network effects in countries adoption of ifrs

Electronic copy available at: http://ssrn.com/abstract=1590245

Copyright © 2010, 2011 by Karthik Ramanna and Ewa Sletten

Working papers are in draft form. This working paper is distributed for purposes of comment and discussion only. It may not be reproduced without permission of the copyright holder. Copies of working papers are available from the author.

Network Effects in Countries’ Adoption of IFRS Karthik Ramanna Ewa Sletten

Working Paper

10-092


Network Effects in Countries’ Adoption of IFRS*

Karthik Ramanna Harvard Business School

[email protected]

and

Ewa Sletten MIT Sloan School of Management

[email protected]

This draft: June 20, 2011

Abstract

If the differences in accounting standards across countries reflect relatively stable institutional differences (e.g., auditing technology, the rule of law, etc.), why did several countries rapidly, albeit in a staggered manner, adopt IFRS over local standards in the 2003–2008 period? We test the hypothesis that perceived network benefits from the extant worldwide adoption of IFRS can explain part of countries’ shift away from local accounting standards. That is, as more jurisdictions with economic ties to a given country adopt IFRS, perceived benefits from lowering transactions costs to foreign financial-statement users increase and contribute significantly towards the country’s decision to adopt IFRS. We find that perceived network benefits increase the degree of IFRS harmonization among countries, and that smaller countries have a differentially higher response to these benefits. The results, robust to numerous alternative hypotheses and specifications, suggest IFRS adoption is self-reinforcing, which, in turn, has implications for the consequences of IFRS adoption.

* We thank Michelle Hanlon, Bob Holthausen, S.P. Kothari, Christian Leuz, Sugata Roychowdhury, Doug Skinner, Thor Sletten, Eugene Soltes, Rodrigo Verdi, Ross Watts, Joe Weber, Gwen Yu, and seminar participants at Arizona State, Boston College, Emory, the Harvard Applied Statistics Workshop, Minnesota, MIT, Oregon, UCLA, the 2010 European Accounting Association Annual Meeting, the 2011 Harvard Business School International Research Conference, and the 2010 Penn State Accounting Conference for helpful suggestions and comments on this paper. The paper is based in part on our manuscript, “Why do countries adopt IFRS,” and has also benefited from comments we received on that manuscript, including from John Core, David Hawkins, Paul Healy, Shiva Rajgopal, Eddie Riedl, Terry Shevlin, D. Shores, Suraj Srinivasan, Steve Zeff, and seminar participants at Boston University, UC–Boulder, HBS, the 2009 GIA Conference at UNC–Chapel Hill, Rice, Toronto, and the University of Washington. Any errors are our responsibility.


1

I. Introduction

The International Accounting Standards Board (IASB) was established in 2001 to

develop International Financial Reporting Standards (IFRS). In the period from 2003 through

2008, nearly 50 countries (including the EU countries) mandated IFRS for all listed companies in

their jurisdictions. Further, in that period, another 15 countries either mandated IFRS for some

listed companies or allowed listed companies to voluntarily adopt IFRS. Another group of 16

countries initiated “convergence” projects with IFRS through 2008, and the list of states with

convergence projects planned beyond 2008 includes some of the world’s largest economies such

as China and the U.S.1 There has been considerable research into the consequences of a

country’s IFRS adoption for firms in its jurisdiction (e.g., Armstrong, Barth, Jagolinzer, and

Riedl, 2010; Daske, Hail, Leuz, and Verdi, 2008; DeFond, Hu, Hung, and Li, 2011; and Lang,

Maffett, and Owens, 2010); but less is known about why countries themselves adopt IFRS.2 We

develop and test the hypothesis that the intertemporal shift towards IFRS over local standards

can be explained, in part, by perceived network benefits, i.e., perceived lower transactions costs,

given the network of IFRS adopters. If true, the implication is that worldwide IFRS acceptance is

self-reinforcing, i.e., adoption begets adoption.

Prior literature argues that financial reporting standards have co-developed with a

country’s economic, political, and cultural institutions (e.g., Ball, Kothari, and Robin, 2000; Ball,

Robin, and Wu, 2003; Bushman and Piotroski, 2006; Bradshaw and Miller, 2008; and Hail,

Leuz, and Wysocki, 2010). In particular, the nature of standards comprising a reporting regime

reflects the monitoring and information processing capabilities of existing market institutions

1 Throughout the paper, “convergence” refers to a country’s efforts to reconcile its domestic accounting standards with IFRS, in lieu of directly adopting IFRS as issued by the IASB. Convergence projects often result in adopting IFRS with modifications and exceptions. 2 Hope, Jin, and Kang (2006) provide some preliminary evidence on the determinants of IFRS adoption through 2005 in a sample of 38 countries (including 14 EU countries).

2

such as auditing, securities laws, courts, and financial intermediaries. Under this view of

“institutional complementarities” (Leuz, 2010), a change in a country’s financial reporting

standards should accompany complementary changes in monitoring and information processing

institutions. However, the rapid proliferation in IFRS adoption across countries in the 2003–2008

period has seldom been accompanied by substantial changes in enforcement institutions.3

Holding institutions constant, the staggered adoption of IFRS over time can be due, in

part, to changing perceptions across countries of the benefits of IFRS. IFRS, as a globally

recognized body of standards, is expected to lower transaction costs associated with foreign users

of financial statements. By adopting IFRS, a country trades off status quo national standards that

reflect its market institutions for (in part) a perceived decrease in transactions costs from using

internationally recognized standards. As more jurisdictions with economic ties to a given country

adopt IFRS, the perceived benefits to that country from lowering transactions costs, and thus

from adopting IFRS, increase. Thus, we hypothesize that a country’s adoption of IFRS is related

to the magnitude of its economic relations with other countries that have adopted IFRS through

that date. We refer to this magnitude as the IFRS network value to a country at a given time.

The first step in an empirical test of IFRS network effects is identifying IFRS adoption

among countries. The nature of IFRS adoption by a country (e.g., converging local standards

with IFRS, permitting firms to voluntarily adopt IFRS, full IFRS adoption, etc.) varies across

jurisdictions and time. Thus, our primary IFRS adoption variable is an ordinal reflecting the

variety of possible adoption states. The variable takes three values: “1” for country-years with no

IFRS-related activities; “2” for country-years with convergence projects, country-years in which

voluntary IFRS adoption is permitted, and country-years in which IFRS is required for some

3 See for example, the discussions in Ball, 2006; and Leuz, 2010. Further, the World Bank cites inadequate enforcement as a key challenge for nearly every emerging market that is harmonizing with IFRS (as catalogued in its country-level Reports on Observance of Standards and Codes).

3

listed firms; and “3” for country-years with full IFRS adoption for listed firms. The variable is

constructed for each country-year in a sample of 92 countries for the period 2003–2008.4

To measure the perceived network benefits to a country from adopting IFRS in a given

year, we construct a variable based on the country’s trade with IFRS-incorporating jurisdictions.

Because “trade” is expected to vary with anticipated IFRS adoption, we compute our measure of

perceived network benefits using the predicted value of trade from an autoregressive model with

controls. Our primary empirical tests involve regressing, in a country-year panel, the IFRS

adoption status on perceived network benefits. Because network effects are measured at the

country-year level, we can include country and year fixed effects in the regression. The country

fixed effects allow us to control for cross-sectional variation in country-specific institutional

features that influence the choice of accounting standards (e.g., enforcement institutions). The

use of country fixed effects assumes that these institutions have not systematically changed

across sample countries over our regression period, 2003–2008, as several scholars and the

World Bank have argued (although, we caution that some institutional changes have likely

occurred). The year fixed effects control for time-based explanations that might lead to a

spurious correlation between IFRS harmonization and our network-effects variable.

The inclusion of country and year fixed effects means that our regressions test whether

perceived network benefits can explain some of the intertemporal variation in IFRS

harmonization across countries, holding constant countries’ institutions and overall worldwide

trends. We find that a country’s IFRS adoption status, as measured by our ordinal dependent

variable, is an increasing function of the perceived value of its IFRS network.

We expect a country’s economic size and its relative economic dependence on other

countries to respectively temper and intensify the perceived effects of a growing IFRS network 4 The EU is represented as a single observation among the 92 countries since EU members adopted IFRS jointly.

4

on the choice of accounting standards. Larger countries, due to the size of their markets, are

likely to attract foreign capital and maintain international trade even if they continue using

domestic standards. Thus, we expect larger (higher GDP) countries to be less swayed towards

IFRS adoption by the value of their IFRS networks. On relative economic dependence, we

expect network effects to be of greater importance to countries where trade accounts for a higher

proportion of GDP. Greater dependence on foreign trade accentuates benefits from lowered

transactions costs perceived from IFRS adoption.

The evidence from multivariate tests accounting for country and year fixed effects is

consistent with network effects mattering less to countries with larger GDPs; the evidence on

countries where foreign trade accounts for a larger fraction of GDP is mixed. For countries in the

largest size quartile, an inter-quartile increase in perceived network benefits is associated with

1.5% of the shift from no IFRS-related activities (level “1”) to some harmonization efforts (level

“2”); by contrast, for countries in the smallest size quartile, an inter-quartile increase in perceived

network benefits is associated with 44% of the shift from level “1” to level “2.” These findings

suggest that a country’s relative economic power shapes its response to the increasing worldwide

adoption of IFRS: economically powerful countries are more likely to restrain from adopting

IFRS or tailor their harmonization with IFRS to the capabilities of domestic institutions.5

We address several potential alternative explanations for our results on network effects.

Since our results are robust to country and year fixed effects, the explanations we consider are

those that, like network effects, are at least partially panel-based (as opposed to purely cross-

sectional or time-based). Moreover, we focus on explanations captured by variables associated

with our measure of network effects (and thus, potentially confounding the interpretation of our

5 Anecdotal evidence from China’s process of converging with IFRS is consistent with this finding: Ramanna, Donovan, and Dai (2009) describe various exceptions to IFRS China has carved out as a condition to converging with the global standards.

5

empirical tests). We identify five such explanations. First, recognizing that since the early 2000s

the World Bank has encouraged its client countries to embrace IFRS (World Bank, 1998–2008),

we include in our tests a variable capturing panel variation in the World Bank’s pressure. Our

second and third alternative-explanation tests are based on the idea that IFRS harmonization can

be part of a country’s broader policy response to globalization. We examine whether IFRS

adoption is related to the extent of internationalization of a country’s economy, and whether our

results can be attributed to countries’ differential exposure over time to global information

resources. The fourth test examines whether countries’ IFRS adoption decisions can be explained

by changes in national governments, since IFRS adoption can be part of a new government’s

broader political-economic strategy. Finally, recognizing the role of the Big 4 audit firms in

promoting and lowering the implementation costs of IFRS adoption, we test whether a country’s

decision to adopt IFRS is related to the entry of the Big 4 auditors into that jurisdiction. We find

some evidence consistent with each of the five alternative explanations above, but the evidence

is not mutually exclusive to network effects: network effects remain statistically and

economically significant determinants of IFRS adoption after the inclusion of these variables.

We also perform a number of sensitivity tests. First, because our adoption status variable

inherently assumes an order among, and equal distances between, the various levels of IFRS

adoption, we perform robustness tests using non-ordinal regressands (e.g., hazard and

multinomial logit models). Second, we test the sensitivity of our results to restricting the

definition of a country’s network benefits to trade with only non-EU countries adopting IFRS.

