The interaction of accruals and hedging : do they affect management decisions? : evidence from the European market

(1)

The interaction of accruals and

hedging: do they affect management

decisions?

Evidence from the European Market

Student: Karim Belâbar Student number: 10422943

Word count: 12155 Date: 23-06-2015

Master Thesis

Supervisor: prof. dr. V.R. (Vincent) O' Connell MSc Accountancy and Control 2014/2015, Variant Accountancy & Variant Control Amsterdam Business School

(2)

2

Statement of Originality

This document is written by student Karim Belâbar. He declares to take full responsibility for the contents of this document.

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

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

(3)

3

Abstract

This study examines the association between derivative hedging and accrual management. It is aimed to determine whether firms which do and do not hedge intensively differ in financial attributes/firm characteristics. This study also investigates the extent to which managers of European firms trade off derivative hedging and accrual management. Finally, I investigate whether firms in Europe use discretionary accruals and hedging as substitutes to manage the volatility of earnings. My results suggest that intensive and non-intensive hedgers have different firm characteristics. Secondly, this study finds evidence that derivative hedging and accrual management are both positively associated with each other. This suggests that they are complete/full complements of one another, although this complementariness applies only to the relation between Net derivatives and the absolute value of discretionary accruals. The measurement of the total-derivatives ratio indicates a complementary relationship. Robustness tests indicate a substitution relationship between hedging and accrual management. Finally, I find evidence that both derivative hedging and accrual management are positively associated with earnings volatility.

Key words: IFRS, IAS 39, IFRS9, Hedging, Derivatives, Real smoothing, Accounting smoothing, Accrual management, Earnings volatility.

(4)

4

Acknowledgments

First of all, I would like to express my appreciation to the my mother and sister for being there for me and believing in me, I am very grateful for that. Secondly, I would like to thank (Vincent) O' Connell, who was my supervisor at the Amsterdam Business School, for helpful comments and suggestions on previous versions of this thesis. I am also grateful to Niels Roeleveld, who supervised me during my internship at PwC.

(5)

5

1. Introduction

1.1 Background

At the G20 Summits in 2009, world leaders declared that improvements were needed in financial reporting (Financial Stability Board, 2009). These improvements mostly relate to IAS 39 (Financial Instruments: Recognition and Measurement), as IAS 39 is often criticized for being complex and for its inconsistencies, hence not meeting the risk management activities (Armstrong et al., 2010; Paananen et al., 2012) . Therefore, the IASB has issued IFRS 9 to replace IAS 39. The objective of IFRS 9 is to reflect the effect of an entity’s risk management activities in the financial statements. This includes replacing some of the arbitrary rules by more principle-based requirements and allowing more hedging instruments and hedged items to qualify for hedge accounting. Overall, the introduction of IFRS9 should result in more risk management strategies qualifying for hedge accounting. Though IFRS 9 is not yet endorsed by the European Union, it is expected to be effective from the fiscal year 2018 onwards.

Managers have a number of incentives for engaging in earnings smoothing: increasing management compensation, reducing corporate taxes and debt financing costs, avoiding underinvestment and earnings surprises, and mitigating volatility caused by low diversification (Barton, 2001). Different tools can be used in order to achieve a desired level of earnings. Two of these tools are hedging and accrual management (Barton, 2001; Haushalter, 2000; Pincus & Rajgopal, 2002). Both tools will be discussed below.

Hedging refers to activities intended to offset potential losses or gains (risks) of an investment position. Hedging is one of the main instruments for risk management strategies. A hedge can be constructed from many types of financial instruments, such as stocks, forwards, swaps, options, etc., depending on the risks the investor wishes to offset. Interest rate risk, for instance, the risk that the relative value of an interest-bearing liability, such as a loan or a bond, will worsen due to an interest rate increase, can be hedged by using fixed-income instruments or interest rate swaps. Other potential risks include commodity or volumetric risk, credit risk, Foreign Exchange or volatility risk, and equity risk.

The other tool, accrual management, is defined managing earnings which does not effect the cash flow of the company (Roychowdhury, 2006). While Healy and Wahlen (1998) define earnings management as when management changes the financial reporting of a firm

(7)

7

in favor of contractual outcomes or to mislead stakeholder about the actual financial performance

A few prior studies have examined hedging and accrual management as a means of achieving a desired level of earnings. These studies suggest that managers view hedging and accrual management as partly substitute tools for smoothing earnings (Barton, 2001; Choi, Mao & Upadhyay, 2011; Pincus & Rajgopal, 2002). In other words: other things being equal, managers are indifferent between using hedging activities or accrual management in order to smooth earnings.

Investigating the trade-off between hedging with derivatives and accrual management is important, because of the soon to be replaced IAS 39, which is, as stated above, heavily criticized for its complexity. In the past decades, the International Accounting Standard Board (IASB) regulatory framework for financial reporting on the disclosures of the use of hedging has experienced some fundamental changes. IAS 39 became effective in January 2005, after it was reissued in December 2003 (Deloitte, n.d.). IAS 39 obliges firms - that have to report under IFRS - to annually report derivatives at fair value in their balance sheets and to include the corresponding gains and losses in earnings (unless the derivatives meet explicit hedging accounting rules1). The accounting treatment of IAS 39 impeded managers to use hedging as a

smoothing mechanism for earnings, as it became more difficult to meet the accounting requirements. Osterland (2000) even argues that SFAS No. 133, which is similar to IAS 39 results in increasing earnings’ volatility.

Experts expect its successor, IFRS 9, to make it less difficult for firms to meet the specific hedge accounting rules compared to its predecessor. If managers in Europe use accrual management and hedging as a substitute for mitigating earnings volatility, a shift may take place from accrual management to hedging.

1.2 Research Question

Based on the above-mentioned issues in the accounting treatment of hedging under IAS 39, the objective of this study is to assess and empirically examine whether and how hedging and accrual management activities are used by managers to control earnings volatility. More specifically, I examine whether the substitution between hedging with derivatives and accrual management holds with a European based sample. Consequently, the research questions is as follows:

(8)

8

“To what extent do managers of European firms trade off between derivative hedging and accrual management to alleviate earnings volatility?”

1.3 Motivation

This section discusses the relevance of this study from a practical and an academic point of view. Previous papers have examined managers’ trade-off between multiple earnings management techniques. However, little research has been conducted regarding the trade-off between hedging with derivatives and abnormal accruals. In particular, only one recent study by Choi et al. (2011), not yet published nor edited, examines whether the adoption of a new accounting treatment can alter the trade-off between derivative hedging and accrual management in getting a desired level of earnings.

In comparison to prior studies, this study contributes to existing research and literature in several ways. Firstly, this study re-examines the substitution hypothesis - raised by Barton (2001) - in a different time span. More broadly, I extend the research literature by providing empirical evidence and documenting results. For this, I use almost the same data as prior studies, making my results comparable to earlier studies. To achieve this, I use multiple databases to overcome any data limitation problems. My study will only focus on the fair value, because of lack of disclosed notional values. The findings of Ahmed et al. (2006) suggest that SFAS No. 133 has improved the transparency of derivative use (I except the same for IAS 39, as IAS 39 and SFAS 133 are comparable), which makes it possible to measure a firms hedging ratio in a different manner.

