The effects of IFRS 8 : on segment reconciliations and it determinants : a European study

(1)

The effects of IFRS 8: On segment reconciliations and it

determinants. A European study

Name: Joe Kok

Student number: 10208585

Thesis supervisor: drs. J.F. Jullens Date: August 10, 2018

Word Count: 17176

MSc Accountancy & Control, Accountancy

(2)

Statement of Originality

This document is written by student Joe Kok who declares to take full responsibility for the contents of this document.

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

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

(3)

Abstract

In this thesis I investigate the determinants of segmental disclosure. Specifically, I research the disclosure of segment reconciliations under IFRS 8. Segment reconciliations (GAPs) are required when the aggregated segment income number differs from the corresponding profit & loss number. If this occurs a disclosure is required that reconciles the segment number to the P&L number. These reconciliations exist because IFRS 8 offers discretion in the disclosure of the aggregated segment income measure. If the segmented income measure is equal to the profit & loss number there is no reconciliation; the GAP is zero. The segmented income measure can be bigger than the profit & loss number. The GAP is negative. The segmented income measure can be smaller than the profit & loss number. The GAP is positive. GAP research is relevant because their use and interpretation is criticized. There is also no prior European literature on this subject. I investigate what drives the disclosure of a GAP and what determines the sign of a GAP. Whether it is positive or negative. By logistically regressing a model I find that when a firm discloses a GAP, a firm discloses a positive GAP because of proprietary considerations. Through additional analysis I find that the number of intangibles on the balance sheet play a role in the disclosure of segmental information. These results are important to both researchers and standard setters.

(4)

1 Introduction

This thesis focusses on IFRS 8. IFRS 8 – Operating Segments was issued by the IASB in 2006 and eventually replaced IAS 14(R) in 2009. IFRS 8 was mainly introduced for two reasons. The first was to create international convergence with US GAAP. IFRS 8 was developed in close resemblance with SFAS 131 – Disclosures about Segments of an Enterprise and Related Information (Leung & Verriest, 2015). The second reason was that IAS 14(R) offered managerial discretion up to the point that the IASB was not satisfied with the outcome. Managers could, for example, disclose only one segment even though in reality there were more. The IASB therefore introduced the “management approach” into segment reporting. In their eyes this would increase the decision usefulness of segment reporting for the users of financial statements. IFRS 8 was introduced as a binocular to see through managements eyes (Leung & Verriest, 2015; IASB, 2013).

The post-implementation review of IFRS 8, published in 2013, described three major criticisms. There criticisms were raised by the users, prepares and issuers of financial statements. These groups criticized the disclosure requirements of IFRS 8, the use of non-IFRS measures and the identification and reporting of operating segments (IASB, 2013). On 22 March 2018 the IASB decided to not amend IFRS 8, even though amendments were proposed. Of the three criticisms, this study focusses on non-IFRS measures and one specific disclosure requirement: segment reconciliations.

IFRS 8 provides the option to disclose an aggregated segment income that differs from the income number in the profit and loss (P&L) (Nichols, Street and Cereola, 2013). The aggregated segment income number can be a number that is not defined by an IFRS standard, a so-called non-IFRS measure. Examples are EBIT, EBITDA or operating income (Nichols et al. 2013). If the aggregated segment income number differs from the corresponding P&L number a reconciliation is required. A reconciliation reconciles the aggregated segmental number to the consolidated income number (IASB 2006, Nichols et al., 2013).

The use of non-IFRS measures and reconciliations has been criticized by users and prepares of financial statements. Non-IFRS measures are supposedly confusing and hard to explain. (IASB, 2013). Reconciliations are deemed difficult to understand by investors and prepares find them complex. When reconciliations are incomplete, which means that not all items are allocated to the reported segments, they are described as being of less quality (Nichols et al., 2013; IASB, 2013). The ongoing discussion around non-IFRS measures and segment reconciliations makes reconciliations an interesting topic to research. A second reason, why the topic is of interest, is that reconciliations only exist due to managerial discretion (Nichols et al., 2013). Firms have the discretion to disclose an aggregated segment income number that, in their eyes, is most useful to

(6)

evaluate segmental performance (Crawford, Extance, Heliar and Power, 2012; Nichols et al, 2013). This discretion exists due to the “management approach” introduced in IFRS 8. It is therefore interesting which circumstances drive the disclosure of segment reconciliations. This thesis therefore examines what determines the disclosure of reconciliations under IFRS 8, in this study called a GAP. Specifically, I examine (i) what determines managers to disclose a GAP at all and (ii) if a GAP is present, what determines the sign of that GAP. I therefore try to answer the following research question:

What determines the disclosure of segment reconciliations under IFRS 8?

This research is important because of three reasons. Firstly, GAPs have not been investigated in a European setting. Prior IFRS 8 research focused on the differences between disclosure under IAS 14(R) & disclosure under IFRS 8 (Nichols et al., 2013) and the consequences of IFRS 8 implementation. My thesis focusses on the determinants of disclosure. Whereas in Europe GAPs have not been researched, this is different in the US. SFAS 131 mandates segment reconciliation as well. The results of these papers raised the call for European based research on GAPs (Alfonso, Hollie and Yu, 2012; Crawford et al, 2012; Wang & Ettredge, 2015; Hollie & Yu, 2012; Nichols et al, 2013).

Secondly, the outcome of this research adds to the discussion on the implementation of IFRS 8. The outcome of the post-implementation review provided the IASB with arguments to amend IFRS 8. The IASB decided that these arguments were not sufficient. They did not amend IFRS 8 (IASB, 2013; IASB, 2018). In my thesis I try to see what circumstances drive the disclosure of segment reconciliation. I therefore hope to provide an extra argument in the discussion on IFRS 8 and segment reconciliations in general.

Finally, this research adds to the literature on the determinants of segmental disclosure. Prior research has focused on what drives firms to disclose segmental information quality or quantity (Andre, Filip and Moldovan, 2016); whether segmental reporting is driven by agency or proprietary considerations (Birt, Bilson, Smith and Whaley, 2006; Bujega, Czernkowski and Moran, 2015; Leuz, 2004; Nichols & Street, 2007 Prencipe, 2004) and whether other characteristics, such as firm characteristics, are drivers of segmental disclosure (Alfaraih & Alanezi, 2011; Amado, Albuquerque and Rodrigues, 2018). None of these papers focus on reconciliations. Therefore, I contribute to this literature.

Through the analysis of IFRS 8 literature, the determinants of segmental disclosure and reconciliations I come up with two different hypotheses. Since a GAP can either be positive, negative or equal to zero, I consider two hypotheses. This is based on the three constellations a GAP can take. A GAP can either be positive, negative or equal to zero. The first hypothesis

(7)

focusses on the case of whether a firm discloses a GAP or not. The second hypothesis focusses on the sign of GAP, when a GAP is disclosed. Thus, on the case whether a GAP is negative or positive.

