IAS VERSUS U.S. GAAP: INFORMATION ASYMMETRY-BASED EVIDENCE FROM GERMANY''''S NEW MARKET potx

GAAP firms in the New Market are ex-pected to exhibit similar accounting quality, despite differences in the stan-dards, because they face similar market forces and institutional factors

Printed in U.S.A.

IAS Versus U.S GAAP:

Information Asymmetry–Based Evidence from Germany’s

in the bid-ask spread and share turnover between IAS and U.S GAAP firmsare statistically insignificant and economically small Subsequent analyses of

∗University of Pennsylvania I gratefully acknowledge helpful comments from Anne d’Arcy, George Benston, Phil Berger, Gus De Franco, Robert Holthausen, Peter Knutson, Christian Laux, S P Kothari, Claudia R¨oder, Robert Verrecchia, and especially Ray Ball and the anony- mous referee This paper has benefited from presentations at the American Enterprise Insti- tute, University of California at Berkeley, Columbia University, Harvard University, J W Goethe Universit¨at Frankfurt, MIT, University of Michigan, Stanford University, Tilburg University, the EAA Meetings (Munich) and the EFA Meetings (Barcelona) I would also like to thank Uwe Schweickert (Deutsche B¨orse), Peter Gomber (Deutsche B¨orse), J¨org Hueber (KPMG), Rainer J¨ager (PWC), and IBES for generously providing data for this study, and Tobias Herwig for his excellent research assistance I also gratefully acknowledge financial support by the Wharton Electronic Business Initiative (WeBI).

445 Copyright  , University of Chicago on behalf of the Institute of Professional Accounting, 2003

446 C.LEUZ

analysts’ forecast dispersion, initial public offering underpricing, and firms’standard choices corroborate these findings Thus, at least for New Marketfirms, the choice between IAS and U.S GAAP appears to be of little conse-quence for information asymmetry and market liquidity These findings donot support widespread claims that U.S GAAP produce financial statements

of higher informational quality than IAS

1 Introduction

This study is motivated by the debate about the two leading contenders inthe global competition among financial reporting regimes: U.S generallyaccepted accounting principles (GAAP) and international accounting stan-dards (IAS) This debate, which is summarized in section 2, focuses primarily

on comparisons of the stipulated accounting methods per se There is, ever, little empirical evidence on the standards’ economic consequences incapital markets (see also Pownall and Schipper [1999])

how-In this article I investigate whether firms using U.S GAAP vis-`a-vis IASexhibit cross-sectional differences in several proxies for information asym-metry, such as bid-ask spreads and share turnover I focus on these proxiesbecause reducing information asymmetries and increasing market liquidity

is one, albeit an important, concern in securities and accounting regulation(e.g., Loss and Seligman [2001]) Moreover, in using these proxies, the testsare not restricted to comparisons of summary accounting measures such asearnings but capture differences in financial reporting information morebroadly (e.g., footnote disclosures)

The study exploits a unique setting in which the two competing sets of dards are placed on a level playing field Firms trading in Germany’s NewMarket are required to choose between IAS and U.S GAAP for financialreporting purposes, but face the same regulatory environment otherwise.Potentially offsetting institutional factors such as capital market structure,listing requirements, and enforcement of accounting standards are heldconstant I thereby avoid difficulties of comparisons across firms from differ-ent countries or different capital markets (e.g., Frost and Pownall [1994]).Furthermore, consolidated IAS and U.S GAAP reports are neither the basis

stan-of taxation nor dividend restrictions in company law, focusing the ison on disclosure effects in capital markets

compar-The results indicate that the differences in the bid-ask spread and shareturnover across IAS and U.S GAAP firms are statistically insignificant andeconomically small For instance, the spread regressions suggest that, set-ting statistical significance aside, the effect of U.S GAAP reporting is less

than 3% of the percentage spread Next, I examine analysts’ forecast

disper-sion and initial public offering (IPO) underpricing as alternative proxiesfor information asymmetry and find that the differences are also statisti-cally insignificant Finally, I analyze firms’ standard choices and check forselection bias with two-stage regressions These tests corroborate the otherfindings In summary, the choice of IAS or U.S GAAP appears to be of littleconsequence for information asymmetry and market liquidity