Secondly, this study contributes to the limited empirical research on the decision to use hedging in Europe. Most studies on hedging or accruals management are conducted in the US with US based data samples, in which data on the notional value was examined (Barton, 2001; Choi et al., 2011; Pincus & Rajgopal, 2002). To my knowledge, no European based study documented evidence on the magnitude of the use of hedging. This research addresses an important gap and responds to calls made in prior literature (Barton, 2001; Latridis, 2012). My study seeks to reduce this gap by using a European based data sample from five different countries (thus cross-country or transboundary) and to see whether my findings differ from previous research conducted on hedging with derivatives and accrual management.

Finally, from a practical point of view, it is interesting to study whether the relation between hedging with derivatives and accrual management in Europe is the same as in the US. While hedging with derivatives and accrual management may have proven to be substitutes in the US, such relation may not exist in Europe because of geographical and

(9)

9

national differences. Prior studies by Barton (2001) and Pincus and Rajgopal (2002) were conducted by using a US-based sample and are from 14 years ago. Therefore, this underlying study on European data is necessary, because it examines whether the same outcomes by Barton (2011), Pincus and Rajgopal (2002) and Choi et al. (2011) hold in a different (geographical) environment.

1.4 Structure

This study proceeds as follows: The next section provides an overview of the basic relation between hedging with derivatives, accrual management, and earnings volatility. This overview is followed in section 2 by a description of prior literature concerning hedging and accrual management. Finally, section 2 ends with the development of the research hypotheses. Section 3 explains the sample characteristics and provides an overview of the proposed data and the research methodology. Furthermore, empirical models of hedging with derivatives and accrual management are developed and the dependent and explanatory variables are explained. Finally, section 4 discusses the empirical findings and section 5 presents the conclusion of this study.

(10)

10

2. Literature review and hypotheses

This section starts by defining and discussing managerial incentives and positive accounting theory, which is followed by defining hedging and accrual management and its determinants. Next IAS 39 and its successor IFRS 9 (effective from 2018 onwards) will be discussed thoroughly. This chapter concludes with an overview of literature regarding the interaction between hedging with derivatives and accrual management. From these studies, the hypotheses are derived.

2.1 Managerial incentives

Managers have several incentives to maintain a desired level of earnings. The Positive Accounting Theory (PAT) from Watts and Zimmerman (1986) exemplifies the motives managers have to manage earnings from the future period to the present period and states that managers act out of their own or their company’s interests. This broad view of the PAT is set put in three hypotheses, namely the bonus plan hypothesis, the debt/equity hypothesis, and the political cost hypothesis (Watts & Zimmermann, 1986).

• The bonus plan hypothesis suggests that managers, who are partly compensated by a bonus which depends on or is related to the firm's net income, have incentives to use accounting techniques to shift net income from the future period to the present period. If the company’s net income increases, the managers will have better compensation. In addition, this hypotheses states that managers can decrease the firm’s net income when the earnings are already beneath the bonus threshold. Watts and Zimmermann (1986) call this decreasing of earnings an “earnings bath”.

• The debt/equity hypothesis implies that managers have incentives to use accounting techniques to shift net income from the future period to the present period when a firm is closer to a violation of debt covenants. In other words, manager use these techniques to avoid these covenants, because firms can be penalized (e.g. higher interest rate) when violating these. This view is strengthen by Burgstahler and Dichev (1997). They state that banks offer more favourable conditions for firms that have higher net income than firms that do not.

(11)

11

• The political cost hypothesis implies that political costs can be compulsory due to high profits. It also suggests that managers that are more likely to counter political cost, may have more incentives to use accounting techniques which shift net income from the present period to future periods. This hypothesis is related to the suggestion that managers have incentives to apply discretion over reported figures to avoid political attention. For example, if a firm earns very high profits, it could be noticed by consumers and media, which, consequently may lead to political pressure. This could lead to politicians reacting with new regulations (e.g. increased taxes), which are unfavourable to the firm.

The nature of the relation between managerial incentives and risk management is complex. Firstly, differences can be observed in the chosen method for risk management. Some literature mainly focuses on studies that examine only one instrument used by managers to manage earnings (for example Healy 1985; Dechow & Sloan 1991).However, recent research (e.g. Roychowdhury, 2006; Zang, 2012) documents that managers may use more than one smoothing method to achieve a certain level of earnings. Since there are several techniques to achieve the same result it is likely that managers use several methods simultaneously. Secondly, a difference in the type of earnings smoothing can be seen. Kamin and Ronen (1978) state that a distinction can be made between “real smoothing”2 and “accounting smoothing”. Choi et al. (2011) define real smoothing as a smoothing method “through contractual transactions such as derivatives” and accounting smoothing as a method “through accounting techniques such as discretionary accruals to shift revenues and costs from one period to another”.

2.2 Defining accrual management and hedging/derivatives

Hedging

Paragraph 9 of IAS 39 provides the following hedging instrument definition:

“A designated derivative or (for a hedge of the risk change in foreign currency rates change only) a designated non-derivative financial asset or non-derivative financial liability whose fair value or cash flows are expected to offset changes in fair values or cash flows of a designated hedged item…” (IAS 39:9; Deloitte, n.d.)

2_{Hedging with derivatives is one of the possible techniques under “real” smoothing methods}

(12)

12

Even though reducing investment risk and thereby reducing earnings volatility is the primary objective of hedging (Fok, Carroll & Chiou, 1997), several studies have documented incentives as to why a manager may want to engage in hedging activities. For example, a large firm tends to use derivative hedging to look sophisticated, by which it can distinguish itself from its competitors. In addition, a hedging firm is likely to be more profitable, because it has more financial possibilities. Moreover, a firm that is high leveraged has incentives to use hedging derivatives to decrease its debt financing costs (Barton, 2001).

According to Graham (2002) firms also have two tax incentives for hedging, which are (1) to increase debt capacity and interest tax deductions, and (2) to reduce expected tax liability if the tax function is convex. Graham (2002) finds evidence for the first incentive, namely that use of hedging increases debt capacity, but does not find support for the second (firms do not hedge regarding tax convexity). The latter is controlled for by studies conducted by Barton (2001) and Choi et al. (2011). In both studies, no significant relation is found between the magnitude in use of derivatives or discretionary accruals and tax convexity.

The manager’s portfolio construction is considered to be an important determinant in hedging. For example, Smith and Stulz (1985) examine the relation between management’s risk aversion and the design of compensation contracts proposed by shareholders. The study implies that managers can be stimulated in risk neutral, risk seeking and risk averse behavior depending on how management’s contract design is structured and on how shareholders can use these compensation contracts to monitor managers’ incentives.

Nan (2008) suggests that a firm’s hedging activities can support reducing risk and attenuating earnings management. The activities depend on whether the decision to hedge is contradictable or not. This paper further outlines that a strategy where hedging is encouraged would need an appropriate compensation in which the appeal for earnings management should be mitigated. This motivated strategy, however, may not be sufficient. In contrast, with a strategy where hedging is demotivated but tolerated, earnings management could be efficient in some situations.

The trade-off between smoothing earnings with abnormal returns and derivative hedging to achieve a desired level of earnings will be discussed more thoroughly in section 2.3.