The EU had a dominant role in the establishment (EC, 2000) and subsequent direction (e.g.,

Leone, 2008) of the IASB, so economic relations with the EU can be overriding in explaining the

impact of network benefits on IFRS adoption. Finally, we expand the ordinal characterization of

6

our IFRS adoption variable from three levels to five to explore differences between countries’

IFRS “convergence” projects and other partial adoption strategies. We discuss the results of

these sensitivity tests in Section IV.

Overall, the evidence in this paper suggests that perceived network benefits are an

important determinant of IFRS adoption over time. Since we conceptualize and measure network

benefits ex ante, i.e., as a determinant (not consequence) of IFRS adoption, our evidence does

not speak to whether these network benefits are realized.6 Nevertheless, our findings related to

network effects are important for at least four reasons. First, they add context to findings from

existing firm-level studies on IFRS adoption. Firm-level studies are conditional on countries’

decisions to mandate IFRS, suggesting firm-level studies examine the second stage in what is at

least a two-stage process. Second, the evidence on network effects suggests a country can adopt

IFRS even if its domestically developed accounting standards are particularly well-suited to its

domestic institutions. Third, we find that smaller and, in some cases, more economically

dependent countries are more likely to adopt IFRS because others are doing so. Such countries

also tend to have weaker market institutions (e.g., World Bank, 1998–2008), likely impeding

effective implementation and enforcement of IFRS. Finally, a literature in economics has shown

that when network effects contribute to the dominance of a standard, even superior innovations

in the future may not be implemented (e.g., the QWERTY keyboard; David, 1985; Katz and

Shapiro, 1985). While our evidence cannot establish this will be the case with IFRS, the evidence

is germane in the context of prior research on network effects.

The remainder of this paper is organized as follows. Section II develops our hypotheses.

Section III describes the cross-country dataset on IFRS adoption and our measures of IFRS

6 Daske et al. (2008) show that real capital-market benefits from IFRS adoption occur only in countries with well-developed information and monitoring institutions. Their evidence suggests that the “perceived” network benefit we show as being a determinant of IFRS adoption is likely to actualize in only some countries.

7

network values. That section also describes the research design, including our tests of alternative

explanations. Section IV describes the results and Section V concludes.

II. Hypotheses development

Prior literature argues that a country’s financial reporting standards co-develop with the

economics, politics, and culture of its institutions like auditing, securities laws, courts, and

financial intermediaries. One implication from this literature is that institutional differences

across societies can explain countries’ choices on IFRS (i.e., adopt or not). In this vein, Ramanna

and Sletten (2009) describe a number of institutional hypotheses for why countries do not adopt

IFRS, including quality of local governance standards, international influence, and cultural

distance from the EU. Notwithstanding institutional differences, IFRS has grown in popularity

over six years, as seen by the increasing number of countries attempting to harmonize with IFRS

between 2003 and 2008. Assuming the institutions that shape the accounting standards in a

country have not changed substantially over the 2003–2008 period, it remains to be explained

why so many countries adopted IFRS so quickly. We hypothesize that network benefits from

extant worldwide adoption of IFRS are significant determinants in the time-series decision on

IFRS harmonization by any given country.7

To understand the role of network benefits in IFRS adoption, it is helpful to consider in

parallel the example of a popular network-dependent product, Facebook.8 The value to a user

from adopting Facebook as a communication portal is driven by three distinct sources: (1) the

demand for a communication portal; (2) the features on Facebook (e.g., applications and

customizability); and (3) the number of other Facebook users. The first two sources are direct

7 Währisch (2001) describes how network effects can motivate firms to adopt international accounting systems. 8 A well-developed theoretical literature in economics explores the notion of network effects. See for example, Katz and Shapiro (1985), Liebowitz and Margolis (1994; 1998).

8

benefits from adopting Facebook, while the third source is the value from Facebook’s network.

The role of direct benefits in adopting Facebook can be explained by idiosyncratic preference

functions that capture an individual’s demand for a communication portal and its user features;

the analogy of direct benefits in the case of countries’ IFRS adoption is the institutional

complementarities discussed earlier. The role of network benefits in adopting Facebook can be

explained by perceived lower transactions costs, given the extant group of Facebook adopters. A

similar argument can be made with respect to IFRS and we elaborate on this point below.

IFRS, as a globally recognized body of standards, is expected to lower transactions costs

for foreign users of financial statements. That is, foreign financial statement users already

familiar with IFRS through their own adoption of the standards are expected to incur lower

barriers in analyzing overseas financials prepared under IFRS, which in turn can result in

benefits accruing to entities reporting under these standards. As more jurisdictions with

economic ties to a given country adopt IFRS, the perceived benefits to that country from

lowering transactions costs to foreign users, and thus from adopting IFRS, can increase.

Consequently, we hypothesize that a country’s adoption of IFRS (holding constant institutional

determinants) can be explained, in part, by the magnitude of its economic relations with other

countries that have adopted IFRS through that date. We refer to this magnitude as the value of

the IFRS network to a country at a given time.

The nature of a country’s economy, specifically as it relates to that of other countries, is

likely to temper or intensify the perceived effects of a growing value of its IFRS network on

accounting standards. We exploit this argument and examine how a country’s economic size and

its relative economic dependence on other countries affect its sensitivity to perceived network

benefits. On economic size, we expect larger countries (i.e., countries with higher GDP) to be

9

less swayed towards IFRS adoption by the value of their IFRS networks. Larger countries are

likely to have greater bargaining power due to the size of their markets, and as a result, are likely

to attract foreign capital and maintain international trade even if they continue using domestic

standards. On relative economic dependence, we predict that network effects are of greater

importance to countries where trade accounts for a higher proportion of GDP. Greater

dependence on foreign trade increases benefits from reduced transactions costs expected through

IFRS adoption. The predictions on size and relative economic dependence generate cross-

sectional tests that can help validate the hypothesis on network effects and provide additional

context to the question why countries adopt IFRS.9

III. Metrics and research design

The use of countries as a unit of empirical analysis is common in research in international

finance and corporate governance (see for example, La Porta, Lopez-de-Silanes, and Shleifer,

2008, for a recent review). Nevertheless, its use remains relatively rare in the accounting

literature. Accordingly, we begin this section by laying out the basic assumptions implicit in our

empirical tests at the country level. Country-level decisions reflect a country’s domestic political

economy.10 In using country-level decisions as the unit of empirical analysis, researchers must

assume that the domestic political economies that generated such decisions are not correlated

with their independent variables.11 The equivalent assumption in our study would be that

9 Note the prediction on size is not that it is inversely related to network effects (countries are likely to perceive network benefits regardless of size); rather, the prediction is that perceived network benefits matter less to larger countries. A similar argument applies to relative economic dependence. 10 More specifically, such decisions are the result of actions by special-interest lobbyists and ideology-driven regulators (see Kothari, Ramanna, and Skinner, 2010, for an analysis of political-economic theories as they apply to accounting standard setting). The relative power of lobbyists and regulators in country-level decision making is likely to vary based on the extent of equitable representativeness within countries (e.g., opposing groups are more likely to get a voice in the United States than in Zimbabwe). 11 Every firm-level study makes a similar assumption vis-à-vis the decision making process within a firm.

10

variation in countries’ perceived networks benefits is not systematically related to variation in

their domestic political economies. However, in our study, if we assume the nature of a country’s

domestic political economy remains static through our six-year sample period, our use of country

fixed effects controls for cross-sectional variation in domestic political economies. Further, our

use of government fixed effects in robustness tests can also address changing political economies

as new governments introduce new policies into their jurisdictions.

To test our hypotheses on network effects, we use by-country time-series measures of

both IFRS harmonization and network benefits. The first two sub-sections of Section III describe

these measures. In the remaining sub-sections, we discuss our research design.

Measuring IFRS adoption

Below, we describe the construction of our database of IFRS adoption by country-year.

While our research question concerns the role of network effects in countries’ adoption

decisions, we are unable to obtain reliable information on the decision dates of IFRS adoption for

a broad sample of countries. Consequently, we use actual adoption dates as a proxy for adoption

decision dates. That is, we compile information on the degree of a country’s IFRS response in

every year between 2003 and 2008. We begin our tests in 2003 because we are interested in

IFRS as developed and sponsored by the IASB, and 2002 was the first full year of the IASB’s

existence.12 We end our sample in 2008, because required macroeconomic data to construct the

independent variables in our study were not available beyond 2008 at the time we initiated our

12 We exclude standards promulgated by the IASB’s predecessor the International Accounting Standards Committee (IASC) because there is evidence to suggest the IASC was culturally quite different from the IASB. In particular, while the IASB’s standards are influenced by Pan-European economic interests (given the IASB’s implicit charter from the EU), the IASC’s work was perceived as relatively more Anglo-centric. The IASC was established in 1973, the year the UK joined the European Community. Benston, Bromwich, Litan, and Wagenhofer (2006, p. 229) argue that by this time, existing European Community countries had made significant progress towards accounting harmonization, and the IASC was created to help the UK have a voice in future cross-country standardization.

11

study. Some countries have implemented IFRS since 2008, but our database, being censored at

year-end 2008, will correctly classify these countries as non-adopters for years 2003–2008.

We use three primary sources of data on IFRS adoption: (1) Deloitte & Touche’s

IASplus.com website; (2) a similar internet database from PriceWaterhouseCoopers; and (3) data

from the World Bank’s country Reports on Observance of Standards and Codes (ROSC reports).

Deloitte’s website lists IFRS adoption information for 153 legal jurisdictions (including 30

jurisdictions in the EU/ European Economic Area, EEA), although the information for several

countries is not up-to-date. PriceWaterhouseCoopers’ website provides similar adoption

information as of the end of 2008 for 109 countries (including 26 jurisdictions in the EU/ EEA).

The ROSC reports are comprehensive reviews of individual countries’ accounting and auditing

regulations, and also include discussions of other related issues such as accounting education and

enforcement. The ROSC reports are available for 74 developing countries (the reports are usually

prepared for World Bank client countries), but some of the reports date as far back as 2001 and

cannot be fully relied on for more current information.

In coding a country’s IFRS adoption we use information from all three data sources,

when available. Each data source covers a slightly different set of countries, and their

assessments of the extent of IFRS adoption occasionally differ for those countries they cover in

common.13 For countries with discrepancies among data sources, we opt for the adoption status

suggested by the majority (two) of these sources or, when only two sources are available, we

determine the status using the source that provides more supporting information for its

assessment. When the available information is not sufficient to determine the extent of IFRS

13 While the IASB itself relies on IASplus.com as a data source (IASB, 2008a, b), there are a few cases in which other sources disagree with the website’s assessment of the country’s adoption status. For example, Egypt and Peru are listed on IASplus.com as requiring IFRS for all listed companies, but the PriceWaterhouseCoopers website, the ROSC reports, and at least one (non Deloitte) Big-4 audit partner in each country disagreed with IASplus.com. In these circumstances, we err in favor of the other data sources.

12

adoption or the relevant dates, we supplement it with information from the official websites of

national standard setters, securities exchanges, associations of accountants, and web searches of

newswire archives. Our searches are limited to websites available in English. Finally, we obtain

some additional information from correspondence with the investor relations departments of

national securities exchanges and country managing partners at Big-4 audit firms.

