Amsterdam Business School
The association between
forward looking segment information
and diversification
Name: Thomas Roos Student number: 10072004 Thesis supervisor: Réka Felleg Date: 20 June 2016
Word count: 12383
MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam
Statement of Originality
This document is written by student Thomas Roos 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
Using agency theory, I investigate the relationship between the degree of geographic diversification of a company and the amount of forward looking information on segments disclosed. I also examine if the amount of information disclosed is related to firm value, or value destruction caused by diversification. I conduct this research using samples of almost 400 U.S. firms reporting at least two geographical segments for the years 1996 to 2000. I find evidence supporting the theory that diversification is positively related to the amount of forward looking information disclosed. Together with the evidence that the amount of forward looking information is positively related to firm value, this has some implications. For example, since managers can benefit from increased firm value, they can use forward looking disclosures to reach this. The results are interesting because research on other types of voluntary disclosures find that managers have more incentives to withhold information, whereas I find that managers have incentives to disclose this information.
Contents
1 Introduction ... 5
2 Theoretical framework and hypothesis development ... 6
2.1 Agency theory, firm value, and legislation ... 6
2.2 Diversification... 7
2.3 Voluntary disclosure ... 9
2.4 Forward Looking Disclosure ... 11
2.5 Segment Disclosure... 12
2.6 Diversification and value destruction ... 13
2.7 Statement of Financial Accounting Standard (SFAS) No. 131 ... 14
2.8 Hypothesis development ... 14 3 Methodology ... 16 3.1 Sample description ... 16 3.2 Empirical design ... 18 4 Results... 22 4.1 Descriptive statistics ... 22 4.2 Hypothesis Tests ... 27 4.2.1 Hypothesis 1 ... 27 4.2.2 Hypothesis 2 ... 27 5 Robustness Tests ... 30 5.1 Hypothesis 1 ... 30 5.2 Hypothesis 2 ... 31 6 Conclusion ... 33 7 References ... 35 8 Appendix ... 42
1 Introduction
This study investigates the relationship between the degree of geographic diversification of a company and the amount of forward looking information on segments disclosed. Diversification makes a firm more complex and increases information asymmetry between managers and shareholders. Managers have incentives to withhold voluntary information, for example by aggregating segment information, despite shareholders valuing additional information. Due to diversification however, managers might have an incentive to decrease information asymmetry and agency costs by increasing disclosure of forward looking information. This is interesting because forward looking disclosures are a special type of voluntary disclosure and therefore the results could be different compared to other types (Healy and Palepu, 2001).
Using samples of almost 400 U.S. firms reporting at least two geographical segments for the years 1996 to 2000, I estimate a model used for testing if the percentage of foreign sales is related to the number of forward looking sentences. I also estimate a model to examine if the number of forward looking sentences influences the value of a firm. The models control for determinants of voluntary disclosure and firm value respectively. I find that, while the degree of diversification is positively associated with the number of forward looking sentences, the relationship with firm value is less certain. This study is different from Bens and Monahan (2004), who also investigate the effect of voluntary disclosure, by looking at forward looking information specifically and not in general. While also similar to Berger and Hann (2003), my research focuses on voluntary instead of mandatory disclosures. Any research on voluntary disclosure is interesting, given that the disclosure has value for investors, because managers have considerable discretion on whether or not to disclose. In this setting, where the disclosures contain information about foreign segments, diversification is an important aspect to consider. My results contribute to existing literature by confirming that results for forward looking information might not be the same as for other types of voluntary disclosures. The results might also be of interest to managers since they can benefit from increased firm value and I find that increased forward looking disclosures are related to higher firm value.
The next section outlines the theoretical framework and introduces the hypotheses, followed by the sample construction and description and the empirical models I estimate for testing the hypotheses. Section 4 contains the results of the regressions and the section following it contains the tests conducted to determine the robustness of these results. The sixth and final section summarizes and concludes.
2 Theoretical framework and hypothesis development
2.1 Agency theory, firm value, and legislationTraditional agency theory uses two actors to explain why the total value of a firm is less when managed by a manager as compared to one managed by the sole owner (Jensen and Meckling, 1976). The agent (manager) acts on behalf of the principal(s) (shareholders) in managing their firm, where both actors are expected to be utility maximizers. Because of the latter, it is likely that the agent might diverge in its actions from what is in the best interest of the principal, resulting in costs. Bonding and monitoring costs are incurred by the agent and principal in order to align the agent’s actions with the best interest of the principal. Because it is impossible to create a contract where the manager always acts according to the principal’s preferences, regardless of bonding and monitoring costs, a residual loss for the principal exists.
In a well-known study, Healy and Palepu (2001, p. 2) argue that “demand for financial reporting and disclosure arises from information asymmetry and agency conflicts between managers and outside investors.” Aljifri and Hussainey (2007) have shown that forward looking disclosures can reduce information asymmetry and agency conflicts. Botosan and Harris (2000) find that the decision of firms to no longer disclose segment information increases information asymmetry. This is supported by Bravo (2015), who argues that disclosure reduces uncertainty and thus reduces information asymmetry. Jensen and Meckling (1976) have also argued that voluntary disclosure can reduce the conflicts of interest between the principal and agent. Hope and Thomas (2008) found that reduced monitoring increased managerial willingness to diversify, indicating a greater divergence between the principal’s preferences and the agent’s actions as a result. They argue that the suboptimal decisions made by management are possible because reduced disclosure quality makes it harder for investors to link managerial actions to firm performance because of reduced monitoring capabilities. Berger and Hann (2007) also argue that nondisclosure of segment information can be explained by agency motives.
So, according to agency theory, the value of a firm is reduced because managers do not act in the best interest of the shareholders. Information asymmetry and agency conflicts lie at the heart of this problem and research has shown that disclosures help mitigate the resulting agency costs. However, managers apparently have motives not to disclose when given discretion.
2.2 Diversification
Diversification can be divided into two categories; international (or geographic) diversification and industrial (or product) diversification. An internationally diversified firm is one that has expanded its operations across the borders of its original country into others, this is often measured by the number of different geographic locations (or markets) in which the firm operates. An industrially diversified firm is one that has expanded its operations into new product markets (Hitt et al., 1997).
This distinction is important because research has shown that both influence a company in different ways (Kim and Mathur, 2008). For example, Denis et al. (2002) show that geographic diversification decreases firm value. Hitt et al. (1997) conclude that geographic diversification does not have a linear relationship with firm performance, at first it increases firm performance but at some point this becomes neutral and eventually negative. Furthermore, Hitt et al. (1997) also note that industrial diversification reduces the effects of geographic diversification. This is interesting since it could mean that industrial diversification can create value for a geographically diversified firm. Related to this, Berger and Ofek (1995) find that firm value is reduced for diversified firms of all sizes but this effect is mitigated for those with more related diversification. Nayyar (1993) argues that this might be because a firm can exploit economies of scope for related diversification but not for unrelated diversification.
