-year de�nitions. In ad-

dition, we include interaction terms of the di�erent suspect �rm-years with MAND

and VOL to see if the suspect years have an additional, discriminating e�ect on

REM for at least one type of quarterly reporters. The suspect �rm-year regressions

all have a higher adjusted R2 than the basic regression so the explanatory power

of the model has increased. The adjusted R2s for the suspect �rm-year regressions

including dummies and interactions for low operating performance, high accounting

150

6.7 Results

Table 6.6: Mean equality tests of abnormal cash �ows from operations (ACFO),abnormal production cost (APROD), and abnormal discretionary expenses

(ADISC)

No. of. Obs. Mean H0: Mean(0) vs. Mean(1)

Experimental variables Obs(0) Obs(1) Mean(0) Mean(1) Test t-stat. p-value

Panel A: ACFO

Mandatory QR (Y/N) 9,204 5,078 0.004 -0.005 (0) > (1) 3.89 0.00 ***

Voluntary QR (Y/N) 13,051 1,231 0.000 0.009 (0) < (1) -2.34 0.01 ***

Perf. Low (Y/N) 7,469 6,813 0.050 -0.054 (0) > (1) 49.43 0.00 ***

Acc. EM High (Y/N) 6,857 7,425 0.030 -0.027 (0) > (1) 25.94 0.00 ***

An. Foll. Low (Y/N) 5,015 3,242 0.024 -0.005 (0) > (1) 12.01 0.00 ***

Anti-DR Low (Y/N) 5,498 8,784 0.004 -0.002 (0) > (1) 2.77 0.00 ***

Panel B: APROD

Mandatory QR (Y/N) 7,898 4,490 0.000 0.000 (0) < (1) -0.07 0.47

Voluntary QR (Y/N) 11,282 1,106 0.002 -0.019 (0) > (1) 2.84 0.00 ***

Perf. Low (Y/N) 6,571 5,817 -0.058 0.066 (0) < (1) -30.82 0.00 ***

Acc. EM High (Y/N) 6,071 6,317 -0.026 0.025 (0) < (1) -12.30 0.00 ***

An. Foll. Low (Y/N) 4,732 2,887 -0.025 0.005 (0) < (1) -6.08 0.00 ***

Anti-DR Low (Y/N) 4,615 7,773 -0.052 0.031 (0) < (1) -19.69 0.00 ***

Panel C: ADISC

Mandatory QR (Y/N) 988 982 0.023 -0.025 (0) > (1) 5.51 0.00 ***

Voluntary QR (Y/N) 1,678 292 -0.010 0.049 (0) < (1) -4.80 0.00 ***

Perf. Low (Y/N) 1,211 759 -0.005 0.006 (0) > (1) -1.30 0.90

Acc. EM High (Y/N) 973 997 0.006 -0.008 (0) > (1) 1.58 0.06 *

An. Foll. Low (Y/N) 1,280 438 -0.004 0.007 (0) > (1) -1.02 0.85

Anti-DR Low (Y/N) 473 1,497 0.020 -0.008 (0) > (1) 2.72 0.00 ***

Notes: This table presents mean equality tests of abnormal cash �ows from operations (ACFO),abnormal production cost (APROD), and abnormal discretionary expenses (ADISC) on mandatoryquarterly reporting, voluntary quarterly reporting and di�erent suspect �rm-year de�nitions (Yes= 1, No = 0). Test shown is a one-sided t-test. Variables are de�ned in Table 6.4. *, **, ***indicates signi�cant at the 0.10, 0.05 and 0.01 levels, respectively, using a one-tailed test.

151


earnings management, low analyst following and low anti-director rights are 37.1%,

44.1%, 38.6% and 34.9%, respectively. Almost all controls are signi�cant at the

1% level for all suspect �rm-year regressions. Comparing semi-annual reporters in

suspect �rm-years and non-suspect �rm-years (SUSPECT vs. 0), we �nd signi�-

cantly higher REM levels across all types of suspect �rm-years on a 5% and 1%

level for low anti-director rights and all others, respectively. Also mandatory quar-

terly reporters exhibit signi�cantly higher REM levels in suspect �rm-years across

the di�erent suspect �rm-year de�nitions (SUSPECT + MAND*SUSPECT vs. 0).

As expected, also voluntary quarterly reporters consistently increase their REM

levels compared to non-suspect �rm-years. We document strong evidence in favor

of H2a, i.e., that companies exhibit higher REM levels in suspect �rm-years. We

also �nd a signi�cant additional suspect �rm-year e�ect driven by quarterly report-

ing for all suspect �rm-year de�nitions in voluntary regimes (VOL.QR*SUSPECT,

Table 6.7) and for high accounting earnings management in mandatory regimes

(MAND.QR*SUSPECT, Table 6.7) indicating that suspect �rm-years have a dis-

criminating e�ect on the relation between reporting frequency and REM and are

therefore appropriately chosen.

152

6.7 Results

Table

6.7:Regressionof

mandatory

andvoluntaryquarterlyreporting

aswellas

suspect�rm-yearinteractions

onabnorm

alcash

�owsfrom

operations

(ACFO)

BasicRegression

SuspectFirm-YearRegressions

Perf.Low

AEM

High

An.Foll.Low

Anti-DRLow

Experim

.Var.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Mand.QR(Y

/N)

--0.013

***

--0.004

-0.003

--0.006

--0.008

*

(-3.73)

(-0.93)

(0.79)

(-1.57)

(-1.77)

Vol.QR(Y

/N)

+0.001

+0.015***

+0.007**

+0.005

+0.020**

(0.17)

(2.88)

(2.53)

(1.04)

(2.47)

SY(Y

/N)

--0.041

***

--0.282

***

--0.009

***

--0.008

**

(-6.48)

(-5.04)

(-3.46)

(-2.09)

Mand.QR*SY

--0.009

--0.026

***

-0.001

--0.001

(-1.59)

(-9.30)

(0.23)

(-0.24)

Vol.QR*SY

--0.016

**-

-0.012

**-

-0.021

***

--0.018

**

(-2.31)

(-2.21)

(-3.26)

(-2.08)

Control

Variables

SIZE

-0.004

***

-0.005

***

-0.004

***

-0.003

**-0.004

***

(-4.23)

(-5.95)

(-4.02)

(-1.97)

(-4.27)

BTM

-0.004

**-0.001

-0.004

**-0.001

-0.003

**

(-2.39)

(-0.72)

(-2.54)

(-1.15)

(-2.46)

NI

0.433***

0.387***

0.453***

0.480***

0.433***

(23.21)

(23.26)

(23.46)

(14.52)

(23.05)

LEV

-0.077

***

-0.044

***

-0.115

***

-0.084

***

-0.074

***

(-12.65)

(-5.90)

(-13.20)

(-9.09)

(-11.64)

Continued

onnext

page

153


Regressionof

MAND,VOLandSY

onACFO�Continued

BasicRegression


Perf.Low

AEM

High

An.Foll.Low

Anti-DRLow

Experim

.Var.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

TANG

0.041***

0.043***

0.017***

0.048***

0.040***

(6.99)

(7.44)

(3.59)

(5.69)

(7.04)

SLACK

0.031***

0.030***

0.013***

0.055***

0.031***

(5.26)

(5.14)

(2.98)

(3.14)

(5.32)

Constant

0.065***

0.086***

0.087***

0.027

0.065***

(5.33)

(7.54)

(5.33)

(1.49)

(5.40)

Firm/yearclusters

yes

yes

yes

yes

yes

Obs.

13,843

13,843

13,461

8,202

13,843

Adjusted

R2

34.8%

37.1%

44.1%

38.6%

34.9%

Notes:

Thistable

presents

resultsfrom

regressionanalysesexaminingthee�ectofvoluntary

(VOL)andmandatory

(MAND)quarterly

reportingaswellassuspect�rm

-yearinteractions(SY)onabnorm

alcash

�ow

sfrom

operations(ACFO).TheestimationisaOLSpanel

regressionwithadjusted

standard

errors

forheteroskedasticity,

serial-,andcross-sectionalcorrelationusingatwo-dimensionalcluster

at

the�rm

andyearlevel(asproposedbyPetersen,2009).ItisbasedonEquation6.5(basicregression)andEquation6.6(suspect�rm

-year

regressions).Variablede�nitionsare

inTable6.4.Allvariablesare

winsorizedat1stand99th

percentile.Amountsreported

are

regression

coe�

cients(witht-statisticsin

parantheses).*,**,***indicatessigni�cantatthe0.10,0.05and0.01levels,respectively,usingatwo-tailed

test.

154

6.7 Results

Regarding H2b, i.e., that mandatory quarterly reporters (voluntary quarterly re-

porters) exhibit higher (lower) REM levels in suspect �rm-years (non-suspect �rm-

years), we �nd supporting evidence for the hypothesized pattern. While the manda-

tory REM di�erence compared to semi-annual reporters is very signi�cant in suspect

�rm-years (MAND + MAND*SUSPECT vs. 0) except in case of low analyst follow-

ing , we are unable to detect a signi�cant di�erence between mandatory quarterly

reporters and semi-annual reporters in non-suspect �rm-years (MAND vs. 0). This

e�ect is only slightly signi�cant for �rms from low investor protection countries (at

10% signi�cance level). In contrast, the opposite is found for voluntary quarterly re-

porters. The latter show signi�cantly lower levels of REM in non-suspect �rm-years

(VOL vs. 0) � except for low analyst following �, while there is no signi�cant di�er-

ence to semi-annual reporters in suspect �rm-years (VOL + VOL*SUSPECT vs. 0).

Overall, these results indicate that REM is especially high for mandatory quarterly

reporters in suspect �rm-years, while it is especially low for voluntary quarterly re-

porters in non-suspect �rm-years. Moreover, voluntary quarterly reporters adapt

their REM levels to the level of semi-annual reporters in suspect �rm-years.

For H2c, we test whether both types of quarterly reporters have disproportionally

increased their REM levels compared to semi-annual reporters in suspect �rm-years.