This finding is open to at least two interpretations If accounting dards are important in determining firms’ accounting quality (e.g., Levitt[1998]), this finding implies that IAS and U.S GAAP are comparable sets

stan-of standards, at least in terms stan-of their ability to reduce information metries The evidence does not support claims that U.S GAAP producefinancial statements of higher informational quality than IAS

asym-However, the finding is also consistent with the view that accounting ity is largely determined by firms’ reporting incentives created by marketforces and institutional factors rather than by accounting standards (in par-ticular, Ball, Robin, and Wu [Forthcoming], Ball and Shivakumar [2002]).Based on this view, IAS and U.S GAAP firms in the New Market are ex-pected to exhibit similar accounting quality, despite differences in the stan-dards, because they face similar market forces and institutional factors.Thus, in holding institutional factors constant and varying firms’ standards,this study complements other studies that vary firms’ incentives while hold-ing standards constant (e.g., Ball, Robin, and Wu [Forthcoming], Ball andShivakumar [2002]) Together, the studies provide evidence suggesting thatthe global accounting debate focuses too much on the standards choice andtoo little on market forces and institutional factors, as also argued by Ball[2001]

qual-Although the New Market setting offers several advantages, there arecaveats that should be borne in mind when interpreting the results First,New Market firms are young growth firms that are strongly dependent onequity financing Although information asymmetry and disclosure issues arepertinent for such firms (Smith and Watts [1992]), financial statements (ofany kind) may not be as important or serve special purposes, reducing thepower of my tests Moreover, it is unclear whether the findings generalize tofirms that, for instance, are in more mature industries, have higher leverage,

or trade on other exchanges

Second, the study focuses on a comparison of IAS and U.S GAAP in terms

of information asymmetry in the equity market and hence is limited in scope.There are other important roles of accounting and disclosure standards,such as improving corporate governance, which are not considered in thisstudy Third, New Market firms choose their accounting standards Thus,

my results are only valid to the extent that I have appropriately controlledfor selection bias Finally, the paper does not address the policy question ofwhether a country should accept or switch to IAS.1

The remainder of the paper is organized as follows Section 2 sketchesthe global accounting debate and summarizes prior empirical results.Section 3 describes Germany’s New Market Section 4 delineates the re-search design and section 5 presents the results for bid-ask spreads and shareturnover Section 6 considers alternative information asymmetry proxies and

1 See, for example, Pownall and Schipper [1999], Dye and Sunder [2001], and Sunder [2002] for this debate Note also that Holthausen and Watts [2001] and Ronen [2001] caution about drawing policy implications from market-based tests.

448 C.LEUZ

section 7 examines determinants of firms’ standard choices Section 8 cludes the paper

con-2 The Development of IAS, the Global Accounting Debate,

and Prior Evidence

The International Accounting Standards Committee (IASC) wasfounded in 1973 to set international accounting standards and promotetheir global acceptance In response to criticism of its early standards, theIASC embarked on the Comparability/Improvements Project in 1987 Therevised standards, which became effective in 1995, substantially reducedthe number of accounting choices In July 1995, the IASC and InternationalOrganization of Securities Commissions (IOSCO) agreed on a list of ac-counting issues that needed to be addressed by any set of standards seekingIOSCO’s endorsement for cross-border offerings and listings The ensuingCore Standards Project led to substantial revisions of IAS By March 1999,the IASC completed all but one of those issues and subsequently receivedIOSCO’s endorsement subject to “reconciliation, disclosure and interpreta-tion at the national level” (IOSCO Press Release [May 17, 2000]).These developments and the competition between IAS and U.S GAAP tobecome the global set of accounting standards have led to a debate abouttheir relative quality Since then, several security regulators such as the U.S.Securities and Exchange Commission (SEC) and the Canadian SecuritiesAdministrators (CSA) have asked for feedback on the quality of IAS (e.g.,SEC [2000], CSA [2002])

In the ongoing debate, IAS proponents argue that IAS have improvedsubstantially over the years and that revised IAS are relatively close to U.S.GAAP with minor differences remaining.2For instance, Harris [1995] con-ducts a detailed case study computing earnings and shareholders’ equityfor eight companies under both standards He concludes that firms com-plying with revised IAS provide accounting measures that are essentiallyconsistent with U.S GAAP and that, based on his study, there is no com-pelling evidence that U.S GAAP are superior to IAS Others point out thatdisclosures are sufficient to allow investors to make their own inferences inthose instances where substantial differences remain or IAS permit severalaccounting choices (e.g., Enevoldsen, Jones, and Carsberg [2000]).Overall, the proponents claim that IAS are now of sufficiently high quality.Consistent with this opinion, a recent survey by KPMG [2000] shows thatCFOs of large European corporations view IAS as offering similar qualitywhile being cheaper to implement than U.S GAAP, which are often per-ceived as being too detailed and too complex

