A study about the characteristics of listed Japanese firms that voluntarily adopt IFRS

(1)

Amsterdam Business School

MSc Accountancy and Control, Accountancy track

Faculty of Economics and Business, University of Amsterdam

Master thesis:

A study about the characteristics of listed Japanese

firms that voluntarily adopt IFRS

Ruby Dirkse

Name: Ruby Dirkse

Student number: 10212892

Thesis supervisor: Prof. Dr. V.R. O’Connell Date: June 20, 2016

MSc Accountancy & Control, specialization: Accountancy Faculty of Economics and Business, University of Amsterdam

(2)

Statement of Originality

This document is written by student Ruby Dirkse, who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

This study examines the determinants of voluntary adoption of International Financial Reporting Standards (IFRS) by Japanese listed firms. The study is motivated by the increasing trend of Japanese listed firms that voluntarily adopt IFRS and the ongoing debate about the costs and benefits of IFRS adoption. Based on prior literature and agency theory I expect that: international orientation, size, leverage, ownership diffusion and industry type are important drivers in the accounting standards choice. I examine the determinants of voluntary IFRS adoption by Japanese listed firms (NIKKEI 225 during the time period 2005-2014) by performing a probit regression analysis. In this study I find that international orientation and size help explain Japanese listed firms’ choice of voluntarily adopting IFRS. The firm characteristics leverage, ownership diffusion and industry type do not appear to play a significant role in the decision.

(4)

Acknowledgements

First of all, I like to express my appreciation to my parents, sister and friends for being there for me, for the support and their patience. Specifically, I would like to thank my friend Ernst de Lange for all the educational and fun days studying together in the library. Secondly, I would like to thank my supervisor at the Amsterdam Business School, Vincent O’Connell for the guidance and direction he gave me, I am very grateful for that. Finally, I would like to thank Mohamed el Kourai, who supervised me during the internship at PricewaterhouseCoopers (PwC).

(5)

1 Introduction

Over the last years, regulatory and economic changes have incentivized Japanese listed firms to voluntarily prepare their financial statements according to International Financial Reporting Standards (IFRS). The International Accounting Standards Board (IASB) is largely responsible for developing IFRS as a single set of financial reporting standards that can be used world-wide. The convergence of financial reporting standards is supported by the assertion that a single set of reporting standards provides more transparency and comparability (Amstrong, Barth and Jagolinzer, 2010). Over 130 countries believe that IFRS enhance the transparency and comparability of financial statements and adopt IFRS as financial framework. In Japan IFRS adoption is permitted since 2010 but it is still on a voluntary basis. Accounting-setting bodies in Japan sought to converge Japanese generally accepted accounting principles (Japanese-GAAP) with IFRS. However, in Japan there is no ‘’direct adoption’’ of IFRS but a ‘’cautious convergence approach’’ (Tsunogaya, 2016). Japan tries to harmonise rather than replace the accounting standards. Still, there is an increasing trend of Japanese listed firms that voluntarily adopt IFRS. This study examines the determinants of voluntary IFRS adoption by Japanese listed firms and is motivated by the increasing trend of IFRS adoption in Japan and the ongoing debate about the costs and benefits of adopting IFRS.

IFRS gained international importance after it became mandatory for public firms in the European Union to prepare their financial statements in accordance with IFRS, from January 2005 onward. Before IFRS adoption in the European Union became mandatory, firms already had the possibility to voluntarily adopt IFRS. Studies were conducted to examine the determinants of these ‘early’ adopters (e.g. Dumontier and Raffournier 1998; Cuijpers and Buijink 2005; Gassen and Sellhorn 2006). The prior literature about voluntary IFRS adoption suggests that firms that adopt IFRS tend to be more internationally orientated, are larger, are audited by a Big 4 firm and have more ownership diffusion. There are no findings that firms that adopt IFRS tend to have more capital intensity and are more profitable. Results about whether a firm is more leveraged are mixed. In this study I examine if the determinants international orientation, size and leverage are positively associated with voluntary IFRS adoption for Japanese listed firms.

To provide empirical evidence this study conducts a probit regression analysis. In the model a binary variable is regressed on the explanatory variables that are expected to influence the binary variable (Cuijpers and Buijnk, 2005). For the tests in this study I use Japanese firms that are listed on the NIKKEI 225. The observations are from 2005 until 2014. The results are based on a significance level of 5 %.

(8)

The findings from this study show that: (I) there is a significant positive relation between IFRS adoption and the size of a firm. In particular, the larger a firm the more the firm is likely to benefit from IFRS adoption. (II) There is a significant positive relation between IFRS adoption and the level of international orientation of a firm. In general, firms that have more foreign sales are more likely to comply with IFRS in order to meet demands of all foreign stakeholders. (III) There is no significant positive relation between IFRS adoption and leverage. The relationship between leverage and IFRS adoption was not clear a priori. In this study a positive relation between leverage and IFRS adoption was expected, based on agency theory. But in this study it seems that agency theory does not hold for this relation. The findings are robust to the checks for heteroscedasticity, multicollinearity and independence. Furthermore, the robustness of the model is tested by sensitivity checks of an alternative model and a matched-paired design. Also these results ensure the robustness of the model.

This study contributes in several ways. First, it extends the existing literature on voluntary IFRS adoption. Nowadays, Japan is unique because it is one of the few countries where IFRS adoption is still voluntary. Japan was quite a late bloomer in adopting IFRS, considering it is the fourth largest world economy. However, this offers the opportunity to complement the existing European studies. Second, because IFRS adoption is still voluntary in Japan it allows for a better understanding of the costs and benefits of IFRS adoption. Finally, standard setters and securities regulators in Japan should have (potential) interest in the effects of voluntary IFRS adoption, because there is an increasing trend of IFRS adoption in Japan. In 2010 only ten firms reported under IFRS and in 2015 this number increased to 112 (20% of the TSE market capitalisation).

This study is organised as follows. In section 2, the background information, theory and research questions are described. In section 3 a review on the firm characteristics of voluntary IFRS adoption is provided and the hypotheses are formulated. Section 4 contains the research model and variables and describes the data collection and research sample. Section 5 presents the empirical results. Finally, section 6 presents the conclusions of this study.

2 Background and theory

In this section I first give some background information about the current accounting policy in Japan, describe the objectives of IFRS and go over the ongoing debate about mandatory IFRS adoption in Japan. Second, I explain the agency theory as a theoretical underpinning of this study. Finally, I formulate the research questions.

(9)

2.1 Accounting standards in Japan and the debate about mandatory IFRS adoption

Over the last years, standard setters tried to enhance accounting standards which resulted in a unique accounting situation in Japan. Hans Hoogervorst, Chairman of the International Accounting Standards Board (IASB), describes the situation around the Japanese accounting standards as follow:

‘’The situation in Japan is unique and insightful; companies have a choice between several sets of standards. There has been a significant increase, in a short period of time, in the number of companies voluntarily adopting IFRS. This demonstrates that these companies find IFRS helpful for their international business activities.’’

