Accounting Comparability and Economic Outcomes of

Mandatory IFRS Adoption

Michael Neel*
University of Houston

July 2013

Abstract: This study examines the accounting mechanisms linking the widespread mandatory
adoption of IFRS in 2005 to market and analyst outcomes of adoption. I partition adopter firms
into treatment groups based on changes in both cross-country accounting comparability and
reporting quality around adoption. I find that an increase in firm value and a decrease in forecast
errors and dispersion under IFRS are restricted to adopters that exhibit an increase in
comparability. I also find that the liquidity effect of IFRS is generally broader, however it is still
most pronounced among adopters with an increase in comparability. In contrast, changes in
reporting quality around adoption appear to have only a second-order effect that is generally
limited to those adopters with a concurrent increase in comparability. I continue to find these
results for adopter samples restricted to countries in the EU, with strong enforcement, or that
implemented proactive financial statement reviews in 2005.

*This paper is based on my dissertation. I thank Anwer Ahmed, Mary Lea McAnally, Senyo Tse, Tom Omer and
Gil Sadka for helpful comments. I also thank workshop participants at Texas A&M University, University of
Miami, University of Houston, Michigan State University, Oklahoma State University and University of South
Carolina, as well as participants at the 2012 AAA Annual Meeting for their helpful comments and suggestions. I
gratefully acknowledge funding support from the C.T Bauer College of Business.
 I. INTRODUCTION

A growing literature points to variation in the economic outcomes of the mandatory

introduction of International Financial Reporting Standards (IFRS) (see e.g. Daske et al. [2008];

Li. [2010]; Byard et al. [2011]; Christensen et al. [2012]; Horton et al. [2013]). These papers

generally focus on country-level variation in economic outcomes, presumably driven by

differences in the level of, and changes in, institutional quality and reporting incentives across

adopter countries.1 My study extends this literature and examines the association between

economic outcomes of adoption and concurrent changes in adopters’ financial reporting

practices. Specifically, I focus on the importance of a change in accounting comparability

around adoption, relative to any concurrent change in financial reporting quality, in explaining

economic outcomes of adoption. Using accounting, analyst and market data for a large number

of firms over an eight year period from 2001 to 2008, I examine variation in the empirical

association between mandatory IFRS adoption and several market and analyst proxies for

information asymmetry, conditional on changes in accounting comparability and accounting

quality around the adoption date. After controlling for other influential factors at the firm and

macro level, the results show that economic benefits to IFRS adoption are either restricted to, or

most pronounced among, firms that exhibit an increase in cross-country accounting

comparability. In contrast, changes in firm-specific accounting quality appear to have only

second-order effects that are generally limited to those adopters with a concurrent increase in

comparability.

1
Two exceptions are Byard et al. (2011) and Horton et al. (2013). Byard et al. study the effect of firms’ pre-adoption
characteristics, in addition to country-level institutions, on changes in analyst forecast errors and dispersion around
adoption. Horton et al. (2013) consider variation in the association between mandatory IFRS adoption and analyst
forecast errors, conditional on the portfolios of firms covered by specific analysts.

1
 I focus on the two financial reporting constructs, accounting comparability and

accounting quality, for several reasons. First, improvements in both cross-country accounting

comparability and firm-specific financial reporting quality are primary stated objectives of the

International Accounting Standards Board (IASCF [2001]). Therefore, it is important to

understand their respective roles in shaping the information environment of financial statement

users following adoption. Second, while the two constructs are both desirable qualities of

financial reporting, they differ in the information they can convey about the firm. Comparability

describes the degree of similarity in accounting choices among two or more firms, reflecting “the

relationship between two or more pieces of information” (FASB [1980]). As such, it is different

from commonly studied accounting quality measures, such as discretionary accruals, accrual

quality and income smoothing. These earnings attributes are firm-specific and often computed

independently of the attributes of other firms (De Franco et al. [2011]). Moreover, comparability

is unique in its enhanced ability to facilitate benchmarking across firms. Third, a number of

studies have provided evidence that mandatory IFRS adoption has differential implications for

the two financial reporting constructs; with an average increase in accounting comparability (Yip

and Young [2012]) and an average decrease in accounting quality (Christensen et al. [2008];

Ahmed et al. [2012]; Capkun et al. [2012]). Fourth, prior studies have shown that both

accounting information quality and accounting information comparability are associated with

empirical proxies for information asymmetry (e.g., Welker [1995], Brown and Hillegeist [ 2007],

DeFranco et al. [2011], Lang et al. [2012]). However, no studies that I am aware of attempt to

assess the relative importance of the two financial reporting constructs with respect to each other,

despite conjecture that the two might be related (Barth et al [2012]).

2
 The economic outcome variables that I include in this study are intended to capture

complimentary aspects of changes in information asymmetry around mandatory IFRS adoption. I

focus on four previously studied economic outcomes of adoption: firm valuations, stock

liquidity, analyst accuracy in forecasting income, and analyst agreement in forecasting income. I

use Tobin’s Q to measure firm valuation and the proportion of zero return days to measure stock

liquidity. I measure absolute analyst forecast errors using the average consensus earnings

forecast and analyst forecast dispersion using the standard deviation of analysts’ earnings

forecasts. Together, these should reflect, among other things, the level of information

asymmetry among both sophisticated and unsophisticated investors.

Ex ante, it is unclear how changes in financial reporting practices under IFRS would

influence these outcomes. On the one hand, adopters that exhibit an increase in cross-country

accounting comparability (Yip and Young [2012]) should have enhanced information

environments, reduced information asymmetry among investors, and enhanced economic

outcomes. On the other hand, prior literature reports an average increase in income smoothing

among adopters, pointing to a decrease in reporting quality (Christensen et al. [2008]; Ahmed et

al. [2012]). However, analysts are better able to forecast smoother income streams (He at al.

[2010]), and stock liquidity is greater for firms with characteristics that result in smoother

income (Lang et al. [2012]). Adding to the discussion, Barth et al. (2012) conjecture that even if

comparability increases under IFRS, a potential source of that improvement is increased income

smoothing. Thus, a primary purpose of my study is to isolate, and differentiate between,

economic outcomes related to changes in comparability versus economic outcomes related to

changes in reporting quality.

3
 I use a research design in which I partition mandatory IFRS adopters into four distinct

treatment groups based on the sign of the firm’s concurrent changes in both cross-country

comparability and reporting quality around adoption. I refer to treatment firms with an increase

in comparability as ‘High Comp’ adopters and treatment firms with a decrease in comparability

as ‘Low Comp’ adopters. Similarly, I refer to treatment firms with an increase in reporting

quality as ‘High RQ’ adopters and treatment firms with a decrease in reporting quality as ‘Low

RQ’ adopters. I assess the cross-country comparability of a firm with its industry peers using

three related measures. Each measure follows the same underlying logic of the FASB’s and

IASB’s conceptual frameworks that two firms have more comparable accounting if they report

similar accounting amounts when they experience similar economic outcomes (and report

different accounting amounts when they experience different economic events). Specifically, the

three measures are based on the similarity in firms’: earnings-return relation; relation between

price and equity book value and earnings; and relation between earnings and subsequent year

cash flow (De Franco et al. [2011]; Barth et al. [2012]). I assess the financial reporting quality of

a firm using three measures: income smoothing measured as the firm-level correlation between

accruals and cash flow; accrual quality based on the model in Dechow and Dichev (2002); and

accrual quality as in Wysocki (2009).

My treatment sample includes 1,861 mandatory IFRS adopters. I control for omitted

events that could potentially impact the treatment firms during the sample period with a

benchmark control sample of 6,652 non-adopters. The benchmark sample controls for

confounding macroeconomic events unrelated to IFRS adoption and, accordingly, my research

design draws inferences based on the outcomes of IFRS adopters relative to the benchmark

firms. I use a bifurcated sample period that straddles the mandatory IFRS adoption date of

4
January 1, 2005: a pre-adoption period (2001-2004) in which the treatment sample uses their

home-country domestic accounting standards; and a post-adoption period (2005-2008) in which

the treatment sample uses IFRS

I estimate panel regressions that account for time varying firm characteristics, separate

industry-year fixed effects, and country-fixed effects. In this analysis, I place each adopter into

one of four distinct treatment groups based on both the sign of the firm’s change in comparability

and the sign of the firm’s change in reporting quality around adoption.2 For each of the three

measures of comparability I find a statistically and economically significant increase in firm

value and liquidity, and decrease in forecast error and dispersion, among High Comp adopters.

Moreover, this result holds regardless of how the firm’s reporting quality changes around

adoption. In contrast, in 8/9 test specifications I find no statistically significant change in firm

value, forecast error, or forecast dispersion among Low Comp adopters. Again, this result holds

regardless of how the firm’s reporting quality changes around adoption. I do find that the

liquidity effect of IFRS is generally broader and evident for firms with an increase in either

comparability or reporting quality, however it is still more pronounced among High Comp

adopters.

As an additional test, I compare the size of the IFRS effect on the outcome variables

across the four treatment groups and find that economic outcomes for adopters do not generally

depend on how their reporting quality changes around adoption, but when it does matter the

effect appears to be restricted to adopters with an increase in comparability. In contrast,

increased comparability appears to have a first-order effect on the economic outcomes of

adoption.

2
I asses reporting quality in this analysis using the correlation of accruals and cash flows (i.e. income smoothing).
In additional analysis, reported below, I asses reporting quality using the two measures of accrual quality

5
 To assess the robustness of my results I perform two additional tests. First, I use the

alternative measures, discussed above, to assess reporting quality. Overall, these results confirm

the inferences from my primary tests. Second, I restrict the analysis to adopters in countries with

strong institutions and reporting incentives. An alternative explanation for my primary findings

is that both increased comparability and positive economic outcomes following IFRS adoption

are a result of stronger institutions. If so, then I would not expect to find any effect of

comparability on the economic outcomes of IFRS adoption in those institutional environments. I

continue to find that the increase in firm value and the decrease in forecast errors and dispersion

are restricted to High Comp adopters. This holds for treatment samples restricted to countries in

the EU, countries with strong legal enforcement, and European countries that initiated proactive

reviews in 2005. I find also that the liquidity effects of adoption become broader across

treatment samples as the quality of their institutions increases. Nonetheless, I continue to find

that the liquidity effects of adoption are most pronounced among High Comp adopters.

While I find consistent results for a range of market-based and analyst-based outcomes of

mandatory IFRS adoption, it is possible that any effects I observe are temporary and will not

generalize to later periods or other IFRS adopters. Additionally, I cannot be sure that differences

in the returns and price generation process (i.e. differences in the speed with which price

captures “true” economic income) across countries and industries is not affecting my results, via

their impact of my market-based comparability measures. Also, returns are forward-looking

while accounting income is retrospective. Thus timing differences between returns and net

income may be affecting my results; although similarities and differences in the timeliness of

accounting’s recognition of economic events is, arguably, a primary determinant of whether

6
firms’ accounting is more or less comparable. Moreover, I find consistent results when using a

measure of comparability that is not based on market data.

With these caveats in mind, my study makes several contributions to the literature on

IFRS. To my knowledge, this study is the first to link capital market benefits of mandatory IFRS

adoption to improvements in cross-country comparability. In a related paper, Horton et al.

(2013) study the importance of individual analyst’s portfolios on the impact of IFRS adoption on

forecast accuracy. Specifically, Horton et al. (2013) allow the impact of mandatory adoption on

forecast errors to vary with whether analysts’ portfolios included firms following multiple non-

IFRS GAAP prior to adoption and only IFRS after adoption. The authors find a smaller increase

in forecasts errors for analysts in this group, relative to all analysts on average, and interpret this

result as consistent with a comparability effect on mandatory IFRS adoption. However, the

authors’ results in this analysis fail to find a negative association between mandatory IFRS

adoption and forecast errors, regardless of analysts’ portfolios (their Table 7 Panel B). In

contrast, I show that the consensus forecast error (and forecast dispersion) decreased among

mandatory IFRS adopters with an increase in three explicit proxies for comparability. Given the

growing global importance of IFRS in financial reporting, my study should be of interest to

countries’ policy makers and regulators that have adopted, or are considering the adoption of,

these standards.

My study is novel in that it also isolates the impact of a change in comparability from

concurrent changes in firms’ reporting quality. Prior studies indicate that IFRS adoption may

involve some trade-off between the quality of firms’ financial reporting and the comparability of

that financial reporting with industry peers. My findings highlights that the information

processing benefit to investors of more readily compared financial statements might be larger

7
than any information processing costs imposed by lower overall quality. Finally, my study

illustrates and validates an adaption of the De Franco et al. (2011) measure of comparability that

relies on annual data. This should be of value in international research when data on firms’

quarterly earnings is not readily available.

