A test of market efficiency regarding pension accounting methods

(1)

A test of market efficiency regarding pension accounting methods

Cross-sectional analysis into the impact of alternative pension accounting methods by Dutch listed companies

Master Thesis, MSc Accountancy

University of Groningen, Faculty of Economics and Business

October 14, 2013 Elise Hesp Studentnumber: s1735578 Derde Kostverlorenkade 29-4 1054 TR, Amsterdam Tel.: +31 (0)648350172 E-mail: e.y.hesp@gmail.com

First supervisor university: Prof. Dr. R.L. ter Hoeven RA

Second supervisor university: W. Kevelam RA

Supervisor Ernst & Young: R. Koekkoek RA

(2)

2

Abstract

Given the recent controversial amendment to IAS 19 which eliminates the well-known and applied corridor method and the inconsistent research evidence on market efficiency, it is interesting to test whether the market can actually distinguish between these pension accounting methods. This thesis explores the ability of investors to fully process available pension information when establishing share prices. Cross-sectional analysis between 2005 and 2012 of Dutch listed companies applying defined benefit plans reveals no evidence that the market is fooled by the complex pension accounting methods. Instead, the evidence suggests that investors as a whole can see through the effects of alternative pension accounting methods at the time it becomes publicly available in the firms’ annual report. I find that it does not matter whether a company applies the fair value method or the corridor method, both methods lead to the same stock price reaction. In addition, after splitting the sample in companies that encountered actuarial gains and companies that encountered actuarial losses, I still found no significant difference in the stock price reaction of the two pension accounting methods. This implies that investors accurately assess the accounting effects of the fair value method, respectively the corridor method.

(3)

3

1. Introduction

There’s an ongoing shift of financial reporting standards for listed companies towards fair value reporting, notably the increasing importance of fair value as an accounting measurement attribute (Hitz, 2007). This trend is reflected by an increasing substitution of cost based measures by market-based measures of standard-setting boards such as the FASB and the ISAB. This tendency to fair value reporting also influenced the financial reporting of corporate pension plans. Recently, there have been some major changes to pension accounting options allowed.1

Since 2006, the Financial Accounting Standards Board (FASB) requires companies listed on U.S. Stock exchanges to recognize the fair value of their net pension assets/liabilities on their balance sheet (SFAS 158). The Board motivates the issuance of SFAS 158 by arguing that it improves financial reporting because the information reported by a sponsoring employer in its financial statements is more complete, timely, and therefore, more representationally faithful. Thus, it will be easier for users of those financial statements to assess an employer’s financial position (FSAB, 2006).

Following this implementation and after considerable discussion and some controversy, on the 16th of June 2011, the IASB issued a revised version of IAS 19 ‘Employee Benefits’ (IAS 19R), which is effective from 1 January 2013 for companies that apply IFRS. One fundamental change is that it eliminates the accounting options available under current IAS 192_{and mandates fair value pension accounting, i.e. actuarial gains and losses will be}

recognized immediately in other comprehensive income. Thus, the amendment will abolish the current option for entities to delay recognition of actuarial gains and losses if the net cumulative unrecognized value of actuarial gains and losses do not exceed certain thresholds, the so-called ‘corridor method’ (IAS 19.92-93).

Companies currently using the corridor method, experience direct consequences in their financial position from this amendment. Given the market development in recent years (volatile discount rate and return on share, respectively, real estate) many actuarial losses occurred (Deloitte, 2010). Elimination of the corridor method will then result in a decline of the equity and may have consequences for the possibility of lending money for the company

1_{i.e. to the measurement of the net defined benefit obligation (DBO) for post-employment benefit plans with employee} contributions

2_{Under IAS 19 three options for the recognition of actuarial gains and losses were available: immediate recognition through} OCI; immediate recognition through profit or loss and deferred recognition through profit or loss (i.e., corridor approach)

(5)

5

(Deloitte, 2010). In addition, the valuation of pension obligations rests on financial and demographic assumptions, and small variations in them can cause large changes in the estimates (Fasshauer and Glaum, 2012). A change of 1 % in the discount rate (which is not uncommon) that is used to calculate the present value of the defined benefit obligation (DBO) under IAS 19R, can lead to a shift in the DBO of 25% (Deloitte, 2010). Fair-value measurement of pension obligations can thus induce a high degree of volatility in financial statements, something company managers have long strongly opposed (Fasshauer and Glaum, 2012). However, by only allowing fair value pension accounting the amendment can enhance the comparability of pension accounting information in the financial statements.

The corridor method creates stability of the pension cost in the profit and loss account (Hann et al, 2007). Proponents of the corridor method argue that the long periods for which defined benefit plans are held will allow the actuarial gains or losses to reverse or offset each other (Morais, 2008a). Thus, the smoothed net recognized pension liability is seen as a stable and reliable estimate of the “true” long-term pension obligation (Fasshaur and Glaum, 2012).

However, as financial analysts and other critics have pointed out, the corridor method renders financial reporting incomplete and opaque (Fasshauer and Glaum, 2012). Arguments for the fair-value measurement of pension obligations are that it is simpler and more transparent, i.e. better to understand (JP Morgan, 2006). Fair values reflect all available information about the pension obligation at the balance-sheet date (Fasshauer and Glaum, 2012).

Given these new regulations, the discussion surrounding it and their potential harmful impact on companies, it is interesting to test whether the market can actually distinguish between these pension accounting methods and understand the implications of these alternative methods on income and equity, i.e. whether or not the market is fooled by the accounting effects of these pension accounting methods.3 Two competing descriptions of investor behaviour can be distinguished: the efficient market hypothesis and functional fixation hypothesis. While the efficient market hypothesis assumes that investors as a whole will look through reported net income to the underlying implications of alternative accounting methods, the functional fixation hypothesis argues that investors are fixated on net income and do not take the underlying implications of alternative accounting methods into account.

In recent years, serious questions have been raised about investor rationality and securities market efficiency (Scott, 2012). Prior empirical research concerning market efficiency

(6)

6

provides inconsistent results and indicates that securities are often mispriced relative to their efficient market value. In this light, Scott (2012, p. 186) argues that: ‘Questions of investor rationality and market efficiency are of considerable importance to accountants since, if these questions are valid, the practice of relying on supplementary information in notes and elsewhere to augment historical cost-based financial statements proper may not be completely effective in conveying useful information to investors. Furthermore, if shares are mispriced, improved financial reporting may be helpful in reducing inefficiencies, thereby enabling securities markets to work better’. Translating this to pension accounting, it suggests that if there is a low degree of market efficiency, relying on the unrealized actuarial gains and losses in the notes of the financial statements (e.g. the corridor method) to provide investors with useful information will not be sufficient, indicating a potential benefit of the amendment to IAS 19.