Japanese listed firms have the choice between four different sets of accounting standards: designated IFRS, JMIS, Japanese GAAP and US GAAP. Historically, most Japanese listed firms report under the Japanese GAAP. However, since 2010 there is an increasing trend of IFRS adoption since Japanese listed firms were permitted to report under IFRS. In 2010 only ten firms reported under IFRS and in 2015 this number increased to 112 (20% of the TSE market capitalisation). Since 2007 the Accounting Standards Board of Japan (ASBJ) and the International Accounting Standards Board (IASB) try to eliminate main differences between Japanese GAAP and IFRS1_{. This convergence}

between Japanese GAAP and IFRS is seen by both parties as an important step towards possible mandatory adoption of IFRS in Japan. In Japan there are opponents and proponents for mandatory IFRS adoption, the decision whether to make IFRS mandatory still needs to be made (IAS, 2015).

After the cautious convergence of Japanese GAAP and IFRS by the elimination of the main differences, the two standards are not identical but can be considered equivalent in quality (according to the ASBJ and IASB). Although the quality of the standards are equivalent, there is still an increasing trend in voluntary IFRS adoption in Japan. This illustrates that Japanese listed firms consider disclosure under Japanese GAAP not optimal and believe in an improvement by voluntarily adopting IFRS.

The main objectives of IFRS is to improve reporting quality and improve the usefulness of financial reporting information for the users, by means of increasing transparency and comparability (Amstrong et al, 2010). The economic and regulatory developments have motivated firms worldwide to voluntarily adopt IFRS. According to prior literature, firms voluntarily adopt IFRS to show their commitment with transparency, because often IFRS requires more disclosure than national accounting standards. In section 3.1 I go more in depth about the incentives for voluntary IFRS adoption.

(10)

Even though there is an increasing trend of IFRS adoption in Japan, using IFRS is still not mandatory. In Japan there is a debate going on about making IFRS adoption mandatory for listed firms. There are groups that either support the mandatory adoption of IFRS or various stakeholders that are against mandatory IFRS adoption. Tsunogaya (2016) investigates if members of the standard setting body; Business Accounting Council (BAC), change their opinion about mandatory IFRS adoption across different time periods. The results of this study indicate that there is a higher level of disapproval of mandatory IFRS adoption by representatives from: accounting academics, the manufacturing industry and the Financial Services Agency (FSA) than by representatives of the Japanese Institute of Certified Public Accountants (JICPA). Another result of the study is that over time the level of disapproval became higher.

In Japan there is a high percentage of manufacturing firms (table 2). According to Tsunogaya (2016) Japanese manufacturing firms are against mandatory IFRS adoption. Manufacturing firms prefer Japanese-GAAP to IFRS the arguments include prudence, a long-term perspective, contextual factors and the relationship with taxation. Manufacturing firms recommend a carve-out of procedures and like a development of a specific Japanese-IFRS. In the recommended carve-out procedures of the IASB-IFRS are deleted and modified to Japanese-GAAP. The development of a Japanese-IFRS will result in controversy, the majority of the BAC members express their concerns about the coexistence of different accounting standards (Japanese-GAAP, Japanese-IFRS, IASB-IFRS). It would decrease comparability which is one of the objectives of IFRS. Proponents argue that IFRS adoption is important for globalization of business activities. Mandatory IFRS adoption would increase attractiveness to the Japanese markets and enhance comparability of financial statements. By establishing the Japanese Federation for Accounting Education and Learning (JFAEL) proponents try to improve judgement skills because IFRS is more principle-based than rule-based

2.2 Agency theory

Like described in paragraph 2.1 there is an increasing trend of IFRS adoption by Japanese listed firms. Japanese listed firms that already voluntarily adopt IFRS believe that IFRS adoption improves reporting quality and the usefulness of financial reporting information for the users (FSA, 2015). In the prior literature on this topic (and on voluntary disclosure) the agency theory is considered as an underlying theory (e.g. Dumontier and Raffournier, 1998; Raffournier, 1995). Therefore in this study the agency theory is used as underpinning theory as well.

(11)

parties. The agency relation discussed in this study is the relation between the capital providers (both equity and debt) of a firm (principals) and the management (agents). The agency theory assumes that everyone maximizes his or her own interest. People are driven by self-interest although agents should serve in the best interest of the principals. The agency theory explains that there is information asymmetry between principles and agents because the management (agents) has more information about the value of a firm (Jensen and Meckling, 1979). Information asymmetry causes two kind of problems: adverse selection problems and moral hazard problems. Adverse selection occurs when potential investors have to decide whether to invest in a company regardless of the type of financing (debt or equity). This decision is purely based on the expected ability of the firm to generate future profits. Moral hazard problems can arise when there are information problems between current investors and the management. Investors try to monitor management to let them behave in their best interest. Both the capital providers and management have incentives to reduce information asymmetry. The management (agents) needs capital and the capital providers (principals) want to have a good return on their investment and let their capital be managed well.

To reduce information asymmetry information needs to be provided. Disclosure of financial statements is one way of providing information. Although it is not necessarily the case that more disclosure is better. The degree the principals consider the disclosure creditable can only reduce information asymmetry (Wallace, 1980). All listed firms use a financial framework (e.g. IFRS or Japanese GAAP) to disclose their financial statements. Dumontier and Raffournier (1998) the first study on characteristics for choosing a financial framework, made use of the agency theory. They assume that IAS give financial statements more credibility because IAS increase disclosure. In this way compliance with IAS may function as a monitoring activity or as bonding activity, which reduces moral hazard. Especially, for small investors financial statements is the main/(only) source of information (Epstein and Anderson, 1994). For these reasons Dumontier and Raffournier (1998) assume that compliance with AIS increases with ownership diffusion. Meek, Roberts and Gray (1995) is a study on voluntary disclosure of financial information. The authors argue that voluntary disclosure increases with the level of leverage. For firms with a greater amount of debt agency costs are larger. The reason for higher agency costs is the type of contract/relationship the management has with shareholders and debtholders and the incentives of all parties. Shareholders and debtholders have conflicting incentives. Shareholders are willing to take more risk to make more money and on the other hand debtholders only concerned about whether to get their invested money back. Besides the agency theory I also rely on the prior literature. The prior literature is discussed in depth in paragraph 3.1.

(12)

2.3 Research questions

The objective of this study is to examine which specific firms’ characteristics influence the choice for voluntary IFRS adoption. As described in paragraph 2.1 there is an increasing trend of Japanese listed firms voluntarily adopting IFRS. Over the last years, there is worldwide an ongoing debate about the costs and benefits of IFRS adoption. The unique situation2_{in Japan enables measuring the costs and}

benefits of IFRS adoption. Altogether, this was a main driver for me to write my master thesis on this topic. In this study the general research question is:

What are the determinants for Japanese listed firms to voluntary adopt International Financial Reporting Standards?