(13)

13

Accrual management

The IASB requires firms to prepare an annual report on the basis of Accrual Accounting (Deloitte, n.d.). The International Accounting Standard Committee (IASC) adopted the accrual accounting definition of the Financial Accounting Standards Board (FASB), which is:

“Accrual accounting uses Accrual, deferral, and allocation procedures whose goal is to relate revenues, expenses, gains, and losses to periods to reflect an entity's performance during a period instead of merely listing its cash receipts and outlays.”

Accrual is the difference between net income and the cash flow from operations (Scott, 2012). In other words, if an expense has been incurred in a specific period and where at the end of that period no invoice has been received, it is an “accrual”. Many previous studies on accrual management use discretionary accruals to proxy accrual management. A discretionary accrual is the difference between total accruals and nondiscretionary accruals with nondiscretionary accruals being normal accruals, i.e. the result from normal business activities (Hribar & Collins, 2002). The discretionary accruals, the other of the two constituents, are thus abnormal accruals, arising from business activities which are ‘abnormal’.

Accrual management can be explained as achieving a desired level of earnings by means of accounting choices made to achieve that desired level of earnings (Elliot & Elliot, 2008). Discretionary accrual management as referred to by Choi et al. (2011), can be seen as an accounting smoothing method in which revenues and expenses are matched to the period in which they incur (instead of cash inflow and outflow being demonstrated). The objective of accruals is to facilitate a better understanding of a firm’s performance and financial position. However, previous studies show that accruals are also used to manipulate earnings (Healy & Wahlen, 1999; Defond & Park, 1997), to meet or beat analysts’ forecasts (Robb, 1998; Moehrle, 2002; Kaznik & Mcnichols, 2001; Bartov, 2002), and loan loss provisioning (Beatty et al., 1995; Collins & Hribar. 1995).

2.3 Empirical literature

Zang (2012) implies that it is hard to distinguish between accounting smoothing and real smoothing and believes/implies that managers always prefer a combination of the two methods to manage earnings. This view is convincing because it would be notable if only one single method is used to smooth earnings. Additionally, Graham et al. (2005) share the same

(14)

14

opinion as Zang (2012), but see it from a different perspective. Graham et al. (2005) suggest that the accounting smoothing method is easier in a principle based environment than under a rule based environment, up to a certain level. Besides that, if a company only uses real activities to achieve a desired level of income it would not create value in the long run. As a result, it is important to use both accounting and real smoothing methods to attain a desired level of income (Graham et al., 2005). Zang (2012), however, implies that both the principle based and rule based environment are not decisive for the choice made between smoothing methods, but rather that the choice depends more on respectively the time, the costs and the industry wherein a company is established.

Zang (2012) argues that, firstly, time is important, because real smoothing activities can solely be used during a specific balance sheet date and accounting smoothing is also possible after balance sheet date. One of the possibilities is that managers use accrual management to compensate for the ineffective part of real activities (e.g. derivatives). Secondly, the costs are of importance, because the manager will trade-off the two methods based on their costs. Real smoothing activities entails additional costs. For example, hedging is an expensive activity, so managers will try to keep these costs as small as possible.

Finally, according to Zang (2012), the industry in which the firm is established is of interest. If a firm is established in a highly competitive industry, managers of those firms will rather engage in accounting smoothing than real activities, because they will try not to endanger the businessin a highly competitive industry. All in all, Zang’s study implies that the use of real activities is determined on the basis of different costs, leading to a substitutive relationship between real activities and accrual management. The use of accrual management increases when less real activities are used and, vice versa, its use decreases if real activities are used intensively. This view is strengthened by the study Roychowdhury (2006), in which it is suggested that earnings smoothing is not possible with the use of real activities or only accrual management only.

Barton (2001) examines the use of hedging and accrual management as an objective to maintain a desired level of earnings volatility. More specifically, he examines whether the use of derivatives and discretionary accruals are substitutes for the smoothing of earnings. The author uses the notional amount of derivatives scaled by lagged total assets to measure derivatives and deploys a proxy for measuring accrual management based on the Modified Jones model (Jones, 1991; Dechow, Sloan & Sweeney, 1995).

Barton (2001) finds a significant negative relation between derivatives and discretionary accruals, providing empirical evidence that managers indeed view derivatives

(15)

15

and accrual management as substitute tools for smoothing earnings. Finally, he hypothesizes on SFAS No. 133, which is, as mentioned in 1,3, similar to IFRS 9 and IAS 39. Barton suggests that the implementation of SFAS No. 133 may increase the costs of using derivatives as an earning-smoothing mechanism compared to earnings management. As a result, he expects the implementation of SFAS No. 133 to lead to a decrease in hedging and an increase in earnings management. Moreover, Barton (2001) suggests that larger firms are more likely to hedge than smaller firms, but does not find support for influence of the firm’s size on the magnitude. In other words, a firm’s size influences the decision to hedge, but does need lead to more hedging. This finding is ratified by the study of Choi et al. (2011). These authors conclude the same findings about firm size and leverage for the pre and post-SFAS No. 133 period. While Barton’s findings provide important insights into and implications as to how managers choose to smooth earnings, a crucial shortcoming of his research is the lack of theory, that he also acknowledges himself: “(…) the literature provides little guidance for specifying and identifying a model that explains the tradeoff between hedging and earnings management.” (Barton, 2001, p. 24).

In a similar study, Pincus and Rajgopal (2002) look at whether managers deploy abnormal accrual choices and hedging with derivatives as substitute tools for managing earnings volatility in the oil and gas industry. The authors provide the same findings as Barton (2001), namely that managers perceive the extent of hedging and smoothing with discretionary accruals as substitute mechanisms for smoothing earnings and dampening earnings volatility. Unlike Barton, Pincus and Rajgopal (2002) find evidence for a sequential process (mainly in the fourth quarter), implying that managers of firms will first determine how much they want to hedge without considering the use of discretionary accruals and than, especially in the fourth quarter, trade off derivative hedging and discretionary accruals to attenuate the remaining earnings volatility.

Where the two previous studies examined a period where there was an absence of uniformity regarding derivative accounting, Choi et al. (2011) investigate whether SFAS No. 133 influenced the advantages of the use of hedging with derivatives and accrual management to smooth earnings and re-examines the substitution hypothesis that has been introduced by Barton (2001) based on a sample of S&P 500 firms in the pre- and post-SFAS No. 133 periods. Choi et al. (2011) provide the same findings as Barton (2001) and Pincus & Rajgopal (2002) for the pre-SFAS No. 133 period, namely that during this period there is a (partly) substitutive relation between hedging with derivatives and accrual management. However, in the post-SFAS No. 133 period this (substitutive) relation is weakened, or to put it differently,

(16)

16

is replaced by a complementary relation after the implementation of SFAS No.133. This evidence indicates that the relation between hedging with derivatives and accrual management has been impacted by the new implemented hedge accounting rules (SFAS No. 133). Moreover, they find that in the post-SFAS No. 133 era, earnings volatility significantly increased which is associated with the use of hedging. This finding implies that the use of hedging is not as good as it was in smoothing earnings or an instrument to smooth earnings in the post-SFAS No. 133 period (Choi et al., 2011).