Given the evidence from these and other studies that diversification negatively affects a firm, why would managers still choose to diversify? The existing literature offers several benefits of diversification. For example, Amihud and Lev (1981) state that diversification leads to reduced risk for the firm. However, Amihud and Lev (1981) also note that this risk reduction has no added value for investors in perfect capital markets because they can reduce their own risk by constructing diversified portfolios. Since Denis et al. (2002) argue that investors’ desire for internationally diversified portfolios can be satisfied through international diversification, presence of perfect capital markets is not assumed. As Bens and Monahan (2004) mention, if capital markets are not assumed to be perfect, due to information asymmetry or taxes, the possibility exists for diversification to either create or destroy value. Possibilities for creating value are due to a greater debt capacity, implying a larger tax debt shield (Lewellen, 1971), the ability to offset losses from one segment with the profits from another which further reduces taxes (Majd and Myers, 1986; Berger and Ofek, 1995), enhanced economies of scope and increased managerial coordination (Chandler, 1977), and creation of efficient internal capital markets, reducing underinvestment (Weston, 1970; Stein, 1997; Berger and Ofek, 1995).
According to Rumelt (1974), diversification affects firm value positively, especially related diversification because these firms can better use their skills and resources in the added markets. Nayyar (1993) mentions that value created by related diversification, unlike unrelated diversification, is positively associated with the firms’ reputation in an existing market. Buhner (1987) argues that international diversification potentially offers greater firm growth through prospective market opportunities. Furthermore, Bodnar et al. (1999) find that, controlling for both types of diversification, international diversification significantly increases firm value. Finally, it has been argued that international diversification can lead to increases in value for shareholders through increased flexibility of operations and exploitation of firm specific assets (Kim and Mathur, 2008; Denis et al., 2002).
Diversification also has downsides as is shown in several studies (Comment and Jarrell, 1994; Denis et al., 1997; Lang and Stulz, 1994; Kim and Mathur, 2008; Denis et al., 2002). For example, Denis et al. (1997) find that diversification is negatively related to outside block holder and managerial equity ownership. Consistent with Kim and Mathur (2008), Denis et al. (2002) conclude that the value loss of diversification is nearly the same for both types of diversification and that the costs outweigh the benefits. These results can be explained by various reasons present in the literature. In a study where diversified firms are compared to benchmark single-segment firms, Lang and Stulz (1994) and Berger and Ofek (1995) show that the diversified firm trades at a discount and argue that the costs of information asymmetry exceeds the benefits of internal capital markets. Similarly, the benefits of increased economies of scope and managerial coordination as presented by Chandler (1977) may be less than the information asymmetry between divisions created by diversification (Harris et al., 1982). This was also suggested by Denis et al. (2002) who argue that diversification increases difficulties in monitoring managerial decision making and coordinating corporate policies. Bodnar et al. (1997) suggest that some problems, such as increased difficulty in monitoring managers, are inherent to diversification because diversified firms are more complex due to their operations in different locations. Berger and Ofek (1995) offer misalignment between central and divisional management, suboptimal cross-subsidies, and investing in value-decreasing projects as explanations for the value loss. In another study, Kim and Mathur (2008) argue that internationally diversified firms suffer from local political and local economic regulation and developments and exchange controls.
Many of the above authors argue that their findings indicate that diversification is a result of information asymmetry between shareholders and managers and use agency theory as an explanation. This is strengthened by several studies finding that refocusing activities lead to increased shareholder value (Berger and Ofek, 1995; Comment and Jarrell, 1995; John and Ofek,
1995; Liebeskind and Opler, 1994). Berger and Ofek (1999), for example, study the causes and effects of refocusing activities of firms between 1984 and 1993. They find that a decrease in diversification often happens after corporate control events such as a change in CEO or financial distress. They also see that refocusing firms, when compared to those that did not refocus, had more value-reducing diversification. These results can be interpreted as evidence that diversification exists even when it is not in the best interest of the shareholders and that managers had to be pressured to act more in line with what shareholders prefer. This is similar to the “managerial” explanation for why firms diversify. This theory states that managers diversify in order to decrease their risk of job loss and to protect their professional reputation, which can also be seen as part of agency theory. Morck et al. (1990) also state that diversification makes a firm more unique which in turn makes managers more valuable to the firm and able to demand larger compensation. This then increases agency costs for shareholders since it reduces the threat of dismissal for the manager and allows him to deviate from shareholders’ best interests even more.
Summarizing the theories, diversification can be seen as managers not acting in line with shareholder preferences for multiple possible reasons, and they are able to do this because of information asymmetries between the them and the shareholders. Finally, research has also shown that diversification itself increases agency costs because diversification makes firms less transparent and harder to actively monitor management decisions (Doukas and Pantzalis, 2003; Kim and Mathur, 2008).
2.3 Voluntary disclosure
Voluntary disclosure is potentially important because managers have significant discretion and it contains valuable information to investors (Healy and Palepu, 2001), because of this it has been extensively researched in the past. Examples include requests to the SEC to withhold certain information in the filings (Verrecchia and Weber, 2006), information about customers (Ellis et al., 2012), and forecasts of earnings (Bamber and Cheon, 1998; Li, 2010). These studies show that firms, or managers, apparently have reasons to withhold certain information when given the opportunity. Arya et al. (2010), however, show that firms with multiple geographic segments provide full disclosure on the firm wide level but not on the segment level despite possible negative consequences of full disclosure. They argue that this is because when the firm has a low profit, it wants to disclose this in order to reduce competition and when it has a high profit it does not want to disclose this. No disclosure therefore signals to the competitor that the firm
has a high profit leading to full disclosure as the equilibrium. They find that firms are more likely to withhold information for segments if their characteristics are similar.
It has been proven that managers do use their discretion to withhold information from shareholders and the existing literature has offered several explanations for this. The most common explanation uses the proprietary cost theory, which posits that proprietary information provides information to competitors and therefore could harm the firm’s performance (Wagenhofer, 1990; Hayes and Lundholm, 1996; Ellis et al., 2012). Thus, the higher the proprietary costs of information, the less likely it is that this information will be disclosed. Proprietary costs not only affect the level of voluntary disclosure, but also its nature. For example, Verrecchia (1983) and Nichols (2009) have shown that more good, or less bad, news is associated with increased proprietary costs. Dedman and Lennox (2009) find that, according to managers, firms are less likely to reveal information when a company has many competitors, a high threat of entry or when competition is intense. However, Dedman and Lennox (2009) also note that managers have to make a trade-off about disclosure since more disclosure also provides benefits. For example, profitable firms can increase their market value by elaborating on their results in increased disclosures (Dedman and Lennox, 2009). Litigation costs also provide mixed incentives; threat of legal action because of inadequate disclosures can reduce incentives to provide voluntary disclosures. On the other hand it can increase voluntary disclosure when legal actions are taken because of incomplete disclosures (Healy and Palepu, 2001).
