Applying IFRS in Germany

Determinants and Consequences

Joachim Gassen
Wirtschaftswissenschaftliche Fakultt
Humboldt-Universitt zu Berlin
10099 Berlin, Germany
Phone: +49(0)30-2093 5764
E-mail: gassen@wiwi.hu-berlin.de

Thorsten Sellhorn
Fakultt fr Wirtschaftswissenschaft
Ruhr-Universitt Bochum
Universittsstrae 150
44780 Bochum, Germany
Phone: +49(0)234-32 28300
E-mail: sellhorn@iur.rub.de

First Version, June 2006

This Version, July 2006
Forthcoming in: Betriebswirtschaftliche Forschung und Praxis, 58 (4), 2006.
We thank Karsten Detert for some of the data used in this study.

Applying IFRS in Germany

Determinants and Consequences

Abstract: We address three research questions motivated by the recent ascent of International Financial Reporting Standards (IFRS) in Europe. First, analyzing the determinants
of voluntary IFRS adoption by publicly traded German firms during the period 19982004, we find that size, international exposure, dispersion of ownership, and recent IPOs
are important drivers. Second, using the results from this determinant model to construct
propensity score-matched samples of IFRS and German-GAAP (HGB) firms, we document significant differences in terms of earnings quality: IFRS firms have more persistent, less predictable and more conditionally conservative earnings. Third, analyzing information asymmetry differences between IFRS and HGB firms, we show that IFRS
adopters experience a decline in bid-ask spread of 70 base points and an average of 17
more days with price changes per year. On the other hand, IFRS adopters stock prices
seem to be more volatile. In the light of some important limitations of our study, we discuss IFRS-related research opportunities in post-2005 Europe.
Keywords: IFRS, earnings quality, earnings attributes, information asymmetry, standard
setting, IAS Regulation, Europe, propensity-score matching, voluntary adoption

Introduction
The European Commissions IAS Regulation (EC 2002) made International Fi-

nancial Reporting Standards (IFRS) mandatory for the consolidated financial statements
of firms publicly traded in the European Economic Area from 2005 onwards.1 It aims at
contributing to the efficient and cost-effective functioning of the capital market by ensuring a high level of transparency and comparability of financial reporting. In several countries, however, firms have already been applying IFRS voluntarily before 2005, frequently in compliance with stock exchange regulations. In Germany, for example, several
global players adopted internationally accepted accounting systems, i.e. IFRS and U.S.
Generally Accepted Accounting Principles (U.S. GAAP), as early as 1993.2 Since these
firms were required to continue reporting under national GAAP, costly parallel and
dual accounting resulted.3 In response to the ensuing demand pull for international
accounting, the 1998 Capital Raising Facilitation Act (KapitalaufnahmeerleichterungsgesetzKapAEG) allowed publicly traded parent companies to substitute consolidated
IFRS or U.S. GAAP accounts for HGB (Handelsgesetzbuch German Commercial
Code, the main source of German GAAP) group accounts.4 Six other European countries
(Austria, Belgium, Finland, France, Italy, and Luxemburg) introduced similar legislation.5 In Germany, IFRS or U.S. GAAP were required of firms listed in the now-defunct
1

2
3

The European Economic Area includes the European Union member states as well as Norway, Iceland, and Liechtenstein. Since January 1, 1994, most Single Market legislation applies to all of them.
At the time, IFRS were known as International Accounting Standards (IAS).
Under dual reporting, firms prepared consolidated accounts that simultaneously fulfilled the requirements of German, EU Directives-based GAAP and internationally accepted standards. Under
parallel reporting, firms prepared consolidated IFRS/U.S. GAAP statements in addition to GermanGAAP reports. See Haller (2002, p. 168).
However, contrary to the initial bill dated June 7, 1996, the German legislator eventually shied away
from also opening individual financial statements to IFRS and U.S. GAAP.
The application decrees, however, have never been adopted in France and Italy (Delvaille/Ebbers/Saccon, 2005, pp. 147, 158).

growth-stock segment Neuer Markt, launched in March 1997. The small-cap segment
SMAX adopted this requirement in 2001. In 2003, the quality segment Prime Standard
inherited the IFRS/U.S. GAAP requirement. These developments resulted in many of
Germanys public firms preparing consolidated IFRS accounts before 2005.
Considering the European Commissions goals in introducing the IAS Regulation,
three broad research questions arise: First, does adoption of IFRS rules ensure a high
level of transparency and comparability? If so, assessing the extent to which highquality, comparable financial reporting contributes to the efficient and cost-effective
functioning of the capital market begets two further questions: Second, how does financial reporting quality influence information asymmetry? And third, how does information
asymmetry relate to stock prices, e.g. to the firms cost of equity capital?
With respect to the first question, research is required that assesses the quality of
financial reporting outcomes under IFRS and its determinants. Financial reporting quality
is an elusive concept. Since accounting serves different purposes, it seems futile to attempt a comprehensive definition of financial reporting quality. Besides serving as a contracting device, the function of financial reporting is to reduce information asymmetries
between market participants. Therefore, most of the academic accounting literature
adopts a decision-usefulness perspective to financial reporting quality,6 which is consis-

In this paper, we follow prior literature in focusing on the decision-usefulness role of financial reporting. This does not imply, however, that we view its contracting role as less important. See Watts
(2006) for a recent discussion of this topic. The theoretical concept of economic income is an alternative, neutral benchmark against which to assess the outcomes of a given financial reporting system
(e.g., Beaver, 1998, p. 3). However, due to its lack of operationality and observability (Schipper/Vincent, 2003), it cannot be used in empirical research, but is rather employed as a frame of reference in prescriptive-normative accounting research.

tent with the IASBs and FASBs stated objectives of financial reporting standards.7 This
paper addresses the connection between accounting standards and the quality of accounting outcomes. According to the SEC, there is a strong link between high-quality standards and high-quality accounting outcomes: The standards must be of high quality
they must result in comparability and transparency, and they must provide for full disclosure. Investors must be able to meaningfully analyze performance across time periods and
among companies (Levitt 1998, p. 81).8 On the other hand, quality as well as other
financial reporting properties9 may not be determined solely by the financial reporting
standards followed. Rather, since standards provide discretion, financial reporting quality
is also shaped by the incentives of those preparing (and auditing) the financial information.10 Consequently, the relative role of standards versus incentives in determining financial reporting properties, including quality, is one of the most fundamental questions
in accounting research.
Second, financial reporting quality, especially the quality and quantity of disclosure (e.g., Leuz/Verrecchia, 2000), is expected to be negatively associated with information asymmetry. To the extent that IFRS reporting increases financial reporting quality,

10

Refer to the IASBs Framework for the Preparation and Presentation of Financial Statements, par. 12,
and the FASBs Concepts Statements No. 1, par. 32-54, and No. 2, par. 30-32. The U.S. Securities and
Exchange Commission (SEC) concurs with these views (Levitt, 1998, p. 81).
Pownall/Schipper (1998) review earlier research related to the SECs consideration of IFRS and to
these three criteria.
While we view comparability as another important financial reporting property, we do not further
address it in this paper. We do, however, discuss in section 4 research opportunities for addressing the
comparability dimension of IFRS reporting.
Accordingly, Ball/Robin/Wu (2003, p. 236) view the focus on standards as substantially and misleadingly incomplete, because financial reporting practice under a given set of standards is sensitive to the
incentives of the managers and auditors responsible for financial statement preparation. Following
this reasoning, installing uniform, high-quality financial reporting standards (de jure harmonization) is
unlikely to be sufficient for achieving comparable, uniformly high-quality financial reporting outcomes (de facto harmonization). Ewert/Wagenhofer (2005) show that tighter accounting standards increase earnings quality, but that this effect is potentially outweighed by detrimental effects of earnings
management activity.

information asymmetry should decrease under IFRS. This should be reflected in measures of information asymmetry such as bid-ask spread, share turnover, stock return volatility, analyst forecast dispersion, and IPO underpricing. The third question calls for theoretical links between information asymmetry and investors resource allocation decisions.
Such links have been provided by, for example, Easley/OHara (2004), OHara (2003),
and Lambert/Leuz/Verrecchia (2006). Several empirical papers including Easley/Hvidkjaer/OHara (2002) and Francis et al. (2005) show associations between different proxies for information asymmetry and the cost of equity capital.
In this paper we condense and extend prior findings regarding the effect of IFRS
on the German market by addressing the first two questions presented above. Despite the
popular notion that high quality standards lower the cost of capital (Levitt, 1998, p.
82),11 we do not analyze the effects of IFRS adoption, an accounting phenomenon, on the
cost of capital, an asset pricing phenomenon, since we believe that this relation is not yet
well understood (e.g., Ewert, 1999). When linking financial reporting quality to the cost
of capital, we view the cost of capital as a second-order effect, following information
asymmetry, the first-order effect. Rather, we initially estimate a determinant model of the
IFRS adoption decision for German firms which voluntarily adopted IFRS over the period 1998 to 2004. Using the results from this model, we construct a propensity scorematched sample of control firms in order to address the problem of self-selection bias.
We then compare selected earnings attributes of these two samples over time in order to
test whether the change to IFRS altered the quality of the financial reporting information.

