Voluntary disclosure, mandatory disclosure, and cost of capital*

Transcription

1 Voluntary disclosure, mandatory disclosure, and cost of capital* Jing He Assistant Professor of Accounting Department of Accounting & MIS University of Delaware Marlene A. Plumlee 1 Professor of Accounting David Eccles School of Business University of Utah Marlene.plumlee@utah.edu Jennifer (He) Wen Assistant Professor of Accounting Department of Accounting University of Missouri-St. Louis Wenhe@umsl.edu May 2018 Abstract We examine the association between firms cost of capital and their voluntary and mandatory disclosures. We incorporate two mandatory disclosure types that arise within financial reporting environments: mandatory disclosure of periodic reports that are realizations of ex-ante reporting systems (periodic mandatory disclosures) and mandatory disclosure of specific corporate events (event-driven mandatory disclosures). To capture firms voluntary and event-driven mandatory disclosure quality, we employ information firms provide via 8Ks filings. To capture period mandatory disclosures, we employ earnings quality measures derived from the literature. Consistent with endogenous relations predicted by theory, we document that voluntary disclosure and both types of mandatory disclosure are correlated, although only event-driven mandatory disclosures are significant in models that explain voluntary disclosure. We find that the cost of capital is influenced by these disclosures, in separate regressions as well as when they are incorporated in a single model. These inferences are largely unchanged when all three aspects of firms disclosures are included in a single model. We find that voluntary disclosure is related to both types of mandatory disclosure, although controlling for them does not impact the association between voluntary disclosure and cost of capital. Also, event-driven mandatory disclosures are significant aspects of firms reporting environment. KEYWORDS Voluntary disclosure, mandatory disclosure, event-driven mandatory disclosure, periodic mandatory disclosure, earnings quality, cost of capital 1 Corresponding author. * We appreciate very helpful discussions with Rachel Hayes, Beatrice Michaeli, Xiaoxia Peng, and Jordan Schoenfeld and comments from Michael Neel (AAA discussant), David Plumlee, Peter Pope (the editor), three anonymous reviewers, and participants at the 2017 AAA annual meeting. 1

2 1. INTRODUCTION We examine the association between voluntary and mandatory disclosure, and the conditional and unconditional pricing effects of voluntary disclosure, given these associations. Our analysis includes voluntary disclosure and two types of mandatory disclosure: periodic mandatory reports produced by firms ex-ante reporting system and event-driven mandatory filings required when specific corporate or firm events occur. Our proxy for periodic mandatory reports is the common factor generated by factor analysis performed on three commonly used earnings quality measures. We follow Francis et al. (2008) in constructing this measure and concur with their view of earnings quality as capturing the precision of the earnings signal emanating from the firm s financial reporting system (pg. 54) Our primary proxies for voluntary disclosure and event-driven mandatory disclosure are based on a technique developed and validated in Cooper, He, & Plumlee (2018) (CHP hereafter), that uniquely partitions information from firms 8-K filings into information required by the SEC to be disclosed and voluntary information. This finer partitioning of firms disclosures allows a deeper examination of the relationships among earnings quality, cost of capital and financial disclosure. Whether mandatory and voluntary disclosure are complements or substitutes represents an ongoing issue in the analytical disclosure literature. For example, some analyses show that the link between voluntary and mandatory disclosure is conditional on whether the disclosure has either financial or real externalities (Dye, 1990), the purpose of the mandatory disclosure (Gigler & Hemmer, 1998), and the features of the mandatory disclosure environment (Einhorn, 2005). Other studies analyze settings where the structure of the mandatory disclosure system is endogenously determined, based on expectations around subsequent receipt of a private signal (e.g., Stocken & Verrecchia, 2004; Freidman, Hughes, & Michaeli, 2018a). As we discuss in Section 2, conclusions from this literature lead to conflicting expectations about the expected relationship between voluntary and mandatory disclosure, due in part to assumptions regarding endogeneity of the mandated disclosures. This body of work motivates our first two hypotheses. A related issue is the link between voluntary disclosure and a firm s cost of capital. Many theoretical and empirical studies examine the association between a firm s cost of capital 2

3 and the quality of its disclosure. 2 The findings from these literatures are also somewhat mixed, though they are generally consistent with voluntary disclosure being inversely related to the cost of capital. These findings motivate our third hypothesis. An unresolved issue in the literature is whether studies that provide evidence of the link between voluntary disclosure and the cost of capital capture the impact of voluntary disclosure or whether the voluntary disclosure measures capture more primitive aspects of firms reporting environment. The concern is that, while an empirical link between voluntary disclosure and cost of capital is established, the true relationship is between the underlying aspects of the reporting environment that are correlated with voluntary disclosure. For example, several empirical studies explore how the inclusion of earnings quality in the analysis impacts the association between voluntary disclosure and the cost of capital (e.g., Francis et al., 2008; Bhattacharya et al., 2012; Heflin, Moon, & Wallace, 2016). 3 We expand this research by explicitly examining how firms mandatory disclosure is related to voluntary disclosure and how controlling for mandatory disclosure affects our understanding of the voluntary disclosure/cost of capital link. A unique aspect of our analysis is the inclusion of two types of mandatory disclosure. Our empirical analysis begins by examining the relationships between voluntary disclosure and periodic mandatory disclosure, and event-driven mandatory disclosure. We document positive correlations between voluntary disclosure and both types of mandatory disclosure. Once we control for firm characteristics expected to predict voluntary disclosure, however, the link between periodic mandatory disclosure and voluntary disclosure becomes insignificant. Event-driven mandatory disclosure continues to provide significant explanatory power. In contrast to the findings in Francis et al. (2008), our results suggest that voluntary disclosures are unrelated to the quality of a firm s mandatory reporting of periodic events, instead they appear to be responses to firm specific corporate events that trigger mandatory disclosure. We also find that greater voluntary disclosure is associated with the cost of capital, 2 For example, the theoretical literature includes Diamond & Verrecchia (1991), Kim & Verrecchia (1994), Zhang (2001), Bertomeu, Beyer, & Dye (2011), Cheynel (2013), Clinch & Verrecchia (2015). Empirical studies in this literature include Botosan & Plumlee (2002), Gietzmann & Ireland (2005), Hail & Leuz (2006), Francis, Nanda, & Olsson (2008), Baginski & Rakow (2012), Bhattacharya, Ecker, Olsson, & Schipper (2012). This is only a partial list. 3 In these studies, earnings quality proxies for the information environment. 3