Usefulness of the financial statements to their users is central in the objective of the IASB4. In addition, the boards motivation for issuing a revision to IAS19 is that the financial statements will be more useful to investors (IAS 19(2011).BC70). Given the focus on usefulness of accounting information and the inconsistent evidence surrounding securities market efficiency, it is interesting to empirically test whether the efficient market hypothesis holds for pension accounting or that there lies more truth in functional fixation. Thus, in this study I will test if the utilization of alternative accounting methods leads to a different reaction of the securities market, in particular, whether the securities market reacts differently to companies applying the corridor method and companies applying the fair value method. The central research question therefore is: ‘Does the pension accounting method which a company applies affect the stock price reaction?’

The remainder of this thesis is structured as follows. The next section (Section 2) provides the theoretical background for this study and introduces the hypotheses. Following this, in Section 3 I outline the research design of this study, including the sample selection process and the research method. In section 4 the research results are presented and section 5 contains the conclusion and discussion.

4_{The objective of the IASB is: ‘to develop, in the public interest, a single set of high quality, understandable, enforceable} and globally accepted financial reporting standards based upon clearly articulated principles. These standards should require high quality, transparent and comparable information in financial statements and other financial reporting to help investors, other participants in the world’s capital markets and other users of financial information make economic decisions’ (http://www.iasplus.com/en/standards/standard56).

(7)

7

2. Theory

In this section I relate my research to the existing literature and develop my hypotheses. First, I start with a brief description of IAS 19 and the issued revision (2.1). Following this, the position of this study in the research spectrum will be discussed (2.2). Thereafter, an overview of the two opposite views of how stock prices are set will be presented, namely the efficient market hypothesis (2.3) and the functional fixation hypothesis (2.4). Research related to these hypotheses will be discussed (2.5), after which the hypotheses are developed (2.6).

2.1 IAS 19

Since 2005, listed European companies are mandated to report their consolidated financial statements in accordance with IFRS. The adoption of the same standards for all listed European companies was seen as an important instrument for achieving more transparent, consistent and comparable financial information at an international level (Morais, 2008a). However, the International Accounting Standard of interest in this study (IAS 19) allowed companies to choose between 3 pension accounting options for defined benefit plans that vary in the way they recognize the actuarial gains and losses: immediate recognition through profit and loss; immediate recognition through OCI and deferred recognition through profit or loss (i.e., corridor approach).

This accounting treatment of defined benefit plans has long been criticised by companies, analysts, investors and other users (Morais, 2008b). The main concerns are that this standard may decrease the comparability of the financial statements for pension accounting information (Morais, 2008a and 2008b) and creates a potential for managers’ opportunistic behaviour (Morais, 2008a).

As a reaction, on the 16th of June 2011, the IASB issued a revision to IAS 19 ‘Employee Benefits’ (IAS 19R), which is effective from 1 January 2013 for companies that apply IFRS. Earlier application is permitted and should be disclosed if used. One fundamental change is that it mandates fair value pension accounting, i.e. the full balance sheet recognition of defined benefit plans. Actuarial gains and losses will be recognized immediately, but not in a way that they have an impact on the profit and loss statement (Fasshauer and Glaum, 2008a). Instead, actuarial gains and losses will be recognized in other comprehensive income. This means balance sheets will contain the net surplus or deficit in a plan, being the fair value of plan assets less the present value of the defined benefit obligation. IAS 19R eliminates the well-known and applied corridor method currently in IAS 19. Under the corridor approach,

(8)

8

entities could defer recognition of actuarial gains and losses if the net cumulative unrecognized value of actuarial gains and losses do not exceed certain tresholds (i.e., the corridor). If the accumulated unrecognized actuarial gains and losses do not exceed 10% of the greater of the defined benefit obligation or the fair value of plan assets (the corridor), they need not be recognized – although the entity may choose to do so (IAS 19.92-93). This amendment can enhance the comparability of pension accounting information in the financial statements, but introduces more balance sheet volatility.

2.2 Position of the study in the research spectrum

As mentioned in the introduction, usefulness of financial statement information to their users is central in the IASB’s motivation for issuing the revision to IAS 19. This must come as no surprise since decision usefulness is a deeply grounded approach in financial accounting and the frameworks of major accounting standards setting bodies, like the FASB and the IASB, are based on the decision usefulness approach.

Thus, central in this study is the decision usefulness approach. In the decision usefulness approach, a distinction is made between the following three approaches: (i) the decision model approach; (ii) the decision maker-aggregate market behaviour approach; and (iii) the decision-maker individual approach (Vergoossen, 1994). Since I test the stock price reaction to alternative pension methods, which is an aggregate market response, this research study falls within the second approach. This approach bases the relationship between aggregate market behaviour and accounting information on the theory of capital market efficiency, i.e. prices fully reflect all publicly available information (Vergoossen, 1994). Thus, a lack of securities market efficiency could bias the results of market-based research studies. Since most research relating to pension accounting information is based on this research approach and given their inconsistent results, it’s important to test whether this proposition of market efficiency holds for pension accounting information. Three types of market-based research studies can be distinguished: (i) information content studies; (ii) studies into the market impact of accounting choice and (iii) studies into the market impact of accounting change (Vergoossen, 1994). The information content studies examine whether stock prices are affected by the announcement of some event, for example an earnings announcement. The second and third types of studies correspond with research studies testing the impact of different accounting methods, respectively an accounting method change, on aggregate market behaviour. Clearly, the current study belongs to studies which are concerned with the market impact of accounting choice. The issue is whether the market is sophisticated enough

(9)