The remainder of the paper is organized as follows. Section 2 provides background and

develops my hypotheses. In sections 3 I discuss my sample and research design. Section 4

describes the data and presents the results. Section 5 concludes.

II. BACKGROUND AND HYPOTHESIS

The IASB Framework for the Preparation and Presentation of Financial Statements

describes comparability as the qualitative characteristic that enables users to evaluate the relative

financial position and performance of different entities and lists comparability as one of three

financial statement characteristics to be required under IFRS (IASCF [2001]). Further, the

FASB asserts that investors, lenders and other creditors should benefit from the increased

comparability that would result from internationally converged accounting standards (FASB

[2010]). Moreover, the SEC has promoted a single set of high-quality globally accepted

accounting standards, which would reduce country-by-country disparities in financial reporting,

according to current Chairman Mary L. Shapiro.3 Moreover, a pre-IFRS practitioner survey of

national accounting rules concluded that requiring EU companies to prepare their financial

statement in accordance with IFRS, as well as adoption and general convergence in other

3
SEC press release 2010-27 (Feb. 24, 2010) Available at http://www.sec.gov/news/press/2010/2010-27.htm

8
countries, would lead to a “significant improvement in financial reporting transparency and

comparability” (GAAP 2001 [2001]). 4

These improvements in financial reporting quality and accounting comparability, if

realized, were expected to increase the quality of firms’ disclosures and likely provide several

capital market and information environment benefits. Prior research finds that disclosure quality

is negatively associated with information asymmetry (Welker [1995], Healy et al. [1999], and

Leuz and Verrecchia [2000]) and that this relationship is primarily caused by a lower likelihood

that investors discover and trade on private information when disclosure quality is higher (Brown

and Hillegeist [2007]). Ex ante, mandatory IFRS adoption was expected to improve firms’

disclosure quality through both an increase in financial reporting quality and an increase in cross-

country accounting comparability. However, the evidence to date is mixed and indicates that

reporting quality decreased, or was unchanged, among mandatory adopters relative to non-

adopters (Christensen et al. [2008]; Ahmed et al. [2012]). This decrease in reporting quality

suggests that the previously documented reduction in information asymmetry and increase in

liquidity and market valuations following IFRS adoption (Daske et al [2008], Li [2010]) are not

due to improvements in reporting quality, per se.

In contrast, recent research provides evidence that is consistent with anticipated and

actual comparability benefits to the mandatory introduction of IFRS in 2005. Armstrong et al.

(2010) examine price reactions surrounding events that increased the probability of eventual

IFRS adoption in Europe and document price appreciation among EU firms that already have

high quality information. The authors attribute this appreciation to anticipated comparability

effects. Using a similar approach, Drake et al. (2010) document an increase in liquidity

4
The accounting firms preparing the report include Andersen, BDO, Deloitte Touche Tohmatsu, Ernst & Young,
Grant Thornton, KPMG and PricewaterhouseCoopers.

9
following actual mandatory IFRS adoption that the authors attribute to comparability benefits.

Brochet et al. (2011) find that abnormal returns to insider information decline in the U.K.

following mandatory IFRS adoption and attribute this to an increase in comparability among the

U.K. firms and non-U.K. firms. Additionally, U.S. mutual funds altered their investment

portfolios based on reductions in the cross-country heterogeneity of accounting standards

(DeFond et al. [2011]). More recently, Yip and Young (2012) provide evidence that mandatory

IFRS adoption improves cross-country information comparability measured as the similarity in

accounting functions, degree of information transfer, and similarity of the information content of

earnings and of the book value of equity.

An increase in cross-country accounting comparability should facilitate comparisons

across firms and improve the information available to investors. In turn, this increase in publicly

available information should reduce the level of information asymmetry and increase liquidity

through both a decrease in the relative level of informed trading (Brown et al. [2004]; Brown and

Hillegeist [2007]) and a reduction in the incentives for private information searches (Verrecchia

[1982]; Diamond [1985]). Moreover, an increase in a firm’s comparability with non-domestic

peers could improve its liquidity through an increase in foreigners’ willingness to invest (Merton

[1987]. Similar reasoning suggests that a decrease in information asymmetry due to more

comparable accounting will, ceteris paribus, increase firms’ valuations through a reduction in the

cost of equity capital (Healy et al. [1999]; Easley et al. [2002]; Easley at al. [2005]). My first

two hypotheses (in alternative form) are:

H1: The liquidity effects of mandatory IFRS adoption are explained by changes in cross-
country accounting comparability.

H2: The valuation effects of mandatory IFRS adoption are explained by changes in
cross-country accounting comparability.

10
 Next, I test the importance of improvements in cross-country comparability following

IFRS adoption on previously documented improvements in financial analysts’ information

environment (Byard et al. [2011], Horton et al. [2013]). In a U.S. setting, De Franco et al. (2011)

find support for their hypothesis that higher comparability results in a higher quality information

set that allows analysts to better evaluate firms’ economic performance and better understand

how that economic performance will be reflected in firms’ accounting performance. Moreover,

the higher level of publicly available information for firms with greater comparability should

reduce the information asymmetry between analysts as they place greater weight on that public

information, relative to private information. To the extent that analysts use similar forecasting

models, this greater reliance on the same public information should reduce disagreement among

analysts (De Franco et al. [2012]). My third hypothesis (in alternative form) is:

H3: The analyst information environment effects of mandatory IFRS adoption are
explained by changes in cross-country accounting comparability.

III. SAMPLE SELECTION AND RESEARCH DESIGN

Sample Selection

I obtain accounting data from Global and North America Compustat, market data from

Global Compustat and The Center for Research in Security Prices (CRSP), and analyst forecast

data from I/B/E/S. I use a bifurcated sample period that straddles the mandatory IFRS adoption

date of January 1, 2005: a pre-adoption period (2001-2004) in which the treatment sample uses

their home-country domestic accounting standards; and a post-adoption period (2005-2008) in

which the treatment sample uses IFRS.

11
 To form the treatment sample of mandatory IFRS adopters, I first select all non-financial

public firms present on Compustat Global during 2001 to 2008.5 I exclude firms that are not

present for the entire eight-year sample period and firms that Compustat codes as adopting IFRS

in a year other than the mandatory fiscal year. I also exclude firms with missing data to estimate

COMPACCT and ρ(ACC, CF), the least restrictive measures of comparability and reporting

quality (discussed below), for the entire sample period. Finally, I exclude firms with missing

data to calculate Tobin’s Q and its associated control variables (discussed below) for the entire

sample period. My final constant treatment sample contains 1,861 firms (14,888 firm-years)

from 23 countries that adopted IFRS for the first time when it became mandatory in 2005.

I control for confounding events that could potentially impact the treatment firms during

the sample period with a benchmark sample of control firms. To form the benchmark sample, I

first select all non-financial public firms present on Compustat Global or Compustat North

America during 2001 to 2008 that are from countries that do not adopt IFRS during the sample

period. I exclude firms that are not present for the entire eight-year sample period and firms that

Compustat identifies as preparing their financial statements in accordance with IFRS during any

part of the sample period. Finally, I exclude firms with missing data to calculate Tobin’s Q and

its associated control variables for the entire sample period. My final constant benchmark

sample contains 6,652 firms (53,216 firm-years) from 18 countries. The benchmark sample

controls for confounding macroeconomic events unrelated to IFRS adoption and, accordingly,

my research design draws inferences based on the outcomes of IFRS adopters relative to the

benchmark firms. Table 1 Panel A provides greater detail about the sample selection for

5
I restrict the potential sample of treatment firms and control firms (discussed below) to countries included in Daske
et al. (2009) to increase the comparability of my results with theirs.

12
treatment firms. Panel B (Panel C) reports the sample composition by country for the treatment

(benchmark) firms.

Comparability Measures

I use three related measures to assess cross-country accounting comparability. Although

the measures differ in terms of their specific inputs and outputs, they all follow the same

underlying logic of the FASB’s and IASB’s conceptual frameworks that two firms, i and j, have

more comparable accounting if they report similar accounting amounts when they experience

similar economic outcomes (and report different accounting amounts when they experience

different economic events).

CompAcct

For my first measure I follow De Franco et al. (2011) who use an earnings-return

regression to estimate a firm’s mapping between economic income and accounting income. De

Franco et al. (2011) follow Ball et al. (2000) and maintain that share returns measure economic

income and use a time-series of 16 quarterly earnings-return observations. Because of data

constraints unique to an international setting, I use four years of annual data and estimate the

following equation at the firm level for both the pre-IFRS (2001-2004) and post-IFRS (2005-

2008) periods: 6

Earningsit = i + i Returnit + it . (1)

6
De Franco et al. (2011) estimate comparability among U.S. firms with readily available quarterly data. This data
are generally unavailable for non-U.S. firms.

13
 The subscript i refers to firm i and the subscript t refers to year t. Earnings is earnings

before extraordinary items scaled by market value of equity nine months prior to the fiscal year-

end. Return is the buy-and-hold percentage stock return from nine months prior to the fiscal

year-end to three months after the fiscal year-end. I require each firm to have available date for

the entire sample period (2001-2008) and winsorize the top and bottom 1% of the distributions of

Earnings and Return to reduce the influence of outliers. The coefficients ( i and i) are the

estimate of the accounting function for firm i during the four years included in each regression

and reflect how economic income (i.e., Return) is reflected in accounting income (i.e., Earnings).

Similarly, the accounting function for firm j is reflected by j and j, estimated using the

earnings and return for firm j.

The similarity of the functions for firm i and firm j represents the comparability of their

accounting. To estimate the similarity in functions, I predict firm i and firm j’s earnings using

their own function but assuming they experience the same economic income (i.e., Return).

Specifically, I calculate:

E(Earnings)iit = ̂ i + ̂ i*Returnit, (2)

E(Earnings)ijt = ̂ j + ̂ j*Returnit. (3)

E(Earnings)iit is the predicted earnings of firm i using firm i’s function and firm i’s return

in period t, and E(Earnings)ijt is the predicted earnings of firm j using firm j’s function and firm

i’s return in period t. Using the same return to compute both predicted earnings holds constant

economic income.

Next I compute the accounting comparability between firm i and firm j (Comp1ij) in both

the pre-IFRS and post-IFRS period as the negative value of the average absolute difference

14
between the predicted earnings using firm i’s and firm j’s accounting functions. I require that

firms i and j be in the same two-digit SIC code, share the same fiscal year-end date, and be from

different countries:7

C -1 x∑ - E( ) . (4)

Larger (i.e. less negative) values for COMP1ij indicate greater cross-country

comparability of firm i with firm j. Finally, I compute a firm level measure of accounting

comparability for both the pre-IFRS and post-IFRS periods by aggregating over all of the firm i –

firm j combinations for a given firm i. Specifically, I compute CompAccti as the median

COMP1ij for all firms j with firm i.8 I use this firm level measure in all analysis.

CompPrice

For my second measure I use a regression of stock price on earnings and equity book

value to estimate a firm’s mapping between economic outcomes and accounting amounts (see,

for example, Barth et al. [2012]). I use four years of annual data and estimate the following

equation at the firm level for both the pre-IFRS (2001-2004) and post-IFRS (2005-2008) periods:

Priceit = i + 1i Earningsit + 2i BVit + it . (5)

Price is year-end stock price scaled by stock price nine months prior to the fiscal year-

end9. Earnings is earnings before extraordinary items scaled by market value of equity nine

months prior to the fiscal year-end. BV is the year-end book value of equity scaled by market

7
I follow De Franco et al. (2011) and exclude holding firms and limited partnerships which I identify based on a
word search of firms’ Compustat name (e.g., Holding, Group, LP, etc.)
8
My CompAcct measure is the annual equivalent to De Franco et al. (2009)’s CompAcctInd measure.
9
I deflate current price by past price to mitigate the effect of scale issues related to differences in terms of general
share price levels (Brown et al. [1999]; Lang et al. [2006])

15
value of equity nine months prior to the fiscal year-end. As with CompAcct above, the similarity

of the functions for a given firm i and firm j represents the comparability of their accounting.