While the previous theories offer explanations for why managers might withhold information, agency theory can be used to explain why managers would voluntarily supply information. Ellis et al. (2012) and Bravo (2015) argue that voluntary disclosures can be used to reduce information asymmetries between management and shareholder, which in turn can reduce the cost of capital and risk, or improve liquidity (Ellis et al., 2012; Kothari et al., 2009; Clarke, 1983; Diamond, 1985; Hviid, 1989; Diamond and Verrecchia, 1991; Hayes and Lundholm, 1996; Francis et al., 2005; Verrecchia and Weber, 2006). Healy and Palepu (2001) find that managers are held responsible for firm performance and that poor performance is associated with higher CEO turnover. They argue that managers can withhold information to reduce the monitoring capabilities of shareholders to reduce threat of dismissal. This means that managers might increase agency costs by withholding information for personal motives. It has been argued that managers will disclose private information if the benefits (meaning a reduction in agency costs) exceeds the costs of not disclosing (Prencipe, 2004). Chan and Watson (2011) argue that managers of diversified companies have more private information that can be useful to investors and therefore a greater demand is placed on them to disclose this information.
2.4 Forward Looking Disclosure
Forward looking information is based on expectations about the future state of the firm and its performance (Alkhatib, 2014). Forward looking disclosures can contain both financial and non-financial information, such as forecasts of revenues, sales volume, and cash flows or risk respectively (Alkhatib, 2014). In the United States specifically, disclosure of forward looking information is encouraged through regulation. The safe harbor regulation protects provision of forward looking information by considering information not to be fraudulent unless the statement was not disclosed in good faith or based on unreasonable expectations (Bravo, 2015). The idea behind this is that this should lower litigation costs and thus result in increased disclosures since managers are expected to supply information if the benefits of this exceed the costs (Baginski et al., 2004).
According to Healy and Palepu (2001), forward looking information is a specific type of voluntary disclosure and might not generalize to other types of voluntary disclosure. Perhaps because of this it has extensively been researched in the past. Aljifri and Hussainey (2007) find that debt ratio and profitability affect the level of disclosure, which is confirmed by Alkhatib (2014). Baginski et al. (2014) investigate the frequency and nature of voluntary disclosure in the context of proxy contests and find that proxy contests increase disclosure levels and that news during these periods is less negative. Bamber and Cheon (1998) find that firms with higher proprietary costs and greater legal liability exposure issue less specific forecasts.
Forward looking disclosures are similar to other types of voluntary disclosure in that they also provide benefits to a company. For example, according to Aljifri and Hussainey (2007), absence of forward looking information may result in investors basing their decisions on wrong information from other sources and thus increased disclosure helps investors in making better investment decisions (Bujaki et al., 1999). Furthermore, it has been argued that managers can use forward looking information to convey their abilities to investors (Trueman, 1986). These are all reasons for companies (or managers) to increase forward looking disclosures, there are however also authors that provide evidence for decreased disclosures. This is shown by Kothari et al. (2009), who find that managers strategically delay bad news, and Bujaki et al. (1999), who show that bad news is dominated by good news. Kasznik (1999) argues that firms might influence performance to match forecasts, however these are difficult to predict accurately because the future is uncertain. Inaccurate information can potentially harm a managers’ reputation, which could reduce pay and the ability to influence future market expectations, or even complicate
future job search (Williams, 1996; Hutton and Stocken, 2009; Zamora, 2009; Baginski et al., 2014).
2.5 Segment Disclosure
Despite investors valuing general voluntary disclosure, full disclosure of available information is not an equilibrium. Due to its nature, voluntary disclosure of segment information can be viewed as supporting other disclosures (Aitken et al., 1997). Withholding segment information is often done by aggregating information. For this reason, prior literature has mainly focused on aggregation of segment information with the reasoning that disaggregated information has more value to investors because it improves earnings predictions (Aitken et al., 1997). Berger and Hann (2003) show that with the introduction of SFAS No. 131, concerning disclosure of segment information, more information was revealed which affected firm valuation; they conclude that this is due to improved monitoring caused by the new standard. This is supported by the findings from Herrmann and Thomas (2000). Harris (1998) investigates if industry competition affects business segment reporting and shows that segments operating in less competitive industries are less likely to be reported separately. They argue that profits are more likely to be higher in less competitive industries and managers want to withhold this information because they view this as proprietary information. Additionally, Hayes and Lundholm (1996) show that disclosure of segments depends on the earnings persistence of the segment and the correlation of profitability between the segments. Arya et al. (2010) state that full disclosure is likely at the firm level, but not for separate segments and that if segment characteristics are similar the information on segments is more likely to be withheld. The latter occurs, according to Aitken et al. (1997), because firm wide disclosures are less informative. Arya et al. (2010) rightly note that withholding segment information not only depends on managers’ desire to do this but also on the viability of this choice. Bens and Monahan (2004) show that increased segment disclosure is positively related to firm value, indicating that investors prefer more information and thus withholding it could be seen as agency costs. Consistent with this, Berger and Hann (2007) find that agency problems partially explain a desire to withhold segment data.
Summarizing, shareholders value disaggregated segment information more than aggregated information and thus value firms that provide disaggregated information higher. However, managers have incentives to withhold information through aggregation which increases agency costs.
2.6 Diversification and value destruction
Disclosures have been proven to be an effective tool for investors to monitor managerial actions and thus reduce agency costs (Bens and Monahan, 2004; Healy and Palepu, 2001). However, as Jensen and Meckling (1976) have found, monitoring management decisions becomes harder as the firm becomes more complex and previous research has shown that international diversification increases complexity (Callen et al., 2005; Thomas, 1999). Furthermore, international diversification has also been related to increased information processing problems for managers (Birkinshaw et al., 2001; Egelhoff, 1991; Kim and Mauborgne, 1995; Schulz, 2001). Berger and Ofek (1999) have shown that management of refocusing firms were forced to refocus by investors, supporting the theory that diversification is a result of agency costs. Other studies have found that diversified firms are valued lower than benchmark single-segment firms, implying that the costs of diversification outweigh the benefits for investors, also known as the diversification discount (Lang and Stulz, 1994; Berger and Ofek, 1995).
The diversification discount has been linked to information asymmetry between managers and shareholders (Lamont and Polk, 2001b; Meyer et al., 1992), or agency problems (Denis et al., 1997; Berger and Ofek, 1999). Berger and Ofek (1999) also document that firms suffering from the greatest diversification discount are more likely to attract external pressure and divest as a result. This indicates that the value loss caused shareholders to take action and force the firm to refocus (Denis et al., 1997). The diversification discount has also been researched in relation to voluntary (and segment) disclosure. For example, higher disclosure quality has been found to reduce managerial actions that destroy shareholder value (Kanodia and Lee, 1998; Healy and Palepu, 2001). Bens and Monahan (2004) also find that excess value is positively related to ratings of voluntary disclosure. This is further supported by Berger and Hann (2003) who show that an increase in the diversification discount is a result of less segment disclosures. According to Berger and Hann (2007), the above suggests that managers face costs from segment disclosures that show underperformance. This is consistent with Hope and Thomas (2008), who find that excess value is significantly lower for firms deciding not to disclose. According to Leuz and Verrecchia (2000), higher quality disclosure also reduces the information asymmetry between managers and shareholders and thus reduces the discount.
So, diversification is an agency problem and because of that diversification reduces firm value for shareholders. While diversification makes a firm more complex and increases information asymmetry, disclosures can reduce the value destruction of diversification by reducing information asymmetry.
