An examination of the confirmation hypothesis in the context of SFAS 131 and geographical segment reporting on profitability

(1)

Amsterdam Business School

An examination of the Confirmation Hypothesis in the context of SFAS 131

and geographical segment reporting on profitability

Name: Thaisia Elisabeth de Waal Student number: 5942659

Thesis supervisor: dr. Sanjay Bissessur & dhr. Wim Janssen Date: 15-06-2016

Word count: 14.632

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

(2)

1

Statement of Originality

This document is written by student Thaisia Elisabeth de Waal 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)

2

Abstract

I examine the Confirmation Hypothesis that audited backward-looking financial outcomes and forward-looking disclosures are complements in the context of SFAS 131. Under SFAS 131 managers of US firms have the option to stop providing audited accounting information on geographical profitability. I predict and find that firms who stop reporting geographical segment profits (the nondisclosers) reduce the number of forward-looking disclosures on foreign operations in their MD&A sections. I find that these firms reduce both good and bad news. Furthermore, I find that before SFAS 131 became effective, the nondiscloser firms disclosed more bad news compared to the firms that choose to continue providing accounting information on geographical segment profits. This suggests that nondisclosers gave up reporting segmented profits because they viewed the disclosure of bad news as undesirable. Moreover, I find that nondisclosers reduce forward-looking disclosures on segment sales rather than segment profits, suggesting that disclosers of segment profits are incentivized to provide forward-looking information on segment sales.

I also investigate whether disclosers exhibit a higher market reaction to their voluntary disclosures than nondisclosers. The results are not significant. One explanation is that the market does not incorporate forward-looking disclosures to the same degree as non-forward-looking news. Finally, I find support for the prediction that disclosers have higher forecast accuracy compared to nondisclosers in the post-SFAS period because their forward-looking disclosures on foreign operations are relatively more verifiable (since they continue to report geographical segment profits in their financial statements which are subject to audit verification), making their forward-looking disclosures more effective at improving the accuracy of analysts’ forecasts.

(4)

3

1. Introduction

In this thesis, I examine the “confirmation” hypothesis as proposed by Ball, Jayaraman, and Shivakumar (2012) within the context of Statement of Financial Accounting Standard 131

Disclosures about Segments of an Enterprise and Related Information. The “confirmation” hypothesis

maintains that audited, backward-looking financial outcomes and forward-looking disclosures are complements, rather than substitutes (Ball, Jayaraman, and Shivakumar, 2012, p.136-137). It has been shown in prior literature that private information disclosure is untruthful in equilibrium (Crawford and Sobel, 1982). Since forward-looking information is privately known only to managers of a firm, voluntary disclosures of this information are uninformative on a stand-alone basis because managers, who may have interests which diverge from the interests of the company, cannot guarantee to be truthful in these disclosures. Therefore, in order to fulfill the informational role of private forward-looking disclosures, managers need a mechanism to credibly communicate forward-looking information to investors. According to the confirmation hypothesis, audited backward-looking financial outcomes can be helpful in this respect. Audited financial statements ensure that reported actual financial outcomes are accurate and independent of managers, thus providing outsiders with a means of evaluating the ex post informativeness and truthfulness of past management disclosures. In turn, the expectation that actual outcomes will be correctly and independently reported in future financial statements, serves to discipline managers to make more truthful voluntary disclosures ex ante (Ball, Jayaraman, and Shivakumar, 2012, p.137). Thus, the a priori commitment of managers to independent verification of financial outcomes by an independent audit firm allows them to credibly disclose private forward-looking information that is otherwise costly for investors or auditors to verify.

The issuance of Statement of Financial Accounting Standard No.131 Disclosures about Segments

of an Enterprise and Related Information (hereafter, SFAS 131) provides an interesting setting to test

the confirmation hypothesis. SFAS 131 was issued by the Financial Accounting Standards Board (hereafter, FASB) in June 1997 and was effective for fiscal years commencing after December 15, 1997 (Berger and Hann, 2003, p.167). SFAS 131 replaced SFAS 14 Financial Reporting for Segments

of a Business Enterprise (hereafter, SFAS 14) on segment disclosures, which was originally issued in

1976. There are reasons to believe that segment disclosures may be particularly important to investors (AIMR Position Paper, 1993, p.59-60). Segment disclosures are an important source of useful information for investors about the operations of a firm, and this may be especially true in the case of foreign operation segment reporting since several studies provide evidence of higher information asymmetry associated with foreign operations compared to domestic operations (Hope and Thomas, 2008, p.599). Indeed, according to the American Institute of Certified Public

(6)

5

Accountants (hereafter, AICPA) “financial information about business segments often is as important as

information about the company as a whole” for users analyzing a company involved in diverse

businesses (FASB, 1997, p.20).

Under the previous standard on segment disclosures, SFAS 14, firms disclosed segment information by line-of-business and geographic area using the so-called “industry approach,” with no immediate link to the internal organization of the company or to the internal decision-making of managers (Herrmann & Thomas, 2000, p.288). Practice has demonstrated, however, that the term “industry” is subjective (FASB, 1997, p.23) and a major concern with SFAS 14 was that the discretion in the definition of “industry” allowed firms to present considerably less segment information to external users than what was reported internally (Berger & Hann, 2003, p.161). As a result disclosures were often highly aggregated, which diminished the usefulness of the reported segment information for decision-making purposes. Indeed, users of financial statement expressed concerns that the flexibility in applying the segment definition criteria in SFAS 14 resulted in less useful information (Herrmann & Thomas, 2000, p.288). For instance, in their 1993 Position Paper, the Association for Investment Management and Research (hereafter, AIMR) expressed its view that reported financial statement data should be much more disaggregated and that it considered SFAS 14 to be “inadequate” in that respect (FASB, 1997, p.19). Furthermore, the Special Committee on Financial Reporting, formed in 1991 by the AICPA, requested that companies should have to, inter alia: disclose a greater number of segments than reported under SFAS 14; provide more information for each segment; and specify segments in such a way that they correspond to internal management reports (FASB, 1997, p.20). Taking investors’ and financial analysts’ concerns into account, the FASB issued SFAS 131, according to which public firms have to define segments as internally viewed by managers. This means that, contrary to the “industry approach” used under SFAS 14 to define reportable segments, these segments are now determined based on the so-called “management approach,” which is based on how a firms’ management organizes segments internally within the firm for making operating decisions and for evaluating performance (FASB, 1997, p.6). The change in segment reporting requirements under SFAS 131 made a relatively significant impact on the disclosure of segment information (Herrmann & Thomas, 2000, p.287). Hermann & Thomas (2000) demonstrate that after the introduction of SFAS 131 there has been an increase in the number of firms providing segment disclosures, and companies were disclosing more items for each operating segment. Furthermore, the proportion of country-level geographic segment disclosures was shown to have increased, while the proportion of broader, regional, geographic area segment disclosures decreased for enterprise-wide disclosures (Herrmann & Thomas, 2000,

(7)