9

not to be fooled by ‘cosmetic’ accounting differences (Vergoossen, 1994). Lev and Ohlson (1982) summarized a number of early studies into the field of market efficiency. They observed that many studies in the 1970s generally pointed to investors’ ability to adjust for differences in a ‘few well-known and clearly disclosed accounting techniques’, such as different depreciation methods. However, are investors able to adjust for differences with when the accounting techniques become more complex? In 2000, Chen and Schoderbek examined how investors assessed the income effects of the deferred tax adjustment caused by the August 10, 1993 increase in the corporate income tax rate (from 34% to 35%). SFAS No. 109 required firms to include the net adjustment, arising from the remeasurement of their deferred tax assets and liabilities, in their 1993 third-quarter earnings. Because this one-period adjustment could have been estimated using deferred tax information in the notes to financial statements, fully rational investors would be able to react to it as a transitory earnings component (Chen and Schoderbek, 2000). However, Chen and Schoderbek (2000) found that investors impounded the tax adjustment into security prices at the same rate as recurring earnings, despite their different implications for future cash flows. This result is consistent with the functional fixation hypothesis. On the other hand, they also found that investors separated unusual items from recurring earnings and did not discount them into security prices, which is inconsistent with the functional fixation hypothesis. This could be due to the fact that unusual items are routinely disclosed and better known to investors than the tax adjustment (Chen and Schoderbek, 2000). In addition, their tests revealed that investor fixation was more pronounced for firms with income-increasing tax adjustments than for firms with income-decreasing adjustments. Chen and Schoderbek (2000) explain this by arguing that SFAS No. 109 is a relatively new standard and the predecessor standards rarely permitted recognition of net deferred tax assets, thus deferred tax liabilities were better known to investors. Their findings imply that investors might not be able to adjust for accounting numbers that result from more complex rules. Given the changes in regulation and the complex nature of pension accounting, it is interesting to test whether the efficient market hypothesis holds for pension accounting.

2.3 The efficient market hypothesis

The efficient market hypothesis became the dominant paradigm in the 1970s due to its apparent empirical ability to be concise and consistent in its description of stock price reaction

(10)

10

to new information (Hand, 1990).5_{In addition, its value also lies in the ability to provide an}

analytical tool with relatively explicit assumptions and relatively explicit, simple and rich implications (Ball, 1972).

A market in which prices at all times fully reflect available information is called efficient (Fama, 1970). Three levels of market efficiency can be distinguished: the weak form, the semi-strong form and the strong form efficiency. This distinction was introduced into literature by Fama in 1970 and is now a commonly accepted and widely spread distinction. If the market responds efficiently to the information set contained in the history of past securities prices the market is considered to be efficient at the weak level. The semi-strong form describes a securities market where the securities prices fully reflect all information that is publicly known. The market is efficient at the strong level if security prices reflect all information, thus including inside information. Research into the aggregate market response to pension accounting information contained in published financial statements can be looked upon as a semi-strong form test. So in future, when we refer to market efficiency, we mean semi-strong efficiency.

Important to note is that even if a securities market is efficient, this does not mean that all participants in it are said to act in a rational manner or believe the market to be efficient. The efficient market hypothesis relates to the behaviour of all participants in it, i.e. the aggregate market response. As Archibald put it, ‘even if the market as a whole were shown to act in accordance with the efficient market hypothesis this would not preclude the possibility that many individual investors acted in accordance with the naïve investor hypothesis’.6

The efficient market hypothesis recognizes that the market will react to useful information from any source, including financial statements information (Scott, 2012). In addition, the form of disclosure does not matter, because there are enough rational, informed investors to quickly and unbiasedly incorporate any reasonable form into the efficient market price, thereby price protecting investors who may not wish to conduct their own in-dept analyses (Scott, 2012). If this is true in reality, we should observe that stock prices respond in predictable ways to new information. This is because rational, informed investors will revise their expectations about future firm performance and share returns on the basis of accounting information (Scott, 2012). Revised beliefs trigger buy/sell decisions, thus leading to changes in stock prices. If there is no information content, there would be no belief revision, no

5_{Hand (1990) also argues that the simplicity and rationality underlying the efficient market hypothesis make it an} economically attractive theory.

(11)

11

resulting buy/sell decisions, and hence no associated price changes. Thus efficiency implies that it is the information content of disclosure, not the form of disclosure itself that is valued by the market (Scott, 2012). Meaning that supplementary information in financial statement notes or elsewhere is just as useful as information in the financial statements proper.

In line with this reasoning, Ball (1972) argues that in an efficient market, no market reaction to an accounting change would be observed unless the change actually conveys new information concerning real factors. He makes a distinction between real factors that influence an entities’ income and accounting factors that allow entities to select from a variety of possible incomes to report.7 Thus, information concerning accounting factors should not produce any stock price response

In addition, Beaver (1973) argues that as long as firms disclose their selected accounting method and any additional information needed to convert from one method to another, the market can see through to the ultimate cash flow and dividend implications regardless of which accounting policy is actually used for reporting. Thus, an efficient market is not fooled by differing accounting policies when comparing firm’s securities (Beaver, 1973).

Translating this to pension accounting, it suggests that it does not matter whether a firm adopts the fair value method or the corridor method, because investors are sophisticated enough to see through the effects of these differing accounting policies, i.e. they are not fooled by cosmetic accounting differences between firms’ accounting policies. In particular, the balance sheet does not have to present fair values of net pension liabilities. If the unrealized gains and losses are presented in the notes to the financial statements – which is mandatory – investors can deduce the fair value for themselves in case the organization applies the corridor method. Thus, according to the efficient market hypothesis the pension accounting method which the organization applies does not affect the stock price.

2.4 The functional fixation hypothesis

The origins of the functional fixation hypothesis can be traced back to psychology literature. Psychologists found that an individual’s prior use of an object in a function dissimilar to that required by a present problem would prevent the individual from discovering a new use for the object (Ashton, 1976). In 1966 the hypothesis was introduced in accounting literature by Ijiri, Jaedicke and Knight. They argued that if the outputs from

7_{See Ball (1972) for a more thorough discussion. He argues that the distinction between real and accounting effects on} income is far from being clear, since changes in accounting techniques can be responses to real variables (such as changes in expected future inventory prices), and they can also induce real effects (such as changes in taxable income).

(12)

12

different accounting methods are called by the same name (such as profit) people who do not understand accounting well tend to neglect the fact that alternative methods may have been used to prepare the outputs. Translating this to pension accounting, it suggests that the choice of the pension accounting method influences the decisions made by investors. However, the analogy with the psychology literature is not an exact parallel as functional fixation in the psychology literature is focused on fixation on the functions of objects, rather than fixation on accounting outputs ignoring the underlying accounting methods (Ashton, 1976).

Essentially, functional fixation is a hypothesis about individuals; individual investors interpret earnings numbers in a certain way disregarding the accounting procedures used to calculate them (Vergoossen, 1994). Several hypotheses can be distinguished that belong under the header of the functional fixation hypothesis, which are strongly interrelated and overlapping hypotheses.

