Externalities of Disclosure Regulation: The Case of Regulation FD
Michael Crawley,a Bin Ke,b and Yong Yuc
ABSTRACT We use Regulation Fair Disclosure (REG FD) to examine a relatively neglected but important effect of disclosure regulation: externalities. REG FD applies to all publicly traded U.S. firms, but foreign firms cross-listed on U.S. stock exchanges are explicitly exempt. Despite the exemption, we find that many cross-listed firms voluntarily adopt REG FD as part of their disclosure policies. We hypothesize that REG FD imposes two externalities on cross-listed firms. First, following REG FD previously disadvantaged U.S. investors have a lower demand for shares of cross-listed firms that continue to follow a selective disclosure policy. Second, REG FD creates an information spillover effect on cross-listed firms whose values and cash flows are correlated with those of U.S. firms. We find evidence of both effects in cross-listed firms’ voluntary REG FD adoption decisions. Further, relative to non-adopters, cross-listed firms who voluntarily adopt REG FD exhibit a significant reduction in the information asymmetry component of cost of capital. REG FD adopters are also more likely than non-adopters to switch to open disclosure post REG FD. These results suggest that cross-listed firms’ voluntary REG FD adoption represents a credible commitment to increased disclosure transparency. Keywords: Regulation Fair Disclosure; disclosure externalities; cross-listed firms JEL codes: G, K2, M4, N2
December 8, 2010 We thank Kris Allee, Rob Bloomfield, Zhihong Chen, Fred Choi, Dain Donelson, Yuyan Guan, Jeff Hales, Steve Huddart, John Jiang, Bob Lipe, Steve Kachelmeier, Bill Kinney, Sidney Leung, Craig Nichols, Sheridan Titman, and workshop participants at Boston University, City University of Hong Kong, Cornell University, FASRI Roundtable, Michigan State University, Pennsylvania State University brownbag seminar, Singapore Management University, University of Texas at Austin, and the 2009 AAA annual meeting for helpful comments and Laura Abrahamson and Mallory Valverde for able research assistance. The paper was formerly titled “why do cross-listed firms voluntarily adopt regulation fair disclosure?”. a
Kelley School of Business, Email:
[email protected].
Indiana University,
Bloomington,
IN
47405.
Tel:
+1
812-855-0951.
b
Nanyang Business School, Nanyang Technological University, S3-01B-39, 50 Nanyang Avenue, Singapore 639798. Tel: +65 6790-4832. Fax: +65 67913697. Email:
[email protected]. c
McCombs School of Business, University of Texas at Austin, 1 University Station, Austin, TX 78712. Tel: +1 512471-6714. Email:
[email protected].
I. INTRODUCTION Disclosure plays a central role in many countries’ securities regulatory frameworks. Over the past decade, a flurry of significant securities laws and regulations (e.g., Regulation Fair Disclosure and the Sarbanes Oxley Act of 2002) and the world-wide mandatory adoption of IFRS have generated a strong interest in understanding the economic consequences of disclosure regulation. While a large body of empirical research has examined firm-specific costs and benefits of disclosure regulation (see, e.g., Greenstone et al. 2006; Bushee and Leuz 2005; Gintschel and Markov 2004; Zhang 2007; Gao et al. 2007), 1 Leuz and Wysocki (2008) conclude after surveying the extant literature that there is a paucity of evidence on the externalities of disclosure regulation. Understanding the externalities of disclosure regulation is important because such externalities provide a possible justification for the necessity of disclosure regulation. Firmspecific effects of disclosure regulation are undoubtedly relevant for evaluating the economic consequences of disclosure regulation. However, the mere existence of firm-specific net benefit from disclosure is not sufficient to justify mandatory disclosure. This is because when the firmspecific net benefit is positive, firms ought to have an incentive to voluntarily provide the disclosure and thus the necessity of mandatory disclosure regulation becomes less clear. Unfortunately, Leuz and Wysocki (2008) indicate that debates about disclosure regulation often incorrectly focus on firm-specific costs and benefits of disclosure choices rather than the aggregate effects of disclosure regulation. The objective of this study is to use Regulation Fair Disclosure (REG FD) as a specific setting to understand the economic forces that determine the degree of externalities associated
1
See Leuz and Wysocki (2008) for a detailed review of this literature.
1
with disclosure regulation. REG FD prohibits management of all publicly traded U.S. firms from sharing material nonpublic information with select investors, particularly financial analysts and institutional investors. However, American Depository Receipts (ADRs) traded on the U.S. stock exchanges (referred to as cross-listed firms), which represent an important and growing segment of the U.S. financial market, are explicitly exempt from the regulation. 2 Despite the exemption, a 2004 survey of 143 large cross-listed firms conducted by Broadgate Capital Advisors, Value Alliance and the Bank of New York reported that 54% of the respondents voluntarily adopted REG FD as part of their disclosure policies. Why did so many cross-listed firms voluntarily follow REG FD? We conjecture that cross-listed firms’ decision to voluntarily adopt REG FD represents externalities resulting from the mandatory implementation of REG FD by U.S. firms. It is reasonable to assume that managers of both U.S. firms and cross-listed firms optimized their disclosure policies prior to the passage of REG FD. Managers that expected a positive (negative) net benefit from an open disclosure policy should have chosen an open (selective) disclosure policy and investors should have priced the firms’ shares accordingly. REG FD forced all U.S. firms to follow an open disclosure policy. The evidence from existing research indicates that REG FD has been effective in reducing the private communication between company management and sophisticated investors (such as financial analysts and institutional investors) and increasing the level playing field among investors with respect to access to management’s private information (see, e.g., Gintschel and Markov 2004; Francis et al. 2006; Ke et al. 2008).
2
From 1990 to 2000, the number of foreign firms listed on NYSE and NASDAQ grew steadily from about 170 to over 750 with cumulative trading volume in these firms reaching more than $750 billion. As of 2002, foreign firms listed on NYSE represented nearly 17% of all NYSE listings (see Coffee 2002).
