Amsterdam Business School
Motives to include or exclude segmented forward looking
disclosures in the MD&A: evidence from U.S. multinationals
Name: Simone Schalkwijk Student number: 10873287
Thesis supervisors: dr. W. H. P. Janssen, dr. R. Felleg Date: 20 June 2016
Word count: 13461
Faculty of Economics and Business, University of Amsterdam MSc Accountancy & Control, specialization Accountancy Research field: financial accounting
Statement of Originality
This document is written by student Simone Schalkwijk 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.
Abstract
In addition to the assumption that firms fully disclose all private information to reduce information asymmetry, prior research proposes an agency costs and a proprietary costs incentive to report segmented forward-looking disclosures. First, proprietary costs influence the number of segmented forward-looking disclosures containing good news. Second, agency costs influence the number of segmented forward-looking disclosures containing bad news. Furthermore previous studies state that firms are constrained in the discretion of the tone of forward-looking disclosures by the risk of litigation. When earnings are volatile this results in a move towards reporting non-earnings related forward-looking information. I study the effect of these incentives on forward-looking disclosures on foreign operations using data of 418 U.S. firms from 1996 to 2000. To the best of my knowledge my research is the first to combine segment reporting with forward-looking disclosures.
I fail to find a significant relationship between proprietary costs and the amount of segmented forward-looking disclosures containing good news. I find that firms with agency costs, in opposite of my expectation, provide more segmented forward-looking disclosures containing bad news. I also fail to find a significant relationship between earnings volatility and the amount of non-earnings related forward-looking disclosures. However I do find that proprietary costs and agency costs have a significant positive relation with the amount of non-earnings related forward-looking disclosures.
Keywords: Segment reporting, forward-looking disclosures, reporting incentives, proprietary costs, agency costs, litigation costs, non-earnings related disclosures.
Contents
1 Introduction ... 5
2 Theoretical framework and hypothesis development ... 9
2.1 Forward looking disclosures ... 9
2.2 Segment reporting ... 12
2.3 Proprietary costs ... 13
2.4 Agency costs ... 15
2.5 Non-earnings related disclosures ... 17
3 Research method... 18
3.1 Sample selection ... 18
3.2 Variables ... 18
3.2.1 Dependent variable ... 18
3.2.2 Main independent variables of interest ... 19
3.2.3 Control variables ... 21
3.3 Research model... 22
4 Results... 24
4.1 Descriptive statistics ... 24
4.2 Correlations and univariate analysis ... 26
4.3 Regression analyses ... 27 4.3.1 Hypothesis 1 ... 27 4.3.2 Hypothesis 2 ... 28 4.3.3 Hypothesis 3 ... 29 5 Conclusion ... 31 References ... 34 Appendices ... 42
1 Introduction
Companies become increasingly more international and diversified. This generates an increased demand for information to evaluate the performance of a firm. Specific information on the segments of firms is important because performance, risk and potential growth might differ significantly between geographical areas and businesses (FASB, 1997). The basic assumption that serves the incentive for managers to disclose all privately known information is the reduction of information asymmetry and lower cost of capital (Diamond & Verrecchia, 1991; Healy & Palepu, 2001; Milgrom, 1981; Verrecchia, 2001). Disclosing forward-looking information is known to offer investors additional information about the current state and future prospects of the firm. Therefore investors are better able to predict future cash flows and consequently investors are better able to correctly valuate the firm (Hussainey, Schleicher, & Walker, 2003; Hussainey & Walker, 2009). Disclosing segmented information is known to increase the understanding of investors about the types of business activities that a firm performs and the economic environments in which a firm does business (FASB, 1997). This should help users of the financial statements to get a better grip on the performance of the firm, to make it easier to predict the future cash flows of the firm and to allow them to make more informed judgments about an entire firm (FASB, 1997). This led me to focus my study on segmented forward-looking disclosures as an opportunity to provide additional information.
Several incentives exist for managers to disclose or withhold these types of additional information. Prior research shows that the decision to disclose additional information is based on a trade-off between the benefits of reducing asymmetry and the costs of informing the competition with important strategic information (Kent & Ung, 2003). In this thesis I examine proprietary costs, agency costs and litigation risk incentives for managers to disclose or withhold segmented forward-looking disclosures. This research is motivated by the increased attention of the Financial Accounting Standards Board (hereafter, FASB) to develop accounting standards that support high quality forward-looking disclosures (Bozanic, Roulstone, & Van Buskirk, 2013; Securities and Exchange Commission, 2003, 1989). Furthermore the Securities and Exchange commission (hereafter, SEC) views the current quality of forward-looking disclosures as insufficient (Cole & Jones, 2005). My study is relevant as it aims to identify if the proposed incentives indeed play a role in the decision to disclose segmented forward-looking information. If so, standard setters might consider to adjust the guidelines on forward-looking disclosures to prevent these incentives from causing information asymmetry.
Another motivation for this study is a gap in the literature. Previous studies have examined incentives to disclose additional information in both management forecast settings and segment reporting settings. Mixed results came out of these studies. With regard to the literature on forward-looking disclosures managers have a tendency to keep bad news private as a consequence of agency problems (Kothari, Shu, & Wysocki, 2009). This leads to disclosure levels that are not aligned with the preferences of shareholders (Kothari et al., 2009). However when expected litigation costs are high or reputation damage is foreseen, managers will be more conservative in their forward-looking disclosures, resulting in a tendency to disclose bad news and withhold good news (Trueman, 1997). With regard to the proprietary costs incentive Luo, Courtenay and Hossain (2006) find that proprietary costs weaken the positive relation between the voluntary disclosure level of the firm and the amount of information on future prospects of their stock return. This suggests that the presence of proprietary costs limits the amount of forward-looking information relative to the total of voluntary disclosures of the firm. However Kent and Ung (2003) fail to find a significant effect of proprietary costs on the amount of earnings-related forward-looking disclosures in the annual report.
With regard to segment reporting the incentives to provide additional information are also agency costs and proprietary costs. Berger and Hann (2007) and Bens, Berger and Monahan (2011) find that managers withhold segmented information that shows lower abnormal profits. Only Leung and Verriest (2014) find no evidence that agency costs influence the extent of aggregation in the reporting of segments. Regarding proprietary costs much more evidence of withholding segment information exists. Hayes and Lundhom (1996) find that if companies have two segments with disparate results, they will report only aggregate information to prevent revealing proprietary information to competitors. Additionally Nagarajan and Sridhar (1996) found that firms tend to mix sensitive segmented information with other information when they are required to provide segment disclosures to deter entry of potentially new competitors. Piotroski (1999) finds evidence that firms that experience declining profitability and that have less variability in cross-segment profitability are more likely to increase information on segments. Leuz (2003) concludes that when a market is easy to enter by competitors, when cross-segment profitability is varied and when firms perform better it is less likely that managers voluntarily disclose segment data. Similar results are found by Leung and Verriest (2014) and Bens et al. (2011). As with the agency costs incentive again some researchers fail to find a significant effect (Berger & Hann, 2007; Botosan & Stanford, 2005; M. L. Ettredge, Kwon, Smith, & Stone, 2006). While the proprietary costs incentive is extensively reviewed in the segment disclosure setting, it is not well documented in the literature on forward-looking disclosures. Furthermore the agency
cost incentive does not always proof to be of significant influence in the decision to disclose additional information. The litigation risk incentive is not studied in the context of segment disclosures. This might be because managers can exercise less discretion in the reporting of segments since these disclosures are largely regulated. Hence this study provides new insights in the incentives to disclose additional information in the form of segmented forward-looking disclosures.