In this study it is not possible to identify all firm specific characteristics of Japanese listed firms that voluntarily adopt IFRS. Therefore, I identify three firm specific characteristics of voluntary IFRS adoption. I derive the characteristics from the prior literature, which is explained further in the literature review (chapter 3). The three characteristics I examine in this study are: international orientation, leverage and size. The research questions concerning the characteristics are:

Research question I: To what extent is international orientation associated with the voluntary adoption of IFRS in Japan?

Research question II: To what extent is size associated with the voluntary adoption of IFRS in Japan? Research question III: To what extent is leverage associated with the voluntary adoption of IFRS in Japan?

In the section hypotheses development in paragraph 3.2, I make assumptions whether the characteristics are positively or negatively associated with IFRS adoption. And after performing the regression analyses, the predicted negative/positive relationships can be supported or rejected.

(13)

3 Literature review and hypothesis development

3.1 Prior research: determinants for voluntary IFRS adoption

In this study I examine why Japanese listed firms voluntarily adopt IFRS. To make assumptions I rely on the existing literature and the agency theory. The existing literature I use has to be similar to this study. This means the following conditions have to be met: (I) in the study firms could voluntarily adopt IFRS and (II) all firms are listed.

In this study I decide to examine only listed firms because up to April 2015 only two unlisted firms voluntarily adopt IFRS in Japan (FSA, 2015). In prior studies listing status is examined as independent variable that influences the choice for accounting standards. The study of Grey and Street (2002) examines eleven different characteristics of firms that adopted IAS in 1998. One of these characteristics is listing status. The authors find that listing status is positively significantly associated with IAS adoption. Especially when firms are internationally listed, the propensity is higher to comply

with IAS.

The first study about what firms motivate to voluntarily adopt IAS was conducted by Dumontier and Raffournier (1998). They argue that the establishment of the International Accounting Standards Committee (IASC) in 1973 had a large impact all over the world. More companies were likely to comply with IAS, especially in countries which had old national accounting standards. In general older national accounting rules are less stringent than IAS. The expectation in this study therefore was that although the switch should be very costly it also results in a more beneficial situation. In Switzerland forty percent of the listed firms made use of IAS in their consolidated accounts. This large percentage was a driver of the study of Dumontier and Raffournier to investigate the drivers of voluntary IAS adoption. Because this was the first study in the research area the authors used characteristics that were drawn upon studies about the determinants of voluntary disclosure3: listing status, international orientation, size, ownership structure, leverage, capital intensity, profitability, and auditors’ reputation. The study excluded banks and insurance companies because of their specific disclosure requirements. The analysis shows a significant positive relationship for: size, international orientation, listing status, auditor type, and ownership diffusion. There was no significant relationship found for: leverage, profitability and capital intensity. The results demonstrate the major role that outsiders play in the firms’ choice for accounting standards. Swiss firms were likely to show their commitment with higher reporting quality to outsiders.

3_{The basis of Dumontier and Raffournier (1998) is the work of Raffournier (1995) who studies the determinants for voluntary financial}

(14)

More studies followed about determinants for the choice of a specific set of accounting standards. Cuijper and Buijnk (2005) examine the determinants of voluntary adoption of non-local GAAP for firms domiciled in the European Union. They examine the annual reports of firms using IAS or US GAAP (non-local GAAP) in 1999. One finding of the study is that the amount of firms using non-local GAAP by non-financial firms was relatively low. It seemed that the adoption of standards other than the local standards doesn’t bring any net economic benefits. Although, the authors find that the firms choosing to report under non-local GAAP have some common characteristics. The firms that use US GAAP or IAS are more likely to: be listed on US exchange or the EASDAQ exchange, are more internationally orientated by means of having more geographically dispersed operations, and are larger on average than firms using local GAAP. Another finding of the study is that firms adopting non-local GAAP are more likely to be domiciled in countries with lower quality of financial reporting standards and in countries where it is explicitly allowed to report under IAS ( e.g. Switzerland).

Gassen and Sellhorn (2006) was one of the first studies that investigates the characteristics of voluntary IFRS adoption by German listed firms. They investigate German firms during the period 1998 until 2004 (until the last year IFRS adoption in Europe was still voluntary for listed firms). Besides determining the motivations for voluntary IFRS adoption, the authors also try to compare the characteristics of IFRS firms with the characteristics of German-GAAP firms. The results indicate that adopters have more persistent, less predictable and more conservative earnings, which means that earnings are of a higher quality. Also they find that IFRS firms experience a lower level of information asymmetry on the German equity market. This is beneficial for IFRS firms. But on the other hand, the authors also find that the level of share price volatility is significantly higher at IFRS firms. Knowing this, the authors try to determine the actual motivations of IFRS adoption. The conclusion of their study is that voluntary IFRS adoption is significantly influenced by: size, international exposure, and dispersion of ownership. They also find that especially young firms that went public in the mid-1990s are more likely to adopt IFRS.

Christensen, Lee, Walker & Zeng (2015) is the most recent study about voluntary IFRS adoption. The authors suggest that IFRS adoption per se does not change accounting quality, only when a firm has specific characteristics IFRS adoption can improve accounting quality. The authors study German listed firms and make a distinction between the groups, voluntary IFRS adopters (with incentives to change) and mandatory IFRS adopters (without incentives to change). The authors suggest that voluntary IFRS adopters (‘early’ adopters) did so because the management benefits from IFRS adoption. Furthermore, they find no accounting quality improvement for firms that postponed

(15)

IFRS adoption up to 2005 (‘late adopters’). IFRS adoption is only beneficial when a company has specific incentives/characteristics for adoption. According to the authors the ‘early’ IFRS adopters have the following characteristics: have higher growth rates, issue more equity and debt securities, have greater sales, are larger and listed on more exchanges, and are more likely to be audited by a big 4 company.

Besides all prior studies on voluntary IFRS adoption in Europe, there is also a survey conducted by the FSA about IFRS adoption in Japan. The first part of this survey describes the motivation of Japanese listed firms to voluntarily adopt IFRS. The reasons for the sixty-five surveyed firms were (ranked in the level of importance): it contributes to business management, it improves comparability, it makes explanations to foreign users easier, it gives a better reflection of performance, it makes finance from abroad smoother and several other reasons. The surveyed firms give further explanations for their reasons in the report (FSA, 2015).

The studies on voluntary IFRS adoption described above identify the determinants of firms that voluntarily adopt IFRS. I use these studies together with the agency theory to shape my hypotheses.

3.2 Hypotheses development

In this paragraph I shape the hypotheses. I explain why international orientation, size and leverage can be identified as drivers for voluntary IFRS adoption by Japanese listed firms. A theory I rely on is the agency theory (2.2) and the hypotheses are developed on the basis of the prior literature (3.1).

3.2.1 International orientation

Prior literature on voluntary IFRS adoption implies that firms which are more internationally orientated experience more benefits from IFRS adoption. Therefore according to the existing literature it is reasonable that a firm that is more internationally orientated has a higher propensity to adopt IFRS (André, Walton, & Yang 2012; Christensen et al, 2012; Cuijpers and Buijink, 2005; Dumontier and Raffournier,1998; Gassen and Sellhorn, 2006). In the prior studies international orientation is measured as: the percentage of foreign sales, the number of overseas subsidiaries, or on which (and the number of) stock exchanges firms are listed. In this study I measure international orientation (like most studies) by the percentage of foreign sales.