The sample of my research is derived from the EuroStoxx 600. My main sample includes 167 firms over the time-period 2011-2015. This results in 835 firm-year observations. My sub-sample, used for the second hypothesis, includes 517 firm-year observations. Through logistic regression I regress the determinant variables on my dummy variable 𝐺𝐴𝑃𝐷𝑢𝑚 This variable takes

the value of 1 when a GAP is disclosed, and 0 when a GAP is not disclosed. I find no significant relationship between my proxies for the agency and proprietary considerations and 𝐺𝐴𝑃𝐷𝑢𝑚. This

can be interpreted as that firms are not influenced by agency or proprietary considerations to disclose a GAP. I therefore, find no evidence for my first hypothesis.

In the analysis of my sub-sample I find evidence for my second hypothesis. I find a significant relationship between the disclosure of a positive GAP and proprietary considerations. In the light of the IASB’s (2018) decision to not amend IFRS 8, these findings are important. They show that firms, in certain circumstances, might disclose imperfect information. These findings might be important for the users of the financial statements who criticize reconciliations and require IFRS 8 amendments.

I perform additional analysis to see if my results are robust. Instead of the proxies for the proprietary consideration, profit margin and research & development expenses, I use the Herfindahl Index. This does not change my results. Suggesting my results are robust.

In my second additional analysis I add two variables to my initial model. I add variables for the amount of goodwill and the number of other intangibles on the balance sheet. This changes my initial results. Both the disclosure of a GAP and the sign of a GAP are influenced by the added variables. This is in line with the findings of Wang & Ettredge (2015).

In the next section I discuss the relevant literature. Section 3 contains the hypothesis development and my methodology. Section 4 contains my main analysis and the additional analysis. I discuss my results in section 5 and conclude my thesis in section 6.

(8)

2 Literature review

In this section I will discuss the literature on IFRS 8, segmental disclosure and segmental reconciliations.

2.1 IFRS 8/Segmental reporting

2.1.1. The differences between IAS 14(R) and IFRS 8

Introduced in November 2006, IFRS 8 replaced IAS 14 and became mandatory for companies from January 1, 2009 (Leung & Verriest, 2015, p. 278; IASB 2006). Table 1 shows the main differences between IAS 14(R) and IFRS 8 (Lenormand & Touchais, 2014; Nichols et al. 2013).

Table 1 provides information on the shift that IFRS 8 introduced. As explained in the introduction, IFRS 8 introduced the “management approach”. Table 1 shows this. On the right side of the table, the CODM is apparent in all three major differences: identification of segments, measurement of segment information and disclosures. The role of the CODM in segmental

Table 1: The main differences between IAS 14(R) and IFRS 8

IAS 14(R) IFRS 8

Identification of segments - Segments are identified based upon whether segmental risks are geographical, or business based;

- Existence of both primary and secondary segments;

- Need to identify that segment that earns the majority in sales.

- Segments are identified based upon the internal reporting system which is used by the Chief Operating Decision Maker (CODM);

- Only one sort of segment;

- No need to identify major segment, when not used for evaluation by the CODM

Measurement of segment

information - Segment information prepared in conformity with the accounting policies in the financial statements

- Segment information prepared in the light of the CODM. The information necessary for the CODM to evaluate segments determines the segmental disclosure. Even if this does not comply with the accounting policies.

Disclosure - Disclose all segmental information that

is required by IAS 14(R) Examples are: segmental revenues, assets, liabilities, capital expenditure, profit.

- Disclose only segmental profit and assets. Additional items only need to be disclosed when used for evaluation by the CODM.

- Disclosure is required when a firm earns more than 10% of its revenue from one client. This needs to be disclosed with the operating segment that accounts for this revenue.

This table describes the main differences between IAS14(R) and IFRS 8. The main differences can be divided in three categories: identification of segments, measurement of segment information and disclosure requirements. Based upon Lenormand & Touchais, 2014 and Nichols et al., 2013.

(9)

reporting has increased after the introduction of IFRS 8. For example, when the CODM evaluates a company by three segments, under IFRS 8, three segments are identified. The identified segment is based upon the internal processes used by the CODM. Under IAS 14(R), identification of segments was risk-based. Primary segments are identified as, the segments that hold the highest risks (business based) or when segments are more geographical diverse (geography based) (Lenormand & Touchais, 2014). Secondary segments are segments that are less diverse and hold less risk.

Measurement of segments is, under IFRS 8, based upon the CODM’s evaluation method. This differs from IAS 14(R), under which segmental measurement follows the accounting policies that apply to the rest of the financial statements. The CODM’s way of evaluation and measuring the segment does not have to comply with the accounting policies of the financial statements (Lenormand & Touchais, 2014).

The disclosure requirements of IFRS 8 are less strict than those of IAS 14(R). IFRS 8 only mandates the disclosure of segment profit and assets. Besides these disclosures, IFRS 8 mandates the disclosure of revenue information on the segment level when the revenue from one client accounts for more than 10% of the total revenue. A firm needs to disclose which segment obtained this revenue. All the other segment information only has to be disclosed when it is relevant in the CODM’s decision making process. IAS 14(R) requires more strict disclosure of segment information. See Table 1 for relevant examples.

2.1.2. IFRS 8 and recent research on IFRS 8

IFRS 8 mandates a firm to report its operating segments and defines an operating segment as a component of an entity (IASB, 2008):

a) “that engages in business activities from which it may earn revenues and incur expenses, b) whose operating results are regularly reviewed by the entity’s chief operating decision maker

(CODM) to make decisions about resources to be allocated to the segment and assess its performance, and

c) for which discrete financial information is available.”

The definition of an operating segment reveals the focus of IFRS 8 on the CODM. This corresponds with Table 1. The IASB introduced the CODM and the management approach as part of an ongoing convergence between IFRS and US GAAP. The CODM concept and the rest of the standard, closely resemble SFAS 131 (Leung & Verriest, 2015). The CODM, or the management approach, was introduced for two reasons: to increase the usefulness of segmental disclosure and consulted users favored this method (Leung & Verriest, 2015). The IASB favored

(10)

the idea of the managerial approach because literature on SFAS 131 showed that the managerial approach would result in enhanced disclosure of segmental information. Examples of enhanced disclosure are: better alignment with economic reality and the disclosure of more segments. A change to the managerial approach also meant enhanced consistency with other sections of the financial statements (Nichols et al, 2013).

Besides the managerial approach, IFRS 8 introduced other ideas within segmental reporting. One idea that was introduced, was the use of non-IFRS measures. This follows from the managerial approach. The focus on the CODM in IFRS 8, means a focus on managerial information. Therefore, segmental information is based upon managerial information. This information is not necessarily IFRS-based information. Examples of non-IFRS measures are EBIT or EBITDA (Leung & Verriest, 2015; Nichols et al., 2013). Another idea was that the identification of segments should not be geographically, or business based. Instead the operations of the firms drive the identification of segments (Nichols et al., 2013).

Four years after the introduction of IFRS 8 the IASB published the post-implementation review (PIR). The IASB identified three major problems with the standard. The first problem was that IFRS 8 did not created the change the IASB expected (IASB, 2013). There was, for example, no significant increase in the number of segments disclosed or increase in the value relevance of this information. This complies with what Nichols et al. (2012) describe. They describe that researchers found no change in the number of operating segments disclosed (Nichols et al, 2013; Bujega, Czernkowski and Bowen, 2012; Crawford et al., 2012; Nichols et al, 2012). A reason might be the managerial discretion in IFRS 8 (IASB, 2013). The second problem is the use of non-IFRS measures. According to the critique in the PIR non-IFRS measures need to be defined more properly (Nichols et al., 2013). The third problem concerns reconciliations. Preparers deem reconciliations clear and easy to comply with but also confusing. There are no strict rules on how to prepare reconciliations. Regulators believe that reconciliations are not prepared properly. Investors state that reconciliations are difficult to understand and have a low usefulness because it is not clear what the reconciliation precisely represent (IASB, 2013). There is no empirical or academic research to ground these statements (Nichols et al., 2013). Recent research on IFRS 8, published after the PIR, has partly focused on the same issues as before the PIR.