To examine the proprietary costs, agency costs and litigation risk hypothesis I perform an analysis on forward-looking sentences (hereafter, FLS) on foreign operations in the management discussion and analysis (hereafter, MD&A) section of the 10K filings of 418 U.S. firms from year 1996 to 2000. These FLS are subtracted by automatic language processing software and analysed using a logistic ordinary least squares (hereafter, OLS) model. The research model used is that of Alfonso, Hollie and Yu (2012) modified to the setting of forward-looking segment disclosures. This means that three variables are excluded and the variable to proxy for proprietary costs is replaced. I predict that proprietary costs influence the number of segmented forward-looking disclosures containing good news negatively and that agency costs influence the number of segmented forward-looking disclosures containing bad news also negatively. Furthermore I predict that firms experiencing litigation risk move towards reporting non-earnings related forward-looking information. However I fail to find a significant relationship between proprietary costs and the amount of segmented forward-looking disclosures containing good news. I find that firms with agency costs, in contrast to my expectation, provide more segmented forward-looking disclosures containing bad news. I also fail to find a significant relationship between litigation risk and the amount of non-earnings related forward-looking disclosures. However I do find that proprietary costs and agency costs have a significant positive relation with the amount of non-earnings related segmented forward-looking disclosures.
My results contribute to existing literature in several ways. First, my results add to the existing literature as there is scarcity in the literature with regard to the proprietary incentive to disclose additional information in the management forecast setting. Although the agency costs hypothesis is extensively studied in both management forecast setting and segment reporting settings, results are a mixed. Furthermore little is known about the incentives to disclose non-earnings related forward-looking information. My results show that proprietary and agency costs are determinants for the decision to disclose non-earnings related segmented forward-looking disclosures. To the best of my knowledge this is never studied before. Last, the examination of forward-looking disclosures in the MD&A including segment characteristics provides new insights in the existing incentives to disclose additional information. This is because
forward-looking information in the MD&A is accompanied by backward-forward-looking information in the annual report. This makes these segmented forward-looking disclosures inherently more credible (Ball, Jayaraman, & Shivakumar, 2012). My results indicate that managers might prefer to reduce information asymmetry above reducing proprietary costs an agency costs which is contradicting to the results of studies examining segment disclosures and forward-looking disclosures in isolation. My results therefore do not indicate that there is need for increased regulation for segmented forward-looking disclosures.
The remainder of this thesis is organized as follows. Chapter 2 provides a theoretical framework and explains how the hypotheses are derived. Chapter 3 outlines the sample selection, data and methodology. Chapter 4 presents the findings. Lastly chapter 5 concludes.
2 Theoretical framework and hypothesis development
An extensive body of literature exists on the incentives of managers to disclose additional information in a quasi-mandatory or quasi-voluntary setting. Because forward-looking information is the most significant information that firms disclose voluntary this will be the starting point of my theoretical framework. The remainder of this chapter focuses on segment disclosures as a way to provide or withhold additional information and provides accompanying hypotheses.
2.1 Forward looking disclosures
Early theories predict that managers voluntary disclose information because of market valuation concern (Diamond & Verrecchia, 1991; Healy & Palepu, 2001; Milgrom, 1981; Verrecchia, 2001). The MD&A section of the annual report provides one way for managers to offer investors additional information about the current state and future prospects of the firm. For public companies it is obligatory to include a MD&A in their 10-K filings (Securities and Exchange Commission, 1980). As argued by Tavcar (1998), the MD&A component of the financial section is the most important part and is most read by outsiders. Knutson (1992) states that financial analysts in United States mostly rely on the MD&A in their analyst forecast. Since the content of the MD&A remains largely unregulated this disclosure setting can be seen as quasi-mandatory or quasi-voluntary (Beyer, Cohen, Lys, & Walther, 2010). Earlier studies have shown that management forecasts are the most informational of all accounting sources (Beyer et al., 2010). A number of studies have shown that these disclosures can provide credible information for investors about the future performance of a firm (Bryan, 1997; Clarkson, Kao, & Richardson, 1994, 1999). Barron, Kile and O’Keefe (1999) examine the predictive value of MD&A disclosures by analysing the relationship between the quality of MD&A sections as rated by the SEC and the quality of analysts’ forecasts in terms of errors and dispersion in the years 1987, 1988 and 1989. They find that the accuracy of analyst forecasts increases as the amount of forward-looking disclosures on investments and operations increases. Additionally Hussainey, Schleicher and Walker (2003) find that earnings-related forward-looking disclosures are helpful for investor to predict the earnings of the firm in the following year more accurately. Hussainey and Walker (2009) conclude that the cumulative earnings of a firm in the following three years are also better to predict for investors when forward-looking information of earnings is disclosed. Disclosing forward-looking information in the MD&A has the advantage that the credibility of these disclosures is inherently higher because the information complements the audited backward-looking information from the annual report (Ball et al., 2012).
The quantity and quality of forward looking disclosures in the MD&A is however not straight forward and has been debated by regulators and users (Securities and Exchange Commission, 2003; Tavcar, 1998). The SEC supports the need for more high quality forward-looking disclosures and has provided some guidelines about the content of MD&A sections. These guidelines stimulate firms to report material issues that could impact firm’s performance in terms of liquidity, capital resources or future operations like trends, events, obligations, plans, and risks (Securities and Exchange Commission, 2003, 1989). The SEC set three requirements for the disclosure of forward-looking information (Securities and Exchange Commission, 2003). First, the circumstance should be “presently known” by management. Second, the circumstance is “reasonably likely” to occur according to management. Lastly, the management expects “material effects”. However, managers have considerable discretion in their decision to disclose forward-looking information since both the terms “reasonably likely” and “material” are not clearly defined by the SEC (Li, 2010). These guidelines are also not found to be effective since MD&A disclosures are still deficient according to the SEC (Cole & Jones, 2005). More specifically, boilerplate statements and immaterial details in general overrule substantial information (Securities and Exchange Commission, 2003). Several other reasons support the concern that forward-looking MD&A disclosures might not be informative (Muslu, Radhakrishnan, Subramanyam, & Lim, 2014). First, the information provided in the MD&A is in general not timely. Other channels like press releases and management forecasts are timelier. Second, the content of the MD&A is only reviewed by auditors on consistency with the annual financial statements of the firm so the information is less credible and could be biased. The reliance on these disclosures for investors is therefore less. Furthermore because future performance is uncertain, it is difficult to provide accurate information.