First, I calculate the fitted stock price for firm i and firm j using their own function but the

earnings and equity book value of firm i. Second, I compute the comparability between firm i

and firm j (COMP2ij) in both the pre-IFRS and post-IFRS period as the negative value of the

average absolute difference between the fitted stock prices. Larger (i.e. less negative) values for

COMP2ij indicate greater cross-country comparability of firm i with firm j. Third, I compute a

firm level measure of for both the pre-IFRS and post-IFRS periods by aggregating over all of the

firm i – firm j combinations for a given firm i. Specifically, I compute CompPricei as the

median COMP2ij for all firms j with firm i.

CompCF

My third measure of comparability does not rely on market data and uses a regression of

subsequent year’s cash flow on earnings to estimate a firm’s mapping between economic

outcomes and accounting amounts (see, for example, Barth et al. [2012]). I use four years of

annual data and estimate the following equation for both the pre-IFRS (2001-2004) and post-

IFRS (2005-2008) periods:

CFit+1 = i + i Earningsit + it . (6)

CFit+1 is cash flow in year t + 1 scaled by market value of equity three months after fiscal

year-end t, for firm i. Earningsit is earnings before extraordinary items scaled by market value of

equity nine months prior to the fiscal year-end, for firm i in year t. As with the two

comparability measures above, the similarity of the functions for a given firm i and firm j

represents the comparability of their accounting. First, I calculate the fitted subsequent year cash

16
flow for firm i and firm j using their own function but the earnings of firm i. Second, I compute

the comparability between firm i and firm j (COMP3ij) in both the pre-IFRS and post-IFRS

period as the negative value of the average absolute difference between the fitted cash flows.

Larger (i.e. less negative) values for COMP3ij indicate greater cross-country comparability of

firm i with firm j. Third, I compute a firm level measure of for both the pre-IFRS and post-IFRS

periods by aggregating over all of the firm i – firm j combinations for a given firm i.

Specifically, I compute CompCFi as the median COMP3ij for all firms j with firm i.

Dependent Variables

I approximate my measure of market valuation, Tobin’s Q (Q), using the ratio of market

value of assets to book value of assets. I measure market liquidity as the proportion of trading

days with zero daily stock returns out of all trading days during the firm’s fiscal year

(ZERO_RET). Infrequent trading, identified by days with no price movement, reflects illiquidity

in a firm’s stock due to relatively high trading costs that require new information to accumulate

over longer horizons before informed trading affects price (Lesmond et al. [1999]). Moreover,

the zero return measure permits me to use a large sample because stock prices are readily

available and measured consistently across markets relative to other measures such as bid-ask

spreads and volume (Lang et al. [2012]). A smaller proportion of days with zero returns

indicates greater liquidity.

I obtain analyst forecast data from I/B/E/S. The analyst forecast error (AFE) for each

firm-year equals: |Mean Forecast EPS – Actual EPS| / Price. Where Mean Forecast EPS is the

last I/B/E/S mean annual EPS forecast prior to the earnings announcement date, Actual EPS is

the actual annual EPS reported by I/B/E/S, and Price is the last available price reported by

17
I/B/E/S prior to the fiscal year-end. Analyst forecast dispersion (AFD) for each firm-year equals:

Standard Deviation of EPS Forecasts / Price. I multiply AFE and AFD by 100 so that they

reflect the percentage of Price.

Variables to Measure Reporting Quality

I want to isolate any impact of comparability on the economic outcomes of adoption from

any concurrent impact of reporting quality. Accordingly, I use two measures of reporting quality

that prior research indicates were impacted by mandatory IFRS adoption (e.g. Christensen et al

[2007], Chen et al. [2010], Ahmed et al. [2012]) and a third related measure proposed by

Wysocki (2009).

My primary measure of reporting quality, ρ(ACC, CF), is the firm-level correlation

between total accruals scaled by beginning total assets and cash flow from operations scaled by

beginning total assets. As firms increase accruals to create reserves during periods of high cash

flows, and draw down those reserves during periods of low cash flows, accruals and cash flows

will be negatively correlated. Moreover, Barth et al. (2012) conjecture that higher levels of

income smoothing following mandatory IFRS adoption (Ahmed et al [2012]) are a potential

source of increased comparability following adoption. I calculate ρ(ACC, CF) using four years

of annual data in both the pre-IFRS and post-IFRS periods. Larger (i.e. less negative) values of

ρ(ACC, CF) indicate less income smoothing.

A second measure of reporting quality, AQ1, is the standard deviation of residuals from

the pooled regression of accruals on prior year, current year, and subsequent year cash flows

(Dechow and Dichev [2002]; Francis et al. [2005]) during both the pre-IFRS and post-IFRS

periods:

18
ACCit = + 1 CFit-1 + 2 CFit + 3CFit+1 + it. (7)

ACC is the ratio of total annual accruals to beginning total assets. CF is the ratio of

annual operating cash flow to beginning total assets. I calculate the standard deviation of the

residuals using four years in both the pre-IFRS and post-IFRS periods. Smaller values of AQ1

indicate higher quality accruals.

A third measure of reporting quality is based on Wysocki (2009) who conjectures that the

Dechow-Dichev (2002) model will assign higher accounting quality to firms that systematically

engage in opportunistic earnings management and smoothing activities compared to firms that do

not engage in those activities. This follows from the negative correlation between current

accruals and cash flows embedded in the model. I following Wysocki (2009) and calculate an

alternative measure of accrual quality, AQ2, equal to the ratio of the standard deviation of

residuals from equation (7) and the standard deviation of residuals from a version of equation (7)

that includes only CFit as an explanatory variable. Higher values of AQ2 indicate higher quality

accruals.

Test Design

I use OLS to test whether changes in comparability around mandatory IFRS adoption

explain capital market and analyst outcomes of adoption. I classify treatment firms into four

groups based on the sign of the pre-post adoption change in comparability (using either

CompAcct, CompPrice, or CompCF, alternately) and reporting quality (using ρ(ACC, CF)) as

follows:10

10
I form partitions using the accrual quality measures, AQ1 or AQ2, as the measure of reporting quality in later
additional analysis.

19
 (1) a High COMP-High RQ group [ΔComparability > 0 and Δρ(ACC, CF) > 0 ],

(2) a High COMP-Low RQ group [ ΔComparability > 0 and Δρ(ACC, CF) ≤ 0 ],

(3) a Low COMP-High RQ group [ ΔComparability ≤ 0 and Δρ(ACC, CF) > 0 ], and

(4) a Low COMP-Low RQ group [ ΔComparability ≤ 0 and Δρ(ACC, CF) ≤ 0 ].

High (Low) COMP groups exhibit an increase (decrease) in comparability and High

(Low) RQ groups exhibit a decrease (increase) in income smoothing. This design accounts for

potentially systematic variation in how both comparability and reporting quality vary around

adoption, as well as their individual and joint impact on adoption outcomes. For example, if

increased comparability is the dominant driver of increased market valuations following IFRS

adoption, then I expect to observe an increase in valuations for all High Comp firms (i.e. High

COMP-High RQ and High COMP-Low RQ), but not for Low Comp firms (i.e. Low COMP-High

RQ and Low COMP-Low RQ). Conversely, if increased reporting quality is the dominant driver,

then I expect to observe an increase in valuations for all High RQ firms (i.e. High COMP-High

RQ and Low COMP-High RQ), but not for Low RQ firms (i.e. High COMP-Low RQ and Low

COMP-Low RQ). Moreover, this research design permits me to test for a moderating effect for

the change in comparability conditional on the change in reporting quality, as well as a

moderating effect for the change in reporting quality conditional on the change in comparability.

Finally, this design permits me to test for joint effects. For example, increased valuations may

only occur for those firms that exhibit an increase in comparability achieved through a reduction

in income smoothing.

To test H1, H2 and H3, I estimate the following two models pooled across all countries in

both the treatment and benchmark samples over the period 2001 to 2008:

20
Dependent Variableit = + 1IFRSit + jControlsit + Country FE + Industry-Year FE + it, (8)

Dependent Variableit = + 1IFRS * DHCOMP-HRQit + 2IFRS * DHCOMP-LRQit (9)

+ 3IFRS * DLCOMP-HRQit + 4IFRS * DLCOMP-LRQit + jControlsit

+ Country FE + Industry-Year FE + it.

The subscript i refers to firm i and the subscript t refers to year t. Dependent Variable

refers to Q, ZERO_RET, AFE, or AFD. IFRS is a dummy variable coded one for treatment firms

during the post-adoption period (i.e. 2005-2008), and zero otherwise. DHCOMP-HRQ is a dummy

variable coded one if a treatment firm is from the High COMP-High RQ group; similarly,

DHCOMP-LRQ, DLCOMP-HRQ and, DLCOMP-LRQ are dummy variables indicating a treatment firm is from

the High COMP-Low RQ, Low COMP-High RQ, and Low COMP-Low RQ groups, respectively.

I include country fixed effects, as well as, separate year fixed effects for each two-digit SIC

industry. This multi-trend specification eliminates shocks to the dependent variables common to

all firms (both treatment and benchmark) within each industry. This control prevents, for

example, potentially spurious results due to an exogenous shock to the dependent variables being

clustered among those industries in which treatment firms are more likely to exhibit an increase

in comparability or reporting quality. Equation (8) provides a baseline, while equation (9) tests

the impact of comparability and reporting quality changes around adoption. All continuous

variables are winsorized at the bottom and top 1% of their distributions and all test statistics are

based on robust standard errors clustered at the country level.

I include controls that are relevant to each dependent variable. For tests of Tobin’s Q (Q)

I control for firm total assets, leverage and asset growth. I measure total assets (ASSETS) as the

fiscal-year-end total assets in $US; leverage (LEV) as fiscal-year-end long-term debt scaled by

21
total assets; and asset growth (ASSET_GR) as the percentage annual change in fiscal-year-end

total assets.

For tests of the proportion of zero return days (ZERO_RET) I control for firm market

value of equity, share turnover and return variability. I measure market value of equity (MVE) as

the fiscal-year-end market value of common equity in $US; share turnover (TURNOVER) as the

fiscal year $US trading volume divided by MVE; and return variability (RET_VAR) as fiscal-year

standard deviation of monthly stock returns (I lag MVE, TURNOVER, and RET_VAR by one

year).

For tests of analyst forecast errors (AFE) and dispersion (AFD) I control for firm market

value of equity, book-to-market, analyst coverage, and forecast horizon. I measure market value

of equity (MVE) as above; book-to-market (BTM) as fiscal-year-end book value of common

equity scaled by market value of common equity; analyst coverage (COVERAGE) as the number

of estimates included in the last I/B/E/S mean annual EPS forecast prior to the earnings

announcement date; and forecast horizon (DAYS) as the number of days between the I/B/E/S

earnings announcement date and the last I/B/E/S mean EPS forecast prior to the earnings

announcement date. For the above tests, I use natural log transformations of ASSETS, MVE,

TURNOVER, RET_VAR, COVERAGE, and DAYS.

My final three control variables account for the similarity between firms with respect to

size, leverage, and growth options. I include these variables to control for the possibility that

both a firm’s information environment and accounting comparability (De Franco et al. [2011])

are associated with its similarity to industry peers. For each firm-year, I compute the ratio of the

smaller value of total $US assets to the larger value of total $US assets using each of its industry

peers. ASSET_RATIO equals the firm-year median value for the ratios. I follow an identical

22
procedure using leverage to compute LEV_RATIO and using market-to-book to compute

MTB_RATIO. Given that the measures are increasing in the similarity between a firm and its

industry peers, I expect ASSET_RATIO, LEV_RATIO and BTM_RATIO to be positively

(negatively) associated with Tobin’s Q (proportion of zero returns, forecast errors, and forecast

dispersion).

IV. SAMPLE DESCRIPTION AND RESULTS

Estimation of Comparability Measures

Table 2 reports descriptive statistics related to my estimation of the three comparability

measures (i.e. CompAcct, CompPrice, and CompCF). Panel A describes the variables used to

estimate equations (1), (5), and (6). The treatment sample consists of 14,888 firm years, evenly

balanced between the pre- and post-IFRS periods. Panel B presents descriptive statistics for the

3,722 estimations of equation (1) used to compute CompAcct. I estimate the equation twice for

each treatment firm, using four firm-years from either the pre- or post-IFRS periods. The mean

β1 coefficient is 0.05, indicating the expected positive relation between returns and earnings, and

the mean R2 is 44%. Using four years of quarterly data, De Franco et al. (2012) estimate a mean

β1 coefficient and R2 of 0.02 and 12%, respectively. Thus, my use of annual data appears to

generate sufficient explanatory power and to identify a larger, average, association between

returns and earnings. Panel C reports the 3,722 estimations of equation (5) used to compute

CompPrice. The mean β1 and β2 coefficients are 2.25 and 0.72, indicating that both earnings and

equity book value are positively associated with stock price, and the mean R2 is 76%. Panel D

reports the 3,722 estimations of equation (6) used to compute CompCF. The mean β1

coefficient is -0.26, indicating a negative average relation between current earnings and

23
subsequent cash flows, and the mean R2 is 34%. This is likely due to negative serial correlation

in the cash flow component of earnings.11

As I final test of the ability of equations (1), (5) and (6) to predict the relevant outcome

variable, I compute the predicted value of each outcome variable for each firm-year. Panel E

reports the correlation between actual and predicted values for earnings, price and next year’s

cash flows. The Pearson correlation between actual and predicted earnings, price and next year’s

cash flows are 0.84, 0.92 and 0.78, respectively. Spearman correlations are similar. Together

with the mean coefficient estimates and regression R2, this indicates that the three models due a

reasonable job of capturing the firm-specific association between the proxies for economic

events and the proxies for accounting amounts.