Existing literature highlights that more internationally orientated firms: are more motivated to comply with IFRS in order to meet foreign investor demands, are motivated to comply with foreign exchange regulations, and try to reduce restatement costs and try to offer more transparency. (André

(16)

et al 2012; Cuijper and Buijink 2005).

Cuijpers and Buijnk (2005) describe that international operating firms have more heterogeneous stakeholders, than firms that are more nationally orientated. They assume that a more heterogeneous group of stakeholder demands for a common measure and that internationally orientated firms therefore are more likely to offer standardization in their financial statements and disclosures. Furthermore, when a firm has more heterogeneous stakeholders, it is time- and cost consuming for a firm to constantly explain national-GAAP to important international stakeholders. Murphy (1999) explains that especially the level of heterogeneity of international stakeholders plays a role in the choice for reporting standards and not the heterogeneity of national stakeholders. When a firm operates in several countries a firm is required to report to various constituents. More authorities need to be satisfied with understandable accounting standards. IFRS are in general more understandable than a national-GAAP.

Sequential to the results and the argumentation described in the prior literature, I assume that international orientation is positively associated with voluntary adoption of IFRS. And therefore the first hypothesis is formulated as follow:

H1: Internationally orientated Japanese listed firms are more likely to adopt IFRS

3.2.2 Size

I follow the prior literature in analysing firms’ characteristics that influence the choice of voluntary IFRS adoption. In almost all prior studies the variable size is included, both as a determinant or a control variable. In this study I include size as determinant for voluntary IFRS adoption. According to the prior literature the propensity to adopt IFRS is higher for larger firms (Christensen et al, 2012; Cuijpers and Buijink, 2005; Dumontier and Raffournier,1998; Gassen and Sellhorn, 2006). In most studies size is measured as the natural logarithm of market capitalisation. There are also studies that measure size as: a natural logarithm of total assets, a natural logarithm of total sales and by the number of employees. In this study I measure size (like most studies) as a natural logarithm of market capitalisation.

According to Watts and Zimmerman (1986) firm size is positively related to political costs. Therefore according to Dumontier and Raffournier (1998) larger firms are more likely to adopt IAS. Compliance with IAS generates confidence for authorities because IAS is more transparent and makes earnings management more difficult. Furthermore the authors suggest that it is less costly to comply with IAS for larger firms. In fact, larger firms are already likely to produce this information for

(17)

internal purposes. Also they assume that there is something like a fixed component to disclosure costs, which decreases ‘’cost of unit per size’’. Also according to André et al (2012) higher political costs lead to voluntary IFRS adoption for larger firms. The authors infer that larger firms have more financial resources which permit the change to IFRS. Furthermore larger firms are more likely to be more internationally orientated. To show this interrelationship I include the correlations between the independent variables (table 6).

Based on the results and the argumentation of the prior literature I assume that size and voluntary IFRS adoption are positively associated. The formulation of the second hypothesis is therefore stated as follow:

H2: Larger firms are more likely to adopt IFRS

3.2.3 Leverage

In line with the prior literature, I include the variable leverage as a determinant for voluntary IFRS adoption. In prior studies, the results about the role of leverage in voluntary IFRS adoption are mixed. Meek et al (1995) expect that a higher level of leverage increases voluntary disclosures. they argue that agency costs increase at firms with proportionally more debt. In contrast, Zarzeski (1996) expects a negative relation between voluntary disclosure and the level of leverage. He assumes that firms with respectively high debt ratios are located in countries with strong bank relationships, and have access to more private information. Stakeholders in such countries therefore rely more on private rather than public information. Murphy (1999) also expects a negative relation because lower leveraged firms are more dependent on equity capital this means firms are subjected to shareholders’ demand for more disclosed information.

Because of the mixed results of the role of leverage, I use agency theory as underpinning theory in this study (like: Dumontier and Raffournier, 1998). Dumotier and Raffournier (1998) assume that IAS give financial statements more credibility for debtholders. Adopting AIS may function in this way as monitoring activity or as bonding activity and reduces moral hazard. Like in prior studies, I measure leverage with the total-debt-on-total-assets ratio. (André et al, 2012; Cuijpers and Buijink, 2005; Gassen and Sellhorn, 2006).

(18)

3.3 Control variables

To isolate the effect of the independent variables (paragraph 3.2) on voluntary adoption of IFRS I include control variables. The control variables I include in this study are: industry and ownership structure. In this paragraph I briefly outline the control variables.

The first control variable is industry, the level of voluntary adoption of reporting standards may differ between industries because of industry specific accounting regulation, competition and/or proprietary costs (Meek et al 1995). These factors lead to different costs and benefits for voluntary adoption of a specific financial framework. According to Tsunogaya (2016) and Gassen and Sellhorn (2006) it can be assumed that manufacturing firms (SIC below 4000) are less likely to voluntarily adopt IFRS. Therefore, in this study I include a dummy variable for industry and I expect that manufacturing firms are less likely to voluntarily adopt IFRS in Japan.

The second control variable is ownership dispersion. This variable is a driver in the choice for a financial framework in the prior literature on listed firms (Dumontier and Raffournier 1998; Cuijpers and Buijink 2005; Gassen and Sellhorn 2006). Firms with many inside stakeholders have less incentives to disclose high quality information than firms that really rely on outside stakeholders (Cuijpers and Buijnk, 2005). The distance between capital providers (principals) and the management (agents) in firms with many outside stakeholders is larger and therefore there is a higher demand for more disclosure. Ding, Richard and Stolowy (2008) describe the change from a stakeholders view to a more shareholder view in different phases. In the phases the authors describe that over time more distance came between capital providers and the management. The distance results in changes made by the IASB in order to meet the demands (more disclosures) of the capital providers. In this study I therefore expect that firms with more ownership dispersion are more likely to adopt IFRS.

4 Research design

In this section I describe the research model, variables, and the data collection.