4 although the sign of the association varies by the cost of capital measure (implied cost of equity, bid-ask spread, or cost of debt) and voluntary disclosure measure (8-K-based measure or management forecasts) employed. When we use the implied cost of equity or cost of debt to capture the cost of capital and our primary measure of voluntary disclosure, the unconditional association with the cost of capital is positive or insignificant. 4 When we use bid-ask spreads to proxy for cost of capital or rely on management forecasts to capture voluntary disclosure, the unconditional relation between voluntary disclosure and the cost of capital is significantly negative. We also document that firms with a greater number of eventdriven mandatory filings have higher costs of capital, and that firms with higher quality periodic mandatory disclosure have lower cost of capital, regardless of the cost of capital proxy employed. When we include all three disclosure variables in the model, we find that doing so does not affect the unconditional relation between the mandatory disclosure proxies and the various measures of the cost of capital. In most cases, we also find that the inclusion of the either or both of the mandatory disclosure proxies does not impact the unconditional relation between voluntary disclosure and the cost of capital. These findings are consistent with voluntary and mandatory disclosures having an impact on firms cost of capital, but do not support the contention that earnings quality subsumes the voluntary disclosure quality and that excluding earnings quality in the analysis leads erroneous conclusions. Ultimately, our findings extend those in Francis et al. (2008) and Heflin et al. (2016). We show that even though periodic mandatory disclosures and event-driven mandatory disclosures are correlated with voluntary disclosures and with the cost of capital, the documented association between voluntary disclosure and the cost of capital are unaffected by the inclusion of the mandatory disclosure variables. Our study contributes to the literature in several distinct ways. First, we link earnings quality and periodic mandatory disclosure, which forms a more comprehensive framework within which to examine the associations between firm disclosure and firms cost of capital. Second, we increase our understanding 4 Documenting a positive association between cost of capital and voluntary disclosures is inconsistent with many theoretical models, which generally predict a negative relation between voluntary disclosure and cost of equity capital due to a reduction in information asymmetry (e.g., Amihud & Mendelson, 1986; Diamond & Verrecchia, 1991) or in estimation risk (e.g., Barry & Brown, 1985). If, however, public information is a complement rather than a substitute for private-information gathering, more public information will reduce market liquidity as it intensifies adverse selection some models (e.g., Kim & Verrecchia, 1994). 4

5 of the link between voluntary and mandatory disclosure by providing empirical evidence of these relationships. Third, we expand the literature around the conditional impact of voluntary disclosure on the cost capital by explicitly examining how firms mandatory disclosure, partitioned into periodic mandatory disclosure and event-driven mandatory disclosure, is related to voluntary disclosure and how controlling for both types of mandatory disclosure affect our understanding of the voluntary disclosure/cost of capital link. Fourth, we demonstrate that the relation between voluntary disclosure and the cost of capital is robust to the inclusion of both types of mandatory disclosure. Ultimately, even though mandatory disclosure is associated with both voluntary disclosure and the cost of capital, it does not subsume the relation between voluntary disclosure and the cost of capital. Finally, we illustrate how CHP s 8K-based voluntary and mandatory disclosure measures can be used to enhance our understanding of firms disclosure environments. The rest of the paper proceeds as follows. Section 2 discusses the analytical literature that investigates the relation between voluntary and mandatory disclosure and the relation between disclosure and a firm s cost of capital. Based on these models, we generate testable hypotheses. Sections 3 and 4 describe the primary empirical proxies (disclosure measures and the cost of capital measures) and describe the sample and presents descriptive statistics. Section 5 presents our empirical results. Section 6 presents robustness tests and section 7 concludes. 2. HYPOTHESES ON THE RELATIONS AMONG VOLUNTARY DISCLOSURE, MANDATORY DISCLOSURE, AND THE COST OF CAPITAL The purpose of this study is to provide empirical evidence that enhances our understanding of the role of voluntary disclosure in a firm s cost of equity capital, conditional on the endogenous nature of voluntary disclosure. Our examination is motivated by and builds on the discussion and analysis provided by Francis et al. (2008). We begin with a review of papers that examine the interdependencies between a firm s voluntary and mandatory disclosures. Our analysis departs from prior work by explicitly considering two mandatory disclosure types that arise within the financial reporting environment: periodic 5