IAS opponents, although acknowledging that IAS have improved erably, argue that many important differences between the two standards

consid-2 See, for example, E MacDonald, “U.S Firms Likely to Balk at SEC Move to Ease Listing of

Foreign Companies,” Wall Street Journal, February 18, 2000, p A3.

remain and that U.S GAAP are still superior because IAS are less rigorous,are less detailed, afford more flexibility, or require fewer disclosures.3TheFinancial Accounting Standards Board (FASB [1999]), for instance, points

to more than 250 key differences in four categories: recognition, ment, permissible alternatives, and lack of requirements or guidance.4Based

measure-on this comparismeasure-on, the FASB cmeasure-oncludes that IAS are of lower quality thanU.S GAAP.5

In summary, there are opposing views on the quality of IAS relative toU.S GAAP, but the debate focuses on the stipulated accounting methodsthemselves rather than any empirical evidence Prior studies focus largely oncomparing foreign and U.S GAAP financial statements, for instance, withrespect to the value relevance of accounting earnings (e.g., Alford et al.[1993]) Other studies use Form 20-F reconciliations to assess the compara-bility and quality of foreign GAAP relative to U.S GAAP (e.g., Amir, Harris,and Venuti [1993]) Using an information-asymmetry approach, which en-compasses a broad set of disclosures, Leuz and Verrecchia [2000] compareGerman firms following U.S GAAP or IAS with firms following GermanGAAP

A few studies explicitly address the relative quality of IAS and U.S GAAPand hence are particularly pertinent to the current debate on global ac-counting standards Harris and Muller [1999] examine Form 20-F recon-ciliations from IAS to U.S GAAP They find that, based on reconciliationmagnitudes, IAS are closer to U.S GAAP than other foreign GAAP but thatreconciliation items are incrementally value relevant.6They interpret theirfindings as evidence that IAS and U.S GAAP accounting measures are notsubstitutes However, as Pownall and Schipper [1999] point out, evidencefrom 20-F reconciliations is unlikely to be representative of IAS (or foreignGAAP) firms not seeking U.S listings.7 Ashbaugh and Olsson [2002] ex-amine non-U.S firms quoted on London’s SEAQ They find that IAS andU.S GAAP earnings and book values of equity are equally value relevantbut that the relative value relevance depends on the valuation model used.Although these results are consistent with my findings, the tests are based on

3See, for example, E MacDonald, “U.S Accounting Board Faults Global Rules,” Wall Street Journal, October 18, 1999, p A1; L Berton, “Countdown to Harmonization,” Institutional In- vestor , June 1999, pp 25–26; J Garten, “Global Accounting Rules? Not So Fast,” Business Week,

April 5, 1999, p 26; M McNamee, “Can the SEC Make Foreign Companies Play by Its Rules?”

Business Week, March 6, 2000, p 46; G Imhoff, “Compromising Standards Threatens ism,” The Dividend, Spring 1999, pp 22–25.

Capital-4 The International Forum on Accountancy Development (www.ifad.net) provides updated comparisons.

5See E MacDonald, “U.S Accounting Board Faults Global Rules,” Wall Street Journal,

Octo-ber 18, 1999, p A1.

6 Davis-Friday and Rueschhoff [2001] examine seven firms before and after the IASC’s parability Project and find that IAS net income and shareholders’ equity are more significantly related to market value after the revision of IAS.

Com-7 Besides, prior research provides little evidence that investors actually use Form 20-F onciliations See Saudagaran and Meek [1997] for a discussion.

Finally, several recent studies find that accounting quality is determinedprimarily by market forces and institutional factors, rather than account-ing standards (e.g., Ball, Robin, and Wu [Forthcoming], Leuz, Nanda, andWysocki [Forthcoming]) These findings suggest that standards per se donot have a major impact on accounting quality and that the global account-ing debate focuses perhaps too much on the standards.