11

A survey conducted by Pellens/Tomaszewski (1999, p. 203) documents that, at the time, 45% of German IFRS or U.S. GAAP adopters expected lower costs of capital. Daske (2006) reports mixed empirical evidence on this issue.

Finally, we address the impact of IFRS adoption on the equity market by investigating
measures of information asymmetry for both samples.
This paper continues as follows: In the second section, we discuss the theoretical
background and selected prior research findings related to the three research questions we
address. In the third section, we provide evidence for the German market in the period
1998-2004 on the determinants of IFRS adoption, on the differences in earnings quality
between IFRS and German GAAP (HGB), and on the information asymmetry effects of
IFRS versus German GAAP reporting. The fourth section concludes and indicates future
IFRS-related research opportunities in post-2005 Europe.

2
2.1

Extant research on the determinants and consequences of IFRS adoption

Determinants of IFRS adoption
Our first analysis elucidates the determinants of voluntary IFRS adoption. In order

to understand the market forces behind the internationalization of financial reporting,

several previous papers have analyzed why firms voluntarily choose to apply international financial reporting standards instead of, or in addition to, their local GAAP. The
general hypothesis is that firms adopt non-local GAAP because it is beneficial for them to
do so. Examining the factors associated with 211 firms listed on the London Stock Exchange in 1993 voluntarily applying IFRS or U.S. GAAP, Ashbaugh (2001) identifies the
number of foreign listings and the higher degree of disclosure and information standardization relative to local GAAP as the main drivers. With regard to the choice between
IFRS and U.S. GAAP, her analysis shows that firms are more likely to adopt IFRS when
participating in seasoned equity offerings and when IFRS are less costly to implement
because they require fewer accounting policy changes from local GAAP relative to U.S.
7

GAAP. The adoption of U.S. GAAP is mainly driven by the SEC filing requirements
faced by firms listed on a U.S. stock exchange. In a more recent study, Cuijpers/Buijink
(2005) address a similar research question using 133 non-financial IFRS or U.S. GAAP
adopters listed and domiciled in the EU. They focus on the year 1999 because more recent accounting choices may be influenced by the anticipation of mandatory IFRS reporting from 2005 onwards. Similar to Ashbaugh (2001), the authors find that foreign listing
(in the U.S. and/or on the EASDAQ exchange in Brussels) is an important driver. Furthermore, the geographical dispersion of a firms operations is positively associated with
non-local-GAAP adoption, while financial reporting quality under local GAAP negatively affects that choice. Unsurprisingly, where national regulation allows non-local
GAAP as an alternative to local GAAP, firms are more likely to adopt them due to lower
costs.12

2.2

Earnings quality
In order to assess whether different accounting standards lead to differences in fi-

nancial reporting quality, a definition of financial reporting quality is required. In line

with prior literature, we focus on the narrower concept of earnings quality, since earnings
is a summary measure on which many users of financial reporting information rely.
Schipper/Vincent (2003) distinguish, inter alia, between two groups of earnings quality
constructs derived from (1) the time-series properties of earnings and (2) the role of accruals in transforming cash flows into earnings, respectively:13 First, the time series of

12

13

Other studies on the determinants of IFRS adoption include Dumontier/Raffournier (1998), ElGazzar/Finn/Jacob (1999), Murphy (1999), and Renders/Gaeremynck (2005).
Much of the following is based on Schipper/Vincent (2003, pp. 99-103). Francis et al. (2004) define
and apply most of Schipper/Vincents (2003) concepts in the context of an empirical study, providing
an overview of the related literature.

earnings is reflected in persistence, predictability, and variability. Perfectly persistent

(sustainable) earnings (e) follow a random walk (et = et-1 + ), i.e. earnings innovations
(shocks) tend to remain in the earnings series permanently. Therefore, present realizations of earnings are assumed to have large implications for future earnings realizations.
Predictability implies that past financial reporting outcomes improve users ability to
forecast items of interest.14 In the context of earnings, past earnings help predict future
earnings. In some contexts, accounting conservatism has been argued to be a quality attribute of earnings, which serves important economic functions (e.g., Watts, 2003).
Ball/Kothari/Robin (2000, p. 2) argue that timeliness and conservatism together capture
much of the commonly used concept of financial statement transparency. Consequently,
empirical measures of conservatism have been used to operationalize financial reporting
quality (e.g. Ball/Robin/Wu, 2003).15
Second, earnings quality constructs are also derived from the accruals and cash
components of earnings.16 Measures such as the ratio of cash from operations to earnings
are based on the view that the accruals process itself decreases earnings quality by reducing the closeness-to-cash of the earnings number. Other measures try to separate a specific, discretionary or unexpected subset of accruals from normal accruals, where the
former is viewed as detrimental to earnings quality. This group of earnings quality measures includes the change in total accruals, which assumes that the normal portion is constant over time, and discretionary accruals estimated using the Jones (1991) model and its
extensions, which assume that the discretionary portion is a linear function of accounting
14

15

16

Bricker et al.s (1995) content analysis of 479 sell-side analyst reports suggests that analysts assess
earnings quality in terms of near-term predictability, focusing on some measure of core earnings.
However, Holthausen (2003, p. 279) questions whether conservatism, e.g. the degree observed in the
U.S., should be equated with a level of quality desirable for other countries.
See Schipper/Vincent (2003, pp. 101-103).

fundamentals such as revenues. A related metric developed by Dechow/Dichev (2002)

models accruals as a function of past, present and future cash flow from operations, interpreting firm-specific accruals estimation errors as inverse measures of earnings quality.
A third set of quality measures adopts a standard-setting perspective: Since standard setters are interested in empirical research to underpin their decisions, researchers
have attempted to operationalize earnings quality in terms of relevance and reliability
(Schipper/Vincent, 2003, pp. 103-104). Value relevance research is a case in point, where
the overall explanatory power and coefficient estimates from regressions of returns/stock
prices on earnings and other accounting measures are interpreted as evidence of the combined relevance/reliability of the financial reporting information (e.g. Barth/Beaver/Landsman, 2001). However, the standard-setting implications of value relevance research are heavily debated (e.g. Holthausen/Watts, 2001). Empirically found value relevance is difficult to interpret because the measure does not support conclusions about
causality. Also, value relevance relies on market data, which makes this metric susceptible to differences in market efficiency.
Given our suggested operationalization of earnings quality: Why would we expect
earnings quality to differ between IFRS and HGB firms? Since earnings quality is a function of accounting standards, institutional environment and incentives, we first have to
assume that, ceteris paribus, accounting standards quality has an impact on earnings
quality. Although we believe accounting standards quality to be unobservable (since there
is no objective, interpersonal rule to judge the quality of standards with), we propose
three different reasons for quality differences between IFRS and German GAAP: (1) the
fundamental character of the standards, (2) the objectives of the standards, and (3) the