6

p.287). However, with the introduction of SFAS 131 the number of firms reporting earnings by geographic area has declined greatly, since – contrary to the rules under SFAS 14 - this item is no longer required to be disclosed for firms reporting on a basis other than geographic area (Herrmann & Thomas, 2000, p. 287). Thus, although geographic segments are reported at a more detailed level under SFAS No. 131, the disclosures may actually be less informative if investors use geographic earnings as an important source of information (Herrmann & Thomas, 2000, p.299). Before SFAS 131 was introduced, all US-based multinational firms were required to disclose sales, total assets, and earnings by geographic area. Although all multinational firms are still required to disclose geographical sales and long-lived assets, under SFAS 131 most firms have the

option of whether to disclose geographical earnings. As a result, some firms reduce the amount of

accounting information in their financial reports under SFAS 131 by giving up geographical segment reporting on profitability, while continuing segment reporting on sales (Hope and Thomas, 2008). Indeed, Herrmann and Thomas (2000) document that only 16% of the companies in their sample continue to report geographic earnings after implementation of SFAS 131 (Hope, Kang, Thomas, and Vasvari, 2009, p.422). The question arises whether these firms also reduce their forward-looking disclosures on foreign operations compared to firms that continue with geographical segment reporting on profitability (as predicted by the confirmation thesis), or whether managers instead increase the amount of forward looking disclosures on foreign operations to compensate for the incurred information loss. Therefore, the main research question of this paper is as follows:

“Do firms that subsequent to the introduction of SFAS 131 give up geographical segment reporting on profitability simultaneously reduce their forward-looking disclosures on foreign operations, and how does this affect

the usefulness of their forward-looking disclosures on foreign operations to investors and analysts?”

This question is relevant for various reasons. Firstly, we know relatively little on how accounting quality is associated with forward-looking disclosures. In particular, there is little literature on qualitative (rather than quantitative) forward-looking disclosures. However, it has been shown that narrative information contributes to disclosure not only through the clarification of quantitative financial measures, but also through the identification of value generation drivers not clearly represented in financial statements (Beretta and Bozollan, 2007, p.2). Indeed, it has been acknowledged that narrative sections of annual reports rank highly as an information source (Athanasakou and Hussainey, 2014, p. 227). Bozanic et al (2015) also state that other forward-looking disclosures than quantitative management earnings forecasts represent an important and under-studied component of firm disclosures. My dataset consists of forward-looking sentences

(8)

7

on foreign operations which are often qualitative in nature and that have been extracted from Management Discussion and Analysis sections of 10-K filings of multinational US firms, using automated language processing techniques. The MD&A is arguably the most read and most important component of the financial section (Li, 2010, p. 1050).

Secondly, answering this research question will provide more insights in the validity of the confirmation hypothesis by examining the relation between geographical segment information and forward-looking disclosures. If firms that subsequent to the introduction of SFAS 131 give up geographical segment reporting on profitability also reduce their forward-looking disclosures on foreign operations, this indicates that the confirmation hypothesis is valid and that audited financial information is complementary to forward-looking disclosures. If however, firms that give up geographical segment reporting on profitability after SFAS 131 increase their forward-looking disclosures on foreign operations, this might indicate that managers use forward-forward-looking disclosures to compensate for the loss of information that they incur by giving up geographical segment reporting. Furthermore, this could also indicate that managers take advantage of the inverifiabality of the forward-looking disclosures and increase positive forward-looking disclosures. There is limited research on the role of forward-looking disclosures in providing a mechanism to compensate for the loss of accounting information. For instance, Muslu, Radhakrishnan, Subramanyam and Lim (2014) find that firms generate more forward-looking disclosures when their stock prices exhibit lower informational efficiency, indicating the importance of voluntary forward-looking information in compensating for poor information. According to this view, firms which reduce the amount of accounting information by giving up segment reporting on profitability, may attempt to compensate for the loss of information by increasing the amount of voluntary forward-looking disclosures. The present thesis contributes to this literature by examining the relation between low quality geographical segment accounting information (operationalized as firms which decide to discontinue segment reporting on profitability after the introduction of SFAS 131) and voluntary forward-looking disclosures on foreign operations.

Thirdly, there is limited research on how the ex post verifiability of forward-looking disclosures through audited financial statements affects the usefulness of these forward-looking disclosures to analysts and/or investors. Voluntary disclosure is expected to be particularly relevant to stock market investors and financial analysts. Ball, Jayaraman, and Shivakumar (2012) expect and find that if independent verification of financial statements is intended to improve the reliability of voluntary disclosures, one should observe a greater stock market reaction to voluntary disclosures that are associated with greater financial statement verification levels (Ball,

(9)

8

Jayaraman, and Shivakumar, 2012, p.140). Furthermore, Bozzolan, Trombetta, and Beretta (2009) find that financially verifiable forward-looking disclosures are more effective than unverifiable disclosures at improving accuracy and reducing dispersion of analysts’ forecasts. However, Bonsall, Bozanic and Merkley (2013) find evidence of a significant wedge between the way the market incorporates forward-looking news versus non-forward-looking news, with the market response to forward-looking news being roughly 75 percent smaller than the response to non-forward-looking news. In this paper, I will examine whether forward-looking disclosures of firms that continue to report geographical segment profit, relative to firms that give up geographical segment profit, i) experience a greater stock market reaction and b) are more effective at improving accuracy of analysts’ forecasts.

My results show that, conform the predictions of the confirmation hypothesis, firms who stop reporting geographical segment profits (nondisclosers) reduce the number of forward-looking disclosures on foreign operations compared to firms that continue reporting accounting information on geographical segment profits (disclosers). I find that these nondiscloser firms reduce both good and bad news. Moreover, my results show that the nondisclosers disclosed more bad news in the pre-SFAS 131 period compared to the disclosers, suggesting that nondisclosers gave up reporting segmented profits because they viewed the disclosure of bad news as undesirable. Furthermore, I also find that compared to disclosers, nondisclosers reduce forward-looking disclosures on segment sales rather than segment profits, suggesting that disclosers are disciplined to provide forward-looking information on segment sales. I do not, however, find evidence that

disclosers experience a higher stock market reaction to their disclosures. A possible explanation is

that the market does not incorporate forward-looking news to the same extent as non-forward-looking news. Finally, I find support for the prediction that disclosers have higher forecast accuracy compared to nondisclosers because their forward-looking disclosures on foreign operations are relatively more verifiable (since they continue to report geographical segment profits), making their forward-looking disclosures more effective at improving the accuracy of analysts’ forecasts. The paper proceeds as follows. Chapter 2 provides an overview of existing literature and hypotheses development. Chapter 3 describes the research design and regression models. Chapter 4, section 4.2.1., presents the empirical results for the hypothesis that nondisclosers reduce the number of forward-looking sentences on foreign operations compared to disclosers. Section 4.2.2. distinguishes between good and bad news, while section 4.2.3. distinguishes between sentences on sales and profits. Chapter 5 examines whether the stock reaction of disclosers is greater than of

nondisclosers, while Chapter 6 examines whether disclosers exhibit a higher forecast accuracy

(10)

9

2. Literature Review and Hypotheses Development

2.1 Forward-Looking Disclosures and the Information Environment

In the United States, the Securities and Exchange Commission (hereafter, SEC) mandates that publicly traded firms provide a narrative called the Management Discussion and Analysis (hereafter, MD&A) in their 10-K filings. The aim of the MD&A is to level the informational playing field by giving market participants an opportunity to look at a firm "through the eyes of

management by providing an historical and prospective analysis” (Barron, Kile, and O’Keefe, 1999, p.78).