6 reports that are a realization of ex-ante reporting systems and event-driven mandatory filings. We rely on several theoretical studies (e.g., Gietzmann & Trombetta, 2003; Jorgensen & Kirschenheiter, 2003; Bertomeu & Magee, 2015; Friedman, Hughes, & Michaeli, 2018a) as a basis for our arguments. After we describe the expected relations between voluntary disclosure and the two aspects of mandatory disclosure, we detail our predictions of the association between cost of capital and the quality of firms disclosures, conditional on the endogenous nature of the voluntary disclosure. (i) Voluntary and Mandatory Disclosures A broad theoretical literature explores the expected association between the quality of voluntary disclosures and the quality of a firm s information environment. Early work, which views voluntary disclosure quality as exogenous, finds that voluntary disclosure is used to mitigate information asymmetry (e.g., Grossman 1981). Treating voluntary disclosure quality as exogenous, however, runs counter to intuition and evidence that suggests that the quality of voluntary disclosure is likely to be related to the underlying information environment (e.g., Coller & Yohn, 1997; Chen, Defond, & Park, 2002) A number of theoretical studies take various approaches to modeling the interdependencies between voluntary disclosure quality and the quality of the underlying information environment. In some settings, the models demonstrate that an increase in the quality of the information environment leads to an increase in the probability of voluntary disclosure. 5 The complementary relationship is driven by the market s inference when a manager does not disclose private information; when the manager is more likely to be informed or has higher quality information, the market is more likely to interpret nondisclosure as bad news (e.g., Dye, 1985; Verrecchia, 1990). Thus, a firm s disclosure threshold i.e., the lowest realization the manager is willing to disclose decreases as the quality of the information environment increases, and firms with higher (lower) quality information disclose more (less). This complementary association between disclosure and the quality of the information system is not universal, 5 This literature includes Verrecchia (1983, 1990), Dye (1985), Einhorn (2005). 6

7 however. Penno (1997) shows that if the probability a manager is informed is lower when the information quality is higher, disclosure and information quality may serve as substitutes. The paper considers several variations of the relation between information quality and the probability of being informed and shows that under multiple scenarios there are regions where voluntary disclosure and information quality will be substitutes. Several theoretical studies endogenize the process by which mandatory disclosures are generated, with a focus on the determination of the disclosure rules and of the reporting systems that produce disclosures (e.g., Stocken & Verrecchia, 2004; Bertomeu & Magee, 2015; Friedman et al. 2018a, 2018b). Stocken & Verrecchia (2004) show that when managers anticipate they will have private information that is not captured within the firms financial reporting system and can manipulate the financial reports, they may not choose the most precise reporting system. Bertomeu & Magee (2015) first model a setting where mandatory disclosure policies are formed in response to managers preferences and political compromise. 6 They demonstrate that the resulting policies mandate asymmetric reporting and that certain adverse (below the median) news be disclosed. Next they expand the model to include the presence of a voluntary disclosure channel and explore the impact it might have on both the endogenously determined mandatory disclosure policies and voluntary disclosure. The expanded model predicts the (1) survival of the asymmetric nature of the mandatory disclosure rules and (2) substitutive nature of voluntary and mandatory disclosure. Similarly, Friedman et al. (2018a) model the endogenous determination of firms reporting systems and voluntary disclosures. Their study focuses on the ex ante design of the reporting system when managers have expectations about the receipt of private information and its disclosure subsequent to the release of realized values from the reporting system. In this setting, the likelihood a firm provides voluntary disclosure is linked to the properties of its reporting system, which is in turn linked to the probability that the firm receives private information. Specifically, they find that firms with a higher (lower) probability of receiving and disclosing private information could select a 6 As noted by the authors, the assumptions that underlie this model including the managers influence on the policies obtained are consistent with stylized facts of the rule-setting process in accounting. 7

8 less (more) informative reporting system. These studies provide analysis that link mandatory disclosure policies and reporting systems, which we contend are related to a specific type of mandatory disclosures. We draw from these models and current financial reporting outcomes to motivate the examination of two types of mandatory disclosure disclosure based on periodic reports that are a realization of an exante reporting system and disclosure of specific events. Bertomeu & Magee (2015) and Friedman et al. (2018a) analyze links with mandatory disclosure policies and reporting systems, which we contend are more closely related to the first type of mandatory disclosure. Empirically, we provide independent proxies for the quality of each mandatory disclosure type, and investigate the relations between voluntary and mandatory disclosure, conditional on the mandatory disclosure type. 7 We present hypotheses on the expected relation between voluntary disclosure and both types of mandatory disclosure below. The first hypothesis draws on the findings from Bertomeu & Magee (2015) and Friedman et al. (2018a) and predicts a substitutive association between a firm s periodic mandatory disclosure and its voluntary disclosure: H1a: Voluntary disclosures are decreasing in the quality of the periodic mandatory reports provided by a firm s accounting system. The second hypothesis predicts an association between mandatory disclosures triggered by to specific corporate events and voluntary disclosure. It is less clear from the literature whether this type of 7 This perspective is supported both theoretically and empirically. For example, Gigler & Hemmer (1998) emphasize the confirmatory role of mandatory reports, where mandatory disclosures are verifiable but noisy and, possibly, late signals of managers private information. Friedman et al. (2018a) describe a setting where voluntary disclosure is influenced by the realized reports generated by the reporting system. These descriptions comport well with reported earnings. Empirical analysis that documents a link between mandatory and voluntary disclosures centers on earnings-quality-related mandatory disclosures (e.g., Francis et al., 2008; Ball, Jayaraman, & Shivakumar, 2012; Li & Yang, 2016). Francis et al. (2008) examine the link between an accruals-based earnings quality measure and voluntary disclosure; their primary findings are consistent with a complementary relation between mandatory and voluntary disclosure. Ball et al. (2012) frame their empirical analysis as an examination of the confirmation hypothesis, where audited financial statements increase the credibility of voluntary disclosures. Consistent with the predications provided by Gigler & Hemmer (1998), Ball et al. predict and find a complementary association between mandatory and voluntary disclosures. Finally, Li & Yang (2016) use changes in the reporting system due to the adoption of IFRS to explore the impact of mandatory disclosure on voluntary disclosures. They examine the channels by which IFRS adoption might impact voluntary disclosures, including reduced pressure to provide voluntary disclosure when more is known a priori about a firm due to higher quality mandatory reporting (Verrecchia, 1990). They document an increase in voluntary disclosure after the adoption of IFRS, inconsistent with the substitutive relation predicted by Verrecchia (1990). Each of these studies employs mandatory disclosure proxies that are based on the outcome of a reporting system: earnings quality, audited financial statements, and the adoption of IFRS. 8