Thus, based on the prior literature, the economic substance of the ences between IAS and U.S GAAP remains an open and largely empiricalquestion

differ-3 The New Market in Germany

The New Market (or Neue Markt) was launched in March 1997 as a newGerman stock market segment geared toward small- and medium-size com-panies in innovative and fast-growing industries Within a few months ofits inception, it became Europe’s most successful equity market for growthfirms, both in terms of market capitalization and number of listings.8How-ever, along with other growth and technology markets, the New Market suf-fered a severe downturn in recent years Responding to this market trend,Deutsche B¨orse decided to reorganize its markets and stop singling outgrowth and technology stocks in a segment In 2003, New Market firms arebeing reassigned to two newly created market segments, called Prime andGeneral Standard The Prime Standard segment inherits the strict listingand disclosure requirements of the New Market, which go substantially be-yond those of the General Standard.9In addition, new securities regulation

is under way to address enforcement problems that have become apparentduring the New Market period.10

8See, for example,“European Stockmarkets: A German Coup,” The Economist, January 9,

1999, pp 69–71; R Zimmermann, “Eine Flasche Champagner zum dritten Geburtstag,” cial Times Deutschland, March 9, 2000, p 20; “Amerikas Anleger werden vielf¨altig gesch ¨utzt,” Frankfurter Allgemeine Zeitung , Septermber 5, 2001, p 208; “Neuer Markt,” Frankfurter Allgemeine Zeitung (Supplement), March 6, 2001.

Finan-9See “Neuer Markt Closure, Comment & Analysis,” Financial Times, September 27, 2002,

p 11.

10 The Deutsche B¨orse tightened the rules several times to address problems as they became apparent For instance, in March 2001, it introduced rules that (1) require disclosure of all share transactions by managers, board members, and the company itself; (2) standardize and extend quarterly financial reports; and (3) increase penalties for rule violations, including fines

up to 100,000 EUR and delisting See A Kueppers, “Following in the Shadow of Nasdaq, Neuer

Markt Sees 76% Plunge in its Stocks,” Wall Street Journal, March 20, 2001, p C10; S Ascarelli,

“German Exchange Unplugs Neuer Markt,” Wall Street Journal, September 27, 2002, p A12.

The New Market’s rules were deliberately chosen to exceed traditionalGerman listing and disclosure requirements because New Market firms arecharacterized by substantial uncertainty about business prospects and man-agement expertise.11 To describe briefly the key rules, the New Market’sRegelwerk stipulates that, at the IPO, firms are three years of age, have a min-imum free float of 20%, and commit to a six-month lock-up period.12Thereare extensive and detailed disclosure requirements for the IPO prospectus(Regelwerk, §4) In particular, firms have to provide comparable financialstatements for three previous fiscal years Subsequently, firms have to pre-pare and publish annual financial statements no later than three monthsafter the fiscal year end (Regelwerk, §7) Financial statements have to be inaccordance with either IAS or U.S GAAP In addition, firms must publishquarterly reports within two months after each quarter and hold at least oneanalyst conference per year.

The Regelwerk also requires that annual financial statements be audited.The enforcement of either IAS or U.S GAAP lies primarily with the auditors,whose legal liability has increased considerably since the amendment of

§323 HGB (German Commercial Code) in April 1998 In addition, auditorsand directors could face criminal prosecution for misleading or fraudulentfinancial statements (§§331 and 332 HGB) It is also possible to sue fordamages in civil courts, but only after a conviction in criminal proceedings.However, as U.S.-style shareholder litigation or SEC-like monitoring does notexist in Germany, enforcement is unlikely to be as strong as in the UnitedStates For this reason, I exclude New Market firms with U.S listings fromthe sample The concern is that differential enforcement could otherwisebias the tests However, I note that weak enforcement likely reduces thepower of my tests

Mitigating this concern, Glaum and Street [Forthcoming] find that thecompliance of New Market firms with required disclosures is on averagereasonably high.13They also show that firms with U.S listings exhibit highercompliance levels than do firms without such listings However, once thiseffect is controlled for, the compliance levels of IAS firms are only slightlylower (≈2%) and only marginally significant (at the 8% to 9% level).14Thus,any remaining bias in my tests appears to be small and in favor of U.S GAAPfirms

Finally, as my empirical tests are based on bid-ask spreads and shareturnover, I briefly describe the market microstructure of the New Market

11 See, for example, V Fuhrmans, “Playing by the Rules: How Neuer Markt Gets Respect,”

Wall Street Journal, August 21, 2000, p A3; “Strenges Regelwerk sorgt f ¨ur hohe Transparenz am Neuen Markt,” Frankfurter Allgemeine Zeitung , September 25, 2000, p 32.