Comparing the increase of REM for mandatory quarterly reporters vis-à-vis semi-

annual reporters from non-suspect to suspect �rm-years (MAND*SUSPECT vs. 0),

we �nd a signi�cant e�ect only for high accounting earnings management �rms. In

turn, analyzing the disproportional REM increase for voluntary quarterly reporters

(VOL*SUSPECT vs. 0) indeed reveals that the di�erence is larger in suspect �rm-

years. The corresponding F -tests are all signi�cant, at least on the 5% level. Overall,

this can be interpreted as supporting evidence for our hypothesis H2c, i.e., that

quarterly reporters exhibit a disproportional increase in REM in suspect �rm-years

for voluntary quarterly reporters. The evidence for mandatory quarterly reporters

is somewhat smaller. One reason for this limited increase for mandatory quarterly

reporters might be their overall high level of REM. Increasing marginal costs for

REM might be a potential explanation for this result.

6.7.2 Additional analyses on ACFO

We perform two additional analyses on the di�erences between mandatory and vol-

untary quarterly reporters in suspect �rm-years (AA1, AA2). We summarize these

155


Table

6.8:F-testsforhypothesesH2a,H2b,andH2c

forabnorm

alcash

�owfrom

operations

(ACFO)(based

onTable6.7)

SUSPECT=

0SUSPECT=

1

F-Tests

forHypotheses

Ex.Sgn

Perf.Low

AEM

HighAn.Low

ADRLow

Perf.Low

AEM

High

An.Low

ADRLow

H2a

SYvs.

0-

-0.041

***-0.282

***

-0.009

***-0.008

**

SY+

MAND*SYvs.

0-

-0.050

***-0.308

***

-0.008

**-0.009

*

SY+

VOL*SYvs.

0-

-0.056

***-0.294

***

-0.030

***-0.026

**

H2b

MANDvs.

0-

-0.004

0.003

-0.006

-0.008

*

MAND+

MAND*SYvs.

0-

-0.013

***-0.023

***

-0.005

-0.009

***

VOLvs.

0+

0.015***0.007**

0.005

0.020**

VOL+

VOL*SYvs.

0+

-0.001

-0.005

-0.016

*0.002

H2c

MAND*SYvs.

0-

-0.009

-0.026

***

0.001

-0.001

VOL*SYvs.

0-

-0.016

**-0.012

**-0.021

***-0.018

**

Notes:

This

table

presents

F-tests

basedtheregressionsofmandatory

andvoluntary

quarterly


-year

interactionsonabnorm

alcash

�ow

from

operations(ACFO)thatcorrespondto

hypotheses

H2a,H2bandH2c.

Detailsoftheregressions

are

inTable6.7.Variablede�nitionsare

inTable6.4.Amountsreported

are

tested

coe�

cients.*,**,***indicatessigni�cantatthe0.10,

0.05and0.01levels,respectively,usinganF-test(two-sided).

156

6.7 Results

analyses in Table 6.9.

Following the above described pattern of higher REM in mandatory regimes and

lower REM in voluntary regimes compared to semi-annual reporters, we test whether

mandatory quarterly reporters consistently experience higher REM levels than vol-

untary reporters across all �rm-years (AA1 in Table 6.9). Regarding the non-suspect

�rm years (MAND vs. VOL) we �nd that the tested di�erence is negative and signif-

icantly di�erent from zero (except for high accounting earnings management). Also

in suspect �rm-years (MAND + MAND*SUSPECT vs. VOL + VOL*SUSPECT)

there is a signi�cant negative delta (except for low analyst following). We conclude

that mandatory quarterly reporters consistently (overall, in non-suspect �rm-years,

as well as in suspect �rm-years) exhibit higher REM levels than voluntary quarterly

reporters. This result is in line with expectations and recon�rms our �ndings from

H2b.

Moreover, we want to test whether the overall level e�ect in suspect �rm-years

(H2a) is stronger for voluntary than for mandatory quarterly reporters. In H2c,

we found that voluntary reporters have increased REM compared to semi-annual

reporters relatively more signi�cantly than mandatory quarterly reporters. However,

this does not allow us to conclude on the absolute REM increase. Looking at AA2 in

Table 6.9 (VOL*SUSPECT vs. MAND*SUSPECT), we �nd that the absolute delta

di�ers by suspect �rm-year. Volutary quarterly reporters have a higher increase

than mandatory reporters in a high accounting earnings management environment

(signi�cant at 1% level). Mandatory reporters exhibit signi�cantly higher marginal

REM increases compared to voluntary reporters when analyst following and anti-

director rights are low (1% and 10% signi�cance level, respectively). We conclude

that, except for a high accounting earnings management environment, mandatory

reporters have the highest absolute REM level e�ect comparing non-suspect and

suspect �rm-years.

6.7.3 Additional analyses on APROD and ADISC

ACFO is a consolidating measure and therefore determines the overall level of REM.

As supporting evidence and to better understand if REM is driven by revenue or

cost manipulation, we also want to test our hypotheses for the cost-related REM

measures APROD and ADISC individually. Regression results for APROD and

ADISC regarding Equation 6.5 and 6.6 are summarized in Tables 6.10 and 6.11,

157


Table

6.9:F-testsforadditionalanalyses

forabnorm

alcash

�owfrom

operations

(ACFO)(based

onTable6.7)

SUSPECT=

0SUSPECT=

1

F-Tests

forAdditional

AnalysesEx.Sgn

Perf.Low

AEM

HighAn.Low

ADRLow

Perf.Low

AEM

HighAn.Low

ADRLow

AA1MANDvs.

VOL

--0.018

***-0.004

-0.011

**-0.027

***

MAND+

MAND*SY

--0.012

*-0.017

***

0.010

-0.011

***

vs.

VOL+

VOL*SY

AA2VOL*SYvs.

MAND*SY

+0.006

-0.014

***

0.022***

0.017*

Notes:

This

table

presents

F-tests


andvoluntary

quarterly


-year


alcash

�ow

from

operations(ACFO)thatcorrespondto

additionalanalyses.Detailsoftheregressionsare

inTable

6.7.Variablede�nitionsare

inTablerefELV:App.Amountsreported

are

tested

coe�

cients.*,**,***indicatessigni�cantatthe0.10,0.05

and0.01levels,respectively,usinganF-test(two-sided).

158

6.7 Results

respectively.

Looking at the results for the basic regressions for APROD (�st column in Table

6.10) and ADISC (�rst column in Table 6.11), we �nd supporting evidence for H1a

and H1b. As with ACFO, for ADISC the mandatory quarterly reporting coe�cient

is signi�cantly negative with a t-statistic of -2.02 (5% signi�cance level for one-sided

test), while the voluntary coe�cient is positive on a t-statistic of 1.37, which is

signi�cant on a 10% con�dence level (one-sided t-test). APROD shows the expected

sign for voluntary quarterly reporting with a t-statistic of -2.49 (1% signi�cance

level for one-sided test). The coe�cient for mandatory quarterly is insigni�cant

(t-statistic of -0.87). Overall, also taking into account the univariate results for

APROD and ADISC from Table 6.6, the results provide supporting evidence for our

hypotheses H1a and H1b. Speci�cally, we �nd some evidence that overproduction

might be a more prevalent form of REM in voluntary regimes, whereas a reduction

of discretionary expenses seems to better explain REM in mandatory quarterly

reporting environments.

159


Table

6.10:Regressionof

mandatory

andvoluntaryquarterlyreporting

aswellas


onabnorm

alproduction

cost(A

PROD)

BasicRegression


Perf.Low

AEM

High

An.Foll.Low

Anti-DRLow

Experim

.Var.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Mand.QR(Y

/N)

+-0.006

+-0.011

+-0.017

+-0.027

***

+0.099***

(-0.87)

(-1.32)

(-1.61)

(-2.69)

(8.09)

Vol.QR(Y

/N)

--0.026

**-

-0.062

***

--0.036

***

--0.055

***

--0.046

**(-2.49)

(-4.52)

(-3.49)

(-4.08)

(-2.05)

SY(Y

/N)

+0.089***

+0.274***

+0.023**

+0.157***

(7.88)

(6.97)

(2.07)

(13.84)

Mand.QR*SY

+-0.007

+0.019*

+0.001

+-0.194

***

(-0.48)

(1.86)

(0.04)

(-11.90)

Vol.QR*SY

+0.044***

+0.022

+0.068***

+-0.051

**(2.83)

(1.43)

(3.62)

(-2.06)

Control

Variables

SIZE

0.007***

0.010***

0.008***

0.017***

0.002

(4.19)

(6.03)

(4.80)

(2.80)

(1.34)

BTM

0.018***

0.012**

0.018***

0.009***

0.017***

(2.70)

(2.10)

(2.71)

(4.05)

(3.44)

NI

-0.367

***

-0.262

***

-0.383

***

-0.383

***

-0.382

***

(-13.47)

(-11.24)

(-14.42)

(-6.68)

(-14.36)

LEV

0.155***

0.087***

0.191***

0.232***

0.105***

(5.61)

(3.74)

(6.07)

(8.77)

(4.47)

TANG

-0.011

-0.016

0.014

-0.005

0.002

(-0.85)

(-1.28)

(1.34)

(-0.32)

(0.16)

SLACK

-0.034

-0.034

*-0.014

-0.083

***

-0.038

*Continued

onnext

page

160

6.7 ResultsRegressionof

MAND,VOLandSY

onAPROD�Continued

BasicRegression


Perf.Low

AEM

High

An.Foll.Low

Anti-DRLow

Experim

.Var.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

(-1.57)

(-1.66)

(-0.73)

(-3.45)

(-1.94)

Constant

-0.140

***

-0.189

***

-0.169

***

-0.199

***

-0.129

***

(-5.48)

(-6.92)

(-6.39)

(-6.10)

(-5.86)

Firm/yearcluster

yes

yes

yes

yes

yes

Obs.

12,203

12,203

11,971

7,578

12,203

Adjusted

R2

8.3%

11.1%

10.7%

11.3%

14.3%

Notes:

Thistable

presents

resultsfrom

regressionanalysesexaminingthee�ectof

voluntary

(VOL)andmandatory

(MAND)quarterly

reportingaswellas

suspect�rm


alproductioncost

(APROD).TheestimationisaOLSpanelregression

withadjusted

standard

errors

forheteroskedasticity,serial-,andcross-sectionalcorrelation

usingatwo-dimensionalcluster

atthe�rm

and

yearlevel(asproposedbyPetersen,2009).ItisbasedonEquation6.5(basicregression)andEquation6.6(suspect�rm

-yearregressions).

Variablede�nitionsare


winsorizedat1st

and99th

percentile.Amounts

reported

are

regressioncoe�

cients

(witht-statisticsin


test.

