The Effect of IFRS on Corporate Risk Taking / /

(1)

The Effect of IFRS on

Corporate Risk Taking

Master Thesis

MSc International Financial Management Christoph Schauer

S2786893

Supervisor: Dr Halit Gonenc Assessor: Dr Hein Vrolijk Date of handing in: 19/06/2015

Word count (excl. abstract, references, and tables): 14,167

/

university of

(2)

II

Abstract

This paper examines the effect of the mandatory adoption of IFRS on corporate risk taking. To this end, difference-in-difference analyses of the effect of IFRS on risk taking are performed, making use of an extensive panel dataset with firm-level data of more than 32,500 firms from 47 countries. The empirical results are in line with the majority of the theoretical arguments developed in this paper, which draw from information asymmetry, portfolio, and agency theories. Overall, a positive effect of the adoption IFRS on corporate risk taking is estimated. This result is robust over a range of different sample periods and model specifications, including a propensity score matching model.

(3)

III

1. Introduction

The International Financial Reporting Standards (IFRS) were designed by the International Accounting Standards Committee (IASC), and its successor body, the International Accounting Standards Board (IASB), as "a single set of high quality, understandable, and enforceable global accounting standards that lead to the reporting of transparent and comparable information in general purpose financial statements. Therefore, IFRS are presumed to (1) increase the transparency and comparability of financial reports, (2) be of higher quality than domestic GAAP, and (3) reduce information asymmetry between firm insiders and outsiders" (Drake et al. 2010, p.7).

The first country to adopt IFRS as mandatory accounting standard was Singapore in 2003, requiring all listed an non-listed companies incorporated in Singapore to report both its consolidated and separate financial statements under IFRS. Singapore was followed in 2005 by Australia, Hong Kong, Norway, the Philippines, South Africa and Switzerland, and all member states of the European Union. Within the EU, all listed firms are now required to report its consolidated financial statements under IFRS. Today, IFRS have become the dominant reporting standard worldwide, with approximately 90 nations and territories fully conforming with IFRS, that is, requiring all listed companies to report at least their consolidated statements under IFRS, according to the American Institute of Certified Public Accountants (AICPA).1

The widespread adoption of IFRS across the globe over the last decade constitutes a significant change for all companies that had to switch standards as well as for all financial analysts, investors, and any other consumers of information from public financial statements. It is therefore no surprise that this event spawned a wealth of research analyzing the impact of IFRS on a wide variety of variables. Prominent examples thereof include – to name only a number of studies relevant in the context of this paper – are Armstrong et al. 2010 for the effect of IFRS on stock prices, Barth et al. 2008 on earnings management, Beuselinck et al. 2009 on stock price synchronicity, Brochet et al. 2013 on abnormal returns to insider purchases, Chan et al. 2013 on credit ratings, Byard et al. 2011, Drake et al. 2011, Landsman et al. 2012, and Tan et al. 2011 on foreign analyst following and analysts' forecast accuracy, Covrig et al. 2007, Florou & Pope 2012, and Brüggemann et al. 2011 on foreign investments, Florou & Kosi 2008 on the cost of debt, Hail & Leuz 2007 and Daske et al. 2008 on the cost

1

(5)

2 of equity, Naranjo et al. 2014 on external financing, or Lang et al. 2010 on earnings co-movements across borders.

However, to the best knowledge of the author, no study has yet examined potential influence of the adoption of IFRS on corporate risk taking. This paper aims to fill this gap. To do so, first several theoretical arguments are established in order to connect IFRS with corporate risk taking. Then five hypotheses are derived from these arguments, and consequently tested by employing a difference-in-difference analysis, using firm-level panel data of more than 32,500 firms from 47 countries.

This paper proceeds as follows: Chapter 2 develops the theoretical arguments and provides an overview of the relevant literature in the process. Chapter 3 explains the methodology for the empirical analysis. Chapter 4 first introduces the data used for the analysis, then proceeds with presenting the main result, the secondary findings, and several robustness tests for the main result. Chapter 5 concludes and offers some suggestions for future research.

2. Literature Review and Hypothesis Development

2.1. IFRS, Information Asymmetries, and Risk Taking

The connection between IFRS and corporate risk taking in this section is developed in three steps: First, the influence of information asymmetries on the cost of capital is introduced. Second, the role of IFRS in reducing information asymmetries and, consequently, the cost of capital, is explained. To close the argument, the expected effect of a) the cost of capital and b) estimation risk on risk taking is established.

Information Asymmetries and the Cost of Capital

(6)

3 operations of the firm, and external investors - outside shareholders or creditors.2 Insiders have an information advantage over outsiders regarding the true value of the company's assets, its creditworthiness, the expected returns and the risks of its investments, its growth prospects, and so forth. Insiders can exploit this advantage to their own benefit by attempting to sell shares and bonds for more than their true worth. Investors, however, are aware of insiders' incentive to overprice these securities. Consequently, they will only buy stocks and bonds at a discount.

Information asymmetries turn the market for financing into a lemons' market à la Akerlof 1970: For non-opportunistic insiders, their information advantage turns into disadvantage with regards to accessing external financing for their firms. They are only able to obtain financing for their company at conditions that are less favorable than they would be in the absence of these information asymmetries. In other words, less transparent financial markets lead to a higher cost of debt and equity capital (Armstrong et al. 2011, Bhattacharya et al. 2012, Easley & O'Hara 2004, Hughes et al. 2007, Leuz & Verrecchia 2000 & 2005, Myers & Majluf 1984, Opler et al. 1999).

Even in the absence of any possibility of opportunistic overpricing, a poor information environment makes it more difficult for investors to assess the true value of shares and debt, thereby making them more risky in their perception. That is, the estimation risk increases with the degree of information asymmetry between corporate insiders and external investors, leading risk averse investors to demand a higher risk premium (Easley & O'Hara 2004, Hail & Leuz 2007, Hughes et al. 2007, Kumar et al. 2008, or Lambert et al. 2007).

A second consequence of superior information of insiders such as the management, vis-à-vis outside shareholders and creditors is that it enables the former to run the company in a way that may not be in the best interest of the latter - information asymmetries give rise to agency conflicts. This will be the topic of section 2.3.

2_{Creditors, in this context, is referring to holders of public debt, such as bonds, for who the information from}

(7)

4

IFRS and Information Asymmetries

Firms' financial statements are an important source of information for outside investors who are not closely involved with the day-to-day operations of a company. Proponents of the mandatory adoption of IFRS as a universal accounting standard argue that IFRS improves both the quality and the comparability of financial statements.