2.7 Statement of Financial Accounting Standard (SFAS) No. 131
An aspect relevant for my research is a change in legislation which affects the diversification discount. In 1997, SFAS No. 131 was issued and became effective for fiscal years starting from 16 December 1997. SFAS No. 131, like its predecessor SFAS No. 14, deals with segment reporting. SFAS No. 14 required segment information to be reported using line-of-business information. This way of reporting, the industry approach, was criticised because the general definition of “industry” allowed for considerable aggregation of segment information (Berger and Hann, 2003). The new standard introduced the management approach with the intention to solve this problem. This approach requires reporting of segments in the same way the firm is organised internally, allowing more information on the internal management strategy (Herrmann and Thomas, 2000; Berger and Hann, 2003). Herrmann and Thomas (2000) investigate the differences of segment disclosures between the two standards and find that the new standard has had a significant impact. Berger and Hann (2003) further find that SFAS No. 131 has resulted in an increase in reported segments and disaggregated information. They also find that the information disclosed because of SFAS No. 131 decreases firm value, especially for firms with previously hidden segments. This indicates that SFAS No. 131 is successful in revealing agency problems related to diversification (Berger and Hann, 2003). Related to this, Bens and Monahan (2004) find that increased voluntary disclosure quality is associated with a smaller discount. So the introduction of SFAS No. 131 has revealed and resolved agency problems, and reduced information asymmetries between management and shareholders, thereby affecting firm value.
2.8 Hypothesis development
Agency theory assumes that managers do not act in the best interest of shareholders because both actors are utility maximizing and it is impossible to write perfect contracts. Because of these incomplete contracts information asymmetry between investors and managers exists and agency conflicts arise. Diversification is an example of such a conflict because investors are better at diversifying their investments on their own yet managers diversify anyway. As research has shown, disclosure of voluntary information can be used to decrease information asymmetry by improving the monitoring capabilities of shareholders. However, research has also shown that managers have incentives to withhold voluntary information, for example to reduce the threat of dismissal or because the private information is considered to be proprietary. Forward looking information has been recognised as a special type of voluntary information which is why reasons for withholding voluntary information might not be valid for forward looking information. Also,
as Chan and Watson (2011) argue, shareholders of diversified companies might demand more disclosure because managers possess more value relevant private information. Since diversification increases the demand for disclosure, and managers are expected to disclose if the benefits of disclosing exceeds the costs, I propose the following hypothesis:
H1: Increased geographic diversification is associated with more forward looking information on segments.
Diversification, being an agency conflict, causes value destruction, leading to a lower firm value. This could be because investors are better able to diversify their portfolio on their own or because diversification makes a firm more complex and increases information asymmetry. Because investors value disclosures, mandatory and voluntary, they can reduce information asymmetry and the resulting value destruction. If forward looking information on segments is also valued by investors, its disclosure would mitigate the value destruction of diversification, therefore I propose the following hypothesis.
H2: More forward looking information on segments is associated with a smaller value loss due to geographic diversification.
3 Methodology
3.1 Sample descriptionIn order for me to test my hypotheses I need data available in the Compustat database and hand-collected data on the number of forward looking sentences in the MD&A section of 10-K filings of US multinationals that report at least two geographical non-US segments for the years 1996 to 2000. The selection, collection, and coding of these sentences was done as described by Janssen (2015). For the sentences a distinction was made between those about domestic and foreign operations and investments. A further classification was made for sales, costs, profits, and investments including or excluding currency related sentences. For each of these sentences it was also recorded if they contained good, bad, or no news. For my hypotheses I am only interested in the number of sentences on foreign operations and investments in general. As can be seen in Table 1, this results in 418 firms for each of the years from 1996 to 2000 with a total of 820 sentences being coded to be about foreign operations and investments and 2,090 firm-year observations.
Table 1 – Sample Construction
This table presents the characteristics of the different samples, showing the number of firms per year, the total number of forward looking sentences and the number of observations.
Sample size
Sample type Firms FLS Obs.
Original sample, used in combination with HITT 418 820 2,090 Sample used for testing Hypothesis 1, dropped observations
because of missing EPS data 390 760 1,950
Sample using the Herfindahl Index, dropped observations are
because of inaccurate Indices 388 756 1,940
Sample using the first value measure (V1), observations dropped to
calculate V1 367 740 1,835
Sample using the second value measure (V2), observations dropped
to calculate V2 367 740 1,835
Sample using the third value measure (V3), observations dropped
to calculate V3 362 749 1,810
The data gathered from Compustat originally contained 88,843 firm-years for all observations for the fiscal years 1995 to 2002 to include all necessary data to compute the variables. I merged this with the forward looking sentence data and dropped all observations not present in both datasets, resulting in 2,090 remaining firm-year observations. Observations were dropped if there was no Earnings Per Share available for the years 1995 to 2000, 1995 is included because the change in Earnings Per Share required for 1996 includes 1995 Earnings Per Share. This results in 140 observations being dropped and leads to the final sample for the first
hypothesis of 1,950 firm-year observations with 760 sentences being classified as about foreign operations and investments. Table 1 shows the sample construction and the number of forward looking sentences per year. The data has been Winsorized on the 1 and 99 percent level when deemed necessary to prevent distorting effects by outliers.
The sample for the second hypothesis deviates from the above sample in a number of ways. Table 1 shows the final firm-year observations per sample. Some observations were dropped because the data needed to calculate the variables was not present (INVEST and SURPRISE, see Table 2 for descriptions). The sample used as a basis for calculating firm value consists of 2,090 observations and 820 sentences. Observations were dropped when calculating the firm value (V1) because of missing data (255 dropped, 1,835 remaining).
3.2 Empirical design
The first hypothesis is aimed at trying to determine if the degree of diversification is positively related to the amount of forward looking disclosures on segments. Using Hope and Thomas (2008) as a basis, I modified their model according to Baginski et al. (2014) to include the relevant control variables and estimate the following empirical model for testing the first hypothesis:
𝐹𝐿𝑆𝑖,𝑡 = 𝛽0+ 𝛽1𝐷𝐼𝑉𝑖,𝑡+ 𝛽2𝑑𝐸𝑃𝑆𝑖,𝑡+ 𝛽3𝑃𝐸𝑅𝐹𝑖,𝑡+ 𝛽4𝐿𝐸𝑉𝑖,𝑡+ 𝛽5𝐴𝑁𝐴𝐿𝐹𝑂𝐿𝐿𝑖,𝑡 + 𝛽6𝑆𝐼𝑍𝐸𝑖,𝑡 + 𝜀𝑖,𝑡
( 1 )
Where:
Table 2 – Variable descriptions
This table contains descriptions and definitions of the variables used in testing the hypotheses with information on whether a variable has been Winsorized at a specific level. The first box mentions variables only used for the first hypothesis, the second box those used for both hypotheses and the third box those only used for the second hypothesis.