One incentive for managers to disclose private forward-looking information, positive or negative, is the reduction of information asymmetry and the potentially lower cost of capital of the firm (Diamond & Verrecchia, 1991; Healy & Palepu, 2001; Verrecchia, 2001). However managers also use discretion in choosing to disclose good or bad news to meet the information demand of investors and to adjust their results to reach the level of their forecasts (Kasznik, 1996). Managers have a tendency to keep bad news private as a consequence of agency problems leading to disclosure levels that are not aligned with the preferences of shareholders (Kothari et al., 2009). This type of behaviour of management is in line with impression management literature. Impression management as defined by Clatworthy and Jones (2001: 311) is the attempt by management ‘to control and manipulate the impression conveyed to users of accounting information’. Yuthas, Rogers and Dillard (2002) propose that impression management is a way
to strategically manipulate the behaviour and perceptions of stakeholders. In practice the theory of impression management should lead to a biased positive tone in disclosures as managers hope that the rosy presentation of the firm will lead to an increase in remuneration and job security (Clatworthy & Jones, 2003). Empirical evidence shows that managers also opportunistically release good news before raising capital or extracting rents to increase stock prices (Aboody & Kasznik, 2000; M. H. Lang & Lundholm, 2000). Kothari et al. (2009) explore stock price reactions to examine the relative timeliness of good and bad news disclosures. Their research concludes that managers reveal good news in a timely fashion but withhold bad news based on career concerns.
When managers withhold bad news disclosures or disclose good news which is later found to be overly optimistic, the firm is expected to incur litigation cost (Trueman, 1997). The model of Trueman (1997) predicts that the magnitude of the bias of the tone varies with the expected cost of litigation. Managers are constrained because outsiders have the ability to access if the assertions of management are realistic true the subsequent financial report (Rogers & Stocken, 2005). When outsiders discover the positive disclosure bias later, the disclosure reputation of the manager might also suffer (Rogers & Stocken, 2005). Therefore a rational manager will weight off the expected benefits from inflated positive disclosures against the possibility that his reputation will reduce (Rogers & Stocken, 2005). As argued by Skinner (1994) and Graham, Harvey and Rajgopal (2005), the reputation is more harmed when earnings surprise in a negative way than when the earnings surprise in a positive way. This difference in impact on reputation can result into a decrease in liquidity and share price. This might also be followed by a decrease in the bonus and job security of management. So when expected litigation costs are high or reputation damage is foreseen, managers will be more conservative in their forward-looking disclosures, resulting in a tendency to disclose bad news and withhold good news (Trueman, 1997).
Managers might also face incentives to reduce releasing private forward-looking information because of proprietary costs (Verrecchia, 1983). These costs arise when competitors use the proprietary information to copy the strategy of the firm. This will lead to increased competition for the disclosing firm resulting in lower profits. The response of traders on perceived withholding of forward-looking information is assumed to be less negative when a firm operates in a highly competitive market because traders are aware of the increased proprietary costs (Verrecchia, 1983). Empirical literature on the effect of proprietary costs on forward-looking disclosures is limited. Kent and Ung (2003) study the impact of proprietary costs on the amount of earnings-related forward-looking disclosures in the annual report. Their sample includes the
annual reports of 100 Australian companies in 1991 and 1992. They find no significant effect. Luo, Courtenay and Hossain (2006) find that proprietary costs weaken the positive relation between the voluntary disclosure level of the firm and the amount of information on future prospects of their stock return. This suggests that the presence of proprietary costs limits the amount of forward-looking information relative to the total of voluntary disclosures of the firm. So according to previous studies firms have multiple incentives to provide or withhold forward-looking information. Most of these studies focus on agency problems as an incentive to disclose or withhold forward-looking information. One basic assumption is that firms disclose all private forward-looking information to reduce information asymmetry. However firms also have incentives to present the firm in such way as to impress outsiders. This assumes a biased positive tone in disclosures. On the other side litigation costs remediate the opportunity to bias the tone and the provision of information. Firms with proprietary costs might withhold information from competitors to protect their competitive position, even if they have to give up transparency and lower cost of capital. One of the assumptions that distinguishes the proprietary cost hypothesis from other incentives to disclose voluntarily is the assumption that there are no agency problems (Healy & Palepu, 2001). This means that the proprietary cost hypothesis predicts that voluntary disclosures are always credible. This hypothesis therefore focuses on economic factors that constrain managers to disclose all available information.
2.2 Segment reporting
A second opportunity for firms to provide useful information to investors to better predict future cash flows is segment reporting. Segment reporting entails, as described in Financial Accounting Standards (SFAS) No. 131 issued by the FASB, the reporting of certain information on operating segments in separate sets of financial statements (FASB, 1997). This includes information on products and services that the firm delivers, the geographic areas in which a firm does business and the biggest customers of the firm. This is required for public business enterprises. The objective of segment reporting is to provide aggregated information on the types of business activities that a firm performs and the economic environments in which a firm does business (FASB, 1997). This should help users of the financial statements to get a better grip on the performance of the firm, to make it easier to predict the future cash flows of the firm and to allow them to make more informed judgments about an entire firm (FASB, 1997). The Association for Investment Management and Research (hereafter, the AIMR) states in their 1992 position paper also the importance of segment data (Knutson, 1992). They argue that it is essential for investors to create the knowledge and understanding of the economic behaviour of
the different segments of a multifaceted firm. Without aggregation it is difficult for investors to predict the future cash flows of a firm (Knutson, 1992). This is because one weak segment could lead to the liquidation of a whole firm. The association also wrote in their position paper that even if weaknesses are absent, the dissimilar streams of cash flow between different segments implies disparate risks and brings out unique values. This implies that there is need for segmented forward-looking disclosures.
Before the amendment of Financial Accounting Standards No. 131, the Statement of Financial Accounting Standard (SFAS) No. 14 prescribed firms to provide segmented information on profits. Firms had to disclose segment information by line-of-business and geographic area linked to their industry. With this way of accounting there was no direct link between disclosed segment information and the information that was used for internal decision-making and the internal organization of the firm (Herrmann & Thomas, 2000). This approach caused firms to report much less segment information than what was reported internally because there was a lot of discretion in the way industry was defined. The standard itself already addresses this issue as noted in the paragraph purpose of segment information (SFAS 14, para 76):
“Information prepared in conformity with this Statement may be of limited usefulness for comparing a segment of one enterprise with a similar segment of another enterprise.”
Due to lobbying of the AIMR the FASB changed the method of accounting for segment profits (Street, Nichols, & Gray, 2000). The FASB issued the Statement of Financial Accounting Standards No. 131 in 1997 which became effective for fiscal periods beginning after December 15, 1997 (FASB, 1997). According to Ettredge, Kwon, Smith and Stone (2006) the new standard led to increased transparency in the profitability of the segments reported. However firms with high proprietary costs are still able to withhold information on differences in segment profitability because of competitive harm concerns.