I will start with the mechanical hypothesis. This hypothesis was prior to the accounting research in the mid-1960s a common paradigm for the relation between a firm's accounting earnings and its stock price (Hand, 1990). In sharp contrast with the efficient market hypothesis, which argues that the market will react to useful information from any source, the mechanical hypothesis states that the market solely reacts to the reported earnings number produced by the firm’s financial accounting system (Hand, 1990). It argues that there is a mechanical relationship between earnings and share prices (Watts and Zimmerman, 1986). Hand (1990) makes in his study a distinction between the mechanical hypothesis and the traditional functional fixation hypothesis, where the mechanical hypothesis represents the precursor of the functional fixation hypothesis. He argued that the latter permits the possibility of two types of investors; sophisticated investors who can see through the implications of differing accounting policies, and unsophisticated ones who do not. Despite the acknowledgement of sophisticated investors, the functional fixation hypothesis does assume that the unsophisticated investors always end up pricing securities (Hand, 1990). Alternatively, the mechanical hypothesis assumes that all investors are unsophisticated.

Strongly related to this hypothesis are the monopolistic (e.g. Ball, 1972) and naïve investor (e.g. Archibald, 1972) hypotheses. The monopolistic hypothesis argues that the market takes accounting data at face value and that accountants possess a monopolistic influence over the data used by the market, since it is assumed that either competing data sources do not exist or (if they exist) are not used (Ball, 1972). The monopolistic hypothesis implies that the market cannot completely distinguish real and accounting effects and that it is therefore misled by accounting changes, reacting to each in a like fashion (Ball, 1972).

(13)

13

The naïve investor hypothesis perceives that the majority of the investing public is relatively unsophisticated with respect to financial reporting and does not appreciate the nature of the methods and principles that underlie the financial statements (Archibald, 1972). Thus, while unable to understand the impact of different accounting methods, investors base their valuations of an entity on measures like net income, thereby neglecting the accounting procedures used to calculate that net income. Consequently, the market in total reacts in a naïve manner to accounting information (Vergoossen, 1994).

Each of the above hypotheses highlights a slightly different aspect but all support the central theme in functional fixation. That is, individual investors interpret accounting information without regard to the rules used to produce it (Ball 1972; Hand, 1990; Watts and Zimmerman 1986). It predicts that investors are fixated on earnings and can be systematically misled by firms' accounting methods and choices (Hand, 1990).

Since there are so many similarities between these hypotheses most research studies describing functional fixation only mention the functional fixation hypothesis and in the description of that hypothesis they summarize the different aspects of the hypotheses above.

The functional fixation hypothesis is enforced by the common practice of measuring market values by multiplying firms’ earnings per share by a current rule-of-thumb multiplier that seems reasonable under current conditions (Keller and Zeff, 1969). Increases in reported net income (regardless of the source) are reflected by a proportionate increase in a firm’s share price. Henceforth, an increase in reported net income may lead to an increase in share price even though it is directly attributable to a change in accounting method (Archibald, 1972).

By now the implication of functional fixation for pension accounting should be clear. If the functional fixation hypothesis is correct with regard to pension accounting, the market is expected to focus on the earnings number produced by the financial statements. Since the fair value method affects other comprehensive income and the corridor method does not, the pension accounting method which the organization applies is expected to affect the stock price reaction.

2.5 An overview of related research

As mentioned in the introduction, in recent years, serious questions have been raised about whether it can be assumed that securities markets are efficient and investors behaviour is rational on average (Scott, 2012). That is, there is evidence that contradicts market efficiency. Some research has confirmed the EMH and argues that investors can ‘see through’ the effects

(14)

14

of different accounting policies (Ball, 1972). More recently, this conclusion has been challenged.

Despite the difficulties of designing experiments to test the implications of decision usefulness, accounting research has established that security market prices do respond to accounting information (Scott, 2012). Most convincing and widely researched is the evidence for the relation between stock prices and earnings. The first significant evidence of stock price reaction to earnings announcements was provided by Ball and Brown in 1968. Since then, many studies have documented a positive contemporaneous association between stock returns and earnings, which is generally attributed to earnings' ability to summarize value relevant information (Sloan, 1996). However, there is research evidence that investors do not correctly use available information in their forecasts of future earnings performance (Ou and Penman 1989; Bernard and Thomas 1990; Hand 1990; Maines and Hand 1996). This evidence raises the possibility that the well documented association between earnings and stock returns may, in part, reflect investors' naive fixation on reported earnings, rather than earnings' ability to summarize value relevant information (Sloan, 1996).

In addition, unlike the impressive evidence of market reaction to earnings, it has been difficult to find direct evidence of usefulness of other financial statement information (balance sheet and supplementary information). The market does not seem to respond to other financial statement information as strongly as it does to earnings information (Scott, 2012).

Scott (2012) points to various reasons for this lack of a strong market response: methodological difficulties, low reliability, availability of alternative information sources, or failure of efficient markets theory. Bernard (1993) argues that evidence inconsistent with market efficiency suggests either an incomplete initial response to earnings announcements (Bernard and Thomas, 1989) or an overreaction (Ou and Penman, 1989) or functional fixation (Hand, 1990). For example, Hand (1990) found a stock price reaction to the reannouncement of the accounting gain arising from a debt-equity swap. Since the swaps produced an immediate accounting gain that amounted to about 20% of earnings for the quarter in which the swap was undertaken, sophisticated investors would have seen this gain at the initial swap announcement. Thus, his finding contradicts market efficiency. He explained his results by the fixation of unsophisticated investors on reported accounting numbers. In addition, Sloan (1996) investigated whether stock prices reflect information about future earnings contained in the accrual and cash flow components of current earnings. His results indicate that earnings performance attributable to the accrual component of earnings exhibits lower persistence than earnings performance attributable to the cash flow component of earnings. However, his

(15)

15

results also indicate that stock prices act as if investors "fixate" on earnings, failing to distinguish fully between the different properties of the accrual and cash flow components of earnings. Stock prices do not reflect this information fully until it impacts future earnings. Consequently, firms with relatively high (low) levels of accruals experience negative (positive) future abnormal stock returns that are concentrated around future earnings announcements (Sloan, 1996). The above findings indicate that investors might not be able to adjust for accounting numbers that result from more complex rules.

2.6 Hypotheses

Accounting rules for corporate pensions are complicated and opaque (Jin et al., 2006). Despite the complexity surrounding pension accounting methods, as shown above with accounting information in general, numerous empirical studies have demonstrated that pension accounting information is significantly associated with share prices (Glaum, 2009). This seems to imply that pension accounting information is used by the investors in their decision making process. Taking it one step further, Jin et al. (2006) found evidence supporting the efficiency of capital markets with regard to pension accounting. They examined whether systematic equity risk of US firms as measured by beta from the capital asset pricing model reflects the risk of their pension plans. Their regression tests indicate that equity betas of firms do appear to accurately reflect the betas of their pension assets and liabilities. Thus, they concluded that ‘the stock market seems to process the available information without bias despite the practical difficulties of deciphering corporate pension accounts’.

