Firm-Specific Industries∗
Stefan Lewellen† Yale University
March 26, 2012
Abstract Standard industry classifications are “fixed,” in the sense that industry definitions are required to be transitive and disjoint. This paper introduces firm-specific industries (FSIs) that relax the transitivity constraints inherent in traditional classifications. Using hand-collected data from firms’ public SEC filings, I show that FSIs provide a material improvement over standard industry classifications at explaining the variation in corporate capital structure and stock returns. As such, FSIs can potentially be used as a building block towards addressing numerous open questions within finance, industrial organization, marketing, and other fields.
JEL: C1, C52, G11, G12, G14, G34 ∗
I thank Andrew Metrick for helpful comments. I also thank Lora Chow, Byron Edwards, Tate Harshbarger, Yoonie Hoh, Siddharth Jain, Connie Liu, Candice Manatsa, Aayush Upadhyay, Tim Wang, Roger Yang and Doris Zhao for excellent research assistance. † Ph.D. Candidate, Financial Economics. Email:
[email protected]. Phone: +1 (203) 535-6562. Address: 135 Prospect Street, P.O. Box 208200, New Haven, CT, USA, 06520-8200.
1
1
Introduction
Industries play a central role in economic theory. From product market competition and production inputs to vertical and horizontal integration, numerous papers have shown that firms’ corporate and financial decision-making processes are endogenously related to the health and structure of their industry.1 Given the strong theoretical link between industry dynamics and corporate decisions, industry controls and other industry-related measures are a common feature of many empirical tests within finance, economics, marketing, and other fields. However, accurately mapping the product market space within a large, integrated economy such as the United States economy is not a simple task, and standard industry classifications exacerbate the problem by employing rigid definitions that allow for little flexibility in grouping together firms that possess similar economic characteristics.2 In this paper, I attempt to design a more accurate map of product market competition within the U.S. economy by relaxing the rigid constraints of traditional classifications. Using hand-collected data from over 35,000 annual 10-K filings, I design firm-specific industries, or FSIs, that allow each firm to be matched with a unique industry that best represents its product market outputs. Unlike traditional industry definitions, FSIs are neither transitive nor disjoint, meaning that while firm A may reside in firm B’s industry, the converse need not be true. The goal of this paper is simply to introduce FSIs and showcase a number of potential applications of FSIs. Specifically, I show that FSIs explain a significantly greater fraction of the variation in stock returns and leverage than traditional industry classifications. However, FSIs can potentially be used to address a number of important questions in finance and other fields. For example, Lewellen (2012b) uses FSIs to address an important open question in the CEO compensation literature. To construct FSIs, I exploit a federal law (the Securities Act of 1933) requiring firms to provide an unbiased list of their primary product market competitors in their 10-K filings. I collect these lists from more than 36,000 10-K filings from 2002 to 2008, totalling more 1
Examples include Brander and Lewis (1986), Maksimovic (1988), Dixit (1989), Bolton and Scharfstein (1990), Maksimovic and Zechner (1991), Hopenhayn (1992), Williams (1995), Miao (2005) and Spiegel and Tookes (forthcoming). 2 Most modern industry classification standards are based on product market outputs. However, the Standard Industrial Classification (SIC) system developed by the U.S. government uses a Byzantine hodgepodge of product market outputs, production inputs, and other factors to assign firms to industry classifications.
2
than 400,000 competitor names. I then match competitor names by hand to the merged CRSP/Compustat database. While most firms are legally required to report their primary product market competitors, a sizable minority of firms do not report specific competitor names in their 10-K reports. Since not all firms report competitors, I examine a number ways to construct “industries” for these firms which are are described in more detail within the text. When all is said and done, however, I am able to assign FSIs to firms comprising more than 90% of the total market capitalization in the merged CRSP/Compustat database. A simple comparison of Coca-Cola and PepsiCo provides a glimpse of the potential improvement offered by FSIs over traditional industry definitions. Coca-Cola and PepsiCo clearly compete with each other in the beverage market, but unlike Coca-Cola, PepsiCo earns a substantial fraction of its sales and earnings from its salty snacks division. Traditional industry classifications assign Coca-Cola and PepsiCo to the beverage industry, which means that PepsiCo is typically matched with firms that only represent about half of its actual earnings. In contrast, FSIs allow PepsiCo to matched with both Coca-Cola and salty snack manufacturers, while Coca-Cola can be matched with PepsiCo without also being matched with salty snack manufacturers. Like traditional classification standards, FSIs also allow industry definitions to change over time. When I compare FSIs to “traditional” industry classifications such as SIC codes, I find that FSIs explain a materially higher fraction of the variation in stock returns and leverage than traditional industry classifications. In panel data regressions from July 2003 to June 2010, I find that FSIs explain approximately 15% more variation in stock returns than Standard Industrial Classification (SIC) industry definitions, which are by far the most widely-used industry definitions in practice (Bhojraj, Lee, and Oler (2003)). FSIs offer an even better improvement (33% to 45%) over traditional industry classifications at explaining variation within firms’ capital structures. I also examine FSI definitions to ensure that firms are not strategically selecting the competitors that they report in their SEC filings. For example, if executive compensation is indexed to industry performance, managers may have an incentive to report peers with low recent stock returns or poor operating performance in their list of competitors. Furthermore, managers typically do not have to report peers at all; under most conditions, the 3