The SEC argues that this type of information improves investors' assessments of future cash flows by improving earnings forecasts, and that financial statements alone may be "insufficient for

an investor to judge the quality of earnings and the likelihood that past performance is indicative of future performance" (Barron, Kile, and O’Keefe, 1999, p.78). With regard to forward-looking information,

the SEC's traditional policy was to prohibit its disclosure. Over the years, however, this policy stance has changed significantly to the point where the SEC encourages, and in some areas even requires, the disclosure of forward-looking information (Romajas, 1993, p.245). In particular, the SEC has emphasized investors’ greater need for forward-looking disclosures, rather than for disclosures about past events (Muslu, Radhakrishnan, Subramanyam and Lim, 2014, p.932).

However, voluntary disclosure theories are ambivalent about whether managers disclose information that is actually useful to investors. On the one hand, the “informativeness perspective” predicts that managers disclose value-relevant information. On the other hand, the “opportunism perspective” predicts that company disclosures are largely shaped by managers’ motives (Muslu, Radhakrishnan, Subramanyam and Lim, 2014, p.934). In their paper, Muslu, Radhakrishnan, Subramanyam and Lim (2014) study the informational properties of forward-looking disclosures in the MD&A sections of 10-K filings made with the SEC. They find that firms make more forward-looking MD&A disclosures when their stock prices have lower informational efficiency, in other words when their stock prices poorly reflect future earnings information. This, according to the authors, seems to underline the importance of voluntary forward-looking information in compensating for poor information. According to this view, firms which reduce the amount of accounting information, may attempt to compensate for the loss of information by increasing the amount of voluntary forward-looking disclosures.

2.2 The “Confirmation” Hypothesis as proposed by Ball, Jayaraman and Shivakumar

The confirmation hypothesis maintains that audited, backward-looking financial outcomes and disclosure of managers’ private forward-looking information are complements since independent audit disciplines managers and hence enhances disclosure credibility. In their paper, Ball,

(11)

10

Jayaraman and Shivakumar (2012) emphasize the interaction between the contracting and informational roles of financial reporting as complementary mechanisms. According to the authors, the primary role of financial reporting is to supply auditable backward-looking financial outcomes for use in efficient contracting with the firm, including the provision of a mechanism through which managers can credibly commit to disclose private, forward-looking information to users. Financial statements contain backward-looking data which is verifiable at relatively low cost by an independent audit firm. However, financial reporting is a relatively inefficient mechanism for communicating managers’ forward-looking information to outsiders, since forward-looking information is relatively costly to verify independently. Thus, while independently verifiable information consists primarily of backward-looking actual outcomes, managers’ private information consists primarily of forward-looking expectations that are not independently verifiable (Ball, Jayaraman and Shivakumar, 2012, 139). In equilibrium unverifiable disclosures are by themselves untruthful and hence uninformative (Crawford and Sobel, 1992). Therefore, managers need a mechanism to credibly committing to be truthful in their forward-looking disclosures. It is here were the “confirmatory role” of financial reporting comes into play. Reporting backward-looking financial outcomes that are accurate and independent of managers provides outsiders with a means of evaluating the ex post informativeness and truthfulness of past management disclosures. In turn, the expectation that actual outcomes will be accurately and independently reported, disciplines managers to make more truthful and informative voluntary disclosures ex ante. Ball, Jayaraman and Shivakumar (2012) indeed find that the resources which firms allocate to audit verification are an increasing function of the extent of their management forecasting activity. Moreover, Ball, Jayaraman and Shivakumar (2012) also examine the stock market reaction to forward looking disclosures. They expect and find a greater stock market reaction to voluntary disclosures that are associated with greater financial statement verification levels.

2.3 Related Literature on the Confirmation Hypothesis

Ball, Jayaraman and Shivakumar (2012) note that the confirmation hypothesis is closely related to models developed by Gigler and Hemmer (1998), Stocken (2000) and Lundholm (2003). These studies also investigate, although in different settings, how verifiable accounting information increases the credibility of non-verifiable disclosures. However, these studies do not specifically evaluate the role of audit verification to signal the credibility of voluntary disclosures (Ball, Jayaraman and Shivakumar, 2012, p.140). In this section I will elaborate on the above-mentioned studies, as well as some other more recent literature that relates to the confirmation hypothesis.

(12)

11

Frank Gigler and Thomas Hemmer (1998) show how mandatory reporting complements voluntary disclosures of private information by playing a confirmatory role in an agency setting where voluntary disclosures are motivated by the desire to achieve efficient contracting. They emphasize the "confirmatory role" of mandatory financial reports and use “frequency” of mandatory financial disclosures to highlight the difference between this view of reporting and the prevailing view of mandatory disclosures as primary sources of information. They claim that since audited statements can be used to evaluate the truthfulness of management’s past voluntary disclosures, they may be useful in creating an environment in which management can credibly communicate more relevant information which is not directly verifiable.

Stocken (2000) investigates disclosures and credibility in the context of financing an investment project. In a single-stage game no communication occurs, but in the repeated game, the manager is able to develop reporting credibility and communication results. The main finding of Stocken is that the manager almost always truthfully discloses his private information provided that the manager's discount factor is sufficiently high, the accounting system that generates the mandatory accounting report is sufficiently useful for assessing the credibility of the manager's disclosure, and the review phase is sufficiently long. Therefore, the article shows that in the absence of a mechanism to enforce verifiability, voluntary disclosures are not credible and therefore ignored by the market. Lundholm (2003) reaches a similar conclusion examining a setting where a firm encounters adverse selection in the equity market and relies on voluntary disclosures to alleviate the problem. Lundholm (2003) studies how historical financial reporting can serve as an ex post check on more timely voluntary disclosures, and demonstrates that such a check can be sufficient to ensure that the voluntary disclosures are credible most of the time.

More recently, Athanasakou and Hussainey (2014) investigate the credibility of forward-looking performance disclosures in the narrative sections of annual reports as perceived by investors. They measure the frequency of forward-looking performance disclosures (FLPD’s) using a score that counts the number of FLPD’s in annual report narrative sections. They find that in the presence of managerial incentives investor reliance on forward looking performance disclosures increases with the quality of earnings reported in the audited financial statements. Their results suggest that firms derive a benefit in terms of higher credibility for their narrative disclosures from having a reputation for high quality earnings.