10

standard-setting process. First, since the German HGB is more strongly principles-based
than IFRS, and since empirical results seem to indicate that principles-based standards
produce better earnings quality (e.g. Webster/Thornton, 2005), we posit the assumption
that HGB standards result in higher earnings quality. Second, although consolidated financial statements under both IFRS and German GAAP are geared exclusively towards
decision usefulness (e.g., Leuz, 2003, p. 453; 297 HGB), German-GAAP group accounts are usually derived from individual accounts, which in Germany are the basis for
taxation and legal dividend restrictions. Since the prudence principle is dominant in German accounting rules ( 252 HGB), German GAAP group accounts might in some cases
reflect prudent accounting to the extent that the consolidation process did not revoke prudent recognition and measurement policies (e.g., Sellhorn/Gornik-Tomaszewski, 2006).
This coexistence of multiple purposes in German GAAP can be argued to decrease earnings quality relative to the single-purpose IFRS. Third, German GAAP are established in
the public-sector political-legal system, while IFRS are promulgated by a private-sector
standard-setting body. It has been argued that procedural properties of the rule-making
process influence the unobservable quality of the resulting rules and, thus, earnings quality (e.g., Pellens/Flbier/Gassen, 2006, pp. 24-28). To the extent that the IASBs due
process is superior to the German legislative procedure in terms of equal involvement of
all stakeholders, neutrality of the standard-setting body, presence of a measurable objective derived from an agreed-upon, consistent conceptual framework, and monitoring and
enforcement of the principles of fair standard-setting by an independent authority, IFRS
can be expected to be higher-quality than German GAAP.

11

Prior research already investigated the impact of accounting standards on earnings

quality. While some studies on international differences in earnings quality either analyze
samples pooled across countries with different standards and different institutional environments (e.g., Ball/Kothari/Robin, 2000; Leuz/Nanda/Wysocki, 2003), others examine
the joint effect of a low-quality institutional environment combined with a high-quality
set of standards (e.g., Ball/Robin/Wu, 2003). Our setting is comparable to Leuz (2003)
and Bartov/Goldberg/Kim (2005) in that we examine the effects of different sets of standards followed by two groups of firms that face very similar institutional environments
and, thus, financial reporting incentives. Similarly, van Tendeloo/Vanstraelen (2005) and
Goncharov (2005) recently took advantage of the German setting to show that German
IFRS firms do not engage significantly less in earnings management than their HGB
counterparts.

2.3

Information asymmetry
Theory suggests that higher financial reporting quality reduces the dispersion of

information (1) between managers and outside investors and (2) between potential buyers
and sellers of shares (e.g., Leuz/Verrecchia, 2000).17 Since IFRS adoption can be viewed
as a voluntary commitment to increased disclosure (e.g., Leuz/Verrecchia, 2000; Cuijpers/Buijink, 2005), we test whether our IFRS and HGB samples differ significantly on
selected dimensions of information asymmetry. Theory suggests that information asymmetry is positively associated with the cost of capital (e.g. Easley/OHara, 2004; Lambert/Leuz/Verrecchia, 2006).

17

While Leuz/Verrecchia (2000, p. 92) actually refer to the notion of increased levels of disclosure to
be interpreted as either an increase in the quantity of disclosure or an increase in the quality of disclosure (or both), we argue that an increase in earnings quality should have qualitatively the same effect.

12

Whether or not this relation holds in practice is an empirical question (e.g., Ewert,
1999). Consistent with the notion that information asymmetry differences are a first-order
effect of differences in financial reporting, Leuz (2003) exploits a unique setting that presented itself at the German Neuer Markt. Comparing information asymmetry measures
between two otherwise homogeneous sets of firms reporting under either IFRS or U.S.
GAAP, he shows that differences between IFRS and U.S. GAAP firms in terms of bidask spread, share turnover and other information asymmetry proxies are statistically insignificant and economically small. Examining the effects of voluntary non-local GAAP
adoption by EU firms, Cuijpers/Buijink (2005) fail to find significant benefits in terms of
the cost of equity capital and stock return volatility. However, they document a positive
effect of non-local GAAP adoption on analyst following. Results for analyst forecast dispersion differ between early and late adopters.

Evidence on the determinants and consequences of IFRS adoption

In this section, we provide year-specific evidence for German firms in the period

1998-2004 on the determinants of IFRS adoption, on the differences in earnings quality

between IFRS and German GAAP reports, and on the effects of IFRS versus HGB reporting on information asymmetry.

3.1

Sample selection and determinants of IFRS adoption

In order to assess the extent and impact of IFRS adoption, we use various sources

to obtain time-series information on the accounting standards followed by German public

firms: First, for the early adopters, we rely on the data provided by Bonse (2004). Second,
we use data about the Neuer Markt presented by Kting (2001). Third, we extensively

13

hand-collect data, especially for the periods 2000 onwards. Finally, where we are unable
to otherwise obtain data, we rely on the Worldscope item Accounting Standards Followed.
[Insert Table 1 around here]
Table 1 shows the distribution of the application of HGB, IFRS and U.S. GAAP
over time. We code dual accounting as HGB accounting, because in this case, compliance
with IFRS is limited by conflicting HGB requirements. In line with prior research, we
document a monotonic incline of IFRS application over time, with rapid growth around
the enactment of the KapAEG. The number of U.S. GAAP adopters also increases during
the 1990s, but later decreases as the IAS Regulation was discussed and finally adopted.
For our further analysis, we focus on the period 1998-2004. In 1998, publicly traded
German corporations were allowed to substitute IFRS consolidated financial statements
for group accounts prepared under German GAAP for the first time. Previously, firms
were forced to incur the costs of dual or parallel reporting, which implies that early and
late (with respect to the year 1998) adopters differ systematically. In addition, IFRS
sample sizes in the years prior to 1998 are too small for the annual regressions we conduct. As we focus on voluntary adopters, we do not include the year 2005. Given the
sample period 1998-2004, our sample comprises purely voluntary adopters along with
firms required to apply IFRS or U.S. GAAP under stock exchange regulations. Since the
latter firms could have chosen a market segment not requiring international standards at
all or could have chosen U.S. GAAP over IFRS, we view them as voluntary IFRS
adopters nonetheless. We also include firms listed on the now-defunct Neuer Markt.
Excluding them does not materially alter our results.

14

We start our sample selection procedure with the universe of German public firms
covered by Worldscope as well as by Thompson Financial Datastream. The values of all
non-truncated variables are winsorized at the bottom and top percentiles. Since we analyze IFRS adoption, we eliminate all U.S. GAAP firms. We also exclude financial institutions because of the fundamental differences in their financial accounting relative to nonfinancial firms. We further limit our sample to observations for which sufficient data is
available to carry out all of our analyses. These cuts substantially reduce our sample size
compared to the numbers in Table 1 and bias our sample towards the larger and more
visible firms on which the international accounting debate concentrates. The resulting
sample size and properties are reported in Table 2.
[Insert Table 2 around here]
The descriptive statistics in Panel A and B reveal substantial differences between
IFRS firms and HGB firms: Not only do fewer IFRS firms operate in the manufacturing
sector; they are also larger, have a higher proportion of foreign sales, have more diverse
ownership structures, are listed on more stock exchanges, have a higher propensity of
being listed on the U.S. capital market, and more frequently went public in 1995 or later.
Since our sample firms self-select into applying IFRS, they can be expected to be
systematically different from other firms which do not choose to voluntarily comply with
IFRS. To address the resulting selection bias (see also Leuz, 2003; Leuz/Verrecchia,
2000), we apply a propensity-score matching procedure, which can be viewed as a similar
but alternative approach to the two-stage Heckman procedure applied by prior literature.18

18

The propensity-score matching technique is a well-accepted matching methodology in the natural and
social sciences. Refer to Rosenbaum/Rubin (1983) for the seminal paper and to Smith/Todd (2001) for
a discussion of the methodological and conceptual differences between the propensity score matching
technique and the two-stage Heckman approach.