A qualitative analysis of the extent of each country’s IFRS adoption leads us to conclude

that this process cannot be described with a simple binary variable. Consequently, we classify

countries into five different groups based on their degree of IFRS harmonization as described

below. First, following the adoption classification on the IASplus.com website, we create four

groups: full adopters (when IFRS as issued by the IASB is required for all listed companies);

IFRS required for some listed companies; IFRS permitted for (at least some) listed companies;

and non-adopters (when IFRS is not permitted for any listed companies). We then refine this

classification to include an additional category for countries with an IFRS “convergence” project

as discussed in Appendix A.

For the purposes of our primary empirical analysis, we reclassify the five categories of

IFRS harmonization described above into three ordinal levels: (1) non-adopters; (2) countries

with convergence projects, countries allowing voluntary IFRS adoption, and countries requiring

IFRS for some listed companies; and (3) full IFRS adopters. The ordinal levels represent

distance from full adoption of IFRS as issued by the IASB. We agglomerate the middle three

categories into a single level (“partial adopters”) because it is difficult to objectively rank these

categories relative to each other. Our primary dependent variable, the three-level ordinal

estimated for every country-year in the dataset, is denoted, , .

13

To test the robustness of our ordinal rankings, in additional analyses we estimate both

multinomial logit regressions, where we compare each harmonization category to IFRS non-

adopters, and hazard regressions, where we treat only countries fully embracing IASB standards

as IFRS adopters. Further, in subsequent OLS-based robustness tests, we explore the sensitivity

of our results to a five-level ordinal classification as follows: (1) non-adopters, (2) convergence

project, (3) voluntary adoption, (4) required for some listed firms, and (5) full adopters.

Since we are interested in the determination of financial reporting requirements for listed

companies, we exclude from our database all jurisdictions that do not have stock exchanges. We

also exclude those jurisdictions for which the World Bank does not report gross domestic

product (GDP) data. The World Bank’s World Development Indicators (WDI) database is our

source for GDP data. Additionally, we exclude jurisdictions without dyadic trade data as

reported by the International Monetary Fund (IMF). These data are required to calculate our

measure of network benefits (in the following subsection). In our final sample, the 27

jurisdictions that compose the member states of the EU are treated as one observation.14 The

reason for this treatment is that in 2002 the EU made a joint decision to adopt IFRS by 2005 (EC,

2008) i.e., adoption of IFRS by EU member states was not made country-by-country. The data

selection procedure described above yields a final sample of 92 countries and 552 country-years.

The countries are listed in Appendix B.

Measuring perceived network benefits

As noted earlier, the network benefits from IFRS harmonization for a given country i in

year t are the economic benefits perceived from commercial relations with the existing base of

14 EU countries are coded as being full adopters between 2005 and 2008, and as having convergence projects with IFRS in 2003 and 2004. Our coding is based on the extent of IFRS harmonization for the majority of EU members.

14

countries already harmonized with IFRS. We determine “the existing base of countries already

harmonized with IFRS” using the country-year data on IFRS adoption described in the previous

sub-section. Measuring “the economic benefits perceived from commercial relations” with this

base of countries is more challenging. One relatively simple proxy for the perceived benefits is

the aggregated value of existing trade with the subset of IFRS adopting countries. This

aggregated value can be constructed from dyadic (bilateral) trade data obtained in panel form

from the IMF’s Direction of Trade Statistics (DOTS) database. Under this approach,

, ,

∈

… 1

Where , is the dyadic trade volume between country i and country j in year t-1,

i ∉ J, and J is the subset of all jurisdictions that have mandated IFRS for all listed firms (i.e., full

adoption), as of year t-1.15

The problem with this approach is that existing trade data incorporate the effects of

country i’s imminent IFRS harmonization as well. That is, since we measure harmonization

effective of the implementation date, rather than the announcement date, existing trade data are

likely to reflect at least some of the realized consequences of country i’s impending

implementation.16 For our purposes, we ideally want to separate these realized effects from any

measure of economic benefits perceived prior to the harmonization decision. The realized effects

can result in either a boost or a decline in existing trade from its latent level (i.e., its level absent

15 Conceptually, one can also measure network benefits using foreign direct investment or foreign equity portfolio investment. We do not use these measures because dyadic-level data on these measures are available for only a handful of very large countries. Moreover, the use of trade in measuring perceived IFRS network benefits is consistent with accounting’s role in contracting. For example, Ramanna et al. (2009) identify “anti-dumping” lawsuits over exports by Chinese manufacturers as a major reason for China’s decision to converge with IFRS. These lawsuits usually allege that Chinese manufacturers are selling (or “dumping”) goods at prices below cost in overseas markets. By adopting IFRS, Chinese manufacturers hoped to provide more reliable data on their costs, thus justifying their low prices on exports and accordingly boosting exports. 16 E.g., Márquez-Ramos (2009) finds evidence of bilateral exports in the E.U. increasing after IFRS adoption.

15

impending implementation), depending on the country’s institutional features that facilitate

contracting efficiency, including auditing, securities laws, courts, and financial intermediaries.

For example, if IFRS harmonization decreases contracting efficiency, imports will decrease

because customers in country i will suffer worse credit terms (anticipating deteriorating

reporting) and exports will decrease because of less reliable measures of financial strength

among country i’s suppliers. This situation will result in a decrease in overall trade.

Alternatively, if IFRS harmonization increases contracting efficiency, countries can similarly

experience an increase in overall trade. 17,18

In order to obtain a cleaner measure of perceived benefits, we construct a proxy that is

based on the lagged value of trade and, thus is less likely to include the real effects of a country’s

impending IFRS harmonization. Specifically, we regress existing dyadic trade on its lagged

value in a panel with country and year fixed effects and use the predicted value of trade from the

regression as our measure of perceived network benefits. That is, we estimate the following

regression in the panel of all country-dyad-years in our dataset.

, ∗ , ∗ . . ∗ . .

∗ 1 . . , … 2

Where , is the dyadic trade volume between country i and country j in year t-1

(scaled by country i’s total trade in year t-1), and . The predicted value of existing dyadic

trade from this regression, , , accounts for secular increases in trade from year t-2,

controlling for country- and year-specific effects. Assuming anticipated IFRS harmonization in 17 Evidence in DeFond et al. (2011) supports this intuition, albeit in the context of investment flows. In studying U.S. mutual fund ownership among 14 IFRS adopting countries, the authors find either increases or decreases in fund ownership depending on the “credibility” of IFRS implementation across those countries. 18 Confounding the realized and perceived benefits from IFRS harmonization is especially troublesome if the realized effects are negative (i.e., decrease trade) because a country is unlikely to adopt IFRS unless it expects a net benefit from the adoption. In other words, measuring network benefits using existing trade data is particularly problematic for countries with weak monitoring and information processing institutions, where harmonization is likely to decrease contracting efficiency.

16

year t does not confound trade data from year t-2, the measure , is a better proxy for

IFRS network benefits than the raw trade value , for the reasons discussed above. As

in equation (1), we estimate the actual network benefits for a given country i in year t as the

aggregated value of , across the subset of countries already fully adopting IFRS.

, , … 3∈

Where i ∉ J and J is the subset of all jurisdictions that have mandated IFRS for all listed

firms (i.e., full adoption), as of year t-1.

In calculating , , trade with countries pursuing convergence, countries

permitting voluntary adoption, and countries requiring only partial adoption are not included. As

noted earlier, “convergence” is a loosely defined term that can result in different sets of standards

in different countries. Further, voluntary and partial adoption efforts are not equivalent to full

adoption because they require different levels of adaptation to IFRS among complementary

institutions such as auditing. Nevertheless, in robustness tests we explore the sensitivity of our

measure of network benefits to including trade with countries harmonizing less than fully with

IFRS. Specifically, we define an alternative network benefits measure _ , .

_ , , … 4∈

Where i ∉ J and J is the subset of all jurisdictions that have: (i) converged with IFRS; (ii)

made IFRS available for voluntary use; (iii) mandated IFRS for at least some listed firms; or (iv)

mandated IFRS for all listed firms (i.e., full adoption), as of year t-1.

Additionally, we define another measure of network benefits that restricts the set J

(where i ∉ J) to only non-EU countries mandating full IFRS adoption. Specifically,

17

_ , , … 5∈

_ , excludes trade with EU countries and allows us to test whether

economic relations with non-EU countries can explain the impact of network benefits on IFRS

adoption. The EU had a dominant role in the establishment of the IASB (EC, 2000) and EU

countries continue to exercise discretion over the Board.19 Moreover, EU countries adopted IFRS

jointly, potentially providing the critical mass of commitment to IFRS needed to generate

network benefits. Given the EU’s central status vis-à-vis IFRS, it is possible that benefits

perceived from relations with the EU are overriding in explaining network effects.20

Research design based on OLS

Our primary tests are ordinary least squares (OLS) regressions of the three-level ordinal

ranking of IFRS harmonization on network benefits. The regressions are estimated in the panel

of all countries i and years t. Standard errors are clustered by country. We predict to be

positive and statistically significant.

, ∗ , ∗ . . , … 6

We expect a country’s economic size and its relative dependence on trade to qualify the

relation between IFRS harmonization and network benefits (i.e., in equation (6)). We measure

economic size using the quartile rank of a country’s GDP (in constant year 2000 U.S. dollars);

the data are obtained from the WDI. Trade dependence is measured as the quartile rank of the

ratio of total foreign trade to GDP; total foreign trade data are obtained from the IMF’s DOTS

19 For example, the IASB, in the wake of declining financial markets in 2008, allowed financial institutions using IFRS to opt for a one-time reclassification of certain investment assets, thus avoiding write-downs. Some in the financial press have suggested the IASB made this decision in response to pressure from the EU (e.g., Leone, 2008). 20 Because of their joint decision to adopt IFRS, EU countries are represented as a single observation among 92 countries in the sample. However, a country may trade with some EU members and not others. In order to reflect this, all measures of network benefits discussed above include trade with individual EU countries.

18

database. The economic size and trade dependence variables are denoted q(GDP) and

q(Trade/GDP), respectively. The formal statement of the OLS regression with size and trade

cross-sectional effects estimated over the dataset of all countries i and years t is as follows.

, ∗ , ∗ , ∗ ∗ ,

∗ / ∗ . . , … 7

We predict and to be positive and statistically significant, and to be negative and

statistically significant. Regression (7) does not include main effects for q(GDP) and

q(Trade/GDP) since we include country fixed effects.

As noted earlier, we define alternative measures of network benefits: specifically,

_ , and _ , . Accordingly, we also estimate regression (7) replacing

, with these measures.

The use of country fixed effects in equations (6) and (7) allows us to hold constant the

institutions that influence a country’s choice of accounting standards. In the initial set of tests,

we exclude year fixed effects because the regressions are intended to test whether our measure of

network benefits explains part of the intertemporal variation in IFRS adoption. More specifically,

since network benefits are in part determined by which countries have adopted IFRS in the past,

and are thus correlated with the average level of adoption over time, the inclusion of time-based

controls can extract that part of the variation in IFRS adoption that is attributable to network

effects. That said, it is important to establish that our results on network effects are robust to the

inclusion of time-based controls, particularly since there are numerous possible time-based

explanations for the increasing IFRS adoption by countries over our sample period. Accordingly,

we also estimate equations (6) and (7) with year fixed effects. In doing so, we note that, even

after including such time-based controls, the network effect can still be identified by our network

19

variable because, both conceptually and empirically, network effects are cross-sectional and

time-series in nature (as discussed earlier).