2.3 Proprietary costs
As previously described, proprietary costs may play a role in the decision to report certain forward-looking information (Kent & Ung, 2003). The FASB (2001) points out three factors in information disclosure that impact the degree of proprietary costs. Firstly, it is the type of information disclosed. Secondly, the level of detail of the disclosure impacts the extent of proprietary costs. Lastly, the timing of disclosure is a determinant of the proprietary costs of a firm. In this research I specifically look at the effect of proprietary costs of firms on the amount of detail in the disclosure of forward-looking information by firms.
The biggest change between SFAS 131 and SFAS 14 is the requirement that information should be disclosed on a country-level basis. SFAS 14 prescribed firms to report information by geographic segment. However firms could use discretion in the definition of geographic segment. Therefore the majority of firms reported their foreign operations by groups of countries. This aggregation could be by continents, multi-continents and hemispheres for example. Motivation for this particular change is that country-level disclosures are more easy to interpret and more useful for analysts to determine the international risks that firms are exposed to (SFAS 131, para 105). Country-level disclosures represent the highest level of detail of the foreign performance of a firm. Disclosing high level aggregated segment information could reveal important strategic information (Ettredge, Kwon, & Smith, 2002). Elliot and Jacobson (1994) argue that firms reporting abnormal segment revenues might be harmed by competitors who follow the strategy of the successful firm. This leads to increased competition and more pressure on profits. According to Ettredge et al. (2002) firms that are perceived to suffer most competitive harm from making segmented disclosures are more likely to lobby against the exposure draft. Tsakumis, Doupnik and Seese (2006) support this incentive to lobby as they find that firms that experience greater competitive harm associated with segment disclosure, provide less information on the revenues of foreign operations on a country level basis.
Several other studies find that proprietary costs affect the reporting of segments of firms. Hayes and Lundhom (1996) analyse a sample of nearly 7,000 Compustat firms in the time period 1985-1991 and find that if companies have two segments with disparate results, they will report only aggregate information to prevent revealing proprietary information to competitors. Additionally Nagarajan and Sridhar (1996) found that firms tend to mix sensitive segmented information with other information when they are required to provide segment disclosures to deter entry of potentially new competitors. Piotroski (1999) finds evidence that firms that experience declining profitability and that have less variability in cross-segment profitability are more likely to increase information on segments. Leuz (2003) examines a sample of German firms and concludes that when a market is easy to enter by competitors, when cross-segment profitability is varied and when firms perform better it is less likely that managers voluntarily disclose segment data. Similar results are found by Leung and Verriest (2014). Lastly Bens et al. (2011) also find that proprietary costs drive the level of aggregation of segment disclosures using the plant level of firms in confidential US Census data.
Some studies find no effects of proprietary costs on management’s incentive to provide segmented information. Berger and Han (2007) examine the reporting of segmented information of lines of businesses by U.S. companies. The empirical results of their study are not supportive
of the assumption that managers withhold segmented information when profits are abnormal. Botosan and Harris (2000) explore the effect of a competitive environment on the issuance of quarterly reports including segmented information. The analysis is performed on 107 multi-segment firms in the U.S. that report on three or more business multi-segments in their annual report of the years 1987-1994. They found no reliable differences in the quarterly reports of firms in different competitive environments.
Most studies find that firms experience competitive harm when they have to provide additional information that is more detailed for example in segment disclosures. Several of these researchers state that especially the separate presentation of high performing segments is sensitive to proprietary costs. This is because competitors can use this information to enter these profitable segments. The literature on forward looking disclosures is scarce on the proprietary costs incentive of disclosure. However Verrechia (1983) does indicate that investors react less negative to the withholding of private information when they know that the firm is experiencing proprietary costs. This implies that managers can reduce competitive harm by keeping proprietary information private with relatively less consequences on the reaction of investors. Therefore I expect that managers have an incentive to report less segmented forward-looking disclosures containing good news when they are subject to proprietary costs. Accordingly I state my first hypothesis as follows:
H1: The amount of segmented forward-looking disclosures containing good news decreases as proprietary costs of the firm increase.
2.4 Agency costs
Segment information is an example of a type of detail-related disclosure that, next to proprietary costs, could lead to increased agency costs for a firm (Elliott & Jacobson, 1994). Various studies find that multi-segment firms have a higher diversification discount relative to firms operating in a single segment (Berger & Ofek, 1995; Lamont, 1997; L. H. Lang & Stulz, 1993; Shin & Stulz, 1998). Similar results found in these studies imply that internal funding within diversified firms in terms of lines of businesses is suboptimal when funds are transferred across segments. In general managers use discretion in choosing investment projects (Stulz, 1990). This leads to too little investments in projects with a positive net present value and too much investment in poor performing projects (Stulz, 1990). This type of dysfunctional behaviour is caused by information asymmetries between managers and shareholders and inefficient use of free cash flows (Stulz, 1990). According to Berger and Ofek (1995) the value of a firm can change as a result of a firm’s diversification strategy. They found that dysfunctional behaviour in terms of segment
investments decisions decreases firm value. Second they found a decrease in firm value when segments are cross subsidized. This means that low performing segments are funded by high performing segments. Lamont (1997) finds similar results and states that funds are allocated in a suboptimal manner when segments are cross-subsidized. The author concludes this after a study on the capital expenditures of 26 diversified, multi-segment firms that were affected by a macroeconomic shock in the years 1985 and 1986. Finally the study of Shin and Stulz (1998) adds to these statements by analysing the investments of the smallest and largest segments of a sample of Compustat firms in the time period 1980 to 1992. They show that the investments of a certain segment of a firm depend significantly on the performance and cash flows of other segments within the firm. So segments often cannot operate independently.
These agency problems influence the extent of withholding segmented information by managers. Berger and Hann (2007) analyse a sample of 796 U.S. firms by restating and comparing the segment disclosures of the firms between the pre- and post-adoption period of SFAS No. 131 in order to analyse the motives to report excess or low segmented profits. The results of their study point out that managers withhold segmented information that shows lower abnormal profits. Greater disclosure might reveal underperformance which could be seen as value destruction and therefore the chance of increased corporate governance and control mechanisms to monitor and discipline the underperforming manager rises. Bens et al. (2011) find similar results. Only Leung and Verriest (2014) find no evidence that agency costs influence the extent of aggregation in the reporting of segments.
While previous studies focus on the effect of agency costs on segmentation in lines of business Hope and Thomas (2008) focus on the effect of segment reporting on the behaviour of managers. They analyse a sample of firms that were no longer required to report earnings by geographic segments under SFAS 131. They find that firms that stopped disclosing earnings by geographic area experienced a decrease in firm value, an increase in foreign sales and a decrease in the profit margin of foreign countries compared to firms that continued to report earnings by geographic areas. They conclude that less transparency in the reporting of geographic segments enables managers to engage in value-destroying empire building because of difficulties in the monitoring of the firm due to absence of segmented information. These findings support the agency costs incentive to withhold segmented forward-looking disclosures.