9 mandatory disclosure will have a complementary or substitutive association with voluntary, however. Thus, we predict an association although we do not provide an expectation of the sign of the relationship. H1b: Voluntary disclosures are associated with the quality of the event-driven mandatory reports. (ii) Voluntary Disclosure and the Cost of Capital A sizable body of theoretical research supports a link between the cost of capital and disclosure. One stream of this literature suggests that, in a market with differentially-informed investors, adverse selection impacts the cost of capital. To reduce this influence firms are willing to commit to higher levels of disclosure to reduce information asymmetry between investors and firms cost of capital (e.g., Grossman, 1981; Verrecchia, 1983; Diamond & Verrecchia, 1991; Easley & O Hara, 2004). Another stream of literature suggests that estimation risk drives the link between cost of capital and disclosure. In this setting, information improves investors assessments of the parameters of the distribution of the future payoffs, which reduces the uncertainty of those payoffs and will lead to a decrease in the cost of capital (Barry & Brown, 1985; Coles & Lowenstein, 1988; Coles, Lowenstein, & Suay, 1995). Cheynel (2013) shows that firms that voluntary disclose have lower costs of capital than firms that do not. In contrast with much of the theoretical literature, this negative association exists in the cross-section of expected returns, consistent with empirical models frequently employed. Regardless of the mechanism that links disclosure and the cost of capital, many of the prior findings demonstrate that firms with more disclosure will face a lower cost of capital than firms with less disclosure. This prediction is not universal, however. For example, Kim & Verrecchia (1991) show that the opposite relation might exist: firms with more disclosure have a higher cost of capital. In their setting, increased disclosure motivates investors to seek additional private information, which leads to an increase in information asymmetry between informed and uninformed investors and a higher cost of capital. Another analytical study (Zhang, 2001) demonstrates that, conditional on the setting, the link between the cost of capital and disclosure could be positive or negative, depending on the factors that cause variation 9

10 in disclosure levels. His model suggests that a negative association between disclosure level and the cost of capital exists only when variation in disclosure is due to disclosure costs. When the primary drivers of variation in disclosure are other factors (e.g., variability in earnings or variability in liquidity shocks) a positive association between disclosure levels and the cost of capital will obtain. 8 Prior empirical studies that examine the relation between voluntary disclosures and the cost of capital generally support a negative association, although this is not always the case. Several studies examine the link between voluntary disclosures based on AIMR scores and the cost of capital both equity and debt (e.g., Welker, 1995; Sengupta, 1998; Botosan & Plumlee, 2002; Heflin et al. 2016). Botosan & Plumlee (2002) regress the implied cost of equity on total AIMR scores and three related AIMR component scores that capture three aspects of a firm s disclosures (annual report, other publications, and investor relations). They fail to document a significant relation when total scores are used to capture voluntary disclosure. When they employ the component scores to capture voluntary disclosure, however, the quality of the annual reports (other publications) is negatively (positively) associated with the cost of capital. Heflin et al. (2016) replicate this analysis using a subset of the Botosan & Plumlee sample and document similar results. Sengupta (1998) and Heflin et al. (2016) document a negative association between the cost of debt and AIMR scores. In addition, Botosan (1997) and Francis et al. (2008) use hand-collected data from 10-Ks and document that more forthcoming firms have lower cost of capital, although Botosan (1997) documents this relation only for firms with low analyst following and Francis et al. s findings disappear or are weaker when they control for earnings quality. We note that when alternative proxies for the cost of capital (e.g., bid-ask spreads, illiquidity measures) are employed, the findings are generally consistent with a negative association between these measures and proxies for higher quality disclosure (e.g., Welker, 1995; Coller & Yohn, 1997; Heflin et al., 2016). The theoretical and empirical findings from prior studies are not clear-cut, with analytical and 8 There also are several theoretical studies that argue that information asymmetry is diversifiable in large economies and should not affect the cost of capital (e.g., Hughes, Liu, & Liu, 2007; Christensen, de la Rosa, & Feltham 2010). The no link result is frequently limited to settings with perfectly competitive markets. 10