Alternative explanations

While a number of factors likely contribute to IFRS adoption by countries, our focus in

this study is network effects. Thus, in this subsection, we focus on explanations that potentially

confound our inferences on network effects (explanations likely captured by variables associated

with our measure of network effects). Moreover, in the context of testing the network effects

hypothesis with country and year fixed effects, the alternative explanations we examine are those

that, like network effects, are at least partially panel-based. We identify five such explanations.

The first focuses on the role of the World Bank in encouraging its client countries to

embrace IFRS. The World Bank, through its periodic ROSC reports, evaluates the status of a

country’s corporate governance institutions (including accounting institutions), and then makes

recommendations on how that country can progress towards more internationally consistent

governance practices (including through IFRS harmonization). The incidence of an ROSC report

may thus result in both IFRS adoption and changes in trade, which will be captured by our

network effects variable. If so, an association between IFRS adoption and the network effects

variable can be due to World Bank pressure rather than network effects. To control for this

possibility, we include in robustness tests an indicator variable _ , to denote whether a

World Bank ROSC report was issued for a given country “i” prior to the year “t” in question

(i.e., in year t-1 or before).

The second alternative explanation we test is based on the idea that IFRS adoption can be

part of a country’s broader policy response to globalization. That is, as a country’s economy

20

becomes increasingly global over our sample period, the country may initiate associated policy

responses such as IFRS harmonization. In this case, the association between IFRS adoption and

our network effects measure can be attributed to the growing internationalization of an economy.

To address this possibility, we include as a control variable the ex-ante share of international

trade in a country’s GDP. This variable, Trade/GDP, is already described in the previous sub-

section since quartile ranks of Trade/GDP are used in our cross-sectional tests of the effect of

trade dependence on network effects.21

The third alternative explanation is also related to globalization: individual countries’

policy responses to globalization (including IFRS harmonization) may vary in time due to

differential exposure to international information resources. Implicitly, the greater the exposure

to international information, the more likely constituents in a country become interested in

globalization. Moreover, differential exposure to international information may affect trade, thus

confounding the interpretation of our network effects measure. To mitigate this possibility, we

include a variable that proxies for countries’ access over time to international information

resources: the prevalence of Internet usage in a country-year (Info_Globalizationi,t). The data are

from the WDI database.

Fourth, we examine whether countries’ IFRS adoption decisions can be explained by

changes in national governments. That is, the adoption of IFRS and corresponding changes in

our network effects variable can simply reflect the policy of a new government to be more

global. To address this explanation, we rerun our regressions with government fixed effects in

lieu of country fixed effects. Each new government in our panel dataset, as denoted by a new

“head of government” in the CIA World Factbook, is identified by a unique fixed effect (the CIA

21 In the presence of country fixed effects, the coefficient on Trade/GDP captures the impact of the variable’s by-country mean-adjusted value on IFRS adoption. This can, in essence, be interpreted as the impact of the growing internationalization of an economy on IFRS adoption.

21

identifies “heads of governments” as presidents in executive republics, prime ministers in

Westminster-style governments, and sovereigns in absolute monarchies).

The fifth alternative explanation we test is based on the hypothesis that a country’s IFRS

adoption and concurrent changes in trade are related to the decision of the Big 4 auditors to enter

that jurisdiction. The entrance of the Big 4 into a country can be associated with IFRS adoption

either because the firms lobby for such outcome, or because the implementation costs of IFRS

for that country decrease with the presence of auditors experienced in IFRS. We address this

hypothesis by including in our regressions an indicator variable, BigAuditor, equal to one for

country-years in which all three of the following Big 4 audit firms have offices located in the

country: Ernst & Young, KPMG, and PWC. The data are collected from the firms’ websites and

through private correspondence with the firms. We have been unable to collect similar

information from Deloitte.22

Research design based on the multinomial logit and hazard models

Our method of coding harmonization levels for the OLS regressions has its limitations: it

requires that we determine the relative gaps between the levels a priori rather than a posteriori,

as would be possible using multinomial logit or other models. We use OLS in the main tests

because country fixed effects are not feasible in the other models. We test the robustness of our

ordinal harmonization levels by estimating both multinomial logit and hazard regressions. In the

multinomial logit regressions, we use IFRS non-adopters as the base case and estimate the

probabilities that countries choose each of two other IFRS harmonization categories (i.e., partial

22 Our results are robust to an alternative measure of BigAuditor: the proportion of the three audit firms (Ernst & Young, KPMG, and PWC) with offices located in a given country in a given year.

22

adoption or full adoption). We also interact network benefits with the economic size and trade

dependence proxies to test the cross-sectional hypotheses on network effects.

In the hazard regressions, full adoption of IFRS is modeled as the “failure” event, i.e., the

dependent variable is the time to full IFRS adoption. The hazard regressions estimate the impact

of network benefits on full adoption, accounting for the relative timing of that adoption. Let “T”

be the “survival” time, i.e., the number of years until full IFRS adoption. Then, T=1 if a country

adopts IFRS in 2003, T=2 if for adoption in 2004, and so on. Since the data are censored in 2008,

the highest possible value for T is six. For countries that do not adopt by 2008 or that harmonize

only partially with IFRS, T is set to six and the dependent variable is treated as censored. In

additional hazard tests, we relax the latter assumption and treat any harmonization effort at

convergence or above (e.g., voluntary adoption) as a “failure” event. We estimate two different

versions of the hazard model: the counting process hazard model (Therneau and Grambsch,

2000) and the more-familiar Cox proportional hazard model. On average, the two methods are

expected to, and do in our case, yield similar results; thus, we only report results based on the

counting process model.23 A positive coefficient on a covariate in the hazard model implies that

the likelihood of IFRS full adoption is increasing in the covariate. The primary independent

variables in the hazard analyses are variously: one measure of network benefits (e.g., , ,

_ , , etc.) and its interactions with the economic size and trade dependence proxies.

Both in the hazard and in the multinomial logit regressions, parameter estimates are

obtained from MLE, so these specifications do not include country fixed effects (due to the

limited number of observations, models with country fixed effects do not converge). We do,

23 The counting-process method is based on a discrete formulation of the dependent variable (i.e., survival time T), which is likely to be more appropriate for our data since we measure T in years (the Cox model assumes T is continuous). Moreover, in situations where the range of possible values for T is low (as in the case of our data where T varies between 1 and 6), Box-Steffensmeier and Jones (2004) argue the counting-process method is likely to be more appropriate for hazard analyses.

23

however, include year fixed effects and variables that control for alternative explanations.

Standard errors are clustered at the country level; in the case of the hazard model we use the

process described in Lin and Wei (1989).

IV. Results

Descriptive results

Table 1 presents the distribution of IFRS harmonization categories across the 552

observations in the country-year panel. The rows in the table correspond to the three different

adoption statuses described in Section III: (1) non-adopters, (2) countries with convergence

projects, countries allowing voluntary IFRS use, countries requiring IFRS for some listed

companies, and (3) full adopters. Countries across the three sub-categories in (2) are presented

separately. The columns represent the six years in the panel, 2003–2008.

INSERT TABLE 1 HERE.

The proportion of non-adopters, 61 countries, is highest in 2003; the proportion decreases

to 30 countries by 2008. In contrast, the number of full adopters, i.e., countries requiring IFRS

(as issued by the IASB) for all listed firms in their jurisdictions, grows from 18 countries in 2003

to 32 countries in 2008. The EU, having adopted IFRS fully in 2005, is represented as a single

observation in this category effective 2005. The number of countries with IFRS convergence

projects is also increasing through the time-series, from six in 2003 to 15 in 2008. The proportion

of countries permitting voluntary IFRS use and the proportion of countries requiring IFRS for

some listed companies are non-monotonic in the time-series. For example, there are five

countries permitting voluntary IFRS use in 2003, 10 in 2006 and 2007, and eight in 2008. These

proportions are likely to vary over time because countries may choose to “ease in” to IFRS

24

harmonization by first allowing voluntary IFRS use and/or requiring IFRS for some listed

companies, before moving to full-scale IFRS adoption.


Table 2 provides descriptive statistics for our measures of perceived network value by

adoption year and by the three levels of IFRS harmonization. Panel A reports mean and median

values of our primary measure of network benefits , , i.e., economic benefits perceived

from countries that fully adopted IFRS. , has increased substantially between 2003 and

2008 across all categories of IFRS harmonization, although the increase is not always consistent

or monotonic. The average values of , differ across the various categories of IFRS

harmonization, although no pattern is apparent in these differences. To interpret the average

values in Table 2, consider for example, the mean , for non-adopters in 2003: 0.019.

This value implies that 1.9% of total trade among non-adopters in 2003 came from countries that

fully adopted IFRS as of 2002. Table 2, Panel B, reports mean and median values of the

alternative measures of perceived network benefits, _ , and _ , .

The average values of _ , are generally higher than those of , , while the

average values of _ , are the same or lower than those of , .


In Table 3, we report the mean and median values of GDP and Trade/GDP. These

variables are used to generate the in-sample quartile rank variables q(GDP) and q(Trade/GDP)

that are the bases for the cross-sectional tests of network effects using country size and economic

dependence. Trade/GDP is also used to test the alternative hypothesis that IFRS adoption can be

attributed to the internationalization of a country’s economy. In addition to GDP and

Trade/GDP, we report the average values of countries’ foreign-equity portfolio investments to

GDP, i.e., FEPI/GDP. This variable can also be used as a proxy for a country’s economic

25

dependence in cross-sectional tests; however, poor availability in the panel limits its use. The

means and medians of GDP, Trade/GDP, and FEPI/GDP in Table 3 are reported by the three

levels of IFRS harmonization and years. The values corresponding to a given year are lagged

values, i.e., the mean GDP value for non-adopters in 2003 is actually the mean from year 2002.

This is because lagged-values are used in the regression analyses, i.e., to explain IFRS

harmonization in 2003, we use size and economic dependence from 2002 (and so on). In Table 3,

the average values of GDP appear to vary across time within the IFRS harmonization categories.

This phenomenon is driven more by countries’ movements across harmonization categories (due

to changes in their harmonization responses) than by changes in the values of GDP for a given

country over time. Unlike with GDP, there appears to be no consistent through-time relation

between Trade/GDP and IFRS harmonization status.

Table 3 also shows the mean and median values of WB_Report, Info_Globalization, and

BigAuditor—three independent variables used in subsequent tests of the alternative explanations

to network effects. By construction, the average values of WB_Report increase over time; but

there is no discernible difference between its average values across the three IFRS harmonization

categories. Info_Globalization is also increasing over time, but only among non-adopters and full

adopters; the decreasing trend in Info_Globalization across partial adopters is likely due to

countries in this category transitioning to full adoption. The median value of BigAuditor is one

across all categories in Table 3, which is expected given the international presence of the Big 4.

OLS regression results

Table 4 presents the results of OLS regressions of the ordinal IFRS harmonization status

on network benefits. The OLS regressions in Panel A include country fixed effects, and

26

statistical inferences are based on country-clustered standard errors (inferences are robust to

clustering by country and year). There are four columns to Table 4. The first column in Panel A

is the result of the regression specified in equation (6) in Section III, i.e., without interactive

effects for country size and economic dependence. The next three columns present the results of

regressions specified by equation (7), i.e., with interactive effects. Panel B is organized similarly,

except all regressions include year fixed effects.