Most studies in both the forward-looking disclosures literature and the segment reporting literature indicate that agency costs lower the reporting of additional information. According to the studies concerning segment disclosures non-disclosure might occur because managers do not
want to present the detailed information of poor performing segments to investors. This is consistent with agency costs incentives to report forward-looking information which also predicts that managers tend to withhold or delay the disclosure of bad news. This leads to the expectation that managers will withhold segmented forward-looking bad news because revealing bad performance decreases numeration and job security and might increase external monitoring. Therefore I state the following hypothesis:
H2: The amount of forward-looking disclosures containing bad news decreases as the agency costs of the firm increase.
2.5 Non-earnings related disclosures
Previous research focused primarily on earnings-related forward-looking disclosures. However not all forward-looking disclosures are related to earnings. The study of Bozanic et al. (2013) specifically examines the incentive to report non-earnings related disclosures. They state that managers disclose more non-earnings-related relative to earnings-related forward-looking information when investor uncertainty is high. This is because managers are less confident about future earnings in unstable periods. Therefore they prefer to provide information over which they have control like expansion of factories or projected investments in R&D.
The incentive to disclose non-earnings related disclosures seems to be driven by litigation costs (Kent & Ung, 2003). With respect to segment reporting, litigation occurs when segments are aggregated while these should be disaggregated. With respect to forward-looking disclosures, the risk of litigation increases when earnings become more volatile because future performance will be harder to predict then. According to Kent and Ung (2003) firms experiencing volatile earnings report less earnings related disclosures because of litigation concerns. This suggest, together with the results of Bozanic et al. (2013), that firms prefer to provide forward-looking information over which they have control. This could be for example reporting plans for future investments. This can be seen as a non-earnings related disclosure which has little impact on the risk to be litigated. Therefore I expect that firms with higher litigation risk disclose more non-earnings related segmented forward-looking disclosures and therefore I state my third and last hypothesis as follows:
H3: The amount of segmented non-earnings related forward-looking disclosures increases as the litigation risk for the firm increases.
3 Research method
This section outlines the sample selection, the data and the methodology.
3.1 Sample selection
The dataset consists of 418 US firms that engage in geographical segment reporting from 1996 to 2000 resulting in 2,090 observations. The criteria for firms to be included in the analysis is that firms had to report at least two geographical non-US segments for each of the five years. The dataset is created in accordance with the research project of dr. W.H.P. Janssen. Except for the dependent variable, data was collected from Compustat. The number of observations dropped from 2,090 to 2,065 due to unavailable data on total assets and total liabilities for 25 firms in 2001. This data was necessary to construct the measure of litigation risk.
3.2 Variables
3.2.1 Dependent variable
To collect data on segmented forward-looking disclosures I joined a research project led by dr. W.H.P. Janssen. The amount of FLS on foreign operations is used as the dependent variable to proxy for the amount of segmented forward-looking disclosures. For the selection of the FLS on foreign operations automatic language processing software was used. This software extracted these sentences from MD&A sections of 10-K filings for US multinationals from 1996-2000. The first requirement for extraction states that the sentence should be forward-looking. This means that the sentence should contain any of the words “will”, “should”, “can”, “could”, “may”, “might”, “expect”, “anticipate”, “believe”, “plan”, “hope”, “intend”, “seek”, “forecast”, “objective”, “goal” or “project” (Li, 2010). The next requirement for extraction states that the sentence should be about a foreign country or region. This means that extraction was based on a string out of country list, region list, country adjusted list and region adjusted list all provided by ResearchGate. List items related to the United States were deleted from the initial ResearchGate lists. The third and last requirement for extraction states that the sentence should be about operations. According to sales operations this involves the words “revenue” and “sales”. For costs the FLS should include “cost” or “expense” or the strings: “SG&A” and “COGS”. Furthermore foreign operations could lead to profits. These sentences are extracted if the sentence contains: “profit”, “loss”, “earnings”, “income”, “margin”, “EBIT”, “EBITDA”, “EBITA” or “EPS” or it contains a sequence of strings: “operating results”, “results of operations” or “operating performance”. Lastly the words “expenditure”, “invest”, “investment”, “acquire” and “acquisition” were used
to filter relevant sentences as these indicate operations on investments. Although the automatic language processing software extracted relevant sentences, there were a lot of unsuccessful matches. That is why these sentences were classified manually as successful vs. unsuccessful matches based on the previously described requirements. Lastly, positive, negative and neutral/no news sentences were distinguished manually.
Based on the hypotheses the FLS were categorized in different types of FLS. First, the positive FLS involve an increase in sales/profits and a decrease in costs. Second, the negative FLS involve a decrease in sales and profits and an increase in costs. The last category is the non-earnings related FLS. These involve FLS on investment and disinvestment plans.
3.2.2 Main independent variables of interest
The first main independent variable of interest is included to test hypothesis 1 and serves as a proxy for proprietary costs. It is also included in the research model of hypothesis 2 and 3. Previous researchers examining segment reporting use the Herfindahl index or an industry concentration ratio to proxy for proprietary costs (Berger & Hann, 2007; Botosan & Stanford, 2005; Harris, 1998). The general assumption arising in these studies is that firms in concentrated markets earn abnormal profits and therefore are perceived to have greater competitive harm in disclosing forward-looking segment disclosures. Tsakumis, Doupnik & Seese (2006) state that potential competitive harm is not dependent on industry concentration ratios because these ratios ignore the dependency of an individual firm’s revenues related to geographic region. This is highly relevant in considering segmented forward-looking disclosures (Tsakumis, Doupnik, & Seese, 2006). To measure proprietary costs I follow the approach of Tsakumis et al. (2006). They identify a proxy for exposure to proprietary costs that measures the firm-specific impact of geographic pressures faced by firms of making detailed segment reconciliations on foreign operations. This measure (FOR_PCT) is calculated as the percentage of an individual firm’s total sales accounted for by its foreign sales (foreign sales/total sales). When the relative part of foreign sales increases to the total amount of sales of the firm, the firm is expected to have greater exposure to competitive harm when disclosing forward-looking information on foreign operations. Firms with higher values on FOR_PCT are expected to have higher proprietary cost from reporting forward-looking information on foreign operations. Therefore the expected sign of the coefficient on FOR_PCT is negative for the research model of hypothesis 1. Based on the model of Verrechia (1983) proprietary value is not only limited to positive disclosures. His model does not specify a specific type of information. Therefore the variable FOR_PCT is also
included in the research model of hypothesis 2 and 3 with the expectation that proprietary costs will lower the amount of segmented forward-looking disclosures.