15

We construct propensity score-matched samples of firms that do and do not adopt IFRS,
respectively, for each year of the sample period: Each IFRS firm is matched by a HGB
firm which, according to the results of estimating determinant model (1) in Panel C of
Table 2, has the same propensity to adopt IFRS but chose not to. In order to include only
close matches in our sample, we impose a caliper of 0.1, meaning that the differences in
propensity scores between matched IFRS and HGB firms are below 10%. Using this
technique, we ensure that differences in the subsequent analyses are not caused by differences underlying the decision to voluntarily adopt IFRS. The matching procedure thus
controls for relevant sample differences other than the accounting standards followed,
allowing for a test of the relative role of rules versus incentives in determining financial
reporting outcomes. In effect, we construct our treatment (IFRS firms) and control samples (HGB firms) so that, conceptually, we can expect the matched firms to differ only in
terms of the accounting standards followed and of the effects of this difference.
Since this matching technique depends crucially on our determinant model capturing all aspects of the voluntary IFRS adoption by German firms, we rely on the results of
prior research (presented in section 2) for constructing model (1).19 The results of the
annual estimations complement the results of prior research: Younger, larger U.S-listed
firms with a larger proportion of foreign sales and more dispersed ownership tend to have
a higher propensity of adopting IFRS.
[Insert Table 3 around here]

19

We apply a set of robustness tests to ensure our model captures the IFRS decision: We include more
industry intercepts, different measures of financing needs or liquidity constraints, and alternative
measures of capital market access. We find our results to be stable. The alternative specifications of
model (1) do not qualitatively influence our subsequent results.

16

The propensity score-matched sample constructed based on the estimates of

model (1) is presented in Table 3. Panel A reports the distribution of our sample across
time and accounting standards followed. Carrying out our matching procedure separately
for each year ensures that for each year every treatment (IFRS) firm is matched by a control (HGB) firm of the same year. It is important to note that the resulting samples are no
longer representative of all HGB and IFRS firms. Comparing the average market capitalizations of the matched HGB sample in Table 3 to the original HGB sample in Table 2 we
find that the former firms are on average (median) roughly 50% larger than the latter. In
contrast, the firms in the propensity score-matched IFRS sample are smaller on average
than the original IFRS sample firms in Table 2.20 Thus, we can draw only relative, not
absolute inferences from the analyses conducted in the following sections. Panel B and C
of Table 3 compare the two propensity score-matched samples along the variables required for the analyses in section 3.2. Generally, there are few differences across the two
samples. Interestingly, however, the only significant difference IFRS firms report lower
earnings than comparable HGB firms is a direct result of the accounting system followed.

3.2

Earnings quality effects of IFRS adoption

In defining our measures of earnings quality, we rely on prior literature. Mainly

following Francis et al. (2004), we use accrual quality, persistence, predictability, value
relevance and conservatism as our measures of earnings quality. The Dechow/Dichev
(2002) measure of accrual quality is derived from the mapping of current working capital

20

To verify the effectiveness of the propensity score-matching technique, we retest the propensity scorematched sample with regard to the univariate variable differences documented in Panel B of Table 2
and find them to be insignificant (not tabulated).

17

accruals into last-period, current-period, and next-period cash flows,21 as indicated by

model (2) presented in Panel A of Table 4. Intuitively, this regression model tries to capture the expected accruals given a specific cash flow stream. The residuals from estimating model (2) capture unexpected working capital accruals and, thus, are a measure of
accrual quality: The lower the absolute level of unexpected working capital accruals, the
higher accrual quality. Although Dechow/Dichev (2002, pp. 43-44) expect firm-specific
estimations of their model to be superior, they also report within-industry and pooled
results. While we agree with Dechow and Dichevs argument, we use pooled estimation
since our time series and samples are too small to estimate firm-specific or industryspecific models. Therefore, our results must be interpreted with caution. The results of
the regressions presented in Panel A are as predicted by theory and generally very similar
to the results presented in prior research: Working capital accruals are positively correlated with prior- and subsequent-period operating cash flows and negatively correlated
with current-period cash flows. The sample differences are evaluated in Panel E of Table
4 along with the other earnings attributes. The results for our measure of accrual quality,
the absolute of the residual of model (2) deflated by its dependent variable,22 are higher
for IFRS, indicating that the accrual quality of IFRS firms is lower than that of comparable HGB firms. However, the differences are insignificant (two-sided significance for
mean: 0.445; median: 0.201).
[Insert Table 4 around here]
21

22

Dechow/Dichevs (2002, p. 37) model of accruals focuses on working capital accruals because cash
flow realizations related to working capital generally occur within one year, making both the theory
and the empirics more tractable.
We use this concept throughout this section whenever we investigate differences in the goodness of fit
across samples. It yields very similar results compared to investigating the standard deviation of the
residuals directly (e.g. Francis et al., 2004, p. 980) and is much less sensitive to scale differences
caused by sample differences in the dependent variable. See Gu (2004) for an excellent discussion of
this topic.

18

Persistence is the slope coefficient from a regression of current earnings on lagged

earnings (model (3)). Given that persistent earnings tend to follow a random walk process, the coefficient of fully persistent earnings should be one. The results of model (3) are
presented in Panel B of Table 4. The coefficient on NIBE_TAt-1 is larger for the IFRS
sample. As Panel E shows based on an interacted version of model (3) (not tabulated),
this difference is significant. Thus, we conclude that IFRS firms tend to have more persistent earnings than comparable HGB firms.
Model (3) also yields information about the predictability of earnings. Earnings
are predictable if they are persistent and exhibit low earnings shocks. Since earnings
shocks are captured as residuals in model (3), the absolute deflated level of residuals is
our measure of predictability: The lower the absolute deflated residuals, the higher the
predictability of earnings. Panel E shows that IFRS firms earnings tend to be less predictable since the median of the absolute deflated residuals is significantly larger for
IFRS firms. The difference in the mean is insignificant (two-sided significance: 0.144).
Model (4) represents a model of value relevance comparable to, among others,
Francis/Schipper (1999), as it investigates the proportion of stock returns which can be
explained by earnings information. We use 15-month returns in order to ensure that the
earnings information, some of which is published after fiscal year end, is incorporated
into stock returns. Panel C reports the results. Again, all estimates are as economically
expected, with the R of the value-relevance regression being somewhat higher for the
IFRS sample. Since it is useless and misleading to compare R values across samples
(Gu, 2004), we again compare the absolute deflated residuals (Panel E). As the residuals
capture the proportion of stock returns not explained by accounting data, higher levels of

19

absolute residuals indicate lower levels of value relevance. The residuals are lower for the
IFRS sample than for the matched HGB sample, indicating that the value relevance of
IFRS firms is higher. However, these differences are insignificant (two-sided, mean:
0.884; median: 0.452).23
Our last measure of earnings quality, conservatism, is based on the asymmetric
timeliness approach pioneered by Basu (1997). It is measured as the difference between
the slope coefficient on negative returns and the slope coefficient on positive returns in a
reverse regression of earnings on returns (model (5)). The results are reported in Panel
D. The good news coefficient (2) is lower for IFRS firms and the bad news coefficient (3) is higher, indicating that IFRS firms exhibit more conditionally conservative
earnings than HGB firms. Panel E reports the differences (3 - 2) for the two samples.
We assess the significance of the difference in this metric between the two samples by
estimating a fully interacted version of model (6) for the pooled sample (not tabulated).
The significance is estimated by an F-Test, which tests whether the difference of the interacted versions of 3 and 2 is above zero. We find the difference to be significant, but
only marginally so (two-sided significance: 0.097).
Taken together, this section provides evidence that, in a comparable sample of
German IFRS and HGB firms, IFRS firms tend to have more persistent, less predictable
and more conservative earnings than their HGB counterparts. Assuming that, first, our
determinant model of IFRS adoption successfully captures the systematic differences
between IFRS and HGB firms prior to IFRS adoption and, second, our earnings attributes
capture the opaque concept of earnings quality, the question arises how to weight the dif23

Following a different methodology, Bartov/Goldberg/Kim (2005) find a significant higher value relevance of IFRS earnings for the time-period 1998-2000.

20

ferent earnings attributes in our combined judgment of earnings quality. From a reliability
and relevance perspective, low absolute earnings shocks (high predictability) are not necessarily an indicator of good earnings quality, since earnings shocks are the only relevant
component of persistent earnings. We thus conclude this section, based on the assumptions made above and notwithstanding further caveats presented in section 4, by cautiously stating that our results provide some limited evidence that the adoption of IFRS
has had a positive impact on the earnings quality of German firms. The next section investigates whether this earnings quality improvement has also had an impact on the information asymmetry in the German capital market.