Sample Description

Table 3, panel A reports descriptive statistics for dependent and control variables for the

pooled sample. The mean (median) Tobin’s Q is 2.0 (1.2) and the average sample firm has

13.1% of daily stock returns equal to zero. The mean (median) forecast error equals 1.73%

(0.30%) of price, while the mean and median forecast dispersion equal 0.73% and 0.24% of

price, respectively.

Table 3, Panel B reports descriptive statistics for the comparability and reporting quality

variables for treatment firms in both the pre- and post-mandatory-adoption windows. The mean

(median) CompAcct increases from -0.150 to -0.095 (-0.096 to -0.057) following adoption. This

11
In further analysis, I decompose current earnings into cash flows and accruals and estimate firm-year regressions
of future cash flows on current earnings and current cash flows. The mean coefficient for both accruals and cash
flows is negative. However, t-tests show that only the mean coefficient on cash flows is significantly negative. This
suggests that the negative mean coefficient I find in the firm-level estimations of equation (6) is primarily due to
negative serial correlation in cash flows (see for example, Dechow 1994) that is not completely offset by current
accruals.

24
positive change in comparability is consistent with the findings in Yip and Young (2012) and

indicates that the average annual cross-country prediction error decreased from 15% of price to

about 9% of price. Among firms within a single country (i.e. U.S.) using the same accounting

standard (i.e. U.S. GAAP), De Franco et al. (2011) find a mean (median) quarterly prediction

error based on their variable CompAcctInd of about 2.5% (1.5%) of price, which equate to a

mean (median) annual prediction error of about 10% (6%) of price. These values are very

similar to the mean (median) of 9.5% (5.7%) I obtain among firms that report using IFRS. 12 I

also find that the mean (median) CompCF increases from -0.250 to -0.234 (-0.180 to -0.160)

following adoption. CompPrice also increases following adoption, although only the median

difference is significant.

The mean (median) ρ(ACC, CF) decreases from -0.65 to -0.72 (-0.88 to -0.91) consistent

with an average increase in income smoothing following IFRS adoption. Moreover, the size of

the decrease is similar to the decrease of 0.05 observed by Ahmed et al. (2012) in their pooled

analysis. The two accruals quality measures provide conflicting evidence on how reporting

quality changes around IFRS adoption. AQ1, based on the Dechow-Dichev model, indicates a

very small decrease. However, only the median change is marginally significant. In contrast,

AQ2, based on Wysocki (2009), indicates a significant mean and median increase in accrual

quality. These results suggest that the three reporting quality measures are picking up distinct

aspects of financial reporting. Overall, the dependent variables, comparability measures, and

reporting quality measures fall within plausible ranges.

Panel C reports frequencies for the partitions I use in my primary analysis. My

identification strategy results in treatment group sizes that allow for meaningful comparisons.

12
Yip and Young (2012) use a semi-annual version of the De Franco et al. (2011) measure but deflate net income
using total assets. Therefore a direct comparison to the values in their study is not possible. However, their Table 2
indicates a mean (median) cross-country prediction error of about 28% (4%) of assets for firms in the same industry.

25
For the maximum treatment sample, the percentage of adopters in the HCOMP-HRQ group is

33%, 20%, and 24% when I measure comparability using CompAcct, CompPrice, and CompCF,

respectively. The percentage of firms in the HCOMP-LRQ group is 44%, 32%, and 34%. The

percentage of firms in the LCOMP-HRQ group is 13%, 26%, and 22%. And the percentage of

firms in the LCOMP-LRQ group is 10%, 22%, and 20%.

Results

For each outcome variable I first estimate a baseline regression using equation (8) to

measure the average IFRS effect, reflected by the coefficient estimate on IFRS. Next, I estimate

equation (9) to measure the IFRS effect within treatment subsamples that I form using a 2 by 2

identification scheme. IFRS is interacted with four dummy variables (DHCOMP-HRQ, DHCOMP-LRQ,

DLCOMP-HRQ, DLCOMP-LRQ) that identify the four treatment groups based on the sign of the pre-post

adoption change in comparability and reporting quality (i.e. High Comp-High RQ; High Comp-

Low RQ; Low Comp-High RQ; and Low Comp-Low RQ).13,

Analysis of Valuation Effects

Table 4 reports the results from multiple regression analyses of equations (8) and (9) with

Q as the dependent variable. In these and remaining tests I report OLS coefficient estimates, t-

statistics (in parentheses) clustered by country, and p-values for tests of differences across

coefficients.

In column (1), the estimated coefficient on IFRS is positive and significant, indicating

that on average mandatory adopters exhibit an increase in firm value compared to the benchmark

13
In my primary analysis I assume that a decrease (increase) in income smoothing indicates an increase (decrease)
in reporting quality. I perform tests with the other two measures of reporting quality, AQ1and AQ2, in additional
analysis.

26
firms (Daske et al. [2008]). In columns (2) to (4), the estimated coefficients on the interaction of

IFRS with DHCOMP-HRQ and DHCOMP-LRQ are positive and significant, indicating that adopters with

an increase in comparability exhibit an increase in valuation, regardless of how reporting quality

changes around adoption. In contrast, the estimated coefficients on the interaction of IFRS with

DLCOMP-HRQ and DLCOMP-LRQ are insignificant across all three measures of comparability. Thus,

only adopters with an increase in comparability following adoption exhibit a positive IFRS effect

on firm value. I also test the equality of the primary coefficients to shed light on whether changes

in reporting quality might still be important even if they do not exhibit a dominant effect. First, I

confirm my finding above that the effect of IFRS adoption on Q varies with comparability.

However, the results also indicate that an increase in reporting quality does increase the

valuation effect of IFRS among adopters whose comparability also increases. As a whole, these

results indicate that improvements in comparability are a primary mechanism behind the effect

of IFRS adoption on firm value and support H1.

Analysis of Liquidity Effects

Table 5 reports the results using ZER0_RET as the dependent variable. In column (1), the

estimated coefficient on IFRS is negative and significant, indicating that on average mandatory

adopters exhibit a positive IFRS effect on liquidity (i.e. a decrease in the proportion of zero

return days) (Daske et al. [2008]). In columns (2) to (4), the estimated coefficients on the

interaction of IFRS and DHCOMP-HRQ and DHCOMP-LRQ are negative and significant, indicating that

adopters with an increase in comparability exhibit an increase in liquidity, regardless of how

reporting quality changes around adoption. Moreover, additional tests confirm that High Comp

adopters exhibit a larger increase in liquidity, relative to Low Comp adopters.

27
 In contrast to the results for valuation above, the estimated coefficients on the interaction

of IFRS and DLCOMP-HRQ are also negative and marginally significant. Moreover, the estimated

coefficients on the interaction of IFRS and DLCOMP-LRQ are also negative, but only marginally

significant when I measure comparability using CompCF. Thus, I find evidence that the average

liquidity effect of IFRS generalizes to adopters with either an increase in comparability or an

increase in reporting quality. However, in 5/6 specifications I fail to find evidence that an

increase in reporting quality positively affects liquidity. Taken as a whole, these results indicate

that changes in reporting quality do not appear to explain the liquidity effects of IFRS, however,

comparability still matters. For instance, when I measure comparability using CompAcct and

focus only on adopters with a concurrent increase in reporting quality (i.e. High RQ), the

decrease in zero return days associated with increased comparability is about 0.021 (i.e. 0.054 -

0.033), or about 44% of the average decrease of 0.048. Overall these results support H2, but

indicate that other factors not modeled in my analysis also impact the liquidity effect of IFRS.

Analysis of Analyst Forecast Effects

Table 6 Panel A reports the results using AFE as the dependent variable. In column (1),

the estimated coefficient on IFRS is negative and significant indicating a negative average effect

of IFRS on forecast errors.14 Consistent with the positive economic outcomes documented above,

the estimated coefficients on IFRS*DHCOMP-HRQ and IFRS*DHCOMP-LRQ in columns (2) to (4) are

all negative and significant, indicating that adopters with an increase in comparability exhibit a

14
The negative average IFRS effect on forecast errors that I find is not consistent with the insignificant effect
documented in Byard et al. (2011). However, and consistent with my results, Byard et al. (2008) do find that
forecast errors decrease for adopters. It is possible that the difference in the two results stems from the fact that they
use early voluntary IFRS adopters as a control sample, while I use non-adopters as a control. For example, Capkun
et al. (2012) find that voluntary adopters exhibit changes in accounting properties around the mandatory adoption
date (i.e. 2005) that are very similar to changes observed for mandatory adopters, while Ahmed et al. (2012) fails to
find a similar change among non-adopters.

28
decrease in forecast errors, regardless of how reporting quality changes around adoption. I also

find weak evidence that treatment firms with a decrease in comparability (measured using

CompCF) and reporting quality experience a decrease in forecast errors. Overall, the evidence

indicates that reduced forecast errors are generally restricted to adopters with an increase in

comparability; although greater income smoothing can, in at least one specification, result in

more accurate forecasts. Additional tests reaffirm these findings and also indicate that among

adopters with increased comparability, the reduction in forecast errors is most pronounced

among those firms with a concurrent increase in income smoothing. Panel B reports results

using AFD as the dependent variable. The results are similar to, but stronger than, those I find

with AFE. IFRS adoption appears to only be associated with a decrease in forecast dispersion

among adopters with an increase in comparability. Overall these result support H3, but also

indicate that changes in reporting quality appear to matter at the margin.

Taken together, the multivariate analysis shows that the economic outcomes of IFRS

adoption, measured using market valuations, stock liquidity, and analysts’ information

environment, are heterogeneous across firms. Using a research design that partitions adopter

firms into groups based on the sign of the concurrent changes in both comparability and

reporting quality, I provide evidence that positive market outcomes to IFRS adoption are

stronger among, and often restricted to, firms that exhibit an increase in comparability. These

results indicate that increased comparability appears to have a first-order effect on the economic

outcomes of adoption. In contrast, changes in reporting quality appear to have second-order

effects that are generally restricted to those adopters with increased comparability.

29
Additional Analysis

Alternative Proxies for Reporting Quality

Table 7 reports the results from multiple regression analyses of equation (9) using Q,

ZERO_RET, AFE, and AFD as the dependent variable in columns (1), (2), (3), and (4),

respectively. When forming the four treatment groups, I use CompAcct to measure

comparability in all tests. Panel A presents the results when I measure reporting quality using

AQ1 based on the Dechow-Dichev (2002) model. As with my primary tests above, increased

firm value and decreased forecast errors and dispersion are restricted to adopters with an increase

in comparability. Also similar to my primary tests, increased liquidity is restricted to adopters

with an increase in either comparability or reporting quality. Panel B presents the results when I

measure reporting quality using AQ2 based on the Wysocki (2009) model. The coefficient

estimates are very similar in sign, magnitude and, significance to what I find for AQ1 above.

Overall, these results confirm the inferences from my primary tests.

Alternative Treatment Samples Based on Country-Level institutions

The treatment firms that I use in my primary analysis are domiciled in countries with

substantially different institutional environments. Prior literature finds that adopters in countries

with higher quality institutions exhibit more pronounced positive economic outcomes of

adoption (Daske et al. [2008]; Li [2010]; Byard et al. [2011]; Christensen et al. [2012]). Thus, an

alternative explanation for my primary findings is that both increased comparability and positive

economic outcomes following IFRS adoption are a result of stronger institutions. If so, then I

would not expect to find any effect of comparability on the economic outcomes of IFRS

adoption in those institutional environments.