2
We hypothesize that REG FD imposes two non-mutually exclusive externalities on crosslisted firms. First, following the passage of REG FD, the information asymmetry among external investors should be reduced in U.S. firms and thus previously disadvantaged U.S. investors (particularly retail investors) are likely to find it more attractive trading in U.S. firms than in cross-listed firms (denoted as investor demand effect). Consistent with this prediction, Bushee et al. (2004) find that following REG FD the amount of individual investor trading increased during conference calls for U.S. firms that previously held closed calls relative to U.S. firms that previously held open calls. There is also evidence that following REG FD, U.S. firms experienced declines in return volatility and increases in both trading volume and informational efficiency, consistent with a shift toward a more transparent information disclosure environment (see, e.g., Heflin et al. 2003; Bailey et al. 2006). If cross-listed firms choose not to follow REG FD, they risk losing investors to U.S. firms, which in turn will reduce their stock liquidity and increase their expected returns in markets that are not perfectly competitive (see Merton 1987; Fishman and Hagerty 1989; Pastor and Stambaugh 2003). The risk of investor loss is particularly high for cross-listed firms because their major investor clientele is U.S. retail investors (Matthew et al. 2007). As a result, crosslisted firms should face the pressure to follow REG FD, especially those firms that have a larger U.S. retail investor clientele. 3 Second, U.S. firms’ implementation of REG FD may create an information spillover on cross-listed firms (denoted as information spillover effect). Assuming that firm values and cash flows of U.S. firms are correlated with those of cross-listed firms, previously disadvantaged
3
Disadvantaged investors may also take other actions to pressure cross-listed firms to adopt REG FD, such as increased shareholder activism and the refusal to purchase products from cross-listed firms. Such actions should be easier to justify in the post-REG FD period due to U.S. firms’ implementation of REG FD.
3
investors can use U.S. firms’ open disclosures mandated under REG FD to update their beliefs about cross-listed firms’ firm values and cash flows and would assume the worst in the event of no disclosure by cross-listed firms. As shown in Dye and Sridhar (1995), this revision in investors' perception should induce cross-listed firms to voluntarily follow REG FD in order to distinguish themselves from other cross-listed firms with worse information, resulting in an unraveling of managers’ private information (Grossman 1981; Milgrom 1981). Therefore, in the absence of disclosure costs, cross-listed firms should find it optimal to pool with U.S. firms by voluntarily switching from a selective disclosure policy to an open disclosure policy. We use both email and telephone surveys to obtain information on cross-listed firms’ REG FD adoption decision. 43% (181/422) of the active cross-listed firms responded to our survey and 40% (72/181) of the respondents stated that they had voluntarily adopted REG FD as part of their formal disclosure policy, a finding consistent with the 2004 Broadgate survey. In terms of economic significance measured using total assets, the 72 REG FD adopters represent approximately 36% of the 422 cross-listed firms in our initial sample and 60% of the 181 crosslisted firms in our final sample. We use the institutional ownership of ADR shares as an inverse proxy for the investor demand effect. We use the percentage of foreign sales and the stock return synchronicity between cross-listed firms and U.S. firms per Piotroski and Roulstone (2004) as two alternative proxies for the information spillover effect. Consistent with our expectations, we find that the probability of REG FD adoption is higher for cross-listed firms that have a smaller institutional ownership, a higher percentage of foreign sales and a higher stock return synchronicity. Consistent with the existing voluntary disclosure literature, we also find that the probability of REG FD adoption is lower for cross-listed firms that are expected to gain less from a voluntary
4
open disclosure policy, i.e., those firms that have a lower investment opportunity set, a lower demand for external financing, and greater managerial agency problems. We find evidence that cross-listed firms’ voluntary adoption of REG FD represents a credible commitment to increased disclosure transparency. First, relative to the non-adopters, adopters exhibit a reduction in the bid-ask spread and an increase in share turnover. Second, adopters are more likely than non-adopters to switch to open conference calls after the voluntary REG FD adoption. Specifically, we find that the majority of both adopters and non-adopters held closed conference calls prior to the passage of REG FD. Following the passage of REG FD, the majority of the adopters but not the non-adopters switched to open conference calls. This study makes several important contributions. First, we are one of the few studies that examine the externalities of disclosure regulation. While the earlier disclosure literature has examined intra-industry information transfers associated with earnings announcements (see, e.g., Foster 1981), very few studies have investigated the externalities of disclosure regulation. 4 One exception is Bushee and Leuz (2005). They examine a 1999 SEC regulatory change that mandates all domestic firms quoted on the over-the-counter bulletin board to file periodic financial reports with the SEC. They find the regulatory change resulted in positive externalities for domestic firms previously filing with the SEC. Anand et al. (2006) also find weak evidence that the Sarbanes and Oxley Act imposes externalities on Canadian firms not listed in the U.S. However, neither Anand et al. (2006) nor Bushee and Leuz (2005) examine the specific mechanisms through which the externalities of disclosure regulation work.
4
Even though we use cross-listed firms, which are exempt from REG FD, to illustrate the externalities of disclosure regulation, it is important to note that externalities of disclosure regulation also exist among firms subject to a common regulation. For example, one firm’s disclosure of information mandated by a regulation could be also useful to the valuation and investment decisions of another firm that is subject to the same regulation (see Leuz and Wysocki 2008).
5
Second, we contribute to the understanding of cross-listed firms’ voluntary disclosure incentives. There appears to be a presumption in the extant literature that cross-listed firms do not have an incentive to follow U.S. laws from which they are exempt (see, e.g., Licht 2003; Gomes et al. 2007; Francis et al. 2006; Chen et al. 2010). We show that this presumption may be premature. Our study should be of interest to researchers who wish to use cross-listed firms as a control group. In the context of REG FD, the evidence from our study suggests that cross-listed firms are not homogenous with respect to their voluntary compliance with U.S. securities laws and thus using all cross-listed firms as a control sample may reduce a researcher’s ability to detect REG FD-related effects for publicly traded U.S. firms. Third, our study is relevant to the legal bonding versus reputational bonding debate in the cross listing literature. The legal bonding hypothesis (see Coffee 1999; Stulz 1999) attributes at least a portion of the benefit of a U.S. cross listing (e.g., an increase in stock market valuation or access to outside finance) to more stringent U.S. securities laws, stronger SEC enforcement power, and a greater threat of litigation by minority shareholders. However, recent research by Siegel (2005) raises questions on the legal bonding hypothesis. 5 Siegel (2005) finds that the SEC and minority shareholders have not effectively enforced U.S. laws against cross-listed firms who violated U.S. laws. However, Siegel (2005) finds that cross-listed Mexican firms that abide by the U.S. laws without exploiting minority shareholders during an economic downturn (i.e., firms with a clean reputation) are able to receive more privileged long-term access to outside finance than those that have a bad reputation. Siegel (2005) argues that reputational bonding better explains the success of U.S. cross listings than legal bonding. Our results lend further support to Siegel’s reputational bonding hypothesis. More importantly, we identify the economic forces that
5
See Karolyi (2006) and Benos and Weisbach (2004) for a comprehensive survey of the cross listing literature.