Variable Name Description
DIV = The degree of diversification of a firm
dEPS** = The change in annual earnings, defined as Earnings Per Share (EPS) in year t deducted by EPS in year t-1
PERF** = The performance of the firm, defined as Return on Equity (ROE) LEV* = The leverage of the firm, defined as total liabilities divided by total assets ANALFOLL = Defined as the number of analysts following the firm
SIZE = The size of the firm, defined as the log of total assets
FLS = The number of forward looking sentences on foreign operations and investments for firm i in year t
* Winsorized at 99%
** Winsorized at 1% and 99%
The operationalization of the theoretical constructs is as follows; FLS is the chosen proxy for the amount of forward looking information on foreign operations and investments and DIV is the degree of diversification of a company. Contrary to previous research on voluntary disclosures I use the number of forward looking sentences instead of AIMR rankings as a measure. This is a better proxy for the theoretical construct since AIMR rankings are a measure of general disclosure quality. AIMR rankings have also been criticized to be biased by opinions and personal relationships of analysts (Bamber and Cheon, 1998). The measure to proxy for the
degree of diversification (DIV) of a firm is FORSALE, which is foreign sales as a percentage of total sales and is commonly used in accounting literature (Hitt et al., 1997).
The control variables, described in Table 2, follow from previous research and have been proven to be associated with forward looking disclosures. For example, dEPS was used by Baginski et al. (2014) to adjust for the change in annual earnings since it was found to influence disclosures (Kasznik and Lev, 1995; Skinner, 1994). Performance (PERF), defined as return on equity, has also been associated with disclosure which is why I include it as a control variable (Miller, 2002). Following Hope and Thomas (2008), leverage (LEV) is included as a control variable. Additionally, Prencipe (2004) argues that, due to agency costs, firms have greater incentives to disclose segment information as leverage increases. It is argued that a greater number of analysts following a firm is positively associated with more detailed disclosures, thus, to control for the firm’s information environment ANALFOLL is included as the number of analysts following the firm (Schipper, 1991; Denis et al., 1997; Botosan, 1997; Baginski et al., 2014). Finally, firm size (SIZE) is included since it is a proxy for the overall disclosure level of a firm and to control for any firm size effects (Hope and Thomas, 2008; Denis et al., 1997; Baginski et al., 2014; Jones, 2007; Hitt et al., 1997). Not included in equation (1), but present in the analysis are year dummies for each of the years.
In order to test the first hypothesis I run an OLS regression on the clustered panel data. The first hypothesis examines if there is a positive relationship between geographic diversification and the amount of forward looking information on segments. This hypothesis is considered to be confirmed if 𝛽1 is positive, since this means that, as diversification increases, the amount of forward looking information increases as well. Because I argue that the degree of diversification is positively related to the amount of forward looking information disclosed, I expect the coefficient of FORSALE to be significantly positive. Because the positive relationship between leverage and the amount of disclosure has been confirmed in prior research, I expect 𝛽4 to be significantly positive (Salamon and Dhaliwal, 1980; Bradbury, 1992; Mitchell et al., 1995; Giner et al., 1997).
The second hypothesis more linearly follows a specific model, the model was introduced by Bens and Monahan (2004). In this model, firm (or excess) value is the dependent variable and the independent variable of interest is the amount of forward looking information. The model (2) attempts to test if there is a negative relationship between the amount of forward looking information and the value loss of diversification:
Where:
Table 2 - Continued
VALUE = The excess value of diversification of a firm
INVEST* = The investment opportunities of a firm, defined as the capital expenditures divided by sales NUMSEG = The number of foreign geographic segments reported by the firm
PROFIT** = Profitability of a firm, defined as net income as a percentage of total assets SALEGROW* = The sales growth, defined as sales for year t divided by sales for year t-1
SURPRISE* = The earnings surprise, defined as the absolute value of EPS difference between t and t-1 divided by the stock price at the beginning of the year
* Winsorized at 99% ** Winsorized at 1% and 99%
The variables are explained in Table 2. A number of the variables (FLS, LEV, ANALFOLL, and SIZE) are the same as those used in equation (1), and others (INVEST, NUMSEG, SALEGROW, and SURPRISE) follow directly from Bens and Monahan (2004). The notable differences between this model and the one used by Bens and Monahan (2004) are in firm value and profitability. Profitability, defined as net income as a percentage of sales, has an interesting relationship with voluntary segment disclosure. Prencipe (2004) argues that, since higher profitability indicated higher investment quality, a firm has greater incentives to disclose if it is highly profitable. However, this disclosure could increase competitive costs. Existing literature does not agree on the direction of the relationship between profitability and segment disclosure, some find it is positive (Giner et al., 1997; Saada, 1998), others find a negative relationship (Kelly, 1994; Leuz, 1999).
𝑉𝐴𝐿𝑈𝐸𝑖,𝑡 = 𝛽0+ 𝛽1𝐹𝐿𝑆𝑖,𝑡+ 𝛽2𝐿𝐸𝑉𝑖,𝑡+ 𝛽3𝐴𝑁𝐴𝐿𝐹𝑂𝐿𝐿𝑖,𝑡+ 𝛽4𝑆𝐼𝑍𝐸𝑖,𝑡
+ 𝛽5𝐼𝑁𝑉𝐸𝑆𝑇𝑖,𝑡+ 𝛽6𝑁𝑈𝑀𝑆𝐸𝐺𝑖,𝑡+ 𝛽7𝑃𝑅𝑂𝐹𝐼𝑇𝑖,𝑡+ 𝛽8𝑆𝐴𝐿𝐸𝐺𝑅𝑂𝑊𝑖,𝑡 + 𝛽9𝑆𝑈𝑅𝑃𝑅𝐼𝑆𝐸𝑖,𝑡+ 𝜀𝑖,𝑡
For measuring the value loss of diversification I employ a value measure (V1) used by Bodnar et al. (1997), being the market value of equity minus the book value of equity, divided by total sales. This value measure is similar to those used in Errunza and Senbet (1981) and Kim and Lyn (1986). V1 is a deviation from the original model used by Bens and Monahan (2004) because I was unable to calculate firm value using their measure. As with equation (1), year dummies are included and additionally, because the sample covers the transition period from SFAS No. 14 to SFAS No. 131 and this has proven to affect the value destruction of a firm, a dummy which equals 1 for the fiscal years subject to SFAS No. 131 to control for this has also been included.
As with the first hypothesis I run an OLS regression on the clustered panel data. The second hypothesis examines if there is a negative relationship between the value destruction due to geographic diversification and the amount of forward looking information on segments. This hypothesis is considered to be confirmed if 𝛽1 is positive, meaning the value of a firm is higher when more forward looking sentences are provided. Because of the mixed evidence described above concerning profitability, no prediction is made about the direction of 𝛽7.
4 Results
4.1 Descriptive statistics
The descriptive statistics for the variables of interest are presented in Table 3, Panel A contains statistics for the original sample. The other panels contain the statistics for the samples used for testing, Panel B pertains to the sample (H1) containing all variables for estimating equation (1) and Panel C contains the statistics of the samples used for estimating equation (2) using V1. The correlation of the variables for the various samples is presented in Table 4.
Table 3 –Summary Descriptive Statistics
Panel A pertains to the original sample and is included for comparison, Panel B represents the statistics for the sample used to test the first hypothesis. Panels C and D show the statistics for samples V1 and V3 respectively, the second value measure is also included in Panel C since the statistics for both samples are similar.