11 empirical evidence of both negative and positive relations. Our hypothesis on the association between disclosure and cost of capital is unsigned. H2: Variation in the level of a firm s voluntary disclosure is associated with variation in its cost of capital. In the next section we discuss the expected associations between cost of capital and voluntary disclosure, conditional on two types of mandatory disclosure. (iii) Mandatory Disclosure and the Cost of Capital As discussed in section (i), whether managers decide to provide less, or more, voluntary disclosure when faced with higher quality or more mandatory disclosure hinges on whether voluntary disclosure serves as a substitute for or complement to mandatory disclosure. Regardless of the underlying relation, voluntary disclosure is generally portrayed as a response to information quality, which leads to a prediction that voluntary disclosure has a second-order effect on the cost of capital. Consistent with this intuition, we expect voluntary disclosure to have a second-order effect on the cost of capital, although we look to the periodic mandatory reports provided by a firm s accounting system as providing the first-order effect. We rely on the same empirical proxy (earnings quality) as Francis et al. (2008) to capture the quality of mandatory reports provided by a firm s accounting system; we argue that earnings quality reflects firms underlying information quality that is reflected in firms periodic mandatory reports. 9 We also expect event-driven mandatory disclosures have a first-order effect. 10 As discussed earlier, we predict that the quality of periodic mandatory reports provided by a firm s accounting system and voluntary disclosure have a substitutive relation. Thus, even if voluntary disclosure is negatively associated with the cost of capital, an empirical model that fails to control for the quality of periodic mandatory reports could report a positive relation between voluntary disclosure and the cost of capital due to the negative correlation 9 This view is consistent with prior studies, including Ball et al. (2012) and Li & Yang (2016). In addition, Francis et al. (2008) state that earnings quality is the precision of the earnings signal emanating from the firm s financial reporting system. 10 Consistent with the potential role of event-driven mandatory disclosures in pricing, McMullin, Miller, & Twedt (2018) document that price formation improves subsequent to changes in the SEC s reporting requirements around firm events. 11

12 between the omitted variable and voluntary disclosure. Additionally, failing to control for the quality of event-driven mandatory disclosures could influence the documented association between voluntary disclosure and the cost of capital. In this case, we predict an association between voluntary disclosure and event-driven mandatory disclosures, although it is unclear whether these disclosures serve as complements or substitutes. If voluntary disclosure and event-driven mandatory disclosure have a substitutive relation, the impact described above would again obtain. If, however, voluntary disclosure and event-driven mandatory disclosure serve as complements then increased mandatory disclosure would lead to increased voluntary disclosure. In this case, failing to control for the first-order effect of mandatory disclosure in a setting where voluntary disclosure is negatively associated with the cost of capital would still lead to a negative relation. Even so, without including the appropriate control for eventdriven mandatory disclosure we are unable to assess whether the documented relation is due to voluntary disclosure or the omitted variable. Ultimately, we are unable to predict the conditional relation between voluntary disclosure and the cost of capital. Irrespective of the association between voluntary disclosure and periodic mandatory disclosure and event-driven mandatory disclosure, however, providing empirical evidence of the association between voluntary disclosure and the cost of capital requires the inclusion of both types of mandatory disclosure. We state our formal prediction of the relation between voluntary disclosure and the cost of capital, conditional on mandatory disclosure as: H3: Controlling for the level of periodic mandatory disclosure and event-driven mandatory disclosure, variation in a firm s voluntary disclosure is associated with variation in its cost of capital. 3. EMPIRICAL MEASURES OF DISCLOSURE AND COST OF CAPITAL In this section we detail the construction of the primary empirical proxies used in our analysis: voluntary disclosure, both types of mandatory disclosure, and the cost of capital. Section (i) below provides a discussion of our voluntary and event-driven mandatory disclosure proxies, which are based on data from 12

13 firms 8-K filings. Sections (ii) and (iii) describe the proxies for periodic mandatory disclosure and the cost of capital, respectively. (i) Voluntary and Event-Driven Mandatory Disclosure Research that examines voluntary disclosure issues uses a number of proxies for firm-level voluntary disclosure, including (1) management earnings forecasts, (2) conference calls, (3) non-gaap earnings, (4) scores based on analysts or researchers assessments, and (5) 8-K-based proxies. 11 We elect to use an 8-K-based voluntary disclosure proxy developed and validated in CHP in our primary analysis for the following reasons. First, the CHP measure is available for a broad range of firms (publicly-listed firms over a twelve-year period) and therefore, the sample is not restricted to firms selected by external data providers or limited to firms covered by analysts or selected by researchers. It also is based on a broad range of voluntary disclosures that are reported in 8-K filings instead of relying on disclosures provided through a single channel (i.e., management earnings forecasts, conference calls, or non-gaap earnings). Limiting the analysis to a single disclosure channel ignores other firm disclosures that could be substitutes, which could lead to potential bias toward firms that choose one channel to communicate versus another. Third, CHP also develop an 8-K-based measure of mandatory disclosure, which serves as our proxy form event-driven mandatory disclosure. Unlike the voluntary disclosure literature, there are fewer obvious candidates to serve as proxies for event-driven mandatory disclosure. CHP calculate their disclosure measures for all publicly-listed firms that file 8-Ks with the Securities and Exchange Commission (SEC) between 2005 and As noted by Carter & Soo (1999) and Lerman & Livnat (2010) and exploited by CHP, information reported in 8-K filings can be categorized as (1) mandatory disclosures (disclosures about corporate events the SEC requires firms to report if the event happens to the firm) or (2) voluntary disclosures that reflect information firms choose 11 See (1) Baginski & Rakow, 2012, (2) Tasker, 1998; Brown, Lo, & Hillegeist. 2004; (3) Brown, Christensen, Elliott, & Mergenthaler, 2012, (4) Lang & Lundholm 1993; Botosan & Plumlee, 2002; Francis et al., 2008 and (5) Leuz & Schrand 2009; Schoenfeld 2017; Bourveau, Lou, & Wang