6
cause cross-listed firms to voluntarily bond with tougher U.S. securities laws from which they are exempt. Fourth, our study provides timely and relevant information to the SEC who indicated in the final rule of REG FD that it would undertake a comprehensive review of cross-listed firms’ reporting requirements. Although we cannot directly assess the consequences of mandating all cross-listed firms to adopt REG FD, our study provides the first ex-ante archival evidence on the major costs and benefits involved in cross-listed firms’ existing disclosure choices and helps the SEC better predict the economic consequences of imposing REG FD on all cross-listed firms. The rest of the paper is organized as follows. The next section describes the sample selection procedures and survey results. Section III discusses the regression results on the determinants of cross-listed firms’ REG FD adoption decision. Section IV shows the results on the effect of voluntary REG FD adoption on the information asymmetry component of cost of capital. Section V reports the analysis of cross-listed firms’ open conference calls in the pre- and post-REG FD periods. Section VI provides a series of sensitivity checks to rule out alternative explanations. Section VII concludes.
II. SAMPLE SELECTION PROCEDURES AND SURVEY RESULTS The Sample of Cross-Listed Firms We use several data sources to identify our cross-listed firm sample, including CRSP, Compustat, Citibank, and the Bank of New York. The initial sample includes all cross-listed firms that were listed on the NYSE, AMEX or NASDAQ in the form of American Depository Receipts (ADRs) at some point in time from January 1, 2000 to December 31, 2005. 6 As REG
6
We do not consider level I ADRs that trade over the counter or SEC Rule 144A private placements to qualified institutional buyers because information for those firms is scarce. We do not expect these firms to adopt REG FD
7
FD took effect in October 2000, we exclude cross-listed firms that were delisted prior to January 1, 2000. As we started our data collection in 2006, we exclude cross-listed firms that became listed after December 31, 2005. We exclude the few Canadian ADRs because most Canadian firms are directly listed in the U.S. In addition, Canada had a “tipping” rule in effect designed to limit selective disclosure prior to REG FD and thus REG FD may have little effect on crosslisted Canadian firms’ voluntary disclosure behavior. 7 The resulting sample contains 552 unique cross-listed firms, of which 422 were active and 130 were inactive as of the end of 2006 when we started our survey.
An Analysis of Survey Responses To increase the survey response rate, our survey questionnaire asks two factual questions about REG FD (see Appendix A for a sample copy of the survey instrument). We conducted our survey in two stages. First, we emailed our survey questionnaire to all 522 cross-listed firms’ investor relations representatives in four rounds of emails. We obtained the email addresses from the cross-listed firms’ U.S. or home country web sites. The first round emails were sent on November 20, 2006 and the last round emails were sent on March 9, 2007. Second, starting in the middle of March 2007, we started to make phone calls to the firms that did not respond to our four rounds of email inquiries. We first called a firm’s U.S. office and then the headquarters at its home country during business hours. We followed the same survey questionnaire in Appendix A for the phone survey.
because their investors should be generally sophisticated and thus they should not face the same REG FD related investor pressure as the ADRs that trade on the major stock exchanges. 7
In 2002 the Canadian Securities Administrators (CSA) issued NP 51-201 (Disclosure Standards) to further interpret and clarify some issues in the “tipping” rule.
8
Of the 552 cross-listed firms, 203 firms responded to our inquiry, of which 10 firms refused to answer our questions and 4 firms provided unusable answers. This leaves us with 189 usable firms, of which only 8 (or 6% of all inactive cross-listed firms) are inactive cross-listed firms. Because of the inactive cross-listed firms’ low response rate and the high likelihood that required financial data on inactive cross-listed firms are unavailable, we dropped the 130 inactive cross-listed firms from our analyses. As a result, the usable response rate for the active crosslisted firms is 43% (181/422), which is high compared with the typical 10-15% survey response rate documented in recent surveys of financial executives (e.g., Graham et al. 2005). As we do not have the survey responses for all of the 422 active cross-listed firms, we also examine the effect of this survey response bias on our empirical results in Appendix B. We find no evidence that the survey response bias causes a material effect on our inferences. Of the 181 active respondents, 75 answered the email inquiry, and 106 answered the phone inquiry (79 provided the answers on the phone immediately and 27 provided the answers in an email following the phone call). Appendix C provides the distribution of the 422 active cross-listed firms and the number of firms that responded to our survey by country. 72 of the 181 respondents (40%) claimed to have voluntarily adopted REG FD. This result suggests that a significant percentage of cross-listed firms chose to voluntarily follow REG FD, consistent with the 2004 Broadgate survey. Due to missing values for some of the regression variables, we lose another 3 firms and therefore our final sample contains 178 active cross-listed firms, of which 70 firms claimed to have adopted REG FD. Regarding our second survey question, the adoption date is always the REG FD effective rate (i.e., October 23, 2000) for the adopters listed on a U.S. stock exchange before October 23,
9
2000. The adoption date is the U.S. listing date for the adopters listed on a U.S. stock exchange after October 23, 2000.
Survey Response Accuracy A common concern with survey research is the accuracy of survey responses. This concern could be more acute in our case because we did not include additional questions to cross-check the validity of the responses to our two factual survey questions. However, it is important to note that conducting a standard survey with multiple questions is not feasible in our setting because the typical response rate for a standard survey is around 10-15% (see Graham et al. 2005). Given that the population of active cross-listed firms is only 422 during our sample period, conducting a standard survey would result in a useable sample that is too small for our study. 8 We address potential survey response errors in two ways. First, in Sections IV-V, we directly use cross-listed firms’ stock market liquidity and ex post open disclosure behavior before vs. after the REG FD adoption date to check the validity of our survey responses. To the extent that our survey responses are reliable and credible, both stock market liquidity and the likelihood of open disclosure should be increased in the post-REG FD period for adopters relative to non-adopters. However, if some survey respondents provided incorrect or intentionally biased responses, we should be less likely to find the predicted effects in Sections IV-V. Second, for the 178 cross-listed firms included in our final sample, we directly cross check the accuracy of the survey responses with publicly available data sources, including the FACTIVA data base, which covers all the major news wires, newspapers and magazines,
8
We wish to thank Bill Kinney for the suggestion of using a short survey to increase the response rate.
10
company web sites, SEC filings, and available conference call transcripts. We discuss the details of this analysis in Section VI.
III. DETERMINANTS OF CROSS-LISTED FIRMS’ REG FD ADOPTION Variable Definitions and Predictions Proxies for the Investor Demand and Information Spillover Effects Investor demand: REG FD was designed to level the playing field among investors of publicly traded U.S. firms with regard to information access to company management. Following REG FD, previously disadvantaged U.S. investors should find it more attractive trading in U.S. firms than in cross-listed firms. Hence after the passage of REG FD we expect all cross-listed firms to face some pressure to follow REG FD. For example, Remond (2000) reports that after the passage of REG FD, many cross-listed foreign firms felt obligated to follow REG FD in order to keep pace with U.S. firms’ disclosure standards. Leading advisors to ADR firms (e.g., Bank of New York Co., depositary bank for 65% of the world's depositary receipts) also recommended that their ADR clients at least operate in the spirit of REG FD, if not the letter (see, e.g., Remond 2000; Platt 2000; Rosenbaum 2001). However, the investor demand for adopting REG FD is likely higher for firms that have a larger base of U.S. retail investors, who were often at a disadvantage relative to analysts and institutional investors in information access to company management prior to the passage of REG FD. 9 Because the size of a cross-listed firm’s U.S. retail investor base is not observable, we
9
In theory the investor demand effect should apply to all previously disadvantaged retail investors, including home country retail investors. However, in practice the investor demand effect is likely much smaller for home country retail investors because the latter face greater transaction costs (e.g., foreign currency exchange risk and a host of issues associated with setting up and managing a foreign stock trading account) in shifting their stock investments from cross-listed firms to U.S. firms.