2.4 Hedge accounting

On November 19, 2013, a new version of IFRS 9 Financial instruments was issued by the IASB (IASPlus, n.d.). IFRS 9 represents a major landmark in IASB’s project to reform and simplify the accounting of financial instruments (i.e. hedge accounting). IFRS 9 introduces the new hedge accounting requirements and replaces IAS 39 “Financial Instruments: Recognition and Measurement”.

The new standard consist of three parts. Part 1 addresses the classification and measurement of financial assets. Part 2 covers Loan Loss Provisioning and part 3 deals with general Hedge Accounting. The IASB decided not to take within the scope of general hedge accounting into account in IFRS 9 yet. The IASB has decided to define Macro Hedge Accounting as a separate (sub)project. Initial proposals about Macro Hedge Accounting still need publication. Therefore, in this thesis, only general hedge accounting is applicable.

IFRS 9 is mandatory from fiscal years beginning after the 1st_{of January, 2018. It}

should be noted, however, that IFRS 9 is not yet endorsed for use in the EU, but it is highly probable that it will be approved.

In a rapport published by IFRS (2014) it is suggested that under the under IFRS 9, hedge accounting and risk management will be more aligned. This will enhance the understandability of a firms risk management activities. IFRS implies that transforming from rule-based under IAS 39, to principle based under IFRS 9. One of the benefits if this shift from rule-based to principle based, is that component of non-financials item can be hedged as well. IFRS (2014) explains that IFRS 9 does not discriminate between types, but rather looks to the possibility to identify the risk component. Furthermore, IFRS (2014) suggests that the implementation cost will reduce, because IFRS 9 does not require the same amount of analysis as that is compulsory under IAS 39. This indicates that hedging will be less costly under IFRS 9, and may lead to more hedging. According to (Barton, 2001; Zang, 2012), the cost of the smoothing tool can be decisive as to whether to use it or not.

(17)

(18)

18 2.5 Hypotheses development

According to Fok et al. (1997) hedging’s main objective is to reduce earnings volatility, through reducing investment risk. However, previous empirical hedging studies have provided several incentives for firms to hedge and documented specific hedging and non-hedging firms characteristics and suggest that specific firm characteristics may affect the decision to hedge or not to hedge (please refer to the previous paragraphs in this chapter).

In addition, earlier research implied that leverage and firm size might be of influence on the decision to hedge. Large firm likely have lower hedging costs because of the essential means to use these financial instruments (Sami & Welch, 1992). While Barton (2001) suggests that larger firms are more likely to hedge than smaller firms, he does not find support for the influence of firm size on the magnitude of hedging. In other words, firms size influences the decision to hedge, but does need lead to more hedging (Barton, 2001).

This study tries to investigate in what way intense hedgers vary from non-intense hedgers in terms of firm characteristics and try to assess how this might be of influence. Therefore my first hypothesis is:

Building on the aforementioned, evidence is found that hedging and accrual management are (partial) substitutes regarding the smoothing of earnings in the pre-SFAS No. 133 period (e.g. Barton 2001; Pincus & Rajgopal, 2002; Choi et al., 2011). The articles in which this effect was investigated, used US based samples. However, no studies conducted with a European based sample could be found. Therefore, I would like to re-examine the substitute hypothesis and test if this is still the case in another continent (non US) by trying to collect evidence from the European market (e.g. IFRS 9, IAS 39).

The reasoning behind this substitution hypothesis is explained by Barton (2001), where he implies that meaning that if the use of derivative hedging is to become more effective or less costly than accrual management, managers would be incentivized to use hedging instead of accrual management to achieve a desired level of earnings and vice versa. Therefore the second hypothesis is:

H1: Firms that use hedging intensively are likely to display significantly different financial attributes compared to firms that do not use hedging intensively

(19)

19

As discussed in the section above, derivative use and accrual management are partial substitutes for smoothing earnings in the pre-SFAS No. 133 period (e.g. Barton 2001; Pincus & Rajgopal 2002; Choi et al., 2011) and have a complementary relation in the post-SFAS No. 133 period (Choi et al., 2011). Based on Choi et al.’s research, I expect that the use of hedging with derivatives under IAS 39 leads to more use of accrual management as a means of smoothing earnings.

My second hypothesis is an alternative hypothesis to hypothesis 1, because both are mutually exclusive. There are different possible outcomes for the relation between derivative hedging and accrual management, where four possible outcomes will be exemplified in table 1 below.

Table 1 Possible outcomes Dependent

variable

|TACC| Derivatives Relation Tested in Association 1 Negative Negative Substitutes H2a Association 2 Negative Positive |TACC| complements

Derivatives

H2b

Association 3 Positive Positive Complements Not hypothsized Association 4 Positive Negative Derivatives complements

|TACC|

Not hypothsized

As mention in section 2.3, Barton. Pincus and Rajgopal (2002) and Choi et al. (2011) have found evidence for the substitution hypothesis, where hedging with derivatives and accrual management are both negatively associated with on another. However, Choi et al. (2011) argues that in the post SFAS No. 133 period the relation between the two smoothing tools is

H2a: Managers of European firms use hedging with derivatives and discretionary accruals as substitute mechanisms to manage earnings volatility.

H2b: Ceteris paribus, hedging with derivatives is positively associated with the use of accrual management

(20)

20

attenuated and became have a complementary relationship, where only one complements the other. Although, there is no empirical support, in theory derivative hedging and accrual management can be complete complements. This would be the case if they are both positively associated with each other, hence this means that if a firm uses more hedging the firm would manage more accruals as well and vice versa.

In general, under SFAS No. 133, derivatives (both assets and liabilities) must be reported at fair value on the balance sheet3, while gains and losses resulting from changes in this fair value, should be directly reported as earnings. Consequently, SFAS No. 133 significantly increased the volatility of earnings (which is associated with the use of hedging) and, at the same time, made hedging a less effective tool for managers to smooth earnings (Choi et al., 2011).

As SFAS 133 and IAS 394 are two very similar standards introduced to ensure companies correctly recognise and report on the value of derivatives, the third prediction is that the use of hedging under IAS 39 would as well be (positively) associated with earnings volatility. Therefore the last prediction is.

3 Unless the derivatives meet specific hedge accounting rules. For a specification of these rules, please refer to IAS 39.

4FAS 133 (“Accounting for Derivative Instruments and Hedging Activities) is part of Generally Accepted Accounting Practice in

the US, while IAS 39 is part of International Financial Reporting Standards (IFRS)_.

H3: Ceteris paribus, the use of hedging with derivatives is positively associated with earnings volatility

(21)

21

3. Research Methodology

3.1 Sample

An archival quantitative data study will be conducted by using a population of the companies listed on the Euronext 100.5 The selection criteria are as follows:

• Firstly, I identify all Euronext 100 firms for the research period (the fiscal year from 31 December 2011 through 31 December 2012), during which the firm discloses the fair value of its derivative holdings;

• Secondly, I check whether the companies are listed on the Compustat global annual database and whether the statement of cash flows data reported under IAS 39 is available for calculating accruals (Collins & Hribar, 2002).6

• Finally, the sample consists of nonfinancial nonregulated firms that belong to the Euronext 100 index and firms with available.