listing of competitors is a voluntary exercise. As such, agency-related considerations may impact managers’ decisions to report a list of competitors and/or impact the competitors that are named, in which case FSIs may do a poor job of representing “actual” industries. Thankfully, I find little evidence that managers are either strategically choosing to report their competitors or are strategically listing such competitors. However, I also find that the Gompers, Ishii, and Metrick (2003) governance index (the G-index) is negatively related to firms’ decision to report competitors. This result suggests that while agency-related considerations may not drive managers to mis-report (or fail to report) their competitors, firms with better corporate governance may have better accounting transparency in their financial statements, complementing the findings of Ferriera and Laux (2007). Many other papers have examined the role of industries within asset pricing and corporate finance. In the asset pricing literature, Moskowitz and Grinblatt (1999) argue that industries explain the momentum effect, while Grundy and Martin (2001) find the opposite result. Novy-Marx (2009, 2011) argues that industries can explain the value premium. Hou and Robinson (2006) find that firms in concentrated industries earn low excess returns, while Giroud and Mueller (2011) find that firms in concentrated industries with poor corporate governance earn lower returns than firms with strong corporate governance. On the corporate finance side, MacKay and Phillips (2005) find that most of the variation in capital structure occurs within firms and industries rather than between industries (see Graham and Leary (2011) for a survey of this literature).3 In a related paper, Leary and Roberts (2011) find that firms often change their capital structures to mimic changes in their peers’ capital structures. Most closely related to this paper are the works of Hoberg and Phillips (2010a,b) and Rauh and Sufi (2012). Hoberg and Phillips (2010a,b) construct textbased network industry classifications (TNICs) from firms’ business descriptions in their 10-K filings. Like FSIs, TNICs are firm-specific in nature and can change over time. However, I find that FSIs do a far better job than TNICs at explaining variation in stock returns (leverage tests involving TNICs are forthcoming). Rauh and Sufi (2012) use data from CapitalIQ that broadly mimics FSIs, but the CapitalIQ data only contains a single cross3
For brevity, I omit a full citation of earlier empirical papers that focus on industries or product market competition such as Chevalier (1995a,b) and Phillips (1995).
4
section. Hence, unlike FSIs, the CapitalIQ data cannot vary over time. Thus, FSIs not only overcome the problems with traditional industry classifications, but they also overcome problems with other studies that utilize firm-specific industry definitions. This paper is organized as follows. Section 2 provides an overview of the motivation behind FSIs along with some examples and a review of the relevant literature. Section 3 provides a description of the data and some basic summary statistics. Section 4 examines whether managers strategically choose to report competitors. Section 5 examines the ability of FSIs to explain stock returns, while Section 6 examines the ability of FSIs to explain corporate capital structures. Section 7 concludes.
2
Motivation and Related Literature
2.1
Motivation
Industry dynamics play an important role in finance, economics, marketing, and numerous other fields. However, most empirical examinations of industry dynamics make the implicit assumption that industries are “correctly” defined, in the sense that industry definitions do an accurate job of grouping together firms with similar economic characteristics. While early industry classification standards often focused on grouping firms together based on production inputs, most modern industry classifications focus on grouping firms based on product market outputs. How good of a job do existing industry classifications do at accurately mapping the product market space within the economy? This important question forms the motivation for the current analysis. Figure 1 displays the traditional definition of an industry classification. This definition applies to each of the widely-used industry classification standards in finance and economics, namely the Standard Industrial Classification (SIC) and North American Industry Classification Standard (NAICS) systems developed by the U.S. government, the Global Industry Classification Standard (GICS) system developed by MSCI Barra and Standard & Poor’s, and the 48-industry standard developed by Fama and French (1997). The key feature of Figure 1 is that traditional industry classification definitions are transitive in nature: if firm
5