However, also with respect to pension accounting, evidence supporting the functional fixation hypothesis can be found. Coronado and Sharpe (2003) found evidence indicating that pension information from the profit and loss account is more value relevant than the balance sheet pension information and that the funded status is not value relevant. They explain their results by arguing that investors were fixated on earnings and profit. Landsman and Ohlson (1990) found that pension accounting information predicts future stock returns, which also contradicts market efficiency. In addition, Franzoni and Marin (2006) argue that the market significantly overvalues firms with severely underfunded pension plans. Their results show that these companies earn lower stock returns than firms with less underfunded or overfunded pension plans for at least 5 years after the first emergence of the underfunding. They explain their results by arguing that investors do not anticipate the impact of the pension liability on future earnings, and they are surprised when the negative implications of underfunding

(16)

16

ultimately materialize. They conclude that underfunded firms have poor operating performance, and they earn low returns, although they are value companies (Franzoni and Marin, 2006). Picconi (2006) found similar evidence. Their results indicate that the off-balance-sheet portion of the pension plan’s funded status and the projected benefit obligation are predictive of future returns while the on-balance-sheet portion of the funded status is not. That is, ceteris paribus, companies with high projected benefit obligations and high off-balance-sheet liabilities experience significant negative returns in subsequent periods. This implies that investors do not accurately assess the long-run cash flow and earnings implications of these off-balance-sheet pension disclosures (Picconi, 2006). He concluded that ‘investors can completely process the pension information that has already been recognised in income, but fail to fully impound the valuation impact of pension liabilities disclosed only in footnotes’.

Given this inconsistent research evidence, I formulate a null hypothesis as well as an alternative hypothesis. My null hypothesis is consistent with the efficient market hypothesis and argues that whether a company applies the fair value method or the corridor method does not affect the stock price reaction. My alternative hypothesis is consistent with the functional fixation hypothesis and argues that the pension accounting method applied does affect the stock price reaction.

H01: The pension accounting method which a company applies does not affect the stock price

reaction.

Ha1: The pension accounting method which a company applies does affect the stock price

reaction.

If the functional fixation hypothesis is correct with regard to pension accounting, a distinction should be made between companies that encountered actuarial gains and companies that encountered actuarial losses. Under functional fixation, actuarial gains and losses have opposite effects on other comprehensive income and, consequently on the direction of the stock price reaction. More specifically, since fair value pension accounting influences other comprehensive income directly, investors can be expected to make a more positive or negative evaluation of the market value of the firm depending on whether there are actuarial gains or losses, respectively. If there are actuarial gains, applying the fair value method instead of the corridor method leads to a relatively higher other comprehensive

(17)

17

income. Investors can be expected to make a more positive evaluation of the market value of the firm, leading to a relatively higher stock price reaction. Thus, fair value pension accounting is more favourable when actuarial gains are expected. Conversely, if actuarial losses are more likely, applying fair value pension accounting leads to a relatively lower other comprehensive income and stock price reaction and applying the corridor method is more favourable. Thus, putting companies that encountered actuarial gains and companies that encountered actuarial losses in the same sample, as is the case with hypothesis 1, creates the possibility that the opposite stock price reactions balance each other out on average, thus leading to the same stock price reaction, showing no evidence of functional fixation even if investors are functionally fixated. To eliminate this potential bias, the second and third hypotheses focus specifically on the actuarial gains, respectively the actuarial losses.

H02: When a company encounters an actuarial gain, applying the fair value method or the

corridor method lead to the same stock price reaction.

Ha2: When a company encounters an actuarial gain, applying the fair value method instead of

the corridor method leads to a relatively higher stock price reaction.

H03: When a company encounters an actuarial loss, applying the fair value method or the

corridor method lead to the same stock price reaction.

Ha3: When a company encounters an actuarial loss, applying the fair value method instead of

(18)

18

3. Research design

3.1 Data sample

The data sample consists of all firms with available data that are listed on the NYSE Euronext Amsterdam.8_{Since 2005 European listed companies are required to conform to}

IFRS, thus the research is conducted from 2005 till 2012. Only firms that were already listed on the Amsterdam Stock Exchange in 2005 and since then succeeded to stay listed till 2012 are taken into account. This leads to a sample of 41 companies followed over 8 years, thus 328 company years. See Appendix A for a specification of these companies. To arrive at the final dataset I made the following adjustments, see Table 1.

Table. 1 Dataset selection

Initial dataset 328

financial year does not correspond with calendar year 16

no defined benefit pension plans 40

actuarial gain / loss not specified in annual report 1

no material effect 82

Final dataset 189

exact publication date annual report known 144

exact publication date annual report not known (replaced by KvK deposit) 45

In line with Vergoossen (1994), I adjust the dataset set to include a threshold for materiality. The actuarial gains or losses must have a material effect on total comprehensive income or equity, i.e. total comprehensive income of the year in which the actuarial gain or loss occurred or equity at the end of that year must be affected by at least five percent. This materiality criterion is important, because actuarial gains or losses with an immaterial effect on income or equity are not expected to be relevant in investment analysis (Vergoossen, 1994). The following formula was used:

Materiality criterion = ACTGLt and/or ACTGLt must be > 5% TCIt EQUITYt

ACTGL : Actuarial gain or loss encountered by a company during the financial year TCI : Total comprehensive income during the financial year

EQUITY : Group equity at the end of the financial year

t : financial year the actuarial gain or loss occurred

(19)

19

For each company for each year I manually gathered the actuarial gain or loss, the net income and other comprehensive income, and the equity at the end of the year reported in the annual report. I measured the total comprehensive income by adding other comprehensive income to net income. The actuarial gain or loss was measured as follows. If a company applied the fair value method, I used the recognized actuarial gain or loss in the profit & loss account, respectively in other comprehensive income during that year. If a company used the corridor method, I used the unrecognized actuarial gain or loss disclosed in the supplementary information to the balance sheet. The corridor method specifies that if the accumulated unrecognized actuarial gains and losses do not exceed 10% of the greater of the defined benefit obligation or the fair value of plan assets (the corridor), they need not be recognized – although the entity may choose to do so (IAS 19.92-93). In reality, none of the companies that applied the corridor method recognized the actuarial gain or loss if it did not exceed the 10% of the greater of the defined benefit obligation or the fair value of plan assets. If the actuarial gain or loss exceeded the corridor and thus the amount by which it exceeded the corridor must be recognized in the profit and loss account, I added that amount to the unrecognized amount specified in the notes to calculate the actuarial gain or loss encountered by the company during that year.