161


Table

6.11:Regressionof

mandatory

andVoluntary

quarterlyreporting

aswellas


onabnorm

aldiscretionaryexpenses(A

DISC)

BasicRegression


Perf.Low

AEM

High

An.Foll.Low

Anti-DRLow

Experim

.Var.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Mand.QR(Y

/N)

--0.034

**-

-0.040

**-

-0.032

--0.023

--0.185

**

(-2.02)

(-2.36)

(-1.52)

(-1.45)

(-2.18)

Vol.QR(Y

/N)

+0.034

+0.033

+0.024

+0.057**

+0.088*

(1.37)

(1.27)

(1.04)

(2.37)

(1.93)

SY(Y

/N)

--0.039

**-

-0.051

-0.003

--0.013

(-2.27)

(-0.50)

(0.12)

(-0.60)

Mand.QR*SY

-0.021

--0.009

--0.035

-0.161*

(0.71)

(-0.67)

(-0.92)

(1.79)

Vol.QR*SY

-0.012

-0.023

--0.050

--0.057

(0.27)

(1.26)

(-0.63)

(-1.23)

Control

Variables

SIZE

0.002

0.001

0.001

-0.001

0.002

(0.38)

(0.25)

(0.15)

(-0.11)

(0.49)

BTM

-0.006

-0.003

-0.004

-0.003

-0.006

(-1.08)

(-0.50)

(-0.67)

(-0.70)

(-1.05)

NI

-0.056

-0.097

-0.050

-0.048

-0.061

(-0.93)

(-1.39)

(-0.78)

(-0.82)

(-1.00)

LEV

0.124***

0.138***

0.112***

0.119***

0.128***

(3.17)

(3.47)

(2.87)

(2.63)

(3.27)

TANG

-0.061

-0.058

-0.061

-0.089

**-0.058

Continued

onnext

page

162

6.7 ResultsRegressionof

MAND,VOLandSY

onADISC�Continued

BasicRegression


Perf.Low

AEM

High

An.Foll.Low

Anti-DRLow

Experim

.Var.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

Ex.Sgn.Coe�.

(-1.26)

(-1.21)

(-1.29)

(-1.97)

(-1.21)

SLACK

0.173***

0.178***

0.156***

0.176***

0.174***

(5.44)

(5.63)

(4.04)

(4.24)

(5.45)

Constant

-0.072

-0.058

-0.054

-0.006

-0.075

(-1.23)

(-1.02)

(-0.85)

(-0.08)

(-1.27)

Firm/yearcluster

yes

yes

yes

yes

yes

Obs.

1,954

1,954

1,910

1,718

1,954

Adjusted

R2

8.6%

8.9%

8.2%

10.9%

9.2%

Notes:

Thistable

presents

resultsfrom

regressionanalysesexaminingthee�ectof

voluntary

(VOL)andmandatory

(MAND)quarterly



aldiscretionary

expenses(A

DISC).

Theestimationis

aOLSpanel

regressionwithadjusted

standard

errors

forheteroskedasticity,

serial-,andcross-sectionalcorrelationusingatwo-dimensionalcluster

at

the�rm

andyearlevel(asproposedbyPetersen,2009).ItisbasedonEquation6.5(basicregression)andEquation6.6(suspect�rm

-year

regressions).Variablede�nitionsare


winsorizedat1stand99th

percentile.Amountsreported

are

regression

coe�

cients(witht-statisticsin


test.

163


Similar to ACFO, we also perform speci�c F -test to test our hypotheses H2a�H2c,

which are documented in Table 6.12 (Panel A for APROD, Panel B for ADISC).

With respect to APROD (Panel A), we �nd consistent support for REM using

overproduction in line with H2a�H2c. We want to highlight two exceptions that

might reveal interesting insights on the use of the di�erent REM methods in di�er-

ent reporting regimes. First, looking at H2b, mandatory quarterly reporters exhibit

lower REM using production manipulation compared to semi-annual reporters, espe-

cially in suspect �rm-years (the MAND + MAND*SUSPECT vs. 0 test reveals sig-

ni�cant negative signs for low performance, low analyst following, low anti-director

rights). This implies that the overall REM e�ect measured by ACFO must stem

disproportionally from other e�ects like, e.g., sales manipulation or cutting discre-

tionary expenses. One potential explanation might be that managers in mandatory

reporting regimes might �nd it harder to use production manipulation to escape the

additional burden from mandatory disclosure. They might �nd it easier to, e.g.,

engage in price discounts or a reduction in discretionary expenses. The second ex-

ception is an interesting adverse pattern of APROD for low anti-director rights. It

seems that low investor protection does a�ect APROD di�erently than the overall

measure ACFO. Similarly to the argument in mandatory regimes, a potential ex-

planation could be that given the higher concentration of power, insiders (managers

and institutional investors) are rather hesitant to increase production levels as this is

probably a rather slow and costly measure. They might also prefer to give selective

discounts to customers or cut back travel expenses. Despite these two deviations,

APROD overall provides strong additional evidence for hypotheses H2a�H2c. Fur-

thermore, they provide evidence for varying patterns in the use of di�erent REM

methods across the interim reporting regimes.

Looking at ADISC (Panel B), the inference on H2a�H2c is more di�cult due to

the lack of signi�cance as a result of the relatively low number of observations (max.

1,954). The additional dummy variables of the suspect �rm-year regressions reduce

the signi�cance levels of the individual coe�cients further. Overall, most of the

coe�cients have the expected signs. For H2a, for example, 10 out of 12 coe�cients

have the expected sign, 5 of which are signi�cant. For H2b all coe�cients have the

expected sign. Again, 5 coe�cients are signi�cantly di�erent from zero. While we

�nd some supporting evidence for the use of discretionary expense reduction in the

basic regression, we refrain from excessively interpreting our moderately signi�cant

164

6.8 Sensitivity analyses

results in the suspect �rm-year regressions. Knowing that another more general

caveat for this measure is that advertising expenses are not observable in Europe and

it only includes R&D and SG&A expenses in our case, we gain only limited insights

on the importance and pattern of discretionary expense reduction in explaining

REM.

6.8 Sensitivity analyses

In order to check the overall robustness of our inferences we perform a set of (unt-

abulated) speci�cation test on our results. We test the sensitivity of our results

towards di�erent sample periods and compositions.

First, we perform all of our tests on a smaller sample from 2007�2009 including

11,503 observations because one can argue that � while most member states started

to immediately implement the Transparency Directive after its adoption by in the

EU in late 2004 � there was a transition period until January 2007, when the Trans-

parency Directive had to be �nally enacted by all member states. As a consequence

the number of observations decreases to 8,905 �rm-years. Still, our results are not

a�ected, neither for the basic regression (Equation 6.5) nor for the suspect �rm-year

regressions (Equation 6.6) � rather on the contrary. For low analyst following the

voluntary quarterly reporters now also exhibit signi�cantly (at 5% level) lower REM

in non-suspect �rm-years.

Second, we test to exclude those �rms that stem from countries with voluntary

quarterly reporting, but are suspect to mandatory quarterly reporting due to cross-

listing on the Prime Market in Austria or Prime Standard in Germany. This does

not a�ect our results.

Third, as mentioned before our results are robust towards excluding �rms origi-

nally classi�ed as mandatory quarterly reporters based on a speci�c listing segment

requirement (Prime Market in Austria, Prime Standard in Germany). Since the de-

cision to enter the segment is voluntary in the �rst place, one could argue that the

disclosure requirements of the segment are only partly mandatory. When excluding

these �rms from our sample, our results are similar.

Fourth, in our main analyses, we add further control variables (leverage, tangi-

bility, slack) on top of the controls used in the basic regression of Roychowdhury

(2006) which turn out to be highly signi�cant in most of our regressions. Limiting

165

6 The Real Business E�ects of Quarterly ReportingTable

6.12:F-testsforhypothesesH2a,H2b,andH2c

forabnorm

alproduction

cost(A

PROD)andabnorm

aldiscretionaryexpenses(A

DISC)(based

onTables6.10

and6.11)

SUSPECT=

0SUSPECT=

1

F-Tests

forHypotheses

Ex.Sgn

Perf.Low

AEM

HighAn.Low

ADRLow

Perf.Low

AEM

HighAn.Low

ADRLow

PanelA:APROD

H2a

SYvs.

0+

0.089***

0.274***

0.023**

0.157***

SY+

MAND*SYvs.

0+

0.082***

0.292***

0.023**

-0.037

***

SY+

VOL*SYvs.

0+

0.133***

0.296***

0.090***0.107**

H2b

MANDvs.

0+

-0.011

-0.017

-0.027

***0.099***

MAND+

MAND*SYvs.

0+

-0.018

*0.001

-0.026

**-0.095

***

VOLvs.

0-

-0.062

***-0.036

***

-0.055

***-0.046

**

VOL+

VOL*SYvs.

0-

-0.017

-0.014

0.012

-0.097

***

H2c

MAND*SYvs.

0+

-0.007

0.019*

0.001

-0.194

***

VOL*SYvs.

0+

0.044***

0.022

0.068***-0.051

**

PanelB:ADISC

H2a

SYvs.

0-

-0.039

**-0.051

0.003

-0.013

SY+

MAND*SYvs.

0-

-0.018

-0.060

-0.032

0.148*

SY+

VOL*SYvs.

0-

-0.027

***-0.028

***

-0.047

***-0.070

**

H2b

MANDvs.

0-

-0.040

**-0.032

-0.023

-0.185

**

MAND+

MAND*SYvs.

0-

-0.019

-0.040

***

-0.058

-0.025

VOLvs.

0+

0.033

0.024

0.057**

0.088*

VOL+

VOL*SYvs.

0+

0.044

0.047

0.007

0.031

H2c

MAND*SYvs.

0-

0.021

-0.009

-0.035

0.161*

VOL*SYvs.

0-

0.012

0.023

-0.050

-0.057

Notes:

This

table

presents

F-tests


andvoluntary

quarterly


-year


alproductioncost

(APROD)andabnorm

aldiscretionary

expenses(A

DISC)thatcorrespondto

hypotheses

H2a,

H2bandH2c.

Detailsoftheregressionsare

inTable

6.10and6.11.Variable

de�nitionsare

inTable

6.4.Amounts

reported

are

tested

coe�

cients.*,**,***indicatessigni�cantatthe0.10,0.05and0.01levels,respectively,usingan

F-test(two-sided).