In column (1) of Table 4 Panel A, the coefficient on the network benefits proxy,

, , is positive and statistically significant. The interpretation is that the degree of IFRS

harmonization is increasing in perceived network benefits, after controlling for institutional

determinants through fixed effects.24 In columns (2) through (4) of Table 4, we test for the cross-

sectional effects of country size and country economic dependence on the relation between IFRS

harmonization and network benefits. Across the three columns, the proxies for network benefits

are , , _ , , and _ , , respectively. The coefficient on the

non-interacted network benefits proxy in Table 4 is positive and statistically significant in

columns (2) and (3). A similar result is obtained on the coefficient of network benefits interacted

with q(Trade/GDP) only in column (3). The interpretation is that a country’s dependence on

foreign trade accentuates the impact of network benefits on IFRS harmonization only when such

benefits are defined to include trade with countries converging with IFRS. In contrast, the

interaction of network benefits with q(GDP) yields a negative and statistically significant

coefficient across both columns (2) and (3), consistent with country size attenuating the impact

24 The same inferences as the ones from column (1) of Table 4, using , as the network benefits proxy, can also be drawn using _ , , and _ , (results not reported).

27

of network benefits on IFRS harmonization when such benefits are defined to include economic

relations with EU member states.

To evaluate the economic significance of network benefits for countries’ IFRS

harmonization choices, we calculate the change in IFRS adoption status due to shifting the value

of , from the 25th to 75th percentile. We use our primary model with interaction terms

(reported in column (2)) for this analysis (not tabulated). For a country with average size and

average trade dependency, the shift in , from the 25th to 75th percentile leads to a 0.34

increase in IFRS adoption status (i.e., 34% of the shift from no IFRS-related activities to some

harmonization efforts or from some harmonization efforts to full IFRS adoption). However, our

cross-sectional results show that this effect is stronger (weaker) for countries that are smaller

(larger): for a country in the lowest (highest) size quartile with average trade dependency, a shift

of two quartiles in the value of , leads to a 61% (6%) increase in IFRS adoption status.

Perceived network benefits, when defined to exclude economic relations with the EU,

i.e., _ , in column (4) of Table 4, are not significantly associated with countries’

ordinal IFRS harmonization responses. Moreover, there is no differential association between

IFRS harmonization and network benefits for larger countries. However, among countries more

dependent on foreign trade, non-EU network effects significantly predict IFRS harmonization

responses. Overall, results from the OLS regressions suggest network benefits expected to accrue

from economic relations with the EU are dominant in explaining IFRS harmonization, especially

among more economically independent countries.25

25 The anecdotal evidence on Japan’s motivation to harmonize its GAAP with IFRS is consistent with this finding: for example, Skinner (2008, p. 220) notes that harmonization attempts in Japan arose from pressure to “convince” the EU “that Japanese GAAP was ‘equivalent’ to IFRS,” since “Japanese companies rely heavily on European capital markets for external debt financing.”

28

Panel B of Table 4 is equivalent in all respects to Panel A, except that it also includes

year fixed effects. As discussed earlier, the inclusion of year fixed effects is likely to weaken the

power of our network effects measure, but is necessary to show that our network effects results

are not simply due to time-based explanations.26 As seen in Panel B, the inclusion of time-based

controls dampens, but does not eliminate, the explanatory power of network effects. In the partial

model (column 1), the inclusion of time-based controls renders network effects insignificant; but

in the full models, with the interactions on size and trade dependency (columns (2) and (3)), the

coefficients on network effects remain statistically significant after including time-based

controls. Moreover, in the full models, the interaction of network effects with size is also

significant; the interaction of network effects with trade dependency is significant in column (3),

but not in column (2) (as in Panel A). Although not directly reported in Panel B, the year fixed

effects are also themselves statistically significant in the regressions. To evaluate the economic

significance of network benefits in Panel B, consider the model in column (2). For a country with

average size and average trade dependency, the shift in , from the 25th to 75th

percentile leads to a 23% increase in IFRS adoption status. For a country in the lowest (highest)

size quartile with average trade dependency, a shift of two quartiles in the value of ,

leads to a 44% (1.5%) increase in IFRS adoption status. Overall, the results suggest that network

effects can explain, in part, decisions to adopt IFRS, and that the effect can be identified even

after controlling for other time-based explanations.

Robustness to alternative explanations

Table 5 presents the results of tests of the five alternative explanations discussed in

Section III. Table 5 has two panels: the panels are identical in all respects other than that Panel A 26 The results reported hereafter are robust to using a time-trend variable in lieu of year fixed effects.

29

excludes year fixed effects and Panel B includes them. There are six columns to each panel of

Table 5. In each successive column, we test one of the five alternative explanations discussed

earlier; in the sixth column, we test all five alternative explanations simultaneously. The

regression specification being tested throughout Table 5 is the full model, i.e., including the

interactive effects of size and trade dependency on perceived network benefits. The proxy for

network benefits throughout Table 5 is , .


The coefficient on , throughout Table 5 is positive and statistically significant,

suggesting that the main effect of perceived network benefits is robust to the various alternative

explanations discussed in Section III. Also, the interaction of network benefits with q(GDP)

yields a negative and statistically significant coefficient throughout Table 5, consistent with

country size attenuating the impact of network benefits on IFRS harmonization. In contrast, the

coefficient of network benefits interacted with q(Trade/GDP) is positive and statistically

significant in only one column of Table 5, providing weak evidence, at best, for the hypothesis

on trade dependency. In untabulated tests, we re-estimate all models from Table 5 adding the

interaction of the network benefits measure with the square of q(Trade/GDP). The goal of these

tests is to examine whether there are nonlinearities in the ability of trade dependency to explain

cross-sectional variation in network effects: countries with the highest levels of trade dependency

may have idiosyncratic competitive advantages in world trade (e.g., being natural-resource rich)

that make them less sensitive to IFRS network effects. We find the first-order effect of

q(Trade/GDP) is significantly positive, while the second-order effect is significantly negative,

consistent with diminishing importance of IFRS network benefits in trade-heavy economies.

Turning back to the results reported in Table 5, in particular, the alternative hypotheses,

_ , and Info_Globalization are positive and statistically significant only when year

30

fixed effects are excluded, lending only modest support to the World Bank hypothesis and the

hypothesis on countries’ differential access to international information sources. In contrast,

BigAuditor is positive and statistically significant even with the inclusion of year fixed effects.

Further, in the model that includes all proxies for alternative hypotheses and year fixed effects

(column 6 of Panel B), BigAuditor remains statistically significant. In that model, Trade/GDP is

also positive and statistically significant. A propos the inclusion of government fixed effects in

lieu of country fixed effects, such fixed effects do not affect inferences on network effects.

Overall, the evidence in Table 5 is consistent with network effects being significant determinants

of the intertemporal shift from domestic accounting standards to IFRS after controlling for

alternative explanations. To evaluate the economic significance of network benefits after

controlling for alternative explanations, consider the full model in column (6) of Panel B. For a

country with average size and average trade dependency, the shift in , from the 25th to

75th percentile leads to a 21% increase in IFRS adoption status. For a country in the lowest

(highest) size quartile with average trade dependency, a shift of two quartiles in the value of

, leads to a 44% increase (2% decrease) in IFRS adoption status.

Hazard and multinomial logit regression results

Table 6 presents the results from the counting-process hazard model estimated with

standard errors clustered at the country level. In columns (1), (2), and (3) we measure network

benefits with , . In columns (4) and (5) we examine the robustness of the results in

column (2) to using different measures of network benefits, specifically: _ , , and

_ , . In the hazard regressions presented in columns (1), (2), (4), and (5) we

define the “failure” event as full adoption of IFRS, i.e., the dependent variable is the time to full

31

IFRS adoption. To examine the sensitivity of our results to an alternative definition of “failure,”

in column (3) we recode the dependent variable to represent the time to IFRS harmonization at

any level at or above a convergence project. Since we cannot estimate the hazard model with

country fixed effects, we supplement as regression controls the un-interacted variables q(GDP)

and q(Trade/GDP). This latter variable is used in lieu of Trade/GDP to test the alternative

hypothesis on the impact of internationalization of a country’s economy on IFRS adoption (the

results are invariant to using Trade/GDP instead).


Table 6 provides evidence that the likelihood of IFRS harmonization (at full adoption or

at or above a convergence project) increases in perceived network benefits. Moreover, consistent

with our predictions, the effect of network benefits on the likelihood of IFRS harmonization is

weaker for larger countries, as demonstrated by the statistically significant negative coefficient

on Nw*q(GDP) in columns (2) and (3). We find no statistically significant coefficients on the

network variables in column (4), where network effects are defined to include trade with all

countries that at a minimum initiated an IFRS convergence project. Thus, full adopters of IFRS

seem to perceive network benefits as arising mostly from sharing standards with countries that

adopt actual IFRS (as developed by the IASB), rather than a modified set of standards (as is the

case in convergence projects). Our inferences regarding the value of IFRS network excluding EU

countries (column (5)) are similar to those based on the OLS regression. In Table 6, we do not

find any evidence that a country’s trade dependence affects the importance of perceived network

benefits for a country’s IFRS adoption. Finally, as indicated by a statistically significant

coefficient on BigAuditor in columns (1) through (5), IFRS harmonization is positively

associated with the presence of Big 4 audit firms.

32

Table 7 reports results from the multinomial logit analysis with standard errors clustered

at the country level. The dependent variable in our multinomial logit model is IFRS

harmonization. Harmonization is coded as one of three categories: non-adoption, partial

adoption, and full adoption. The multinomial logit model does not assume any ordering among

the harmonization levels. Instead, it compares the partial adoption and full adoption categories to

the base-line decision, non-adoption. Consequently, the results from the multinomial logit allow

us to assess the importance of network effects for each category of adopters. As with the hazard

model, multinomial logit regressions exclude country fixed effects, but include the un-interacted

variables q(GDP) and q(Trade/GDP). Also as before, the latter variable is used in lieu of

Trade/GDP to test the alternative hypothesis on the impact of an economy’s internationalization

on IFRS adoption (results are invariant to using Trade/GDP instead). The results from the basic

model excluding interaction terms of Nw with q(GDP) and q(Trade/GDP) and using our primary

measure of network benefits ( , ) can be found in column (1) of Table 7. In this model

we find that the likelihood of full IFRS harmonization increases in perceived network benefits;

the coefficient on , is positive and statistically significant only for full IFRS adopters.


Columns (2) through (4) of Table 7 contain the results from regressions of the expanded

model that, in addition to the main effect of network benefits, allows us to test our cross-

sectional hypotheses. In the three columns, we report the results of estimating this expanded

model using alternative measures of network benefits based on different definitions of

“incorporating” IFRS: specifically, , , _ , ,, and _ , .

Our results from the cross-sectional models show that perceived network benefits

generally increase the probability a country fully adopts IFRS. Column 3, where we define

33

network benefits to include trade with countries that at least converge with IFRS, is an exception:

here, we find that the coefficient on _ , is not significant for full adoptions. This

result is consistent with inferences obtained from the hazard model and suggests that countries

adopting IFRS fully are more discriminating in their perception of network benefits. Such

countries are less likely to expect harmonization benefits from partial adoptions, including IFRS

converging countries, because the nature of such adoptions is likely idiosyncratic. The evidence

in Table 7 suggests network effects do not play a significant role in partial IFRS adoption.