To measure the stock price reaction I used the closing stock price on the day of publication of the annual report as well as the closing stock price the day before publication of the annual report. To eliminate causes that are outside the scope of the company, I adjusted this stock price reaction for the general market mood. This is reflected by the multiple indices at the NYSE Euronext. If a company was listed at the AEX in a year, I used the AEX index to filter out potential general influences on the firms’ stock price that are not attributable to the disclosed information. If a company was listed at the AMX or AsCX, I used the Amsterdam Midkap-Index, respectively the Amsterdam Smallcap-Index. These indices are commonly used as a benchmark for the development of the stock price. As well as with the firms’ stock price reaction, I used the closing index the day of publication of the annual report as well as the closing index the day before publication of the annual report. If the NYSE Euronext Amsterdam was closed the day before publication, I used the last day before the day of publication that it was open. To gather the closing stock price for each firm and for each index I used Bloomberg. The stock price reaction was calculated on the basis of the following formulae:

(20)

20 Pt – Pt-1

ΔP = Pt-1

It – It-1 It-1

ΔP : Stock price reaction of company x P : closing stock price of company x I : closing index of AEX, AMX or AsCX t : day of publication of the annual report

t-1 : day before the publication date of the annual report

For the exact publication date of the annual report I searched the internet, the corporate website and the annual report itself. In 45 cases the exact publication data was not known. For these cases I used the KvK deposit date of the annual report as a proxy for the publication date of the annual report. During the gathering of the publication date, it came to my attention that in most cases considerable time has passed between the exact publication date of the annual report and the KvK deposit date of the annual report. The KvK deposit date sometimes being even a month later than the exact publication date of the annual report to the public. This could cloud the results, since it can be expected that there is no stock price reaction to this KvK deposit considering the information present in the annual report is already known. To eliminate this potential bias I run all tests both including all data (incl. KvK deposit) as well as without data from which the exact publication date is not known and thus the KvK deposit date is used as a proxy instead (excl. KvK deposit). I will maintain this distinction throughout this thesis.

3.2 Research method

There are different tests to measure whether there is a significant difference between two independent groups: a test, a Mann-Whitney test and the Kolmogorov-Smirnov test. The T-test is the most powerful T-test to establish whether the means of two groups are different from each other. This is because the T-test uses more information, in particular, it uses the actual values instead of the rank order of the values. The T-test uses the means of an interval- or ratio variable, where the assumption is that the observations for both groups are from a random sample from a normally divided population. If these requirements are not met, then the other tests must be used to establish a difference. Stock price reaction is a ratio variable so the first precondition is met. However, the normality of the data samples is not established

(21)

21

yet. Thus, before deciding which research method to use to test the hypotheses, I will examine whether the observed data distribution matches with a normal distribution. Normality of data can be examined through a variety of ways: (i) through descriptive statistics as the skewness and the kurtosis; (ii) through plotting a graph, like a histogram with a normal curve or a normal-distribution diagram like a normal-quantile plot or a normal probability plot; (iii) or you can test the normality by using a One-Sample Kolmogorov-Smirnov test (at least an ordinal variable) or a Chi-Square test (at least a nominal variable). Given that the One-Sample Kolmogorov-Smirnov test leaves no room for interpretation as is the case with visual graphs, it is the most reliable one and thus I use this test to explore the normality of the data samples. The results of this test are displayed below in Table 2.

Table. 2 Output One-Sample Kolmogorov-Smirnov test

N Komogorov-Smirnov Z Asmp. Sig.

(2-tailed) Sig. (1-tail)

Normal distribution*

H1 Fair Value Incl. KvK 57 1,314 0,063 Yes

Excl. KvK 49 1,243 0,091 Yes

Corridor Incl. KvK 132 1,930 0,001 No

Excl. KvK 96 1,742 0,005 No

H2 Actuarial gains Fair Value Incl. KvK 13 1,004 0,133 Yes

Excl. KvK 11 1,023 0,123 Yes

Corridor Incl. KvK 41 1,057 0,107 Yes

Excl. KvK 26 1,017 0,127 Yes

H3 Actuarial losses Fair Value Incl. KvK 44 0,626 0,414 Yes

Excl. KvK 38 0,537 0,468 Yes

Corridor Incl. KvK 91 1,654 0,004 No

Excl. KvK 70 1,539 0,009 No

* assume normal distribution if ρ > 0.05

For hypothesis 1 and 3 the results show that the sample of companies applying the fair value method is in both cases (incl KvK as well as excl. KvK) normally distributed. However, the sample of companies that apply the corridor method is not normally distributed. Since a requirement of the T-test is that both samples have a normal distribution, this test cannot be used.

For hypothesis 2 both samples seem to have a normal distribution meaning that the T-Test will be used. Because only with this division of the total sample normality seems to be the case and because the two samples consist of a limited amount of data, it seems that the case

(22)

22

for normality and accordingly the T-Test is not that strong. Thus, I will also perform additional tests to see whether the results of these tests correspond.

The Mann-Whitney test and the Kolmogorov-Smirnov test are both non-parametric tests that are used when the requirements of the T-Test are not met. Literature regarding which of the two methods to use slightly favours the Mann-Whitney test in case of large sample sizes and the Kolmogorov-Smirnov test with very small sample sizes (Siegel and Castellan, 1988). However, the Mann-Whitney test is by far the most popular of the two-independent-samples tests. It tests whether the two samples are equivalent in location. The Kolmogorov-Smirnov test is a more general test that detects differences in both the locations and shapes of the distributions. If the shapes of the two sample distributions differ, then the null hypothesis will be rejected even if the locations of both samples do not differ. The shape of the distribution can be approximated by the skewness and the kurtosis statistics. Since the sample distributions used to test the hypotheses differ in shape, the Kolmogorov-Smirnov test is not appropriate and thus, will not be used. See Appendix B for the evidence on skewness and kurtosis.