166

6.9 Conclusion

our regressions to only the controls used by Roychowdhury does not in�uence the

results. On the contrary, the explanatory power of the experimental variables in-

creases, especially for low analyst following and low anti-director rights. However,

as expected, the overall explanatory power of the model decreases by 2�4%-points.

Moreover, using an alternative SIZE de�nition (natural logarithm of the market

value of equity instead of the natural logarithm of total assets) does also not a�ect

our conclusions.

Fifth, we use Fama-French industry classi�cation to de�ne our industries and re-

quire 15 observations per industry-year in our main analyses. As another sensitivity

check, we also test our models using a 2-digit SIC code, requiring 30 observations per

industry-year. This does not change our results dramatically. While the maximum

number of observations increases slightly, the adjusted R2 of the model remains

similar. All the relevant coe�cient signs hold, only the individual signi�cance levels

decrease slightly on a few experimental variables.

6.9 Conclusion

We investigate the real business e�ects of reporting frequency in mandatory and

voluntary quarterly reporting regimes using a sample of 15 EU countries. We �nd

that mandatory quarterly reporting is associated with higher REM than semi-annual

reporting. This e�ect is particularly strong when company performance is below

industry average, accounting earnings management is high, analyst coverage is low,

and in countries with low minority shareholder protection (�suspect �rm-years�). In

contrast, voluntary quarterly reporting is associated with lower REM compared to

semi-annual reporters in �non-suspect �rm-years�. In suspect �rm-years, however,

voluntary quarterly reporters exhibit disproportionally higher and thus comparable

REM levels to semi-annual reporters. This suggests that primarily �good� �rms

increase disclosure frequency voluntarily but need to signi�cantly increase REM

levels under adverse conditions to compensate for the increased transparency.

The underlying setting in our study is unique because all included countries ex-

hibit a common minimum regulatory and institutional base that minimizes deterring

e�ects usually present in cross-country studies in international accounting (i.e., all

countries use IFRS, have comparable levels of enforcement and a common minimum

transparency standard through EU Transparency Directive). As the reporting reg-

167


ulation between countries only di�ers in few respects, the most important of which

is reporting frequency, the sample setup allows to reasonably isolate the frequency-

related e�ect on our dependent variable real earnings management.

This paper complements the existing literature by analyzing the e�ect of reporting-

frequency on real earnings management. Previous research has extensively covered

only the capital market e�ects and accounting earnings management implications

of reporting frequency so far. We also contribute to the �real e�ects� literature and

the non-U.S. focused international accounting literature as suggested by Leuz and

Wysocki (2008).

Although we �nd evidence for interim reporting frequency-induced real earnings

management, our analysis does not permit us to draw conclusions on the overall

desirability of mandatory versus voluntary reporting regimes. As we investigate the

e�ects of reporting frequency regimes solely on the �rm/manager level, there are

probably many unobserved e�ects that inhibit drawing conclusion on the macro-

level. This could be one area of potential further research.

168

Bibliography

Aders C, Wagner M. 2004. Kapitalkosten in der Bewerungspraxis: Zu hoch für die NewEconomy und zu niedrig für die Old Economy. FinanzBetrieb 6: 30�42.

African Securities Exchanges Association. 2008. Annual statistics.URL http://www.africansea.org/ASEA/Members_Annual_Statistics.aspx

Armstrong CS, Barth ME, Jagolinzer AD, Riedl EJ. 2010. Market reaction to the adoptionof IFRS in Europe. The Accounting Review 85: 31�61.

Ashbaugh H, Pincus M. 2001. Domestic accounting standards, international accountingstandards, and the predictability of earnings. Journal of Accounting Research 39: 417�434.

Ashbaugh H, War�eld TD. 2003. Audits as a corporate governance mechanism: Evidencefrom the German market. Journal of International Accounting Research 2: 1�21.

Ashbaugh-Skaife H, Collins DW, Kinney WRJ, LaFond R. 2008. The e�ect of SOX internalcontrol de�ciencies and their remediation on accrual quality. The Accounting Review83: 217�250.

Ashbough-Skaife H, Gassen J, LaFond R. 2006. Does stock price synchronicity represent�rm-speci�c information? The international evidence. Working Paper .

Auer KV. 1996. Capital market reactions to earnings announcements: Empirical evidenceon the di�erence in the information content of IAS-based earnings and EC-Directives-based earnings. European Accounting Review 5: 587�623.

Auer KV. 1998. Der Ein�uÿ des Wechsels vom Rechnungslegungsstandard auf die Risikopa-rameter von schweizerischen Aktien. Zeitschrift für betriebswirtschaftliche Forschung 50:129�155.

Baetge J, Berndt H, Bruns H, von Colbe WB, Coenenberg AG, Korst H, Lederle H, MetzeA, Ordelheide D, P�ngsten H, Reinhard H, Schnicke C, Schwitters J, Seeberg T, Siepe G,Urban P, Weber H, Weismüller A, von Wysocki K. 1995. German accounting principles:An institutionalized framework. Accounting Horizons 9: 92�99.

Ball R. 2006. International Financial Reporting Standards (IFRS): Pros and cons forinvestors. Accounting and Business Research 36: 5�27.

Ball R, Kothari SP, Robin A. 2000. The e�ect of international institutional factor scoreson properties of accounting earnings. Journal of Accounting and Economics 29: 1�51.

169

http://www.africansea.org/ASEA/Members_Annual_Statistics.aspx

Bibliography

Ball R, Robin A, Wu JS. 2003. Incentives versus standards: Properties of accountingincome in four East Asian countries. Journal of Accounting and Economics 36: 235�270.

Balsam S, Bartov E, Marquardt C. 2002. Accruals management, investor sophistication,and equity valuation: Evidence from 10-Q �lings. Journal of Accounting Research 40:987�1012.

Balz U, Bordemann H. 2007. Ermittlung von Eigenkapitalkosten zur Unternehmensbew-ertung mittelständischer Unternehmen mithilfe des CAPM. FinanzBetrieb 9: 737�743.

Banz RW. 1981. The relationship between return and market value of common stocks.Journal of Financial Economics 9: 3�18.

Barth M, Konchitchki Y, Landsman WR. 2006a. Cost of capital and �nancial statementstransparency. Working Paper .

Barth M, Landsman WR, Lang M. 2008. International accounting standards and account-ing quality. Journal of Accounting Research 46: 467�498.

Barth M, Landsman WR, Lang M, Williams CW. 2006b. Accounting quality: InternationalAccounting Standards and US GAAP. Working Paper .

Bartov E, Cohen A. 2007. Mechanisms to meet/beat analyst earnings expectations in thepre- and post-Sarbanes Oxley years. Working Paper .

Bartov E, Goldberg SR, Kim M. 2005. Comparative value relevance among German, U.S.,and International Accounting Standards: A German stock market perspective. Journalof Accounting, Auditing & Finance 20: 95�119.

Basu S. 1983. The relationship between earnings yield, market value, and return for NYSEcommon stocks: Further evidence. Journal of Financial Economics 12: 129�156.

Becker C, DeFond ML, Jiambalvo J, Subramanyam KR. 1998. The e�ect of audit qualityon earnings management. Contemporary Accounting Research 15: 1�24.

Beiker H, Steiner M. 1993. Überrenditen und Risiken kleiner Aktiengesellschaften. Einetheoretische und empirische Analyse des deutschen Kapitalmarktes von 1966 bis 1989.Köln: Botermann u. Botermann.

Bekaert G, Harvey CR, Lundblad C. 2007. Liquidity and expected returns: Lessons fromemerging markets. Review of Financial Studies 20: 1783�1831.

Belsley DA, Kuh E, Welsch RE. 1980. Regression Diagnostics. New York: Wiley.

Bens DA, Nagar V, Skinner DJ, Wong MHF. 2003. Employee stock options, EPS dilution,and stock repurchases. Journal of Accounting and Economics 36: 51�90.

170

Bibliography

Berk JG. 1995. A critique of size-related anomalies. Review of Financial Studies 8: 275�286.

Berkowitz MK, Qiu J. 2001. Common risk factors in explaining Canadian equity returns.Working Paper .

Bhandari LC. 1988. Debt/equity ratio and expected common stock returns: Empiricalevidence. Journal of Finance 43: 507�28.

Bhattacharya U, Daouk H, Welker M. 2003. The world price of earnings opacity. TheAccounting Review 78: 641�678.

Bhojraj S, Libby R. 2005. Capital market pressure, disclosure frequency-induced earnings-cash �ow con�ict, and managerial myopia. The Accounting Review 80: 1�20.

Black F, Jensen MC, Scholes M. 1972. The capital asset pricing model: Some empiricaltests. In Jensen MC (ed.) Studies in the Theory of Capital Markets. New York: Praeger.

Botosan CA. 2006. Disclosure and the cost of capital: What do we know? Accounting andBusiness Research - International Accounting Policy Forum 36: 31�40.

Botosan CA, Plumlee MA. 2002. A re-examination of disclosure level and the expectedcost of equity capital. Journal of Accounting Research 40: 21�40.

Brown LD, Caylor ML. 2005. A temporal analysis of quarterly earnings thresholds: Propen-sities and valuation consequences. The Accounting Review 80: 423�440.

Brown NC, Pott C, Wömpener A. 2008. The e�ect of internal control regulation on earningsquality: Evidence from Germany. Working Paper .

Brown P, Tarca A. 2005. A commentary on issues relating to the enforcement of Inter-national Financial Reporting Standards in the EU. European Accounting Review 14:181�212.

Burger A, Feldrappe T, Ulbrich P. 2006. Auswirkungen der Umstellung von HGB auf IFRSauf zentrale Kennzahlen des Jahresabschlusses. Praxis der internationalen Rechnungsle-gung 2: 134�141.

Burger A, Fröhlich J, Ulbrich P. 2004. Die Auswirkungen der Umstellung von HGB aufIFRS auf wesentliche Kennzahlen der externen Unternehmensrechnung. Kapitalmarkto-rientierte Rechnungslegung 4: 353�366.

Burger A, Schäfer S, Ulbrich P, Zeimes M. 2005. Die Umstellung der Rechnungslegung nachIFRS 1 - Empirische Analyse und Bewertung der Neuregelungen zur Erstanwendung derIFRS. Die Wirtschaftsprüfung 58: 1193�1200.