The second main independent variable is a proxy for agency costs to test hypothesis 2 and is also included in the research model of hypothesis 1 and 3. This variable is a dummy variable equal to 1 if current period earnings are negative and 0 if earnings are zero or positive. I label this variable LOSS. This variable is based on the research of Schleicher and Walker (2010) who suggest that it is a major disappointment for investors if firms report a loss. Especially when this loss is considered to be permanent by outsiders. A permanent loss raises questions on the capabilities of the management team to continue the firm as a going concern. Therefore the manager’s main concern will be to convince investors that the strategy of the management team is still suitable to generate positive rewards in the future. So managers will try to avoid the reporting of a loss which might result in increased asymmetry because losses encourage managers to bias the tone of forward-looking disclosures upwards. This information asymmetry is expected to increase agency costs. Because of the increased likelihood of financial distress within loss firms, detailed segment disclosures will also help the market to assess the firm’s viability as a going concern. Accordingly, managers of loss firms will avoid detailed disclosure of segment information (Alfonso, Hollie, & Yu, 2012; Hope & Thomas, 2008). This situation is very different in firms making profits because their financial results already show that the firm is managed competently. So profit firms will put less emphasis on communicating positive prospects through forward-looking disclosures or have less incentives to withhold segmented information (Alfonso et al., 2012; Schleicher & Walker, 2010). According to the results of these previous studies I expect a negative sign on the coefficient of LOSS in the research of model 2. In the first model the sign of the coefficient on LOSS is expected to be positive because Schleicher and Walker (2010) state that firms with losses bias the tone of forward looking disclosures upwards. In the third model the sign of this coefficient is expected to be negative because future performance is harder to predict for loss firms (Collins, Maydew, & Weiss, 1997; Hayn, 1995). So non-earnings related forward-looking disclosures are more credible in these kind of situations. Common practice is to use the cross subsidization of segments to proxy for agency costs (Alfonso et al., 2012; Berger & Hann, 2007). Due to resource constraints however the dummy for loss firms is despite the crudeness the best compromise to proxy for agency costs.
The third main independent variable is a proxy for litigation risk to test hypothesis 3 and is also included in the research model of hypothesis 1 and 2. Following Kent and Ung (2003) I assume that the risk of litigation is higher when earnings are more volatile. The formula below represents the calculation of this variable.
𝐸𝐴𝑅𝑁_𝑉𝑂𝐿𝑡= [𝐸𝑞𝑢𝑖𝑡𝑦(𝑡+1)− 𝐸𝑞𝑢𝑖𝑡𝑦(𝑡)] + [𝐷𝑒𝑏𝑡(𝑡+1)− 𝐷𝑒𝑏𝑡(𝑡)] Total Assets
Following Bozanic et al. (2013) I expect that firms with more volatile earnings provide more segmented non-earnings related forward-looking disclosures. So I expect a positive sign for the coefficient of EARN_VOL in research model 3. In model 1 and 2 I expect a negative sign for the coefficient of EARN_VOL because Kent and Ung (2003) report that firms with higher earnings volatility provide less earnings related disclosures.
3.2.3 Control variables
To control for other relationships than between proprietary costs, agency costs, investor uncertainty and the number of FLS on foreign operations, control variables are added to the research model. Following Alfonso et al. (2012) I include the number of business segments reported (BUS_SEG) and the number of geographic segments reported (GEO_SEG) to control for the complexity of the firm. It is reasonable that the number of segments in the firm affects the disclosing behaviour of managers as segmented FLS can only exist in firms with reportable segments. Therefore the expectation is that there is a positive relation between the amount of segments and the number of segmented FLS on foreign operations. The number of financial analysts following the firm is also expected to influence the extent of segmented forward-looking disclosures as Alfonso et al. (2012) state that the change in the accounting standard for segment reporting was mainly driven by financial analyst. They find that the number of financial analysts following the firm is significantly positively associated with the level of aggregation in the reporting of the segments of the firm. Therefore I include the variable number of financial analysts following the firm (FIN_FOL) to control for this effect. According to Buzby (1975) and Diamond and Verrecchia (1991) larger size firms have a higher disclosure level overall. This leads to the inclusion of the variable firm size (LN_ASSETS), which is the natural logarithm of firm’s assets. The expected sign of this coefficient is positive for all three models. Furthermore the variable leverage (LEVERAGE) is included which is the ratio of total debt to total assets of the firm. According to Jensen and Meckling (1976) firms with more leverage have a higher level of monitoring by creditors and therefore have larger agency costs. Additional disclosures are essential role for highly leveraged firms to reduce these costs because they help reassuring creditors that the firm will not bypass their covenant claims (Hossain, 1994). Following is a dummy variable (BIGN) which is included to control for the type of auditor that the firm chose. This variable takes on the value 1 if the firm is audited by a Big N audit firm in a given year and takes on 0 otherwise. This variable is included because prior literature argues that the auditor has important influence in the quality of the reporting of the firm (Hail, 2002). Empirical results
suggest that firms reviewed by large audit firms on average provide financial statements with higher quality than firms reviewed by small audit firms (Becker, DeFond, Jiambalvo, & Subramanyam, 1998; McNally, Eng, & Hasseldine, 1982). The previously described variables all are expected to lead to more segmented forward-looking disclosures on foreign operations overall.
The last control variable is return on assets (ROA). Previous research reports that managers try to protect revealing proprietary information, for example abnormal profits, to competitors by withholding segmented information (Harris, 1998; Hayes & Lundholm, 1996). This implicates that profitable firms will report less segmented forward-looking disclosures since segmented information has more proprietary value. Firms with low profits perceive to experience less competitive harm by disclosing segmented forward-looking information. For hypothesis 1 I expect a negative relation between ROA and the amount of positive segmented FLS. Following Berger and Hann (2007) I expect a positive coefficient for ROA in the research model of hypothesis 2. Berger and Hann (2007) find that firms with higher agency costs try to hide segments that are less profitable by providing less segmented information in the pre-SFAS No. 131 period. Firms with a high return on assets are more likely to have segments that are well performing because they are relatively less asset intensive or generate relatively high profits, so managers have less incentives to withhold segmented information. Managers of a firm with a lower ROA use more discretion in reporting segmented information to hide poor performance (Alfonso et al., 2012). This is also consistent with the results of Hope and Thomas (2008). In the third research model ROA is included to prevent an omitted variable bias since ROA is correlated with LOSS.
3.3 Research model
To estimate the number of FLS on foreign operations I make a few adjustments in the model of Alfonso et al. (2012). In accordance with Tsakumis et al. (2006) I use the variable percentage foreign sales (FOR_PCT) to proxy for proprietary costs instead of the herfindahlindex (HERF) that Alfonso et al. (2012) use. I do not include the variables industry-adjusted return on sales (ROS) and accruals (ACC) since they are more an indicator of backward-looking segment disclosures. The variable growth (MB) proxied by the market value of equity divided by the book value of equity is excluded because data on the market value of equity is not available for the first two years of my sample.
The variables are included in the following ordinary least squares (OLS) model. This model will test all three hypotheses using different types of FLS as the dependent variable.