By agreeing on a single international accounting standard, IFRS is expected to increase the comparability of financial statements across borders. If financial statements of firms from different countries follow the same rules, if they are written in the same accounting language, analysts are better able to understand the statements in all their intricacies. Consequently, IFRS facilitates analyzing foreign financial statements. In this way, IFRS is expected to decrease the information asymmetry between corporate insiders and outside foreign investors.

IFRS is also generally presumed to provide more transparent, more extensive, generally higher-quality accounting information than many of the domestic standards it replaced. Domestic and foreign investors alike will benefit from better disclosure through IFRS. For the argument along these lines, see e.g. Armstrong et al. 2010, Ball 2006, DeFond et al. 2011, Doupnik & Perera 2011, Drake et al. 2010, Houqe et al. 2013, Horton et al. 2013, or Leuz & Verracchia 2000.

The possibly most direct way of empirically assessing whether the improvement in comparability of financial statements through IFRS reduces the information disadvantage for analysts is to look at the accuracy with which they predict future stock prices and their propensity to follow (foreign) stocks in the first place. Byard et al. 2011, De Franco et al. 2011, Drake et al. 2010, Hogdon et al. 2008, Houqe & Easton 2013, Horton et al. 2013, Jiao et al. 2012, Landsman et al. 2012, Preiato et al. 2009, and Tan et al. 2011 all examine the influence of the adoption of IFRS on either or both of these variables, and find a positive effect of IFRS here.

(8)

5 effect of IFRS on market liquidity, Florou & Kosi 2008 and Moscariello et al. 2014 for IFRS reducing the cost of debt, Chan et al. 2013 for IFRS improving credit ratings, and Armstrong et al. 2010 for the stock market reaction to adopting IFRS.

A number of studies finds improvements in the quality of disclosure to be the (or a) driver for the effect of IFRS on their respective variable of interest: Horton et al. 2013, Houqe et al. 2013, and Jiao et al. 2012 for analysts' stock and earnings forecast accuracy, Bischof 2009 for the quality of disclosure in the banking sector, Beneish et al. 2012 for foreign investments, and Yip & Young 2012 for a several accounting variables.

Further evidence for IFRS having a positive effect on variables closely associated with the information environment of outside investors, be it driven by improved quality or comparability, is provided by e.g. Landsman et al. 2012 for abnormal announcement return volatility and trading volume around earnings announcements, Beuselinck et al. 2009 and Santana et al. 2014 for stock price synchronicity, and Houqe et al. 2012 and Dimitropoulus et al. 2013 for earnings quality.

IFRS and the Cost of Capital

As IFRS has been demonstrated to reduce information asymmetries between corporate insiders and outside investors, and information asymmetries in turn are positively linked to the cost of capital, IFRS is expected to reduce firms' cost of capital. Many studies derive this hypothesis and find empirical evidence supporting it. Examples here include Daske et al. 2008 & 2013, Florou & Kosi 2008 & 2015, Hail & Leuz 2007, Houqe et al. 2013, Kim et al. 2011, Lee et al. 2008, Li 2010, Moscariello et al. 2014, Shi & Kim 2007, and Silva & Nardi 2014.

The Cost of Capital and Corporate Risk Taking

(9)

6 more costly for a firm if the cost of external financing is high. In the limit, if this cost is prohibitively high, recovering these losses is impossible, and the firm faces default. Consequently, the cost of capital should have a negative influence on corporate risk taking: The cheaper access to external financing is, the cheaper it is for a firm to recover potential losses resulting from taking risks, and thus the more risk it will be willing to take. As the adoption of IFRS has been shown to reduce the cost of capital, IFRS is therefore expected to increase corporate risk taking.

 Hypothesis 1a: IFRS increases corporate risk taking

This argument borrows from the literature on corporate cash holdings: The precautionary motive – the ability to feather unforeseen losses – is one of the principal motives to hold cash, according to Keynes 1936, and the cost of capital is one of the main determinants of corporate cash holdings (Opler et al. 1999). If the cost of capital decreases, firms will either reduce their cash holdings, take more risks, or do a combination thereof, to arrive at the same level of risk as before.

Unfortunately, there is little research focusing on the effect of the cost of external financing on corporate risk taking. This may be because such research is likely be fraught with severe endogeneity issues. The cost of capital will affect corporate risk taking; however, the other way round, risk taking will also affects the cost of capital: Taking higher risks increases the default risk, which will reflect in higher risk premiums. Still, Bruno & Shin 2014 find that the liquidity of the global financial system, which can be interpreted as the availability and cost of financing, is positively associated with corporate risk taking.

IFRS, Estimation Risk, and Corporate Risk Taking

(10)

7 Consequently, if an investor wants to keep the aggregate risk in her portfolio – as she perceives it – constant, then she will want to substitute this "loss" in estimation risk with higher "real" risk. In this way, investors will induce (or allow) firms to take greater risks. The hypothesis derived from this argument is the same as in the preceding section:

2.2. IFRS, Shareholder Diversification, and Risk Taking

Portfolio Theory

The essence of portfolio theory, originally developed by Markowitz 1952, is that risk averse investors demand a premium over the return of a risk-free rate for holding risky assets (for example as expressed mathematically in the CAPM). Investors can diversify away this risk to some extent by holding a portfolio with assets with volatilities of returns that are not correlated, although some undiversifiable risk usually remains. Consequently, changes that facilitate diversification reduce the risk premium an investor demands in order to be willing to hold a stock in her portfolio.

Investors' Home Bias

(11)

8

IFRS and Portfolio Diversification

One reason why it is more difficult for investors to evaluate foreign companies is that these firms oftentimes report in a different accounting standard. IFRS reduces these differences in accounting standards to zero between all the countries that adopt it, as already approximately 90 nations and territories did until 2015 (although how an accounting standard is applied and enforced matters too; see section 2.4.). Thus, the adoption of IFRS leads to a reduction in the costs of acquiring and processing information about foreign companies for investors for a large set of countries – evidenced, for instance, by IFRS increasing foreign analyst following and the accuracy of their evaluations, as shown in chapter 2.1. In this way, IFRS should increase investors' willingness to undertake foreign investments. A vast literature offers empirical support for this prediction: Amiram 2010, Brüggemann et al. 2011, DeFond et al. 2011, Florou & Pope 2012, Hamberg et al. 2013, Landsman et al. 2012, and Shima & Gordon 2011 all find that IFRS increases foreign equity investments. Covrig et al. 2007 arrive at the same conclusion when examining voluntary adopters of IFRS, and find that the effect is larger in poor information environments, i.e. where the information asymmetry reduction through IFRS is greater. Kim et al. 2011 and Beneish et al. 2012 estimate the same positive effect on foreign debt holdings.