Franzen & Weißenberger (2018) research the effects of the introduction of the managerial approach under IFRS 8. They test whether German firms experience differences in the level of information asymmetry and forecast accuracy after the mandatory adoption of IFRS 8. They use a sample from 2007 until 2010. Through a differences-in-difference methodology, using a control group and a treatment group, Franzen & Weißenberg (2018) find no change in information

(11)

asymmetry and forecast accuracy for mandatory adopters. The results are partly problematic because the control group also discloses under IFRS but has chosen to adapt IFRS 8 earlier. Therefore, there is a big similarity on the research level.

Cereola, Nichols and Street (2017) examine the influence of the adoption of IFRS 8 on blue-chip1 companies in Europe, Australia and New Zealand. In contrast to earlier research

(Nichols et al. 2013; Bujega et al. (2012), Cereola et al. (2017) find that firms do change disclosures after the adoption of IFRS 8. They found that firms provide more geographic specific information about their segments when you compare IAS 14R disclosures to IFRS 8 disclosures. They also find evidence for the managerial discretion of IFRS 8. Cereola et al. (2017) describe that in 55% of their sample segment revenue is only disclosed when a certain materiality threshold is met. This implies that managers only disclose the segmental information they deem important.

2.1.3 Concluding remarks

IFRS 8 is introduced in 2009. It introduced the CODM, which meant a shift to a managerial approach in segment reporting. This also explains the major differences between IFRS 8 and IAS 14(R) The post-implementation review of IFRS 8 stated criticisms on the standard which criticized its managerial approach, the use of non-IFRS measures and reconciliations. Research after the post-implementation review has mostly focused on the effects of the managerial approach on segmental reporting. Besides research on the consequences of segmental standards, there is also an abundance of research on the determinants of segmental reporting.

2.2 Determinants of segmental reporting

The managerial discretion, offered by IFRS 8, gives managers the opportunity to influence disclosure decisions at the segmental reporting level. Prior research has offered two main reasons for managers to base their decisions on: proprietary considerations and agency problems (Andre et al., 2016).

2.2.1 Proprietary considerations

Proprietary considerations are based upon proprietary costs. Proprietary costs, in segmental reporting, exist due to the disclosure of segmental information. The disclosure of segmental information reveals proprietary information. Proprietary information is information that tells competitors how valuable (or profitable) a segment could be and whether that segment is worth investing in. The disclosing company provides information and might be affected by the entry of

(12)

competitors (Berger & Hann, 2007; Hayes & Lundholm, 1996; Andre et al., 2016).

Prior studies, based on US data, support proprietary considerations. When SFAS 142 was

still in effect, segments in less competitive industries were not reported as much as in other industries. This changed with the introduction of SFAS 131. Andre et al. (2015) describe that under SFAS 14 firms underreported the number of segments by combining them into one segment. SFAS 131 initiated multi-segment disclosure (Andre et al., 2016). Prior studies show that firms try to hide profitable segments in less concentrated industries (Botosan & Stanford, 2005). The same results are found when focusing on high concentrated industries. Under SFAS 14, firms in these industries disclose fewer segments. This suggests that managers want to hide the segment performance. The introduction of SFAS 131 led to an increase of segment disclosure for these sort of firms (Andre et al. 2016).

Leung & Verriest (2015) review literature on the determinants of segmental reporting and describe that there is mixed evidence on the proprietary consideration. Harris (1998) and Bens, Berger and Monahan (2011) show that proprietary costs drive the aggregation of segments. It leads to less segments. Botosan and Harris (2000) contrast these findings and Ettredge, Kwon, Smith and Stone (2006) find that there is no relation between the number of segments disclosed and competition. Wang, Ettredge, Huang and Sun (2011) investigate the relationship between segmental disclosure and the influence of agency costs and proprietary costs. They use the Herfindahl Index (HHI), abnormal profits and capital intensity as proxies for proprietary costs. They find that firms with higher proprietary costs disclose less segmental information.

In a European setting Leuz (2004) and Prencipe (2004) investigate respectively Germany and Italy. Both find that firms tend to disclose less segments when they face high proprietary costs. They both use different proxies for the proprietary costs and segments, which might question the comparability of these studies. Nichols and Street (2007) go further by examining firms from all over the world. Their sample includes, European, American and Asian firms who adopted IFRS. They found a negative relationship between proprietary costs and segmental reporting. Which means that when proprietary costs are apparent, firms disclose less segmental information.

Andre et al. (2016) investigate determinants of segmental reporting with a sample of 270 European firms. They distinguish between three different disclosure groups: Under-disclosure, Over-disclosure and Box-tickers. They can make this distinction because of the discretional space that IFRS 8 offers. The Box-tickers comply with the standard, the Under-disclosers disclose fewer segmental lines and the Over-disclosers disclose more segmental lines than required by the standard. Which means that Under-disclosures provide less segmental information and

(13)

disclosers provide too much information. Andre et al. (2016) find that Under-disclosers provide less segmental information because of higher market concentration and the fear of more entries. In this case the control group for the Under-disclosers is the Over-disclosures. The results of Andre et al. (2016) are in line with prior literature: proprietary considerations lead to the disclosure of less segmental information.

2.2.2 Agency considerations

The agency consideration posits that managers hide segments with a low profit, because it could result in more monitoring by the principal. This results in more controlling costs for the manager (Berger & Hann, 2007; Andre et al. 2016; Leung & Verriest, 2015). Bens et al (2011) find evidence that this agency position is apparent in firms with multiple segments. Berger and Hann (2007) are one of the first to describe the agency consideration as a determinant for segmental disclosure in a US setting. They research whether the change from SFAS 14 to SFAS 131 results in a change in the number of segments disclosed. They find that before the implementation of SFAS 131 firms disclosed more segments, because the change of SFAS 14 to SFAS 131 results in the presence of stronger agency motives. An interpretation of these results is that firms tend to disclose more when agency costs are present. Agency costs have therefore a different influence on segmental disclosure than proprietary costs.

Wang et al. (2011) complement this by using different proxies for agency costs: high abnormal accruals and the availability of more free cash flows. They propose these proxies, because in firms where agency costs are higher managers attempt to keep agency costs high by merger & acquisition activities. Free cash flows make these activities possible, because there is more room for managers to invest. Managers invest in projects that do not have the highest net present value. The main reason to do this is to keep the size of the company big and therefore make the company complex. This makes segmental disclosure more difficult. Wang et al. (2011) explain that high abnormal accruals suggest lower earnings quality and suggests the manipulation of accounting data. Wang et al. 2011 find that higher agency costs, not necessarily lead to less segmental disclosure.