𝐿𝑁_𝐹𝐿𝑆 (1, 2, 3)
= 𝛼 + 𝛽1FOR_PCT + 𝛽2𝐿𝑂𝑆𝑆 + 𝛽3𝐸𝐴𝑅𝑁_𝑉𝑂𝐿 + 𝛽4𝐵𝑈𝑆_𝑆𝐸𝐺 + 𝛽5𝐺𝐸𝑂_𝑆𝐸𝐺 + 𝛽6𝐹𝐼𝑁_𝐹𝑂𝐿 + 𝛽7𝐿𝑁_𝐴𝑆𝑆𝐸𝑇𝑆 + 𝛽8𝐿𝐸𝑉𝐸𝑅𝐴𝐺𝐸 + 𝛽9𝐵𝐼𝐺𝑁 + 𝛽10𝑅𝑂𝐴 + 𝜀
Additionally all regressions contain year and industry dummies. Standard errors are robust to correct for heteroskedasticity and clustered by firmid. Following Alfonso et al. (2012) I do not correct any outliers.
Following the hypotheses I run three regressions with different types of FLS. First, using the number of positive FLS as the dependent variable to test the proprietary costs hypothesis. Second, the number of negative FLS is used as the dependent variable to test the agency costs hypothesis. Lastly the number of FLS on investments will be used as the dependent variable to test the litigation risk hypothesis. As the distribution of the different categories of FLS is severely skewed the distribution of the dependent variable is transformed in the natural logarithm of FLS. If 𝛽1 has a negative coefficient in research model 1 than it is consistent with my hypothesis that proprietary costs have a negative effect on the amount of forward-looking segment disclosures containing good news. If 𝛽2 has a negative coefficient in research model 2 than it is consistent with my hypothesis that agency costs have a negative effect on the amount of forward-looking segment disclosures containing bad news. If 𝛽3 has a positive coefficient in research model 3 than it is consistent with my hypothesis that litigation risk has a positive effect on the amount of non-earnings related forward-looking segment disclosures.
4 Results
This section presents the results from the three research models used to test my hypotheses. This chapter starts with the descriptive statistics followed by an analysis of the correlations between the variables. Lastly the results of the regression models are discussed.
4.1 Descriptive statistics
Table 1 provides the distribution of observations between the different industries. The manufacturing firms represent the biggest part of the observations with 71% . This is followed by the services industry which represent 18% of the observations. The other industries account for 0.2% to 3.6% of the observations.
Table 1: Number of observations per industry
Industry SIC # observations
Agriculture, Forestry & Fishing 0100 – 0999 10
Mining 1000 – 1499 72
Construction 1500 – 1799 6
Manufacturing 2000 – 3999 1473
Transportation & Public Utilities 4000 – 4999 42
Wholesale Trade 5000 – 5199 75
Retail Trade 5200 – 5999 19
Services 7000 – 8999 363
Public Administration 9100 – 9999 4
Table 2 outlines the distribution of the different categories of FLS. Because of the severe skewness of the distribution of FLS as indicated in table 2, the dependent variable is transformed in the natural logarithm of FLS.
Table 2: Frequency table per category FLS # of FLS 0 1 2 3 4 5 Dependent variable H1: good news FLS 1,896 (91.82%) 137 (6.63%) 25 (1.21%) 5 (0.24%) 1 (0.05%) 1 (0.05%) Dependent variable H2: bad news FLS 1,865 (90.31%) 154 (7.46%) 36 (1.74%) 8 (0.39%) 1 (0.05%) 1 (0.05%) Dependent variable H3: non-earnings FLS (91.04%) 1,880 (7.75%) 160 (0.92%) 19 (0.29%) 6 (0%) - (0%) -
Notes: Bold numbers represent the number of observations reporting a certain amount, indicated in the first row, of the category of FLS. Between brackets is the percentage of the total observations that report a certain amount, indicated in the first row, of the category of FLS.
Table 3 presents the descriptive statistics of all the variables included in the analysis. In comparison with the full sample of Alfonso et al. (2012) my sample is a bit smaller because it covers 5 years instead of 7. The number of firms is also larger in the sample of Alfonso et al. (2012). Their sample covers 1,202 firms instead of the 418 firms in my sample. The size of the firms and the number of segments within the firms are comparable in both samples. The same holds for the average of firms audited by a big N audit firm and the average number of analysts following the firms. Lastly the earnings volatility and the return on assets of the firms in both samples is very similar. However the differences in sample size also lead to a number of
differences in the characteristics of the samples. Compared to my sample the sample of Alfonso et al. (2012) has slightly less loss firms, 11% in their sample and 19% in my sample, and on average less leveraged firms, 22% compared to 56%. Overall the samples look to a great extent similar.
Table 3: Descriptive statistics
Variable Observations Mean Median Std. Dev. min max
Dependent variables # good news FLS 2065 0.10 0 0.38 0 5 # bad news FLS 2065 0.13 0 0.43 0 5 # non-earnings FLS 2065 0.10 0 0.36 0 3 LN good news FLS 2065 0.07 0 0.23 0 1.79 LN bad news FLS 2065 0.09 0 0.27 0 1.79 LN non-earnings FLS 2065 0.07 0 0.22 0 1.39 Independent variables FOR_PCT 2065 40% 39% 18% 0% 100% LOSS 2065 0.19 0 0.40 0 1 EARN_VOL 2065 0.13 0.05 0.42 (0.82) 7.11 Control variables BUS_SEG 2065 2.55 1 2.11 1 17 GEO_SEG 2065 3.04 3 1.56 1 21 FIN_FOL 2065 6.94 3 9.16 0 51 LN_ASSETS 2065 6.59 6.57 1.95 1.32 12.99 LEVERAGE 2065 56% 53% 26% 0% 285% BIGN 2065 0.98 1 0.13 0 1 ROA 2065 6% 8% 18% (202%) 387%
4.2 Correlations and univariate analysis
Table 4 shows the Pearson correlation between all variables included in the analysis. In a univariate analysis only the proxy for proprietary costs seems to have a significant positive effect in research model 2 and 3 compared to the other main variables of interest agency costs and litigation risk. This is further examined in a multivariate regression described in section 4.3. I find one correlation coefficient that is above -0.6 and that is the correlation between ROA and LOSS which is logical because loss firms have low profitability reflected in a low return on assets. The correlation between LN_ASSETS and FIN_FOL is also relatively high as expected but is not expected to be a problem. Several other correlation coefficients are significant. However since these values are low no multicollinearity occurs in the database. The variance inflation factor (VIF) value is 1.45 which is below 4 and all tolerance levels of the variables are above 0.1 so this also indicates that there is no multicollinearity.