The potential benefits in terms of portfolio diversification by increasing the share of foreign assets are substantial. As assets from different countries tend to be subject to shocks that are (at least to some degree) specific to their countries, the risks of these securities are less correlated with the risks of the domestic stocks already in the portfolio (see e.g. Stulz 1999 for a more detailed argument).

Portfolio Diversification and Risk Taking

(12)

9 Paligorova 2010 support this hypothesis directly; Barry et al. 2011 and Garcia-Kuhnert et al. 2015 find that institutional investors – which are typically large and well diversified – impose riskier strategies on banks. In addition, a positive effect of IFRS on foreign investments, which were demonstrated above to increase after its adoption, has been found by An et al. 2014, Boubakri et al. 2013, Koernioadi 2012, and Nguyen 2012.

The argument is different, but the expected effect of IFRS is the same as in section 2.1: By facilitating international portfolio diversification, IFRS will lead now better diversified investors to induce firms to take greater risks.

Qualification: IFRS, Integration of Financial Markets, and Systemic Risk

A stronger integration, a larger connectedness of financial markets means that they become more homogeneous as well. In more integrated markets, shocks that may otherwise hit markets in only one or few countries are transmitted more easily to other countries. That is, earnings, returns, or risks of firms across the globe are more correlated with each other. From an investor's point of view, the risks of individual assets are more strongly correlated, and there is greater portion of undiversifiable risk in a global portfolio (while not being direct sources, this argument is inspired by Bruno & Shin 2014, Lang et al. 2010, and Stiglitz 2010). Bruno & Shin 2014 find that common global financial conditions induce greater synchronization of risk-taking across regions.

IFRS, harmonizing accounting standards and increasing investment flows across many borders, may be a step into this direction. Lang et al. 2010 support this argument: They find that the introduction of IFRS increased earning co-movements across countries. That is, IFRS seems to have increased the impact of common shocks on firms (or at least their earnings). If true, then IFRS increasing the integration of international financial markets would counteract its positive effect on risk taking through facilitating portfolio diversification. Within the framework of portfolio theory, depending on whether the positive or negative effect prevails on balance, IFRS could instead increase aggregate portfolio risk and thus, in turn, have a negative effect on corporate risk taking.

(13)

10

2.3. IFRS, Agency

Conflicts

, and Risk Taking

Agency Conflicts and Corporate Governance

Agency theory dates back to Jensen & Meckling 1976. Two agency conflicts within a firm are of interest in the context of this paper. The first conflict is between the management of a firm and its shareholders: Managers are tasked by shareholders, the owners of the firm, to maximize the value of the firm, but managers are self-interested human beings too, and their personal interests are not necessarily aligned with those of the owners of the firm: Maximizing shareholder value. The second conflict is between shareholders and debtholders: Shareholders want to maximize the value of the firm, while debtholders want to maximize the likelihood that their debt is serviced.

IFRS & Agency Conflicts between Shareholders and Managements

An information disadvantage of shareholders gives rise to agency conflicts in the first place. If shareholders were perfectly informed, there would be little scope for managers to act against their interests, as shareholders would notice deviations immediately and be able to punish managers accordingly. Thus, the better shareholders are informed about the activities of the management, the less room there is for agency conflicts to manifest.

Financial statements are an important source of information here for outside shareholders who are not closely involved with the operations of the company. More extensive and transparent financial statements help shareholders to inform themselves about the actions of the management (Hail & Leuz 2006, La Porta et al. 2006). As outlined in section 2.1., the by now widespread adoption of IFRS is widely regarded as such a step towards reducing these information disadvantages for corporate outsiders, with a wealth of empirical evidence buttressing this notion. IFRS should therefore reduce the cost of monitoring the management for shareholders, thereby helping shareholders to impose their preferences better on the management. To put it differently, IFRS is expected to reduce agency conflicts and improve corporate governance (e.g. Hail et al. 2014).

(14)

11 purchases decreased after the adoption of IFRS in the UK, which can (also) be interpreted as IFRS curbing expropriatory insider trading at the expense of outside investors. Hail et al. 2014 find that IFRS reduced dividend payments, which they interpret as IFRS improving corporate governance: Where shareholders can trust managers more to re-invest profits in a way that maximizes shareholder value, they insistent less on profits being paid out immediately as dividends.

The general assumption is that managers are more risk averse than shareholders when it comes to deciding upon corporate investment strategy. They want to take fewer risks than would be optimal for maximizing shareholder value. This notion traces back to Jensen & Meckling 1976, Fama 1980, Amihud & Lev 1981, or Hrishleifer & Thakor 1992, with some more recent evidence provided by e.g. John et al. 2008 and Pathan 2009. One reason for this tendency is that a large part of managers' personal income and wealth is tied to the company they work for, be it in the form of stock options, expected future salaries, reputation, and future career opportunities). In portfolio theory terms, their personal portfolios are not well diversified. Thus, managers seek to limit the risk of their firms by pursuing overly cautious investment strategies.

Agency Conflicts between Shareholders and Managements and Risk Taking

The ability of shareholders to limit managers' inclination towards lower risk taking, i.e. their ability to impose their preferences on managers, depends on their power relative the management. As outlined above, better corporate governance empowers shareholders here, and corporate governance is positively associated with corporate risk taking. John et al. 2008 in their seminal study on this topic provide more in-depth arguments here, as well as empirical evidence in support. Laeven & Levine 2009 find a positive effect of shareholder power and Paligorova 2010 for the legal protection of shareholder rights on corporate risk taking as well, to name just two more examples.

In summary, by facilitating monitoring and control, adopting a more extensive, more transparent accounting standard such as IFRS helps shareholders to impose their preferences for risk on managements, thus increasing corporate risk taking.

(15)

12 As introduced above, the general assumption in the literature is that managers take less risks than shareholders want. However, this does not preclude the possibility that managers take excessive risks in certain cases. Overconfidence in their ability to assess risks, to forward an argument from the behavioral economics, would be an example here, or the desire to achieve overly ambitious performance targets, but there may be many other motivations as well. Where managers take excessive risks, IFRS empowering shareholders should lead to the opposite result: Shareholders will use their greater power to impose lower risk taking on managements. The effect of IFRS on corporate risk taking is therefore expected to be dependent on the level of risk taking in a firm prior to its adoption of IFRS. In firms where pre-IFRS risk taking is high, the positive (negative) effect of IFRS will be weaker (stronger) than in firms where it is low. The effect of the interaction between IFRS and pre-IFRS risk taking on post-IFRS risk taking will be negative.