(13)

12 2.4 Positive and Negative Forward-Looking Disclosures

As Ball, Jayaraman and Shivakumar (2012) mention in their work, the confirming role of audit verification of financial statements may be relevant to both positive and negative forward-looking disclosures. Early studies on management forecasts reveal that negative forward-looking sentences (or “bad” news) are intrinsically more credible than positive forward-looking sentences (or “good” news), since managers prefer to release good news because this influences their performance assessment in a positive way (Ball, Jayaraman and Shivakumar, 2012, p.139). For instance, Skinner (1994) argues that managers have legal incentives (namely incentives to reduce their exposure to litigation risk) and reputational incentives to preempt the announcement of large negative earnings surprises (“bad news”). Besides short-term quantitative forecasts, voluntary disclosure can also take the form of qualitative disclosures about for example the firms’ operations. It has been shown that these qualitative forms of voluntary disclosures tend to increase when firm performance improves, but tend to decline when firm performance deteriorates. This suggests that qualitative forms of voluntary disclosures are selectively used to convey good news about the firm (Roychowdhury and Sletten, 2012, p. 1680).

If it is true that bad news is inherently more credible than good news, then audit verification is less relevant for bad news than for good news. However, more recent studies suggest that managers have incentives to take “big baths” in order to create reserves for the future, which would imply that bad news is not necessarily more credible than good news and arguably could even be more biased (Rogers and Stocken, 2005). This may be the case in particular with regard to segment reporting under SFAS 131, which induces companies to provide more disaggregated segment information, since managers have an incentive to aggregate segment information and conceal segment profits because of proprietary costs (Berger and Hann, 2007). Consistent with the notion that bad news is not necessarily less biased than good news, Kothari et al. (2009) find that management on average delays the release of bad news to investors. Dye (1985) argues that managers will not disclosure nonproprietary information when investors are unsure about the kind of information held by management, whether it is due to the nonexistence of information or due to its adverse content. This means that managers have an incentive to suppress bad news privately available to them, because investors are not aware of the information (Dye, 1985, p.127). Verrecchia (1983) also argues that bad news will be disclosed only if the costs of disclosure are low enough or if the information asymmetry between managers and investors is sufficiently high. Although this discussion about the credibility of bad versus good news means that it is difficult to make a directional hypothesis, it is worthwhile to investigate whether the confirmation hypothesis holds equally for good and bad forward-looking disclosures.

(14)

13 2.5 Financial Analysts and Forward-Looking Disclosures

Ball, Jayaraman and Shivakumar (2012) propose that abnormal audit fees (their proxy for resources committed to independent auditing) allow managers to signal their commitment in providing accurate accounting information, such that managers are not able to conceal management earnings forecast errors. Therefore, I will also test whether forward-looking disclosures of firms that continue segment reporting on profitability are useful to financial analysts in forecasting earnings. Segment data is considered to be important for financial analysts. For instance, paragraph 44 of SFAS 131 cites the AIMR (now CFA Institute) Position Paper (AIMR, 1993, p.59-60): “[Segment data] is vital, essential, fundamental, indispensable, and integral to the

investment analysis process. Analysts need to know and understand how the various components of a multifaceted enterprise behave economically (…). There is little dispute over the analytic usefulness of disaggregated financial data.”

Few previous studies have dealt with this issue. Bozzolan, Trombetta, and Beretta (2009) study the effect of forward-looking disclosures on analysts' forecast properties, in particular accuracy and dispersion. According to the classical ‘unraveling result’ it is “the exogenous or

endogenous credibility of the information disclosed that allows different types of firms to separate and avoid the uninformative pooling equilibrium” (Bozzolan, Trombetta, and Beretta, 2009, p.438). They find

support for the theoretical prediction that verifiable disclosures (which they define as forward-looking sentences which explicitly disclose the direction and measurement of the expected impact) are more effective than unverifiable (where no direction and measurement of expected impact is given) disclosures at improving accuracy and reducing dispersion of analysts' forecasts. Their work therefore shows the value of verifiable versus unverifiable forward-looking disclosures, indicating that verifiability is important for forward-looking disclosures to be informative to analysts. Furthermore, Bryan (1997) using a sample of 250 US firm observations demonstrated that when the information contained in the MD&A, particularly the discussions of future operations and planned capital expenditures, is used as supporting the interpretation of the financial statements, forward-looking disclosures can assist investors in assessing short-term prospects. Kieso and Weygandt (1995) also asserted that forward-looking information is helpful to investors in their forecasts.

In relation to the confirmation hypothesis, one would expect forward-looking disclosures of firms that continue to report geographical segment profit (and thus have verifiable segment information on profitability) relative to firms that give up geographical segment profit, to be more effective at improving accuracy of analysts’ forecasts. In their work, Bozzolan, Trombetta, and Beretta (2009) define financially verifiable looking disclosures as those

(15)

forward-14

looking sentences which include i) the economic sign that communicates the direction of the expected impact upon the future performance of the firm, and ii) the measures used in order to quantify/qualify the expected impact (Bozzolan, Trombetta, and Beretta, 2009, p.444). In my study, however, I will define verifiability of forward-looking sentences on geographical segment profitability on the basis of whether these forward-looking sentences on geographical earnings are disclosed by the firm under SFAS 131 and thus ex post audited and verified by an independent audit firm. Thus, firms that continue to report geographical segment profit under SFAS 131 are considered to have verifiable disclosures with regard to their segment earnings information, whereas firms that give up segment reporting on profitability are unverifiable since that information will not be independently verified by an audit team.

2.6 The Impact of a Change in Segment Reporting under SFAS 131

I will examine the confirmation hypothesis in the context of segment reporting. There are reasons to believe that segment disclosures may be particularly important to investors (AIMR Position Paper, 1993, p.59-60). This may especially be the case for foreign operation segment reporting as several studies provide evidence of higher information asymmetry associated with foreign operations compared to domestic operations (Hope and Thomas, 2008, p.599). On top of that, with the substantial increase in multinational operations of firms in today’s globalized economy, it has become increasingly important for investors to gather information about a firms’ foreign operations. Therefore, it becomes even more critical for both management and investors or analysts to understand better the impact of foreign operations on the overall firm performance (Hope, Kang, Thomas, and Vasvari, 2009, p.421).

In the US, the introduction of SFAS 131 represented a significant change in geographic segment disclosure requirements for US-based multinational firms, which provides me with a natural experiment to examine the confirmation hypothesis. SFAS 131 Disclosures about Segments of

an Enterprise and Related Information, replacing SFAS 14 Financial Reporting for Segments of a Business Enterprise, was issued by FASB in June 1997 and became effective for fiscal years commencing

after December 15, 1997 (Berger and Hann, 2003, p.167). Under the rules of SFS 131, public enterprises have to define segments “as internally viewed by managers.” This means that segments are now determined based on the so-called “management approach,” which is based on the way management organizes segments internally for making operating decisions and for evaluating the operating performance of its business units (FASB, 1997, p.6). By following this approach, SFAS 131 makes managerial decisions more transparent. Whereas the term “industry,” which was used under the industry approach of SFAS 14, has been shown to be subjective in practice, segments

(16)

15

based on an existing internal structure as demanded under the “management approach” adopted under SFAS 131, should be less subjective (FASB, 1997, p.23). Therefore, the adoption of SFAS 131 significantly reduced managers’ discretion to aggregate segments by requiring firms to define segments in a manner consistent with their internal organizational structures (Cho, 2015, p.676).