11
use INSTITUTIONOWN (defined below) as an inverse proxy for small investors’ pressure for cross-listed firms to follow REG FD. INSTITUTIONOWN is defined as the fraction of crosslisted firm’s ADR shares owned by U.S. institutional investors as disclosed in the Spectrum database. 10 , 11 For cross-listed firms listed on a U.S. stock exchange prior to the REG FD effective date, INSTITUTIONOWN is measured at the end of the calendar quarter prior to the REG FD effective date. For cross-listed firms listed on a U.S. stock exchange after the REG FD effective date, INSTITUTIONOWN is measured at the end of the calendar quarter following the firm’s U.S. stock listing date. We predict the coefficient on INSTITUTIONOWN to be negative. Information spillover: U.S. firms’ implementation of REG FD may also induce crosslisted firms to voluntarily follow REG FD in order to avoid being labeled by investors as firms with worse information. This information spillover effect is expected to increase with the correlation between U.S. firms’ and cross-listed firms’ values and cash flows (see Dye and Sridhar 1995). We capture this correlation in two ways. First, we use a cross-listed firm’s percentage of annual foreign sales as disclosed in Compustat (denoted as FOREIGNSALE) to proxy for this correlation. The idea is that as a cross-listed firm has more foreign sales, its value and cash flows are more likely to be correlated with those of U.S. firms because it is very likely that a significant portion of cross-listed firms’ foreign sales are in the U.S. 12 For cross-listed firms listed on a U.S. stock exchange prior to the REG FD effective date, FOREIGNSALE is
10
An alternative and potentially preferred definition of INSTITUTIONOWN is to include the ownership of home country institutional investors. Unfortunately, ownership data for home country institutions are not readily available.
11
To make sure INSTITUTIONOWN is not a proxy for the size of the ADR float relative to a cross-listed firm’s worldwide market cap, we also include the latter as a control in Table 2 and obtain similar inferences (untabulated). The coefficient on this new control variable is insignificant. 12
An alternative and better proxy for the correlation between U.S. firms’ and cross-listed firms’ value and cash flows is the percentage of a cross-listed firm’s U.S. sales. Unfortunately many cross-listed firms do not separately disclose U.S. sales.
12
measured at the end of the fiscal year prior to the REG FD effective date. For cross-listed firms listed on a U.S. stock exchange after the REG FD effective date, FOREIGNSALE is measured in the fiscal year prior to or in the year of the firm’s U.S. stock listing date. We expect the coefficient on FOREIGNSALE to be positive. Second, we measure the correlation between U.S. firms’ and cross-listed firms’ values and cash flows using the stock return synchronicity (denoted as SYNCHRONICITY) derived from the following firm-specific model in Piotroski and Roulstone (2004, 1123): Ri ,t = α + β1 MARETi ,t −1 + β 2 MARETi ,t + β 3 INDRETi ,t −1 + β 4 INDRETi ,t + ε i ,t
(1)
where i=firm index t=time index R=the weekly return of a cross-listed firm. MARET=the weekly value weighted average market return of all U.S. firms. INDRET=the weekly value weighted average return of all U.S. firms in the same two-digit SIC code as the cross-listed firm. For cross-listed firms listed in the U.S. prior to the REG FD effective date, equation (1) is estimated using weekly returns from CRSP over a one-year period that ends one day prior to the REG FD effective date. For cross-listed firms listed in the U.S. after the REG FD effective date, equation (1) is estimated using weekly returns over a one-year period that starts with the firm’s U.S. listing date. We require a minimum of 45 weeks for the estimation. SYNCHRONICITY is defined as the natural logarithm of
R2 , where R2 is the coefficient of determination from (1 − R 2 )
the estimation of equation (1). Higher values of synchronicity imply a stronger correlation in
13
firm values and cash flows between a cross-listed firm and U.S. firms in the same industry. We expect the coefficient on SYNCHRONICITY to be positive.
Proxies for Other Voluntary Disclosure Determinants Conditional on the REG FD-driven investor demand and information spillover effects, cross-listed firms’ REG FD adoption decision should also depend on the common costs and benefits of voluntary disclosure identified in the existing literature: investment opportunities, demand for external equity finance, managerial agency costs, information asymmetry between management and outside investors, proprietary costs, and disclosure complexity (see, e.g., Healy and Palepu 2001; Bushee et al. 2003). 13 Below we introduce the proxies for each of the above mentioned voluntary disclosure determinants. Unless stated otherwise, all the financial data required for the subsequent regression variables are drawn from Compustat, Worldscope, CRSP, or hand-collected from companies' SEC filings. Unless stated otherwise, the following control variables are measured at the end of the fiscal year prior to the REG FD effective date for crosslisted firms listed on a U.S. stock exchange prior to the REG FD effective date, and measured in the fiscal year prior to or in the year of the firm’s U.S. stock listing for cross-listed firms listed on a U.S. stock exchange after the REG FD effective date. Investment opportunity set and demand for external equity financing: Prior research finds that the benefit of increased disclosure is larger for firms that have more investment opportunities that cannot be satisfied by internally generated cash flows and debt financing. The definitions of both the investment opportunity set and demand for external equity financing follow Durnev and Kim (2005) and Demirguc-Kunt and Maksimovic (1998). A firm’s 13
We do not consider litigation risk as a determinant because as noted in Bushee et al. (2003), under existing fraudon-the-market rules, individuals can sue regardless of whether they heard the information directly or not.