 Hypothesis 2: The interaction between IFRS and pre-IFRS risk taking on risk taking

is negative

Other than through reducing the cost of monitoring directly, there are two additional corporate-governance-related ways in which IFRS may influence corporate risk taking. First, uniform accounting standards facilitate firm evaluations across borders and thus takeovers by foreign investors. The implicit or explicit threat to sell is a disciplinary tool for shareholders, as a change of ownership often entails a replacement of the management (Stulz 1999). To put it differently, IFRS should reduce the entrenchment of managers, and less entrenched managers tend to take more risks (Chintrakarn et al. 2013). Second, IFRS opens companies up not only to foreign investors who want to take them over, but also to large non-majority shareholders. Foreign large shareholders will be more likely to be outside investors, without vested interests in the firm (Stulz, 1999), and are oftentimes institutional investors without controlling interests. The presence of multiple large shareholders has been shown to have a positive influence on corporate risk taking (e.g. Boubaker et al. 2012, Koerniadi et al. 2014, and Paligorova 2010).

IFRS, Debtholders, and Risk Taking

(16)

13 informed decisions about whether and to which conditions to buy bonds or other debt of a firm. They will also be able to better monitor the extent to which a firm complies with whatever promises it made when issuing the debt - for example about the riskiness of its investments, which has a direct influence on the likelihood of a firm being able to service its debt.

Debtholders generally prefer firms to take less risks than shareholders (e.g. Ross et al. 2011). The reason is that debt (typically) has fixed returns, and thus debtholders do not profit from returns on investments that are higher than what is necessary in order to service the debt. That is, above this threshold, they do not benefit from the higher expected returns that are associated with undertaking riskier investments, while still bearing (some of) the risk if the investments turn sour. Creditor rights, which can be regarded as proxy for debtholder power, have been found to have a negative influence on corporate risk taking (e.g. Acharya et al. 2011, Claessens et al. 2000, King & Wen 2011, and Paligorova 2010; see also below for more detail).

Thus, as the adoption of IFRS is not only expected to empower not shareholders, but also debtholders, IFRS may also have a negative effect on corporate risk taking:

 Hypothesis 1b: IFRS decreases corporate risk taking

(17)

14  Hypothesis 3: The effect of the interaction between IFRS and creditor protection on

risk taking is negative

2.4. Accounting Standard Differences & Strength of Enforcement

The purpose of this section is to introduce two aspects that are connected to all three strands of theory previously introduced: Namely, that IFRS is expected to have stronger effects where a) IFRS constitutes a greater change in accounting systems and b) where accounting rules are better enforced.

Accounting Standard Differences

Any effects of IFRS resulting from an increase in the comparability or quality of financial statements will depend on how different from IFRS the preceding national accounting standards were in a country. Countries with preceding national accounting standards that are very different from IFRS experience a greater increase in comparability of financial reports than countries with very similar standards. At the same time, if IFRS is the best practice in financial reporting regulation, removing any differences to IFRS in an accounting standard would constitute an increase in quality (if such a gold standard exists in the first place). Either way, if the accounting standard in place before the adoption of IFRS was very similar to IFRs, then sweeping changes should not be expected – the greater a reduction of accounting standard differences IFRS constitutes, the greater will be the effect. This argument, and the corresponding evidence to support it, can be found in a number of papers, such as Byard et al. 2011, Chan et al. 2013, Florou & Pope 2012, Naranjo et al. 2014, and Tan et al. 2009.

Apart from IFRS, for instance Ashbaugh & Pincus 2001, Bae et al. 2008, Bradshaw et al. 2009, and Guan et al. 2006 find that the accounting differences between national accounting standards, or between national standards and IFRS/IAS, are negatively correlated with foreign analyst following and their forecast accuracy. Yu & Wahid 2014 find the same negative relation between accounting distance and international portfolio holdings.

(18)

15 accounting standards were more different. This pattern should hold up regardless of whether the effect of IFRS is ultimately driven by a reduction of the cost of capital, estimation risk, or agency conflicts, or by facilitating international portfolio diversification.

 Hypothesis 4: The effect of IFRS on corporate risk taking is stronger where adopting

IFRS constitutes a larger change in accounting standards

Strength of Enforcement

The strength of the enforcement of accounting rules is an aspect of the overall quality of the regulatory and legal institutional environment of a country. The general argument here is that where the strength of enforcement of accounting rules is low, the usefulness of public financial statements as tool for investment decisions is limited. If new accounting rules are adopted in a country, but the enforcement of these rules is weak, then investors will not be very confident in the truthfulness of any new information disclosed through the new accounting rules either. Consequently, a substantive impact of a change in rules, or an entire set of rules such as IFRS, cannot be expected. The strength of enforcement of accounting rules in a country has been identified as an important factor for explaining the strength of the effect of IFRS on a variety of variables by a large number of studies: Examples here include Beneish et al. 2012, Beusenlick et al. 2009, Byard et al. 2011, Chen et al. 2013, Christensen et al. 2013, Daske et al. 2008, Florou & Pope 2012, Hogdon et al. 2008, Holthausen 2009, Landsman et al. 2012, Li 2010, Naranjo et al. 2014, and Preiato et al. 2009. The argument above is based on these studies as well.

This pattern should apply to the effect of IFRS on corporate risk taking as much as it does to the various variables examined in the above studies.

(19)

16

3. Research Design

3.1. Regression Model

This section introduces the empirical approach used to estimate the effect of the mandatory adoption of IFRS on corporate risk taking. The baseline regression model is a difference-in-differences (DiD) model on annual firm-level panel data. Several different specifications of this model, as well as two models with a propensity score matched sample (PSM) are then used to examine hypotheses 2 to 5, and to test the results for robustness. The baseline model is the following Equation 1:

denotes the standard deviation of the return on assets of firm in year over the past

5 years, a standard proxy for corporate risk taking. is a dummy variable, taking the

value of 1 if a firm reports under IFRS in year and 0 otherwise. are the

selection of firm- and country-level control variables. These variables are defined in section 3.1. and are firm- and year-fixed effects. Firm-fixed effects are the appropriate choice over industry and country fixed effects here, as is a popular combination in the risk taking literature as well (used e.g. by Acharya et al. 2011, An et al. 2014, Boubakri et al. 2013, and Faccio et al. 2011), for two reasons: First, there are likely to be many unobserved variables that influence corporate risk taking which differ between firms and are (approximately) time-invariant. Second, what is in the focus is the change of risk taking within a firm. If country- and industry-fixed effects are used instead, the explanatory power of all models estimated in this paper and the statistical significance of the included variables is lower (not shown). Year-fixed effects capture trends in risk taking over time that affect all firms equally. As will be shown in section 4.1., their inclusion is necessary due to the development of risk taking over time.