Following earlier reviewed literature (Burgstahler & Dichev, 1997; Barton, 2001; Zang, 2012), regulated industries (SIC 4400 - 4999) and financial institutions (SIC 6000 – 6999) are excluded from the sample, because, first, these firms are governed by specific regulation and, second, because of the unique operating characteristics of these firms and not always comparable with other industries (Achleitner, Guenther, Kaserer, & Siciliano, 2014).

My initial sample obtained from the Datastream database included 202 firm-year observations. I excluded regulated, financial firms and firms with incomplete data. Of the remaining firms, I hand-collected all fair value derivative holdings (when available) from financial statement footnotes to classify the usage of derivatives for the sample firms. To keep the collection of data practicable, I collected derivative holdings for the following research period, the fiscal year from 31 December 2011 through 31 December 2012.

The variables in the dataset are constructed using information from three sources. Information on financial statement required for the empirical investigation was collected from the company.info annual report database and company websites. All the other data are from the Compustat Global and Datastream database.7

5_{Almost all data in this research is from Compustat. If other data is used, this will be mentioned.} 6_{For a fiscal year ending after July 1988, SFAS No. 95 requires to present a statement of cash flows.} 7_{See Appendix A for a complete overview}

(22)

22

Table 2 presents the sample selection procedure and sample selection criteria used to determine the final sample. First financial firms are excluded from the sample. Second, regulated firms are removed. Third, Firm that with missing data are removed from the sample. Finally, firms that are not reporting under IFRS are excluded from the sample. This results in a final sample size of 62 firms and 124 firm-year observations.

Table 2 SAMPLE SELECTION

Sample Selection Procedure Number of firm

observations Total number of firms listed on the Euronext as of January 1, 2012 101

Delisting financial firms (-/-) 21

Delisting regulated firms (-/-) 16

Delisting non-IFRS firms (-/-) 1

Less: Firms with missing observations (-/-) 1

Final Sample Size 62

Results in 124 firm-year observations

3.2 Empirical model and variable measures

Accrual management measurement

Following prior literature, to proxy for accrual management the cross-sectional model of discretionary accruals will be used. For every industry (classified by its two-digit SIC) the annual model is estimated to capture discretionary accruals, being my proxy for accrual management. As mentioned in 2.2,, discretionary accruals are the difference between firms actual accruals and the normal level of accruals. I estimate these by using the following modified cross-sectional Jones model by Dechow et al. (1995)presented below:

Accrualsi,t / Assetsi,t-1 =

α

*(1/ Assetsi,t-1) +

β

1*[(ΔRevi,t - ΔRevi,t / Assetsi,t-1]

(23)

23

Hypothesis 1

In section 2.3.2. of this study, hypothesis 1 predicts that “Firms that use hedging intensively are likely to display significantly different financial attributes compared to firms that do not use hedging intensively”. This hypothesis is tested by a PROBIT regression on my entire sample of 124 firm-years observations to estimate equation (2), where a dummy variable is coded 1 if firm i holds a derivative position and coded 0 otherwise. I assess whether intense hedging firms and firms that are not have different firm characteristics by the following equation:

HEDGER =

ϒ

0 +

ϒ

‘CONTROLSit +

ε

it (2)

Unlike Barton (2001) and Choi et al. (2011), I will not calculate the inverse MILLS ratio, because I do not have to correct for potential self-selection bias in the simultaneous equations (2) and (3). All companies that can be considered a HEDGER, disclose the fair value amount.8

8_{Under SFAS No. 133, companies have a choice in disclosing the notional value or the fair value of}

their financial derivative

position, therefore there can be self-selection bias.

Where, for firm i and at fiscal year t, TACC represents total accruals defined as:

TACC = Total accruals for i and at fiscal year t, measured as EXBI i,t - CFO i,t, where EXBI is the

earnings before extraordinary items and discontinued operations (annual Compustat Global item “IBC”) and CFO is the operating cash flows (from operating activities) taken from the statement of cash flows (annual Compustat Global item “OANCF”);

Assets = total assets in year t–1

(annual Compustat Global item “AT”);

ΔRev = change in revenue from year t−1 to year t (REVt− REVt−1)

(annual Compustat Global item “SALE”);

ΔRec = change in net accounts receivables from year t-1 to year t (RECt− RECt−1)

(annual Compustat Global item “RECT”); PPE = gross property, plant and equipment in year t

(annual Compustat Global item “AT”);

(24)

24

Hypothesis 2

Following Barton (2001) and Choi et al. (2011), I test hypotheses 2 and 3, where DERIVATIVES and |TACC| are endogenously determined and the two-stage least square (2SLS)9 estimation procedure is used to test the following simultaneous equations:

DERIVATIVESit =

α

0 +

α

1|TACC|it +

α

hdgCONTROLSit +

ε

it (3)

and

|TACC|it =

_β

0 +

β

1DERIVATIVESit +

_β

tacc‘CONTROLSit +

ε

it (4)

where, CONTROLS is compiled/consist of multiple control variables and accounts for the various incentives of managers to retain a preferred level of earnings volatility, by using derivative hedging and accrual management. The control variables in this study are based on Barton (2001) and Choi et al. (2011). However, some control variables are excluded due to data limitations.10

Control Variables11

Explanatory/Control variables common to equations (2), (3) and (4): Leverage = Total debt scaled by total assets

R&D = R&D expenses scaled by total assets

R&D*Leverage = RD multiplied by leverage, both scaled by total assets Distress = Altman z-score to calculate the probability of bankruptcy Foreign sales ratio = Foreign sales scaled by total sales

Explanatory/Control variables common to HEDGER (2) and DERIVATIVES (3) equations: Size = Ln of total assets

Dividend yield = Common (cash) dividends scaled by market value of equity Short-term debt = Short-term debt scaled by total debt

Cash cycle = Cash conversion cycle

9_{Dependent variables will be tested on endogeneity (hausman test), if rejected an OLS regression will}

be used.

10_{Analysts and lifo reserve, tax convexity and CEO compensation controls.} 11_{Table x gives a complete overview of all the control variables}

(25)

25

Explanatory/Control variables common to |TACC| (4)equation: Dividend payout = Common (cash) dividend scaled by EBIT Flexibility = Root mean squared of |TACC| [equation (1)]

|OCF| = Absolute value of the ratio of operating cash flow scaled by total assets

Hypothesis 3

I follow prior literature to measure earnings volatility (Dechow & Dichev, 2002; Choi et al., 2011). Earning volatility is defined as “the standard deviation of Net Earnings before Extraordinary Items” (Compustat global item “IBC”), from years t-5 to t. According to Choi et al. (2011), earnings volatility is estimated as a function of DERIVATIVES, |TACC| and firm specific control variables. Many of these control variables are also used in the other equations and this test as well.

The following variables Firm size, ROA, BM-ratio, Capital intensity, Leverage, Dividend payout ratio, Foreign sales ratio and control for year effects based.

Where the specific variables for this test are specified below: BM-ratio = Book value of equity – Market value of equity Capital intensity = Capital expenditures / Total Assets

ROA = EBIT12 / Total Assets

12_{Choi et al. (2011) uses Data item 172 (Net income), this data item is not available in the compustat}

(26)

26

4. Results

4.1 Descriptive statistics

The summary statistics of the full sample of 124 firms-year observation is reported in Panel A of table 3. The variables in the dataset are all winsorized at the 1st and 99th percentile. Table 3 shows the relevant descriptive statistics of the variables used in the regression analysis and shows the median, mean and standard deviation for each variable. The median13 is the middle score when scores are ranked, the mean is the average score and the standard deviation is a measurement of the average spread.