The change in segment reporting requirements under SFAS No. 131 made a relatively significant impact on the disclosure of segment information (Herrmann & Thomas, 2000, p.287). Two important changes to the disclosure of geographic information disclosure occurred (Hope, Kang, Thomas, and Vasvari, 2009, p.422). First, SFAS 131 encourages firms to disclose country-level geographic information for “material” countries, while allowing immaterial countries to be aggregated into a single “Other Foreign” segment. Second, SFAS 131 allows enterprises to no longer disclose earnings for secondary segments (Hope, Kang, Thomas, and Vasvari, 2009, p.422). Firms that define their primary operating segments on another basis than geographical segment (such as line of business) are only required to disclose sales and assets for geographic operations, but no earnings.

Hermann & Thomas (2000) demonstrate that after the introduction of SFAS 131 there has been an increase in the number of firms providing segment disclosures, and companies were disclosing more items for each operating segment. Furthermore, the proportion of country-level geographic segment disclosures was shown to have increased, while the proportion of broader geographic area segment disclosures decreased for enterprise-wide disclosures (Herrmann & Thomas, 2000, p.287). However, with the introduction of SFAS 131 the number of firms reporting geographical earnings has declined greatly (Herrmann & Thomas, 2000, p.287), since following the passage of SFAS 131 most multinational firms were no longer required to disclose earnings by geographic area (for example net income in Brazil or net income in East Asia). Whereas prior to SFAS 131 all multinational firms were required to disclose sales, total assets, and

earnings by geographic area, under SFAS 131 most firms now have the option of whether or not

to disclose geographic earnings (although all firms are still required to disclose geographic sales and long-lived assets). So while firms must still disclose total foreign earnings as part of SEC regulation paragraph 210.4-08(h), the mere option of disclosing earnings at a less geographical level under SFAS 131 could hinder the ability of shareholders to monitor actions of managers related to geographical operations (Hope and Thomas, 2008, p.592). Thus, although geographic segments are reported at a finer level under SFAS No. 131, the disclosures may actually be less informative if investors rely on geographic earnings as a source of information (Herrmann & Thomas, 2000, p.299).

(17)

16

Hope and Thomas (2008) test the agency theory hypothesis in the context of geographic earnings disclosures and SFAS 131. The authors compare firm performance for a group of disclosers versus a group of non-disclosers in the pre- and post-SFAS 131 periods. Hope and Thomas find that non-disclosure of geographical earnings is associated with a significant increase in foreign sales growth and a significant decrease in foreign profit margin. Furthermore, overall firm values of non-disclosers, measured using Tobin’s q-ratio, are found to be significantly lower than those of disclosers in the post-SFAS 131 period. The lower firm value is consistent with investors detecting the value reducing decisions of managers of non-disclosing firms. Importantly, they do not find a similar pattern in firm performance for domestic operations in the post-SFAS 131 period or for foreign operations in the pre-SFAS 131 period.

Hope, Kang, Thomas, and Vasvari (2009) study the impact of the change in geographical segment disclosure requirements on US-based multinational enterprises from an investor’s perspective. They establish a link between cross-sectional differences in geographic segment disclosure practices and the valuation of foreign earnings. First of all, they compare the pricing of foreign earnings in the post-SFAS 131 period between firms that increase their number of geographic segments and firms that do not. Second, since most firms define their operating segments along industry lines, most firms under SFAS 131 are no longer required to disclose geographic earnings, and most indeed do not (Hope, Kang, Thomas, and Vasvari, 2009, p.425). Therefore, the authors suggest that disclosure of geographic earnings in the post-SFAS 131 period represents higher quality disclosures and will therefore result in higher quality earnings. The findings suggest that increased geographic segment disclosures lead to higher valuations for foreign earnings, and that these valuations are value relevant.

2.7 Hypotheses Development

Considering the studies presented in the previous sections, I test the confirmation hypothesis, which states that giving up geographical segment reporting on profitability will lead to a lower credibility of forward looking disclosures on foreign operations, in the context of SFAS 131. If credibility is crucial for the informational role of forward looking disclosures, firms that give up geographical segment reporting on profitability are likely to reduce their forward-looking disclosures on foreign operations. In contrast, firms that choose to continue to report their geographical segment profit signal higher credibility compared to firms that give up geographical segment reporting, and thus will have a higher level of forward-looking disclosures on foreign operations. This leads to the first hypothesis:

(18)

17

Hypothesis 1a: Firms that subsequent to SFAS 131 stop geographical segment reporting on profitability, also reduce the number of forward-looking disclosures on foreign operations compared to firms that continue with geographical segment reporting on profitability.

Alternatively, it could also be, counter to the predictions of the confirmation hypothesis but conform the findings of Muslu, Radhakrishnan, Subramanyam and Lim (2014), that firms which reduce the amount of accounting information following SFAS 131 increase their forward-looking disclosures in order to, for instance, compensate for the loss of accounting information.

Furthermore, I will differentiate between positive and negative forward-looking disclosures. Since prior literature is inconclusive as to the differential credibility of bad versus good news, I will not make a directional hypothesis in this respect. This leads to the following hypothesis, stated in the null form:

Hypothesis 1b: Firms that following SFAS 131 give up geographical segment reporting on profitability, equally reduce negative and positive forward-looking sentences compared to firms that continue geographical segment reporting.

In my second test, I will examine the credibility of forward-looking disclosures of firms which continue segment reporting on profitability versus firms which give up geographical segment reporting for investors. If additional verification of financial statements is intended to improve the reliability of voluntary disclosures, one should observe a greater stock market reaction to voluntary disclosures that are associated with greater financial statement verification levels. Hypothesis 2: The stock market reaction to firms’ voluntary disclosures is higher for firms that continue geographical segment reporting on profitability compared to firms that stop geographical reporting on profitability.

Finally, I will examine the effects for financial analysts. With regard to financial analysts and the confirmation hypothesis, one would expect forward-looking disclosures of firms that continue to report geographical segment profit (which are thus more verifiable since the information will be audited ex post by an independent audit firm) to be more effective at improving accuracy of analysts’ forecasts compared to firms that stop reporting geographical segment profits. This leads to my final hypothesis:

(19)

18

Hypothesis 3: Analysts’ forecast accuracy is higher for firms that continue geographical segment reporting compared to firms that give up geographical segment reporting.

Alternatively, it could be that no significant results are found for hypothesis 2 and 3. Bonsall, Bozanic and Merkley (2013), for instance, find evidence of a significant wedge between the way the market incorporates forward-looking news versus non-forward-looking news, with the market response to forward-looking news being roughly 75 percent smaller than the response to non-forward-looking news. So it could very well be that investors do not incorporate the news contained in the forward-looking disclosures. Another potential explanation if no results are found is that analysts may have had access to most of the SFAS 131 segment profit information even before the new standard. If analysts already had access to all the segment information made public by mandating the management approach, analyst forecast accuracy would be unaffected by SFAS 131.