14
investment opportunity set (INVEST_OPP) is defined as the 2-year geometric average of the annual percentage growth in net sales. A firm’s demand for external equity financing (EXTERNAL_FIN) is defined as the difference between the firm’s actual growth rate and the sustainable growth rate with retained earnings and short-term and long-term debt financing that maintain a constant debt-to-assets ratio. The actual growth rate is the 2-year geometric average of the annual growth rate in total assets and the sustainable growth rate is the 2-year average of ROE/(1-ROE), where ROE is the return on equity. Both INVEST_OPP and EXTERNAL_FIN are winsorized at the 1% and 99% percentiles. As Durnev and Kim (2005) note, these two proxies are measured using ex ante information and thus do not suffer from the same degree of endogeneity as other proxies for the investment opportunity set (e.g., the book-to-market ratio) and the demand for external equity financing (e.g., the future ex post realized equity financing). We expect the coefficients on both variables to be positive. Managerial agency costs: Following Gong et al. (2010), we measure the managerial agency costs using the extent of divergence between the voting rights and cash flows rights of the company management group (including officers, directors, and their immediate family members) who are controlling shareholders of the firm. 14 We define a dummy variable CONTROL_WEDGE that equals one if a cross-listed firm’s management group represents the largest block holder of the firm by voting rights and its voting rights exceed cash flow rights. We define a dummy variable CONTROL_NOWEDGE that equals one if a cross-listed firm’s 14
We collect the voting rights and cash flows rights as follows. First, we match our cross-listed firm sample with the datasets used by Faccio and Lang (2002) and Lins (2003), which contain the voting rights and cash flow rights of the firm management group for many non-U.S. firms. Second, for the cross-listed firms that are not included in the above two studies, we hand collect the voting rights and cash flow rights of the firm management group using the approach described in Lins (2003). The data sources include SEC filings (proxy statements, 20-F, 40-F, or 10-K), company web sites, stock exchange web sites, as well as the other data sources listed in Faccio and Lang (2002), and Lins (2003). For the firms cross listed prior to REG FD, we use the year 2000 data if available and the most recent year data if year 2000 is not available. For the firms cross listed after REG FD, we use the year prior to the listing year if available and the most recent available data otherwise.
15
management group represents the largest block holder of the firm by voting rights and its voting rights do not exceed its cash flow rights. We expect management of CONTROL_WEDGE=1 firms to have both the ability and incentive to expropriate minority shareholders. Because increased disclosure transparency is expected to reduce management’s private control benefits (see, e.g., Shleifer and Wolfenzon 2002; Greenstone et al. 2006), we expect the coefficient on CONTROL_WEDGE to be negative. While management of CONTROL_NOWEDGE=1 firms is controlling shareholders and thus have the ability to expropriate minority shareholders, it does not possess voting rights in excess of its cash flow rights and thus should have no incentive to expropriate minority shareholders. Hence, we expect the coefficient on CONTROL_NOWEDGE to be insignificant. Information asymmetry: As noted at the beginning of this Section, all cross-listed firms face the threat of losing investors to publicly traded U.S. firms following the passage of REG FD. Fishman and Hagerty (1989) argue that it is costly for an investor to study the disclosure of every firm and thus an investor who does not follow a stock will be less likely to spend time studying and trading on the firm’s public disclosure; hence, losing investors will reduce a firm’s stock price efficiency in less than perfectly competitive markets (see also Merton 1987). In addition, a significant loss of investors’ interest in a stock may also reduce analysts’ incentive to cover the stock, which would further exacerbate the information asymmetry problem. Therefore, crosslisted firms that suffer from a higher information asymmetry between management and outside investors should be more likely to adopt REG FD. 15 We use the natural logarithm of total assets (LN(ASSETS)) and the natural logarithm of one plus the number of analysts following the firm from IBES (LN(ANALYST)) as proxies for the degree of information asymmetry. We expect the 15
However, U.S. firms’ open disclosure under REG FD should mitigate this information asymmetry to the extent that there is a positive correlation between U.S. firms’ and cross-listed firms’ values and cash flows.
16
coefficients on both variables to be negative. None of the regression coefficients in the REG FD adoption regression are affected if we drop either LN(ASSETS) or LN(ANALYST). Proprietary costs: Similar to Bushee et al. (2003), we assume that relative to a policy of selective disclosure, an open disclosure policy increases a firm’s risk of leaking proprietary information to product market competitors. A common proprietary cost proxy used in prior accounting research (e.g., Harris 1998) is the industry Herfindahl index, which intends to capture the degree of industry-level product market competition. However, relying on recent industrial organization studies, Karuna (2007) argues that product market competition is better captured using three dimensions, including product substitutability (SUBSTITUTION), market size (MKTSIZE), and entry costs (ENTRYCOST). Accordingly, we use the three dimensions of the product market competition rather than the industry Herfindahl index to measure proprietary costs. 16 Following Karuna (2007), SUBSTITUTION is the sum of sales in an industry (4-digit SIC) divided by the sum of operating costs in the same industry. MKTSIZE is the sum of sales in an industry (in millions of U.S. dollars). ENTRYCOST is the average gross PPE (in millions of U.S. dollars) in an industry weighted by each firm’s sales in the same industry. We express both MKTSIZE and ENTRYCOST in the natural logarithm form in the later regression model to reduce skewness. Higher values of SUBSTITUTION and ENTRYCOST and lower values of MKTSIZE represent lower industry competition. Because our sample firms are cross-listed foreign firms and industry segment data are not always available, we make a few simplifying assumptions in computing the three proprietary cost proxies. First, if a cross-listed firm derives 50% or more of its sales from its home country, we use the firm’s home country industry data to measure the three proxies. In addition, because 16
Inference is similar if using the industry Herfindahl index as a proprietary cost proxy. The coefficient on the industry Herfindahl index is insignificant in the REG FD adoption regression.
17
cross-listed firms rarely disclose industry segment data, we assume that all of the cross-listed firm’s sales are derived from its primary 4-digit SIC code industry. Second, if a cross-listed firm derives less than 50% of its sales from the home country, we assume that the cross-listed firm’s primary product market competition resides in the U.S. and therefore we use the U.S. 4-digit SIC code industry segment data. 17 A priori the effect of proprietary costs on REG FD adoption is not clear cut. However, prior research (see, e.g., Harris 1998) finds that firms that operate in less competitive product markets are less willing to provide public disclosures that may hurt their competitive advantages. Thus, we predict the coefficients on SUBSTITUTION and ENTRYCOST to be negative and the coefficient on MKTSIZE to be positive. Financial disclosure complexity: Existing research suggests that small (or naïve) investors are not as sophisticated as large (or institutional) investors in processing financial information (see, e.g., Battalio and Mendenhall 2005; Bhattacharya 2001; Bhattacharya et al. 2007; Walther 1997; Malmendier and Shanthikumar 2007; Mikhail et al. 2007). In a study of the determinants of open versus closed conference calls, Bushee et al. (2003) find that firms that have more complex financial information are less likely to choose open conference calls. Their result is consistent with the idea that company management has a fear that the public disclosure of complex financial information to all investors will increase the likelihood of naïve investors’ misinterpretation of the complex disclosure and lead to unnecessary stock return volatility. Following Bushee et al. (2003), we use three proxies to capture the complexity of a firm’s financial disclosure: HIGH_TECH, STD_REV, and INTANGIBLE. HIGH_TECH is a dummy that is equal to 1 for firms in a high-tech industry (SIC codes: 2833-36, 3612-13, 362117
Results are similar if we use only U.S. industry segment data to measure the three proprietary cost proxies for all cross-listed firms.