30
 I test this alternative explanation and Table 8 presents the results when using CompAcct

and ρ(Acc, CF) to measure comparability and reporting quality, respectively. I continue to find

that the increase in firm value and the decrease in forecast errors and dispersion are restricted to

High Comp adopters. This holds for treatment samples restricted to countries in the EU,

countries with strong legal enforcement, and European countries that initiated proactive reviews

in 2005.15 In contrast, I find that the liquidity effects of adoption become broader across

treatment samples as the quality of their institutions increases. Among EU adopters, I only find

evidence of increased liquidity when comparability increases. Among adopters in countries with

strong enforcement, the liquidity effect extends to firms with an increase in either comparability

or reporting quality. And among adopters in countries that initiated proactive reviews, the

liquidity effect extends to the entire treatment sample. Nonetheless, I continue to find that the

liquidity effects of adoption are most pronounced among firms with increased comparability,

confirming my primary tests. 16

V. CONCLUSION

I examine the accounting mechanisms linking the widespread mandatory introduction of

IFRS in 2005 to market and analyst outcomes of adoption (i.e. firm value, stock liquidity, and

analyst forecast errors and dispersion). I use a simple design to partition adopter firms into four

treatment groups based on the sign of the firm’s concurrent changes in both comparability and

reporting quality around adoption. To control for confounding events that could potentially

15
I include Norway in the treatment sample of EU countries. I identify a country as having strong legal enforcement
if its score on the Kaufman et al. (2007) Rule of Law measure for 2005 is above the sample median. Strong
enforcement countries include Australia, Austria, Denmark, Finland, Germany, Luxembourg, The Netherlands,
Norway, Sweden, Switzerland, and the UK. Countries initiating proactive reviews in 2005 include Germany,
Finland, The Netherlands, Norway and the UK (Christensen et al. [2012]).
16
I repeat the analysis using the omitted treatment firms. Specifically, I use treatment samples from: 1) Non-EU
countries, 2) Weak enforcement countries, or 3) countries that did not initiate proactive reviews in 2005. The results
(untabulated) support my primary inferences.

31
impact the treatment firms during the sample period, I use a benchmark control sample of non-

adopters. I show that, across all adopters, an increase in firm value and a decrease in forecast

errors and dispersion under IFRS are restricted to firms that exhibit an increase in comparability.

Although I find that the liquidity effect of IFRS is generally broader, it is still most pronounced

among adopters with an increase in comparability. In contrast, changes in reporting quality

around adoption appear to have only second-order effects that are generally limited to those

adopters with a concurrent increase in comparability. Overall, these results suggest that

improvements in cross-country comparability have a first-order effect on firms’ information

environments. Moreover, the evidence suggests that the information processing benefit to

investors of more readily compared financial statements following IFRS adoption is larger than

any information processing costs imposed by a concurrent decrease in the overall quality of

financial reporting.

32
 REFRENCES

AHMED, A., M. NEEL and D. WANG. ‘Does Mandatory Adoption of IFRS Improve
Accounting Quality: Preliminary Evidence?’ Forthcoming Contemporary Accounting
Research (2012).
ARMSTRONG, C. S., M. BARTH, A. JAGOLINZER and E. RIEDL. ‘Market Reaction to the
Adoption of IFRS in Europe.’ The Accounting Review 85 (1) (2010): 31-61.
BALL, R., S.P. KOTHARI, and A. ROBIN. ‘The Effect of International Institutional Factors on
Properties of Accounting Earnings.’ Journal of Accounting and Economics 29 (2000): 1-
51.
BARTH, M., W. LANDSMAN, M. LANG and C. WILLIAMS. ‘Are IFRS-Based and US
GAAP-Based Accounting Amounts Comparable?’ Journal of Accounting and Economics
54 (2012): 68-93.
BROCHET, F., A. JAGOLINZER and E. RIEDL. ‘Mandatory IFRS Adoption and Financial
Statement Comparability.’ Forthcoming Contemporary Accounting Research (2011).
BROWN, S., S. HILLIGEIST and K. LO. ‘Conference Calls and Information Asymmetry.’
Journal of Accounting & Economics 37 (2004): 343–366.
BROWN, S. and S. HILLIGEIST. ‘How Disclosure Quality Affects the Level of Information
Asymmetry.’ Review of Accounting Studies 12, No.2-3 (2007): 443-477.
BROWN, S., K. LO, and T. LYS. ‘Use of R2 in Accounting Research: Measuring Changes in
Value Relevance over the Last Four Decades.’ Journal of Accounting and Economics 28
(1999): 83-115.
BYARD, D., L. YING and Y. YU. ‘The Effect of Mandatory IFRS Adoption on Financial
Analysts’ Information Environment.’ Journal of Accounting Research 49 (1) (2011).
CAPKUN, V., D. COLLINS, W. DANIEL, and T. JEANJEAN. ‘Does Adoption of IAS IFRS
Deter Earnings Management?’ Working paper (2012).
CHEN, H., Q. TANG, Y. JIANG and Z. LIN. ‘The Role of International Financial Reporting
Standards in Accounting Quality: Evidence from the European Union.’ Journal of
International Financial Management & Accounting 21 (3) (2010): 220-278.
CHRISTENSEN, H., E. LEE, and M. WALKER. ‘Cross-sectional Variation in the Economic
Consequences of International Accounting Harmonization: The Case of Mandatory IFRS
Adoption in the UK.’ The International Journal of Accounting 42 (2007): 341-379.
CHRISTENSEN, H., E. LEE, and M. WALKER. ‘Incentives or Standards: What Determines
Accounting Quality Changes Around IFRS Adoption?’ Working paper (2008).
DASKE, H., L. HAIL, C. LEUZ, and R. VENDI. ‘Mandatory IFRS Reporting Around the
World: Early Evidence on the Economic Consequences.’ Journal of Accounting
Research 46 (5) (2008):1085-1142.
DECHOW, P. ‘Accounting Earnings and Cash Flows as Measures of Firm Performance: The
Role of Accounting Accruals.’ Journal of Accounting and Economics 18 (1994):3-42.

33
DECHOW, P. and I. DICHEV. ‘The Quality of Accruals and Earnings: The Role of Accrual
Estimation Errors.’ The Accounting Review 77 (Supplement) (2002): 35-59.
DEFOND, M., X. HU, M. HUNG and S. LI. ‘The Impact of Mandatory IFRS Adoption on
Foreign Mutual Fund Ownership: The Role of Comparability.’ Journal of Accounting
and Economics 51 (2011): 240-258.
DE FRANCO, G., S.P. KOTHARI and R.S. VERDI. ‘The Benefits of Financial Statement
Comparability.’ Journal of Accounting Research 49 (4) (2011).
DIAMOND, D. ‘Optimal Release of Information by Firms.’ Journal of Finance 40 (1985), 1071-
1094.
DRAKE, M., L. MYERS and L. YAO. ‘Are Liquidity Improvements Around the Mandatory
Adoption of IFRS Attributable to Comparability Effects or to Quality Effects?’ Working
paper (2010).
EASLEY, D., S. HVIDKJAER, and M. O'HARA. ‘Is Information Risk a Determinant of Asset
Returns?’ Journal of Finance 57 (2002):2185-2221.
EASLEY, D., S. HVIDKJAER, and M. O'HARA. ‘Factoring Information into Returns.’
Working paper (2005).
FASB. ‘Statement of Financial Accounting Concepts No. 8.’ 2010.
FRANCIS, J., R. LAFOND, P. OLSSON, and K. SCHIPPER. ‘Costs of Equity and Earnings
Attributes’. The Accounting Review 79 (2004): 967-1010.
GAAP 2001-A Survey of National Accounting Rules. 2001.
HE, W., B.SIDHU, and H.TAN. ‘Income Smoothing and Properties of Analyst Information
Environment.’ Working paper (2010).
HEALY, P., A. HUTTON, and K. PALEPU. ‘Stock performance and intermediation changes
surrounding sustained increases in disclosures.’ Contemporary Accounting Research 16
(1999): 485-520.
HORTON, J., G. SERAFEIM, I. SERAFEIM. ‘Does Mandatory IFRS Adoption Improve the
Information Environment?’ Contemporary Accounting Research 30 (2013): 388-423.
IASCF. ‘Framework for the Preparation and Presentation of Financial Statements.’ 2001.
KAUFMANN, D., A. KRAAY, and M. MASTRUZZI. ‘Governance Matters VI: Aggregate and
Individual Governance Indicators 1996-2006.’ The World Bank, (2007).
LANG, M., K. LINS, and M. MAFFETT. ‘Transparency, Liquidity, and Valuation: International
Evidence on When Transparency matters Most.’ Journal of Accounting Research 50 (3)
(2012): 729-774.
LANG,M., J. RAEDY, and W. WILSON. ‘Earnings Management and Cross Listing: Are
Reconciled Earnings Comparable to US Earnings?’ Journal of Accounting and
Economics 42 (2006): 255–83.
LESMOND, D., J. OGDEN, and C. TRZCINKA. ‘A New Estimate of Transaction Costs.’ The
Review of Financial Studies 12 (5) (1999): 1113-1141.

34
LEUZ, C., and R. VERRECCHIA. ‘The Economic Consequences of Increased Disclosure.’
Journal of Accounting Research 38 (2000): 353-386.
LI, S. ‘Does Mandatory Adoption of International Financial Reporting Standards in the
European Union Reduce the Cost of Equity Capital?’ The Accounting Review 85 (2)
(2010): 607-636.
MERTON, R. ‘A Simple Model of Capital Market Equilibrium with Incomplete Information.’
Journal of Finance 42 (1987), 483–510.
VERRECCHIA, R. ‘Information Acquisition in a Noisy Rational Expectations Economy.’
Econometrica 50 (1982): 1415-1430.
WELKER, M. ‘Disclosure Policy, Information Asymmetry, and Liquidity in Equity Markets.’
Contemporary Accounting Research 11 (1995): 801-27.
WYSOCKI, P. ‘Assessing Earnings and Accruals Quality: U.S. and International Evidence.’
Working paper, 2009.
YIP, R. and D. YOUNG. ‘Does Mandatory IFRS Adoption Improve Information
Comparability?’ Accounting Review 87 (5) (2012):1767-1789.

35
 Table 1: Sample Composition
Panel A: Treatment Sample Selection
Non-financial firms from Compustat Global 9,231

Delete: Firms with missing years from 2001-2008 (5,198)

Delete: Firms that do not adopt IFRS in mandatory year (985)

Delete: Firms with missing data to estimate COMPACCT and ρ(ACC, CF) (1,152)

Delete: Firms with missing data for Tobin's Q and associated control variables (35)

Number of firms in treatment sample 1,861

Panel B: Treatment Sample Composition by Country

No. of
Percent No. of Firms Percent
Firms
Australia 184 9.9% Luxembourg 5 0.3%
Austria 13 0.7% Norway 52 2.8%
Belgium 33 1.8% Philippines 18 1.0%
Czech Republic 1 0.1% Poland 25 1.3%
Denmark 36 1.9% Portugal 19 1.0%
Finland 60 3.2% South Africa 76 4.1%
France 259 13.9% Spain 61 3.3%
Germany 166 8.9% Sweden 129 6.9%
Greece 48 2.6% Switzerland 26 1.4%
Hong Kong 79 4.2% The Netherlands 80 4.3%
Ireland 24 1.3% United Kingdom 357 19.2%
Italy 110 5.9% Total 1,861 100%

Panel C: Benchmark Sample Composition by Country

No. of
Percent No. of Firms Percent
Firms
Argentina 25 0.4% Malaysia 334 5.0%
Brazil 70 1.1% Mexico 35 0.5%
Canada 477 7.2% Peru 6 0.1%
Chile 51 0.8% South Korea 196 2.9%
Columbia 5 0.1% Sri Lanka 4 0.1%
Egypt 5 0.1% Taiwan 155 2.3%
India 83 1.2% Thailand 165 2.5%
Indonesia 85 1.3% United States 2,980 44.8%
Israel 17 0.3% Total 6,652 55%
Japan 1,959 29.4%
Panel A shows my sample selection for treatment firms and panel B shows the distribution of treatment firms across
countries. The treatment sample consists of 1,861 firms that switched from their domestic accounting standards to IFRS
for fiscal years beginning on or after January 1, 2005. Panel C shows the distribution of control firms across countries.
The control sample consists of 6,652 firms that do not use IFRS over the sample period.