120 out of the 124 firm-year observation (60 out of the 62 firms) in the sample make use of derivative hedging, this is 96,8 %, which is a huge part of the sample. The sample firms have on average an amount of € 498 million Total derivatives and an average Net amount of € 28.1 million derivatives and for the same variables a median of € 128 million for Total derivatives and € -0.25 for the Net derivatives. This would mean that more than 50% of our full sample has more liability than asset derivatives. The control/exogenous variables of the sample have small standard deviations (except for CashCycle).

Foreign sales is a considerable part of the firm total sales mean (mean = 0.68) and (median 0.741), firms are substantially leveraged (mean leverage = .229) and (median leverage = .219) and pay out a substantial part of there earnings before income and taxes (mean dividend pay-out ratio = .237) and (median dividend pay-out ratio = .22).

Table 2 Full sample descriptive statistics

13_{I have an even number of observations, this would mean that the median is the average of the two}

(27)

27

Variables count Mean Median Std. Dev. Skewness Kurtosis

HEDGER 124 0.500 0.500 0.502 949.810 338.405 0.029 0.008 0.023 0.018 0.242 0.030 0.115 0.027 0.006 1.101 0.206 108.877 0.045 0.034 1.188 0.143 2.418 0.067 0.038 0.000 1.000 Total derivatives 124 498.059 127.500 3.031 11.762 Net derivatives 124 28.813 -0.250 3.687 24.868 TOTAL_FVderivative_ratio 124 0.016 0.009 6.031 43.226 NET_FVderivative_ratio 124 -0.000 -0.000 -0.326 5.695 |TACC| 124 0.024 0.018 1.888 6.802 Dividend Yield 124 0.030 0.027 1.079 4.186

Foreign Sales ratio 124 0.680 0.741 -1.149 3.882

Earnings volatility 124 0.025 0.016 3.990 22.882 Leverage 124 0.229 0.219 0.294 3.302 RD 124 0.018 0.003 1.991 6.648 RD*Leverage 84 0.005 0.003 1.791 5.539 Size 124 9.657 9.668 0.071 2.223 Short-term debt 124 0.259 0.196 1.498 4.677 Cash Cycle 124 114.569 76.848 1.444 5.252 |OCF| 124 0.085 0.079 0.764 4.051 Capital Intensity 124 0.041 0.033 2.340 10.350 Distress 124 2.197 2.016 1.434 5.359 Dividend Payout 124 0.237 0.220 1.641 7.376 BM ratio 124 2.416 1.764 3.788 21.634 Flexibility 124 0.140 0.132 0.838 3.432 ROA 124 0.078 0.073 0.349 2.697 N 124 4.2 Results hypothesis H1 Mean test

The Skewness test for dependent variables did not reject the null hypothesis. This indicates that the dependent varibales are not normally distributed. Therefore, a Wilcoxon rank-sum test is used (which is a non-parametric test). A T-test (which is a parametric test) is also used to determine whether the means of the non-intensive hedgers differ from the intense hedgers.

The results of both the Wilcoxon rank-sum test and the t-test show that both intensive hedgers and those that are not differ in Leverage and Size at the 5% significance level. The Cash Cycle is also significant at the 5% level according to the T-test and at the 10% significance level according to the Wilcoxon rank-sum test. These results strengthen the predictions for hypothesis 1. The result are presented in the table below.

(28)

28

Probit regression

A PROBIT regression was employed to examine the characteristics of intensive and non-intensive hedgers. To determine whether non-intense hedging firms and intense-hedging firms have a different relationship with firm characteristics, 9 variables were used. These are Leverage, R&D intensity, R&D*leverage, Distress, foreign-sales ratio, Firm size, Dividend yield, Short-term debt, and Cash-Conversion Cycle. As mentioned previously, a dummy variable is used for the dependent variable. Intense hedgers are coded 1, and non-intense hedgers are coded 0. This method provides insight into the conditional relationships between intensity of hedging and firm characteristics.

Panel B of table 4 reports the estimation results of equation (2) from the probit model. The likelihood ratio chi-square of 26.038 and the p-value of 0.004 indicates that the model that is used is statistically significant as a whole. In other words, it fits significantly better than a model with no predictors. RD*leverage excepted, the results support predictions based on previous research. Leverage (Beta of 4.913) and Size (Beta 0.403) are both positive significant at the 1% level. RD significant positive (Beta 24.588) and RD*leverage significant negative (Beta -109.817) at the 5% level. Leverage is significantly positive, this means that the more leveraged a firm is, the more likely it is to be an intensive hedger. The same is true for Size. The positive significant relationship indicates that the larger a firm is, the more likely it is to be an intensive hedger. The results for Leverage and Size are as predicted, and they support Hypothesis 1.

Although not directly hypothesized, RD, RD*leverage and Cash Cycle contribute to our understanding of hedging firms. RD*leverage is negatively significant. This means that the higher this ratio is, the more likely the firm is to be a non-intense hedger. This was not expected. Prior research suggested that RD*leverage is positively related to the decision to hedge (Barton, 2001; Choi et al, 2011); but the results of this study imply that this is true for the non-intense hedgers. Overall, these results are consistent with H1 and support the hypothesis that intense hedging and non-intense hedging firms are likely to display significantly different financial attributes.

(29)

29

Table 4, Wilcoxon rank-sum test, T-test and Probit model results

Panel A Wilcoxon rank-sum test and T-test

Non-intense hedger Intense hedger Wilcoxon

rank-sum test T-test

Variables Mean Sd Median Mean Sd Median Statistic P T-value P(t-value)

TOTAL ratio 0.004 0.003 0.004 0.028 .038 0.019 -9.605 0.000 -4.820 0.000 NET ratio -0.000 0.003 0.000 -0.001 .011 -0.000 0.405 0.686 0.398 0.692 |TACC| 0.025 0.020 0.018 0.023 .026 0.017 1.339 0.180 0.407 0.684 Foreign Sales 0.685 0.229 0.761 0.676 .257 0.736 -0.005 0.996 0.216 0.830 Leverage 0.208 0.095 0.213 0.250 .129 0.223 -1.726 0.078 -2.070 0.041 RD 0.018 0.030 0.003 0.018 .024 0.007 -1.184 0.263 -0.163 0.871 RD*Leverage 0.003 0.006 0.001 0.003 .005 0.001 1.042 0.297 0.010 0.992 Size 9.387 1.076 9.290 9.928 .066 10.07 -2.734 0.006 -2.812 0.006 Dividend yield 0.030 0.018 0.027 0.030 .018 0.028 -0.010 0.992 -0,076 0.939 St Debt 0.249 0.196 0.196 0.270 .217 0.196 -0.315 0.735 -0.564 0.574 Cash Cycle 95.31 88.21 72.70 133.8 24.0 102.0 -1.659 0.097 -1.993 0.048 |OCF| 0.084 0.043 0.078 0.086 .047 0.080 -0.247 0.805 -0.309 0.758 Distress 2.334 1.249 2.064 2.059 .117 1.832 1.472 0.141 1.295 0.198 Dividend Payout 0.229 0.153 0.218 0.245 .132 0.220 -0.912 0.362 -0.611 0.542 Earn volatility 0.021 0.021 0.013 0.028 .037 0.017 -1.654 0.098 -1.219 0.225 Flexibility 0.145 0.062 0.133 0.134 .072 0.129 1.339 0.180 0.923 0.358