Burgstahler D, Dichev I. 1998. Earnings management to avoid earnings decreases andlosses. Journal of Accounting and Economics 24: 99�126.

171

Bibliography

Bushee BJ, Leuz C. 2005. Economic consequences of SEC disclosure regulation: Evidencefrom the OTC bulletin board. Journal of Accounting and Economics 39: 223�264.

Bushman RM, Piotroski JO, Smith AJ. 2004. What determines corporate transparency?Journal of Accounting Research 42: 207�252.

Butler M, Kraft A, Weiss IS. 2007. The e�ect of reporting frequency on the timeliness ofearnings: The cases of voluntary and mandatory interim reports. Journal of Accountingand Economics 43: 181�217.

Carhart MM. 1997. On persistence in mutual fund performance. Journal of Finance 52:57�82.

Chen L, Petkova R, Zhang L. 2008. The expected value premium. Journal of FinancialEconomics 87: 269�280.

Clarkson PM, Guedes J, Thompson R. 1996. On the diversi�cation, observability, andmeasurement of estimation risk. Journal of Financial and Quantitative Analysis 31:69�84.

Clarkson PM, Thompson R. 1990. Empirical estimates of beta when investors face esti-mation risk. Journal of Finance 45: 431�453.

Co�ee JC. 2005. A theory of corporate scandals: Why the U.S. and Europe di�er. OxfordReview of Economic Policy 21: 198�211.

Cohen DA, Dey A, Lys TZ. 2008. Real and accrual-based earnings management in thepre- and post-Sarbanes Oxley periods. The Accounting Review 83: 757�787.

Coles JL, Loewenstein U, Suay J. 1995. On equilibrium pricing under parameter uncer-tainty. Journal of Financial and Quantitative Analysis 30: 347�364.

Committee of European Securities Regulators. 2003. Standard no. 1 on �nancial informa-tion: Enforcement of standards on �nancial information in europe.URL http://www.europefesco.org/v2/default.asp

Conover WJ, Johnson ME, Johnson MM. 1981. A comparative study of tests for ho-mogeneity of variances, with applications to the outer continental shelf bidding data.Technometrics 23: 351�361.

Cook RD. 1979. In�uential observations in linear regression. Journal of the AmericanStatistical Association 74: 169�174.

Covrig VM, Defond ML, Hung M. 2007. Home bias, foreign mutual fund holdings, andthe voluntary adoption of International Accounting Standards. Journal of AccountingResearch 45: 41�70.

Cready WM, Gurun UG. 2007. Earnings as lagging economic indicators for market return.Working Paper .

172

http://www.europefesco.org/v2/default.asp

Bibliography

Cuijpers R, Buijink W. 2005. Voluntary adoption of non-local GAAP in the EuropeanUnion: A study of determinants and consequences. European Accounting Review 14:487�524.

Cuijpers R, Peek E. 2010. Reporting frequency, information precision and private infor-mation acquisition. Journal of Business, Finance & Accounting 37: 27�59.

Das S, Shro� PK, Zhang H. 2009. Quarterly earnings patterns and earnings management.Contemporary Accounting Research 26: 797�831.

Daske H. 2006. Economic bene�ts of adopting IFRS or US-GAAP: Have the expected costof capital really decreased? Journal of Business, Finance & Accounting 33: 329�373.

Daske H, Hail L, Leuz C, Verdi RS. 2007. Adopting a label: Heterogeneity in the economicconsequences of IFRS adoptions. Working Paper .

Daske H, Hail L, Leuz C, Verdi RS. 2008. Mandatory IFRS reporting around the world:Early evidence on the economic consequences. Journal of Accounting Research 46:1085�1142.

Daske H, Wiesenbach K. 2005. Praktische Probleme der zukunftsorientierten Schätzungvon Eigenkapitalkosten am deutschen Aktienmarkt. FinanzBetrieb 7: 407�419.

Davidson R, Goodwin-Stewart J, Kent P. 2005. Internal governance structures and earningsmanagement. Accounting and Finance 45: 241�267.

Dechow PM, Dichev ID. 2002. The quality of accruals and earnings: The role of accrualestimation errors. The Accounting Review 77: 35�59.

Dechow PM, Ge W, Larson CR, Sloan RG. 2010a. Predicting material accounting mis-statements. Contemporary Accounting Research (forthcoming) .

Dechow PM, Ge W, Schrand C. 2010b. Understanding earnings quality: A review ofthe proxies, their determinants and their consequences. Journal of Accounting andEconomics 50: 344�401.

Dechow PM, Kothari SP, Watts RL. 1998. The relation between earnings and cash �ows.Journal of Accounting and Economics 25: 133�168.

Dechow PM, Sloan R. 1991. Executive incentives and the horizon problem: An empiricalinvestigation. Journal of Accounting and Economics 14: 51�89.

Dechow PM, Sloan RG, Sweeney AP. 1995. Detecting earnings management. The Account-ing Review 70: 193�225.

Dechow PM, Sloan RG, Sweeney AP. 1996. Causes and consequences of earnings manipu-lation: An analysis of �rms subject to enforcement actions by the SEC. ContemporaryAccounting Research 13: 1�36.

173

Bibliography

DeFond ML, Hung M, Trezevant R. 2007. Investor protection and the information contentof annual earnings announcements: International evidence. Journal of Accounting andEconomics 43: 37�67.

DeFond ML, Jiambalvo J. 1991. Incidence and circumstances of accounting errors. TheAccounting Review 66: 643�655.

DeFond ML, Jiambalvo J. 1993. Factors related to auditor-client disagreements overincome-increasing accounting methods. Contemporary Accounting Research 9: 415�431.

DeFond ML, Park CW. 2001. The reversal of abnormal accruals and the market valuationof earnings surprises. The Accounting Review 76: 375�404.

Degeorge F, Patel J, Zeckhauser R. 1999. Earnings management to exceed thresholds.Journal of Business 72: 1�33.

Deloitte. 2010. IFRS: What should boards and audit committees be considering now?URL http://www.iasplus.com/dttpubs/1001ifrsauditcommittees.pdf

Delvaille P, Ebbers G, Saccon C. 2005. International �nancial reporting convergence:Evidence from three Continental European countries. Accounting in Europe 2: 137�164.

Deutsche Prüfstelle für Rechnungslegung eV. 2008. Annual activity report for the periodfrom January 1 through December 31, 2007.URL http://www.frep.info/docs/jahresberichte/2007_tb_pruefstelle.pdf

Deutsche Prüfstelle für Rechnungslegung eV. 2009. Annual activity report for the periodfrom January 1 through December 31, 2008.URL http://www.frep.info/docs/jahresberichte/2008_tb_pruefstelle.pdf

Deutsches Aktieninstitut and PricewaterhouseCoopers AG. 2009. Erfahrungen mit DPR-Prüfungen: Ergebnisse einer Umfrage unter Führungskräften im Rechnungswesen kapi-talmarktorientierter Unternehmen.URL http://www.dai.de/internet/dai/dai-2-0.nsf/dai_publikationen.htm

Deutsches Aktieninstitut (DAI). 2005. Verhalten und Präferenzen deutscher Aktionäre:Eine Befragung privater und institutioneller Anleger zu Informationsverhalten, Divi-dendenpräferenz und Wahrnehmung von Stimmrechten. DAI Study .

Diamond DW, Verrecchia RE. 1991. Disclosure, liquidity, and the cost of capital. Journalof Finance 46: 1325�1359.

Dichev ID, Skinner DJ. 2002. Large-sample evidence on the debt covenant hypothesis.Journal of Accounting Research 40: 1091�1123.

Djankov S, La Porta R, Lopez-de Silanes F, Shleifer A. 2008. The law and economics ofself-dealing. Journal of Financial Economics 88: 430�465.

174

http://www.iasplus.com/dttpubs/1001ifrsauditcommittees.pdf

http://www.frep.info/docs/jahresberichte/2007_tb_ pruefstelle.pdf

http://www.frep.info/docs/jahresberichte/2008_tb_ pruefstelle.pdf

http://www.dai.de/internet/dai/dai-2-0.nsf/dai_publikationen.htm

Bibliography

Durnev A, Kim EH. 2005. To steal or not to steal: Firm attributes, legal environment,and valuation. Journal of Finance 60: 1461�1493.

Easley DA, O'Hara M. 2004. Information and the cost of capital. Journal of Finance 59:1553�1583.

Easton PD. 2004. PE ratios, PEG ratios, and estimating the implied expected rate ofreturn on equity capital. The Accounting Review 79: 73�95.

Easton PD. 2006. Use of forecasts of earnings to estimate and compare cost of capitalacross regimes. Journal of Business, Finance & Accounting 33: 374�394.

Easton PD, Monahan SJ. 2005. An evaluation of accounting-based measures of expectedreturns. The Accounting Review 80: 501�538.

Ecker F, Francis J, Kim I, Olsson PM, Schipper K. 2006. Returns-based representation ofearnings quality. The Accounting Review 81: 749�780.

Edelstein RH, Liu P, Tsang D. 2008. Real earnings management and dividend payoutsignals: A study for U.S. real estate investment trusts. CAAA 2008 Annual ConferencePaper .

Eleswarapu VR, Thompson R, Venkatarman K. 2004. The impact of regulation fair disclo-sure: Trading costs and information asymmetry. Journal of Financial and QuantitativeAnalysis 39: 209�225.

Elsas R, Krahnen JP. 2004. Universal banks and relationships with �rms. In Krahnen JP,Schmidt RH (eds.) The German �nancial system. Oxford: Oxford University Press.

Enriques L, Volpin P. 2007. Corporate governance reforms in Continental Europe. Journalof Economic Perspectives 21: 117�140.

Ernst & Young. 2009. Interim management statements: Achieving comparability inEurope?URL http://ey.mobi/Publication/vwLUAssets/Interim_management_statements_

-_achieving_comparability_in_Europe/$FILE/EY_Interim_management_

statements.pdf

Ernstberger J, Haupt H, Vogler O. 2011. The role of sorting portfolios in asset pricingmodels. Applied Financial Economics, Forthcoming .

Ernstberger J, Hitz J, Stich M. 2010. Enforcement of accounting standards in Europe:Empirical evidence for the two-tier mechanism in Germany. Working Paper .

Ernstberger J, Vogler O. 2008. Analyzing the German accounting triad: 'Accountingpremium' for IAS/IFRS and US GAAP vis-à-vis German GAAP? The InternationalJournal of Accounting 43: 339�386.