On the cross-sectional variables, we find some evidence that network effects play a

weaker role in full adoption decisions of larger countries. In particular, the coefficient on

Nw*q(GDP) is statistically significant and negative for full adoptions in column (2). For full

adoptions, there is no evidence that the importance of perceived network benefits to a country

varies with its trade dependence as measured by Nw*q(Trade/GDP). Finally, we find that both

partial and full IFRS adoptions are positively related to the presence of Big 4 audit firms in all

four specifications of Table 7.

Overall, the evidence from our multinomial logit analysis indicates that perceived

network effects impact the likelihood of a country fully adopting IFRS. Our hazard model tests

further confirm these insights. Both models provide some support for our cross-sectional

prediction that the impact of network effects on IFRS harmonization is mitigated for larger

countries, but there is no evidence that network effects vary with a country’s trade dependence.

Robustness tests: OLS regressions with five levels of IFRS adoption status

In untabulated tests, we examine the robustness of our OLS regression results to using an

ordinal with five (rather than three) levels of IFRS adoption as the dependent variable. The five

34

levels are discussed in Section III. These robustness tests confirm our inferences from Tables 4

and 5: perceived network benefits are positively and significantly associated with IFRS adoption

status and more important for smaller countries. Inferences from the robustness tests are

unaffected by the inclusion of year fixed effects and controls for alternative explanations.

V. Conclusion

We develop and test the hypothesis that network effects are a significant factor in the

time-series growth in IFRS harmonization across countries. Network effects refer to perceived

lower transactions costs given the community of IFRS adopters worldwide. We find the degree

of IFRS harmonization in a country is an increasing function of the perceived value of its IFRS

network, particularly when network benefits are defined to include relations with EU member

states. The results suggest IFRS adoption is self-reinforcing. In cross-sectional tests, we explore

how the nature of a country’s economy, specifically its economic size and trade dependence, is

likely to affect the relation between network benefits and IFRS harmonization. We find evidence

consistent with network effects mattering less to countries with larger GDPs and, only in some

cases, to countries where trade accounts for a smaller fraction of GDP. These findings suggest

that the role of perceived network benefits in countries’ IFRS harmonization decisions is weaker

for countries with more bargaining power (larger countries).

The presence of network effects in the adoption of IFRS is significant because it means a

country can adopt IFRS even if its domestically developed accounting standards are particularly

well-suited to its domestic institutions. Moreover, if network effects contribute to the adoption of

IFRS, they can sustain its eventual dominance even in the presence of technologically superior

35

innovations (David, 1985).27 The evidence in this paper can also complement the growing

literature on the determinants and consequences of firms’ IFRS adoption. Firms’ adoption

decisions are conditional on countries’ decisions to harmonize with IFRS, suggesting that the

analysis of why countries adopt IFRS is important. Our evidence also implies that countries with

low bargaining power are more susceptible to adopting IFRS because others are doing so,

consistent with such countries being less distinctive in their approach to IFRS harmonization.

Ironically, low bargaining power countries also tend to have weak market institutions, implying

IFRS implementation in these countries is less likely to be effective.

The concept of network effects has recently been used to explain several accounting-

related phenomena, such as the adoption of stock-option compensation plans (Kedia and

Rajgopal, 2009) and the decision to expense options in the income statement (Reppenhagen

2010). We document that network effects influence accounting-related decisions more generally,

suggesting network theory can be used broadly in the literature on accounting and corporate

governance choice. For example, the theory of IFRS network benefits tested herein in the context

of countries can equivalently be applied to firms. Firm-level research on the determinants of

IFRS adoption may consider how perceived network benefits interact with economic incentives

in shaping adoption choices. Future firm-level research can also investigate whether network

benefits increase the attractiveness of voluntary IFRS adoptions to multinational as opposed to

domestic corporations.

27 For example, David notes how the QWERTY keyboard has remained the world standard despite longstanding experimental evidence (dating back to at least the 1940s) supporting alternative keyboard designs as superior.

36

References

Armstrong, C., M.E. Barth, A.D. Jagolinzer, and E.J. Riedl. 2010. Market reaction to the adoption of IFRS in Europe. The Accounting Review 85: 31–61.

Ball, R. 2006. International Financial Reporting Standards (IFRS): pros and cons for investors. Accounting & Business Research 36: 5–27.

——–, S.P. Kothari, and A. Robin. 2000. The effect of international institutional factors on properties of accounting earnings. Journal of Accounting & Economics 29: 1–51.

——–, A. Robin, and J. Wu. 2003. Incentives versus standards: Properties of accounting income in four East Asian countries. Journal of Accounting & Economics 36: 235–270.

Benston, G.J., M. Bromwich, R.E. Litan, and A. Wagenhofer. 2006. Worldwide financial reporting: The development and future of accounting standards. New York: Oxford University Press.

Box-Steffensmeier, J.M., and B.S. Jones. 2004. Event history modeling: A guide for social scientists. New York: Cambridge University Press.

Bradshaw, M. and G. Miller. 2008. Will harmonizing accounting standards really harmonize accounting? Evidence from non-U.S. firms adopting U.S. GAAP. Journal of Accounting, Auditing and Finance 23: 233–263.

Bushman, R. and J. Piotroski. 2006. Financial reporting incentives for conservative accounting: The influence of legal and political institutions. Journal of Accounting & Economics 42: 107–148.

Daske, H., L. Hail, C. Leuz, and R. Verdi. 2008. Mandatory IFRS reporting around the world: early evidence on the economic consequences. Journal of Accounting Research 46: 1085–1142.

David, P. 1985. Clio and the Economics of QWERTY. American Economic Review 75: 332–337.

DeFond, M., X. Hu, M. Hung, and S. Li. 2011. The impact of mandatory IFRS adoption on foreign mutual fund ownership: The role of comparability. Journal of Accounting & Economics 51: 240–258.

EC (Commission of the European Communities). 2000. EU Financial Reporting Strategy: The Way Forward. Brussels, Belgium: EC. http://www.iasplus.com/resource/cec.pdf.

——–. 2008. Table of decisions of Member States and EEA Countries concerning the use of options in the IAS Regulation. Brussels, Belgium: EC. http://ec.europa.eu/internal_market/accounting/docs/ias/ias-use-of-options_en.pdf.

Hail, L., C. Leuz, and P. Wysocki. 2010. Global accounting convergence and the potential adoption of IFRS by the U.S. (Part I): Conceptual underpinnings and economics analysis. Accounting Horizons 24: 355–394.

37

Hope, O., J. Jin, and T. Kang. 2006. Empirical evidence on jurisdictions that adopt IFRS. Journal of International Accounting Research 5: 1–20.

International Accounting Standards Board (IASB). 2008a. The IASC Foundation’s Annual Report 2007. London, UK: IASB. http://www.iasb.org/NR/rdonlyres/99A983BD-191F-474B-87A9-E25BB9D7093C/0/IASCF_annual_report_2007.pdf.

——–. 2008b. IASB and the IASC Foundation: Who we are and what we do. London, UK: IASB. http://www.iasb.org/NR/rdonlyres/0A5A767C-E7DE-49E5-8B12-499F62F8870C/0/WhoWeAre_Final12508.pdf.

——–. 2009. IASC Foundation Annual Report. London, UK: IASB. http://www.iasb.org/NR/rdonlyres/473C971C-2C13-444F-A94F-706C996F645C/0/Annualreport20092.pdf.

Katz, M.L., and C. Shapiro. 1985. Network externalities, competition, and compatibility. American Economic Review 75: 424–440.

Kedia, S., and S. Rajgopal. 2009. Neighborhood matters: The impact of location on broad based stock option plans. Journal of Financial Economics 92: 109–127.

Kothari, S., Ramanna, K., Skinner, D., 2010. Implications for GAAP from an analysis of positive research in accounting. Journal of Accounting & Economics 50: 246–286.

La Porta, R., F. Lopez-de-Silanes, and A. Shleifer. 2008. The economic consequences of legal origins. Journal of Economic Literature 46: 285–332.

Lang, M., M. Maffett, and E. Owens. 2010. Earnings comovement and accounting comparability: The effects of mandatory IFRS adoption. Working Paper #11-03, Simon School of Business.

Leone, M. 2008. “Spineless?” UK pressure targets fair value weakening. CFO.com, November. http://www.cfo.com/article.cfm/12586836?f=related.

Leuz, C. 2010. Different approaches to corporate reporting regulation: how jurisdictions differ and why. Accounting and Business Research 40 (3): 1–28.

Liebowitz, S.J., and S.E. Margolis. 1994. Network externality: An uncommon tragedy. Journal of Economic Perspectives 8: 133–150.

——– and ——–. 1998. Network externalities (effects). In The New Palgrave's dictionary of economics and the law, ed. Peter Newman. London: Palgrave MacMillan.

Lin, D.Y., and L.J. Wei. 1989. The robust inference for the Cox proportional hazards model. Journal of the American Statistical Association 84: 1074–1078.

Márquez-Ramos, L. 2008. The effect of IFRS adoption on trade and foreign direct investments. Working Paper #19, International Trade and Finance Association.

38

Ramanna, K., G.A. Donovan, and N. Dai. 2009. IFRS in China. Harvard Business School Case 110-037.

Ramanna, K., and E. Sletten. 2009. Why do countries adopt International Financial Reporting Standards? Working Paper #09-102, Harvard Business School.

Reppenhagen, D.A. 2010. Contagion of accounting methods: evidence from stock option expensing. Review of Accounting Studies 15: 629–657.

Skinner, D. 2008. The rise of deferred tax assets in Japan: The role of deferred tax accounting in

the Japanese banking crisis. Journal of Accounting & Economics 46: 218–239.

Therneau, T.M., and P. M. Grambsch. 2000. Modeling survival data: Extending the Cox model. New York: Springer-Verlag.

Währisch, M. 2001. The evolution of international accounting systems. Frankfurt am Main: Lang.

World Bank. 1999–2008. Reports on Observance of Standards and Codes. http://www.worldbank.org/ifa/rosc.html.

39

Appendix A, Defining the category for countries with an IFRS “convergence” project

Through a qualitative analysis of IFRS adoption patterns across countries, we encounter jurisdictions that have made some progress towards IFRS harmonization, but whose domestic standards continue to differ from IFRS as issued by the IASB. In particular:

1. Some countries develop national accounting standards that are based on, but not

identical to, IFRS. For example, the Institute of Chartered Accountants of Nepal lists the various Nepalese Accounting Standards (NAS) together with their corresponding IFRS equivalents. NAS are produced in English and appear to be derived from IFRS; however, not every IFRS has an equivalent NAS and extant NAS differ in wording from IFRS.

2. Other countries claim to have adopted IFRS but carve out certain standards or rules, often supplementing IFRS with local exceptions. For example, Ramanna et al. (2010) describe various exceptions to IFRS that China has carved out as a condition to converging with the global standards, including standards around consolidated financial reporting, fair-value accounting, and reversal of impairments.