A is in the same industry classification as firms B and C, then firms B and C must also be in the same classification. Traditional industry classification standards are also disjoint: if firm A is in firm B’s industry, then firm B must be in firm A’s industry as well. While transitive, disjoint industry definitions are convenient for research purposes, their extreme rigidity is often at odds with the actual nature of product market competition within the economy. For example, Coca-Cola and PepsiCo clearly compete with one another in the beverage arena. However, while Coca-Cola only sells beverages and beverage-related products (such as syrups), PepsiCo generates nearly half of its sales and operating profits from its salty snacks businesses (Frito-Lay North America, Quaker Foods North America, and Latin American Foods). Thus, unlike Coca-Cola, PepsiCo’s competitors should include snack manufacturers. However, Coca-Cola and PepsiCo’s SIC classification (“Bottled and Canned Soft Drinks and Carbonated Waters,” SIC code 2086) only contains large beverage manufacturers (Coca-Cola, PepsiCo, the Coke and Pepsi bottling groups, Dr. Pepper Snapple Group, Monster Beverage Corp, National Beverage Corp, and Heckmann Corp). PepsiCo’s salty snack competitors are listed in SIC code 2096 (“Potato Chips, Corn Chips, and Similar Snacks”), but PepsiCo is not included in the salty snack industry. In most cases, traditional industry classifications assign firms like PepsiCo to the industry that represents the firm’s largest source of revenue or profits, or alternatively assigns the firm to a hodge-podge industry containing all multi-division conglomerates. Unfortunately, both approaches mischaracterize PepsiCo’s actual product market competitors and hence lead to industries that may not reflect the actual product market map within the economy. Traditional industry definitions also have difficulty separating businesses that may appear to be related to each other but nonetheless have different economic exposures. For example, consider a West Texas restaurant chain and a New York City restaurant chain. While both firms are in the restaurant business, the importance of the oil industry to the West Texas economy suggests that the West Texas chain will likely be affected by the health of the oil industry, while the New York chain may be affected by the health of the financial services sector (this example is adopted from Chevalier (2004); see Stein (2003) for a further discussion). Traditional industry definitions would typically group our hypothetical West Texas and New York City restaurant chains into a broad “restaurant” category. However, neither 6
chain competes directly with one another, and while both chains are broadly exposed to common factors such as food prices, their local economic exposures are significantly different. Hence, separate “West Texas restaurants” and “New York City restaurants” industries would be ideal in this scenario. As a related example, suppose that GlobalCo and LocalCo both produce bubble gum, but LocalCo only sells its gum within a specific locality or region – say, Nebraska – while GlobalCo sells gum across the globe in a variety of local and regional markets. While LocalCo may face significant economic exposure to the health of the local market, GlobalCo may be able to largely diversify away the effects of an economic shock to a single local market. As such, LocalCo’s competitors may include both gum makers and other local businesses within its region, whereas GlobalCo’s economic exposure to local businesses in LocalCo’s area may be close to zero. Again, traditional industry classifications may fail to capture this sublety. To overcome the problems with traditional industry classifications, this paper introduces firm-specific industries, or FSIs. FSIs differ from traditional industry classifications in that industries are defined on a firm-by-firm basis based on the specific economic characteristics of the firm in question. Hence, FSIs are neither transitive nor disjoint. This yields two immediate benefits which are depicted graphically in Figure 2. First, a single firm can reside within multiple FSIs spanning many “traditional” industry classifications. FSIs can also be asymmetric: if firm A points to firm B as a competitor, firm B need not point to firm A as a competitor. As such, FSIs overcome the largest hurdles faced by existing classification standards at accurately mapping the product market space within the U.S. economy. To construct FSIs, I examine public firms’ regulatory filings with the Securities and Exchange Commission (SEC). As described in the following section, firms often provide a list of their direct competitors in their public filings. For example, in its 2008 10-K filing, Coca-Cola lists PepsiCo, Dr. Pepper Snapple Group, Nestle, Group Danone, Kraft Foods, and Unilver as its primary competitors. In contrast, PepsiCo reports competitors separately for its beverage and snack divisions. In the beverage division, PepsiCo reports Coca-Cola, Cadbury Schewppes, and Nestle as its primary competitors, while in the snack division, PepsiCo reports Kraft Foods, Proctor & Gamble, General Mills, Kellogg’s, Campbell Soup, ConAgra, and Snyder’s as its competitors. These self-reported product market competitors 7
nicely illustrate the benefits of FSIs: (i) PepsiCo is able to report snack manufacturers as competitors while staying in the same industry as Coca-Cola; (ii) Coca-Cola can stay in the same industry as PepsiCo without having snack manufacturers in its industry; and (iii) even within the beverage industry, Coca-Cola and PepsiCo can list different competitors, potentially relating to Coca-Cola’s focus (at the time) on bottled water versus PepsiCo’s focus on juice products through its Tropicana subsidiary. None of these outcomes would be possible using traditional industry classification standards. While FSIs should do a much better job of grouping together firms with similar product market characteristics than traditional industry classifications, FSIs are not without their drawbacks. In the case of Coca-Cola and PepsiCo, for example, the “Bottled and Canned Soft Drinks and Carbonated Waters” SIC classification includes Coca-Cola and PepsiCo’s bottling operations (which at the time were separate, publicly-traded companies) in addition to Coca-Cola and PepsiCo. Coke and Pepsi bottlers are clearly exposed to many of the same economic factors that drive the performance of Coca-Cola and PepsiCo, but Coke and Pepsi do not generally compete with their bottlers – indeed, they typically have long-term vertical supply-chain agreements that substantially tie their performance to that of their bottlers. As such, Coca-Cola and PepsiCo do not list their bottlers as competitor firms, even though the bottlers clearly share the same product market space and are subject to many of the same economic forces as the two “parent” companies.