12 The Regelwerk is available online at http://deutsche-boerse.com/nm/.

13 Mean and median compliance ratios are 83.7% and 85.9%, respectively Bradshaw and Miller [2002] show that, even for U.S firms, compliance ratios are on average below 100% But the compliance ratios are not directly comparable.

14 D’Arcy and Grabensberger [Forthcoming] find similar results comparing the compliance

of IAS and U.S GAAP firms with the New Market’s quarterly reporting rules.

452 C.LEUZ

Shares are traded simultaneously on the floor and on an electronic tradingplatform (Xetra) Floor trading is organized as an auction system only Theelectronic trading system, which is a hybrid between an auction and market-maker system, allows traders to post limit orders Spreads arise from the bestbid and best ask of all limit orders, not just the quotes of the market makers.Each New Market firm has to name at least two designated sponsors(Betreuer) that act as market makers and promote liquidity They providebinding bid and ask limit orders (or quotes) for the three daily auctions andupon request by a market participant (with a maximum response time of

120 seconds).15 The sponsors’ minimum quote volume is 20,000 EUR andtheir quoted spreads must not exceed 4% The exchange monitors spon-sor performance (since October 1999) and publish ratings (since January2000) Prior studies suggest that sponsors have a stabilizing, but not domi-nating, role in the New Market, and in particular they facilitate larger trades(Theissen [1998], Gerke and Bosch [1999])

4 Research Design and Information Asymmetry Proxies

Economic theory suggests that information asymmetries between tial buyers and sellers of firm shares introduce adverse selection into sec-ondary share markets and hence reduce market liquidity (e.g., Copelandand Galai [1983], Glosten and Milgrom [1985]) Information asymmetriesare costly to firms, as investors adjust prices to compensate for holdingshares in illiquid markets Increasing the level or precision of disclosureshould reduce the likelihood of information asymmetries between investorsand increase market liquidity (e.g., Diamond and Verrecchia [1991]) Thus,information asymmetry proxies should reflect, among other things, firms’accounting quality In principle, the measures can capture any differencebetween IAS and U.S GAAP and should account for trade-offs among recog-nition, measurement, and footnote disclosures Consistent with these hy-potheses, Welker [1995], Healy, Hutton, and Palepu [1999], and Leuz andVerrecchia [2000] provide evidence that information asymmetry and liquid-ity proxies are associated with firms’ disclosure and accounting policies.Furthermore, security market regulators emphasize the role of high-quality accounting standards in leveling the playing field among investorsand increasing investor confidence, that is, in reducing information asym-metries and increasing liquidity (e.g., Sutton [1997]) For instance, Levitt[1998, p 81] states, “High[er] quality accounting standards result in greaterinvestor confidence, which improves liquidity, [and] reduces capital costs”(see also FASB [1999, p 3])

poten-The preceding discussion suggests information asymmetry–based tests as away to assess the economic substance of the global accounting debate and, inparticular, the arguments put forth in favor of U.S GAAP The tests examine

whether, ceteris paribus, firms employing U.S GAAP exhibit less information

15For details, see Designated Sponsor Guide, http://deutsche-boerse.com/nm/.

asymmetry and higher market liquidity than firms using IAS If the ences between IAS and U.S GAAP firms turn out to be insignificant, theneither IAS and U.S GAAP are comparable in reducing information asym-metries, consistent with the argument of IAS proponents, or standards donot primarily determine informational quality, as suggested by Ball, Robin,and Wu [Forthcoming].16

differ-As the ceteris paribus condition is critical for the test, I examine New

Mar-ket firms that are incorporated in Germany and that do not trade abroad.For these firms, country- and market-specific factors, such as the enforce-ment of accounting standards and nonaccounting disclosure requirements,are held constant Furthermore, consolidated IAS or U.S GAAP reports donot have immediate tax or dividend implications In Germany, taxation andlegal dividend restrictions are not based on consolidated (or group) finan-cial statements Under the Commercial Code, such statements serve purelyinformational purposes