175

http://ey.mobi/Publication/vwLUAssets/Interim_management_statements_-_achieving_comparability_in_Europe/$FILE/EY_Interim_management_statements.pdf



Bibliography

European Commission. 2002. Agreement on International Accounting Standards will helpinvestors and boost business in EU.URL http://europa.eu/rapid/pressReleasesAction.do?reference=IP/02/

827&format=HTML&aged=1&language=EN&guiLanguage=en

Ewert R, Wagenhofer A. 2005. Economic e�ects of tightening accounting standards torestrict earnings management. The Accounting Review 43: 1101�1124.

Fama EF, French KR. 1992. The cross-section of expected stock returns. Journal ofFinance 47: 427�465.

Fama EF, French KR. 1993. Common risk factors in the returns on stocks and bonds.Journal of Financial Economics 33: 3�56.

Fama EF, French KR. 1996a. Multifactor explanations of asset pricing anomalies. Journalof Finance 51: 55�84.

Fama EF, French KR. 1996b. The CAPM is wanted, dead or alive. Journal of Finance51: 1947�1958.

Fama EF, French KR. 1997. Industry cost of equity. Journal of Financial Economics 43:153�193.

Fama EF, French KR. 1998. Value versus growth: The international evidence. Journal ofFinance 53: 1975�1999.

Fama EF, French KR. 2006. The value premium and the CAPM. Journal of Finance 61:2163�2185.

Fama EF, MacBeth JD. 1973. Risk, return and equilibrium: Empirical tests. Journal ofPolitical Economy 81: 607�637.

Fan Y, Barua A, Cready WM, Thomas WB. 2010. Managing earnings using classi�cationshifting: Evidence from quarterly special items. The Accounting Review 85: 1303�1323.

FASB, IASB. 2009. Joint statement of the FASB and IASB to rea�rm commitment tomemorandum of understanding.URL http://www.ifrs.org/NR/rdonlyres/0AE63429-BCF3-461A-8B5D-3948732CDDC2/

0/JointCommunique_October2009FINAL4.pdf

Fohlin C. 2005. The history of corporate ownership and control in Germany. In MorckRK (ed.) A history of corporate governance around the world. Chicago and London:University of Chicago Press.

Francis J, LaFond R, Olsson PM, Schipper K. 2004. Costs of equity and earnings attributes.The Accounting Review 79: 967�1010.

Francis J, LaFond R, Olsson PM, Schipper K. 2005a. The market pricing of accrualsquality. Journal of Accounting and Economics 39: 295�327.

176

http://europa.eu/rapid/pressReleasesAction.do?reference=IP/02/827&format=HTML&aged=1&language=EN&guiLanguage=en

http://europa.eu/rapid/pressReleasesAction.do?reference=IP/02/827&format=HTML&aged=1&language=EN&guiLanguage=en

http://www.ifrs.org/NR/rdonlyres/0AE63429-BCF3-461A-8B5D-3948732CDDC2/0/JointCommunique_October2009FINAL4.pdf

http://www.ifrs.org/NR/rdonlyres/0AE63429-BCF3-461A-8B5D-3948732CDDC2/0/JointCommunique_October2009FINAL4.pdf

Bibliography

Francis JR, Khurana IK, Pereira R. 2005b. Disclosure incentives and e�ects on cost ofcapital around the world. The Accounting Review 80: 1125�1162.

Frankfurter GM, McGoun EG. 2001. Anomalies in Finance: What are they and what arethey good for? International Review of Financial Analysis 10: 407�429.

Franks J, Mayer C. 1997. Corporate ownership and control in the UK, Germany, andFrance. In Chew DH (ed.) Studies in international corporate �nance and governancesystems. Japan and Europe: Oxford University Press.

French KR. 2009. Data library.URL http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.

html

Fudenberg D, Tirole J. 1995. A theory of income and dividend smoothing based on incum-bency rents. Journal of Political Economy 103: 75�93.

Gassen J, Ashbaugh-Skaife H. 2009. Can audit reforms a�ect the information role of audits?Evidence from the German market. Contemporary Accounting Research 26: 867�898.

Gassen J, Sellhorn T. 2006. Applying IFRS in Germany: Determinants and consequences.Betriebswirtschaftliche Forschung und Praxis 58: 365�386.

Gebhardt W, Lee CMC, Swaminathan B. 2001. Toward an implied cost of capital. Journalof Accounting Research 39: 135�176.

Gietzmann MB, Trombetta M. 2003. Disclosure interactions: Accounting policy choiceand voluntary disclosure e�ects on the cost of raising capital. Accounting and BusinessResearch 33: 187�205.

Glaum M, Street DL. 2003. Compliance with the disclosure requirements of Germany'sNew Market: IAS versus US GAAP. Journal of International Financial Managementand Accounting 14: 64�100.

Glaum M, Thomaschewski D, Weber S. 2006. Auswirkungen des Sarbanes-Oxley Acts aufdeutsche Unternehmen: Kosten, Nutzen, Folgen für US-Börsennotierungen. Studien desDeutschen Aktieninstituts 33.

Goldman E, Slezak SL. 2006. An equilibrium model of incentive contracts in the presenceof information manipulation. Journal of Financial Economics 80: 603�626.

Goncharov I. 2005. Earnings management and its determinants: Closing gaps in empiricalaccounting research. Frankfurt/Main: Peter Lang Pub.

Goncharov I, Zimmermann J. 2006. Do accounting standards in�uence the level of earningsmanagement? Evidence from Germany. Die Unternehmung: Swiss Journal of Business,Research and Practice 61: 371�388.

177

http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html

http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html

Bibliography

Gore P, Pope P, Singh A. 2001. Non-audit services, auditor independence and earningsmanagement. Working Paper .

Graham JR, Harvey CR. 2001. The theory and practice of corporate �nance: Evidencefrom the �eld. Journal of Financial Economics 60: 187�243.

Graham JR, Harvey CR, Rajgopal S. 2005. The economic implications of corporate �nan-cial reporting. Journal of Accounting and Economics 40: 3�73.

Gri�n JM. 2002. Are the Fama and French factors global or country speci�c? Review ofFinancial Studies 15: 783�803.

Gunny K. 2005. What are the consequences of real earnings management? Working Paper.

Habib A. 2006. Information risk and the cost of capital: Review of the empirical literature.Journal of Accounting Literature 25: 127�168.

Hackethal A. 2004. German banks and banking structure. In Krahnen JP, Schmidt RH(eds.) The German �nancial system. Oxford: Oxford University Press.

Hackethal A, Schmidt RH, Tyrell M. 2005. Banks and German corporate governance: Onthe way to a capital market-based system? Corporate Governance - An InternationalReview 13: 397�407.

Hail L. 2002. The impact of voluntary corporate disclosures on the ex-ante cost of capitalfor Swiss �rms. European Accounting Review 11: 741�773.

Hail L, Leuz C. 2006. International di�erences in the cost of equity capital: Do legalinstitutions and securities regulation matter? Journal of Accounting Research 44: 485�531.

Haller A. 2002. Financial accounting developments in the European Union: Past eventsand future prospects. European Accounting Review 11: 153�190.

Haller A, Eierle B. 2004. The adaptation of German accounting rules to IFRS: A legislativebalancing act. Accounting in Europe 1: 27�50.

Haller A, Ernstberger J, Kraus C. 2006. Extraterritorial impacts of the Sarbanes-OxleyAct on external corporate governance: Current evidence from a German perspective.Corporate Ownership & Control 3: 113�127.

Haller A, Walton P. 2003. Country di�erences and harmonization. In Walton P, Haller A,Ra�ournier B (eds.) International Accounting. London: Thomson Learning.

Harris MS. 1998. The association between competition and managers' business segmentreporting decisions. Journal of Accounting Research 36: 111�128.

Harvey CR. 1991. The world price of covariance risk. Journal of Finance 46: 111�157.

178

Bibliography

Healy PM, Wahlen JM. 1999. A review of the earnings management literature and itsimplications for standard setting. Accounting Horizons 13: 365�383.

Heckman JJ. 1978. Dummy endogenous variables in simultaneous equation systems. Econo-metrica 46: 931�951.

Hofstede GH. 1984. Culture's consequences: International di�erences in work-related val-ues. Beverly Hills: Sage Publications.

Holthausen RW. 2003. Testing the relative power of accounting standards versus incentivesand other institutional features to in�uence the outcome of �nancial reporting in aninternational setting. Journal of Accounting and Economics 36: 271�283.

Holthausen RW. 2009. Accounting standards, �nancial reporting outcomes and enforce-ment. Journal of Accounting Research 47: 447�458.

Holthausen RW, Watts RL. 2001. The relevance of the value-relevance literature for �nan-cial accounting standard setting. Journal of Accounting and Economics 31: 3�75.

Hope O. 2003. Disclosure practice, enforcement of accounting standards and analystsforecast accuracy: An international study. Journal of Accounting Research 41: 235�272.

Hughes J, Liu L, Liu J. 2007. Information asymmetry, diversi�cation, and cost of capital.The Accounting Review 82: 705�729.

Hung M. 2000. Accounting standards and value relevance of �nancial statement: Aninternational analysis. Journal of Accounting and Economics 30: 401�420.

Hung M, Subramanyam KR. 2007. Financial statement e�ects of adopting InternationalAccounting Standards: The case of Germany. Review of Accounting Studies 12: 623�657.

IAS Plus. 2011. Use of IFRS by jurisdiction.URL http://www.iasplus.com/country/useias.htm

IFRS Foundation. 2011a. The move towards global standards.URL http://www.ifrs.org/Use+around+the+world/Use+around+the+world.htm

IFRS Foundation. 2011b. About the IFRS Foundation and the IASB.URL http://www.ifrs.org/The+organisation/IASCF+and+IASB.htm

Jackson HE, Roe MJ. 2009. Public and private enforcement of securities laws: Resource-based evidence. Journal of Financial Economics 93: 207�238.

Jagannathan R, Wang Z. 1996. The conditional CAPM and the cross-section of expectedreturns. Journal of Finance 51: 3�53.

179

http://www.iasplus.com/country/useias.htm

http://www.ifrs.org/Use+around+the+world/Use+around+the+world.htm

http://www.ifrs.org/The+organisation/IASCF+and+IASB.htm

Bibliography

Jain PK, Kim JC, Rezaee Z. 2006. The Sarbanes-Oxley Act of 2002 and market liquidity.Financial Review 43: 361�382.