3. Still other countries rely on “static” IFRS—that is, they use a version of IFRS that was effective as of some prior year. For example, PriceWaterhouseCoopers reports that current IFRS is neither required nor permitted for listed companies in Thailand, but that “Thai GAAP follows the 2005 version of IFRS,” with certain exceptions.

The countries described above cannot be properly classified as having adopted IFRS as

issued by the IASB. Accordingly, we reclassify the countries as IFRS “convergence” countries, where “convergence” represents a country’s efforts to reconcile its domestic accounting standards with IFRS, in lieu of directly adopting IFRS as issued by the IASB. Creating the “convergence” category results in differences between our score of IFRS adoption and that on Deloitte’s IASplus.com: for example, Nepal is classified as a “full adopter” on IASplus.com.

40

Appendix B, List of countries in the dataset Argentina Guyana Niger Armenia Haiti Oman Australia Honduras Pakistan Azerbaijan Hong Kong Panama Bahamas India Papua New Guinea Bahrain Indonesia Paraguay Bangladesh Iran Peru Barbados Israel Philippines Belarus Ivory Coast Qatar Benin Jamaica Russia Bermuda Japan Saudi Arabia Bolivia Jordan Singapore Bosnia and Herzegovina Kazakhstan South Africa Brazil Kenya Sri Lanka Burkina Faso Korea (South) Switzerland Canada Kuwait Syria Chile Kyrgyzstan Tajikistan China Laos Tanzania Colombia Lebanon Thailand Costa Rica Macedonia Togo Croatia Malawi Trinidad and Tobago Cuba Malaysia Tunisia Dominican Republic Mali Turkey Ecuador Mauritius Ukraine Egypt Mexico United Arab Emirates El Salvador Moldova United States European Union Morocco Uruguay Fiji Mozambique Venezuela Georgia Nepal Zambia Ghana New Zealand Zimbabwe Guatemala Nicaragua

41

Appendix C, Variable definitions

Variable DescriptionDependent variable: IFRS adoption status1. Non adopter Countries with no IFRS harmonization activity

A. Convergence project Countries attempting to reconcile their domestic accounting standards with IFRS, without directly adopting IFRS

B. Voluntary adoption Countries permitting at least some listed firms in their jurisdiction to adopt IFRS as issued by the IASB

C. Required for some Countries requiring IFRS as issued by the IASB for some listed firms in their jurisdiction

3. Full adoption Countries requiring IFRS as issued by the IASB for all listed firms in their jurisdiction

Primary independent variable: Network value of IFRSNetwork Predicted prior-year fraction of a country's trade with all countries

that have IFRS adoption status "3"

Network_Conv Predicted prior-year fraction of a country's trade with all countries that have IFRS adoption status "2A", "2B", "2C" or above

Network_Non-EU Predicted prior-year fraction of a country's trade with all non-EU countries that have IFRS adoption status "3" or above

Other independent variablesq(GDP) In-sample quartile rank of gross domestic product

q(Trade/GDP) In-sample quartile rank of the ratio of foreign trade to GDP

FEPI/GDP Ratio of foreign equity portfolio investment to gross domestic product

WB_Report Time-series indicator to denote whether a World Bank ROSC report was issued for the given country prior to the year in question

Info_Globalizaion Percentage of Internet users

BigAuditor Indicator equal to one for country-years in which Ernst & Young, KPMG, and PWC have offices located in the country.

2. Partial adoption

42

Table 1, IFRS adoption status in the country-year panel “Non adopter” refers to countries with no IFRS harmonization activity. “Partial adoption” refers to applying IFRS with exception or only for some firms in the economy and includes: “Convergence project,” “Voluntary adoption,” and “Required for some.” “Convergence project” refers to countries attempting to reconcile their domestic accounting standards with IFRS, without directly adopting IFRS. “Voluntary adoption” refers to countries permitting at least some listed firms in their jurisdiction to adopt IFRS as issued by the IASB. “Required for some” refers to countries requiring IFRS as issued by the IASB for some listed firms in their jurisdiction. “Full adoption” refers to countries requiring IFRS as issued by the IASB for all listed firms in their jurisdiction. Adoption Status 2003 2004 2005 2006 2007 2008 Total

1. Non adopter 61 54 48 44 36 30 2732. Partial adoption

A. Convergence project 6 8 8 11 14 15 62

B. Voluntary adoption 5 7 9 10 10 8 49

C. Required for some 2 4 3 3 4 7 23

3. Full adoption 18 19 24 24 28 32 145

Total 92 92 92 92 92 92 552

43

Table 2, Descriptive statistics for measures of perceived network benefits “Non adopter” refers to countries with no IFRS harmonization activity. “Partial adoption” refers to applying IFRS with exception or only for some firms in the economy and includes: “Convergence project,” “Voluntary adoption,” and “Required for some.” “Convergence project” refers to countries attempting to reconcile their domestic accounting standards with IFRS, without directly adopting IFRS. “Voluntary adoption” refers to countries permitting at least some listed firms in their jurisdiction to adopt IFRS as issued by the IASB. “Required for some” refers to countries requiring IFRS as issued by the IASB for some listed firms in their jurisdiction. “Full adoption” refers to countries requiring IFRS as issued by the IASB for all listed firms in their jurisdiction. Network is the predicted fraction of a country’s prior-year trade with all countries that have IFRS adoption status “3” (i.e., full adoption). Network_Conv is the predicted fraction of a country’s prior-year trade with all countries that have IFRS adoption status “2” or above. Network_Non-EU is the predicted fraction of a country’s prior-year trade with all non-EU countries that have IFRS adoption status “3.” Med. denotes median. Panel A

2003 2004 2005 2006 2007 2008 Adoption Status Variable Mean Med. Mean Med. Mean Med. Mean Med. Mean Med. Mean Med.

1. Non adopter Network 0.019 0.010 0.024 0.013 0.025 0.014 0.086 0.060 0.082 0.059 0.092 0.059

2. Partial adoption Network 0.015 0.011 0.027 0.019 0.026 0.019 0.117 0.060 0.129 0.106 0.124 0.079

3. Full adoption Network 0.039 0.015 0.048 0.041 0.047 0.046 0.095 0.072 0.087 0.072 0.097 0.079 Panel B

2003 2004 2005 2006 2007 2008 Adoption Status Variable Mean Med. Mean Med. Mean Med. Mean Med. Mean Med. Mean Med.

1. Non adopter Network_Conv 0.112 0.082 0.119 0.093 0.136 0.103 0.158 0.108 0.157 0.112 0.213 0.125 Network_Non-EU 0.019 0.010 0.023 0.011 0.024 0.012 0.033 0.016 0.032 0.017 0.038 0.023

2. Partial adoption Network_Conv 0.145 0.080 0.158 0.077 0.163 0.105 0.175 0.128 0.210 0.159 0.222 0.155 Network_Non-EU 0.015 0.011 0.025 0.018 0.023 0.013 0.032 0.021 0.037 0.024 0.041 0.027

3. Full adoption Network_Conv 0.063 0.045 0.073 0.063 0.104 0.076 0.121 0.087 0.120 0.091 0.149 0.115 Network_Non-EU 0.039 0.015 0.047 0.041 0.047 0.046 0.063 0.065 0.054 0.050 0.052 0.056

44

Table 3, Descriptive statistics for other independent variables “Non adopter” refers to countries with no IFRS harmonization activity. “Partial adoption” refers to applying IFRS with exception or only for some firms in the economy and includes: “Convergence project,” “Voluntary adoption,” and “Required for some.” “Convergence project” refers to countries attempting to reconcile their domestic accounting standards with IFRS, without directly adopting IFRS. “Voluntary adoption” refers to countries permitting at least some listed firms in their jurisdiction to adopt IFRS as issued by the IASB. “Required for some” refers to countries requiring IFRS as issued by the IASB for some listed firms in their jurisdiction. “Full adoption” refers to countries requiring IFRS as issued by the IASB for all listed firms in their jurisdiction. GDP is gross domestic product. Trade/GDP is the ratio of foreign trade to gross domestic product. FEPI/GDP is the ratio of foreign equity portfolio investment to gross domestic product. WB_Report is a time-series indicator to denote whether a World Bank ROSC report was issued for the given country prior to the year in question. Info_Globalization is the percentage of Internet users. BigAuditor is an indicator equal to one for country-years in which Ernst & Young, KPMG, and PWC have offices located in the country. Med. denotes median.

Adoption Status Variable Mean Med. Mean Med. Mean Med. Mean Med. Mean Med. Mean Med.1. Non adopter GDP 360.86 17.77 415.19 21.69 451.72 22.73 507.43 23.76 573.34 24.97 708.79 52.17

Trade/GDP 0.607 0.527 0.699 0.569 0.822 0.674 0.933 0.779 1.053 0.806 1.198 0.870FEPI/GDP -0.0006 0.0000 0.0007 0.0000 0.0022 0.0000 0.0024 0.0002 0.0016 0.0002 . .

WB_Report 0.033 0 0.094 0 0.255 0 0.302 0 0.314 0 0.379 0Info_Globalization 11.09 3.36 13.57 4.58 13.37 6.58 14.92 8.45 17.39 7.91 21.01 14.33

BigAuditor 0.644 1 0.654 1 0.617 1 0.605 1 0.600 1 0.552 1

2. Partial adoption GDP 155.03 80.68 127.64 24.92 178.07 59.15 164.37 60.33 244.23 99.59 228.76 52.02Trade/GDP 0.731 0.599 0.757 0.655 0.824 0.722 0.858 0.754 0.943 0.805 1.081 0.879FEPI/GDP 0.0004 0.0000 0.0009 0.0000 0.0009 0.0000 0.0030 0.0000 0.0060 0.0042 . .


BigAuditor 0.846 1 0.842 1 0.895 1 0.913 1 0.926 1 0.897 1

3. Full adoption GDP 15.27 9.17 17.72 10.71 38.94 11.67 40.72 12.53 35.02 10.53 53.17 11.47Trade/GDP 0.743 0.709 0.788 0.727 0.933 0.764 1.048 0.882 1.190 1.228 1.362 1.343FEPI/GDP -0.0024 0.0000 0.0006 0.0000 0.0025 0.0000 0.0076 0.0002 0.0071 0.0000 . .


BigAuditor 0.882 1 0.875 1 0.905 1 0.905 1 0.800 1 0.833 1

20082003 2004 2005 2006 2007

45

Table 4, OLS regressions of IFRS adoption status The dependent variable is IFRS adoption status which takes the value of 1 for non-adopters, 2 for partial adoption, and 3 for full adoption. Panel A presents OLS regressions including country fixed effects, while Panel B presents similar regressions including both country and year fixed effects. In columns (1) and (2), Nw is Network. In column (3), Nw is Network_Conv. In column (4), Nw is Network_Non-EU. Network is the predicted fraction of a country’s prior-year trade with all countries that have IFRS adoption status “3” (i.e., full adoption). Network_Conv is the predicted fraction of a country’s prior-year trade with all countries that have IFRS adoption status “2” or above. Network_Non-EU is the predicted fraction of a country’s prior-year trade with all non-EU countries that have IFRS adoption status “3.” q(GDP) is the in-sample quartile rank of gross domestic product. q(Trade/GDP) is the in-sample quartile rank of the ratio of foreign trade to gross domestic product. Standard errors are clustered by country (inferences are robust to clustering by country and year). Numbers in italics are p-values (two-tailed). Panel A – OLS regressions with country fixed effects

***, **, and * denote statistical significance at the two-tail 99%, 95%, and 90% confidence levels, respectively. The reported R Square is the Incremental R Square not accounting for explanatory power from country fixed effects.