36
 Table 2: Comparability Measures
Panel A: Descriptive Statistics for Variables Used in Regressions to Estimate Comparability Measures

Variable N Mean Std. Dev. 10th Percent Median 90th Percent

Earningst 14,888 0.01 0.23 -0.17 0.05 0.15
Returnt 14,888 0.11 0.56 -0.53 0.03 0.77
BVt 14,888 0.77 0.71 0.19 0.57 1.56
Pricet 14,888 1.02 0.51 0.47 0.97 1.52
CFt+1 14,288 0.13 0.33 -0.11 0.10 0.41

Panel B: Descriptive Statistics from Estimations of Equation (1) CompAcct ( = α + β1 Re u + ε)

Variable N Mean Std. Dev. 10th Percent Median 90th Percent

Intercept (α) 3,722 0.00 0.18 -0.17 0.05 0.11
β1 coefficient 3,722 0.05 0.27 -0.11 0.03 0.23
Regression R2(%) 3,722 44.10 32.02 1.19 42.05 89.22

Panel C: Descriptive Statistics from Estimations of Equation (5) CompPrice ( ce = α + β1 + β2 BV + ε)

Variable N Mean Std. Dev. 10th Percent Median 90th Percent

Intercept (α) 3,722 0.41 2.14 -0.99 0.35 1.82
β1 coefficient 3,722 2.25 41.59 -10.19 1.49 20.45
β2 coeff c e 3,722 0.72 10.24 -2.81 0.55 4.19
Regression R2(%) 3,722 75.63 26.60 31.92 86.26 99.50

Panel D: Descriptive Statistics from Estimations of Equation (6) CompCF (CFt+1 = α + β1 + ε)

Variable N Mean Std. Dev. 10th Percent Median 90th Percent

Intercept (α) 3,572 0.14 0.42 -0.19 0.11 0.50
β1 coefficient 3,572 -0.26 5.03 -2.94 -0.18 2.63
Regression R2(%) 3,572 34.09 30.65 0.91 25.19 83.80

Panel E: Correlations between Dependent Variables and Predicted Values Obtained from Equations (1), (5) and (6)

Pearson Correlation Spearman Correlation

E[Earningst] E[Pricet] E[CFt+1] E[Earningst] E[Pricet] E[CFt+1]
Earningst 0.843 0.849
Pricet 0.916 0.879
CFt+1 0.783 0.743
This table reports descriptive statistics related to my estimation of three comparability measures (CompAcct, CompPrice,
and CompCF). Earnings is earnings before extraordinary items scaled by market value of equity nine months prior to the
fiscal year-end. Return is the buy-and-hold percentage stock return from nine months prior to the fiscal year-end to three
months after the fiscal year-end. Price is year-end stock price scaled by stock price nine months prior to the fiscal year-end.
BV is the year-end book value of equity scaled by market value of equity nine months prior to the fiscal year-end. CFt+1 is
cash flow in year t + 1 scaled by market value of equity three months after fiscal year-end t. Each regression is estimated
for each treatment firm in both the pre- and post-IFRS periods using four years of data. The expectation operator, E[ ],
indicates the predicted value obtained from the relevant regression.

37
Table 3: Descriptive Statistics
Panel A: Dependent and Control Variables
N Mean Std. Dev. 10% 25% Median 75% 90%
Q 68,104 2.004 3.269 0.774 0.945 1.217 1.787 3.068
ZERO_RET 54,016 11.86% 13.12% 0.79% 2.42% 7.22% 16.80% 29.24%
AFE 24,184 1.72% 5.35% 0.03% 0.09% 0.30% 0.97% 3.15%
AFD 24,165 0.73% 1.58% 0.04% 0.08% 0.24% 0.65% 1.66%
ASSET_RATIO 68,104 0.182 0.117 0.026 0.082 0.180 0.267 0.342
LEV_RATIO 68,104 0.766 0.206 0.624 0.745 0.814 0.860 0.903
MTB_RATIO 68,104 0.373 0.437 0.149 0.307 0.464 0.569 0.628
ASSETS 68,104 2,093 5,998 17 71 283 1,092 4,365
LEV 68,104 0.236 0.246 0.000 0.041 0.195 0.349 0.500
ASSET_GR 68,104 0.093 0.383 -0.173 -0.052 0.031 0.137 0.340
MVE 54,106 2,236 7,400 20 57 221 1,018 4,211
RET_VAR 54,106 0.137 0.095 0.053 0.074 0.110 0.168 0.250
TURNOVER 54,106 1.241 1.944 0.072 0.206 0.582 1.400 2.973
BTM 24,184 0.634 0.506 0.184 0.314 0.513 0.798 1.203
COVERAGE 24,184 7.90 6.31 1.00 3.00 6.00 11.00 17.00
DAYS 24,184 18.31 18.70 6.00 8.00 14.00 25.00 28.00

Panel B: Comparability and Reporting Quality Variables for Treatment Firms: Pre- and Post-Mandatory-Adoption Windows

Preperiod Postperiod Change (p-value)

Mean Mean Mean t-test
Firms [Median] [Median] [Median] [Rank Sum test]
(a) (b) (b) - (a)

CompAcct 1,861 -0.149 -0.095 0.054 <0.01

[-0.096] [-0.057] [0.039] [<0.01]

CompPrice 1,861 -1.107 -1.099 0.008 0.81

[-0.793] [-0.748] [0.045] [0.05]

CompCF 1,786 -0.250 -0.233 0.017 0.01

[-0.179] [-0.161] [0.018] [<0.01]

ρ(Acc, CF) 1,861 -0.653 -0.719 -0.066 <0.01

[-0.877] [-0.908] [-0.031] [<0.01]

AQ1 1,700 -0.049 -0.050 -0.001 0.71

[-0.035] [-0.038] [-0.003] [0.04]

AQ2 1,700 1.217 1.244 0.027 0.08

[1.149] [1.185] [0.036] [<0.01]

Continued

38
Table 3: Continued
Panel C: Frequencies for High (Low) Comp and High (Low) RQ Partitions for Treatment Firms
CompAcct CompPrice CompCF
Freq Pct. Freq Pct. Freq Pct.
High COMP 1,428 77% 975 52% 1,037 58%
Low COMP 433 23% 886 48% 749 42%
High RQ 857 46% 857 46% 824 46%
Low RQ 1,004 54% 1,004 54% 962 54%

High COMP-High RQ 612 33% 381 20% 430 24%

High COMP-Low RQ 816 44% 594 32% 607 34%
Low COMP-High RQ 245 13% 476 26% 394 22%
Low COMP-Low RQ 188 10% 410 22% 355 20%
Panel A provides pooled descriptive statistics for the dependent and control variables. The maximum treatment sample consists
of 1,861 firms (14,888 firm-years) that switched from their domestic accounting standards to IFRS for fiscal years beginning on
or after January 1, 2005. The maximum control sample consists of 6,913 firms (55,304 firm-years) that do not use IFRS over
the sample period. Q is ratio of Market Value of Assets to Book Value of Assets. ZERO_RET is the proportion of trading days
with zero daily stock return during the firm’s fiscal year. AFE is the absolute forecast error, (Actual Earnings-Mean Forecast | /
Stock Price). AFD is the dispersion of analysts’ forecasts, (Standard Deviation of Forecasts Stock Price). I multiply AFE and
AFD by 100. ASSET_RATIO is the median ratio of the smaller value of total $US assets to the larger value of total $US assets
using a firm’s industry peers.
LEV_RATIO is the median ratio of the smaller value of leverage (i.e. of Long-term Debt to Total Assets) to the larger value of
leverage using a firm’s industry peers. MTB_RATIO is the median ratio of the smaller value of MTB (i.e. ratio of Market Value
of Common Equity to Book Value of Common Equity) to the larger value of MTB using a firm’s industry peers. ASSETS is
fiscal year-end total assets in $US. LEV is the fiscal year-end ratio of Long-term Debt to Total Assets. ASSET_GR is the
percentage annual change in Total Assets. MVE is the fiscal year-end Market Value of Common Equity in $US. RET_VAR is the
Annual Standard Deviation of Monthly Stock Returns. TURNOVER is the annual US$ Trading Volume divided by Market
Value of Common Equity. BTM is the fiscal year-end ratio of Book Value of Common Equity to Market Value of Common
Equity. COVERAGE is the total number of analysts included in the consensus estimate to compute AFE. DAYS is the number of
days between the forecast date used to compute AFE and the earnings announcement date.

Panel B provides a comparison of the comparability and reporting quality measures for treatment firms between the pre-
adoption (2001-2004) and post-adoption (2005-2008) windows. CompAcct measures the cross-country similarity in a firm to its
industry peers with respect to the earnings-return relation. CompPrice measures the cross-country similarity in a firm to its
industry peers with respect to the relation between price and book value and earnings. CompCF measures the cross-country
similarity in a firm to its industry peers with respect to the relation between earnings and subsequent year cash flow. See
section 3.2 for an extended discussion of how I calculate these variables. ρ(ACC, CF) is the firm-level four-year correlation
between Total Accruals scaled by Total assets and Cash Flow from Operations scaled by Total Assets. AQ1 is the standard
deviation of four years of residuals from the pooled regression of accruals on prior year, current year, and subsequent year Cash
Flow from Operations (Dechow and Dichev [2002]). AQ2 is the ratio of AQ1 to the standard deviation of residuals from a
version of the Dechow and Dichev (2002) model that includes only current Cash Flow from Operations as an explanatory
variable (Wysocki [2009]). All variables are winsorized at the 1% and 99% levels to mitigate the influence of outliers.

Panel C presents frequencies, and percent of total, for the partitions I use in the analysis. High (Low) Comp indicates treatment
firm with an increase (decrease) in comparability between the pre- and post-IFRS periods. High (Low) RQ indicates treatment
firm with an increase (decrease) in reporting quality between the pre- and post-IFRS periods. I measure reporting quality using
ρ(ACC, CF), my proxy for income smoothing, consistent with my primary analysis.

39
 Table 4: Effect of Mandatory IFRS adoption on Tobin’s Q, Conditional on Changes in
Comparability and Reporting Quality Around Adoption
Various Comparability Measures
CompAcct CompPrice CompCF
(1) (2) (3) (4)
IFRS 0.238*
(1.81)
IFRS*DHCOMP-HRQ 0.399*** 0.399*** 0.446***
(2.99) (3.22) (3.12)
IFRS*DHCOMP-LRQ 0.262** 0.278** 0.318**
(2.03) (2.21) (2.44)
IFRS*DLCOMP-HRQ 0.041 0.216 0.121
(0.25) (1.37) (0.83)
IFRS*DLCOMP-LRQ -0.125 0.058 -0.006
(-0.80) (0.37) (-0.05)
Asset Ratio -3.834*** -3.829*** -3.838*** -3.854***
(-4.88) (-4.87) (-4.90) (-4.94)
Leverage Ratio -1.193 -1.191 -1.193 -1.187
(-1.10) (-1.10) (-1.10) (-1.09)
BTM Ratio -0.000*** -0.000*** -0.000*** -0.000***
(-9.01) (-9.00) (-9.00) (-9.04)
Ln(Assets) -0.447*** -0.447*** -0.447*** -0.449***
(-4.51) (-4.51) (-4.51) (-4.56)
Leverage 2.406*** 2.410*** 2.406*** 2.424***
(2.87) (2.88) (2.87) (2.90)
Asset Growth 0.625*** 0.623*** 0.624*** 0.626***
(5.80) (5.69) (5.76) (5.77)

Test for Differences across Coefficients (p-values):

Increase versus Decrease in Comparability

IFRS*DHCOMP-HRQ > IFRS*DLCOMP-HRQ <0.01 0.02 <0.01
IFRS*DHCOMP-LRQ > IFRS*DLCOMP-LRQ <0.01 0.02 <0.01
Increase versus Decrease in Reporting Quality
IFRS*DHCOMP-HRQ > IFRS*DHCOMP-LRQ <0.01 0.01 0.05
IFRS*DLCOMP-HRQ > IFRS*DLCOMP-LRQ 0.05 0.03 0.01

Fixed Effects: Country, Ind-year Yes Yes Yes Yes

Adj. R-squared 33.3% 33.3% 33.3% 33.4%
Observations 68,104 68,104 68,104 67,504
This table reports the results of testing the firm valuation effects of mandatory IFRS adoption conditional on the sign of
treatment firms’ change in comparability and reporting quality around adoption. The maximum treatment sample consists of
Continued

40
Table 4: Continued
1,861 firms (14,888 firm-years) that switched from their domestic accounting standards to IFRS for fiscal years beginning on or
after January 1, 2005.The maximum control sample consists of 6,652 firms (53,216firm-years) that do not use IFRS over the
sample period. The sample period includes fiscal years 2001-2008.