With Australian data Birt et al. (2006) find that voluntary disclosures are both influenced by agency and proprietary considerations. All the firms in their sample prepared with a standard that is (almost) the same as IFRS 8. The research of Birt et al. (2006) differs from other studies because they introduce a variable that incorporates both the proprietary and the agency consideration. This variable, named OC, is the product of their proxies for agency considerations (managerial ownership) and proprietary considerations (HHI). What they find is that this product variable has, in every case, a stronger relationship with the disclosure of segmental information than the agency or proprietary proxy.

(14)

Bujega et al. (2015) investigate the effect of IAS 14 and IFRS 8 adoption in Australia. What they find is that after the introduction of those standards firms increased the number of reported segments. They continue their research by looking at what determines this and conclude that this is due to agency considerations. Interestingly enough Bujega et al. (2015) also find that managers have proprietary considerations when it comes to the disclosure of line item information. Suggesting that both proprietary and agency considerations influence segmental disclosure.

2.2.3 Other determinants of segmental disclosure

Agency and proprietary considerations are the two most prominent determinants discussed in prior literature, but there might be other determinants. Possible determinants are firm size, international presence, barriers of entry, auditor influence, profitability or firm age (Leung & Verriest, 2015; Amado, et al. 2018).

Amado, Albuquerque and Rodrigues (2018) explore different explanatory characteristics that could determine segmental reporting. Amado et al. (2018) focus on the number of segments disclosed and the number of mandatory and/or voluntary items that is disclosed per segment. They find, based upon a sample of 91 European firms, that firm size is positively related with more segmental disclosure. An explanation is that firms which are bigger in size, are more complex and therefore require more segmental disclosure.

Alfaraih & Alanezi (2011) find that for Kuwaitan firms, firms who are bigger, more leveraged, more profitable and audited by a Big4 company provide more detailed segmental disclosures. These findings correspond with the results of Prencipe (2004) who interprets the results from a proprietary perspective. Important to note is that the results of Alfaraih & Alanezi are based upon firms who used IAS14(R) and these results might not be transferrable to IFRS 8.

2.2.4 Concluding Remarks

In this paragraph I have given an overview of the research on the determinants of segmental disclosure. The main focus of this literature is that the presence of agency and proprietary costs give the consideration on whether to disclose certain information or not. The literature presents evidence that both the agency and proprietary considerations are present. Recent literature has also tried to focus on different determinants of segmental reporting than just the agency and proprietary consideration.

(15)

2.3 Reconciliation and reconciliation gaps

A reconciliation provides information on how one number can be traced back to another number. A disclosed reconciliation provides information on the steps taken to get from one number to another. An example is 20-K reconciliations. A vast amount of literature has been written about the relevance and informativeness of 20-K reconciliations. 20-K reconciliations are mandatory when a firm is cross-listed in the United States and another country, but reports under IFRS (Kim, Li and Li, 2012). A company has to report the reconciliation of its IFRS numbers to its US GAAP counterpart. This should provide information to investors (Chen & Khurana, 2014).

This thesis focusses on segment reconciliations. IFRS 8 provides the discretionary space to disclose information based upon the management’s view. Segmental information is reflective of the way information flows to the CODM (Nichols et al, 2013). A CODM could reflect the performance of segments different from the performance of the consolidated company. Therefore, the CODM might use different measures: one for the segments (for example EBIT or EBITDA) and one for the consolidated income (for example Operating Income). IFRS 8 mandates that if a different segmented measure is used, a reconciliation must be disclosed (IASB, 2006; Nichols et al. 2013). This is a reconciliation from the aggregated segment income to the consolidated income number. Figure 1 gives an illustration of a segment reconciliation (Alfonso, Hollie & Yu., 2012; Hollie & Yu, 2012).

Figure 13:

This figure shows the general idea of a segment reconciliation. (A) is the sum of the segment income numbers. (B) is the consolidated income number from the P&L and GAP is the difference between those two.

Figure 1 shows that the aggregated segment income (A) number is the summation of all the segment income numbers. The number of segments can vary from 1 to x, based upon how the CODM identifies its segments. The consolidated income number (B) is from the P&L. The arrow between (A) and (B) reflects the difference between these two. When there is a difference between

3_{This figure is based upon the figure of Alfonso et al. (2012, p. 49)}

Segment income 1

Segment income 2

Segment income x

Aggregated segment income number (A) Consolidated income number (B)

Segment reconciliation (GAP): (B)-(A)

(16)

those two, a segment reconciliation must be disclosed. In this thesis, I call segment reconciliations a GAP. Both (A) and (B) can be an IFRS or Non-IFRS measure. (A) is depended on the CODM’s segment evaluation process. This specific point of view influences the sign and magnitude of the GAP (see further). Figure 2 gives a practical example of a reconciliation from Volkswagen’s 2016 financial statements:

Figure 2:

This figure shows the reconciliation from Segment profit or loss to the operating result. From Volkswagens 2016 financial statements.

This example shows that Volkswagen discloses an aggregated segment profit or loss of 8.171. The segment measure they use is a non-IFRS measure (operating profit or loss). This measure differs from the P&L number, operating result. Therefore, a reconciliation is required. The GAP is 7.103 -/- 8.171 = -1.068. The GAP is negative because segment profit is bigger than consolidated income number (see Table 2 below).

The Volkswagen example shows a specific GAP configuration. In this case the reconciliation has three components: unallocated activities, group financing and consolidation. In general, a reconciliation is comprised of three components (Alfonso et al., 2012). I expand on the right side of figure 1 to illustrate this:

(17)

Figure 3:

This figure shows the three components of a GAP: un allocated revenue and expenses, unallocated gains and losses and differences between accounting policies on the segment and consolidated level. Based upon Alfonso et al., 2012.

Based upon figure 3, a GAP is, in general, comprised of three elements: unallocated revenue and expenses, unallocated gains and losses and accounting policy differences at the segment and consolidated level. An example of unallocated revenues and expenses is goodwill impairment. Sometimes an impairment cannot be allocated to the segment level. An example of unallocated gains and losses is when a foreign operation is shared by different segments. Currency translations of this operation are harder to allocate. An example of different accounting policies is the situation where a different measure is used at the segment level. The use of EBITDA at the segment level leads to a GAP that consists of differences in amortization, depreciation & depletion and other elements.

IFRS 8 provides the room to allocate, or not allocate items to segments; to disclose a segment measure that is bigger, smaller or equal to the consolidated number. Since IFRS 8 does not provide detailed guidance on how a GAP should be disclosed, there is a lot of discretionary space for the preparer (IASB, 2006). The IASB (2006) only mandates that in a reconciliation, material items are disclosed. Taken all this into account, a GAP occurs in three ways. Table 2 provides an overview of these three. The GAP could be equal to zero (row 1), smaller than zero (row 2) or bigger than zero (row 3). When a GAP is smaller than zero, the sign of the GAP is negative. When a GAP is bigger than zero, the GAP is positive. The sign of the GAP reflects the CODM’s point of view on how to evaluate4segments (Alfonso et al., 2012; Wang & Ettredge,

2015). When the aggregated segmental income measure is bigger or smaller; a reconciliation is mandated (IASB, 2012).