(1) (2) (3) (4)Nw 1.696 *** 10.619 *** 7.908 *** 3.974

0.000 0.000 0.000 0.341

Nw*q(GDP) . -2.644 *** -2.008 *** -1.2730.000 0.000 0.233

Nw*q(Trade/GDP) . 0.315 0.507 *** 2.434 ***0.118 0.000 0.000

Constant 1.660 *** 1.488 *** 1.298 *** 1.494 ***0.000 0.000 0.000 0.000

Country F.E. Yes Yes Yes YesYear F.E. No No No No

N Obs. 552 514 514 508R Square 0.079 0.178 0.192 0.126

46

Panel B - OLS regressions with country and year fixed effects

***, **, and * denote statistical significance at the two-tail 99%, 95%, and 90% confidence levels, respectively. The reported R Square is the Incremental R Square not accounting for explanatory power from country fixed effects.

(1) (2) (3) (4)Nw 0.424 7.750 *** 5.382 ** 4.916

0.226 0.008 0.036 0.283

Nw*q(GDP) . -2.007 *** -1.578 ** -2.216 *0.005 0.013 0.086

Nw*q(Trade/GDP) . 0.197 0.291 ** 1.232 *0.310 0.040 0.075

Constant 1.523 *** 1.408 *** 1.371 *** 1.438 ***0.000 0.000 0.000 0.000

Country F.E. Yes Yes Yes YesYear F.E. Yes Yes Yes Yes

N Obs. 552 514 514 508R Square 0.208 0.258 0.257 0.233

47

Table 5, OLS regression of IFRS adoption status – including tests of alternative explanations The dependent variable is IFRS adoption status which takes the value of 1 for non-adopters, 2 for partial adoption, and 3 for full adoption. Panel A presents OLS regressions including either country or government fixed effects (as specified in the table), while Panel B presents similar regressions including year fixed effects and either country or government fixed effects. In all columns Nw is Network. Network is the predicted fraction of a country’s prior-year trade with all countries that have IFRS adoption status “3” (i.e., full adoption). q(GDP) is the in-sample quartile rank of gross domestic product. q(Trade/GDP) is the in-sample quartile rank of the ratio of foreign trade to gross domestic product. WB_Report is a time-series indicator to denote whether a World Bank ROSC report was issued for the given country prior to the year in question. Info_Globalization is the percentage of Internet users. BigAuditor is an indicator equal to one for country-years in which Ernst & Young, KPMG, and PWC have offices located in the country. Standard errors are clustered by country or government depending on the type of fixed effects included (inferences are robust to clustering by country and year). Numbers in italics are p-values (two-tailed). Panel A - OLS regressions with country (or government) fixed effects

***, **, and * denote statistical significance at the two-tail 99%, 95%, and 90% confidence levels, respectively.

The reported R Square is the Incremental R Square not accounting for explanatory power from country/government fixed effects.

(1) (2) (3) (4) (5) (6)Nw 9.507 *** 8.109 *** 8.583 *** 10.618 *** 10.989 *** 7.639 ***

0.000 0.002 0.001 0.000 0.000 0.008

Nw*q(GDP) -2.433 *** -1.917 *** -2.260 *** -2.612 *** -2.787 *** -1.870 **0.000 0.005 0.001 0.000 0.001 0.012

Nw*q(Trade/GDP) 0.365 * 0.028 0.255 0.271 0.287 -0.2580.068 0.901 0.205 0.163 0.308 0.331

WB_Report 0.178 ** 0.0290.014 0.761

Trade/GDP 0.380 ** 0.581 **0.030 0.011

Info_Globalization 0.019 *** 0.0160.001 0.201

BigAuditor 0.514 *** 0.424 **0.002 0.020

Constant 1.467 *** 1.215 *** 1.234 *** 1.107 *** 1.485 *** 0.512 **0.000 0.000 0.000 0.000 0.000 0.015

Government F.E. No No No No Yes YesCountry F.E. Yes Yes Yes Yes No No

Year F.E. No No No No No NoN Obs. 514 514 512 514 508 506

R Square 0.195 0.224 0.215 0.200 0.140 0.267

48

Panel B - OLS regressions with country (or government) and year fixed effects

***, **, and * denote statistical significance at the two-tail 99%, 95%, and 90% confidence levels, respectively. The reported R Square is the Incremental R Square not accounting for explanatory power from country/government fixed effects.

(1) (2) (3) (4) (5) (6)Nw 7.736 *** 7.660 *** 7.743 *** 8.265 *** 8.949 *** 9.175 ***

0.009 0.009 0.008 0.005 0.004 0.004

Nw*q(GDP) -2.006 *** -1.924 *** -2.014 *** -2.089 *** -2.348 *** -2.146 ***0.005 0.008 0.005 0.004 0.003 0.006

Nw*q(Trade/GDP) 0.204 0.115 0.192 0.163 0.092 -0.3320.298 0.603 0.328 0.399 0.699 0.224

WB_Report 0.021 0.0330.787 0.710

Trade/GDP 0.122 0.605 **0.508 0.012

Info_Globalization 0.003 0.0160.725 0.295

BigAuditor 0.359 * 0.460 **0.058 0.014

Constant 1.407 *** 1.330 *** 1.375 *** 1.148 *** 1.351 *** 0.454 *0.000 0.000 0.000 0.000 0.000 0.087

Government F.E. No No No No Yes YesCountry F.E. Yes Yes Yes Yes No No

Year F.E. Yes Yes Yes Yes Yes YesN Obs. 514 514 512 514 508 506

R Square 0.258 0.261 0.259 0.268 0.217 0.274

49

Table 6, Counting-process hazard regression of IFRS adoption status In column (1) and (2), Nw is Network and full adoption is defined as the failure event. In column (3), Nw is Network and convergence and higher levels of adoption are defined as the failure event. In column (4), Nw is Network_Conv and full adoption is defined as the failure event. In column (5), Nw is Network_Non-EU and full adoption is defined as the failure event. Network is the predicted fraction of a country’s prior-year trade with all countries that have IFRS adoption status “3” (i.e., full adoption). Network_Conv is the predicted fraction of a country’s prior-year trade with all countries that have IFRS adoption status “2” or above. Network_Non-EU is the predicted fraction of a country’s prior-year trade with all non-EU countries that have IFRS adoption status “3.” q(GDP) is the in-sample quartile rank of gross domestic product. q(Trade/GDP) is the in-sample quartile rank of the ratio of foreign trade to gross domestic product. WB_Report is a time-series indicator to denote whether a World Bank ROSC report was issued for the given country prior to the year in question. Info_Globalization is the percentage of Internet users. BigAuditor is an indicator equal to one for country-years in which Ernst & Young, KPMG, and PWC have offices located in the country. Standard errors are clustered by country. Numbers in italics are p-values (two-tailed).


(1) (2) (3) (4) (5)Nw 5.028 *** 22.185 ** 11.059 * 6.548 21.102

0.001 0.014 0.072 0.419 0.165

q(GDP) -0.854 *** -0.596 *** -0.292 ** -0.539 ** -0.667 *0.000 0.003 0.029 0.045 0.075

Nw*q(GDP) -4.175 ** -2.714 * -1.906 -4.4770.029 0.062 0.265 0.696

q(Trade/GDP) 0.214 0.332 -0.009 0.202 0.3070.235 0.161 0.952 0.470 0.327

Nw*q(Trade/GDP) -0.807 0.729 0.103 1.4100.633 0.346 0.944 0.871

WB_Report 0.461 0.484 0.270 0.350 0.650 *0.159 0.151 0.181 0.284 0.077

Info_Globalization 0.001 -0.001 0.007 0.002 0.0030.942 0.951 0.231 0.860 0.777

BigAuditor 1.305 ** 1.013 ** 1.306 *** 1.243 ** 0.970 *0.013 0.043 0.001 0.024 0.067

N Obs 512 512 512 512 506pseudo R Square 0.172 0.200 0.140 0.153 0.250

50

Table 7, Multinomial logit of IFRS adoption status with non-adoption as the base case In columns (1) and (2), Nw is Network. In column (3), Nw is Network_Conv. In column (4), Nw is Network_Non-EU. Network is the predicted fraction of a country’s prior-year trade with all countries that have IFRS adoption status “3” (i.e., full adoption). Network_Conv is the predicted fraction of a country’s prior-year trade with all countries that have IFRS adoption status “2” or above. Network_Non-EU is the predicted fraction of a country’s prior-year trade with all non-EU countries that have IFRS adoption status “3.” q(GDP) is the in-sample quartile rank of gross domestic product. q(Trade/GDP) is the in-sample quartile rank of the ratio of foreign trade to gross domestic product. WB_Report is a time-series indicator to denote whether a World Bank ROSC report was issued for the given country prior to the year in question. Info_Globalization is the percentage of Internet users. BigAuditor is an indicator equal to one for country-years in which Ernst & Young, KPMG, and PWC have offices located in the country. Standard errors are clustered by country. Numbers in italics are p-values (two-tailed).


(1) (2) (3) (4)

Nw 1.768 4.021 -5.816 20.8330.525 0.778 0.530 0.427

q(GDP) -0.145 -0.086 -0.115 0.1760.597 0.765 0.732 0.630

Nw*q(GDP) -2.204 0.281 -9.6240.517 0.901 0.148

q(Trade/GDP) 0.009 -0.179 -0.348 -0.0940.968 0.504 0.250 0.761

Nw*q(Trade/GDP) 2.100 1.812 * 1.3280.208 0.096 0.797

WB_Report -0.034 -0.032 -0.096 -0.0640.941 0.944 0.839 0.894

Info_Globalization 0.010 0.011 0.013 0.0100.471 0.391 0.280 0.439

BigAuditor 1.847 ** 1.839 ** 1.998 ** 1.693 **0.014 0.015 0.022 0.033

Constant -2.816 *** -2.640 *** -2.182 ** -3.305 ***0.004 0.010 0.019 0.004

Nw 7.732 *** 30.517 ** 4.678 43.579 **0.004 0.012 0.669 0.032

q(GDP) -1.162 *** -0.832 ** -0.815 ** -0.767 *0.000 0.013 0.043 0.073

Nw*q(GDP) -6.626 ** -1.736 -10.4030.020 0.470 0.136

q(Trade/GDP) 0.330 0.334 0.167 0.3790.223 0.306 0.656 0.337

Nw*q(Trade/GDP) 0.424 0.873 1.2520.866 0.661 0.840

WB_Report 0.659 0.634 0.471 0.8460.238 0.260 0.389 0.135

Info_Globalization 0.005 0.006 0.009 0.0050.753 0.744 0.604 0.752

BigAuditor 2.124 *** 1.798 *** 2.086 *** 1.687 **0.001 0.007 0.003 0.014

Constant -1.430 * -2.081 ** -1.737 * -2.422 **0.063 0.016 0.084 0.029

Year F.E. Yes Yes Yes YesN Obs. 512 512 512 506

Pseudo R Square 0.159 0.177 0.154 0.206

Pr(Partial adoption)

Pr(Full adoption)

network effects in countries adoption of ifrs

Documents