(20)

17 The error terms in all models are heteroskedasticity-consistent robust standard errors, which are clustered at the firm level in order to allow for the errors of a firm to be serially correlated over time, as is consistent with the majority of the literature on risk taking (e.g. An et al. 2014, Boubakri et al. 2013, Bruno & Shin 2014, Faccio et al. 2011, and Nguyen 2012). As the mandatory adoption of IFRS happened on a country level (at least after a fashion: firms' fiscal year dates matter for the exact date of IFRS adoption; see section 3.2. for details), error clustering on the country level would be an alternative as well. However, when errors are clustered at the country level instead, statistical significance in all regressions estimated in this paper tends to suffer (not shown).

While endogeneity can always be an issue in this kind of research, it is not likely to be a grave problem here. It is possible that the allocation of countries into the treatment and control group may not be random in terms of information asymmetry: Countries that stand more to gain in terms of better disclosure may be more likely to adopt IFRS. However, while information asymmetries are at the core of the argument, the variable that is actually examined in this paper is not asymmetric information, but corporate risk taking. It is hard to imagine that decision makers were motivated to introduce IFRS in a country because they hoped to influence risk taking of firms in their country in a certain way by doing so – not least because there has been no research on this topic yet.

Furthermore, while political decision makers may have been motivated by any number of reasons to introduce IFRS in their countries, firms did not have this choice: Where IFRS became mandatory, all firms that qualify had to adopt it. As the option to report under IFRS on a voluntary basis existed too in many countries before IFRS became mandatory, self-selection issues will be present within the group of voluntary adopters. These firms are therefore excluded from the sample (see section 4.1. for more detail here). For the identification strategy in the estimations in paper, the introduction of IFRS can therefore be treated as an exogenous shock.

(21)

18 neighbor in the other group based on the pre-IFRS values of its firm-level characteristics. For a detailed description of the PSM methodology, please refer to section 4.3.

3.2. Description of Variables

Dependent Variable: SROA

The dependent variable is the standard deviation of the annual returns over assets ( ) of firm in year over the last 5 years including year . , in turn, is defined as the (winsorized) earnings before interest and taxes (EBITDA) over total assets. The standard deviation of the return over assets over a certain period is the most widely used proxy for corporate risk taking in the literature. is calculated on a rolling basis. Following the recent empirical literature on corporate risk taking (An et al. 2014, Boubakri et al. 2013, Faccio et al. 2011, Paligorova 2010), the used to calculate are country-adjusted by subtracting the mean over all firms within each country and for each year from firms' individual . This is done to remove country-level influences, like a country's economic cycle, from , in order to arrive at a cleaner measure for risk taking as expression of corporate investment decisions (Boubakri et al. 2013, Faccio et al. 2011). SROA is winsorized at the top and bottom 1% of the distribution in order to mitigate potentially distorting influences of outliers. The data source of firms' EBITDA and total assets is Thomson Reuters Worldscope/Datastream.

Independent Variable of Interest: IFRS

(22)

19 date of the official IFRS adoption, and the firm thus only has to report under IFRS for its financial statements published 2006 and later. IFRS always took effect on January 1 in the respective adoption year in all countries. Consequently, all firms with fiscal years starting later than January 1 in the IFRS introduction year will be classified as having adopted IFRS one year later. As data on firms' fiscal years was available only until 2011 in the data set used, firms' fiscal year end dates were extrapolated from their 2011 dates to 2012 and 2013, the last two years in the sample. Countrywide IFRS adoption dates are obtained from Ramanna & Sletten 2014, and updated with dates from the official country profiles of the IFRS Foundation and the IASB where missing.3 The data on firms' fiscal years is from Thomson Reuters Worldscope/Datastream.

Firm-Level Control Variables

The selection of firm-level controls included in all estimations in this paper follows the control variables most frequently used in recent empirical work on corporate risk taking, such as Acharya et al. 2011, An et al. 2014, Barry et al. 2011, Boubakri et al. 2013, Bruno & Shin 2014, Faccio et al. 2011, John et al. 2008, Koernadi et al. 2013, and Paligorova 2010. Like , all firm level controls are winsorized at the 1% level, are for each a firm in year , and their data source is Thomson Reuters Worldscope/Datastream.

is the return over assets of firm, defined as earnings before interest and taxes

(EBITDA) over total assets.

is the size of a firm, measured as the natural logarithm of its total assets in 1000 US$. is the leverage, defined as total debt (common equity plus debt plus preferred stock ) over the book value of total assets.

is the market-to-book ratio, defined as market capitalization over common equity, of a

firm, a frequently used measure in the literature for a firm's growth potential. are a firm's capital expenditures over its total assets.

is the annual growth of a sales, defined as its sales(t)-sales(t-1))/sales(t-1).

3

(23)

20 is the mean of SROA over the past 3 years, by firm, exclusive year . For firms in IFRS adopting countries, this variable measures pre-IFRS risk taking in the adoption year. This variable is used to examine hypothesis 2.

is the mean of SROA over the following 3 years, by firm, exclusive year . For firms in IFRS adopting countries, this variable measures post-IFRS corporate risk taking. This variable is used as dependent variable in one PSM model.

is : The change in corporate risk taking. For firms in IFRS adopting countries, this variable measures the pre-post IFRS change in risk taking. This variable is used as dependent variable in one PSM model. Like , these 3 variables are calculated based on data from Thomson Reuters Worldscope/Datastream.

Country-Level Control Variables

is the annual growth of GDP in year in country , taken from World Bank's World Development Indicators database.

is the index for regulatory quality in a year in country . Regulatory Quality is one

of the World Bank's six Worldwide Governance Indicators, which are compiled by Kaufmann, Kraay & Mastruzzi. is not available for the years 1999 and 2001. For these years, the average value between of the preceding and following in year by country was calculated. Apart from quantifying the quality of the regulatory environment in a country in general, this index is used more specifically as proxy for the strength of enforcement of accounting rules in the literature on IFRS, for example by Christensen et al. 2013, or, as part of a composite index, by Preiato et al. 2009 and Beneish et al. 2012.

is the creditor protection index from Djankov et al. 2008, measuring the extent to which

creditors are legally protected in a country. Table 1 in the appendix reports the scores of each country.

(24)

21 is, the only time variance in results from the reduction of to 0 after IFRS is adopted in a country. This limits the usefulness of this index for panel data analysis. Moreover, its validity is only guaranteed for the years 1998-2004, the period which Bae et al. 2008 examined. There may be additional changes in accounting standards between 2004 and 2013, the last year of the sample used in this paper, which are not captured by this index. However, the scope for such unaccounted changes is limited, more and more countries in this sample were adopting IFRS with every year since 2005, and accounting rules usually do not change that much over time, and should therefore not be of concern.