(23)

23

4. Research results

4.1 Descriptive statistics

The descriptive statistics are presented in Table 3. As mentioned in subsection 3.2, I maintain a distinction between samples including the KvK deposit data (Panel A) and samples excluding the KvK deposit data (Panel B). Since most samples are not normally distributed, I included the median and the 25% and 75% quartile in Table 3. The inter-quartile range is very small compared to the minimum and maximum values, indicating the existence of outliers. This is not surprising since normality of the sample distributions, used to test H1 and H3, is rejected by the One-Sample Kolmogorov-Smirnov test. Notable is the fact that the median (and the mean) of the corridor sample are higher than the median (respectively the mean) of the corresponding fair value sample for every combination of samples. Thus, even if the tests for H2 show a significant difference, this will not be in the predicted direction; namely that applying the fair value method leads to a higher stock price reaction than the corridor method. However, this also suggests that the predicted direction of H3 is correct. In addition, it must be noted that given the small sample size of some samples it is difficult to reject the null hypothesis. This is especially the case with the samples solely consisting of companies that

Table. 3 Descriptive Statistics

Panel A: Inclusive KvK

N Mean St. dev. Minimum 25% Median 75% Maximum

Variable: stock price reaction

H1 Fair value 57 -0,48% 2,23% -10,88% -1,27% -0,37% 0,65% 3,91% Corridor 132 -0,09% 2,95% -11,59% -0,96% -0,09% 0,95% 11,59% Total 189 -0,21% 2,75% -11,59% -1,11% -0,16% 0,85% 11,59%

H2 Fair value 13 -1,71% 3,66 -10,88% -3,08% -1,22% 0,19% 3,12% Corridor 41 -0,44% 2,79% -6,15% -2,16% -0,37% 0,71% 10,73% Total actuarial gains 54 -0,74% 3,04% -10,88% -2,05% -0,41% 0,67% 10,73%

H3 Fair value 44 -0,12% 1,46% -3,47% -0,95% -0,15% 0,68% 3,91% Corridor 91 0,07% 3,02% -11,59% -0,89% -0,03% 1,13% 11,59%

(24)

24 Panel B: Exclusive KvK

N Mean St. dev. Minimum 25% Median 75% Maximum

Variable: stock price reaction

H1 Fair value 49 -0,50% 2,23% -10,88% -1,27% -0,37% 0,65% 3,72% Corridor 96 0,26% 3,12% -11,59% -0,70% 0,13% 1,17% 11,59% Total 145 0,01% 2,86% -11,59% -1,01% -0,13% 0,94% 11,59%

H2 Fair value 11 -1,53% 3,89% -10,88% -1,62% -1,22% 1,21% 3,12% Corridor 26 0,26% 2,86% -4,79% -0,77% -0,31% 1,04% 10,73% Total actuarial gains 37 -0,27% 3,25% -10,88% -1,33% -0,37% 1,08% 10,73%

H3 Fair value 38 -0,20% 1,39% -3,47% -0,97% -0,17% 0,63% 3,72% Corridor 70 0,26% 3,22% -11,59% -0,74% 0,22% 1,20% 11,59%

Total actuarial losses 108 0,10% 2,72% -11,59% -0,91% -0,04% 0,90% 11,59%

encountered actuarial gains. The final dataset consists of much more companies that encountered actuarial losses than companies that encountered actuarial gains, which can be explained by the market development in recent years. Finally, there are no signs of large differences between the statistics of Panel A and Panel B, suggesting that it does not matter whether or not the KvK deposit data is included.

4.2 Hypothesis testing

Table 4 and 5 present the output of the Mann-Whitney test, respectively the T-test. An alpha level of 5% is used for all statistical tests. In order to test the first hypothesis, I compared the stock price reaction of companies applying the fair value method and companies applyingthe corridor method byusing the Mann-Whitney test. An examination of the findings in Table 4 reveals no statistical difference between the stock price reactions of the alternative pension accounting methods. This is true for the sample including KvK deposit (Z= -1.182; ρ= 0.237), as well as the sample excluding KvK deposit (Z=-1.739; ρ= 0.082). The mean rank of the stock price reaction of the fair value sample including KvK deposit is 87.84 (excl. KvK: 64.51), while the companies applying the corridor method had a stock price reaction mean rank of 98.09 (excl. KvK: 77.33). These results suggest that the pension

(25)

25

accounting method does not affect the stock price reaction, thus the null hypothesis cannot be rejected.

Table. 4 Results Mann-Whitney test

Mean Rank Z Asymp. Sig.*

2-tailed 1-tail

H1 Incl. KvK Fair Value 87,84

-1,182 0,237

Corridor 98,09

Excl. KvK Fair Value 64,51

-1,739 0,082

Corridor 77,33

-1,265 0,103

Corridor 29,02

-1,495 0,068

Corridor 20,73

-0,901 0,184

Corridor 70,11

-1,287 0,099

Corridor 57,36

* reject the null hypothesis if ρ < 0.05

The second hypothesis focuses on companies that encountered an actuarial gain. It is formulated to test whether the fair value method and the corridor method lead to the same stock price reaction or that the fair value method leads to a relatively higher stock price reaction. Both a Mann-Whitney test and a T-test were performed. First I will describe the results of the more powerful T-test. For the samples including KvK deposit, homogeneity of variances, as assessed by Levene’s Test for Equality of Variances (ρ = 0.403), is assumed. The results in Table 5 show no significant difference in the stock price reaction of companies applying the fair value method (M= -1.71, SD= 3.66) and companies applying the corridor method (M= -0.44, SD= 2.79); t(52)= -1.32, ρ= 0,097. For the samples excluding KvK deposit, homogeneity of equal variances is also assumed (ρ= 0.466). In addition, the results show no significant difference in the stock price reaction of companies applying the fair value method (M= -1.53, SD= 3.89) and companies applying the corridor method (M= 0.26, SD= 2.86); t(35)= -1.56, ρ= 0.065. The differences between the sample means are likely due to chance instead of the chosen pension accounting method.

The results of the Mann-Whitney test confirm the absence of a significant difference. No significant difference between the companies applying the fair value method and the companies applying the corridor method could be established, both for data including KvK

(26)

26

deposit (Z= -1.265, ρ= 0.103) and data excluding KvK deposit (Z= -1.495, ρ= 0.068). Thus, for companies that encountered actuarial gains, applying the fair value method leads to the same stock price reaction as applying the corridor method. The null hypothesis cannot be rejected.

Table. 5 Results T-test

Levene's Test for Equality of Variances T-test for Equality of Means

F Asymmp. Sig (2-tailed) t df Asymp. Sig (1-tail)*

H2

Incl. KvK 0,71 0,403 -1,32 52 0,097

Excl. KvK 0,54 0,466 -1,56 35 0,065

* reject the null hypothesis if ρ < 0.05

The third hypothesis examines whether, if a company encounters an actuarial loss, applying the fair value method or the corridor method lead to the same stock price reaction or that applying the fair value method leads to a relatively lower stock price reaction. Table 4 reports that also with actuarial losses, the Mann-Whitney test shows no evidence of a difference in stock price reaction between companies applying the fair value method and companies applying the corridor method. This is true for samples including KvK deposit (Z= -0.901, ρ= 0.184) as well as for samples excluding KvK deposit (Z= -1.287, ρ= 0.099). The mean rank of the fair value sample including KvK deposit is 63.64 (excl. KvK: 49.24) and the mean rank of the corridor sample including KvK deposit is 70.11 (excl. KvK: 57.36). Thus, the results suggest that if a company encounters an actuarial loss, it does not matter whether it applies the fair value method or the corridor method. Both lead to the same stock price reaction, showing no proof of functional fixation. Hence, I conclude that the null hypothesis cannot be rejected.