Standard
Variable Name Obs Mean Deviation Median Minimum Maximum
Panel A: Original sample
FLS 2,090 0.392 0.790 0 0 6 FORSALE 2,090 0.401 0.178 0.385 0.00384 1 PERF 2,090 0.0849 0.480 0.127 -2.294 2.407 LEV 2,090 0.554 0.244 0.536 0.0473 1.461 ANALFOLL 2,090 6.900 9.145 3 0 51 SIZE 2,090 6.571 1.950 6.562 1.318 12.99 HITT 2,090 8.2273 2.2350 8.2302 2.4166 14.462 Panel B: H1 sample FLS 1,950 0.390 0.789 0 0 6 FORSALE 1,950 0.398 0.173 0.384 0.00384 1 PERF 1,950 0.0903 0.473 0.127 -2.294 2.407 LEV 1,950 0.550 0.238 0.530 0.0473 1.461 ANALFOLL 1,950 7.157 9.268 3 0 51 SIZE 1,950 6.597 1.947 6.572 1.318 12.99 dEPS 1,950 -0.123 1.846 0.0300 -8.020 7.190 HERF 1,940 0.5168 0.1603 0.4904 0.1516 1
When comparing Panel A and B of Table 3 it shows that no outliers were present in the dropped observations since the minimum and maximum does not change for any of the variables. An interesting point worth mentioning is that while the maximum number of forward looking sentences is 6, the mean is only 0,4. This suggests that most companies do not provide any forward looking sentences (the median is 0, confirming this). The change in annual earnings (dEPS) is negative on average, but the profitability as measured by the return on equity is positive. Finally, the degree of diversification is on average 0,4, meaning forty percent of sales are foreign and the maximum is 1, which means the firm has no domestic (U.S.) sales at all. It is worth noting that for each of the regressions performed one year-dummy was excluded because
of multicollinearity. Panel C shows the statistics for the sample used to test the second hypothesis, a comparison with Panel B shows that there are only minor differences for the similar variables. What is interesting is that the mean (and median) of reported foreign geographic segments is 3 but the maximum is 21. The mean of the firm value measure used is 1.409 with a median of 0.65 which means the distribution is somewhat skewed to the right.
Table 3 - Continued
Standard
Variable Name Obs Mean Deviation Median Minimum Maximum
Panel C: V1&V2 sample
V1 1,835 1.409 4.525 0.650 -0.904 128.0 FLS 1,835 0.403 0.793 0 0 6 LEV 1,835 0.555 0.261 0.534 0.0473 2.851 ANALFOLL 1,835 7.324 9.434 4 0 51 SIZE 1,835 6.682 1.940 6.674 1.318 12.99 INVEST 1,835 0.0681 0.0876 0.0479 0.00122 1.464 NUMSEG 1,835 3.031 1.589 3 1 21 PROFIT 1,835 0.0365 0.229 0.0530 -2.026 7.550 SALEGROW 1,835 1.120 0.308 1.071 0.252 4.347 SURPRISE 1,835 0.102 0.613 0.0246 0.000 22.53 SFAS 1,835 0.545 0.498 1 0 1 V2 1,835 1.895 1.685 1.369 0.0709 18.56 Panel D: V3 sample V3 1,810 26.19 183.4 16.55 -1,356 6,056 FLS 1,810 0.414 0.813 0 0 6 LEV 1,810 0.551 0.259 0.530 0.0473 2.851 ANALFOLL 1,810 7.381 9.475 4 0 51 SIZE 1,810 6.688 1.943 6.673 1.318 12.99 INVEST 1,810 0.0688 0.0889 0.0480 0.00122 1.464 NUMSEG 1,810 3.035 1.594 3 1 21 PROFIT 1,810 0.0374 0.230 0.0538 -2.026 7.550 SALEGROW 1,810 1.127 0.348 1.073 0.252 6.337 SURPRISE 1,810 0.100 0.616 0.0246 0.000 22.53
FLS is the number of forward looking sentences on foreign operations and investments disclosed by the firm. FORSALE is the degree of diversification, measured as the foreign sales as a percentage of total sales. A higher FORSALE means the firm has relatively more foreign sales and is therefore more geographically diversified. PERF is the performance of the firm, measured as return on equity. LEV is the leverage of the firm, defined as total liabilities divided by total assets. The number of analysts following a firm is contained in ANALFOLL, and SIZE measures the size of the company as being the natural log of total assets. Y2000 is one of the included year-dummies and equals 1 for the year it relates to (2000 in this case). The change in annual earnings is captured by dEPS, which is the earnings per share of year t minus the earnings per share in year t-1. V1, V2, and V3 are different measures of capturing firm values, used to determine the value destruction of diversification. V1 is the market value of equity minus the book value of equity, divided by total sales. V2 is the market value of the firm as a percentage of its book value. V3 is the share price divided by the earnings per share. INVEST is capital expenditures divided by total sales. NUMSEG is the number of reported foreign segments by the firm. PROFIT is net income as a percentage of total assets. SALEGROW is sales in year t divided by sales in year t-1. SURPRISE is the absolute value of dEPS divided by the stock price at the beginning of the year. SFAS is a dummy indicating whether the data is from a year subject to SFAS No. 131 regulation (1) or not (0).
Table 4 shows the correlations for the samples for which summary statistics are also provided. Panel A shows the correlations for the original sample (above the diagonal) and the sample used to test hypothesis 1 (below the diagonal). As is visible in Table 4 Panel A, the correlation between the dependent and independent variable is significant, but low. The correlation of the diversification measure used (FORSALE) and firm performance is similar to that found by Hitt et al. (1997), this also applies to diversification and firm size. Other correlations also share similarities with Bens and Monahan (2004), such as the number of analysts following the firm and leverage. The only strong relationship found is between the number of analysts following a firm and firm size, which is consistent with prior literature. Panel B shows that the correlation between the independent and dependent variable is low but significant as well. Most of the correlations are in line with those reported by Bens and Monahan (2004). It is worth noting that only the number of analysts following a firm has a moderate positive correlation with firm value, which means that firms with a higher value also have more analysts following the firm.
Table 4 – Sample Correlations
Panel A shows the correlations for the sample used to test the first hypothesis below the diagonal and the original sample above the diagonal. Because observations were dropped to calculate dEPS the correlation with this variable for the original sample could not be presented. Panel B contains the correlations for the first and second value measure since these are the same, Panel C shows the correlations for the third value measure.