18
29, 3651-52, 3661-69, 3671-2, 3674, 3695, 4812-22, 4832-99, 7370-79). STD_REV is the standard deviation of net sales scaled by the mean sales over the prior 3 years. INTANGIBLE is the percentage of total assets that are intangible. STD_REV is winsorized at the 1% and 99% percentiles while INTANGIBLE is winsorized at the 99% percentile. We expect the coefficients on HIGH_TECH, STD_REV, and INTANGIBLE to be negative. Although we attempt to use different variables for different theoretical constructs, it is likely that some of the variables could be proxies for multiple constructs. For example, INSTITUTIONOWN, a proxy for investor demand, could also be a proxy for information asymmetry. Likewise, the variables for financial disclosure complexity could also capture proprietary costs. As a result, readers should interpret the regression coefficients with this caveat in mind.
Descriptive Statistics and Primary Regression Results Table 1 provides descriptive statistics on the determinants of REG FD adoption for the adopters in column (1) and the non-adopters in column (2). Consistent with our expectations, relative
to
the
adopters,
the
non-adopters
have
higher
institutional
ownership
(INSTITUTIONOWN), lower foreign sales (FOREIGNSALE), lower stock return synchronicity (SYNCHRONICITY), lower investment opportunities (INVEST_OPP), lower demand for external equity finance (EXTERNAL_FIN), and greater divergence of the voting rights and cash flow rights by top management (CONTROL_WEDGE). The other economic determinants do not significantly differ across the two groups. Table 2 reports the logistic regression results on the determinants of the REG FD adoption decision based on the 178 active cross-listed firms that responded to our survey. The
19
two regression models in Table 2 are identical except that the proxy for the information spillover effect is FOREIGNSALE in column (1) and SYNCHRONICITY in column (2). We discuss the regression result in column (1) first. Consistent with the investor demand effect, the coefficient on INSTITUTIONOWN is significantly negative. The significantly positive coefficient on FOREIGNSALE is consistent with the information spillover effect, suggesting that when there is a positive correlation between U.S. firms’ and cross-listed firms’ values and cash flows, crosslisted firms are more likely to pool with U.S. firms by voluntarily providing open disclosure. Consistent with the existing theories of voluntary disclosure (see Healy and Palepu 2001), crosslisted firms are more likely to adopt REG FD when the investment opportunity set is higher, the demand for external equity financing is higher, and the divergence of management’s voting rights and cash flow rights is lower. We find little evidence that proprietary costs (SUBSTITUTION, LN(MKTSIZE), and LN(ENTRYCOST)) and information asymmetry (LN(ASSETS) and LN(ANALYST)) affect cross-listed firms’ REG FD adoption. There is mixed evidence that financial disclosure complexity affects REG FD adoption. Only one of the three proxies for disclosure complexity (i.e., STD_REV) is significant and consistent with our prediction. Overall these results suggest that U.S. firms’ adoption of REG FD creates pressure for cross-listed firms to voluntarily follow REG FD, but cross-listed firms whose management is expected to benefit less from open disclosure are less likely to adopt REG FD. The sample size for the regression in column (2) is smaller due to missing values for SYNCHRONOCITY. With the exception of the significant coefficients on LN(MKTSIZE) and LN(ENTRYCOST) in column (2), the inferences from the regression in column (2) are qualitatively similar to those from the regression in column (1). Consistent with the coefficient on FOREIGNSALE, the coefficient on SYNCHRONICITY is significantly positive. This finding
20
is not surprising because the Pearson (Spearman) correlation between the two information spillover effect proxies is a significant 0.31 (0.34).
Robustness Checks The regression results in Table 2 are robust to a series of untabulated sensitivity checks. First, we exclude the cross-listed firms from UK and Australia because we learned in the process of conducting the survey that around the passage of REG FD both the UK and Australian regulators issued rules similar to REG FD though it is unclear whether those rules are strictly enforced. Second, we include country fixed effects to control for potential omitted country-level variables such as country investor protection. With the exception of the insignificant coefficients on STD_REV and INVEST_OPP, none of the other significant coefficients are affected. Third, we include in the regression a series of dummy variables indicating (1) whether a cross-listed firm is listed on one of the three major U.S. stock exchanges; (2) whether a cross-listed firm is listed on its home country exchange; and (3) whether a cross-listed firm is listed in any foreign stock exchange other than the U.S. Finally, we retain only the cross-listed firms that were listed on a U.S. stock exchange prior to the effective date of REG FD. Eliminating the foreign firms cross-listed in the U.S. after the REG FD effective date helps reduce the concern that our results could be partially due to alternative economic forces such as the mandatory adoption of IFRS across Europe in 2005, which tends to increase disclosure transparency.
IV. THE EFFECT OF VOLUNTARY REG FD ADOPTION ON THE INFORMATION ASYMMETRY COMPONENT OF COST OF CAPITAL Predictions
21
Existing disclosure theories suggest that a credible commitment to increased disclosure transparency should result in a reduction in a firm’s information asymmetry component of cost of capital (see Verrecchia 2001). Hence, a natural question is whether cross-listed firms’ voluntary commitment to REG FD helps reduce the information asymmetry component of cost of capital. 18 Ex ante it is difficult to predict the effect of voluntary REG FD adoption on the adopters’ information asymmetry component of cost of capital. If cross-listed firms’ stated REG FD adoption represents a credible commitment to increased disclosure transparency, we should expect the cross-listed firms who voluntarily followed REG FD to enjoy a significant reduction in the information asymmetry component of cost of capital. Table 2 shows that cross-listed firms’ REG FD adoption can be explained by economic incentives, suggesting that the voluntary adoption should be credible. However, there are also reasons why cross-listed firms’ voluntary REG FD adoption may not be credible. First, the REG FD adoption is voluntary and an adopting firm can revert to a selective disclosure policy at any time without the fear of being prosecuted by the SEC. However, cross-listed firms that are suspected of breaking their REG FD commitment would bear other significant costs, such as loss of management’s reputation and credibility. A few SEC enforcement cases against U.S. firms’ REG FD violations (see Fenwick and West 2005) suggest that detecting REG FD violations is not that difficult because the selective disclosure of material non-public information would inevitably result in significant changes in stock prices, which is observable to all investors.
18
Though untabulated, we also conducted an event study of cross-listed firms’ stock market reactions to the events that led to the final passage of REG FD. We find no evidence of either a positive or negative stock market reaction to the events, suggesting that REG FD on average did not cause any change in shareholder valuation. However, we would like to caution that the lack of significant results could reflect the difficulty of identifying the events relevant to REG FD.