Although that for all sample combinations the corridor sample has a higher mean rank than the corresponding fair value sample, these differences are relatively small and are likely due to chance instead of the chosen pension accounting method.

Finally, the test results show that the same conclusions can be drawn for data including KvK deposit and data excluding the KvK deposit. This confirms the suggestion made in the previous subsection that it does not matter whether or not the KvK deposit data is included in the samples to test the hypotheses.

(27)

27

5. Conclusion and discussion

In this thesis, I examined whether the market can distinguish between the alternative pension accounting methods allowed under IAS 19. Given the discussion surrounding the revision to IAS 19 and the inconsistent research evidence on market efficiency, it is interesting to test investor’s ability to adjust for differences in pension accounting methods between companies. The issue is whether or not the market is fooled by the accounting effects of these alternative pension accounting methods. Cross-sectional analysis between 2005 and 2012 of Dutch listed firms applying defined benefit plans reveals no support for the functional fixation hypothesis. My findings indicate that the securities market seems to react in a non biased way to the applied pension accounting method. I find that the pension accounting method which a company applies does not affect the stock price reaction. Even if the total sample of companies is split between two samples consisting solely of companies that encountered actuarial gains, respectively actuarial losses, I still found no significant difference in the stock price reaction of the two pension accounting methods. Both with actuarial gains and actuarial losses it seems that the market can actually distinguish between these pension accounting methods. I interpret this evidence as being due to investors’ ability to fully process available pension information when establishing share prices. Hence, it does not matter whether a firm adopts the fair value method or the corridor method, because investors are sophisticated enough to see through the effects of these differing accounting policies. Thus, the evidence supports the efficient market hypothesis with regard to the pension accounting information, which is in line with earlier research by Jin et al. (2006). This does not necessarily imply that all investors act in a rational manner or believe the market to be efficient, individual investors could act in accordance with the functional fixation hypothesis. However, overall, the securities market as a whole is shown to act in accordance with the efficient market hypothesis. Hence, I conclude that investors accurately assess the accounting effects of the fair value method, respectively the corridor method, even when faced with the complexity of pension accounting information.

This thesis contains a few limitations, which raise additional issues for future research. One caveat to the results is that there are methodological difficulties. I measured the stock price reaction to the publication of the annual report. If the pension accounting method is of little influence on the stock price reaction compared to other information in the annual report, this could cloud the results. However, much of the information in the annual report that can be expected to have a substantial effect on the stock price reaction (earnings etc.) is already

(28)

28

known through earlier publication, limiting the effect of other information in the annual report on the stock price reaction. In addition, I adjusted the stock price reaction for the general market mood, eliminating factors that are not attributable to the companies’ annual report. Despite efforts to limit influences of other variables on the stock price reaction, these are not wholly eliminated. It is a challenge for future research to develop advanced statistical techniques that could better distinguish between these different effects on the stock price reaction.

Another caveat is the limited sample sizes, which makes it more difficult to reject the null hypothesis. First, in the absence of appropriate databases, I had to gather all data manually which put a limitation on sample size. Second, only 59% (189 of 329) of the initial dataset is actually used in this research. This can largely be explained by companies that applied defined contribution plans instead of defined benefit plans (12%) and companies that encountered an actuarial gain or loss that did not have a material impact on total comprehensive income or equity (25%). Finally, given the market developments in recent years, much more actuarial losses than actuarial gains occurred. Since H2 solely focuses on companies that encountered actuarial gains, this puts another limitation on the sizes of samples used to test the second hypothesis. However, in order to still be able to get a representative sample I used all Dutch listed companies with available data from 2005 till 2012. It is interesting to test whether these results hold with larger sample sizes.

(29)

29

Appendix A

List of companies in initial dataset

The list beneath presents the companies that were already listed on the Amsterdam Stock Exchange in 2005 and since then succeeded to stay listed till 2012. This leads to a sample of 41 companies.

Appendix A: List of Companies included in initial dataset

Company name ISIN Ticker symbol

AEGON NL0000303709 AGN

AHOLD KON NL0006033250 AH

AIR FRANCE -KLM FR0000031122 AF AKZO NOBEL NL0000009132 AKZA ASML HOLDING NL0010273215 ASML

CORIO NL0000288967 CORA

DSM KON NL0000009827 DSM

FUGRO NL0000352565 FUR

HEINEKEN NL0000009165 HEIA

ING GROEP NL0000303600 ING

KPN KON NL0000009082 KPN

PHILIPS KON NL0000009538 PHIA

RANDSTAD NL0000379121 RAND

REED ELSEVIER NL0006144495 REN ROYAL DUTCH SHELLA GB00B03MLX29 RDSA ROYAL IMTECH NL0006055329 IM SBM OFFSHORE NL0000360618 SBMO UNILEVER DR NL0000009355 UNA WOLTERS KLUWER NL0000395903 WKL AALBERTS INDUSTR NL0000852564 AALB ASM INTERNATIONAL NL0000334118 ASM BAM GROEP KON NL0000337319 BAMNB BOSKALIS WESTMIN NL0000852580 BOKA BRUNEL INTERNAT NL0000343432 BRNL

CSM NL0000852549 CSM

EUROCOMMERCIAL NL0000288876 ECMPA

HEIJMANS NL0009269109 HEIJM

NIEUWE STEEN INV NL0000292324 NISTI

NUTRECO NL0010395208 NUO

PHARMING GROUP NL0010391025 PHARM

TEN CATE NL0000375749 KTC

UNIT4 NL0000389096 UNIT4

USG PEOPLE NL0000354488 USG

VASTNED RETAIL NL0000288918 VASTN

VOPAK NL0009432491 VPK

WERELDHAVE NL0000289213 WHA

WESSANEN KON NL0000395317 WES BE SEMICONDUCTOR NL0000339760 BESI

GRONTMIJ NL0010200358 GRONT

ORDINA NL0000440584 ORDI

A test of market efficiency regarding pension accounting methods