(20)

19

3. Methodology

3.1 Sample and Dataset

My sample consists of US multinational firms which are subject to geographical segment reporting around the time SFAS 131 became effective. The initial sample consists of 418 firms, which are US firms that report profit for at least two foreign geographical segments in the two years prior the adoption of SFAS 131. I require that each firm has data in the two years prior (1996-1998) and the three years following (1998-2000) the adoption of SFAS 131. Therefore, the balanced panel dataset covers the period 1996-2000. This allows me to construct a difference-in-difference model, as I am able to perform cross-sectional comparisons between the pre-SFAS 131 period and the post-SFAS 131 period. This design allows me to control for whether any differential performance in the post-SFAS 131 period also existed in the pre-SFAS period. Without this control, it could be that differences in forward-looking disclosures between disclosers and nondisclosers in the post SFAS 131 period existed before the adoption of SFAS 131 because of some other factor.

For the firms in the dataset, forward-looking sentences on foreign operations (foreign sales, profits, costs, investments) were extracted by my supervisor dr. W. Janssen from MD&A sections of 10-K filings, using automated language processing techniques.1_{Since the linguistic content}

analysis is not foolproof, I helped recoding the extracted sentences and manually classify them under various criteria, such as whether the disaggregated information refers to sales, profits, costs or investments; the location of the future disaggregated operation; and the sign of the news (which may be good, bad, or neutral).

The final sample of 418 firms (2.090 firm-year observations) consists of information on the number of forward-looking sentences on foreign segment operations per firm-year, classified under the afore-mentioned criteria. There are two groups of firms. Firms that continue to provide accounting information on geographical segment profitability in both post-SFAS 131 years (1999 and 2000) are labelled as disclosers. In contrast, firms who decide to stop reporting geographical segment profits in any of the post-SFAS 131 years, are labelled as nondisclosers. In my sample, I identify 129 disclosers and 289 nondisclosers.

Furthermore, I supplemented the dataset with additional data from publicly available datasets, such as COMPUSTAT and the IBES database. In order to test hypotheses 2 and 3, I additionally constructed the dependent variables Cumulative Abnormal Returns (using Eventus

1_{My supervisor, dr. W. Janssen, downloaded the 10-K filings from the Edgar database and manually}

extracted the MD&A sections, after which the forward-looking sentences on foreign operations were extracted through linguistic content analysis.

(21)

20

software from CRSP) and Forecast Accuracy (using the IBES database and Compustat). Because of missing values for these two dependent variables, the initial dataset is reduced for hypotheses 2 and 3. This is shown in table 1. Moreover, table 2 presents a comparison of firm characteristics over the three different samples.

Table 1 – Overview Sub-samples

Sub-Samples Number of companies Number of

firm years

% matched

Final annual dataset for hypothesis 1 418 2.090 100%

Less: missing values for CAR (207)

Sample for hypothesis 2 211 1.055 51%

Less: missing values for Forecast Accuracy (161)

Sample for hypothesis 3 257 1.285 62%

This table presents an overview of the sub-samples for the three hypotheses. I require that each firm have data in the two years prior and the three years following the adoption of SFAS 131. The final annual dataset consists of 418 firms (2.090 firm-year observations). This sample is reduced to 211 firms (1.055 firm-year observations) for hypothesis 2 due to missing values for CAR. The sample for hypothesis 3 is 257 firms (1.285 firm-year observations) due to missing values for Forecast Accuracy.

(22)

21

Table 2 – Comparison across the three samples

Panel A: Comparison of variables across samples

Sample 1 Sample 2 Sample 3

Mean Std.dev. Mean Std.dev. Mean Std.dev.

LEV 0.191 0.197 0.165 0.146 0.169 0.140 LIT 0.321 0.467 0.323 0.468 0.333 0.472 LN(ANALYST) 1.339 1.172 2.008 0.872 2.245 0.673 LN(BSEG) 0.651 0.728 0.644 0.730 0.678 0.725 LN(FORSEG) 1.024 0.389 1.018 0.388 1.013 0.340 RATIO_FOREIGNSALES 0.401 0.178 0.395 0.170 0.400 0.156 LN(ASSETS) 6.571 1.950 6.722 1.759 7.117 1.647 ROE 12.342 16.871 11.754 15.578 13.804 14.092

Panel B: Comparison of disclosers and nondisclosers across samples

Number of firms in sample 418 207 161

# Disclosers 129 77 55

# Nondisclosers 289 180 152

% Disclosers in sample 31% 37% 34%

Panel A of this table presents a comparison of several control variables/firm characteristics. Panel B presents a comparison of the number of disclosers and nondisclosers across the three samples, as well as the percentage of disclosers. Appendix A provides a detailed description of the variables.

3.2 Proxies for resources committed to forward-looking disclosures

In their paper on the confirmation hypothesis Ball, Jayaraman and Shivakumar (2012) look at management earnings forecasts as the voluntary disclosure variable and use number, timeliness,

specificity and accuracy of firms’ forecasts as proxies for resources committed to voluntary

disclosure. In this paper however, I focus on a sub-section of forward-looking disclosures, namely qualitative forward-looking disclosures. My dataset consists of forward-looking sentences on foreign operations which are qualitative in nature and that have been extracted from MD&A sections of 10-K filings of multinational US firms, using automated language processing techniques. The MD&A is arguably the most read and most important component of the financial section (Li, 2010, p. 1050). Qualitative forward-looking disclosures are relevant to study since it has been acknowledged in the literature that narrative sections of annual reports rank highly as an information source (Athanasakou and Hussainey, 2014, p. 227). It has been shown that narrative information contributes to disclosure not only through the clarification of quantitative financial measures, but also through the identification of value generation drivers not clearly represented in financial statements (Beretta and Bozollan, 2007, p.2). Therefore, Bozanic

(23)

22

et al (2015) believe that other forward-looking disclosures than just quantitative management earnings forecasts represent an important and under-studied component of firm disclosures.

Because of the qualitative nature of the disclosures studied in this paper, it is not feasible to look at the accuracy, timeliness or specificity of the forward-looking disclosures. Qualitative disclosures are inherently less precise than quantitative point and/or range estimates, and the accuracy of qualitative statements is often difficult to establish. Therefore, following Athanasakou and Hussainey (2014), the proxy for resources committed to forward-looking disclosures on disaggregated foreign operations that I will use in my analysis is the number of forward-looking sentences on disaggregated foreign operations (#FLS_FO). In order to examine whether

non-disclosers decrease the number of forward-looking sentences on disaggregated foreign operations

compared to disclosers, I use the following regression model:

#FLS_FOi,t = γo + γ1SFAS131i,t + γ2G_PROFITi,t + γ3SFAS131*G_PROFITi,t γ4ANALYSTi,t +

γ5ASSETSi,t + γ6LEVi,t + γ7LITi,t + γ8RATIO_FOREIGNSALESi,t + γ9BSEGi,t + γ10FORSEGi,t

+ Σ INDUSTRY + ε

[1a] where #FLS_FO refers to the number of forward-looking sentences on disaggregated foreign operations (calculated as the natural logarithm of one plus the number of forward-looking sentences on foreign operations in the MD&A). G_PROFIT is a dummy variable coded 1 if the firm continues with geographical segment reporting on profitability after the introduction of SFAS 131 (discloser), and zero otherwise (nondiscloser). The dummy variable SFAS131 equals 1 in the case the MD&A is about fiscal periods that end after the 15th_{of December 1998 (when SFAS}

131 became effective), and zero otherwise. The interaction term SFAS131*G_PROFIT is the variable of interest, which allows me to directly link cross-sectional variations in geographic segment disclosures to the introduction of SFAS 131 and the decision to disclose geographical segment profitability or not. I expect γ3 to be positive, since a positive coefficient on γ3 indicates

that nondisclosers (disclosers) decrease (increase) the number of forward-looking sentences on disaggregated foreign operations compared to disclosers (nondisclosers) after the introduction of SFAS 131. If nondisclosers (firms who stop reporting geographical segment profits in their financial statements) reduce the number of forward-looking sentences on foreign operations, this provides evidence for the confirmatory role of financial reporting.