Jegadeesh N, Titman S. 1993. Returns to buying winners and selling losers: Implicationsfor stock market e�ciency. Journal of Finance 48: 65�91.

Jensen MC. 1968. The performance of mutual funds in the period 1945�1964. Journal ofFinance 23: 389�416.

Jones JJ. 1991. Earnings management during import relief investigations. Journal ofAccounting Research 29: 193�228.

Jones KL, Krishnan GV, Melendrez KD. 2008. Do models of discretionary accruals detectactual cases of fraudulent and restated earnings? An empirical evaluation. ContemporaryAccounting Research 25: 499�531.

JSE Limited. 2004. Full compliance with International Financial Reporting Standards from1 January 2005.URL http://www.jse.co.za/docs/listings/ifrs.pdf

JSE Limited. 2008. History/residence.URL http://www.jse.co.za/history_residence.jsp

Kaufmann D, Kraay A, Mastruzzi M. 2007. Governance matters vi: Governance indicatorsfor 1996�2006. World Bank Policy Research Working Paper No. 4280 .

Kilka M, Weber M. 2000. Home bias in international stock return expectations. Journalof Psychology and Financial Markets 1: 176�192.

Kim O, Verrecchia RE. 1991. Trading volume and price reactions to public announcements.Journal of Accounting Research 29: 302�321.

Kim O, Verrecchia RE. 1994. Market liquidity and volume around earnings announcements.Journal of Accounting and Economics 17: 41�67.

Klein A. 2002. Audit committee, board of director characteristics, and earnings manage-ment. Journal of Accounting and Economics 33: 375�400.

Kothari SP, Leone AJ, Wasley CE. 2005. Performance matched discretionary accrualmeasures. Journal of Accounting and Economics 33: 163�197.

Kothari SP, Shanken J, Sloan RG. 1995. Another look at the cross-section of expectedstock returns. Journal of Finance 50: 185�224.

Küting K, Dürr U, Zwirner C. 2002. Internationalisierung der Rechnungslegung in Deutsch-land: Ausweitung durch die Unternehmen des SMAX ab 2002. KapitalmarktorientierteRechnungslegung 2: 1�13.

180

http://www.jse.co.za/docs/listings/ifrs.pdf

http://www.jse.co.za/history_residence.jsp

Bibliography

Küting K, Zwirner C. 2007. Analyse quantitativer Reinvermögense�ekte durch die Anwen-dung der IFRS. Kapitalmarktorientierte Rechnungslegung 7: 142�154.

La Porta R, Lopez-de Silanes F, Shleifer A, Vishny RW. 1997. Legal determinants ofexternal �nance. Journal of Finance 52: 1131�1150.

La Porta R, Lopez-de Silanes F, Shleifer A, Vishny RW. 2002. Investor protection andcorporate valuation. Journal of Finance 57: 1147�1170.

Lambert R, Leuz C, Verrecchia RE. 2007. Accounting information, disclosure, and the costof capital. Journal of Accounting Research 45: 385�420.

Lang LHP, Stulz RM. 1994. Tobins q, corporate diversi�cation, and �rm performance.Journal of Political Economy 102: 1248�1280.

Lang MC, Smith Raedy J, Wilson W. 2006. Earnings management and cross listing: Arereconciled earnings comparable to U.S. earnings? Journal of Accounting and Economics42: 255�283.

Lara JMG, Osma GB, Noguer BGA. 2006. E�ects of database choice on internationalaccounting research. Abacus 42: 426�454.

Lesmond DA. 2005. Liquidity of emerging markets. Journal of Financial Economics 77:411�442.

Lesmond DA, Ogden JP, Trzcinka CA. 1999. A new estimate of transaction costs. Reviewof Financial Studies 12: 1113�1141.

Leuz C. 2003a. Discussion of ADRs, analysts, and accuracy: Does cross-listing in theUnited States improve a �rm's information environment and increase market value?Journal of Accounting Research 41: 347�362.

Leuz C. 2003b. IAS versus US GAAP: Information asymmetry-based evidence from Ger-many's New Market. Journal of Accounting Research 41: 445�472.

Leuz C, Nanda D, Wysocki PD. 2003. Earnings management and investor protection: Aninternational comparison. Journal of Financial Economics 69: 505�527.

Leuz C, Verrecchia RE. 2000. The economic consequences of increased disclosure. Journalof Accounting Research 38: 91�124.

Leuz C, Verrecchia RE. 2005. Firms' capital allocation choices, information quality, andthe cost of capital. Working paper .

Leuz C, Wüstemann J. 2004. The role of accounting in the German �nancial system. InKrahnen JP, Schmidt RH (eds.) The German �nancial system. Oxford: Oxford Univer-sity Press.

181

Bibliography

Leuz C, Wysocki PD. 2008. Economic consequences of �nancial reporting and disclosureregulation: A review and suggestions for future research. Working Paper .

Liew J, Vassalou M. 2000. Can book-to-market, size and momentum be risk factors thatpredict economic growth? Journal of Financial Economics 57: 221�245.

Lintner J. 1965. The valuation of risk assets and the selection of risky investments in stockportfolios and capital budgets. Review of Economics and Statistics 47: 13�37.

Loughran T. 1997. Book-to-market across �rm size, exchange, and seasonality: Is there ane�ect? Journal of Financial and Quantitative Analysis 32: 249�268.

Maijoor S, Vanstraelen A. 2006. Earnings management within Europe: The e�ects ofmember state audit environment, audit �rm quality and international capital markets.Accounting and Business Research 36: 33�52.

Markowitz H. 1952. Portfolio selection. Journal of Finance 7: 77�91.

Markowitz H. 1959. Portfolio Selection: E�cient Diversi�cations of Investments. NewYork: John Wiley and Sons.

Mazars. 2010. Transparency Directive Assessment Report.URL http://eng.mazars.de/mazarspage/download/40806/998379/version/3/

file/Transparency_Directive_Assessment_Report.pdf

McVay SE. 2006. Earnings management using classi�cation shifting: An examination ofcore earnings and special items. The Accounting Review 81: 501�531.

Mensah YM, Werner RH. 2008. The capital market implications of the frequency of in-terim �nancial reporting: An international analysis. Review of Quantitative Financeand Accounting 31: 71�104.

Merton RC. 1987. A simple model of capital market equilibrium with incomplete informa-tion. Journal of Finance 42: 483�510.

Moya S, Oliveras E. 2006. Voluntary adoption of IFRS in Germany: A regulatory impactstudy. Corporate Ownership & Control 3: 138�147.

Nagar V, Petacchi P. 2005. An economy-level model of earnings management with endoge-nous enforcement. Working Paper .

Nagy AL. 2005. Mandatory audit �rm turnover, �nancial reporting quality, and clientbargaining power: The case of Arthur Andersen. Accounting Horizons 19: 51�68.

Naumann K. 2000. Financial reporting enforcement mechanisms as an element of corpo-rate governance in Germany and re�ections on their further developments. EuropeanAccounting Review 9: 655�672.

182

http://eng.mazars.de/mazarspage/download/40806/998379/version/3/file/Transparency_Directive_Assessment_Report.pdf

http://eng.mazars.de/mazarspage/download/40806/998379/version/3/file/Transparency_Directive_Assessment_Report.pdf

Bibliography

Nietsch M. 2005. Corporate governance and company law reform: A German perspective.Corporate Governance - An International Review 13: 368�376.

Noack U, Zetzsche D. 2005. Corporate governance reform in Germany: The second decade.European Business Law Review 16: 1033�1064.

O'Hara M. 2003. Presidential address: Liquidity and price discovery. Journal of Finance59: 1335�1354.

Page MJ, Palmer F. 1991. The relationship between excess returns, �rm size and earningson the Johannesburg Stock Exchange. South African Journal of Business Management22: 63�73.

Palmrose ZV, Richardson VJ, Scholz S. 2004. Determinants of market reactions to restate-ment announcements. Journal of Accounting and Economics 37: 59�89.

Peasnell KV, Pope PF, Young S. 2005. Board monitoring and earnings management:Do outside directors in�uence abnormal accruals? Journal of Business, Finance &Accounting 32: 1311�1346.

Pellens B, Tomaszewski C. 1999. Kapitalmarktreaktionen auf den Rechnungslegungswech-sel zu IAS bzw. US-GAAP. In Gebhardt G, Pellens B (eds.) Rechnungswesen undKapitalmarkt. Düsseldorf/Frankfurt (Main): ZfbF Special Issue 41.

Petersen MA. 2009. Estimating standard errors in �nance panel data sets: Comparingapproaches. Review of Financial Studies 22: 435�480.

Pfa� D, Schröer T. 1996. The relationship between �nancial and tax accounting in Ger-many: The authoritativeness and reverse authoritativeness principle. European Account-ing Review 5: 963�979.

Rahman AR, Tay TM, Ong BT, Cai S. 2007. Quarterly reporting in a voluntary disclosureenvironment: Its bene�ts, drawbacks and determinants. The International Journal ofAccounting 42: 416�442.

Ray SD, Savin NE, Tiwari A. 2008. Testing the momentum anomaly. Working Paper .

Roll R. 1977. A critique of the asset pricing theory's tests part i: On past and potentialtestability of the theory. Journal of Financial Economics 4: 129�176.

Ross SA. 1976. The arbitrage theory of capital asset pricing. Journal of Economic Theory13: 341�360.

Roychowdhury S. 2006. Earnings management through real activities manipulation. Jour-nal of Accounting and Economics 42: 335�370.

Sattler RR. 1994. Renditeanomalien am deutschen Aktienmarkt. Aachen: Shaker Verlag.

183

Bibliography

Schilling F. 2001. Corporate governance in Germany: The move to shareholder value.Corporate Governance - An International Review 9: 148�151.

Schipper K. 2005. The introduction of International Accounting Standards in Europe:Implications for international convergence. European Accounting Review 14: 101�126.

Schmidt R. 2004. Corporate governance in Germany: An economic perspective. In KrahnenJP, Schmidt RH (eds.) The German �nancial system. Oxford: Oxford University Press.

Schrimpf A, Schröder M, Stehle R. 2006. Evaluating conditional asset pricing models forthe German stock market. Working paper .