22
Second, even if cross-listed firms’ stated REG FD adoption is credible, the adoption itself may not represent a significant increase in the voluntary disclosure transparency. The cross-listed firms who claimed to have voluntarily followed REG FD could merely reconfirm their existing voluntary open disclosure policy. However, REG FD adopters’ ex post open disclosures shown in Section V suggest that this is not a significant concern. Finally, REG FD may create a chilling effect among managers of both U.S. firms and cross-listed firms who are pressured to follow REG FD. The primary objective of REG FD is to level the playing field among different investors with regard to information access to company management (i.e., increase management’s voluntary disclosure transparency) and does not intend to force management to increase or decrease the total amount of voluntary (open and closed) disclosures to all investors. However, in reality REG FD may also affect the total amount of voluntary disclosures available to all investors. In particular, some commentators in the debate leading to the passage of REG FD argued that REG FD would create a chilling effect among corporate managers and discourage them from disclosing the private information that they would otherwise disclose to select investors. If this were true, REG FD could increase rather than decrease the information asymmetry among investors, especially between insiders and outside investors. Thus, to the extent that a significant chilling effect exists, the previously discussed investor demand and information spillover effects on cross-listed firms would be smaller and therefore we will be less likely to find a negative effect of the voluntary REG FD adoption on the information asymmetry component of cost of capital. The non-adopters’ change in the information asymmetry component of cost of capital around the passage of REG FD is hard to predict due to two offsetting effects. On one hand, due to the investor demand effect resulting from U.S. firms’ adoption of REG FD, the non-adopters
23
should experience a reduction in investors’ interest and therefore a decrease in stock liquidity in less than perfectly competitive financial markets (Fishman and Hagerty 1989). On the other hand, due to the information spillover resulting from U.S. firms’ adoption of REG FD, more public information about non-adopters’ values and cash flows could become available, resulting in an increase in the non-adopters’ stock liquidity and a decrease in their cost of capital following REG FD (see Lambert et al. 2007).
Research Design Following Leuz and Verrecchia (2000), we use two approaches to test the effect of adopting REG FD on the information asymmetry component of cost of capital. The first approach compares the information asymmetry component of cost of capital in the post-REG FD adoption period for the adopters versus non-adopters (the level regression). The second approach compares the change in the information asymmetry component of cost of capital before versus after the passage of REG FD for the adopters versus non-adopters (the change regression). We use the non-adopters (i.e., a difference-in-difference design) to control for confounding factors that affect all cross-listed firms during our sample period. We adopt both the level and change approaches because both have pros and cons. The level regression has a larger sample because we can include the firms cross listed after REG FD’s effective date. The disadvantage of the level regression is that it is more prone to correlated omitted variables with the REG FD adoption decision. To reduce the correlated omitted variable concern for the level regression, we add several control variables (discussed below) and also use the Heckman’s (1978) approach to correct for the self selection bias of the REG FD adoption decision.
24
The advantage of the change regression is that we can use the firm itself as a control for the many unobservable confounding firm fixed effects (e.g., firm characteristics). As REG FD is an exogenous shock to cross-listed firms, the change in the information asymmetry component of cost of capital in our short event window (defined below) before versus after the passage of REG FD for the adopters versus non-adopters is less likely caused by changes in the unobservable confounding factors. However, the change regression has a smaller sample than the level regression because it requires each firm to have data in both pre and post REG FD periods. In addition, the change specification cannot capture the REG FD induced cost of capital effect already reflected in the pre REG FD period. Consistent with Leuz and Verrecchia (2000), the information asymmetry component of a firm’s cost of capital in either the pre-REG FD period or the post-REG FD period is measured using the bid-ask spread, trading volume, and share price volatility. 19 To reduce the confounding effects associated with a firm’s gradual shift in voluntary disclosure, we follow Leuz and Verrecchia (2000) by using a 6-month window to compute the three proxies, but results are similar if we use a 3-month window. Leuz and Verrecchia (2000) argue that the bid-ask spread is the most explicit measure of a firm’s information asymmetry while the other two proxies can be influenced by many other factors unrelated to information, especially the share price volatility. Indeed, Leuz and Verrecchia (2000) find significant results for the bid-ask spread and trading volume but not for the share price volatility. The bid-ask spread (SPREAD) is defined as the average relative closing bid-ask spread from the daily CRSP (i.e., the absolute spread divided by the average of bid and ask) in a 6month window. Trading volume (TURNOVER) is defined as the median daily turnover ratio (i.e., 19
We do not examine the implied cost of capital due to data limitations and the noise of existing implied cost of capital measures (see Easton et. al. 2002; Easton and Sommers 2007).
25
the number of shares traded divided by the total shares outstanding from the daily CRSP) in a 6month window. 20 Share price volatility (VOLATILITY) is defined as the standard deviation of daily stock returns in a 6-month window. All three variables are winsorized at the 99% percentile. For the pre-REG FD period, the three cost of capital proxies are measured over the 6 months before the REG FD effective month. Note that the three proxies in the pre-REG FD period cannot be computed for the firms cross-listed after the REG FD effective date because these firms did not trade on a U.S. stock exchange. For the post-REG FD period, the three proxies are measured over the 6 months after the REG FD effective month if a firm is listed on a U.S. stock exchange prior to the end of the REG FD effective month and over the 6 months after the firm’s initial month of cross listing if a firm is listed on a U.S. stock exchange after the REG FD effective month. Recall that the REG FD adopters that were cross listed prior to the REG FD effective date all adopted REG FD on the REG FD effective date. Following Leuz and Verrecchia (2000), we include several control variables in the level regression to reduce the influence of confounding factors. For the bid-ask spread level regression, we include as controls TURNOVER, VOLATILITY, LN(MARKETCAP), and FREEFLOAT. LN(MARKETCAP) is defined as the natural logarithm of a firm’s total market capitalization (i.e., the sum of the market values of all issues for firms with multiple share issues). For the firms cross listed prior to the REG FD effective date, LN(MARKETCAP) is measured at the year end prior to the 6-month measurement window of the dependent variable. For the firms cross listed after the REG FD effective date, LN(MARKETCAP) is measured at the year end before the firm’s cross listing date. FREEFLOAT indicates the presence of insiders and is defined as the 20
To ensure that the trading volume change around REG FD is not caused by a difference in the frequency of earnings announcement events in the 6 months around REG FD, we also excluded the daily trading volumes in the five trading days around the quarterly earnings announcement dates in computing TURNOVER and obtained similar inference.
