«Large Blocks of Stock: Prevalence, Size and Measurement Jennifer Dlugosz Rudiger Fahlenbrach Paul Gompers Andrew Metrick 22-04 Large Blocks of Stock: ...»
The Rodney L. White Center for Financial Research
Large Blocks of Stock: Prevalence, Size and Measurement
Large Blocks of Stock: Prevalence, Size, and Measurement
Department of Economics and Harvard Business School
Department of Finance, Fisher College of Business
Ohio State University
Harvard Business School
Harvard University and NBER
Andrew Metrick Department of Finance, Wharton University of Pennsylvania and NBER July 2004 We thank Sith Chaisurote, Ashton Hawk, Allison Lowenstein, Jason O’Connor, Colleen Pontious, and Rebecca Yang for excellent research assistance. Gompers and Metrick acknowledge support from NSF grant #SES-0136791. Gompers acknowledges support from the Division of Research, Harvard Business School. The data set built in this paper can be downloaded from http://finance.wharton.upenn.edu/~metrick/data.htm.
Abstract Large blocks of stock play an important role in many studies of corporate governance and finance. Despite this important role, there is no standardized data set for these blocks, and the best available data source, Compact Disclosure, has many mistakes and biases. In this paper, we document these mistakes and show how to fix them. The mistakes and bias tend to increase with the level of reported blockholdings: in firms where Compact Disclosure reports that aggregate blockholdings are greater than 50 percent, these aggregate holdings are incorrect more than half the time and average holdings for these incorrect firms are overstated by almost 30 percentage points. We also demonstrate that our fixes are economically and statistically significant in an analysis of the relationship between firm value and outside blockholders.
2 I. Introduction Large-block shareholders play an important role in corporate governance. For this reason, the presence of such “blockholders” and the size of their holdings is a common explanatory variable in financial research. In just the last few years, a representative sample of such studies includes analyses of the role of blockholders in executive turnover, executive compensation, firm diversification, discretionary expenses, market liquidity, and corporate performance.1 Furthermore, blockholder data is a crucial input in the analysis of the relationship between ownership structure and firm value, where seminal works by Demsetz and Lehn (1985) and Morck et al. (1988) gave rise to a vast and growing literature.
Despite the common use of large shareholder data, there does not exist a clean offthe-shelf database to facilitate research. Many of the papers cited above required their authors to gather their own data. This time-consuming task is necessary because of several weaknesses in the available databases. Of course, decentralized data gathering causes duplication of effort and lack of standardization across projects. Also, because of the large time commitment necessary to clean the data for each firm, most researchers have gathered data for a relatively small number of firms. This paper aims to fill this data gap by documenting the problems with the currently available data, proposing a consistent set of solutions to these problems, and making a “clean” database freely available to all researchers.2 Furthermore, we demonstrate the superiority of clean (vs. raw) data with a representative study on the relationship between outside blockholders and firm value.
1 For examples of papers on these listed topics, see Denis et. al (1997), Ryan and Wiggins (2001), Anderson et.
al (2000), Ang et. al (2000), Singh and Davidson (2003), Heflin and Shaw (2000), Cremers and Nair (2004), and Shivdasani (1993).
2 The database can be downloaded from http://finance.wharton.upenn.edu/~metrick/data.htm.
requirements for public corporations in Regulation 14A and Schedule 14A. Virtually everything we know about blockholders in the United States comes from these disclosure requirements, which are described in detail in Appendix A of this paper. The two main types of data produced by the SEA are for holdings (once per year, reported in the annual proxy statement), and for transactions by corporate insiders and beneficial owners (updated through Forms 3, 4 and 5). While the trading data would appear to provide the most current and comprehensive information, past research has demonstrated that this data is difficult to work with and cannot be relied upon to infer the holdings of individual blockholders (Anderson and Lee (1997a and 1997b), Jeng et al. (2003)). Thus, we focus in this paper on the annual proxy data, which is more reliable and more commonly used by researchers.
Proxy data is available from many sources, including direct electronic access using the SEC’s “Edgar” tool for all corporate filings since the mid-1990s. For large-scale data downloads, however, it is necessary to use a commercial product. The most widely used product is the Compact Disclosure database of Standard & Poor’s. Anderson and Lee (1997a and 1997b) focus their analysis on the holdings of corporate officers and directors, and show that Compact Disclosure accurately reproduces the information in proxy statements for all firms except those with multiple classes of stock. While Compact Disclosure also reproduces data on blockholders from the tables in the proxy statement, there are additional problems with these data. We discuss these problems and their solutions in Section II, and summarize the changes for a large sample of firms from 1996 to 2001.
In Section III, we perform a representative study using both raw Compact Disclosure data and a “clean” data set where these problems have been fixed. We find that the raw
and other control variables, the clean data set is far more likely to yield statistically significant point estimates for the ownership variables. Section IV summarizes our conclusions. Two appendices supplement the text. Appendix A provides details on the 1934 SEA and the disclosure requirements it created, and Appendix B provides details on the construction of our sample.