4.1 Probit model

Existing literature uses a logistic (logit) regression or a probability (probit) regression, to test the hypothesized relationships of firm characteristics and voluntary IFRS adoption. (André et al 2012; Cuipers and Buijnk, 2005; Dumontier and Raffournier, 1998; Gassen and Sellhorn, 2006). In both regressions a binary variable (e.g. yes/no, true/false, agree/disagree) is regressed on the explanatory variables, that are expected to influence the binary variable (Cuijpers and Buijnk, 2005). In this study

(19)

the binary variable is IFRS adoption/no IFRS adoption. Both regressions are appropriate to use in this study, however in this study I believe the probit regression fits best. Hahn and Soyer (2005) investigate the differences between the logit regression model and the probit regression model and when to use which model. In univariate tests the two models essentially give similar results but in multivariate tests this is not always the case. The authors find that the logit regression4 model fits better in the presence of extreme independent variable levels. On the other hand in random effects models with moderate size data sets, model fit is improved by selecting the probit regression model. When there is a moderate data set and random effects, the probit model seems to offer a more consistent advantage over the logit model from the perspective of minimizing deviance and enhancing model fit. The model I use can be considered as a random effects model5 with a moderate size6. The probit model takes the following form:

Summary of symbols: Pr: denotes probability

Y: IFRS adoption (1)/no IFRS adoption (0) X: is a vector regressors which influences size

Φ: the cumulative distribution function of the standard normal distribution β: are estimated by maximum likelihood

4.2

Variables

In this study I use a probit regression to test the relationship of firm characteristics and voluntary IFRS adoption in Japan. The main variable of interest, is a dummy variable that indicates IFRS adoption in Japan. Explanatory variables which are (based on the proposed hypothesis) expected to influence the variable of interest are in this study: IOi, SIZEi and LEVi. The three independent variables are controlled by the variables: OWNi and the dummy variable MANUFi. The information needed to determine the variables is retrieved from Datastream. Below I explain the variables. The choice for

4_{In paragraph 5.4.1 I run a logistic regression as a sensitivity test}

5_{The Nikkei 225 includes companies from all kind of industries and all observations are over a time period of 10 years, therefore}

random effects are expected.

(20)

these measures is based on the relevant literature7 and on data availability. The model in this study takes the following form:

VAi= α+ β1IOi+ β2SIZEi+ β3LEVi+ β4OWNi+ β5MANUFi+ e

VAi= 1 indicates if a Japanese listed firm voluntarily adopts IFRS and 0 means otherwise. Japan has four different reporting standards8, this study only indicates IFRS or otherwise.

IOi= Indicates the international orientation of a firm using the foreign sales-to-total-sales-ratio.

SIZEi9= Indicates firm size using the natural logarithm of market capitalisation

LEVi= Indicates the level of leverage within a firm using the long-term-debt-to-total-capital ratio

OWNi= Indicates ownership structure using the number of closely held shares

MANUFi= Is a dummy variable for the industry of the firm and takes the value 1 if SIC is below 4000, and 0 if otherwise.

4.3 Data collection

The information used to determine the variables is retrieved from Datastream. Studying the specific variables, described in the previous paragraph is based on data availability and existing relevant literature (e.g. Christensen et al, 2012; Cuijpers and Buijink, 2005; Dumontier and Raffournier,1998; Gassen and Sellhorn, 2006). For the tests in this study I use Japanese firms that are listed on the NIKKEI 22510. I choose this sample because the NIKKEI 225 is the most important stock index of the Tokyo Stock Exchange (TSE). The NIKKEI 225 contains (needless to say) 225 companies, which I analyse over the time period 2005-2014. From the 225 listed companies on the NIKKEI 225 there was no information at all for 2 companies, this led to 223 companies and 2230 observations (10 years). Unfortunately, not all data for the remaining 223 companies was available. Mainly information on

7_{In the relevant existing literature the variable AUDi was often included. The variable AUDi is a dummy variable with a value of 1 if a}

firm was audited by a Big 4 and 0 if otherwise. In the sample of this study all IFRS firms were audited by a Big 4 firm and from the non-IFRS firms over 95% were audited by a Big 4 firm. This led to omitted results and therefore I exclude this variable from the model.

8_{In paragraph 2.1 the accounting policy of Japan is described. The different reporting standards are: designated IFRS, JMIS, Japanese}

GAAP and US GAAP.

9_{In paragraph 5.4.2 I perform a sensitivity/additional test with another measure of size which be explained in that paragraph.} 10_{The IFRS firms and the control non-IFRS firms analysed, are listed in appendix 2}

(21)

international sales was missing and the year 2013 was scarce on data, which is clearly visible in table 1. I deleted the observations for which not all data was available. This led to 1584 observations from which 24 observations scored 1 on the dummy variable reporting standards (VAi). After deleting the missing data I searched for outliers and after deleting the outliers there remained 1532 observations of which 24 were IFRS adopters (1,6%). In table 1 the observations per year are illustrated with the reporting standards and the percentage of IFRS adopters, in table 2 all 223 firms are distributed per industry and in table 3 only the IFRS firms (2014) are distributed per industry.

TABLE 1: accounting standards Japanese listed firms (2005-2014)

Year Non-IFRS

(JMIS, Japanese GAAP and US GAAP)

IFRS Total Percentage IFRS

% 2005 156 0 156 0.00 2006 161 0 161 0.00 2007 161 0 161 0.00 2008 163 0 163 0.00 2009 164 0 164 0.00 2010 167 0 167 0.00 2011 167 1 168 0.60 2012 162 3 165 1.8 2013 38 3 41 7.32 2014 169 17 186 9.14 Total 1508 24 1532 1.57

(22)

TABLE 2 : Distribution of Nikkei 225 firms by industry (n=223)

SIC N %

Agriculture, Forestry & Fishing 0100-0999 1 0.45

Mining 1000-1499 3 1.35

Construction 1500-1799 9 4.04

Manufacturing 2000-3999 135 60.54

Transportation and Public Utilities 4000-4900 25 11.21

Wholesale trade 5000-5100 1 0.45

Retail trade 5200-5900 13 5.83

Finance, Insurance, Real Estate 6000-6999 24 10.8

Services 7000-8900 12 5.38

Total 223 100.0

Note: All observations are from 2014 the data is retrieved from Datastream.

TABLE 3: Distribution of Nikkei 225 IFRS adopting firms (2014) per industry (n=17)

SIC N %

Agriculture, Forestry & Fishing 0100-0999 0 0.00

Mining 1000-1499 0 0.00

Construction 1500-1799 0 0.00

Manufacturing 2000-3999 11 64.71

Transportation and Public Utilities 4000-4900 1 5.88

Wholesale trade 5000-5100 0 0.00

Retail trade 5200-5900 4 23.53

Finance, Insurance, Real Estate 6000-6999 0 0.00

Services 7000-8900 1 5.88

Total 17 100.0

(23)

5 Empirical results

In this section I present the results of the regressions. The regressions test the hypothesized relationship of firm’ characteristics and voluntary IFRS adoption in Japan. The first paragraph illustrates the descriptive statistics of the independent variables and the univariate results for both the IFRS adopters and non-IFRS adopters. In the second paragraph the results of the probit regression are presented. Furthermore, I perform robustness checks and additional tests which are described in the third and fourth paragraph.

5.1 Descriptive information and univariate results

In table 4 the descriptive statistics of the variables are illustrated. Panel A presents the descriptive statistics for non-IFRS firms and panel B presents the descriptive statistics for IFRS firms. Table 5 shows the results of a simple two-sample t-test. The t-test assesses the differences between the means of the two groups.

Table 4 illustrates that IFRS firms are more internationally orientated than non-IFRS firms. Almost half of the sales are foreign sales (48.2 %) in contrast to around a third (32.4 %) for non-IFRS adopters. Also, the statistics show that IFRS firms (31.1%) on average have a higher level of leverage than IFRS firms (25.9 %). Furthermore regarding size, IFRS firms seem to be bigger than non-IFRS firms. Results on ownership diffusion show a difference of 5% between non-IFRS firms (16.0%) and for non-IFRS firms (21.0%). Finally, table 4 shows the results about industry type, there is just a little difference between IFRS firms and non-IFRS firms (1%). The distribution between the industries of IFRS firms and non-IFRS firms is almost identical, which is also illustrated in table 2 and table 3.