26
percentage of firm shares that are not closely held. FREEFLOAT is measured at the year-end prior to the 6-month measurement window of the dependent variable. For the trading volume level regression, we include as controls LN(MARKETCAP), VOLATILITY, and FREEFLOAT. For the share price volatility level regression, we include as controls LN(MARKETCAP), FREEFLOAT, and BETA. BETA is the market beta directly from the Worldscope database and is a firm-specific constant over the sample period. We do not include the three control variables (i.e., LN(MARKETCAP), FREEFLOAT, and BETA) in the change regression because FREEFLOAT and BETA are not likely to change significantly over the test period. In addition, we find no evidence from untabulated regression tests that the reduction in the bid-ask spread in Table 4 is caused by stock price appreciation (i.e., LN(MAREKETCAP)) in the post-REG FD period.
The Level Regression Results Primary Regression Results Table 3 reports the level regression results on the effect of REG FD adoption on the information asymmetry component of cost of capital measured by the bid-ask spread (SPREAD), share turnover (TURNOVER), and share price volatility (VOLATILITY) in the post REG FD period. ADOPT is a dummy that equlas one for the REG FD adopters. Following Heckman (1978), we use the model in column (1) of Table 2 to correct for the self selection bias of ADOPT. However, the untabulated OLS regression results are similar. We find that the adopters have a lower bid-ask spread and higher share turnover than non-adopters. However, similar to Leuz and Verrecchia (2000), we find no evidence that the adopters have lower share price volatility than non-adopters, suggesting that VOLATILITY may
27
not be a good proxy for a firm’s information asymmetry component of cost of capital. Many of the control variables are significant and consistent with expectations. For the SPREAD regression, larger firms with higher share turnover have lower bid-ask spread. For the TURNOVER regression, larger firms with higher return volatility have higher share turnover. For the VOLATILITY regression, the only significant control variable is BETA. Overall the coefficients on the control variables are similar to those reported in Leuz and Verrecchia (2000). The only noticeable exception is that the coefficient on FREEFLOAT is never significant in our sample.
Robustness Checks The regression results in Table 3 are robust to a battery of untabulated sensitivity checks. First, to recognize the fact that ADOPT has a joint effect on both SPREAD and TURNOVER regression, we jointly estimate a three-equation system including the REG FD adoption regression, the bid-ask spread regression and the share turnover regression. Because this system is only identified via the functional form, we also eliminate TURNOVER in the SPREAD regression and estimate a reduced form. Second, we control for the number of analysts following in Table 3 in order to rule out the possibility that the effect of ADOPT is due to a difference in analyst following across the adopters and non-adopters. Third, we rerun the regressions in Table 3 after excluding cross-listed firms from the UK and Australia that introduced similar regulations around the passage of REG FD. Fourth, we include a series of dummy variables indicating whether a cross-listed firm is listed on one of the three major U.S. stock exchanges, whether a cross-listed firm is listed on its home country exchange, and whether a cross-listed firm is listed in any foreign stock exchange other than the U.S.
28
Results on Changes of Liquidity Primary Results Table 4 reports the changes of SPREAD, TURNOVER, and VOLATILITY from the 6 months before to the 6 months after the REG FD adoption for the cross-listed firms that were listed on the U.S. stock exchanges in the pre REG FD period. The sample size here is smaller than that in Table 1 because we require each firm to have data in both the pre and post REG FD periods. For the non-adopters, the changes in the three proxies for the information asymmetry component of cost of capital are insignificant except for the significant reduction in share turnover, suggesting that the non-adopters’ share liquidity decreases in the post-REG FD period. This finding suggests that following the passage of REG FD, i.e., investors were less willing to trade in the shares of non-adopters. For the adopters, however, all three measures of the information asymmetry component of cost of capital change significantly in the predicted directions. In addition, a test of the differences in differences in the last column of Table 4 for the adopters versus non-adopters in the two time periods indicates that the changes in the three proxies of the information asymmetry component of cost of capital are significantly different for adopters and non-adopters. Overall, the results in Table 4 are consistent with those from Table 3 and suggest that the adoption of REG FD is a credible commitment to increased disclosure transparency that helps significantly reduce adopting firms’ information asymmetry component of cost of capital.
Robustness checks
29
Although we have used the non-adopters to control for common economic shocks that affect all cross-listed firms, the differences in results shown in Table 4 between the non-adopters and adopters could be caused by some unknown changes other than REG FD that affect the adopters but not the non-adopters. For example, it is possible that even in the absence of REG FD, we may still observe the results in Table 4 due to the adopters’ increased commitment to disclosure transparency over time. To rule out this alternative explanation, we rerun the analyses of ∆LN(SPREAD), ∆LN(TURNOVER), and ∆LN(VOLATILITY) in Table 4 for the nonadopters and adopters in two alternative 12-month time periods: a) the 12-month period prior to the REG FD effective date; and b) the 12-month period subsequent to the REG FD effective date (see Table 5). To the extent that the results in Table 4 are due to the adopters’ gradual increase in commitment to disclosure transparency over time rather than REG FD, the results in Table 5 ought to be similar to those in Table 4. We fail to find support for this alternative explanation in Table 5, suggesting that the results in Table 4 are likely driven by REG FD.
V. CROSS-LISTED FIRMS’ EX POST VOLUNTARY DISCLOSURE: THE CASE OF OPEN CONFERENCE CALLS In this Section we examine whether cross-listed firms who claimed to have adopted REG FD actually followed through on their initial commitment by providing voluntary open disclosure in the post-REG FD period. Following Bushee et al. (2003), we use open conference calls as an ex post validation of a firm’s ex ante open disclosure commitment. 21 We believe it is reasonable to use open conference calls as a proxy for a firm’s ex post voluntary open disclosure
21
6-K filings could be an alternative measure of cross-listed firms’ extent of voluntary open disclosure. We decided not to adopt this measure for two major reasons. First, not all 6-K filings are voluntary. Second, 6-K is not the only method of disclosure endorsed by REG FD.
30
because over the past decade conference calls have become a popular medium publicly traded firms use to communicate non-public information to outside investors. We hand collected the open versus closed conference call data from Thomson Financial’s StreetEvents database over the years 1999-2006. 1999 is the first year that the StreetEvents database has a reasonable coverage of conference calls. For both the pre- and post-REG FD periods separately, OPEN is coded one if a firm always held open conference calls, and zero if the firm held no conference call at all or at least one closed call. 22 A firm is deemed to hold a closed conference call if the conference call either provided no call/web information for investors to listen in or restricted the call to analysts only. Panel A of Table 6 shows the descriptive statistics of OPEN in the pre- and post-REG FD periods for only the cross-listed firms that were listed in the U.S. stock exchanges prior to REG FD’s effective date. Panel B of Table 6 shows the descriptive statistics of OPEN in the post REG FD period for the combined sample of firms that were cross listed pre REG FD and the firms that were cross listed post REG FD. 23 Focusing on the REG FD adopters in Panel A, we find that for the pre-REG FD period only 25% of the adopters followed an open call policy. In the post-REG FD period 75% of the adopters followed an open call policy, which is significantly higher than the percentage in the pre-REG FD period (p value