14 to disclose that the SEC then requires them to report in an 8-K. Relying on the SEC s 8-K filing requirements and the findings in Lerman & Livant (2010), CHP classify 8K item numbers reported within 8-K filings into those that are related to events that the SEC specifies be disclosed via an 8K (mandatory disclosures) and other disclosures where the manager has greater discretion (voluntary disclosures). Specifically, CHP classify items 2.02 (Results of Operations and Financial Condition), 7.01 (Regulation FD Disclosure), and 8.01 (Other Important Events) as voluntary. All other item numbers are classified as mandatory. 12 We use their variable VolDisc_Ct, the total number of items and exhibits classified as voluntary that are reported in a firm s 8Ks issued during a fiscal year, as our proxy for voluntary disclosure (VD_8K) and their variable ManDisc_Ct, as our proxy for event-driven mandatory disclosure (MD_Event). Appendix A provides detailed descriptions of these measures. We perform additional analysis using management forecasts (VD_MF) to proxy for voluntary disclosure, a proxy that is widely used in the literature. VD_MF equals the log of the sum of the values assigned to each management forecast for all forecasts issued during the fiscal year. Each forecast was assigned a value based on the forecast form: one for qualitative forecasts, two for range forecasts and open-ended forecasts, and three for point forecasts (Francis et al., 2008). (ii) Periodic Mandatory Disclosure A number of empirical studies use earnings quality as a proxy for information quality or information risk, which links information quality to the quality of the firm s reporting system. 13 We take this notion a step further and assert that information quality is a function of information generated by the financial reporting system, such that earnings quality proxies for mandatory disclosure related to the reporting system used to generate periodic reports. 12 See Cooper, He, and Plumlee (2018) available at for additional details. 13 Francis et al. (2008) state, we view our earnings quality variable as a proxy for the information quality of the firm s financial information system (footnote 1, pg. 54). Bhattacharya, Ecker, Olsson, & Schipper (2012) examine the direct and mediated paths by which earnings quality impacts the cost equity, where earnings quality serves as a proxy for information risk. 14

15 Consistent with the literature and Francis et al. (2008), we calculate three related measures of earnings quality (accrual quality (EQ AQ ), earnings variability (EQ EV ), and absolute abnormal accruals (EQ AbsAA )). We then calculate EQ CF, a common factor score obtained from a factor analysis of these three measures. 14 EQ CF is our proxy for the quality of the mandatory disclosures generated from the reporting system MD_EQ. (iii) Cost of Capital We include three implied cost of equity measures (r MPEG, r DIV, r STPEG ) for use in our analyses, consistent with prior studies in this literature. r MPEG is based on the modified price-earnings-growth ratio method, following the procedure detailed in Easton (2004). r DIV is estimated based on the target price method identified in several studies (e.g., Botosan, Plumlee, & Wen 2011; Botosan & Plumlee, 2013). We use the I/B/E/S 12-month ahead target price forecast, which mirrors the process employed in Francis et al. (2008) and Heflin et al. (2016). 15 We rely on the findings in Botosan et al. (2011) to support the use of the implied cost of capital measures above. Nonetheless, we recognize that the literature has failed to reach consensus on the best measure of the cost of capital. Consequently, we include additional analysis using alternative cost of capital measures: bid-ask spread and the cost of debt. Bid-ask spread (AvgBAS) is calculated as the average daily quoted bid-ask spread over a fiscal year. Our proxy for the cost of debt (SPRating) is the 14 Specifically, EQ AQ, is the standard deviation of residuals from a regression of total current accruals on cash flow from operations, change in revenues, and gross value of property, plant and equipment using data from year t-10 to year t (e.g., the data from is used to calculate the 2005 EQ AQ ). EQ EV, is the standard deviation of earnings before extraordinary items scaled by total assets. Data from year t-9 to year t is used to estimate the measure for year t. EQ AbsAA, is based on modified Jones (1991) approach. We follow Francis et al. (2008) and use a ten-year rolling data (from year t-9 to year t) to generate average absolute abnormal accruals, AA j. Thus, EQ AbsAA for 2005 is the firm-level average AA j,t over the estimation period ( ). EQ CF, is the common factor score obtained from a factor analysis of EQ AQ, EQ EV, and EQ AbsAA. We multiply this value by a negative one, so a higher value of earnings quality indicates higher quality mandatory disclosure. See Appendix B for a detailed description of these measures. 15 Prior studies use Value Line three-to-five-year-ahead target price forecasts to capture the firm terminal value, while we use I/B/E/S analysts target price one-year-ahead forecasts, due to data availability. These values are qualitatively the same as the r DIV measure employed in Francis et al. (2008), which allows for a meaningful comparison between our results and Francis et al. and related studies. 15