Securities and Exchange Commission (SEC). 2010. Commission statement in support ofconvergence and global accounting standards.URL http://www.sec.gov/rules/other/2010/33-9109.pdf

Sellhorn T, Gornik-Tomaszewski S. 2006. Implications of the IAS Regulation for researchinto the international di�erences in accounting systems. Accounting in Europe 15: 187�217.

Servaes H. 1991. Tobin's q and gains from takeovers. Journal of Finance 46: 409�419.

Sharpe WF. 1964. Capital asset prices: A theory of market equilibrium under conditionsof risk. Journal of Finance 19: 425�142.

Sloan RG. 2001. Financial accounting and corporate governance: A discussion. Journal ofAccounting and Economics 32: 335�347.

Stattman D. 1980. Book values and stock returns. The Chicago MBA: A Journal ofSelected Papers 4: 25�45.

Stein JC. 1996. Rational capital budgeting in an irrational world. Journal of Business 69:429�455.

Sweeney AP. 1994. Debt-covenant violations and managers' accounting responses. Journalof Accounting and Economics 17: 281�308.

Teoh SH, Wong TJ, Rao GR. 1998. Are accruals during initial public o�erings opportunis-tic? Review of Accounting Studies 3: 175�208.

Tesar LL, Werner IM. 1995. Home bias and high turnover. Journal of International Moneyand Finance 14: 467�492.

Timmreck C. 2002. Beta-Faktoren: Anwendungsprobleme und Lösungsansätze. Finanz-Betrieb 4: 300�307.

Tweedie D. 2011. Speech by IASB Chairman: The future of �nancial reporting �Convergence or not?URL http://www.ifrs.org/News/Announcements+and+Speeches/convergence+or+

not+speech.htm

184

http://www.sec.gov/rules/other/2010/33-9109.pdf

http://www.ifrs.org/News/Announcements+and+Speeches/convergence+or+not+speech.htm

http://www.ifrs.org/News/Announcements+and+Speeches/convergence+or+not+speech.htm

Bibliography

Vafeas N. 2005. Audit committees, boards, and the quality of reported earnings. Contem-porary Accounting Research 22: 1093�1122.

Van der Meulen S, Gaeremynck A, Willekens M. 2007. Attribute di�erences betweenU.S. GAAP and IFRS earnings: An exploratory study. The International Journal ofAccounting 42: 123�142.

Van Rensburg P. 2001. A decomposition of style based risk on the JSE. Investment AnalysisJournal 54: 45�60.

Van Rensburg P, Roberts M. 2003. Style variables and the cross-section of JSE returns.Investment Analysis Journal 57: 7�15.

Van Tendeloo B, Vanstraelen A. 2005. Earnings management under German GAAP versusIFRS. European Accounting Review 14: 155�180.

Verrecchia RE. 2001. Essays on disclosure. Journal of Accounting and Economics 32:97�180.

Vitols S. 2005. Changes in Germany's bank-based �nancial system: Implications for cor-porate governance. Corporate Governance - An International Review 13: 386�396.

Wallmeier M. 2000. Determinanten erwarteter Renditen am deutschen Aktienmarkt -Eine empirische Untersuchung anhand ausgewählter Kennzahlen. Zeitschrift für betrieb-swirtschaftliche Forschung 52: 27�57.

Webster E, Thornton DB. 2005. Earnings quality under rules- versus principle-based ac-counting standards: A test of the Skinner hypothesis. Canadian Accounting Perspectives4: 167�192.

Weiÿenberger BE, Stahl AB, Vorstius S. 2004. Changing from German GAAP to IFRS orUS GAAP: A survey of German companies. Accounting in Europe 1: 169�189.

Welker M. 1995. Disclosure policy, information asymmetry, and liquidity in equity markets.Contemporary Accounting Research 11: 801�827.

Wooldridge JM. 2002. Introductory Econometrics. A Modern Approach. Mason: ItpsThomson Learning, 2nd ed. edition.

World Federation of Exchanges. 2008. Statistics on domestic market capitalization.URL http://www.world-exchanges.org/WFE/home.asp?menu=426&document=4770

Wüstemann J. 2001. Ökonomische Theorie gesetzlicher Informationsprinzipien. Tübingen:Mohr Siebeck.

Wurgler J. 2000. Financial markets and the allocation of capital. Journal of FinancialEconomics 58: 187�214.

185

http://www.world-exchanges.org/WFE/home.asp?menu=426&document=4770

Bibliography

Young S. 1999. Systematic measurement error in the estimation of discretionary accruals:An evaluation of alternative modelling procedures. Journal of Business, Finance &Accounting 26: 833�862.

Yu F. 2008. Analyst coverage and earnings management. Journal of Financial Economics88: 245�271.

Zang AZ. 2007. Evidence on the tradeo� between real manipulation and accrual manipu-lation. Working Paper .

Ziegler A, Schröder M, Schulz A, Stehle R. 2007. Multifaktormodelle zur Erklärungdeutscher Aktienrenditen: Eine empirische Analyse. Schmalenbachs Zeitschrift für be-triebswirtschaftliche Forschung 59: 356�390.

186

Oliver Vogler

Dietrich-Bonhoeffer-Straße 6 85567 Grafing bei München

Telefon: +49-80 92-230 2929 E-Mail: [email protected]

Educational Qualifications

Since Apr'09 Ruhr-University, Bochum Doctoral student, Chair for Accounting and Auditing, Prof. Dr. Jürgen Ernstberger

Sep'07 – Dec'07 Humboldt University, Berlin Guest researcher at the Collaborative Research Center 649 ("Economic Risk") of the German Research Foundation (Deutsche Forschungsgemeinschaft, DFG)

Aug’06 – Dec’06 American University, Washington D.C. (USA) “Washington Semester Program“ in International Business and Trade

Oct’05 – Jan'08 University of Regensburg Graduate studies in the Honors Degree Program in Business Administration (Elite Network of Bavaria); Degree: Dipl.-Kfm. (1.5)

Mar’05 – Nov’06 Bavarian Elite-Academy, Munich 4-semester complementary studies for outstanding university students to get qualified for executive tasks in business, finance and economy

Oct’03 – Sep’05 Friedrich-Alexander-University, Erlangen-Nuremberg Undergraduate studies in "International Business Administration“; Degree: Vordiplom (1.9)

Sep’97– Jun’02 Gymnasium, Neutraubling Degree: Abitur (1.2)

Work Experience

Since May’11 OSRAM GmbH, Munich Strategic Project Manager, Strategy-World

Mar'08 – Apr’11 McKinsey&Company, Munich Member of the Corporate Finance Practice (Fellow Senior Associate) Educational leave from April'10 to April’11

Aug'07 – Jan'08 OSRAM Opto Semiconductors, Regensburg Working student, Strategic Marketing and Planning (part time)

Feb'07 – Apr'07 McKinsey&Company, Munich Internship (Fellow Intern)

Apr’03 – Sep’03 BSH Bosch and Siemens Home Appliances GmbH, Regensburg Internship, Project Controlling

Academic Honors

Aug’06 – Dec’06 German Academic Exchange Service (DAAD), Bonn Scholarship for international studies in Washington D.C. (USA)

Feb'05 Fulbright Commission, Berlin Fulbright scholarship for studies in the USA (not exercised)

Apr’04 – Jan'08 German National Academic Foundation (Studienstiftung des Dt. Volkes), Bonn General scholarship

Refereed Articles

"Economic Consequences of Accounting Enforcement Reforms: The Case of Germany" with Jürgen Ernstberger and Michael Stich. European Accounting Review, Conditionally Accepted.

"The role of sorting portfolios in asset pricing models" with Jürgen Ernstberger and Harry Haupt. Applied Financial Economics, Forthcoming 2011.

"Das Fama-French Modell: Eine Alternative zum CAPM - auch in Deutschland". FinanzBetrieb, 11 (07/08), 382-8, July 2009.

"Analyzing the German Accounting Triad – ‘Accounting Premium’ for IFRS and US GAAP vis-à-vis German GAAP?" with Jürgen Ernstberger. The International Journal of Accounting, 43 (4), 339-86, December 2008.

Reply to the Discussion of "Analyzing the German Accounting Triad – ‘Accounting Premium’ for IFRS and US GAAP vis-à-vis German GAAP?" with Jürgen Ernstberger. The International Journal of Accounting, 43 (4), 394-7, December 2008.

"The Value and Accounting Premium for South African Listed Shares" with Jürgen Ernstberger and Christian Heinze. Journal of Economic and Financial Sciences, 2 (2), 187-202, October 2008.

Working Papers

"The Real Business Effects of Quarterly Reporting" with Jürgen Ernstberger and Benedikt Link. July 2011.

Other Publications

"Adapting or Adopting IFRS? – Weltweite IFRS Implementierung weiterhin fraglich" with Benedikt Link. ZfCM Controlling & Management, 2011 (55), 209-210, Summer 2011.

"LED lighting at the crossroads: country road or expressway?" with Florian Wunderlich and Dominik Wee. LEDs Magazine, 2010 (11/12), 31-34, November 2010.

"Betreuung von Studierenden - Ein Werkstattbericht über Beteiligung, Befähigung, Integration und Employability" with Susanne Esslinger. Hochschulmanagement, 4 (3), 77-83, November 2009.

"Module development lights up the future LED value-chain" with Michael Viertler and Dominik Wee. LEDs Magazine, 2009 (9/10), 19-22, September 2009.

Invited Presentations

American Accounting Association, IAS Mid-Year Meeting, Tampa, 2011: “The Real Business Effects of Quarterly Reporting” (Best Paper Award)

European Accounting Association, 31st Annual Congress, Rotterdam, 2008: “Analyzing the German Accounting Triad – ‘Accounting Premium’ for IAS/IFRS and US GAAP vis-à-vis German GAAP?” (Best Paper Award)

Illinois Int’l Accounting Symposium, Honolulu, 2007: “Analyzing the German Accounting Triad with an Enhanced Multifactor Model – ‘Accounting Premium’ for IAS/IFRS and US GAAP vs. German GAAP”

Awards

Best Paper Award in the 2011 International Accounting Section Mid-Year Conference, American Accounting Association, Tampa, 2011

ERIM Best Paper Award in the category International Financial Accounting, European Accounting Association, 31st Annual Congress, Rotterdam, 2008

Grafing, July 10, 2011

Documents

International financial reporting : economic … · Economic Consequences of IFRS Adoption, Enforcement Reforms, and Interim Reporting ... 4 2 The German Accounting riadT