3.3

Information asymmetry effects of IFRS adoption

As in the previous section, the following set of tests also uses the propensity

score-matched sample developed in section 3.1. Following Leuz/Verrecchia (2000), we

adopt the bid-ask spread, share turnover, and share price volatility as proxies for information asymmetry. In addition, we incorporate the percentage of trading days in a fiscal year
with no price change as an additional measure of information asymmetry, since recent
research by Ashbaugh-Skaife/Gassen/LaFond (2006) shows that this metric robustly captures the level of information impounded into prices for a large body of international
markets. Bid-ask spread is possibly the most direct measure of information asymmetry
because the diverse expectations of potential buyers and sellers are reflected in the range
of prices offered and asked. Trading volume captures the liquidity of a firms shares, with
the willingness to trade being inversely related to the information asymmetry among
market participants. The argument regarding share price volatility can go both ways: On
the one hand, idiosyncratic (unsystematic) price changes can be viewed as an indicator of

21

firm-specific information being impounded into prices. On the other hand, research
shows that share price volatility tends to be too high to be explained by firm fundamentals alone. Thus, if noise is one potential explanation for share price volatility, volatility
can be expected to increase as information asymmetry increases.
[Insert Table 5 around here]
The explanatory variables in models (6) to (9) are as those used in
Leuz/Verrecchia (2000). They are aimed at capturing influential factors other than accounting standards followed. Their definitions are given in the notes to Table 5. In line
with Leuz/Verrecchia (2000, p. 110), we include BETA as an additional control in model
(9) because we expect the volatility of a share to be influenced by its systematic risk.
Concerning our measure of bid-ask spread, we note that intraday bid-ask spreads are generally better suited than closing-price bid-ask spreads to capture the level of information
asymmetry, especially since thin trading often occurs in the late trading hours. However,
we use average closing price bid-ask spreads because we do not have access to a sufficiently large sample of intraday bid-ask spread data. The values of the dependent as well
as independent variables are compared across the samples in Panel A and B of Table 5.
Besides bid-ask-spread, all variables exhibit significant differences across the samples:
The shares of IFRS firms have more turnover, more frequent price changes, more volatile
prices and carry more systematic risk than do the shares of comparable HGB firms.
The results of the models are presented in Panel C of Table 5. The signs and magnitudes of the other explanatory variables are as theoretically predicted and generally
significant. The estimation quality of the models is good, indicating a reasonable fit with
the data. Turning to our main variable of interest, the dummy indicating IFRS adoption,

22

we find that IFRS adoption seems to lead to lower bid-ask spreads, more frequent price
changes and more volatility, while there is no significant impact on share turnover. The
differences appear to be economically meaningful as well: They indicate a bid-ask spread
reduction by 70 base points as well as increased liquidity reflected in 17 more trading
days with price changes per year.
Again, these results crucially depend on the validity of model (1), which tries to
capture the determinants of IFRS adoption. If we assume, in contrast to prior literature,
that firms with higher levels of information asymmetry have an incentive not to adopt
IFRS, possibly because the high levels of information asymmetry are the result of an
opaque yet efficient corporate disclosure strategy not captured by the other determinants
in model (1), this might be an alternative explanation for our findings in models (6) to
(9). In addition, as index membership and IFRS adoption are closely related in our German sample, we cannot rule out index membership as an alternative explanation. In order
to further test our results, we construct an alternative sample based on a pre and post
analysis. The sample contains only firms which switch from HGB to IFRS and for which
we have three years of prior and three years of subsequent data. This yields a small sample of 29 firms (172 observations). Comparing our measures of information asymmetry
across time for these firms (not tabulated), we are able to replicate our findings for the
frequency of price changes and volatility, but not for the bid-ask spread. However, pre
and post research designs have fundamental problems, since it is nearly impossible to
control for all changes occurring over time. Therefore, and due to the severely reduced
sample size, we view the propensity score-matched results as our main results for our
information asymmetry analysis. We conclude from this section that the supposedly

23

higher quality of IFRS accounting has had a slight impact on the information asymmetry
related to firms adopting IFRS, as we document a decline in the average bid-ask spread
by 70 base points and 17 more price change days for these firms. However, this decrease
in information asymmetry comes with a price tag: IFRS adopters also experience a significant increase in share price volatility.

Conclusion and future research opportunities in post-2005 Europe

We conclude from our study that the voluntary adoption of IFRS for German

firms is influenced by size, international exposure, and dispersion of ownership. In addition, adopting IFRS was especially attractive for young firms which initially went public
subsequent to the mid-1990s. Comparing a sub-sample of these voluntary IFRS adopters
with comparable German-GAAP (HGB) firms, we find IFRS firms to have more persistent, less predictable and more conditionally conservative earnings. Taken together, we
view these results as indicating that the earnings of IFRS firms are of higher quality. We
also find that IFRS adopters experience lower levels of information asymmetry on the
German equity market relative to their German counterparts: They experience a decline
in bid-ask spread of 70 base points and an average of 17 more days with price changes.
On the other hand, more information also seems to induce more volatility: We find the
level of share price volatility to be significantly higher for IFRS firms.
However, these results come with some important caveats: First, the concept of
earnings quality does not capture accounting quality per se, let alone disclosure quality.
Given its theoretical underpinnings, it gives more weight to the decision-usefulness view
of accounting than it does to the contracting view of accounting. Thus, concluding that
IFRS earnings are of higher quality does not necessarily imply that they are a more effi-

24

cient contracting device. Second, as every other correction scheme for selection bias we
are aware of, our propensity score-matched research design presumes that the determinant model of IFRS adoption captures all relevant variables. If our determinant model
suffers from omitted variables, these variables might present alternative explanations for
our findings regarding earnings quality and information asymmetry. Third, our measures
of earnings attributes and our proxies for information asymmetry are noisy and potentially biased. While noise reduces the power of our tests and thus works against finding
significant differences, bias might render our results useless. Finally, even if our results
are internally valid, their external validity might still be questioned: We investigate IFRS
over a time period in which substantial changes concerning the contents of IFRS, the institutional infrastructure of IFRS standard setting, and the institutional infrastructure surrounding the German capital market and its participants occurred. All these changes
might substantially limit the generalizability of our findings to other time periods and
other institutional infrastructures.
Given the limitations of this study, the need for future research is obvious. In contrast to the time period studied in this paper, the post-IAS Regulation situation in Europe
represents a nearly ideal setting for analyzing the relative roles of standards and institutional environments in determining financial reporting quality and other properties of
financial reporting (Schipper, 2005; Nobes, 2005). This new setting allows researchers to
examine financial reports prepared under identical, high-quality standards by firms facing
vastly different regulatory and firm-specific incentives. Therefore, besides the aspect of
financial reporting quality, it also lends itself to assessing the extent to which the IAS

25

Regulation was successful at bringing about greater comparability of financial reporting

across Europe.
Not unlike previous research, the dimensions along which post-2005 European
IFRS reports will need to be analyzed and compared internationally include accounting
choice, compliance, financial reporting quality, and the impact of financial reporting on
information asymmetry and on the cost of debt and equity capital. Any remaining international variation under harmonized IFRS reporting is likely to be driven by country-level
differences in the degree of investor protection, securities regulation, legal system, financing system, corporate governance, auditing, and enforcement. Furthermore, firmlevel determinants including ownership structure, size, industry, investment opportunity
set, and other incentives will also be influential factors.
However, while the new setting has obvious appeal, to the extent that mandatory
IFRS reporting is shaped by the transitional choices of first-time adopters as well as by
learning-curve effects of financial statement preparers and users, empirical findings in the
initial years may be difficult to interpret and may not be representative of what IFRS reporting will some day look like. Also, the continuing evolution of IFRS (including their
convergence with U.S. GAAP) and capital market regulation will inhibit the development
of a stable European, let alone global, reporting environment in the foreseeable future.
While this may represent bad news to some constituents, it certainly is good news for
researchers: The supply of captivating research questions is unlikely to decline soon.