A positive (negative) coefficient on SFAS131 means that nondisclosers increase (decrease) the number of forward-looking sentences on foreign operations in the post-SFAS period,

(24)

23

whereas a positive (negative) coefficient on G_PROFIT indicates that nondisclosers disclose less (more) forward-looking information in the pre-SFAS 131 period. A positive γ2 could mean that

nondisclosers anticipated the introduction of SFAS 131 and decided to already decrease the number

of forward-looking sentences on foreign operations in the pre-SFAS period in order to limit the consequences of the new rule. In contrast, a negative γ2 seems to indicate that nondisclosers were

“forced” to disclose forward-looking information in the pre-SFAS period and opted to stop doing so when they were given the option under SFAS 131 to decide whether or not to continue with geographical segment reporting on profitability. I expect both γ1 and γ2 to be negative.

It is important to keep in mind that the hypothesis that geographical segment reporting choice and voluntary disclosures are complementary, raises the likelihood that firms simultaneously decide on how many resources to allocate to forward-looking disclosures on geographical segment profitability and whether to continue or stop providing geographical profitability accounting information in their financial statements (Ball, Jayaraman and Shivakumar, 2012, 146). Therefore, I do not aim to infer causality. To reduce the chance that my results are influenced by other factors, I add various control variables to my regression model. In my choice of control variables I follow Ball, Jayaraman, and Shivakumar (2012: 146) – who test the confirmation hypothesis - and Hope and Thomas (2008: 24) - who examine the introduction of SFAS 131 in the context of empire building. I control for the number of analyst following a firm (ANALYST) since firms with more analyst following are more likely to make forecasts. Following Ball, Jayaraman, and Shivakumar (2012), I also include a dummy variable LIT for indicated industries with high litigation risk (i.e., SIC codes 2833-2836, 3570-3577, 3600-3674, 5200-5961, 7370-7374 and 8731-8734). BSSEG (reported number of line of business segments) controls for industry diversification across firms, which may affect overall firm growth and profitability (Hope and Thomas, 2008, 24). I also control for the reported number of foreign segments (FORSEG), controlling for the level of disaggregation of foreign segment reporting across firms. RATIO_FOREIGNSALES controls for the relative importance of foreign operations across the sample (Hope and Thomas, 2008, 25). I also control for the firm characteristics size (ASSETS, the natural log of assets) and leverage (LEV). Appendix A provides a detailed overview of the definitions of the variables. Industry dummies are defined at the 2-digit SIC level.

(25)

24

To test hypothesis 1b, which states that firms which give up geographical segment reporting on profitability after the introduction of SFAS 131 equally reduce negative and positive forward-looking sentences compared to firms that continue geographical segment reporting, I run two regressions with #FLS_FO_BAD and #FLS_FO_GOOD as the dependent variables:

#FLS_FO_BADi,t = γo + γ1SFAS131i,t + γ2G_PROFITi,t + γ3SFAS131*G_PROFITi,t γ4ANALYSTi,t

+ γ5ASSETSi,t + γ6LEVi,t + γ7LITi,t + γ8RATIO_FOREIGNSALESi,t + γ9BSEGi,t +

γ10FORSEGi,t + Σ INDUSTRY + ε

#FLS_FO_GOODi,t = γo + γ1SFAS131i,t + γ2G_PROFITi,t + γ3SFAS131*G_PROFITi,t

γ4ANALYSTi,t + γ5ASSETSi,t + γ6LEVi,t + γ7LITi,t + γ8RATIO_FOREIGNSALESi,t + γ9BSEGi,t

+ γ10FORSEGi,t + Σ INDUSTRY + ε

[1b] Where #FLS_FO_BAD (#FLS_FO_GOOD) is a dummy variable which equals 1 if the forward-looking sentence contains bad (good) news, and zero otherwise. In these regressions I am interested in γ1. If the interaction coefficient is statistically significant and positive in both

regressions, that would mean that nondisclosers reduce the number of forward-looking disclosures for both bad news and good news sentences.

3.3 Measurement of stock reaction to forward-looking disclosures

I use the absolute value of cumulative abnormal return (CAR) as the dependent variable to examine my second hypothesis. Following Ball, Jayaraman, and Shivakumar (2012:142) I measure the cumulative abnormal return (CAR) as the 3-day cumulative return in excess of the value-weighted market return over days -1 to +1 relative to the management forecast date (day 0). I obtain CAR from the database Eventus, by performing a cross-sectional daily event study around the earnings announcements dates (which are obtained from I/B/E/S adjusted database).

Because a number of firm characteristics may explain cross-sectional differences in firm performance, I also report results after adding several control variables that control for the firm’s information environment and firm performance. Following Ball, Jayaraman, and Shivakumar (2012: 150), I control for the market value of equity (MVE), leverage (LEV), market-to-book ratio (MB), the number of analysts following the stock (ANALYST) and the litigation risk indicator (LIT). The reported number of line of business segments (BSSEG) controls for industry diversification across firms, which may affect overall firm growth and profitability (Hope and Thomas, 2008, 24). RATIO_FOREIGNSALES (percentage of sales from foreign operations) controls for the relative importance of foreign operations across my sample firms,

(26)

25

and the reported number of foreign segments (FORSEG) controls for the level of disaggregation of foreign segment reporting across firms. Finally, I control for size (ASSETS) as it is a standard proxy for the firm’s overall disclosure level (Hope and Thomas, 2008, 24). This gives the following regression to test hypothesis 2:

ABS_CARi,t = γo + γ1SFAS131i,t + γ2G_PROFITi,t + γ3SFAS131*G_PROFITi,t + γ4MVEi,t + γ5LEVi,t

+ γ6MBi,t + γ7ANALYSTi,t + γ8LITi,t + γ9ASSETSi,t + γ10BSSEGi,t + γ11FORSEGi,t +

γ12RATIO_FOREIGNSALES + Σ INDUSTRY + ε

[2] The interaction term SFAS131*G_PROFIT is the variable of interest. I expect the coefficient on the interaction term to be positive, since a positive γ3 indicates that nondisclosers have a lower

cumulative abnormal return in the event window compared to disclosers after the introduction of SFAS 131.