Panel A: Original and Hypothesis 1 sample
FLS FORSALE PERF LEV ANALFOLL SIZE
FLS 0.09*** -0.03 -0.04* 0.03 0.04* FORSALE 0.08*** 0.00 0.05** 0.08*** 0.09*** PERF -0.03 0.02 0.05** 0.11*** 0.18*** LEV -0.03 0.02 0.06*** -0.11*** 0.23*** ANALFOLL 0.02 0.1*** 0.10*** -0.10*** 0.52*** SIZE 0.05** 0.11*** 0.18*** 0.22*** 0.53*** dEPS -0.02 0.00 0.19*** -0.07*** 0.00 -0.03
Panel B: First and second value measure sample
V1 V2 FLS LEV ANALFOLL SIZE NUMSEG PROFIT SALEGROW
FLS -0.02** -0.03 LEV 0.26*** 0.27*** 0.01 ANALFOLL 0.37*** 0.31*** 0.02 -0.01*** SIZE 0.18 0.10*** 0.04* -0.30*** 0.52*** NUMSEG 0.10 0.06** 0.05** -0.06* 0.10*** 0.16*** PROFIT 0.11 0.14*** 0.02 0.08*** 0.10*** 0.13*** 0.03 SALEGROW 0.23*** 0.21*** -0.03 0.04*** 0.08*** 0.01 0.02 0.07*** SURPRISE -0.07 -0.08*** -0.01 -0.06*** -0.08*** -0.10*** -0.02 0.14*** -0.07***
Table 4 -Continued Panel C: Third value measure sample
V3 FLS LEV ANALFOLL SIZE NUMSEG PROFIT SALEGROW
FLS -0.05 LEV 0.010 0.01 ANALFOLL 0.11* 0.01 -0.01*** SIZE 0.10** 0.03 -0.30*** 0.52*** NUMSEG 0.11 0.05** -0.06* 0.10*** 0.16*** PROFIT 0.09 0.01 0.08*** 0.09*** 0.12*** 0.03 SALEGROW 0.12* 0.01 0.04*** 0.05** -0.02 0.01 0.06** SURPRISE -0.04 -0.01 -0.06*** -0.08*** -0.1*** -0.02 0.15*** -0.06**
4.2 Hypothesis Tests
4.2.1 Hypothesis 1
Hypothesis 1 tests if there is a positive relationship between the degree of geographic diversification of a firm and the amount of forward looking information on segments it discloses, the results are shown in Table 5. For Hypothesis 1 to be confirmed, the coefficient of FORSALE should be positive and significant. As is shown in Table 5, the coefficient is 0.352 and very significant, indicating that firms with relatively more foreign sales also provide more forward looking sentences. Because FORSALE is foreign sales divided by total sales, a one percent increase in foreign sales with total sales held equal would result in one percent more forward looking sentences. Seeing as the number of forward looking sentences is not very high, a one percent increase is not meaningful. The increase in foreign sales is relevant since it would mean that a company, regardless of the absolute size of sales, with mainly foreign sales is more likely to provide forward looking sentences. The direction of the coefficients and the significance of the various control variables is consistent with Baginski et al. (2014). For example, the coefficient of the change in annual earnings is negative, but not statistically significant. Assuming FORSALE is a fitting proxy for geographic diversification, this would mean that more geographically diversified firms also supply more forward looking sentences. Based on this I accept the first hypothesis.
4.2.2 Hypothesis 2
Hypothesis 2 tests if there is a negative relationship between the amount of forward looking information on segments disclosed and the value destruction caused by geographic diversification, the results are shown in Table 6. For Hypothesis 2 to be confirmed, the coefficients of FLS for the firm value measure should be positive and significant. As Table 6 shows, the coefficient of FLS for V1 is positive (0.257) and significant at 10%. Since firm value is measured as the market value of equity minus the book value of equity divided by total sales, more forward looking sentences are positively associated with either a higher market value of equity compared to its book value or lower sales compared to equity. This would mean that, on average, a firm’s excess value is 25.7% higher than a firm with one forward looking sentence less. The direction of the coefficients of the control variables are generally similar to the results reported for the model used as a basis (Bens and Monahan, 2004). For example, the number of analysts following the firm is positively related to the value measure, which means that firms with less value destruction have more analysts following the firm. Furthermore, the coefficient related
to firm size is negative, which means that larger firms have less value destruction from diversification than smaller firms. However, since the coefficient is low, it might not be economically significant. Based on the results, if V1 accurately measures excess value due to diversification, I find some evidence to support the hypothesis.
Table 5 – Hypothesis 1 Regression Results This table shows the regression results for equation (1) for two samples. The first sample uses FORSALE as the independent variable, where higher values mean greater diversification. The second sample uses HERF as the independent variable, where lower values mean greater diversification. The results for these variables are presented in bold.
Explanatory variables FLS (DIV) FLS (HERF)
FORSALE 0.352*** - (0.104) HERF - -0.307*** (0.113) dEPS -0.00519 -0.00569 (0.00984) (0.00988) PERF -0.0614 -0.0624 (0.0390) (0.0391) LEV -0.162** -0.169** (0.0797) (0.0800) ANALFOLL -0.00240 -0.00259 (0.00234) (0.00235) SIZE 0.0288** 0.0303*** (0.0115) (0.0116) Y1996 0.0222 - (0.0564) Y1997 0.124** 0.101* (0.0564) (0.0563) Y1998 0.193*** 0.194*** (0.0562) (0.0566) Y1999 0.0251 0.0163 (0.0562) (0.0565) Y2000 - -0.0151 (0.0566) Constant 0.0981 0.406*** (0.0895) (0.103) Observations 1,950 1,940 R-squared 0.020 0.019
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Values for year-dummies are missing because the variable was excluded due to multicollinearity.
Table 6 – Hypothesis 2 Regression Results This table shows the regression results for equation (2) for each of the three firm value measures. The results for the independent variables of interest are presented in bold. VARIABLES V1 V2 V3 FLS 0.257* -0.0727 -6.865 (0.131) (0.0457) (5.343) LEV -2.081*** -1.153*** 1.062 (0.441) (0.153) (18.50) ANALFOLL 0.0674*** 0.0506*** 0.233 (0.0134) (0.00465) (0.555) SIZE -0.0493 -0.0139 4.057 (0.0674) (0.0234) (2.811) INVEST 1.221 -1.394*** 42.27 (1.230) (0.428) (50.41) NUMSEG 0.0485 0.0389 1.687 (0.0709) (0.0247) (2.950) PROFIT -0.709 0.554*** 6.819 (0.479) (0.167) (19.92) SALEGROW 1.303*** 0.852*** 22.03* (0.342) (0.119) (12.58) SURPRISE 0.0780 -0.0711 -0.873 (0.177) (0.0614) (7.340) SFAS 0.303 0.0474 -1.800 (0.520) (0.181) (21.66) Y1996 0.208 -0.187 -25.38* (0.614) (0.214) (13.64) Y1997 0.335 -0.0135 - (0.614) (0.214) Y1998 -0.300 -0.204 -24.43 (0.358) (0.125) (20.75) Y1999 - - -16.54 (25.63) Y2000 -0.165 -0.235** -11.38 (0.328) (0.114) (25.62) Constant 0.444 1.399*** -16.89 (0.804) (0.280) (25.22) Observations 1,835 1,835 1,810 R-squared 0.053 0.174 0.009
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Values for year-dummies are missing because the variable was excluded due to multicollinearity.
5 Robustness Tests
5.1 Hypothesis 1Hypothesis 1 examines if there is a positive relationship between the degree of geographic diversification of a firm and the amount of forward looking information on segments it discloses. For testing this hypothesis I have used FORSALE as a measure of geographic diversification and was also used by Hitt et al. (1997). This measure has, however, been criticized by some for being unidimensional, which is why Sullivan (1994) recommended a multidimensional measure. I do not use this multidimensional measure because Ramaswamy et al. (1996) tested this measure and found no support that this measure was better than the one it aimed to replace. In order to test the robustness of my results I do consider another measure of diversification, a sales-based Herfindahl index.