26

References
Ashbaugh, H. (2001), Non-US Firms Accounting Standard Choices, Journal of Accounting and Public Policy, Vol. 20, pp. 129-153.
Ashbaugh-Skaife, H./Gassen, J./LaFond, R. (2006), Does Stock Price Synchronicity Represent Firm-Specific Information? The International Evidence, MIT Sloan Research
Paper 4551-05.
Ball, R./Kothari, S. P./Robin, A. (2000), The Effect of International Institutional Factors
on Properties of Accounting Earnings, in: Journal of Accounting and Economics, Vol.
29, pp. 1-51.
Ball, R./Robin, A./Wu, J. S. (2003), Incentives versus Standards: Properties of Accounting
Income in Four East Asian Countries, in: Journal of Accounting and Economics, Vol.
36, pp. 235-270.
Barth, M. E./Beaver, W. H./Landsman, W. R. (2001), The Relevance of the ValueRelevance Literature for Financial Accounting Standard Setting: Another View, in:
Journal of Accounting and Economics, Vol. 31, pp. 77-104.
Bartov, E./Goldberg, S. R./Kim, M. (2005), Comparative Value Relevance Among German, U.S., and International Accounting Standards: A German Market Perspective, in:
Journal of Accounting, Auditing and Finance, Vol. 20, pp. 95119.
Basu, S. (1997), The conservatism Principle and the Asymmetric Timeliness of Earnings,
in: Journal of Accounting and Economics, Vol. 24, pp. 3-37.
Beaver, W. H. (1998), Financial Reporting An Accounting Revolution, 3rd ed., Upper
Saddle River/New Jersey.
Bonse, A. (2004), Informationsgehalt von Konzernabschlssen nach HGB, IAS und USGAAP Eine empirische Analyse aus Sicht der Eigenkapitalgeber, Frankfurt (Main)
et al.
Bricker, R./Previts, G./Robinson, T./Young, S. (1995), Financial Analyst Assessment of
Company Earnings Quality, in: Journal of Accounting, Auditing & Finance, Vol. 10,
pp. 541-554.
Cuijpers, R./Buijink, W. (2005), Voluntary Adoption of Non-Local GAAP in the European Union: A Study of Determinants and Consequences, in: European Accounting
Review, Vol. 14, pp. 487-524.
Daske, H. (2006), Economic Benefits of Adopting IFRS or US-GAAP Have the Expected Cost of Equity Capital Really Decreased?, in: Journal of Business Finance and
Accounting, Vol. 33, pp. 329373.
Dechow, P. M./Dichev, I. D. (2002), The Quality of Accruals and Earnings The Role of
Accrual Estimation Errors, in: The Accounting Review, Vol. 77 (Supplement), pp. 3559.
Delvaille, P./Ebbers, G./Saccon, C. (2005), International Financial Reporting Convergence: Evidence from three Continental European Countries, in: Accounting in
Europe, Vol. 2, pp. 137-164.

27

Dumontier P./Raffournier, B. (1998), Why Firms Comply Voluntarily with IAS: An Empirical Analysis with Swiss Data, in: Journal of International Financial Management &
Accounting, Vol. 9, pp. 216-245.
Easley, D./OHara, M. (2004), Information and the Cost of Capital, in: Journal of Finance, Vol. 59, pp. 1553-1583.
Easley, D./Hvidkjaer, S./OHara, M. (2002), Is Information Risk a Determinant of Asset
Returns?, in: Journal of Finance, Vol. 57, pp. 2185-2221.
El-Gazzar, S. M./Finn, P. M./Jacob, R. (1999), An Empirical Investigation of Multinational Firms' Compliance with International Accounting Standards, in: The International Journal of Accounting, Vol. 34, pp. 239-248.
European Commission (EC) (2002), Regulation (EC) No 1606/2002 of the European Parliament and of the Council of 19 July 2002 on the Application of International Accounting Standards, Official Journal L 243, 11/09/2002, pp. 1-4.
Ewert, R. (1999), Rechnungslegung, Globalisierung und Kapitalkosten, in: Kostenrechnungspraxis, Special Issue 3/1999, pp. 39-46.
Ewert, R./Wagenhofer, A. (2005), Economic Effects of Tightening Accounting Standards
to Restrict Earnings Management, in: The Accounting Review, Vol. 80, pp. 11011124.
Francis, J./LaFond, R./Olsson, P. M./Schipper, K. (2004), Cost of Equity and Earnings
Attributes, in: The Accounting Review, Vol. 79, pp. 967-1010.
Francis, J./LaFond, R./Olsson, P. M./Schipper, K. (2005), The Market Pricing of Accruals Quality, in: Journal of Accounting and Economics, Vol. 39, pp. 295-327.
Francis, J./Schipper, K. (1999), Have Financial Statements Lost their Relevance? in:
Journal of Accounting Research, Vol. 37, pp. 319-352.
Goncharov, I. (2005), Earnings Management and Its Determinants: Closing Gaps in Empirical Accounting Research. Frankfurt am Main et al.
Gu, Z. (2004), Across-sample Incomparability of R2s and Additional Evidence on Value
Relevance Changes Over Time, Working Paper, Carnegie Mellon University.
Haller, A. (2002), Financial Accounting Developments in the EU: Past Events and Future
Prospects, in: The European Accounting Review, Vol. 11, pp. 153-190.
Holthausen, R. W./Watts, R. L. (2001), The Relevance of the Value-Relevance Literature
for Financial Accounting Standard Setting, in: Journal of Accounting and Economics,
Vol. 31, pp. 3-75.
Holthausen, R.W. (2003), Testing the Relative Power of Accounting Standards versus
Incentives and Other Institutional Features to Influence the Outcome of Financial Reporting in an International Setting, in: Journal of Accounting and Economics, Vol. 36,
pp. 271-283.
Jones, J. J. (1991), Earnings Management During Import Relief Investigations, in: Journal of Accounting Research, Vol. 29, pp. 193-228.
Kting, K. (2001), Bilanzierung und Bilanzanalyse am Neuen Markt, Stuttgart.

28

Lambert, R./Leuz, C./Verrecchia, R. E. (2006), Accounting Information, Disclosure, and

the Cost of Capital, Working Paper, University of Pennsylvania.
Leuz, C. (2003), IAS versus US GAAP: Information-Asymmetry Based Evidence from
Germanys New Market, in: Journal of Accounting Research, Vol. 41, pp. 445-472.
Leuz, C./Verrecchia, R. E. (2000), The Economic Consequences of Increased Disclosure,
in: Journal of Accounting Research, Vol. 38 (Supplement), pp. 91-124.
Leuz, C./Nanda, D./Wysocki, P. D. (2003), Earnings Management and Investor Protection: An International Comparison, in: Journal of Financial Economics, Vol. 69, pp.
505-527.
Levitt, A. (1998), The Importance of High Quality Accounting Standards, in: Accounting
Horizons, Vol. 12, pp. 79-82.
Murphy, A. B. (1999), Firm Characteristics of Swiss Companies that Utilize International
Accounting Standards, in: International Journal of Accounting, Vol. 35, pp.121-131.
Nobes, C. W. (2005), The Survival of International Differences under IFRS: Towards a
Research Agenda, Working Paper, University of Reading.
OHara, M. (2003), Liquidity and Price Discovery, in: Journal of Finance, Vol. 58, pp.
1335-1354.
Pellens, B./Flbier, R. U./Gassen, J. (2006), Internationale Rechnungslegung, 6th ed.,
Stuttgart.
Pellens, B./Tomaszewski, C. (1999), Kapitalmarktreaktionen auf den Rechnungslegungswechsel zu IAS bzw. US-GAAP, in: Gebhardt, G./Pellens, B. (eds.), Rechnungswesen
und Kapitalmarkt, ZfbF Special Issue 41, Dsseldorf/Frankfurt (Main), pp. 199-228.
Pownall, G./Schipper, K. (1999), Implications of Accounting Research for the SECs
Consideration of International Accounting Standards for U.S. Security Offerings, in:
Accounting Horizons, Vol. 13, pp. 259-280.
Renders, A,/Gaeremynck, A. (2005), The Impact of Legal and Voluntary Investor Protection on the Early Adoption of IFRS, Working Paper, Catholic University of Leuven.
Rosenbaum, P. R./Rubin, D. B. (1983), The Central Role of the Propensity Score in Observational Studies for Causal Effects, in: Biometrica, Vol. 70, pp. 41-55.
Schipper, K. (2005), The Introduction of International Accounting Standards in Europe:
Implications for International Convergence, in: European Accounting Review, Vol.
14, pp. 101-126.
Schipper, K./Vincent, L. (2003), Earnings quality, in: Accounting Horizons, Vol. 17, pp.
97-110.
Sellhorn, T./Gornik-Tomaszewski, S. (2006), Implications of the IAS Regulation for
Research into the International Differences in Accounting Systems, Working Paper,
Ruhr-Universitaet Bochum/St. Johns University.
Smith, J. A./Todd, P. E. (2001), Reconciling Conflicting Evidence on the Performance of
Propensity-Score Matching Methods, The American Economic Review, Vol. 91, pp.
112-118.