The two sample t-test, illustrated in table 5 assesses the differences in means between IFRS firms and non-IFRS firms and also shows a level of significance (1%***, 5%**, 10%*). The two-sample t-test shows for IOi a highly significant difference in means between IFRS firms and non IFRS firms (1%). Regarding the second explanatory variable SIZEi, there is also a highly significant difference in means between IFRS firms and non-IFRS firms (1%). In contrast to the first two variables, the results on the difference between the means of the third variable LEVi are not significant at any significance level (>10 %). The two control variables OWNi and MANUFi neither show a significant difference between the means of IFRS firms and non-IFRS adopters.

(24)

TABLE 4 : Descriptive statistics for Japanese listed firms-- IFRS versus non-IFRS group

Panel A: Descriptive Variables for non-IFRS firms (n=1,508)

Variable Mean Std. Dev. Min Max

IOI 0.324 0.238 0 1 SIZEi 8.750 0.499 7.422 10.383 LEVi 0.259 0.164 0 0.732 OWNi 0.210 0.149 0.000 0.897 MANUFi 0.740 0.439 0 1

Panel B: Descriptive Variables for IFRS firms (n=24)

Variable Mean Std. Dev. Min Max

IOI 0.482 0.188 0 0.759 SIZEi 9.083 0.546 7.980 9.967 LEVi 0.311 0.208 0 0.555 OWNi 0.160 0.214 0.004 0.628 0.6275 MANUFi 0.750 0.442 0 1

All observations are over the time period 2005-2014. IOi is the percentage of the foreign sales (WC07101) divided by the total sales (WC01001); LEVi is the total debt (WC03255) divided by the total assets (WC8001); SIZEi is the natural logarithm of market capitalisation of equity LOG(WC08001); OWNi is the percentage of closely held shares (WC08021); MANUFi is a dummy variable with 1 if the firm belongs to a manufacturing firm (SIC is below 4000) and 0 if otherwise. VAi equals 1 if a firm reports under IFRS and 0 if otherwise (WC07536).The codes between the brackets are the codes retrieved from Datastream.

(25)

TABLE 5 : Univariate results for Japanese listed firms—IFRS group versus non-IFRS group Charasteristics Variable t-value Sig. (2-tailed) International orientation IOi -3.224 0.001***

Size SIZEi -3.248 0.001***

Leverage LEVi -1.523 0.128

Ownership diffusion OWNi 1.607 0.108

Industry MANUFi -0.114 0.910

All observations are over the time period 2005-2014. IOi is the percentage of the foreign sales (WC07101) divided by the total sales (WC01001); LEVi is the total debt (WC03255) divided by the total assets (WC8001); SIZEi is the natural logarithm of market capitalisation of equity LOG(WC08001); OWNi is the percentage of closely held shares (WC08021); MANUFi is a dummy variable with 1 if the firm belongs to a manufacturing firm (SIC is below 4000) and 0 if otherwise. VAi equals 1 if a firm reports under IFRS and 0 if otherwise (WC07536).The codes between the brackets are the codes retrieved from Datastream.***, **, * significant differences between Japanese listed IFRS firms and non-IFRS firms (two-sided) at significance level of 0,01, 0,05, 0,1

5.2 Probit regression

To test the hypothesized relationships of firm characteristics and voluntary IFRS adoption in Japan, I use a probit regression. Before I go to the results of the probit regression model, I first examine the correlations11 between all variables. In table 6 the correlation matrix is illustrated.

The correlation matrix presents that the three independent variables: IOi, LEVi and SIZEi are all positively correlated with the choice for reporting standards, however the correlation is quite low. A positive correlation means that the variables influence the choice for IFRS adoption. The correlation of the control variables with the dependent variable are extremely low. MANUFi is as well positively correlated with the dependent variable but OWNi is not. The correlation between the explanatory variables are moderate, this means that the variables contain different information about the likelihood of voluntary IFRS adoption. The strongest correlation is found between IOi and MANUFi (0.427). This means that the two variables contain similar information about voluntary IFRS adoption and manufacturing firms are probably more internationally orientated than non-manufacturing firms. Furthermore, I expected that SIZEi and IOi would be highly correlated because larger firms tend to be

11_{High correlation between independent variables (determinants) can cause problems of multicollinearity . In paragraph 5.3.2 I performed a}

(26)

more internationally orientated (André et al, 2012). In contrast to my expectations, the correlation matrix shows that SIZEi and IOi share only for 17.4% information.

TABLE 6: Pearson Correlation matrix of all variables

VAi IOi SIZEi LEVi OWNi MANUFi

VAI 1.000 IOi 0.081*** 1.000 SIZEi 0.081*** 0.174*** 1.000 LEVi 0.038 -0.123*** -0.221*** 1.000 OWNi -0.040 -0.064** 0.064** -0.174*** 1.000 MANUFi 0.000 0.427*** -0.174*** -0.073*** -0.080*** 1.000

All observations are over the time period 2005-2014. IOi is the percentage of the foreign sales (WC07101) divided by the total sales (WC01001); LEVi is the total debt (WC03255) divided by the total assets (WC8001); SIZEi is the natural logarithm of market capitalisation of equity LOG(WC08001); OWNi is the percentage of closely held shares (WC08021); MANUFi is a dummy variable with 1 if the firm belongs to a manufacturing firm (SIC is below 4000) and 0 if otherwise. VAi equals 1 if a firm reports under IFRS and 0 if otherwise (WC07536).The codes between the brackets are the codes retrieved from Datastream. ***, **, * significant levels between variables of 0,01, 0,05, 0,1.

Table 7 reveals the results of the probit regression. I use the results from the probit regression to conclude whether there is a relationship between firm characteristics and voluntary IFRS adoption in Japan. The three independent variables all have a positive coefficient, this means that: international orientation, leverage and size have a positive relation with voluntary IFRS adoption in Japan. Furthermore the table shows a z-statistic and p-value per variable. The p-value indicates whether results are significant, in the model I include three levels of significance (1%*, 5%**, 10%***). I base my conclusions on a significance level of 5%. A p-value smaller than 0.05 means the hypothesis is supported, conversely a p-value larger than 0.05 means the hypothesis is rejected.

The results illustrated in table 7 show that IOi is positively and significantly associated with the propensity to voluntarily adopt IFRS in Japan. The positive relationship is shown with the positive coefficient (1.038) and the significance is shown with a p-value of 0.01. The findings confirm my expectations. Based on the findings the first hypothesis about the positive relationship between

international orientation and IFRS adoption is supported. This also means the results are consistent

(27)

on voluntary IFRS adoption in other countries12 (André et al, 2012; Christensen et al, 2012; Cuijpers and Buijink, 2005; Dumontier and Raffournier,1998; Gassen and Sellhorn, 2006).