I analyze whether IAS and U.S GAAP firms in the New Market exhibitcross-sectional differences in the bid-ask spread and share turnover, both ofwhich are standard proxies for information asymmetry and market liquidity.Although spread data are not readily available for the New Market, theDeutsche B¨orse kindly provided data from June 1, 1999, to July 31, 1999,and from June 1, 2000, to August 31, 2000 In 1999, the spreads are measuredafter the New Market All Share index had gained 24% from January to theend of May In 2000, it had lost almost 35% from its peak in March to thebeginning of the measurement interval Analyzing both periods serves as

a robustness check that the results are not specific to the New Market’sboom phase However, sharp market movements, such as those of the NewMarket in 1999 and 2000, could also make the proxies more cross-correlatedand less powerful for my purposes because investors are more likely to rely

on marketwide information rather than firm-specific information in thesetimes.17It is therefore noteworthy that the measurement intervals themselvesare comparatively stable periods, where the New Market All Share indexposted only modest returns of 1.8% and 1.3%, respectively

For each stock, the exchange provides an equally weighted monthly erage of all spreads that existed in the electronic trading system (Xetra).18The aggregated data provision precludes a decomposition of the spread intocomponents related and unrelated to information asymmetry There are,however, several institutional features that mitigate this data limitation andmake the Xetra spreads used in this study conceptually appealing proxies for

av-16 These arguments presume that test power is sufficiently high The power issue is therefore explicitly addressed in the robustness checks and should be kept in mind as a caveat when interpreting the results.

17 I thank the anonymous referee for pointing this out.

18 Every (new) spread arising from either a change of the best-bid or the best-ask price is

recorded In 2000, the monthly average for each stock is computed on average from more than

3,000 individual spreads.

454 C.LEUZ

information asymmetry.19First, spreads arise from limit orders posted in the

electronic trading system by all traders, instead of the quotes of a specialist or

a few market makers In such a market, other spread components unrelated

to information asymmetry, such as inventory-holding costs or monopolyrents, should be smaller Empirical comparisons of order- and quote-drivenmarkets support this conjecture (e.g., Huang and Stoll [1996]) Second,traders are charged for order processing separately, and New Market firmspay fees to their designated sponsors to compensate them for their service.For these reasons, order-processing costs are unlikely to be a major spreadcomponent Third, trades are executed automatically, which implies that thequoted spreads are the effective spreads Finally, share prices are quoted withtwo decimal places, which reduces price discreteness and should increasethe power of the proxy

Similar issues arise with respect to share turnover Although adverse tion among investors clearly reduces liquidity, the proxy also reflects factorsunrelated to information asymmetry, such as portfolio rebalancing, liquidityshocks, or changes in risk preferences There is, however, evidence that sup-ports the choice of turnover as an inverse proxy for information asymmetry.For instance, Easley et al [1996] show that the probability of informed trad-ing is decreasing in trading volume and Grammig, Schiereck, and Theissen[2000] confirm their findings for the regular German market segment

selec-5 Cross-Sectional Analysis of Information Asymmetry

and Market Liquidity

5.1 SAMPLE SELECTION AND DESCRIPTIVE STATISTICS

As of April 30, 1999, 90 firms were listed in the New Market I eliminatefirms that are incorporated outside of Germany (6), listed abroad (5), orboth (5) This restriction reduces the sample to 74 firms but ensures thatall of them trade in the same market and operate in the same legal envi-ronment.20 In addition, I eliminate 5 firms that follow German GAAP intheir annual reports.21Thus, the final sample for 1999 contains 69 firms Toincrease the sample, I also perform my analyses using the 246 firms listed

in the New Market as of April 30, 2000 Again, I eliminate firms that are

19 Besides, Clarke and Shastri [2000] demonstrate severe problems with the decomposition of spreads Thus, it is not obvious whether the decomposition increases or decreases measurement error.

20 Some New Market firms are also listed on the Geregelten Markt of regional exchanges

or traded in the Freiverkehr (OTC market) However, the disclosure and listing requirement for these market segments are limited compared with the New Market Moreover, a listing on

the New Market entails a private contract with the Deutsche B¨orse and a registration for the

Geregelten Markt in Frankfurt.

21 In its early days, the New Market allowed some firms to provide German GAAP financial statements for a limited time if they were temporarily unable to prepare them according to IAS or U.S GAAP By April 30, 2000, all firms follow either IAS or U.S GAAP in their annual reports.