29

Van Tendeloo, B./Vanstraelen, A. (2005), Earnings Management under German GAAP

versus IFRS, in: European Accounting Review, Vol. 14, pp. 155-180.
Watts, R. L. (2003), Conservatism in Accounting, Part I: Explanations and Implications,
in: Accounting Horizons, Vol. 17, pp. 207-221.
Watts, R. L. (2006), What has the Invisible Hand Achieved? Working Paper, Massachusetts Institute of Technology.
Webster, E./Thornton, D. B. (2005), Earnings Quality under Rules- versus Principlesbased Accounting Standards: A Test of the Skinner Hypothesis, in: Canadian Accounting Perspectives, Vol. 4, pp. 167-192.

30

TABLE 1: Accounting Standards Followed by German Public Firms 1993-2004

Year
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Total

Accounting Standards Followed

HGB

IFRS

US-GAAP

549
99.82
559
99.64
560
98.94
556
97.54
545
92.06
518
86.77
503
77.74
473
62.48
451
60.54
443
53.70
440
53.33
378
48.84

0
0.00
1
0.18
3
0.53
7
1.23
19
3.21
42
7.04
86
13.29
162
21.40
177
23.76
253
30.67
263
31.88
305
39.41

1
0.18
1
0.18
3
0.53
7
1.23
28
4.73
37
6.20
58
8.96
122
16.12
117
15.70
129
15.64
122
14.79
91
11.76

5,975

1,318

716

Total
550
561
566
570
592
597
647
757
745
825
825
774
8,009

The sample comprises all firms included in the Worldscope universe. The data for Accounting Standards
Followed is in large part hand-collected from annual reports. Where annual reports are unavailable, we use
Worldscope item WC07536 (Accounting Standards Followed). Dual accounting is coded as HGB. Nonitalic numbers are frequency counts, and numbers in italics are percentages for the respective year.

31

TABLE 2: Determinants of Accounting Choice

Panel A: Descriptive Statistics for HGB Firms (n=1,176)
Variable
MANUF
MKTCAP
DEBT_TA
FOR_SALES
CLOSEHELD
NR_EXCHG
USLISTING
IPO

Mean
0.669
809.478
0.219
0.268
0.676
1.709
0.025
0.220

Median
121.455
0.202
0.177
0.713
2.000

Std. Dev
2,773.829
0.186
0.284
0.243
1.182

First Quartile
43.600
0.047
0.000
0.506
1.000

Third Quartile
366.400
0.351
0.512
0.900
2.000

Panel B: Descriptive Statistics for IFRS Firms (n=630)

Variable
MANUF
MKTCAP
DEBT_TA
FOR_SALES
CLOSEHELD
NR_EXCHG
USLISTING
IPO

Mean
0.517***
1,925.339***
0.208
0.352***
0.519***
2.154***
0.090***
0.594***

Median

Std. Dev

190.634*** 4,439.419
0.185
0.180
0.339***
0.299
0.526***
0.268
2.000***
1.462

32

First Quartile
49.728
0.040
0.000
0.311
1.000

Third Quartile
1,044.721
0.326
0.609
0.714
3.000

Panel C: Annual Logistic Determinant Model of IFRS Adoption

(1)

Prob (IFRS = 1 ) = logit ( 0 + 1 MANUF + 2 log( MKTCAP ) + 3 DEBT _ TA + 4 FOR _ SALES +

5 CLOSEHELD + 6 NR _ EXCHG + 7USLISTING + 8 IPO)

Variable
INTERCEPT
MANUF
log(MKTCAP)
DEBT_TA
FOR_SALES
CLOSEHELD
NR_EXCHG
USLISTING
IPO
Rescaled R

Expected Sign
?
+
?
+
+
+
+

Coefficient
-6.118
-0.054
0.394
0.314
1.332
-1.417
-0.041
0.958
1.456
0.359

Significance
0.011
0.872
0.000
0.732
0.000
0.023
0.543
0.084
0.000

Notes: n represents the number of firm-year observations over 1998-2004. MANUF is equal to one if the
firm belongs to a manufacturing industry (SIC below 4000). MKTCAP is the market capitalization of equity (WC08001) at the beginning of the fiscal year, measured in million . DEBT_TA is total debt
(WC03255) divided by total assets (WC02999). FOR_SALES is foreign sales (WC07101) divided by total
sales (WC01001). CLOSEHELD is the percentage of closely held shares (WC08021). NR_EXCHG is the
number of exchanges (domestic and foreign) on which the firms shares are listed (WC05427). USLISTING is one if the firm is directly or indirectly (via ADRs) traded at the New York Stock Exchange. IPO is
one if the firm went public in 1995 or later. IFRS is one if the firm follows IFRS in its group accounts. The
significance of sample differences is assessed by Chi-squared tests for nominal variables and by t-tests
(Wilcoxon tests) for the means (medians) of non-nominal variables. ***/**/* marks two-sided significance
at the 1/5/10% level. The coefficients in Panel C are average coefficients of annual logistic regressions of
model (1) over the period 1998-2004 and are estimated by maximum likelihood. Their significance is assessed by their annual distributions.

33

TABLE 3: Propensity Score-Matched Sample

Panel A: Accounting Standards Followed by Propensity Score-Matched Sample
Year

Accounting Standards Followed

Total

HGB

IFRS

1998
1999
2000
2001
2002
2003
2004

10
28
42
60
80
73
61

10
28
42
60
80
73
61

20
56
84
120
160
146
122

Total

354

354

708

Panel B: Descriptive Statistics for HGB Firms (n=354)

Variable
MKTCAP
NIBE_TA
CFO_TA
RET
NEG

Mean

Median

1,485.498
0.021
0.078
0.089
0.480

173.380
0.029
0.080
0.014

Std. Dev
4,511.109
0.115
0.131
0.518

First Quartile
47.676
0.003
0.020
-0.202

Third Quartile
618.484
0.064
0.136
0.270

Panel C: Descriptive Statistics for IFRS Firms (n=354)

Variable
MKTCAP
NIBE_TA
CFO_TA
RET
NEG

Mean
1,248.330
-0.021***
0.065
0.051
0.497

Median

Std. Dev

151.114
2,905.008
0.019***
0.155
0.076
0.174
0.000
0.617

First Quartile
32.907
-0.045
-0.002
-0.372

Third Quartile
669.900
0.052
0.151
0.295

Notes: n represents the number of firm-year observations over 1998-2004. The sample is constructed by
matching each IFRS observation with a HGB observation based on the propensities derived from the estimation of model (1). This procedure yields a sample of 1,140 observations. To ensure that only closely
matched pairs are used in the analysis, a caliper of 0.1 is imposed, resulting in the smaller sample presented
in the table. MKTCAP is the market capitalization of equity (WC08001) at the beginning of the fiscal year,
measured in million . NIBE_TA is net income before extraordinary items (WC01551) divided by total
assets (WC02999). CFO_TA is CFO divided by total assets. CFO is net income before extraordinary items
minus total accruals, which are equal to change in current assets (WC02201) less change in current liabilities (WC03101) less change in cash and cash equivalents (WC02001) plus change in current debt
(WC03051) less depreciation, amortization and depletion (WC01151). RET is the buy-and-hold return over
the fiscal year. NEG is one if RET is below zero. The significance of sample differences is assessed by Chisquared tests for nominal variables and by t-tests (Wilcoxon tests) for the means (medians) of non-nominal
variables. ***/**/* marks two-sided significance at the 1/5/10% level.