Table 7 also presents that SIZEi is positively and significantly associated with the propensity of voluntary IFRS adoption in Japan. The positive relationship is shown with a positive coefficient (0.508) and the variable is again highly significant with a p-value of 0.004. With a significance level of 0.01 the hypothesis about the positive relationship of size and IFRS adoption is supported. Again the results are in line with the findings in the prior literature about voluntary IFRS adoption (Christensen et al, 2012; Cuijpers and Buijink, 2005; Dumontier and Raffournier,1998; Gassen and Sellhorn, 2006).

The third variable LEVi is not positively significantly associated with the propensity to voluntarily adopt IFRS in Japan. Table 7 reveals a p-value of 0.076 this means the hypothesis can only be supported with a significance level of 0.10. In this study I use (like mentioned) a significance level of 0.05 and therefore the hypothesis about the positive relationship between leverage and IFRS adoption is rejected. I expected a positive relationship between leverage and IFRS adoption based on agency theory. At more levered firms the demand for monitoring the relationship between the shareholders and the creditors is higher (Dumontier and Raffournier, 1998). However in contrast with the agency theory, some prior literature on voluntary IFRS adoption find no significant relationship for leverage as a driver of IFRS adoption (Christensen et al, 2012; Cuijpers and Buijink, 2005; Gassen and Sellhorn, 2006). There can be assumed that the agency theory is possibly not suitable to rely on for this variable. An explanation that IFRS adoption is not associated with leverage can be that creditors do not think IFRS provides higher quality information than for example Japanese GAAP. Like mentioned in paragraph 2.1, IFRS and Japanese GAAP can be found equivalent in quality. Another explanation might be that firms communicate to their creditors on a more private basis whereby private information instead of public information (financial statements) declines information asymmetries.

Regarding the control variables there are no significant results. My expectation for a positive relation between ownership diffusion and IFRS adoption was also based on agency theory. Again, there can be considered if agency theory is suitable for the sample. Looking to the pseudo R2 , the figure of 9.32% shows the explanatory power of the variables regarding IFRS adoption. The number also indicates that besides the variables tested in this study there are other variables/drivers that influence IFRS adoption in Japan.

12_{IFRS adoption became mandatory in 2005 in the EU, before 2005 firms were allowed to use IFRS. Existing literature examines the}

(28)

Summarizing the findings, hypothesis 1 and hypothesis 2 are supported. This means there can be inferred that international orientation and size are significantly and positively associated with IFRS adoption in Japan. Hypothesis 3 is rejected since the result on LEVi is not significant (p-value > 0.05). Also the control variables OWNi and MANUFi are not positively associated with IFRS adoption. Finally, because I base my expectations about leverage and ownership diffusion on agency theory and my results are not in line with my expectations, agency theory might not fit best in this study.

TABLE 7: Probit Regression Analysis for Japanese listed firms—IFRS versus non-IFRS

All observations are over the time period 2005-2014. IOi is the percentage of the foreign sales (WC07101) divided by the total sales (WC01001); LEVi is the total debt (WC03255) divided by the total assets (WC8001); SIZEi is the natural logarithm of market capitalisation of equity LOG(WC08001); OWNi is the percentage of closely held shares (WC08021); MANUFi is a dummy variable with 1 if the firm belongs to a manufacturing firm (SIC is below 4000) and 0 if otherwise. VAi equals 1 if a firm reports under IFRS and 0 if otherwise (WC07536).The codes between the brackets are the codes retrieved from Datastream. ***, **, * significant differences between Japanese listed IFRS firms and non-IFRS firms at significance level of 0,01, 0,05, 0,1.

Observations: 1532

Pseudo R2: 0,0932

5.3 Robustness checks

To check the robustness of the model I test the following assumptions: heteroscedasticity, multicollinearity and independence.

Variable Predicted sign Coefficient z-statistic p-value

IOi + 1.038 2.59 0.010*** SIZEi + 0.508 2.89 0.004*** LEVi + 0.942 1.77 0.076* OWNi - -0.068 -1.15 0.249 MANUFi ? -0.107 -0.57 0.640 Intercept ? -7.139 -4.36 0.000

(29)

5.3.1 Heteroscedasticity

Heteroscedasticity is a problem which occurs in statistical models. Heteroscedasticity describes the behaviour of the variances and the standards deviations of a sample. Heteroscedasticity exists when the variances and standard deviations of a variable measured over time are non-constant. In this section I correct the probit model for heteroscedasticity and analyse the outcomes. I check whether the findings are similar (H1 and H2 are supported and H3 is rejected) and if the probit model does not violate the assumption.

TABLE 8: Probit Regression Analysis for Japanese listed firms- assumption homoscedasticity

All observations are over the time period 2005-2014. IOi is the percentage of the foreign sales (WC07101) divided by the total sales (WC01001); LEVi is the total debt (WC03255) divided by the total assets (WC8001); SIZEi is the natural logarithm of market capitalisation of equity LOG(WC08001); OWNi is the percentage of closely held shares (WC08021); MANUFi is a dummy variable with 1 if the firm belongs to a manufacturing firm (SIC is below 4000) and 0 if otherwise. VAi equals 1 if a firm reports under IFRS and 0 if otherwise (WC07536).The codes between the brackets are the codes retrieved from Datastream.***, **, * significant differences between Japanese listed IFRS firms and non-IFRS firms at significance level of 0,01, 0,05, 0,1

Wald chi2= 25.84 | Prob>chi2 = 0.0001

Comparing table 8 with table 7 we see some differences in z-statistics and p-values. The hypotheses about international orientation (0.005) and size (0.007) are still supported and the hypothesis about leverage rejected. The only thing that immediately catches the eye is the change in p-value of the variable LEVi. In the first probit model the hypothesis could be supported with a significance level of 10% (0.076), after the correction for heteroscedasticity this is not possible anymore (0.123). Furthermore, the test shows a Wald chi –square statistic of 25.84 and a p-value of 0.0001. The Wald chi-square statistic is used to test that at least one predictors’ regression coefficient is not equal to Variable Predicted sign Coefficient z-statistic p-value

IOi + 1.037533 2.82 0.005*** SIZEi + 0.5075093 2.72 0.007*** LEVi + 0.9423203 1.54 0.123 OWNi - -0.06837091 -0.82 0.412 MANUFi ? -0.1074014 -0.44 0.662 Intercept ? -7.138922 -4.00 0.000

(30)

zero. The null hypothesis set is that all regression coefficients across the model are equal to zero. With the p-value of 0.0001 the hypothesis is rejected. It indicates that the model is significant as a whole.

5.3.2 Multicollinearity

Multicollinearity occurs when two or more independent variables are highly correlated within the model. In table 6 an overview of all correlations and their significance level are illustrated. To test whether variables are correlated at an unacceptable level and imply multicollinearity I use the tolerance levels and the VIF values, which are presented in table 9. The tolerance value indicates how much of the variability of an independent variable is not explained by another independent variable in the model. The VIF value is the inverse of the tolerance value. A tolerance value of less than 0.10 - 0.20 and/or a VIF value above 5-10 indicates a multicollinearity problem (Hair, Anderson, Tatham and Black, 1998). This means there is no problem with multicollinearity in the model.