Table 4: Correlation matrix LN good news FLS LN bad news FLS LN non-earnings FLS FOR_ PCT LOSS EARN_ VOL BUS_ SEG GEO_ SEG FIN_ FOL LN_ ASSETS
LEVERAGE BIGN ROA
LN good news FLS 1 LN bad news FLS (0.011) 1 LN non-earnings FLS (0.017) 0.054** 1 FOR_PCT 0.034 0.044** 0.055** 1 LOSS (0.004) 0.061 0.010 (0.015) 1 EARN_VOL (0.011) 0.001 0.012 0.022 (0.106)*** 1 BUS_SEG (0.040)* 0.011 0.010 0.001 (0.082)*** (0.083)*** 1 GEO_SEG 0.027 0.011 0.002 0.269*** (0.032) (0.027) 0.270*** 1 FIN_FOL (0.036) (0.004) 0.060*** 0.074*** (0.188)*** 0.028 0.146*** 0.095*** 1 LN_ASSETS (0.015) 0.007 0.069*** 0.084*** (0.296)*** (0.056)** 0.421*** 0.148*** 0.514*** 1 LEVERAGE 0.034 (0.079)*** 0.026 0.053** 0.177*** (0.129)*** 0.157*** 0.031 (0.117)*** 0.210*** 1 BIGN (0.006) 0.035 0.018 (0.014) (0.084)*** (0.037)* 0.045** (0.002) 0.089*** 0.180*** 0.039* 1 ROA (0.006) (0.017) 0.005 0.012 (0.607)*** 0.090*** 0.032 0.005 0.178*** 0.209*** (0.291)*** 0.062*** 1
4.3 Regression analyses
4.3.1 Hypothesis 1
Table 5 reports the results of the regression model of hypothesis 1. The coefficient on FOR_PCT is not significant therefore hypothesis 1 is not confirmed. As earlier described normally positive news is less credible as viewed by the impression management theory. However since forward-looking disclosures in the MD&A in general are more credible the incentive to withhold proprietary disclosures might be driven out by firm’s incentive to reduce information asymmetry. Furthermore it is not uncommon that the proprietary costs hypothesis is not significant in both management forecast settings and segment reporting settings (Berger & Hann, 2007; Kent & Ung, 2003). The coefficients on BUS_SEG and GEO_SEG are significant when considering a one-tailed test. This means that an increase of one business segment results in reporting on average 0.004% less FLS on foreign operations, ceteris paribus. It is reasonable that firms that are more diversified in terms of business segments would put more emphasis on forward-looking disclosures on business segments instead of foreign operations. With regard to geographical segments, an increase of 1 segment results in reporting on average 0.005%. The economic significance of these variables is limited because the difference in the number of segments should be reasonable high to get an significant difference in the amount of FLS on foreign operations. All other coefficients are insignificant so they do not appear to have an effect on the amount of positive forward-looking sentences on foreign operations.
Table 5: Regression model hypothesis 1 Dependent variable: LN good
news FLS
Prediction Coefficient P-value
Main independent variables
FOR_PCT - 0.038 0.413 LOSS + (0.012) 0.414 EARN_VOL - (0.010) 0.540 Control variables BUS_SEG + (0.004) 0.182 GEO_SEG + 0.005 0.174 FIN_FOL + (0.000) 0.977 LN_ASSETS + (0.005 0.418 LEVERAGE + 0.288 0.444 BIGN + 0.011 0.786 ROA - (0.011) 0.732
Characteristics
Observations 2065
R2 /adjusted R2 0.028/0.017
F-value/sig 3.26/0.000***
Industry and year dummies Yes
Notes: * significant at 0.10 level; ** significant at 0.05 level; *** significant at 0.01 level, based on two-tailed tests. P-values are based on robust standard errors. Observations are clustered by firmid.
4.3.2 Hypothesis 2
Table 6 reports the results of the regression model of hypothesis 2. The prediction of the coefficient on agency costs is a negative sign. The outcome of the regression however shows a significant positive coefficient which means that firms reporting a loss in a given year disclose on average 6.5% more FLS containing bad news on foreign operations, ceteris paribus. This means that hypothesis 2 is not supported. This result could be explained by the big bath theory. This theory predicts that managers might take extra losses to make outperforming in the next year easier (Kirschenheiter & Melumad, 2002). So if the manager biases the tone of the forward-looking disclosures downwards outsiders will lower their expectation regarding the performance of the manager in the next year. Furthermore I find a significant coefficient on leverage and on Big N audit firm. This means that if leverage grows with 1% firms on average will report 0.118% less FLS containing bad news on foreign operations, ceteris paribus. If the firm is audited by a Big N audit firm in a given year, firms on average report 7.8% more FLS containing bad news on foreign operations, ceteris paribus. The results in this research model are
economically significant because relative small changes in profit, the report of a profit instead of a loss, and in choice of audit firm, result in changes approaching 10% of the amount of FLS on foreign operations. The coefficient on leverage might seem small but debt levels might vary significantly between years and firms so this percentage change of the number of FLS on foreign operations will often be multiplied resulting in a bigger magnitude of the effect of leverage.
Table 6: Regression model hypothesis 2 Dependent variable: LN bad
news FLS Prediction Coefficient P-value
Main independent variables
FOR_PCT - 0.086 0.108
LOSS - 0.065 0.015**
EARN_VOL - 0.003 0.835
Control variables
GEO_SEG + 0.001 0.783 FIN_FOL + (0.001) 0.438 LN_ASSETS + 0.010 0.110 LEVERAGE + (0.118) 0.000*** BIGN + 0.078 0.002*** ROA + (0.002) 0.978 Characteristics Observations 2065 R2 /adjusted R2 0.042/0.031 F-value/sig 5.65/0.000***
Industry and year dummies Yes
Notes: * significant at 0.10 level; ** significant at 0.05 level; *** significant at 0.01 level, based on two-tailed tests. P-values are based on robust standard errors. Observations are clustered by firmid.
4.3.3 Hypothesis 3
Table 7 reports the results of the regression model of hypothesis 3. Based on the insignificant coefficient on investor uncertainty it is not possible to reject the null hypothesis. This means I do not find evidence that litigation risk is associated with the amount of non-earnings FLS on foreign operations. This might be due to the lack of variation in the measurement of non-earnings related forward-looking segment disclosures. Very few observations report 2 or more FLS on investments and the majority of observations do not report a FLS on investments at all. I do however find that proprietary costs and agency costs have an effect on the amount of non-earnings FLS on foreign operations. First, the coefficient on proprietary costs is positive and significant. This indicates that firms that are subject to proprietary costs report more non-earnings FLS on foreign operations, ceteris paribus. An explanation for this result is that firms might provide non-earnings related because earnings related disclosures because it is easier to enter into competition if the expected result is quantified. An example is that firms announcing higher expected sales in a certain region will reveal more information about the profitability in a region than firms that announce a factory expansion in a certain region. Because firms have less ability to communicate the future prospects of the firm due to high proprietary value of
earnings-related forward-looking disclosures the manager might want to compensate for this loss of information firms by reporting more non-earnings related forward-looking disclosures. In particular this means that as the part of foreign sales of the firm increases with a portion of 1% the firm on average will disclose 0.676% more non-earnings FLS on foreign operations, ceteris paribus. The coefficient on agency costs is also positive and significant. This means that firms reporting a loss in a given year disclose on average 2.5% more non-earnings FLS on foreign operations, ceteris paribus which is consistent with the results of Alfonso et al. (2012). Loss