Our initial sample of firms consists of firms that are covered by the Investor Responsibility Research Center (IRRC) for both their publication Corporate Takeover Defenses (Rosenbaum 1995, 1998, 2000) and their director’s database which provides details on the board of directors for about 1,500 of the largest U.S. companies. The IRRC’s universe is drawn from the Standard & Poor’s (S&P) 500 as well as the annual lists of the largest corporations in the publications of Fortune, Forbes, and Businessweek. We use the IRRC sample as a starting point because a wide range of governance data is available for this group of companies and our goal is to make this set of data as comprehensive as possible for this group.3 A special subset of the IRRC companies – less than 10 percent in all years – have multiple classes of common stock. For these companies, Anderson and Lee (1997a) showed there are many problems with the Compact Disclosure data, and these problems are very difficult to fix. In this paper, we eliminate all multiple-class companies from the database and 3 The IRRC data has been used as a starting point by Gompers et al. (2003), Cremers and Nair (2004), and Gillan et al. (2003).
companies in the IRRC sample from 1996 to 2001.4 The initial ownership data comes from the Compact Disclosure CD-ROMs. Based on the results of Anderson and Lee (1997a, 1997b) we build our sample from the information on large shareholders that Compact Disclosure derives directly from the proxies and ignore the insider-trading data that is also available through Compact Disclosure. Appendix B provides details on the construction of the initial database.
We next check the initial database by comparing the Compact Disclosure data to the original proxy statements, which we obtain from Livedgar,5 making changes to the ownership percentages of large shareholders where appropriate. All firms in the sample were checked – even those with no reported blockholders in Compact Disclosure. We employ the following general rules when deciding on share ownership. The SEC defines beneficial ownership as either voting or investment power, and sometimes companies report both measures in their proxies. We use voting power as opposed to investment power for our database when a distinction is made between the two. Also, even if individuals disclaim beneficial ownership of some portion of their holdings in the proxy, we treat these holdings as if the individual had the voting power. Under the terms of SEC Rule 13d-3, shares of common stock that may be acquired within 60 days are deemed outstanding for the purposes of computing the percentage of common stock owned by a shareholder. We follow this SEC rule and include these options. In the rare cases of a company having a temporary ownership structure resulting from a recent merger or acquisition, we remove these companies from our 4 The dual-class companies are analyzed in a companion paper, Gompers et al. (2004), where we attempt to build a comprehensive sample of all dual-class companies with any share-class trading on any major exchange in the United States.
5 Livedgar is an online data service, provided by Global Securities Information, Inc., that enables users to obtain source documents as filed with the SEC.
and these firm-years were removed from our sample. Our final sample consists thus of 7,649 firm-years and covers 1,913 unique firms. Table 1 shows summary statistics of our sample firms. The table is based on cross-sectional averages of time-series means.
Many researchers are interested in knowing whether a specific blockholder is an “insider” or an “outsider” to the firm. The role of large shareholders in corporate governance is often treated differently depending on the classification of the shareholder. Since our work required the examination of all blockholders, the marginal cost of coding these classifications was relatively low, so we did so. The results are summarized in Table 2. The possible classifications are (1) officer, (2) director, (3) affiliated entity, (4) ESOP, and (5) outside blockholder. Category (1) includes all officers, even if they are also directors. Category (2) only includes non-officer directors. Category (3) includes any individual, trust, or company whose voting outcome is partially influenced, but not completely controlled, by an officer or director of the company. If the shares are completely controlled by the officer or director, then these shares would be counted under category (1) or (2), respectively. Category (4) is the aggregate number shares held by Employee Share Ownership Plans, but does not include employee shares held through non-ESOP retirement plans (such as non-ESOP 401(k) plans).
Category (5) includes all blockholders not elsewhere classified. This final category makes up about two-thirds of the aggregate amount of blockholding, and will be examined in the analysis of Section III.
Two main biases are introduced if researchers were to work directly with the benchmark Compact Disclosure database: overlaps and preferred shares. The SEC requires that all beneficial owners of more than 5% of a company’s common stock be listed in the proxy, and consequently shares are often double or triple counted under different people or entities.6 While the SEC requires firms to detail the ownership structure of jointly held block in the footnotes, Compact Disclosure ignores all of the footnotes detailing joint or cross ownership of shares and lists every blockholder and ownership percentage exactly as it appears in the summary table of the proxy section “Security Ownership of Management and Certain Beneficial Owners.” This leads to the overlap of reported ownership, which might be either a full overlap or a partial overlap. Examples of these two cases are documented below in Subsection 1. Second, Compact Disclosure sometimes misrepresents preferred shares as common equity ownership. This problem is illustrated below in Subsection 2.
Full overlaps can arise in two types of situations. In the first scenario, two or more blockholders are listed in the ownership table with the same shareholdings and the joint ownership of these shares is disclosed only by the footnotes. In the second scenario, the proxy separates the beneficial ownership of directors and officers from that of large shareholders and Compact Disclosure reproduces entries from both tables without crosschecking identities. Figures 1 and 2 display an example of the latter case. Figure 1 shows the Compact Disclosure data for Coca Cola Co. from the October 1999 CD, and Figure 2 shows an excerpt of the proxy statement from March 4th, 1999 on which the data is based. While 6 See Appendix A for details of these disclosure requirements.
tables in the proxy, the vital information of the proxy footnote is ignored. Figure 1 lists Berkshire Hathaway and Warren Buffett individually as 8.10% owners of the common stock.
Referring to the footnote 4 of Figure 1 (the ownership table from the proxy statement), we find that all of the shares listed under Warren Buffett are owned indirectly through Berkshire Hathaway. Tallying the beneficial ownership percentages without referencing the table footnotes in the proxy would suggest that 22.3% of Coca Cola’s common stock is held by blockholders when actually 14.2% is the correct figure.