T A B L E 1

Descriptive Statistics on Accounting Standard Choices

Panel A: Accounting standard choices in the New Market a

All Firms Listed as Sample for 1999 All Firms Listed as Sample for 2000

of April 30, 1999 (n = 69 firms) of April 30, 2000 (n = 195 firms)

Panel B: Distribution of accounting standards by industry b

Sample for 1999 Sample for 2000

b Panel B is based on the industry classification provided for the New Market by the Deutsche B¨orse (http://www.neuer-markt.de).

incorporated outside Germany (34), listed abroad (11), or both (6) Thus,the final sample for 2000 comprises 195 firms

Panel A of table 1 reports the accounting standard choices of sample firmsand all New Market firms as of April 30, 1999, and April 30, 2000 The panelshows that IAS and U.S GAAP are fairly evenly distributed across firms Thepercentage of U.S GAAP firms is smaller in my samples than in the entiremarket because firms that are eliminated because of their foreign listingoften trade in the U.S and use U.S GAAP Panel B reports firms’ standardchoices by industry Both standards are fairly evenly distributed within mostindustries, except in Media & Entertainment and in Internet industries,where firms seem to prefer IAS and U.S GAAP, respectively Subsequenttests therefore check for industry effects

Table 2 provides descriptive statistics for the dependent variables and firmcharacteristics by accounting standard choice The table reports the averagepercentage spread from June 1, 1999, to July 31, 1999, from June 1, 2000,

to August 31, 2000, respectively, and the median daily share turnover fromMay 1, 1999, to July 31, 1999, and from June 1, 2000, to August 31, 2000,respectively.22I use the median turnover because the median is conceptually

a better proxy for the (normal) level of liquidity trading than the average,

22 Although the exchange provides spread data in 1999 for two months only, I use month intervals for all other variables to have at least 60 trading days to compute them.

by Deutsche B¨orse The spread is expressed as a percentage and computed as the difference between the

best bid and ask divided by the midpoint For each stock, the exchange provides a monthly average of all

spreads that existed in the electronic XETRA trading system (see discussion in section 4) The table reports the average percentage spreads from June 1, 1999, to July 31, 1999, and from June 1, 2000, to August 31,

2000 Turnover, share price, and market value data have been obtained from Datastream Share turnover is expressed as a percentage and computed for 60 trading days as the median of daily volume in Euro divided

by daily market capitalization in Euro from May 1, 1999, to July 31, 1999, and from June 1, 2000, to August

31, 2000, respectively The market capitalization (in millions of Euro) is measured as the average market value of equity from May 1, 1999, to July 31, 1999, and from June 1, 2000, to August 31, 2000, respectively Share price volatility is computed over the same intervals as the standard deviation of daily returns Free float (i.e., 1 minus the percentage of shares closely held) was obtained from B¨orse Online (based on ownership data published by New Market firms) and is measured as of May 1, 1999, and June 1, 2000, respectively The number of days listed in the New Market is measured from the IPO to April 30, 1999, and April 30,

2000, respectively Analyst following (i.e., the number of financial analysts providing earnings forecasts) was obtained from IBES Forecast dispersion is the standard deviation of the seven-month consensus forecast for the fiscal years 1999 and 2000 Asterisks indicate that the means (medians) of IAS and U.S GAAP firms

are significantly different using a two-tailed t-test (Mann-Whitney-Wilcoxon test).

∗p < 1;∗∗p < 05;∗∗∗p < 01.

which may be influenced by a few days of heavy trading around a particularevent Using the average turnover, however, produces similar results

The descriptive statistics show that the New Market’s average spread and

average share turnover are considerably lower in 2000 than in 1999 Thesharp decline in share turnover likely reflects the end of the new economy

boom and the market downturn Continuous efforts of Deutsche B¨orse toimprove market structure and trading efficiency could explain the reduction

in spreads For instance, as described in section 3, the Deutsche B¨orse startedmonitoring sponsor performance in October 1999 and publishing sponsorratings in January 2000, which could have reduced the average spread inthe market As marketwide fluctuations in liquidity and microstructure im-provements apply to all firms, they are unlikely to affect my cross-sectionaltests.23

The differences in the spread and share turnover across U.S GAAP andIAS firms are small and statistically insignificant for both the groups’ meansand medians But as these univariate comparisons do not control for spreadand turnover determinants, they should be interpreted cautiously Table 2also presents descriptive statistics for other firm characteristics, showing thatthe two groups exhibit significant differences in firm size (medians only)and free float