The sales-based Herfindahl index is used to determine robustness because it has often been used in prior literature as a measure of diversification (Comment and Jarrell, 1995; Denis et al., 1997; Kim and Mathur, 2008; Lang and Stulz, 1994). A description of how the index is calculated is provided in the Appendix. Because not all firms have data available for calculating the index I had to drop some observations, resulting in 1,940 observations for 388 firms (see Table 1). The sample is similar to the sample originally used to test the hypothesis and the descriptive statistics of the index (HERF) are shown in Panel B of Table 3. If a firm has one segment its Herfindahl index is 1 and the more diversified it is, the lower the index is. On average the diversification using the index is 0.5, the most diversified company in my sample has an index of 0.15 and the least diversified company has an index of 1. Because the index is opposite in measuring diversification to FORSALE, I also expect the coefficient to be opposite, negative instead of positive. A negative coefficient would mean that, as the index goes up the number of forward looking sentences provided goes down. In other words, as a firm becomes less diversified, it provides less forward looking information on segments. Similar to the original test, I run an OLS regression on the panel data, the results of which are shown in Table 5 under FLS (HERF). Because the sample is similar to the one originally used to test the hypothesis and only the independent variable of interest has been replaced, the coefficients of the control variables are also similar, in direction, magnitude, and significance. Confirming my expectation, the results show that the coefficient of HERF is negative and significant (-0.307). This means that as the index approaches 1, the number of forward looking sentences provided becomes lower. Put in numbers, a fully diversified company provides 0.3 forward looking sentences more than a firm without diversification, a difference which of itself is negligible since it effectively
does not change the amount of sentences provided. The result does confirm the hypothesis stating more diversified companies provide more forward looking information on segments. The result is consistent with the result for the other diversification measure used, strengthening the confirmation of the first hypothesis.
5.2 Hypothesis 2
Hypothesis 2 examines if there is a negative relationship between the amount of forward looking information on segments disclosed and the value destruction caused by geographic diversification. For testing the second hypothesis I have used V1, as used in Bodnar et al. (1997), as a measure for excess value from diversification. In their research they also provide two other measures as alternatives for the one I have used. In order to determine the robustness of my results I have performed the same test using both of these alternative measures.
The first measure (V2) is not only used by Bodnar et al. (1997), but it is also similar to Tobin’s q used by Morck and Yeung (1991). It is defined as the market value of a firm as a percentage of its book value. The sample is similar in size, characteristics, and correlations to the original sample used for the hypothesis, as can be seen in Table 1, 3, and 4 respectively. The sample consists of 1,835 observations, the firm value measure has a mean of 1.895 but a median of 1.369 which means the distribution is somewhat skewed to the right. As with V1, higher values of this measure means a firm has less value destruction due to diversification. For that reason the second hypothesis is considered to be confirmed if the coefficient of FLS is positive, which I expect it to be. Again, I run an OLS regression on the clustered panel data, and the results are presented in Table 6 in the column under V2. What is interesting to note is that, opposite to the original measure used, more investment opportunities decreases firm value. Contrary to what I expected, the coefficient of FLS (-0.0727) is negative but not significant. This means that firms that provide more forward looking sentences have lower firm values. In other words, firms with greater value destruction due to diversification provide more forward looking information on segments, and thus more information on something that reduces firm value.
The second measure (V3) used to determine the robustness of my results is price per share divided by earnings per share. The sample construction is included in Panel D of Table 3, the dropped observations were due to missing values when calculating V3, resulting in a final 1,810 observations. As with the previous two measures, higher values indicate less value destruction as a result of diversification. The characteristics of the control variables are similar to those used in the previous samples, which is also true for the correlations presented in Panel C
the minimum indicates a significant value loss. As with the other value measures, no strong correlations exist between the dependent variable and the independent variables. For hypothesis 2 to be confirmed, firm value should increase with the amount of forward looking sentences disclosed. Since higher values of V3 mean higher firm values, I expect the coefficient of FLS to be positive, which would confirm the hypothesis. For this too I run an OLS regression on the clustered panel data and the results are shown in Table 6 in the column under V3. One of the notable differences with the previous two regressions is that there is only one coefficient which is statistically significant. The coefficient of the variable of interest (FLS) is -6.865, similar to the result for V2 but much larger, which can be explained by the difference in minimum and maximum of the two measures. Again, this result would mean that firms with greater value destruction due to diversification provide more forward looking information on segments, and thus more information on something that reduces firm value.
Based on the original results I find weak support in favour of the second hypothesis. In order to test the robustness of this result I have performed regressions for two alternative measures. Neither of these measures provide evidence in support of the hypothesis, the results, however, are not statistically significant. Because I do not find enough evidence to reject the second hypothesis, the support of the original results still stands.
6 Conclusion
This study examines the effect of diversification on the amount of forward looking information disclosed. I also examine the effect of this disclosure on firm value. Specifically, I examine how the degree of geographic diversification of a firm affects the number of forward looking sentences on segments a firm voluntarily discloses. The second part of my research focuses on the influence of these disclosures on the value destruction caused by diversification.
I start by examining if the amount of foreign sales relative to total sales is positively related to the number of forward looking sentences provided. I find that, as the percentage of foreign sales to total sales increases, the number of sentences provided also increases. This indicates that firms with more diversification also provide more information related to this diversification. This result is robust to another diversification measure commonly used in prior literature. Next, I examine if the number of sentences provided is negatively related to the value destruction caused by diversification. I find weak evidence supporting this, implying that firm value is higher for firms disclosing more forward looking information on segments.
My findings imply that voluntarily disclosed forward looking information can be used to increase firm value, meaning that investors must value this type of information. They also imply that managers provide more information when the firm is more geographically diversified. Since voluntary disclosure of information reduces information asymmetry between managers and investors, my results show that managers use forward looking information to reduce agency costs. This is consistent with Bravo (2015) and Ellis et al. (2012). This study is, to my best knowledge, the first to examine the effect of geographic diversification on this specific type of voluntary disclosures and the findings contradict findings on other types of voluntary disclosure.
My results contradict much research on voluntary disclosures and diversification, which finds that diversification is often associated with less voluntary, or segment disclosures (Dedman and Lennox, 2009; Harris, 1998). This study contributes to the literature on voluntary disclosures, specifically forward looking, by confirming that results for forward looking disclosures might not be the same as for other types of voluntary disclosures. Another contribution is one meant for managers and shareholders of geographically diversified firms. Since managers can benefit from increased firm value through compensation (Chan and Watson, 2011), disclosing more forward looking information could be beneficial to both shareholders and managers.
This study is subject to some limitations. First, the second model I estimate is based on Bens and Monahan (2004), but I could not use the same firm value measure. As a result, although the results are similar, one has to keep the difference in measures used in mind. Second, the data is partly hand-collected which limits the sample size. Classification of the sentences is subject to errors caused by human judgment. As mentioned before, because forward looking information is a specific type of voluntary disclosure it might not generalize to other types. Due to the nature of the data, being sentences about foreign segments, domestic-only firms cannot be included in the research. This is however not a serious limitation because the results can, by design, not be generalized to these firms since they have no foreign segments.