16 Standard and Poor s domestic long-term issuer credit rating. Details of constructing these variables are provided in Appendix A. 4. SAMPLE AND VARIABLE DESCRIPTION AND DESCRIPTIVE STATISTICS We begin our sample in 2005, after firms were required to comply with the SEC s Additional Form 8K disclosure requirements and acceleration of filing date (August 23, 2004). The new guidance expanded the items to be reported in 8Ks and shortened the filing deadline, which is likely to affect firms measured disclosure practices. 16 To be included in our sample, we first require that a firm-year has data to calculate the 8K-based disclosure measures and the earnings quality based disclosure measures from Using firm-years that meet these data requirements, we then form three samples to use in our analysis. These samples differ based on the alternative cost of capital measure used in the analysis. The implied cost of capital sample is limited to firm-years for which we can estimate the implied cost of capital measures and the relevant control variables. The bid-ask spread sample is limited to firm-years for which we can estimate the bid-ask spread and the relevant control variables, and the cost of debt sample is limited to firm-years for which we can calculate the cost of debt and the relevant control variables. Restricting the sample to firm-year observations that have all three measures would result in a small sample. We obtain firm-level financial accounting information from Compustat, stock-return information from CRSP, analyst forecast information from I/B/E/S, and the 8-K-based disclosure measures from CHP. This process generates an implied cost of capital sample (ICC) of 20,228 firm-year observations, a bidask spread sample (BAS) of 27,669 firm-year observations, and a cost of debt sample (COD) of 8,733 firm-year observations. The control variables we include vary by the sample. The control variables for the ICC sample are Beta (estimated from a firm-specific CAPM regression using monthly data for the five preceding 16 The new SEC rule generally requires 8Ks to be filed within four business days after the event date. See CHP for a detailed discussion of the timing. Lerman and Livnat (2010) examine the timeliness of 8K filing for a sample from 2005 through 2007 and find that 8Ks are filed in a timely manner after the enforcement of the new rule. Specifically, they document that 95 percent of material events are filed within four business days of the event dates and that mandatory and voluntary items exhibit similar timeliness. 16

17 fiscal years), MVE (the market value of equity measured at the beginning of the fiscal year), and BTM (the firm s book-to-market ratio measured at the beginning of the fiscal year). We log both MVE and BTM when including it in ICC regressions. The control variables for the BAS sample are AvgVol, (the natural log of the average daily firm trading volume over the prior fiscal year) AvgPrc (the natural log of the split adjusted daily firm price over the prior fiscal year), LnAssets (the natural log of a firm s total assets at the end of the prior fiscal year), and LnBTM (defined above). The control variables for the COD sample include three variables defined above (LnMVE, Beta, and LnBTM) and ROA (return on assets, calculated as operating income before depreciation divided by the beginning of the year total assets), LEV (financial leverage, calculated as the beginning of the fiscal year total liabilities divided by total assets), and IntCov (interest coverage, calculated as pretax income plus interest expense divided by interest expense). Table 1 presents summary statistics for the disclosure, cost of capital, and control variables included in the empirical analyses. Panel A (B, C) presents statistics for the ICC (BAS, COD) sample. To facilitate comparisons across the three samples, we present summary statistics for Beta, MVE, and BTM in each panel. While we present descriptive statistics for each sample, our discussion focuses on the ICC sample. Firms in our sample file an average of just over 30 8K items and exhibits a year, classified as voluntary and classified as mandatory. 17,18 We also document significant variation in the our two proxies for mandatory disclosure: MD_Event and MD_EQ. Our sample mean (median) cost of equity capital values range from percent (10.29 percent) to percent (20.67 percent), depending on the estimation method. These values are consistent with those reported in prior studies. For example, Botosan (1997) reports a mean of 20.1 percent (19.0 percent) for 122 manufacturing firms in 1991; Francis et al. (2008) show a mean (median) value of 16.6 percent (15.6 percent) for 677 firms in their sample during the year of The mean (median) firm in our sample has a market value of equity of $6.56 billion ($1.09 billion), suggesting that our sample includes smaller firms than in Francis et al. 17 The number of 8K items and exhibits exceed the number of 8K filings since firms frequently report both items and exhibits and/or report multiple items in a single 8K. 18 These values are larger than the mean values reported in CHP. They report mean values of K items and exhibits a year, classified as voluntary and classified as mandatory. 17

18 (2008) and firms that are closer to the Compustat firms during the period. Finally, VD_MF, a proxy for voluntary disclosure based on management forecasts, is available for only about half of the firms in our full sample. For the set of firm years for which we have the data, VD_MF has a mean (median) value of (10). Again, we document meaningful variation in this measure. Panel B of Table 1 presents similar statistics for the BAS sample. Generally, the mean/median values of the disclosure variables are consistent with the ICC sample. When we compare firm size, riskiness, and growth (MVE, Beta, and BTM), we see that the BAS sample includes slightly smaller (mean MVE of versus ), equally risky (mean Beta of 1.29 versus 1.30), and slightly higher growth (mean BTM of 0.56 versus 0.52) firm-years than the ICC sample. Average daily price (AvgPrc) has a mean of and median of These values are generally consistent with those reported in prior research. For example, Amiram, Owens, & Rozenbaum (2016) report daily average price of (median of 24.91). Panel C of Table 1 presents analogous statistics for the COD sample. Generally, the mean/median values of the disclosure variables are higher than in the ICC sample. When we compare firm size, riskiness, and growth (MVE, Beta, and BTM), we see that the COD sample includes much larger (mean MVE of 13, versus 6,560.37), less risky (mean Beta of 1.18 versus 1.30), and slightly higher growth (mean BTM of 0.55 versus 0.52) firm-years than the ICC sample. 5. EMPIRICAL RESULTS (i) Tests of the Relation between Voluntary Disclosures and Periodic and Event-Driven Mandatory Disclosures: H1a and H1b. We begin by examining the associations between the three aspects of corporate information flow: voluntary disclosure, periodic mandatory disclosure, and event-driven mandatory disclosure. H1a predicts a substitutive relation between voluntary disclosure and periodic mandatory disclosure (higher-quality periodic mandatory disclosure will lead to lower-quality voluntary disclosure). H1b predicts that 18