4_{With to evaluate I mean: how to evaluate segmental performance. The CODM uses a certain}

measure (in the case of Volkswagen Operating Profit) to evaluate the performance of the aggregated segments.

Consolidated income number (B)

Aggregated segment income number (A)

Segment reconciliation GAP: (B)-(A)

GAP = (C) + (D) + (E)

Unallocated revenue and expenses (C)

Unallocated gains and losses (D)

Segmental accounting policies that differ from consolidated (E)

(18)

Table 2: Occurrences of a GAP Sign of GAP

Aggregated segment income = Consolidated Income measure in P&L _{GAP = 0} _{No sign}

Aggregated segment income > Consolidated Income measure in P&L _{GAP < 0 Sign = -} Aggregated segment income < Consolidated Income measure in P&L _{GAP > 0 Sign = +} This table presents the three occurrences of a GAP

The disclosure of a GAP has been criticized (IASB, 2013). The disclosed segment measure is deemed difficult to understand and interpret. These measures might not be as useful as the IASB assumed. Another critique is that because firms can disclose a GAP they would not allocate material amounts to operating segments (Peter & Pedersen, 2010). Managers use reconciliations to provide information on unallocated items, but the information could be more useful when allocated to segments5(Nichols et al, 2013; IASB, 2013).

2.3.1 Prior research on segment reconciliations

Research on reconciliations has been US-based. This research has mostly focused on the determinants of reconciliation disclosure. There are three prior studies: Alfonso et al. (2012), Hollie & Yu (2012) and Wang & Ettredge (2015). In the rest of this paragraph I will summarize and review these papers.

Hollie & Yu (2012) look at the informativeness of segment reconciliations and which firm characteristics determine the sign of the reconciliation. They define reconciliations differences, in their case called segment reconciliations (SER’s), as the difference between firm-level consolidated earnings and aggregated segment-level earnings. A SER is, in this study, negative when aggregated segment earnings are greater than consolidated earnings. A SER is positive when consolidated earnings are greater than aggregated segment earnings. This corresponds with Table 2. Hollie and Yu (2012) study both cases: GAP disclosure and GAP sign. As their main focus is on the pricing of SER’s they find that investors misprice positive SERs and that, according to the researchers, this is because reconciliation information is hard for investors to interpret and they therefore underestimate the usefulness of reconciliations.

In addition to their main findings, Hollie and Yu (2012) describe that firms who disclose negative gaps have greater earnings, aggregated segment earnings, sales, ROE, ROA, operating cash flows and firm growth. Firms who disclose positive gaps are in general larger and have more accruals. They also find that firms who disclose gaps compared to firms who do not disclose gaps are financially worse off.

5_{The allocation to segments is not mandatory. Because IFRS 8 introduced the CODM the}

(19)

Wang & Ettredge (2015) focus on the determinants of segmental reconciliations and how markets value these reconciliations. They define reconciliations, in their paper called Gaps, as the difference between summed segment earnings and corporate-level income. Just as in the research of Hollie & Yu a Gap can be positive or negative. A negative Gap exists when corporate-level income is less than segment-level earnings and a positive Gap is when corporate-level income is greater than segment-level earnings. This is no different from Hollie & Yu and correspond with Table 2.

Wang & Ettredge (2015) use a sample of 20594 firm years over a timespan of 14 years. They find, using ordinary logistic regression, that proprietary costs drive companies’ decisions to create a Gap. The more competitive the industry, measured by the Herfindahl index, the larger a gap is. They also find that firms with Gap have more goodwill and other intangibles, are more involved in mergers and acquisitions and have more special items. Their explanation is that it is hard to allocate these sort of items (impairment, M&A costs, special items) to segments. According to their results not only agency costs, but also proprietary costs play an important role in the disclosure of Gaps. The results are not change when you compare firms without Gaps to firm with specific sign (positive or negative) Gaps.

Besides their analysis between firms Wang & Ettredge (2015) find that both agency & proprietary costs determine Gaps. This is in line with their expectations. There is no difference in the results when a Gap is negative or positive, so for all the different forms a Gap can take they find the same result.

Alfonso et al. (2012) find similar results but use different variables. Just as Wang & Ettredge (2015) they proxy agency costs with the Berger & Hann (2007) measure of agency costs, but Alfonso et al. (2012) do not use a proxy for proprietary costs. Alfonso et al. (2012) define reconciliations as the difference between firm-level and aggregated segment-level earnings. In line with the other two papers a segment reconciliation (SR) can be positive or negative. It is negative when aggregated segment-level earnings exceed the firm-level and it is positive when firm-level is greater than at the segment-level.

Using a logistic model to test what determines SR disclosure, Alfonso et al. (2012) find that firms with higher agency costs are more likely to not disclose SR’s and that proprietary costs drive the decisions to disclose SR. Two possible reasons are that stronger competition pushes disclosure policies and that most information is already available to competitors. Therefore, omission of information has no effect. A second logistic model is used to test the determinants of positive or negative SR’s. They find that agency costs mostly determine the decision to disclose a positive SR. They explain that firms with higher agency costs are, segmentally speaking, less profitable. Higher

(20)

agency costs provide a reason to report more conservative and this might have an effect up to the

segmental level.

2.3.2 Concluding remarks

In this paragraph I focused on one specific aspect of segmental disclosure: reconciliations. Reconciliations appear in three forms: equal to zero, bigger than zero and smaller than zero. Equal to zero means that no reconciliation has to be disclosed. Bigger than zero means that the consolidated income measure is bigger than the aggregated segment income measure. Smaller than zero means that the aggregated segment income measure is bigger than the consolidated income measure. Three different studies have been performed about reconciliations under SFAS 131. These studies found that agency as well as proprietary considerations influence the decision to disclose reconciliations. These considerations also influence the sign of a GAP.

(21)

3 Hypotheses development and methodology

In this paper I try to answer what determines the disclosure of reconciliations in financial statements. Reconciliations, in this case, relate to segmental reporting. This study focusses on reconciliations from aggregated segment income to the corresponding number in the P&L.

Previous research on segmental reporting and reconciliations have found two main determinants: agency and proprietary considerations. There are also other determinants. Prior literature did primarily focus on these. (Prencipe, 2004; Leuz & Verriest, 2015; Leuz, 2004; Andre et al., 2016; Alfonso et al., 2012, Hollie & Yu, 2012 and Wang & Ettredge, 2015).

When agency costs are apparent, a manager withholds segment information from the principle (Berger & Hann, 2007). The same logic applies to the disclosure of reconciliations. When agency costs are considered a GAP is rather not disclosed. This means that an aggregated segment income number is disclosed that is equal to the consolidated income number.