The dependent variable, Q, is ratio of Market Value of Assets to Book Value of Assets. IFRS is a dummy variable coded one for
treatment firms during the post-adoption period (i.e. 2005-2008), and zero otherwise. I partition my treatment firms based on
the sign their pre-post adoption change in both comparability and reporting quality. High (Low) COMP groups exhibit an
increase (decrease) in comparability and High (Low) RQ groups exhibit a increase (decrease) in reporting quality. DHCOMP-HRQ is
a dummy variable coded one if a treatment firm is from the High COMP-High RQ group; similarly, DHCOMP-LRQ, DLCOMP-HRQ
and, DLCOMP-LRQ are dummy variables indicating a treatment firm is from the High COMP-Low RQ, Low COMP-High RQ, and
Low COMP-Low RQ groups, respectively. I assess comparability using CompAcct, CompPrice, and CompCF in columns (2),
(3), and (4), respectively (See section 3.2 for a discussion of how I compute each comparability measure). I assess reporting
quality using ρ(ACC, CF), the firm-level four-year correlation between Total Accruals scaled by Total Assets and Cash Flow
from Operations scaled by Total Assets. I interpret higher values for ρ(ACC, CF) as indicating higher reporting quality (i.e. a
less negative correlation indicates lower levels of income smoothing).

I include the following control variables. ASSETS is fiscal year-end total assets in $US. LEV is the fiscal year-end ratio of Long-
term Debt to Total Assets. ASSET_GR is the percentage annual change in Total Assets. ASSET_RATIO is the median ratio of the
smaller value of total $US assets to the larger value of total $US assets using a firm’s industry peers. LEV_RATIO is the median
ratio of the smaller value of leverage (i.e. of Long-term Debt to Total Assets) to the larger value of leverage using a firm’s
industry peers. MTB_RATIO is the median ratio of the smaller value of MTB (i.e. ratio of Market Value of Common Equity to
Book Value of Common Equity) to the larger value of MTB using a firm’s industry peers. I include country fixed effects and
separate industry-year fixed effects. All continuous variables are winsorized at the 1% and 99% levels to mitigate the influence
of outliers. I cluster on country to correct for the inflation in standard errors due to multiple observations from the same country.
Estimated coefficients are followed by t-statistics in parentheses. Significance levels at 10%, 5%, and 1%, two-tailed, are
indicated by ∗, ∗∗, and ∗∗∗, respectively. I also report one-tailed p-values from tests for differences across my primary
coefficients of interest.

41
 Table 5: Effect of Mandatory IFRS adoption on Proportion of Zero Return Days,
Conditional on Changes in Comparability and Reporting Quality Around Adoption
Various Comparability Measures
CompAcct CompPrice CompCF
(1) (2) (3) (4)
IFRS -0.048**
(-2.10)
IFRS*DHCOMP-HRQ -0.054** -0.055** -0.056**
(-2.15) (-2.22) (-2.13)
IFRS*DHCOMP-LRQ -0.050** -0.056** -0.053**
(-2.22) (-2.59) (-2.40)
IFRS*DLCOMP-HRQ -0.033* -0.043* -0.039*
(-1.76) (-1.94) (-2.03)
IFRS*DLCOMP-LRQ -0.033 -0.034 -0.038*
(-1.51) (-1.45) (-1.86)
Asset Ratio -0.016 -0.016 -0.015 -0.019
(-0.54) (-0.54) (-0.51) (-0.67)
Leverage Ratio -0.004 -0.004 -0.004 -0.005
(-1.06) (-1.05) (-1.04) (-1.12)
BTM Ratio -0.000 -0.000 -0.000 -0.000
(-0.98) (-0.96) (-0.96) (-1.02)
Ln(Market Valuet-1) -0.023*** -0.023*** -0.023*** -0.022***
(-4.38) (-4.39) (-4.38) (-4.43)
Ln(Return Variabilityt-1) -0.001 -0.001 -0.001 -0.001
(-0.50) (-0.52) (-0.53) (-0.49)
Ln(Share Turnovert-1) -0.022*** -0.022*** -0.022*** -0.022***
(-4.93) (-4.93) (-4.93) (-4.91)

Test for Differences across Coefficients (p-values):

Increase versus Decrease in Comparability

IFRS*DHCOMP-HRQ < IFRS*DLCOMP-HRQ 0.01 0.02 0.03
IFRS*DHCOMP-LRQ < IFRS*DLCOMP-LRQ <0.01 <0.01 <0.01
Increase versus Decrease in Reporting Quality
IFRS*DHCOMP-HRQ < IFRS*DHCOMP-LRQ 0.17 0.59 0.32
IFRS*DLCOMP-HRQ < IFRS*DLCOMP-LRQ 0.48 <0.01 0.43

Fixed Effects: Country, Ind-year Yes Yes Yes Yes

Adj. R-squared 56.2% 56.3% 56.3% 56.3%
Observations 54,016 54,016 54,016 53,448
This table reports the results of testing the stock liquidity effects of mandatory IFRS adoption conditional on the sign of
treatment firms’ change in comparability and reporting quality around adoption. The maximum treatment sample consists of
Continued

42
Table 5: Continued
1,749 firms (13,992 firm-years) that switched from their domestic accounting standards to IFRS for fiscal years beginning on or
after January 1, 2005. The maximum control sample consists of 5,003 firms (40,024firm-years) that do not use IFRS over the
sample period. The sample period includes fiscal years 2001-2008.

The dependent variable, ZERO_RET, is the proportion of trading days with zero daily stock return during the firm’s fiscal year.
IFRS is a dummy variable coded one for treatment firms during the post-adoption period (i.e. 2005-2008), and zero otherwise. I
partition my treatment firms based on the sign their pre-post adoption change in both comparability and reporting quality. High
(Low) COMP groups exhibit an increase (decrease) in comparability and High (Low) RQ groups exhibit a increase (decrease) in
reporting quality. DHCOMP-HRQ is a dummy variable coded one if a treatment firm is from the High COMP-High RQ group;
similarly, DHCOMP-LRQ, DLCOMP-HRQ and, DLCOMP-LRQ are dummy variables indicating a treatment firm is from the High COMP-
Low RQ, Low COMP-High RQ, and Low COMP-Low RQ groups, respectively. I assess comparability using CompAcct,
CompPrice, and CompCF in columns (2), (3), and (4), respectively (See section 3.2 for a discussion of how I compute each
comparability measure). I assess reporting quality using ρ(ACC, CF), the firm-level four-year correlation between Total
Accruals scaled by Total Assets and Cash Flow from Operations scaled by Total Assets. I interpret higher values for ρ(ACC,
CF) as indicating higher reporting quality (i.e. a less negative correlation indicates lower levels of income smoothing).

I include the following control variables. MVE is the fiscal year-end Market Value of Common Equity in $US. RET_VAR is the
Annual Standard Deviation of Monthly Stock Returns. TURNOVER is the annual US$ Trading Volume divided by Market
Value of Common Equity. I lag each control variable by one year. ASSET_RATIO is the median ratio of the smaller value of
total $US assets to the larger value of total $US assets using a firm’s industry peers. LEV_RATIO is the median ratio of the
smaller value of leverage (i.e. of Long-term Debt to Total Assets) to the larger value of leverage using a firm’s industry peers.
MTB_RATIO is the median ratio of the smaller value of MTB (i.e. ratio of Market Value of Common Equity to Book Value of
Common Equity) to the larger value of MTB using a firm’s industry peers. I include country fixed effects and separate industry-
year fixed effects. All continuous variables are winsorized at the 1% and 99% levels to mitigate the influence of outliers. I
cluster on country to correct for the inflation in standard errors due to multiple observations from the same country. Estimated
coefficients are followed by t-statistics in parentheses. Significance levels at 10%, 5%, and 1%, two-tailed, are indicated by ∗,
∗∗, and ∗∗∗, respectively. I also report one-tailed p-values from tests for differences across my primary coefficients of interest.

43
Table 6: Effect of Mandatory IFRS adoption on Analyst Forecast Errors and Dispersion,
Conditional on Changes in Comparability and Reporting Quality Around Adoption
Panel A: Analyst Forecast Error as the Dependent Variable
Various Comparability Measures
CompAcct CompPrice CompCF
(1) (2) (3) (4)
IFRS -1.072***
(-3.55)
IFRS*DHCOMP-HRQ -1.208*** -1.478*** -1.271***
(-3.61) (-4.65) (-3.95)
IFRS*DHCOMP-LRQ -1.734*** -1.725*** -1.769***
(-6.25) (-6.07) (-5.81)
IFRS*DLCOMP-HRQ 0.438 -0.269 -0.515
(0.52) (-0.55) (-1.22)
IFRS*DLCOMP-LRQ 0.593 -0.611 -0.847***
(1.14) (-1.55) (-3.05)
Asset Ratio -6.751*** -6.829*** -6.741*** -6.524***
(-11.08) (-11.54) (-10.74) (-8.72)
Leverage Ratio -0.692 -0.683 -0.690 -0.675
(-1.15) (-1.14) (-1.15) (-1.14)
BTM Ratio -0.166*** -0.165*** -0.165*** -0.160***
(-4.92) (-5.06) (-5.00) (-5.67)
Ln(Market Value) -0.766*** -0.766*** -0.769*** -0.734***
(-12.05) (-12.03) (-11.84) (-9.48)
BTM 1.719*** 1.683*** 1.698*** 1.726***
(10.60) (10.55) (10.72) (10.50)
Coverage -0.394*** -0.396*** -0.390*** -0.386***
(-2.92) (-2.90) (-2.89) (-2.97)
Days 0.371*** 0.369*** 0.371*** 0.356***
(7.52) (7.66) (7.63) (7.58)

Test for Differences across Coefficients (p-values):

Increase versus Decrease in Comparability

IFRS*DHCOMP-HRQ < IFRS*DLCOMP-HRQ 0.03 <0.01 0.03
IFRS*DHCOMP-LRQ < IFRS*DLCOMP-LRQ <0.01 <0.01 <0.01
Increase versus Decrease in Reporting Quality
IFRS*DHCOMP-HRQ < IFRS*DHCOMP-LRQ 0.99 0.90 0.99
IFRS*DLCOMP-HRQ < IFRS*DLCOMP-LRQ 0.40 0.82 0.87
Fixed Effects: Country,Ind-year Yes Yes Yes Yes
Adj. R-squared 20.2% 20.5% 20.4% 20.2%
Observations 24,184 24,184 24,184 24,000

44
Table 5: Continued
Panel B: Analyst Forecast Dispersion as the Dependent Variable
Various Comparability Measures
CompAcct CompPrice CompCF
(1) (2) (3) (4)
IFRS -0.210*
(-1.80)
IFRS*DHCOMP-HRQ -0.298** -0.256** -0.292**
(-2.42) (-2.35) (-2.38)
IFRS*DHCOMP-LRQ -0.354*** -0.300** -0.443***
(-3.25) (-2.59) (-3.68)
IFRS*DLCOMP-HRQ 0.270 -0.100 -0.022
(1.05) (-0.56) (-0.12)
IFRS*DLCOMP-LRQ 0.255 -0.156 0.069
(1.57) (-1.33) (0.53)
Asset Ratio -2.186*** -2.213*** -2.188*** -2.173***
(-10.13) (-10.30) (-10.19) (-10.26)
Leverage Ratio -0.197 -0.195 -0.197 -0.195
(-1.16) (-1.16) (-1.16) (-1.16)
BTM Ratio -0.039*** -0.039*** -0.039*** -0.039***
(-4.93) (-5.18) (-4.99) (-5.03)
Ln(Market Value) -0.310*** -0.310*** -0.310*** -0.304***
(-13.23) (-13.48) (-13.10) (-13.82)
BTM 0.665*** 0.652*** 0.662*** 0.660***
(6.17) (6.26) (6.19) (6.11)
Coverage 0.144*** 0.144*** 0.145*** 0.139***
(5.67) (5.71) (5.66) (5.47)
Days -0.010 -0.011 -0.010 -0.010
(-0.62) (-0.74) (-0.64) (-0.59)

Test for Differences across Coefficients (p-values):

Increase versus Decrease in Comparability

IFRS*DHCOMP-HRQ < IFRS*DLCOMP-HRQ <0.01 0.11 0.05
IFRS*DHCOMP-LRQ < IFRS*DLCOMP-LRQ <0.01 0.06 <0.01
Increase versus Decrease in Reporting Quality
IFRS*DHCOMP-HRQ < IFRS*DHCOMP-LRQ 0.84 0.75 0.99
IFRS*DLCOMP-HRQ < IFRS*DLCOMP-LRQ 0.53 0.66 0.01
Fixed Effects: Country, Ind-year Yes Yes Yes Yes
Adj. R-squared 28.6% 28.8% 28.6% 28.8%
Observations 21,465 21,465 21,465 21,354
Continued

45
Table 6: Continued
This table reports the results of testing the analyst effects of mandatory IFRS adoption conditional on the sign of treatment
firms’ change in comparability and reporting quality around adoption. Panel A reports the results for tests of analyst forecast
errors and panel B reports the results for tests of analyst forecast dispersion. For tests of analyst forecast error in panel A the
maximum treatment sample consists of 816 firms (6,528 firm-years) that switched from their domestic accounting standards to
IFRS for fiscal years beginning on or after January 1, 2005 and the maximum control sample consists of 2,207 firms (17,656
firm-years) that do not use IFRS over the sample period. . For tests of analyst forecast dispersion in panel B the maximum
treatment sample consists of 811 firms (5,720 firm-years) that switched from their domestic accounting standards to IFRS for
fiscal years beginning on or after January 1, 2005 and the maximum control sample consists of 2,198 firms (15,745 firm-years)
that do not use IFRS over the sample period. The sample period includes fiscal years 2001-2008.