3.4 Measurement of financial analysts’ forecasts

I use the accuracy of analysts’ forecast as the dependent variable to examine my third hypothesis. The financial analysts’ data was obtained from the I/B/E/S database and then matched on the basis of 9 digit CUSIPS with the Compustat data (resulting in a 63% match). Since geographical EPS is not available, I must rely on estimates of total EPS. If analysts are predicting a decline in overall firm performance, then it could very well be that the expected declining firm performance relates to foreign operations (Hope and Thomas, 2008, p. 611) since the firms in my sample have substantial foreign operations. As a proxy for the accuracy of analysts’ forecasts, I use “forecast accuracy” defined as, among others, in Bozzolan, Trombetta, and Beretta (2009), Lang and Lundholm (1996), Barron et al. (1999) and Hope (2003a):

Source: Bozzolan, Trombetta, and Beretta, 2009, p.441.

Following Bozzolan, Trombetta, and Beretta (2009), I add the number of analysts following the company (ANALYST), return on equity (ROE) and liabilities/total assets (LEV) as control variables. The number of analysts following the firm (ANALYST) is used as a proxy for the competitive pressure in offering precise forecasts and in looking for information through private channels (Bozzolan, Trombetta, and Beretta, 2009, p.450). To control for firm size, I include the

(27)

26

natural logarithm of total assets (ASSETS). Since I am testing forecast accuracy in the context of SFAS131, I also include BSSEG, RATIO_FOREIGNSALES and FORSEG. This gives me the following regression equation to test hypothesis 3:

Forecast Accuracy = γo + γ1SFAS131i,t + γ2G_PROFITi,t + γ3SFAS131*G_PROFITi,t + γ3ROEi,t +

γ4LEVi,t + γ5ANALYSTi,t + γ7ASSETSi,t + γ8BSSEGi,t + γ8FORSEGi,t +

γ9RATIO_FOEIGNSALES + ε

[3] Again, the interaction term SFAS131*G_PROFIT is the variable of interest. A positive coefficient on γ3 indicates that disclosers have higher forecast accuracy compared to nondisclosers in

(28)

27

4. Results Confirmation Hypothesis

In this chapter I first present the results for the confirmation hypothesis, which in my context predicts that firms that subsequent to SFAS 131 stop geographical segment reporting on profitability, also reduce the number of forward-looking disclosures on foreign operations compared to firms that continue with geographical segment reporting on profitability. I then show the results for hypothesis 1b, where I differentiate between positive and negative forward-looking disclosures. Finally, I will also make a differentiation between forward-forward-looking sentences which relate to sales and those which relate to profit.

4.1. Descriptive Statistics

4.1.1. Summary Statistics

Panel A of table 3 contains the summary statistics for my sample. The dependent variable #FLS_FO has a mean of 0.241, which means that the sample mean of forward-looking sentences on foreign operations per MD&A is 0.241. The average number of “good news” forward-looking sentences on foreign operations per MD&A is 0.102, while the average number of “bad news” forward-looking sentences on foreign operations is 0.125. Furthermore, the average number of forward-looking sentences related to profit is 0.10, while the number related to sales is 0.137. The average reported number of line of business segments in my sample is 2.5, while the mean amount of foreign segments reported is 3.0. Furthermore, the firms in my sample have substantial foreign operations, since 40% of their total sales can be attributed to foreign sales.

4.1.2. Differences in Firm Characteristics between Disclosers and Nondisclosers in the Pre-SFAS 131 period

To explore whether the firm characteristics of nondisclosers differed from those of the disclosers in the pre-SFAS 131 period, Panel B of table 3 shows the results of a “difference in means” test over fiscal year 1997 (which is the year before SFAS 131 became effective). Ideally I would like to find no indication that nondisclosers differ in firm characteristics from disclosers, since this would increase confidence that potential differences in disclosure of forward-looking information can be attributed to the introduction of SFAS 131 rather than to some other (exogenous) factors. The results show that while nondisclosers are significantly larger than disclosers, the nondisclosers do not differ significantly from disclosers on the other firm characteristics tested, such as leverage, number of analysts following the firm, the litigation riskiness of the industry in which they operate, or the number of line segments reported. This increases confidence that any results attributed to the

(29)

28

disclosure/non-disclosure of geographic earnings are reliable. As a robustness check, a similar test was performed for the fiscal year 1996, with qualitatively similar results. Moreover, when a similar test is run in the post-SFAS period, for instance the year 2000, I also find qualitatively similar results, indicating that potential differences in number of forward-looking disclosures are not driven by firm characteristics.

Table 3 - Summary Statistics

Sample consists of 2.090 firm-years from 1996-2000

Panel A: Descriptive Statistics per firm-year (n =2.090)

Variable Mean Std. Dev. 5th_pct _Median ₉₅th_pct

LN(FLS_FO) 0.145 0.332 0 0 0.693 #FLS_FO 0.241 0.607 0 0 1 #FLS_FO_GOOD 0.102 0.381 0 0 1 #FLS_FO_BAD 0.125 0.427 0 0 1 #FLS_PROFIT 0.100 0.346 0 0 1 #FLS_SALES 0.137 0.447 0 0 1 LN(ANALYST) 1.339 1.172 0 1.386 3.219 LEV 0.191 0.197 0 0.154 0.522 LIT 0.321 0.467 0 0 1 LN(BSEG) 0.651 0.728 0 0 1.946 LN(FORSEG) 1.024 0.389 1 1 7 RATIO_FOREIGNSALES 0.401 0.178 0.127 0.385 0.712 LN(ASSETS) 6.571 1.950 3.395 6.562 9.813

Panel B: Difference in means test for fiscal year 1997 (n=418)

Mean Mean Difference St. Error

G_PROFIT = 1 (discloser) G_PROFIT = 0 (nondiscloser) LN(ASSETS) 5.988 6.708 0.720*** 0.1998 LN(ANALYST) 1.187 1.365 0.179 0.1236 LEV 0.184 0.175 - 0.009 0.1851 LITIGATION 0.287 0.336 0.049 0.0495 RATIO_FOREIGNSALES 0.385 0.378 - 0.007 0.0184 LN(BSSEG) 0.294 0.382 0.088 0.0605 LN(FORSEG) 0.850 0.857 0.007 0.0215 EARNVOL 0.067 0.046 - 0.021 0.0129 RETVOL 0.117 0.104 - 0.013 0.0165 n 129 289

This table displays the summary statistics for the full dataset. Panel A shows the summary statistics per firm year observation. Panel B shows the results of a difference in means test for the fiscal year 1997 for various firm characteristics (two-sided t-test). *** p<0.01, ** p<0.05, * p<0.1.

(30)

29 4.1.3. Pearson Correlation Matrix

Table 4 presents the Pearson correlation coefficients between the various variables used in the regression model. The dependent variable LN(FLS_FO) is weakly correlated with control variable RATIO_FOREIGNSALES, meaning that firms with relatively more foreign sales disclose more forward-looking sentences on foreign operations in their MD&A’s. Furthermore, table 4 shows that there is no high correlation between two or more independent variables (except, by its definition, the correlation of SFAS131 and G_PROFIT with the interaction term SFAS131*G_PROFIT).

An examination of the confirmation hypothesis in the context of SFAS 131 and geographical segment reporting on profitability