Both proprietary and agency costs are taken into consideration when a firm decides to disclose a GAP (Wang & Ettredge, 2015; Alfonso et al, 2012). Previous IFRS 8 literature shows that proprietary costs are a reason to disclose less segmental information (Leuz, 2004; Prencipe 2004; Nichols & Street, 2007; Andre et al., 2016). For the disclosure of reconciliations, and extra segmental information, the agency consideration is more apparent. GAPs are more common, when firms have special items (Wang & Ettredge, 2015). There is an incentive to not allocate these special items to the segmental level. The agent does not want to provide the principal with too much information on segmental performance (Wang & Ettredge 2015; Hollie & Yu, 2012). The disclosure of a reconciliation provides room to decrease agency costs. A reconciliation provides extra information to the principal. Based on previous literature my first hypothesis is:

H1: agency costs drive the disclosure of IFRS 8 reconciliation gaps

Prior research shows that both agency and proprietary considerations influence the sign of a GAP (Alfonso et al., 2012; Wang & Ettredge, 2015). According to Wang & Ettredge (2015) disclosure of a negative GAP is influenced by the presence of agency costs. Disclosure of a positive GAP is related to both agency and proprietary consideration. Alfonso et al. (2012) find that the agency costs influence the sign of a GAP. Firms where the agency cost are considered are twice more likely to disclose a positive GAP, which can be explained by abnormal low segmental profits in the light of more managerial oversight (due to the presence of high agency costs). Since all this research is based in the US, it is hard to determine whether in an IFRS setting the agency or proprietary consideration is more apparent. Recent research on segmental disclosure, under IFRS, points to both considerations (Andre et al., 2016; Bujega et al., 2015) Older literature (Leuz, 2004;

(22)

Prencipe, 2004; Birt et al., 2006) shows that under IFRS 8 proprietary considerations are more common. I therefore describe my second hypothesis in the following way:

H2: The disclosure of a positive GAP is driven by proprietary costs. 3.1 Methodology

3.1.1 Dependent variable: GAPDum and GAPPos

Based upon Wang & Ettredge (2015), Hollie & Yu (2012) and Alfonso et al. (2012) I use logistic regression to test my hypotheses. Logistic regression analysis is a statistical method that tries to find a relationship between the dependent and independent variable. The dependent variable is defined in a dichotomous form that is either 1 or 0. The possible relationship between the dependent and the independent variable explains the likelihood that the independent variable impacts the dependent variable to take the form of either 1 or 0.

In this thesis I try to investigate what drives segmental reconciliation. A reconciliation can occur in three different ways (see Table 2). In line with Alfonso et al. (2012) and Wang & Ettredge (2015) I investigate two different cases. For first hypothesis, I research the case where GAP is either 0 or  0. To test this, I created the dependent variable GAPDum. This variable is a dichotomous

variable that takes the value of 1 when GAP is not equal to zero and the value of 0 when the GAP is equal to zero.

In the second case I investigate the sign of a GAP; where GAP > 0 or GAP > 0. In this case I created the dependent variable GAPPos. This variable is a dichotomous variable that takes the

value of 1 when GAP > 0 and the value of 0 when GAP < 0.

To test H1 and H2 I relate the determinants (most importantly agency costs and proprietary costs) to the dichotomous variables GAPDum and GAPPos with the following model:

(1) 𝐺𝐴𝑃_𝐷𝑢𝑚(𝑜𝑟 𝐺𝐴𝑃_𝑃𝑜𝑠)

= 𝛽₀+ 𝛽₁∗ 𝑅𝐷𝑆 + 𝛽₂∗ 𝑃𝑀 + 𝛽₃ ∗ 𝐿𝑁_𝑂𝑊𝑁+ 𝛽₄ ∗ 𝐿𝑁_𝐹𝑂𝐿+ 𝛽₅∗ 𝑅𝑒𝑡𝑉𝑜𝑙_𝑁 + 𝛽₆ ∗ 𝑅𝑂𝐴 + 𝛽₇∗ 𝑆𝑒𝑔𝐼_𝑇𝐴+ 𝛽₈ ∗ 𝐵𝑖𝑔4 + 𝛽₉∗ 𝑆𝑒𝑔 + 𝛽₁₀∗ 𝐿𝑒𝑣 + 𝛽₁₁∗ 𝐿𝑜𝑠𝑠 + 𝛽₁₂∗ 𝑀𝑇𝐵 + 𝜀

3.1.2 Independent variables

Following Andre et al. (2016) and Birt et al (2006), I proxy for agency costs by using a measure for insider ownership LNOWN. The reason to use this proxy is that insiders, who own part of the

company, have reasons to divert resources in such a way that it benefits themselves instead of the stakeholders (Andre et al. 2016; Jensen & Meckling 1976). I use this proxy instead of other agency

(23)

proxies because it is used in European literature on segmental disclosure6 (Andre et al., 2016; Birt

et al., 2006). LNOWN is defined as the natural logarithm7of the percentage of shares that it is owned

by insiders. Insiders are defined as managers who are capable of exerting discretion on segmental disclosure (Andre et al., 2016). When there is no insiders’ ownership the variable is set equal to zero.

Based upon previous research (Alfonso et al., 2012; Wang & Ettredge, 2015) I expect the sign to be positive for H1 and it is unknown what to expect for H2. This means that when an agency consideration is more apparent, the higher the change that a firm discloses a GAP. The reason is that when agency costs are higher, firms disclose a GAP to provide complementary information to the principal.

I proxy the proprietary consideration with two different variables: RDS and PM. I again follow Andre et al. (2016). They proxy proprietary costs by using research and developments costs. R&D costs partly show the amount of money that a company puts into innovation and therefore invests into new business. When competition is strong a firm does not want to invest as much into R&D. Competitors could copy these innovations (Ellis, Fee and Thomas (2012). RDS is defined as R&D expenses in year t scaled by the lagged8 (t – 1) total sales. When no data on the R&D

expenses in year t are available the measure is set to zero.

In accordance with Alfonso et al. (2012) and Ettredge and Wang (2015) I expect this proxy to be negative in the full sample analysis, and negative in the sub-sample analysis. When a firm faces higher proprietary costs, there is no reason to disclose a GAP. It is not mandatory for firms under IFRS to disclose R&D expenses at the segmental level, but it does not mean that the firm wants to provide additional information. For the second hypothesis I expect the sign to be positive. Firms with a proprietary consideration want to disclose a positive GAP because they do not want to provide additional information to competitors about segmental performance.

The second proxy I use for proprietary costs is the profit margin (PM). Soliman (2008) explains that proprietary costs are higher when firms have a greater profit margin. Providing

6_{In SFAS 131 literature on segmental disclosure the standard measure for agency costs, is the}

measure developed by Berger and Hann (2007). This measure proxies for agency costs by determining which segments have abnormal low profits and therefore need excess investment. This measure was not usable because not enough firms disclosed the required information. Therefore, I followed Andre et al. (2016) who conducted a European study of the determinants of segmental disclosure.

7_{The reason to take the natural logarithm is to scale for the differences in insider ownership. The}

percentage of insider ownership is from 0 percent up to almost 43%.

8_{The reason to scale by total lagged sales is that sales in the previous year have an impact on the}

cash flow of that year and therefore play an important role in the determination of innovation investment in the following year.

(24)

disclosures of (segmental) profit margin draws possible new entrants to that specific segment or market. Other competitors might be tempted to copy segment operations due to the high profit margin (Soliman, 2008). In line with Soliman (2008), I define PM as the operating income divided by net sales. I expect the signs to be in the same as the sign of RDS.

3.1.3 Control variables

Prior literature has found a variety of other variables that could influence segmental disclosure and GAP disclosure.