5.2 REGRESSION ANALYSIS

In this section I study cross-sectional differences in the bid-ask spreadand share turnover between IAS and U.S GAAP firms controlling for firmcharacteristics I report the results for 1999 and 2000 using the samplesdescribed in the previous section

5.2.1 Bid-Ask Spreads The bid-ask spread model is based on the extant

literature Prior studies suggest that the percentage spread is negatively ciated with trading volume, firm size, and market-maker competition, andpositively associated with price volatility and insider presence (e.g., Stoll[1978], Chiang and Venkatesh [1988], Glosten and Harris [1988])

asso-To control for these determinants, I use the average share turnover, age market capitalization, and daily share price volatility over the respectiveintervals in 1999 and 2000.24 I also include the firm’s free float as an in-verse proxy for the presence of insiders Reflecting the New Market’s order-driven microstructure, the model does not include a variable for market-maker competition.25To conduct the test developed in the previous section,

aver-23 The fact that spreads and turnover move in the same direction appears to send mixed signals about changes in (aggregate) liquidity Although both proxies are related to the same economic construct, they are also affected by other factors unrelated to information asymmetry (see section 4) As these factors are unlikely to be the same for spreads and turnover, it is conceivable that changes in these other factors are responsible for the joint decrease in spreads and turnover (see also Chordia, Roll, and Subrahmanyam [2001]).

24 See table 2 for details Turnover is used as opposed to unscaled trading volume to avoid multicollinearity problems with market capitalization.

25 In an earlier version, I reported spread regressions including the number of designated sponsors as a control variable These regressions are close to those in table 3 and yield an insignificant coefficient for the number of designated sponsors The latter finding is not sur- prising considering that spreads arise from limit orders posted by all traders, and not only the from market-maker quotes (see also the discussion in section 4) Similarly, controlling for sponsor ratings does not materially affect results.

458 C.LEUZ

I introduce a binary variable into the model indicating the firm’s reportingchoice (U.S GAAP = 1) To the extent that U.S GAAP firms report higherquality information, the dummy variable is expected to exhibit a negative co-efficient indicating lower spreads and hence lower information asymmetryfor U.S GAAP firms

As most analytical models identify multiplicative relations between thespread and its determinants (e.g., Stoll [1978], Glosten and Milgrom[1985]), I estimate a log-linear specification, which is standard in the ex-tant literature.26 Spearman correlations and regression diagnostics based

on Belsley, Kuh, and Welsch [1980] suggest that multicollinearity amongthe independent variables is not a problem

Panel A of table 3 presents the coefficients and t-statistics for ordinary

least squares (OLS) regressions with White-corrected standard errors Theregressions explain at least 75% of the variation in spreads, which is com-parable to prior research In both years, the coefficient on the U.S GAAPdummy is negative, but not statistically significant All other variables havethe expected signs and are highly significant

In summary, there is little evidence that firms employing U.S GAAP havelower bid-ask spreads than firms using IAS Thus, the regressions do notsupport the claim that U.S GAAP are of significantly higher quality thanIAS

5.2.2 Share Turnover The turnover model is also based on the extant

literature Prior studies suggest that share turnover is related to firm size andpositively associated with volatility, institutional ownership, and the inclusion

in a stock index (e.g., Bessembinder, Chan, and Seguin [1996], Tkac [1999],Leuz and Verrecchia [2000])

To control for these determinants, I use the average market tion and daily share price volatility over the respective intervals in 1999 and

capitaliza-2000, and a binary variable indicating whether the firm is included in theNEMAX 50 index Data on institutional ownership are not publicly available

in Germany Instead, I control for the firm’s free float, which is expected

to be positively associated with turnover To conduct the test proposed insection 4, I again include a binary variable representing the firm’s standardchoice (U.S GAAP = 1) To the extent that U.S GAAP firms report higherquality information, the dummy variable is expected to exhibit a positivecoefficient, indicating higher share turnover and hence higher market liq-uidity for U.S GAAP firms

Following the spread model, I use a log-linear specification The sults, however, are similar using a linear specification or rank regressions Ialso check that multicollinearity among the independent variables is not aproblem

re-26 I also estimate rank regressions as a robustness check and find that this specification does not materially alter my results or conclusions.