4. Results

4.1. Sample and Descriptive Statistics

The data source for all firm-level variables is Thomson Reuters Worldscope/Datastream, with country-level variables originating from different sources, such as the World Bank. The time period covered by the sample 1998 to 2013. 1998 was chosen as start date as 1998 is five years before the first country adopted IFRS: Singapore in 2003. 2013 is the last year for which sufficient data was available. 1998 - 2013 is a longer time frame than many other studies on the effects of IFRS use. 2001-2008, for instance, is a typical period (e.g. in Hail et al. 2014 and Naranjo et al. 2014), and many other studies use similar (or shorter) periods (e.g. Chen et al. 2013, Christensen et al. 2013, DeFond et al. 2011, Jiao et al. 2012, Landsman et al. 2012, and Preiato et al. 2009). While using a short sample like 2001-2008 would increase consistency with the previous literature on IFRS, it also precludes examining the effect of IFRS in all countries from the analysis that adopted IFRS after 2008 – such as Argentina, Brazil, Canada, Chile, Mexico, or South Korea. It also results in a heavily EU-dominated sample, as all EU countries adopted IFRS in 2005 already. This paper therefore mostly uses a longer sample. However, for consistency with the literature, the main analysis is repeated in section 4.3. with the 2001-2008 sample period and other shorter periods as well.4

Altogether, 32,659 firms are part of the sample, with up to 16 years of data per firm, equaling 16 observations. As is standard in the literature, financial and utility firms are not part of the

4_{At least to a certain extent, the shorter sample periods generally found in the literature will not have been a}

(25)

22 sample, as these firms tend to follow different patterns and are often subject to different regulations. The firms originate from 47 countries: 37 countries that mandatorily adopted IFRS in between 2003 and 2013. For the countries included in the sample, their IFRS adoption dates, and the number of firms and observations from each country, see Table 1 in the appendix.

In total, up to 266,644 firm-year combinations are used as input for the regression models in this paper (in model 2). This number already excludes all observations where data for any of the firm-level variables in the model and GDP growth is missing. In 49,340 of the 266,644 firm-years, firms report under IFRS. The treatment group, i.e. the firms from countries that adopted IFRS, includes 104,734 observations, the control group – firms from all other countries – 161,910 observations. The panel is unbalanced: Many firms do not exist for the whole 16 year period, and for some firms, data is not available for all years.

The empirical analysis of this papers focuses on effect of the mandatory adoption of IFRS. This is consistent with a plethora of other studies on IFRS effects, such as Beneish et al. 2012, Beusenlinck et al. 2009, Brochet et al. 2013, Byard et al. 2011, Chen et al. 2013, Christensen et al. 2013, Daske et al. 2012, Drake et al. 2010, Florou & Kosi 2015, Florou & Pope 2012, Hail et al. 2014, Houqe et al. 2013, Jiao et al. 2012, Landsman et al. 2012, Lang et al. 2010, Preiato et al. 2009, and Tan et al. 2011, to name just already cited studies.

Firms that voluntarily reported under IFRS before IFRS became the mandatory accounting standard in a country were excluded from the sample. Voluntary adopters are likely to self-select into treatment and control group, thereby distorting results. Firms that mandatorily adopt IFRs do not have such a choice. Unless voluntary adopters are object of research, it is therefore advisable to exclude them in order to focus the analysis on the mandatory, exogenously imposed adoption of IFRS (Christensen et al. 2013, Daske et al. 2008 & 2013, Florou & Kosi 2015, Lang et al. 2010, Leuz & Verrecchia 2000). Excluding voluntary adopters does not substantially reduce the sample size: If they were included in this sample, it would comprise only about 1300 firms more – compared to the 32,659 that are already part of the sample.

(26)

23 between the (winsorized) minimum and the maximum: a SROA of 0.0035 for a number of firms-years, and 1.9287, for Timestrip PLC from the UK in the year 2002, respectively. The mean SROA is 0.1377 across all firms and years, with a substantially lower median of 0.0569. The median mean SROA by firm is close to this value, 0.0669.

When examining the development of SROA over time as well as between firms from countries that adopted IFRS and from the control group, two patterns are visible, as illustrated in Table (and graph) 3. First, in both groups, mean SROA is first increasing over time after 1998, peaking in the years 2004-2005, and then decreasing again. As IFRS was introduced in the EU and a number of other countries in 2005, this year happens to be the year in which about two thirds of all firms in the sample that did adopt IFRS IFRS did so. Looking at this trend alone would lead to the conclusion that adopting IFRS reduced corporate risk taking. What is more, SROA is lower in most years in the IFRS adopter countries than in the control group. A simple cross section analysis (without controls), for example for the year 2008, would therefore also suggest that IFRS reduced risk taking. Only by examining the differences in SROA between IFRS adopters and non-adopters does it become apparent that SROA is not falling as fast in the former group after 2005 as it is in the latter, and in the last years in the sample, SROA in adopting countries even starts to overtake SROA in the control group. This pattern illustrates the need for a DiD approach for estimating the effect of IFRS on corporate risk taking.

The control variables included in the analysis display sufficient variation as well. The range of LEVERAGE, for instance, is equal to the maximum possible range: Many firms have no debt at all over the whole period, and many firms have a debt-to-assets ratio of de facto 1 as well, that is, they are on the brink of bankruptcy. Total assets in 1000 US$, of which the logarithm is taken to calculate SIZE, range from single digit values for many firms to corporate behemoths like Royal Dutch Shell, Volkswagen, and General Electric, with the last displaying total assets of almost $800bn on its balance sheet. The (winsorized) values for return over assets, ROA, range from -4.357 to 0.461, with the mean being at -0.006 and the median at 0.087: The median firm has a positive return on assets, but some firms incur huge losses.

(27)

24 comes at a loss of 4 countries and 9,106 observations (compare model 1 to 2 in Table 5). That said, the majority of results presented in the following section hold up regardless of whether or not REGQ is included as control, as is exemplarily shown in model 2. Table 1 reports the scores of by country (as mean over the 16 year period). ACCDIFF ranges from 0 different items between Singapore's and South Africa's pre-IFRS accounting standards and IFRS, to 17 for Greece's and 18 for Luxembourg's, with an average score of 8.667 (see Table 1).

4.2. Regression Results

Main Results

The results of the estimation of (EQ1) in the basic specification are reported in Table 5, model 1, in the appendix. This is the baseline model of the analysis in this paper. For the firms in the sample, the adoption of IFRS has a positive effect on corporate risk taking. On balance, firms experience an increase in SROA by 0.007 after they report under IFRS. 0.007 is an increase of 2.5% of a standard deviation of SROA (0.2751), which equals an increase by 5.1% at the mean SROA of 0.1377. This is a small effect, but it is statistically significant at the 1% confidence level.