I control for firm characteristics with the Market-To-Book ratio (MTB), analyst following LNFOL return on assets ((ROA) and leverage (Lev) (Andre et al., 2016; Bujega et al., 2015; Leuz &

Verriest, 2015; Alfonso et al. 2012). Following Andre et al. (2016) I control for the MTB ratio to control for firm growth. Andre et al (2016), Leuz & Verriest (2015) control for firm performance by using the return on assets. The higher the ROA the more incentive to provide fewer segmental disclosures, because it could provide information on segment profitability. A higher ROA is therefore a reason that not provide a GAP. ROA is defined as the operating income divided by the total assets. Bujega et al. (2015) control for the voluntary disclosure of segmental information with the use of leverage. When the amount of debt is higher, it means more segmental disclosure is required due to covenants. I follow them. Lev is defined as the total of liabilities divided by the total of assets. The transparency of a firm is found to play a role in segmental disclosure (Leuz & Verriest, 2015). The more transparent a firm is, the more it discloses segmental information. Therefore, I follow Leuz & Verriest (2015) and proxy for transparency by the number of analysts that follow a firm in a specific firm-year.

Additionally, I control for the volatility of return with the variable RetVolN because firms

might disclose segments differently when returns are volatile (Alfonso et al., 2012; Andre et al., 2016). More detailed explanations of these variables are given in Appendix A.

Both Alfonso et al. (2012) and Wang & Ettredge (2015) found that firms, who have an operating loss, rather disclose no GAPs to not gave away too much information but comply with the standard. Andre et al. (2016) describe that a control variable for losses also captures firm performance. I therefore control for a firm-year loss with the variable Loss that is a dichotomous variable equal to 1 when it is a loss-year and 0 when this is not the case.

I control for the information environment with the variable Big4 that is equal to 1 when a firm has a Big4 auditor and 0 when this is not the case. Firms that are audited by Big4 auditors are more likely to report conservatively and therefore provide higher accounting information quality (Andre et al. 2016; Bujega et al., 2015; Alfonso et al, 2012; Leuz & Verriest, 2015).

(25)

control for segmental disclosure. The number of segments has been studied extensively in prior research (Nichols et al., 2013; Nichols et al., 2012; Cereola et al., 2017) and what has been established is that firm are shifting segments to enhance the information environment or to hide certain information. The number of segments is also an indication of firm complexity and therefore controls for firm complexity (Bujega et al. 2015).

Alfonso et al. (2012) control for the segmented income number because it provides a reason to decide whether or not a reconciliation might be required from a managerial perspective. IFRS 8 does mandate the disclosure of an aggregated segment income number as well. They also find that this relationship exists under SFAS 131. I therefore control with the variable SegITA.

(26)

4 Sample and results

4.1 Sample

The segmental data was hand collected due to the lack of information in available databases. DataStream offers partial segmental information, but this data is not sufficient to do the required analysis. The sample is constructed from the Eurostoxx 600. The 300 biggest companies at 01/05/2018 are the basis of my main sample. The years included in the sample are 2011-2015. The main reason is the availability of data in the Osiris database. Osiris database contains the most information on European listed firms.

In line with Andre et al. (2016) and Nichols et al. (2013) I excluded companies active in the financial sector. Based upon their industry classification these are companies that are active in: banking, insurance and other financial services. This amounts to 57 companies. Of the remaining 243 companies, the managerial ownership data in S&P’s Capital IQ was not available for 71 companies in the period 2011-2015. Leading to 172 companies. Of those companies, two companies report with US GAAP and three companies were excluded because upon hand-collecting their segment disclosures were deemed not informative enough. The final sample contains 167 firms, meaning 835 firm years. Table 3 shows, the sample construction.

Table 3: Sample construction

300 biggest companies of Stoxx 600 300

(-) Financial companies - 57

(-) No managerial ownership data - 71

(-) US GAAP reporting - 2

(-) Insufficient IFRS 8 information -3

(=) Total 167

This table describes the sampling procedure

The sample companies are all listed companies from 16 different EU countries, this is summarized in Table 4. The biggest number of companies comes from Great Britain (31,14 percent), followed by France (17.96%). All the other countries represent roughly less than 10 percent of the sample. Almost everything is line with the sample of Andre et al. (2016) except for the German companies within the sample. Out of the 35 German companies

(27)

in the 300 biggest companies of the Eurostoxx 600, 5 are financial service companies and of 22 German companies, no ownership data was available over the full time-period. In line with Andre et al. (2016) the rest of the company distribution fits the Eurostoxx distribution. I use a different industry classification than Andre et al. (2016). The distribution results (see Table 5) show that Healthcare and Industrial Goods and Services are the only classifications with more than 10% of the sample (respectively 10.78% and 16.77%). The sample distribution differs from previous

Table 4: Sample per country

Country Frequency Percentage

Austria 1 0.60 Belgium 2 1.20 Denmark 8 4.79 Finland 6 3.59 France 30 17.96 Germany 6 3.59 Great Britain 52 31.14 Ireland 4 2.40 Italy 6 3.59 Luxemburg 1 0.60 The Netherlands 11 6.59 Norway 4 2.40 Portugal 1 0.60 Spain 9 5.39 Sweden 11 6.59 Switzerland 15 8.97 Total 167 100

Table 5: Sample per industry classification

Country Frequency Percentage

Automobiles & Parts 5 2.99

Basic Resources 11 6.59

Chemicals 8 4.79

Construction & Materials 10 5.98

Food & Beverage 11 6.59

Health Care 18 10.78

Industrial goods & Services 28 16.77

Media 9 5.39

Oil & Gas 9 5.39

Personal & Household Goods 15 8.98

Real Estate 2 1.20

Retail 7 4.19

Technology 11 6.59

Telecommunications 9 5.39

Travel & Leisure 6 3.59

Utilities 8 4.79

Total 167 100

(28)

studies because of the unavailability of data from S&P Capital IQ.

Table 6 provides the descriptive data of the sample. Panel A provides the descriptive statistics of the sample as a whole. Panel B provides the descriptive statistics of all the firms years when the difference between the financial statement measure of income and the segmental measure of income is not equal to zero. Panel C and D show sub-samples of Panel B. Panel C.1 summarizes the statistics when the GAP is bigger than zero and C.2 summarizes the statistics when the GAP is smaller than zero. I winsorized the continuous variables, where the kurtosis and skewness showed high values9, to the 1st and 99th percentile. This is for the independent variables RDS, PM,

RetVolN and ROA.

On average a firm has four different segments, which equals the average of Andre et al. (2016), where the minimum number of segments is 1 and the biggest amount is 14. Interestingly enough there are single-segment firms that disclose a GAP. The average of Big4 is around 94%, which means that most financial statements are audited by a Big4 office. Since Big4 proxies for accounting quality, this could be interpreted as that most of the firms disclose more quality accounting information. On average around 3,5% of a firm is owned by insiders, but the dispersion of insiders’ ownership is big. The reason for low ownership is the fact that the data only contains listed firms. In some companies more than 40% is owners by insiders. Reason might be that founders of the companies still hold shares, or that previous family-owned businesses are now listed, and shares are still mostly owned by family members. Of all the firm-years only 4% is a loss year, with an average profit margin of around 15%. Based upon the dummy GAPDum and Panel B

517 firm-years contain a GAP.