19 voluntary disclosure will be associated with event-driven mandatory disclosure, although we do not predict a sign. We test these hypotheses using both a univariate and multivariate model, which includes firm characteristics drawn from prior studies. Specifically, we include firm size (lnmve) to control for firm-specific factors prior studies suggest influence disclosure policy (e.g., Bamber & Cheon, 1998). We also include the natural log of BTM (lnbtm) to control for the association between growth and voluntary disclosure (Bamber & Cheon, 1998; Nagar, Nanda, & Wysocki, 2003). Lang & Lundholm (1993) show that voluntary disclosure (measured with AIMR scores) is associated with the number of analysts that follow the firm. Based on their finding, we include #Analyst (the number of analysts that follow the firm during the prior fiscal year) and ROA (return on assets calculated as operating income before depreciation divided by total assets) in the model. We include #Segment (the number of segments the firm reports at the end of the prior fiscal year) to control for firm complexity (Nagar et al. 2003). Finally, we include Issue (an indicator variable that equals one if the number of shares outstanding, adjusted for stock dividends/splits, increases by more than twenty percent from year t-1 to year t, and zero otherwise), as Frankel, McNichols, & Wilson (1995) and Lang & Lundholm (2000) document links between equity issuance and a firm s voluntary disclosure. Panel A of Table 2 presents pairwise correlations between VD_8K, MD_Event, MD_EQ and the firm characteristics included in the multivariate model. We show that both types of mandatory disclosure MD_Event and MD_EQ are consistently positively associated with voluntary disclosure, although the association with MD_Event is much larger. The average correlation between VD_8K and MD_Event is 0.393, compared to the average correlation with MD_EQ (0.103). Panel B provides further evidence using multivariate analysis. 19 We present results when MD_EQ is included in the model (column (2)), when MD_Event is included in the model (column (3)), and when both are included (column (4)). We fail to document a significant association between VD_8K and MD_EQ in the multivariate setting (columns (2) and (4)), inconsistent with the univariate setting. In contrast, VD_8K and MD_Event are 19 We estimate our model using the largest of our three samples the BAS sample. 19

20 consistently positively associated (columns (3) and (4)). Our results are inconsistent with H1a, where we predict a substitutive relation between VD_8K and MD_EQ, although they are consistent with our unsigned expectations in H1b. 20 In sum, our results suggest that the quality of a firm s voluntary disclosure is (1) unrelated to the quality of the reports produced from the firm s mandatory reporting system and (2) a complement to event-driven mandatory disclosures. 21 We also find that voluntary disclosure is positively associated with firm size and the number of segments a firm reports and negatively associated with the book-to-market ratio and the number of analyst that follow it. 22 (ii) Test of H2 and H3: Conditional and Unconditional Associations Between Cost of Capital and Voluntary and Mandatory Disclosure The empirical associations between VD_8K and MD_Event and MD_EQ we document are important in interpreting prior work that investigates how voluntary disclosure influences the cost of capital. Furthermore, while we fail to document that MD_EQ is related to VD_8K, there is theoretical support for the endogenous nature of the mandatory reporting system and voluntary disclosures. We examine the empirical association between a firm s cost of capital and its disclosures using a series of regression models. We first provide evidence of the unconditional relation between the cost of capital and voluntary disclosure (H2) and then expand the model to provide evidence of the relation between the cost of capital and voluntary disclosure, conditional on mandatory disclosures. We estimate a series of 20 Failing to document a significant association between voluntary disclosure and earnings quality measures is also inconsistent with Francis et al. 2008, who document a significant complementary relation between EQ and voluntary disclosure based on information provided within 10-K filings. 21 In untabulated analysis, we estimate the Table 2 model without fixed effects and compare the adjusted R 2 s of the base model with and without the two mandatory disclosure proxies: periodic mandatory disclosures (MD_EQ) and event-driven mandatory disclosures (MD_Event). This allows to evaluate the relative importance of these measures in explaining voluntary disclosure. When we add MD_EQ to the base model, the adjusted R 2 of the model increases from 10.8% to 10.9%. In contrast, when MD_Event is added to the base model, the adjusted R 2 increased from 10.8% to 23.1%. 22 Our analysis mirrors the analysis in Francis et al. (2008) (their Table 3), extended by the addition of MD_Event. Our analyses and results differ from theirs, however, in two significant ways. First, while our univariate analysis is consistent with theirs (VD_8K and MD_EQ have a complementary relation), we fail to document a significant association between VD_8K and MD_EQ in the multivariate analysis. Second, we find that our proxy for voluntary disclosure is significantly associated with additional firm controls, relative to their study. We believe that differences in our proxies for voluntary disclosure likely impact these results: our primary proxy for voluntary disclosure is based on data drawn from 8-K filings, while they construct a measure based on disclosures within 10-K filings. 20