This finding supports Hypothesis 1a - IFRS increases corporate risk taking - and is thus in line with the majority of the arguments presented in chapter 2. Consequently, the counterhypothesis, Hypothesis 1b - IFRS decreases corporate risk taking – is not supported. Provided that this model is correctly specified and that IFRS is indeed is an exogenous shock, as was argued in section 3.1., this result is evidence for a positive causal effect of IFRS on corporate risk taking.

(28)

25 conservatively. The coefficient of CAPEX is negative but not statistically significant at any relevant confidence level; in all other estimated models, the sign of the coefficient fluctuates between plus and minus, but the variable is never statistically significant. The expected sign of MB, the market-to-book ratio, is positive, as firms with a high growth potential will undertake many and riskier investments in order to realize this potential. However, in this and in all other estimations, MB is always virtually 0, and never statistically significant either. In general, the coefficients and statistical significance of all controls are relatively robust across all estimated models. The controls variables will therefore not be commented further in the followings sections of chapter 4.

LEVERAGE is positive and significant at the 5% level. Theoretically, highly leveraged firms are expected to take less risks, as their default risk is already high due to their high leverage, and due to debtholders being comparatively influential due to the importance of debt as source of financing. However, SROA is not a perfect measure of corporate risk taking, it is merely the standard deviation of the return over assets. The volatility of these returns increases with leverage, as higher leverage means greater (fixed) debt payments as well as lower (flexible) dividend payments. Consequently, high returns in good times are even higher, and negative returns in bad times are even lower. Leaving any economic theory aside for the moment, due to the way SROA is calculated, SROA will thus increase mechanically with leverage. The positive coefficient of LEVERAGE is therefore no reason for concern regarding possible misspecifications of the model or errors in the data.

REGQ is statistically significant at the 1% level and positive, as expected: A better regulatory environment, a stronger enforcement of regulations such as accounting standards in a country is associated with higher corporate risk taking. REGQ captures a country-level aspect of corporate governance, which is generally positively associated with corporate risk taking. If REGQ is dropped, in model 2, thereby adding the observations from the 4 countries for which REGQ is not available to the sample, results are very similar to model 1: The coefficient of IFRS is 0.006, only slightly lower than in model 1, and still statistically significant at the 5% level.

(29)

26 trends, although the coefficients of the fixed effects do not closely match the pattern illustrate in Table 3 (not shown),.

Pre-IFRS Risk Taking

Hypothesis 2 states that the interaction between IFRS and pre-IFRS risk taking on corporate risk taking is negative: In firms with a high level of risk taking prior to IFRS, the overall positive effect of IFRS will be weaker, or even negative. The results from model 3 to 5 in Table 5 support this prediction.

In model 3 and 4, the sample is divided at the median of pre-IFRS risk taking, IFRS_pre, as defined in section 3.2: The mean SROA of a firm over the 3 years prior to its IFRS adoption. The control group is identical in both models: Firms from countries that did not adopt IFRS, regardless of their values of IFRS_pre in any given year. Therefore, the combined number of observations from both subsamples is not actually larger than the total sample, as adding the numbers from both subsamples might suggest. For firms with low pre-IFRS levels of risk taking, there is a positive effect of IFRS on SROA, and the coefficient of the IFRS dummy is 4 times larger than in the main model: 0.028 compared to 0.007. For firms with high pre-IFRS levels of risk taking, on the other hand, the coefficient of IFRS is negative. In both estimations, IFRS is statistically significant at least at the 5% level. Qualitatively, the result is the same no matter where the sample is exactly divided – be it at the mean, median, or between the 25% most extreme risk takers and all other firms.

(30)

27 In conclusion, the results of model 3 to 5 confirm Hypothesis 2.

These findings provide evidence for three arguments. First, the effect of IFRS on neither cost of capital, nor estimation risk, nor portfolio diversification is dependent on pre-IFRS risk taking. As the influence of pre-IFRS taking is quite strong in the estimated models, this suggests that the reduction of agency conflicts between managers and shareholders due to a lower information disadvantage of shareholders is indeed a main driver behind the effect of IFRS on corporate risk taking. The influence of creditor protection, as is shown in the following section, provides additional evidence for agency conflicts being central to explaining the effect of IFRS on risk taking.

Second, if one wants to draw a particularly bold conclusion here, then model 5 estimates the optimal level of risk taking to be at an SROA of 0.150: At this level of pre-IFRS risk taking, the net effect of IFRS on SROA is 0. Following the theoretical argument, at this level of risk taking shareholders are apparently content with the riskiness of the investments pursued by the management, as they do not use their additional power through the adoption of IFRS to induce changes in risk taking in one way or the other.

Third, in more general terms, the strong and robust influence of pre-IFRS risk taking demonstrates the need to take the level of risk taking in a firm into account when examining the effect of corporate governance-related changes on corporate risk taking. A uniform relationship should not be assumed here. Even if the effect of corporate governance on risk taking is positive on average, as many studies suggest, and as the findings of this paper do not contest in any shape or form, improvements in corporate governance may still reduce corporate risk taking in firms where managements take excessive risks.

Creditor Protection

Hypothesis 3 states that the effect of the interaction between IFRS and creditor protection on risk taking is negative: Where debtholders are well-protected, i.e. relatively powerful vis-à-vis shareholders, they can use the additional information provided by IFRS to impose their preferences for lower corporate risk taking better on firms.

(31)

28 mean and median CPI, which happen to divide the sample at the same point as CPI is a discrete variable. All countries with a CPI of 2 or higher belong to the high CPI subsample. For firms in the low CPI subsample, the effect of IFRS on SROA is positive, whereas it is negative for firms in the high CPI subsample. In both cases, the IFRS dummy is statistically significant at the 1% level. It is noteworthy that the estimated coefficient of IFRS in the low CPI subsample is about 3.5 times larger than in the baseline model 1 based on the complete sample: 0.026 compared to 0.007.

If an interaction term between the IFRS dummy and CPI is incorporated instead, as in model 8, these results are confirmed. The coefficient of IFRS is 0.020, the coefficient of the interaction term -0.006. Both coefficients are statistically significant at the 1% level. At the median CPI of 2, the net effect of IFRS on SROA is 0.008, almost identical to the coefficient of SROA in model 1.

The findings from model 6 to 8 support Hypothesis 3: The generally positive effect of IFRS on corporate risk taking is weaker where creditor protection is strong, and even negative in the countries with a very high CPI scores greater than 3.

This interpretation rests on the assumption that CPI is a good proxy for debtholder power in terms of their ability to influence company policy. While analyzing the validity of this assumption would go beyond the scope of this paper, there is at least one other connection between creditor protection and corporate risk taking that has nothing to with agency issues: Better creditor protection means that creditors are generally better off in case of bankruptcy. That is, from the shareholders' point of view, bankruptcy is more costly as creditors have stronger – or more strongly enforced – claims on the assets of the firm. As the risk of bankruptcy and the cost thereof is one of the costs of corporate risk taking, better creditor protection will reduce corporate risk taking in this way too (Acharya et al. 2011).