Panel B Probit model results

Dependent variable HEDGER (2)

Expected sign Std. error Coefficient

Leverage + (1.672) 4.913*** RD ? (9.937) 24.588** RD*Leverage ? (48.291) -109.82** Distress ? (0.161) 0.170 Foreign Sales ? (0.602) -0.491 Size + (0.132) 0.403*** Dividend yield ? (7.205) -1.062 Short-term Debt ? (0.728) 0.000 Cash Cycle + (0.001) 0.003** Year 2012 (0.246) -0.396 Constant (1.640) -5.263*** NOBS 124 pseudo R2 _0.151 Chi-statistic 26.038 P(Chi2-statistic) 0.004

Regression type Ptobit model

(30)

30 4.3 Results hypothesis H2

Table 5 presents the results of the simultaneous equations (3) and (4), which exhibits two alternate measures for derivative use in its (1) Net derivative ratio and its (2) Total derivative ratio. H2 tested the relation between hedging with derivatives and accrual management by re-examining the substitution hypothesis raised by Barton (2001). The first measurement (1) indicates that |TACC| and Net derivative ratio are positively associated with each other at the 5% significance level. This means that the use of derivatives leads both to more accrual management and vice-versa. This is completely the opposite of the prediction of H2a, which predicts that |TACC| and DERIVATIVES are both negatively associated. It therefore offers no support for a substitution relationship. H2b hypothesizes that the use of derivative hedging leads to more accrual management but not that accrual management leads to more use of derivatives. Therefore, this hypothesize is partly supported. This relationship supports the view that both accrual management and hedging (i.e., real activities) are used to manage earnings (Zang, 2012), but no support is found for the hypothesis that they encourage each other’s use (i.e. complete complements). Barton (2001), Pincus & Rajgopal (2002) and Choi et al. (2001) found a substitution relation in the pre-SFAS No. 133 period, and Choi et al. suggest a complementary relationship. The second measurement (2) implies that Total derivatives are positively associated with |TACC|. Total derivatives is positive and significant at the 1% level. This indicates that the use of derivatives leads to the use of accrual management.

The null hypothesis of no simultaneity cannot be rejected. This applies to all equations; therefore, the model offers poor explanatory power for the dependent variables. The variation of the dependent variables is poorly explained by the independent variables. This applies to all equations. This means that the model is not reliable; therefore, the results are not reliable either. Although endogeneity is proven and taken into account, there may be other problems with this sample of data that need to be taken into account, as almost no control variables is significant and as the P-value of the F statistic is also not significant.

(31)

31

Table 5: Simultaneously equations (3) and (4)

Measurement (1) Measurement (2)

VARIABLES Dependent variable

Expected sign

Net Derivative ratio

|TACC| Total Derivative ratio

|TACC|

NET Derivative ratio +/- 11.519**

5.700

TOTAL Derivative ratio +/- 1.927**

0.631 |TACC| +/- 0.066** 0.322 (0.033) (0.277) Leverage + -0.010 0.131 0.000 0.016 (0.010) 0.118 (0.023) 0.064 RD + -0.092 1.443 0.008 0.085 (0.064) 0.952 (0.193) 0.427 RD*Leverage + 0.309 -4.921 0.053 -0.258 (0.305) 4.042 (0.875) 2.056 Distress + 0.000 -0.007 0.002 0.003 (0.001) 0.010 (0.002) 0.005 Foreign Sales + 0.001 -0.021 0.021 -0.057 (0.004) 0.039 (0.019) 0.026 Size + 0.000 0.005*** (0.001) (0.002) Dividend Yield + 0.096** -0.058 (0.044) (0.080) Short-term Debt + 0.006 -0.016 (0.005) (0.011) Cash Cycle + -0.000 0.000 (0.000) (0.000) Dividend Payout - 0.001 -0.001 0.006 0.021 Flexibility + 0.016 0.093** 0.011 0.040 |0CF| + 0.029 -0.160** 0.022 0.079 _cons 0.023 0.017 0.053 0.032 Year 2012 ? 0.001 -0.010 -0.004 0.009 (0.001) 0.018 (0.005) 0.010 Constant -0.009 -0.053* (0.010) (0.029)

Heteroscedasticity accepted? No Yes

Corrected for heteroscedastic errors No Robust Standard Errors NOBS 120 120 120 120 R2 0.128 0.073 0.107 0.0972 adj. R2 _0.0392 _-0.0026 _0.0161 _0.0233 F-statistic 1.4419 0.97 1.0300 1.32 p(F) 0.1648 0.4725 0.4254 0.02368

Hausman Rejected 0.000 Rejected 0.000

Regression type OLS 2SLS OLS 2SLS

*, ** and *** denote statistical significance at the 10%, 5% and 1% level, respectively

(32)

32 4.4 Results hypothesis H3

Hypothesis 3 predicts that the use of hedging is positively associated with earnings volatility. In other words, it predicts that the use of derivative hedging leads to more volatile earnings. Table 6 presents the results of the OLS regression. Endogeneity was rejected (P value .222). This was tested with the Hausman test.

Consistent with hypothesis 3, the results in table 7 indicate that the use of hedging is positively associated with earnings volatility (at the 1% significance level). This is line with Choi et al. (2011). The difference is that I find support for the whole sample, whereas Choi et al. (2011) find support only for the firms that are involved in low accrual management. Although not hypothesized, table 7 also indicates that the use of accrual management is positively (at the 10% significance level) associated with earnings volatility. This is in line with the results of H2, in which both are positively significantly associated with each other.

Furthermore, the table shows that firm size (at the 5% significance level) and ROA (at the 1% significance level) are both negatively associated with earnings volatility. This can be explained by supposing that the larger firms are less volatile. Foreign Sales is also positively associated (at the 5% significance level). ROA is negatively associated with earnings volatility, which indicates that more profitable firms have less volatile earnings. The results for SIZE and ROA meet the expected prediction and are consistent with prior research (Choi et al. 2011). The R-squared (0.243) is acceptable and indicates that the independent variables explain 24.3% of the proportion of variance in the dependent variable. This indicates the overall strength between the independent and the dependent variables. However, the P-value of the F statistis (0.2902) is not significant and indicates that the model is not reliable.