The dependent variable in panel A, AFE, is the absolute forecast error, (Actual Earnings-Mean Forecast | / Stock Price). The
dependent variable in panel B, AFD, is the dispersion of analysts’ forecasts, (Standard Deviation of Forecasts Stock Price). I
multiply AFE and AFD by 100. IFRS is a dummy variable coded one for treatment firms during the post-adoption period (i.e.
2005-2008), and zero otherwise. I partition my treatment firms based on the sign their pre-post adoption change in both
comparability and reporting quality. High (Low) COMP groups exhibit an increase (decrease) in comparability and High (Low)
RQ groups exhibit a increase (decrease) in reporting quality. DHCOMP-HRQ is a dummy variable coded one if a treatment firm is
from the High COMP-High RQ group; similarly, DHCOMP-LRQ, DLCOMP-HRQ and, DLCOMP-LRQ are dummy variables indicating a
treatment firm is from the High COMP-Low RQ, Low COMP-High RQ, and Low COMP-Low RQ groups, respectively. I assess
comparability using CompAcct, CompPrice, and CompCF in columns (2), (3), and (4), respectively (See section 3.2 for a
discussion of how I compute each comparability measure). I assess reporting quality using ρ(ACC, CF), the firm-level four-year
correlation between Total Accruals scaled by Total Assets and Cash Flow from Operations scaled by Total Assets. I interpret
higher values for ρ(ACC, CF) as indicating higher reporting quality (i.e. a less negative correlation indicates lower levels of
income smoothing).

I include the following control variables. MVE is the fiscal year-end Market Value of Common Equity in $US. BTM is the fiscal
year-end ratio of Book Value of Common Equity to Market Value of Common Equity. COVERAGE is the total number of
analysts included in the consensus estimate to compute AFE. DAYS is the number of days between the forecast date used to
compute AFE and the earnings announcement date. ASSET_RATIO is the median ratio of the smaller value of total $US assets
to the larger value of total $US assets using a firm’s industry peers. LEV_RATIO is the median ratio of the smaller value of
leverage (i.e. of Long-term Debt to Total Assets) to the larger value of leverage using a firm’s industry peers. MTB_RATIO is
the median ratio of the smaller value of MTB (i.e. ratio of Market Value of Common Equity to Book Value of Common Equity)
to the larger value of MTB using a firm’s industry peers. I include country fixed effects and separate industry-year fixed effects.
All continuous variables are winsorized at the 1% and 99% levels to mitigate the influence of outliers. I cluster on country to
correct for the inflation in standard errors due to multiple observations from the same country. Estimated coefficients are
followed by t-statistics in parentheses. Significance levels at 10%, 5%, and 1%, two-tailed, are indicated by ∗, ∗∗, and ∗∗∗,
respectively. I also report one-tailed p-values from tests for differences across my primary coefficients of interest.

46
 Table 7: Alternative Proxies for Reporting Quality
Panel A: Dechow-Dichev Accruals Quality to Proxy for Reporting Quality

Q Zero Ret AFE AFD

(1) (2) (3) (4)
IFRS Variables:
IFRS*DHCOMP-HRQ 0.350** -0.055** -1.583*** -0.324**
(2.61) (-2.31) (-4.79) (-2.68)
IFRS*DHCOMP-LRQ 0.297** -0.050** -1.396*** -0.328***
(2.23) (-2.08) (-4.94) (-2.91)
IFRS*DLCOMP-HRQ 0.030 -0.043* -0.393 -0.011
(0.13) (-1.88) (-0.93) (-0.06)
IFRS*DLCOMP-LRQ -0.059 -0.029 0.914 0.399*
(-0.41) (-1.55) (1.15) (1.79)

Firm-level Control Variables: Yes Yes Yes Yes

Fixed Effects: Country, Ind-year Yes Yes Yes Yes
Adj. R-squared 32.8% 55.9% 20.5% 28.9%
Observations 68,104 54,106 24,184 21,465

Panel B: Wysocki (2009) Accruals Quality to Proxy for Reporting Quality

Q Zero Ret AFE AFD

(1) (2) (3) (4)
IFRS Variables:
IFRS*DHCOMP-HRQ 0.351** -0.059** -1.689*** -0.316***
(2.63) (-2.51) (-5.69) (-2.95)
IFRS*DHCOMP-LRQ 0.291** -0.045* -1.264*** -0.339**
(2.29) (-1.86) (-4.16) (-2.67)
IFRS*DLCOMP-HRQ -0.004 -0.031* 0.123 0.039
(-0.02) (-1.79) (0.17) (0.16)
IFRS*DLCOMP-LRQ -0.051 -0.034 0.778 0.437*
(-0.36) (-1.58) (0.99) (1.97)

Firm-level Control Variables: Yes Yes Yes Yes

Fixed Effects: Country, Ind-year Yes Yes Yes Yes
Adj. R-squared 32.8% 55.9% 20.5% 28.8%
Observations 68,104 54,106 24,184 21,465
This table repeats the primary analysis in tables 4, 5, and 6 using alternative measures of reporting quality. In panel A, I assess
reporting quality using AQ1, the standard deviation of four years of residuals from the pooled regression of accruals on prior
year, current year, and subsequent year Cash Flow from Operations (Dechow and Dichev [2002]). Smaller values of AQ1
indicate higher quality accruals.
Continued

47
Table 7: Continued
In panel B, I assess reporting quality using AQ2, the ratio of AQ1 to the standard deviation of residuals from a version of the
Dechow and Dichev (2002) model that includes only current Cash Flow from Operations as an explanatory variable (Wysocki
[2009]). Larger values of AQ2 indicate higher quality accruals (Sec. 3.4 for further discussion of how I compute AQ1 and
AQ2). I assess comparability in all tests using CompAcct. The maximum treatment sample consists of 1,700 firms (13,600
firm-years) that switched from their domestic accounting standards to IFRS for fiscal years beginning on or after January 1,
2005 and the maximum control sample consists of 6,913 firms (55,304 firm-years) that do not use IFRS over the sample period.
The sample period includes fiscal years 2001-2008. I use Q, ZERO_RET, AFE, and AFD as dependent variables. Control
variables are included but not tabulated. I include country fixed effects and separate industry-year fixed effects. All continuous
variables are winsorized at the 1% and 99% levels to mitigate the influence of outliers. I cluster on country to correct for the
inflation in standard errors due to multiple observations from the same country. Estimated coefficients are followed by t-
statistics in parentheses. Significance levels at 10%, 5%, and 1%, two-tailed, are indicated by ∗, ∗∗, and ∗∗∗, respectively.

48
 Table 8: Alternative Mandatory IFRS Adopter Samples
Q Zero Ret AFE AFD
(1) (2) (3) (4)
Mandatory IFRS Adopters from Countries in the European Union
IFRS*DHCOMP-HRQ 0.314** -0.060* -1.226*** -0.346**
(2.49) (-1.98) (-3.33) (-2.67)
IFRS*DHCOMP-LRQ 0.239* -0.054* -1.723*** -0.388***
(1.88) (-1.95) (-5.79) (-3.61)
IFRS*DLCOMP-HRQ -0.016 -0.035 0.496 0.165
(-0.10) (-1.56) (0.47) (0.60)
IFRS*DLCOMP-LRQ -0.204 -0.035 0.489 0.325*
(-1.37) (-1.35) (0.90) (1.82)
Observations 65,040 51,128 22,896 20,298

Mandatory IFRS Adopters from Countries with Strong Enforcement

IFRS*DHCOMP-HRQ 0.422*** -0.079** -1.059*** -0.279**
(3.06) (-2.29) (-3.03) (-2.33)
IFRS*DHCOMP-LRQ 0.305** -0.072** -1.676*** -0.389***
(2.43) (-2.36) (-8.09) (-3.58)
IFRS*DLCOMP-HRQ 0.085 -0.056** 0.209 0.112
(0.44) (-2.55) (0.16) (0.33)
IFRS*DLCOMP-LRQ -0.194 -0.047 0.453 0.272
(-1.07) (-1.58) (0.80) (1.27)
Observations 62,080 48,368 21,744 19,250

Mandatory IFRS Adopters from Countries that Begin Proactive Reviews in 2005
IFRS*DHCOMP-HRQ 0.214* -0.108** -1.120*** -0.390***
(1.95) (-2.84) (-3.13) (-3.40)
IFRS*DHCOMP-LRQ 0.141* -0.094** -1.744*** -0.416***
(1.83) (-2.59) (-10.08) (-4.57)
IFRS*DLCOMP-HRQ 0.006 -0.074*** -0.371 -0.239
(0.04) (-3.25) (-0.23) (-0.93)
IFRS*DLCOMP-LRQ -0.241 -0.073** 0.104 0.231
(-1.58) (-2.30) (0.15) (0.88)
Observations 58,936 45,392 20,408 18,039

Firm-level Control Variables: Yes Yes Yes Yes

Fixed Effects: Country, Ind-year Yes Yes Yes Yes
This table repeats the primary analysis in tables 4, 5, and 6 using alternative treatment samples. First, I restrict the treatment
sample to firms in the EU (including Norway). Second, I restrict the treatment sample to firms in countries with strong legal
enforcement. I identify a country as having strong legal enforcement if its score on the Kaufman et al. (2007) Rule of Law
measure for 2005 is above the sample median.
Continued

49
Table 8: Continued
Strong enforcement countries include Australia, Austria, Denmark, Finland, Germany, Luxembourg, The Netherlands,
Norway, Sweden, Switzerland, and the UK. Third, I restrict the treatment sample to countries initiating proactive reviews in
2005, including Germany, Finland, The Netherlands, Norway and the UK (Christensen et al. [2012]). I assess comparability
using CompAcct and reporting quality using ρ(ACC, CF) in all tests. The maximum EU treatment sample consists of 1,478
firms (11,824 firm-years), the maximum Strong Legal Enforcement treatment sample consists of 1,108 firms (8,864 firm-
years), and the maximum review initiation treatment sample consists of 715 firms (5,720 firm-years) that switched from their
domestic accounting standards to IFRS for fiscal years beginning on or after January 1, 2005. The maximum control sample
consists of 6,913 firms (55,304 firm-years) that do not use IFRS over the sample period. The sample period includes fiscal
years 2001-2008. I partition my treatment firms based on the sign their pre-post adoption change in both comparability and
reporting quality. High (Low) COMP groups exhibit an increase (decrease) in comparability and High (Low) RQ groups
exhibit a increase (decrease) in reporting quality. DHCOMP-HRQ is a dummy variable coded one if a treatment firm is from the
High COMP-High RQ group; similarly, DHCOMP-LRQ, DLCOMP-HRQ and, DLCOMP-LRQ are dummy variables indicating a treatment
firm is from the High COMP-Low RQ, Low COMP-High RQ, and Low COMP-Low RQ groups, respectively. I use Q,
ZERO_RET, AFE, and AFD as dependent variables. Control variables are included but not tabulated. I include country fixed
effects and separate industry-year fixed effects. All continuous variables are winsorized at the 1% and 99% levels to mitigate
the influence of outliers. I cluster on country to correct for the inflation in standard errors due to multiple observations from the
same country. Estimated coefficients are followed by t-statistics in parentheses. Significance levels at 10%, 5%, and 1%, two-
tailed, are indicated by ∗, ∗∗, and ∗∗∗, respectively.

50