(29)

Table 6: Descriptive statistics

Panel A: descriptive statistics of the full sample

Variable Obs Mean Std. Dev. Min Max

RecTA 835 -0.014 0.033 -0.354 0.217 RDS 835 -0.028 0.049 -0.215 0 LNOWN 835 -1.460 2.319 -6.908 3.910 OWN 835 3.482 8.928 0 49.92 PM 835 0.146 0.130 -0.089 0.800 LNFOL 835 3.114 0.322 1.792 4.127 FOL 835 23.66 7.256 6 62 RetVolN 835 1.835 0.378 1.038 2.926 ROA 835 8.749 7.753 -7.8 36.68 SegITA 835 .112 0.079 -0.159 0.539 Big4 835 0.939 0.240 0 1 Seg 835 4.168 2.119 1 14 Lev 835 .605 0.181 0.107 1.35 Loss 835 0.043 0.203 0 1 MTB 835 3.417 1.214 -120.22 197.83 RecDif 835 -244351.6 1064228 -1.13e+07 4170500 GAPDum 835 0.619 0.486 0 1

This table presents descriptive statistics for the full sample used in the analysis. The sample contains 835 firm years over the time period 2011-2015. See definitions for the variables in Appendix A. OWN, FOL and RecDif are the raw variables of LNFOL, LNOWNand RecTA. All continuous variables are winsorized at the 1st and 99th percentile of their distributions.

Interestingly, the mean of all variables in Panel B shows little to no difference with the means in Panel A. Since for every firm there are five firm years, it is interesting to note that some firms change the segmental income measure between years. 517 firm year observations mean that there are firms that have shifted between providing a measure that is equal to the measure in the P&L and one that is not. The descriptive statistics do not correspond with the findings of Wang et al. (2011), who find that on average Gap-firms disclose more segments. An explanation could be that firms prefer to disclose more segments under US GAAP.

(30)

Panel B: descriptive statistics of the sample where GAP is not equal to zero

Variable Obs Mean Std. Dev. Min Max

RecTA 517 -0.022 0.039 -0.354 0.217 RDS 517 -0.024 .0042 -0.205 0 LNOWN 517 -1.476 2.362 -6.908 3.831 OWN 517 3.499 8.705 0 46.12 PM 517 0.142 0.139 -0.087 0.878 LNFOL 517 3.118 0.319 1.792 3.892 FOL 517 23.70 6.998 6 49 RetVolN 517 1.824 0.360 1.038 2.882 ROA 517 8.082 6.844 -7.8 32.42 SegITA 517 0.113 .0715 -0.045 0.466 Big4 517 0.930 0.254 0 1 Seg 517 4.453 2.253 1 14 Lev 517 0.616 0.167 0.213 1.257 Loss 517 0.044 0.206 0 1 MTB 517 3.024 1.327 -120.22 197.83 RecDif 517 -394649 1330841 -1.13e+07 4170500 GAPPos 517 0.142 0.349 0 1

This table presents descriptive statistics for the sample, where GAP10_{ 0 used in the analysis. The sample contains 517 firm}

years over the time period 2011-2015. See definitions for the variables in Appendix A. OWN, FOL and RecDif are the raw variables of LNFOL, LNOW_N and RecTA. All continuous variables are winsorized at the 1st and 99th percentile of their

distributions.

Panel C.1 and C.2 are both sub-samples of Panel B. There are almost six times more firms that provide a segment income number that is higher than the P&L number. An explanation could be that firms like to show a segment number that shows more potential compared to the P&L number. It could also mean that there are more firm years where it was harder to allocate items from the group level to the segment level.

The mean of PM is lower in Panel C.2 than C.1, which might relate to a lower profit margin being an indication of higher segment income. There is also a big difference in the mean of the MTB variable, which can be explained by the presence of more intangibles, which would

(31)

Panel C.1: descriptive statistics of the sample where GAP is > 0 Panel C.2: descriptive statistics of the sample where GAP is < 0

Variable Obs Mean Std. Dev. Min Max Obs Mean Std.

Dev. Min Max

RecTA 73 0.018 0.031 0.00 0.217 444 0.029 0.036 -0.354 -0.00 RDS 73 -0.025 0.049 -0.205 0 444 -0.024 0.041 -0.205 0 LNOWN 73 -1.910 2.248 -5.809 3.342 444 1.405 2.375 -6.908 3.831 OWN 73 2.599 6.649 0.003 28.28 444 3.647 8.996 0 46.12 PM 73 0.242 0.238 0.018 0.88 444 0.125 0.107 -0.087 0.88 LNFOL 73 3.197 0.333 2.302 3.737 444 3.104 0.315 1.792 3.89 FOL 73 25.73 7.790 10 42 444 23.36 6.810 6 49 RetVolN 73 1.807 0.359 1.038 2.774 444 1.827 0.361 1.038 2.882 ROA 73 1.018 7.419 -4.6 32.42 444 7.737 6.690 -7.8 32.42 SegITA 73 0.097 0.073 0.008 0.363 444 0.116 0.071 -0.045 0.467 Big4 73 0.93 0.254 0 1 444 0.93 0.255 0 1 Seg 73 4.233 2.233 1 13 444 4.489 2.256 1 14 Lev 73 0.612 0.191 0.290 1.257 444 0.615 0.163 0.213 1.116 Loss 73 0 0 0 0 444 0.052 0.222 0 1 MTB 73 0.581 1.786 -120.22 36.85 444 3.426 1.233 -105.96 197.83 RecDif 73 289721.1 737243.3 5 4170500 444 - 507169.4 1372876 -1.13e+07 -1

This table presents descriptive statistics for the sample, where GAP11_{ 0 used in the analysis. The sample contains 517 firm years over the time period 2011-2015. See definitions for the variables in}

Appendix A. OWN, FOL and RecDif are the raw variables of LNOWN, LNFOLand RecTA. Omitted the dummy variable for when GAP is bigger than zero, because it is not informative. All

(32)

be explained by more intangibles for which amortization/impairment is more difficult to allocate to segments (Wang & Ettredge, 2015). More intangibles could lead to a higher difference between market- and book value because it is harder to capture the value of intangibles in accounting numbers (Hulten & Hao, 2008). Panel C.2 shows a higher mean for OWN than C.1. This difference is as expected because a negative GAP would involve more managerial discretion, as managers want to show better segmental results.

Table 7 Panel A and B tabulate the correlations between the different variables for both the full sample and the subsample. The variable PM is significantly correlated with all proxy variables for agency/proprietary costs. A reason could be the overall significance of the profit margin; for example, R&D costs influence the profit margin. The higher the R&D costs, the lower the profit margin. PM is significantly correlated with almost all the control variables and shows the strongest correlation with ROA and SegITA (respectively 88% and 85%), which makes sense as

aggregate segmented income and the profit margin are related, just as the return on assets and the profit margin are connected performance indicators.

Among the control variables, most correlations are significant, but not very strong. The correlation results in the sub-sample show some different correlations between the variables. RDS and the dependent variable GAPPos are not correlated, but LNOWN and GAPPos are. The correlation

with PM is stronger in the sub sample. Interestingly there is no significant correlation between the proxy for agency costs and all the other dependent and the non-control independent variables.

The effects of IFRS 8 : on segment reconciliations and it determinants : a European study