34

TABLE 4: Earnings Attributes

Panel A: Accrual Quality, Yearly Fixed Effects, Clustered Standard Errors
(2)

WCACC _ TAt =

2004

YEAR

i =1998

i ,t

+1CFO _ TAt 1 + 2CFO _ TAt + 3CFO _ TAt +1 + t

Accounting Standards Followed

HGB
IFRS
Variable

Coefficient

Significance

Coefficient

Significance

0.028
0.170
-0.562
0.204
0.580

0.028
0.000
0.000
0.000

0.012
0.081
-0.469
0.177
0.520

0.389
0.056
0.000
0.001

Avg. INTERCEPT
CFO_TAt-1
CFO_TAt
CFO_TAt+1
R2

Panel B: Persistence and Predictability, Yearly Fixed Effects, Clustered Standard Errors
(3)

NIBE _ TAt =

2004

YEAR

i =1998

i ,t

+1 NIBE _ TAt 1 + t

Accounting Standards Followed

HGB
IFRS
Variable

Coefficient

Avg. INTERCEPT
NIBE_TAt-1
R2

Significance

Coefficient

Significance

0.011
0.000

-0.011
0.502
0.294

0.321
0.000

0.027
0.133
0.387

Panel C: Value Relevance, Yearly Fixed Effects, Clustered Standard Errors

2004

(4)

RET15t =

YEAR + NIBE_ MVE + NIBE_ MVE +

i

i ,t

i =1998

Accounting Standards Followed

HGB
IFRS
Variable
Avg. INTERCEPT
NIBE_MVEt
NIBE_MVEt
R2

Coefficient
0.063
0.574
0.290
0.360

Significance

Coefficient

Significance

0.594
0.000
0.081

0.032
0.406
0.164
0.404

0.854
0.006
0.036

35

TABLE 4 (continued)
Panel D: Conditional Conservatism, Yearly Fixed Effects, Clustered Standard Errors
2004

(5)

NIBE_ MVEt =

YEAR + NEG + RET + NEG RET +

i

i,t

i =1998

Accounting Standards Followed

HGB
IFRS
Variable
Avg. INTERCEPT
NEGt
RETt
NEGtRETt
R2

Coefficient

Significance

Coefficient

Significance

0.073
-0.023
0.042
0.335
0.139

0.016
0.553
0.337
0.018

0.057
0.012
-0.064
0.543
0.189

0.316
0.835
0.399
0.000

Panel E: Summary and Significance Tests

Earnings Attribute

Statistic

Accounting Standards Followed

HGB
IFRS

Accrual Quality

Mean t WCACC _ TAt

1.857

2.115

Median t WCACC _ TAt

0.704

0.808

Persistence

0.133

0.502***

Mean t NIBE _ TAt

1.397

1.719

Median t NIBE _ TAt

0.789

0.896***

Predictability
Value Relevance

Mean t RET15t

2.183

2.131

Median t RET15t

0.825

0.757

Conditional Conservatism

3 - 2

0.293

0.607*

Notes: This Table details the regression results of models (2) to (5). Each model is estimated as a yearly
fixed-effect model using OLS and clustered standard errors to account for the heteroskedasticity and autocorrelation caused by the panel structure of the data. The sample is as reported in Table 3. YEAR is a series
of yearly fixed effects. WCACC_TA is working capital accruals, which are equal to change in current assets (WC02201) less change in current liabilities (WC03101) less change in cash and cash equivalents
(WC02001) plus change in current debt (WC03051), divided by total assets. TOTACC_TA is
WCACC_TA less depreciation, amortization and depletion (WC01151) divided by total assets. RET15 is
the buy-and-hold return over a 15-month period ending three months after fiscal year end. NIBE_MVE is
net income before extraordinary items (WC01551) divided by beginning-of-period market capitalization of
equity (WC08001). All other variables are as previously defined. In Panel E, the significance of mean (median) differences is assessed by t-tests (Wilcoxon tests). The significance of regression coefficient differences is assessed by fully interacted regressions of pooled samples (not tabulated) along with t-tests (1) or
F-tests (3 - 2), respectively. ***/**/* marks two-sided significance at the 1/5/10% level.

36

TABLE 5: Accounting Standards and Information Asymmetry

Panel A: Descriptive Statistics for HGB Firms (n=354)
Variable

Mean

Median

Std. Dev

First Quartile

Third Quartile

BAS
TOVER
ZRETURN
VOLA
BETA

0.035
0.006
0.340
0.414
0.431

0.025
0.002
0.254
0.359
0.295

0.039
0.015
0.250
0.234
0.536

0.014
0.001
0.131
0.266
0.060

0.042
0.006
0.531
0.478
0.590

Std. Dev

First Quartile

Third Quartile

0.030
0.017
0.189
0.287
0.668

0.016
0.001
0.096
0.315
0.260

0.041
0.010
0.312
0.715
1.140

Panel B: Descriptive Statistics for IFRS Firms (n=354)

Variable

Mean

BAS
TOVER
ZRETURN
VOLA
BETA

0.033
0.009**
0.237***
0.524***
0.771***

Median
0.025
0.004***
0.165***
0.440***
0.600***

37

Panel C: Impact of IFRS on Information Asymmetry

2004

YEAR

(6)

BAS t =

(7)

TOVERt =

(8)

ZRETURN t =

(9)

VOLAt =

+1 log( MKTCAPt ) + 2VOLAt + 3TOVERt + 4CLOSEHELDt + 5 IFRS + t

i ,t

i
i =1998
2004

YEAR

i ,t

i
i =1998
2004

YEAR

i =1998

i ,t

2004

YEAR

i =1998

+1 log(MKTCAPt ) + 2VOLAt + 3CLOSEHELDt + 4 IFRS + t

i ,t

+ 1 log( MKTCAPt ) + 2VOLAt + 3CLOSEHELDt + 4 IFRS + t

+1 log(MKTCAPt ) + 2CLOSEHELDt + 3 BETAt + 4 IFRS + t

Model
(6)
Variable
Avg. INTERCEPT
log(MKTCAP)
VOLA
TOVER
CLOSEHELD
BETA
IFRS
R2

Coefficient

(7)

Significance

0.076
-0.006
0.046
-0.190
0.027

0.000
0.000
0.000
0.001
0.000

-0.007
0.682

0.011

(8)

Coefficient Significance

(9)

Coefficient

Significance

Coefficient Significance

0.009
0.001
0.005

0.037
0.216
0.055

0.783
-0.049
-0.277

0.000
0.000
0.000

1.026
-0.055

0.000
0.000

-0.016

0.000

0.425

0.000

0.002
0.335

0.205

-0.068
0.800

0.001

-0.060
0.220
0.034
0.889

0.079
0.000
0.069

Notes: This Table details the regression results of models (6) to (9). Each model is estimated as a yearly fixed-effect model using OLS and clustered standard
errors to account for the heteroskedasticity and autocorrelation caused by the panel structure of the data. The sample is as reported in Table 3. YEAR is a series
of yearly fixed effects. BAS is the average bid-ask spread of the closing price at the main exchange over the calendar year. TOVER is the average daily number
of shares traded at all German exchanges divided by the daily number of shares outstanding. ZRETURN is the proportion of trading days where there is no price
change for the stock. VOLA is the volatility of the firms stock over the calendar year. BETA is the coefficient of a daily regression of the firms stock return on
a value-weighted market index. All other variables are as previously defined. In Panel A and B, the significance of mean (median) differences is assessed by ttests (Wilcoxon tests). ***/**/* marks two-sided significance at the 1/5/10% level.

38