Accounting Standard Differences

This section examines whether Hypothesis 4 is true: Is the effect of IFRS on corporate risk taking stronger where its adoption constitutes a larger change in accounting standards?

(32)

29 model is identical to model 1. ACCDIFF is statistically significant at the 1% level and negative, as expected: The adoption of IFRS constitutes a reduction in the accounting differences to IFRS to zero, thus a negative coefficient means that a decrease in accounting differences is associated with an increase in corporate risk taking.

While this result suggests that (changes in) accounting standard differences do indeed matter, it is not evidence for the effect of IFRS depending on the change in accounting differences beyond the introduction of IFRS itself. As adopting IFRS always constitutes a change in ACCDIFF (at least except for Singapore and South Africa, as both their pre-IFRS ACCDIFF scores are 0), the IFRS dummy, too, captures a change in accounting differences. However, to examine whether ACCDIFF has an additional effect beyond the effect of IFRS is not possible within the framework of the methodological framework used here: Including both ACCDIFF and IFRS as variables in a regression, individually or in combination as interaction term, would be problematic, as IFRS and ACCDIFF approximate a linear combination: IFRS essentially equals a truncated (to 1 for all ACCDIFF > 1) and inverted ACCDIFF.

Using IFRS instead of ACCDIFF again, and dividing the sample between countries that experienced a low and a high reduction in ACCDIFF through the adoption of IFRS, it becomes apparent that the effect of IFRS on corporate risk taking is indeed stronger in countries that experienced a higher reduction in ACCDIFF. In models 10 and 11, the sample is divided at the median pre-IFRS ACCDIFF of 6, but dividing at any other quartile leads to qualitatively identical results (not shown). Analogously to model 3 and 4, both subsamples use the same control group of firms from non-adopter countries. In the regression of SROA on IFRS and controls on the low ACCDIFF subsample, the IFRS dummy is positive, with a coefficient of 0.005, but not statistically significant. In the high ACCDIFF subsample, the coefficient of IFRS is 0.011, and statistically significant at the 5% level; it is also larger than the coefficient of model 1 based on the whole sample, 0.007.

(33)

30

The Strength of Enforcement

The World Governance Indicator for regulatory quality, REGQ, is included as control variable as proxy for regulatory quality in general and the strength of enforcement of accounting rules in particular in almost all estimations in this paper. It almost always has a positive, statistically significant effect on corporate risk taking. In addition to the positive influence of REGQ on SROA itself, the effect of IFRS on SROA is also stronger where REGQ is higher, as the results from model 12 and 13 show. Here, the sample is divided at the mean between countries with high and low REGQ prior to the introduction of IFRS (measured as mean REGQ over the years 1998-2004 by country).

The effect of IFRS on SROA is positive in countries with high REGQ, and negative in countries with low REGQ. For the high pre-IFRS REGQ subsample, the effect is stronger than in model 1, the baseline model, as well. In both model 12 and 13, IFRS is significant at least at the 5% level.

This result is consistent with the findings of the many studies cited in section 2.4. that find the quality of enforcement to matter for the strength of the effect of the adoption of IFRS on a range of variables. However, this result is not robust. If the sample division is replaced by an interaction term between REGQ and the IFRS dummy, IFRS*REGQ, as in model 15, then this interaction term does not enter the model significantly. Ignoring statistical insignificance for the moment, the sign of the coefficient of IFRS*REGQ is negative too, contrary to the prediction: The effect of IFRS on risk taking should be stronger where enforcement is stronger.

(34)

31

4.3. Robustness Tests

Shorter Sample Periods: 2001-2008 and 2000-2009

As mentioned in section 4.1., the years 2001 to 2008, i.e. the 4 years before and after the IFRS introduction in the EU, or similar periods, are time frames frequently used in empirical studies on IFRS. For consistency with this literature it is therefore useful to analyze whether the main result of this paper – the positive effect of IFRS on risk taking estimated in model 1 – holds up if the sample is limited to shorter periods, even though this happens at the expense of missing out on the IFRS introduction in all countries that adopted IFRS later.

The models 15 and 16 in Table 7 in the appendix show the results of the same specification as the baseline model 1, except with the sample being limited to the years 2001-2008 and 2000-2009, respectively. Overall, the results are very similar to the output from model 1: In model 15, the coefficient of IFRS is the same, albeit statistical significance is a lower; in model 16, the coefficient of IFRS is a bit higher. The lower significance may be at least in part due to the two samples containing only slightly more than half respectively two thirds of the number of observations of the full sample.

Windows Around the IFRS Adoption Year

In order to focus the analysis on the time around the event of interest, the date of the introduction of IFRS in a country, and to eliminate the potentially distorting influence of events happening some time before or after this date, several studies opted for very short sample periods of 4 to 7 years and examining only one IFRS introduction year – 2005. Examples here include Beuselinck et al. 2013, Brüggemann et al. 2011, DeFond et al. 2011, Naranjo et al. 2014, and Preiato et al. 2009).

(35)

32 and after 2007 are excluded; for IFRS adoption in 2008, it all observations prior to 2006 and after 2010 would be excluded. Apart from limiting the sample size in this way, the specification here differs from the baseline model 1 in two other aspects as well, namely that neither a control group nor year-fixed effects are included. With overlapping time windows, an appropriate control group is not easily defined. As the time frame is very short – max. 5 resp. 7 observations per firm – the scope for time trends that would otherwise captured by year-fixed effects to distort results is limited. Including year-fixed effect would even be problematic here, as they may end up capturing a great deal of the variance actually due to IFRS because there is limited overlap between the different windows. Most countries in the sample adopted IFRS either in 2005 or around 2010 to 2012, thus IFRS is introduced in mostly the same year within each window. In consequence, including year-fixed effects could lead to collinearity problems.

The results of estimating model 17 and 18 are presented in Table 7. With coefficients of 0.006 and 0.008 respectively, IFRS is of the same magnitude as in the baseline model. Statistical significance of the variable is only at the 5% level in both estimations, although that will be at least partially due to the much smaller sample sizes: The full sample contains 9 times more observations than the sample with the 5 year window.

The estimated effect of IFRS on corporate risk taking is very similar in both this and the preceding section. These findings suggest that the estimated effect of IFRS is not sensitive with regards to limitations of the sample size in terms of the years that are included, a choice which is always arbitrary to some degree. In doing so, models 15-18 also confirm Hypothesis 1a and buttress the main finding of this paper: The adoption of IFRS increases corporate risk taking.

The Effect of IFRS on Corporate Risk Taking / /