(33)

33

Table 6 Earnings volatility

Dependent variable Earnings volatility

TOTAL derivative ratio NET derivative ratio

VARIABLES Coefficient Std. Err. Coefficient Std. Err.

|TACC| .2887** (6.883) 0.435* (.226) NET derivative 1.231* (.716) Total derivative -1.7256 (11.173) Leverage .0128 (.070) .015 (.024) Size .0021 (.0028) .005** (.002) ROA -.4536 (2.219) -.263*** (.100) BM -.0053 (.015) -.003 (.005) Capital Intensity .2517* (3.332) .002 (.113) Dividend Payout -.0031 (.0219) .002 (.023) Foreign Sales .0789 (.2266) .041** (.017) Year 2012 -.0067 (.0236) -.003 (.006) Constant -.0220 (.2345) .051** (.023)

Heteroscedasticity accepted? Yes Yes

Heterosc. Test: chi2 _133.07 _133.07

Heterosc. Test: P(chi2₎ _0.0000 _0.0000

Corrected for heteroscedastic errors Robust standard errors Robust standard errors NOBS 120 120 R2 _0.307 _0.307 adj. R2 0.2430 0.2430 F-statistic 1.2139 1.2139 p(F) 0.2902 0.2902 Hausman 0.000 Rejected

Regression type 2SLS OLS

(34)

34 4.5 Additional Robustness and sensitivity tests

Robustness test

For this test, a probit model with endogenous regressors is used. This test is actually a combination H1 and H2. Intense hedger is coded 1, and non-intense hedger is coded 0. It assesses the relation between derivative hedging and accrual management. This method is needed because of the endogeneity of the dependent variables and because of the conflicting findings in H2. Remember that equation (2) indicates whether intensive hedgers and non-intensive hedgers have different firm characteristics, whereas equations (3) and (4) examine the relation between derivative hedging and accrual management.

Table 6 reports a likelihood ratio chi-square of 129.86 with a p-value of 0.0000. This shows that the model that is used is statistically significant as a whole. In other words, it fits significantly better than a model with no predictors. It has similar results to H1, in which Leverage, RD and RDlev exhibited a significant association. This implies that intensive hedgers have a higher leverage and are larger in size, which is consistent with H1. Foreign sales ratio is negatively associated with intensive hedgers at the 5% significance level.

Furthermore, table 6 indicates that |TACC| is negatively associated at the 1% significance level with intensive hedgers. It also indicates that intensive hedgers are negatively associated with |TACC| at the 5% level of significance. This implies that derivative hedging is a substitute, but only for the intensive hedgers. This result is the complete opposite of the findings for H2 in section 4.3.

Sensitivity test

Due to missing value for the variable, R&D expense, two sample approaches were used. For the first method, all missing values for R&D expense were replaced by 0. This was done under the assumption that firms that did not disclose their R&D expense did not have any to disclose. This approach ensured that all variables were equal in sample size and therefore comparable without losing firm-year observations. It resulted in 124 firm-year observations. The second method ensures that all restricted variables have no missing values and that all have the same sample size. This resulted in a loss of 16 firm-year observations—even if firms had all values for the necessary variables—and therefore in a smaller sample size. The use of these two approaches made it possible to test the results sensitivity. The results of the first method were presented section 4.3; therefore, only the results of the second method are reported here.

(35)

35

This test does not lead to other results. It has exactly the same results for the total derivatives, Net derivatives and |TACC| variables. This means that the method of measurement (1), in which all missing variables were replaced by 0, did not lead to significant differences between the results.

Overall, H1 is supported, which examined if intensive hedgers and non-intensive hedgers have different firm characteristics. Among other, I find that Leverage and Size are positively associated with intensive hedgers. My analysis in section 4.3, where the results of Hypothesis 2 are presents results for two types of measurement, the Net derivative ratio and the Total derivative ratio. While the Net derivative ratio and |TACC| where both positively associated, which indicates that they have complete complementary relationship, The Total derivative ratio is also positively associated with |TACC|, but |TACC| is not significantly associated with total derivative ratio. This implies that the use of hedging leads to more use of accrual management, but the use of accrual management does not lead to the use of hedging. Choi et al. states that the use of hedging leads to more volatile earnings and therefor, it leads to more accruals management. The results of H3 indicate the accrual management and derivative hedging are both positively associated with earnings volatility, Finally, robustness testing indicates that there is substitution relationship between derivative hedging and accrual management, but only for the intensive hedgers.

(36)

36

Tabel 6: Robustness Test Dependent variable

VARIABLES Expected sign Intensive Hedger |TACC|

HEDGER -0.135** 0.063 |TACC| +/- -42.610*** 4.338 Leverage + 3.288** 0.220** 1.355 0.112 RD + 16.446* 1.212* 8.541 0.690 RD*Leverage + -67.972* -5.291 41.048 3.224 Distress + 0.007 -0.002 0.099 0.006 Foreign Sales + -0.826* -0.019 0.437 0.027 Size + -0.003* 0.002 Dividend yield + -0.024 0.062 Short-term Debt + 0.000 0.006 Cash Cycle + -0.000* 0.000 Dividend Payout 0.441 0.330 Flexibility -0.939 0.676 Abs0CF 1.752 1.305 _cons 0.711 0.061 0.623 0.041 Year_2012 ? 0.048 -0.013 0.207 0.015 Constant NOBS 124 NOBS 124 Log Likelihood 224.7154 R2 _0.1205 Chi-statistic 129.86 adj. R2 0.0511

p(Chi2_-statistic) _0.0000 _F-statistic _0.0887

p(F)

Wald test 0.002

Hausman 0.000

Regression type Probit model with

endogenous regressors

2SLS

(37)

37

Tabel 7: Sensitivity Analysis Dependent variable

VARIABLES Expected sign NetDerivative

ratio

|TACC| Total Derivative

ratio

|TACC|

NET derivative ratio +/- 10.569**

4.644

TOTAL derivative ratio +/- 1.334***

0.500 |TACC| +/- 0.091** 0.296 (0.041) (0.295) Leverage + -0.011 0.095 0.003 0.031 (0.016) 0.150 (0.038) 0.082 RD + -0.108 1.392 -0.141 0.216 (0.079) 0.925 (0.207) 0.411 RD*Leverage + 0.390 -4.282 -0.301 0.028 (0.383) 3.974 (0.928) 2.059 Distress + 0.001 -0.010 -0.002 0.005 (0.001) 0.011 (0.003) 0.006 ForeignSalesratio + 0.001 -0.027 0.043 -0.065 (0.006) 0.048 (0.030) 0.031 Size + 0.001 0.001 (0.001) (0.003) Dividendyield + 0.054 -0.240 (0.068) (0.152) Short-term Debt + 0.006 -0.011 (0.006) (0.013) Cash Cycle + -0.000 0.000 (0.000) (0.000) Dividend Payout - -0.000 0.011 0.010 0.042 Flexibility + 0.023 0.113 0.014 0.058 |0CF| + 0.028 -0.213 0.026 0.106 _cons 0.041 0.015 0.068 0.037 year2012 ? 0.002 -0.021 -0.005 0.010 (0.002) 0.022 (0.009) 0.011 Constant -0.014 -0.007 (0.015) (0.040)

Heteroscedasticity accepted? No Yes

Heterosc. Test: chi2 _0.04 90.59

Heterosc. Test: P(chi2₎

0.8452 0.0000

Corrected for heteroscedastic errors No Yes NOBS 77 77 77 77 R2 _0.168 _0.1384 _0.133 _0.1577 adj. R2 _0.0268 _0.0226 _-0.0132 _0.0445 F-statistic 1.1905 1.20 1.0080 1.39 p(F) 0.3113 0.127 0.4495 0.2089

Hausman Rejected 0.000 Rejected 0.000

Regression type OLS 2SLS OLS 2SLS

The interaction of accruals and hedging : do they affect management decisions? : evidence from the European market