TABLE 9: Collinearity statistics for multicollinearity

Variable VIF Tolerance level

IOi 1.34 0.744 SIZEi 1.22 0.822 LEVi 1.14 0.880 MANUF 1.34 0.746 OWNi 1.05 0.955 Mean 1.19

All observations are over the time period 2005-2014. IOi is the percentage of the foreign sales (WC07101) divided by the total sales (WC01001); LEVi is the total debt (WC03255) divided by the total assets (WC8001); SIZEi is the natural logarithm of market capitalisation of equity LOG(WC08001); OWNi is the percentage of closely held shares (WC08021); MANUFi is a dummy variable with 1 if the firm belongs to a manufacturing firm (SIC is below 4000) and 0 if otherwise. VAi equals 1 if a firm reports under IFRS and 0 if otherwise (WC07536).The codes between the brackets are the codes retrieved from Datastream

5.3.3 Independence

The last assumption I test is the assumption about independence. The data should not be connected in a way that I haven’t accounted for in the model already. Because I have a moderate size data set (1532 observations from 223 firms), there can be significant correlations between the observations, either over time or across firms. To correct cross-observation correlation I cluster standard errors, which are robust standards errors and need to be corrected for unwanted correlation. By comparing this table

(31)

with table 7 we see some differences in z-statistics and p-values. Still the hypotheses about international orientation and size are supported and the hypothesis about leverage rejected with a p-value larger than 0.05. Thus, the model does not violate the assumption of independence.

TABLE 10: Probit Regression Analysis for Japanese listed firms- assumption independence

All observations are over the time period 2005-2014. IOi is the percentage of the foreign sales (WC07101) divided by the total sales (WC01001); LEVi is the total debt (WC03255) divided by the total assets (WC8001); SIZEi is the natural logarithm of market capitalisation of equity LOG(WC08001); OWNi is the percentage of closely held shares (WC08021); MANUFi is a dummy variable with 1 if the firm belongs to a manufacturing firm (SIC is below 4000) and 0 if otherwise. VAi equals 1 if a firm reports under IFRS and 0 if otherwise (WC07536).The codes between the brackets are the codes retrieved from Datastream. ***, **, * significant differences between Japanese listed IFRS firms and non-IFRS firms at significance level of 0,01, 0,05, 0,1

5.4 Additional tests

To test the model I perform some sensitivity tests. First, I perform the analyses with a logit regression instead of a probit regression. Second, I perform the probit regression in a matched paired design. Finally, I take an alternative measure for size.

5.4.1 Probability versus Logistic

Existing literature uses a logistic (logit) regression or a probability (probit) regression, to test the hypothesized relationships of firm characteristics and voluntary IFRS adoption. (André et al 2012; Cuipers and Buijnk, 2005; Dumontier and Raffournier, 1998; Gassen and Sellhorn, 2006). Like described in paragraph 4.1 I choose to use the probit model because with the sample and the variables in this study the probit model fits best. Because in similar studies authors also use logit regression, I also analyse the outcomes of a logit regression and examine if the results are consistent with the probit regression.

Variable Predicted sign Coefficient z-statistic p-value

(32)

Comparing the results of the logit regression with the results of the probit regression the numbers are not completely identical. However, the conclusions we can draw from the results remain the same. Still, the hypothesis about international orientation and size are supported and the third hypothesis about leverage rejected. Furthermore, the logit regression has an R-square of 0.1022, so that the explanatory power of the probit model is lower (0.0932). The homogeneous results of the both models lead to a more robust conclusion.

TABLE 11: Logit Regression Analysis for Japanese listed firms—IFRS versus non-IFRS

All observations are over the time period 2005-2014. IOi is the percentage of the foreign sales (WC07101) divided by the total sales (WC01001); LEVi is the total debt (WC03255) divided by the total assets (WC8001); SIZEi is the natural logarithm of market capitalisation of equity LOG(WC08001); OWNi is the percentage of closely held shares (WC08021); MANUFi is a dummy variable with 1 if the firm belongs to a manufacturing firm (SIC is below 4000) and 0 if otherwise. VAi equals 1 if a firm reports under IFRS and 0 if otherwise (WC07536).The codes between the brackets are the codes retrieved from Datastream. ***, **, * significant differences between Japanese listed IFRS firms and non-IFRS firms at significance level of 0,01, 0,05, 0,1

Observations: 1532

Pseudo R2: 0,1022

5.4.2 Matched-paired design

In the main analysis I use a sample with 1532 observations from which 24 observations are of firms that adopt IFRS (1.57 %). Although the percentage of firms that adopt IFRS is very small there are some interesting significant results. The first two hypotheses about international orientation and size are supported. The third hypothesis about the relation between IFRS adoption and leverage is not significant and therefore rejected.

In addition to the analyses in the previous paragraphs, I carry-out the same probit regression in a matched-pair design (like Gassen and Sellhorn (2006)). The IFRS firms are matched to a control Variable Predicted sign Coefficient z-statistic p-value

(33)

non-IFRS firm of the same year13. The sample I use is not from 2005-2014 but from 2011-2014, with 24 IFRS adopters and 24 control firms (table 12). However, this means that in the new sample the non-IFRS firms (control group) are not representative for all Japanese non-IFRS firms anymore and therefore it is important to note that I can only draw relative and not absolute inferences from the analysis. I match the firms by the number of employees and the ROE.

The results in the matched-paired design are somewhat different. First, the correlation matrix (of the new sample) shows that all three independent variables are positively correlated with the choice for reporting standards with a much higher number than the main sample (illustrated in table 6). This means the variables in the new sample contain more information about the choice of reporting standards (international orientation: 0.3209, size: 0.099, leverage: 0.1994). Furthermore, after performing the probit regression the results for the variables international orientation and leverage are the same (p-values; international orientation 0.042**, and leverage 0.085*). However, in contrast to the main results, the variable size does not show a positive relation. In the main sample there was a positive significant relation with a significance level of 1% now there is no significant relation at all (p-value: 0.187).

The difference in the p-values of the variable size between the main sample and the matched-paired sample can be explained by the way I match the IFRS firms to the control firms. The IFRS firms are matched to a control firm by the number of employees. However, number of employees is also used as a measure of size in studies about private firms (e.g. André et al, 2012; Kumar, Rajan, & Zingales, 1999). This explains why the results of the matched paired design indicate that size play a minor role in the choice for reporting standards. In table 14 a summary of the hypothesis and results of the main analysis and additional analysis is illustrated.

TABLE 12: Accounting standards followed scored matched-sample

Year IFRS non-IFRS Total

2011 1 1 2

2012 3 3 6

2013 3 3 6

2014 17 17 34

Total 24 24 48

Note: All observations are over the time period 2011-2014 the data is retrieved from Datastream