2.2
Related Literature
The research most closely related to this paper is that of Hoberg and Phillips (2010a,b) and Rauh and Sufi (2012). Hoberg and Phillips (2010a,b) collect business descriptions from firms’ 10-K filings and use textual analysis to construct “text-based network industry classifications,” or TNICs, based on firms’ stated descriptions of their product markets. The authors find that TNICs do a much better job than SIC classifications at explaining a wide range of financial variables such as profitability and leverage. Like FSIs, TNICs are firm-specific in nature and are not transitive or disjoint. Unlike FSIs, however, TNICs are computed using a textual analysis algorithm that may introduce error into industry definitions. In contrast, FSIs are “straight from the horse’s mouth” and should thus contain far less noise than TNIC 8
classifications, a finding that I empirically verify in Section 5. In contrast to Hoberg and Phillips (2010a,b), Rauh and Sufi (2012) use firm-specific industry definitions from the data vendor CapitalIQ, which is a subsidiary of Standard & Poor’s. Like the FSIs examined in this paper, CapitalIQ collects competitor names from firms’ 10-K filings. However, CapitalIQ also supplements their 10-K data with additional competitors gleaned from prospectus filings, annual reports, and surveys of public companies conducted by CapitalIQ analysts. Thus, the CapitalIQ data contains more information than FSIs, which may or may not improve the ability of the CapitalIQ data to correctly ascertain the true product market relationships within the economy. Rauh and Sufi (2012) show that the combination of firm-specific industries and a more accurate definition of firms’ leverage leads to a 50% increase in the fraction of capital structure variation that can be explained by industry effects. Unlike FSIs, however, CapitalIQ’s data consists of a single cross-section which is updated annually to reflect the most recent year’s data. Thus, CapitalIQ data cannot be used for time-series tests without backfilling industry definitions. This is potentially problematic since a material fraction of firms change their competitor lists from year to year. Despite their shortcomings, both of the firm-specific industry classifications described above are vastly superior to traditional classifications, in which industry definitions are transitive and disjoint in nature. However, it is worthwhile to provide a brief overview of the research on traditional industry classifications. Most of the early research within the finance literature focused on the ability of SIC classifications to identify “correct” industries. Clarke (1989) finds that two-digit SIC codes do a better job than three- and four-digit SIC codes at identifying distinct economic markets. In contrast, Kahle and Walking (1996) find that four-digit SIC codes do a better job than two-digit SIC codes at detecting abnormal firm performance. Guenther and Rosman (1994) find that intra-industry monthly stock return correlations are larger using Compustat SIC codes than similar correlations based on CRSP SIC codes. However, given Compustat’s poor historical SIC coverage, Guenther and Rosman (1994) also note that CRSP SIC codes may be more useful than Compustat SIC codes at identifying distinct industries over long time periods. In recent years, the empirical literature has expanded to also examine the properties of 9
the FF48, GICS, and NAICS classification standards. Bhojraj, Lee, and Oler (2003) compare the SIC, NAICS, FF48, and GICS methodologies and find that GICS classifications do a better job of explaining stock returns and accounting variables than other classifications standards. Chan, Lakonishok, and Swaminathan (2007) compare GICS and FF48 classifications and again find that GICS classifications explain a larger fraction of within-industry stock return correlations than FF48 industry definitions. Weiner (2005) studies the performance of SIC, NAICS, FF48, and GICS in nearly a dozen different applications and finds strong support for GICS classifications relative to the other types of industry classifications. Finally, Lewellen (2012a) examines the performance of the SIC, NAICS, FF48 and GICS classifications during the 1990s in an asset pricing context and finds that GICS classifications generally have stronger size and power properties than competing industry classification standards in randomly-drawn samples. Lewellen (2012a) also notes that industry granularity matters in the context of stock return tests: as industries get smaller and smaller, their returns contain an increasingly significant amount of idiosyncratic noise that can reduce the statistical power of industry definitions to explain firm-level returns.
3
Data Description and Summary Statistics
3.1 3.1.1
Data Description Firm-Specific Industry Definitions
Firm-specific industry definitions are constructed using public SEC filings. The SEC requires all issuers of publicly-traded securities in the United States to file an annual report (form 10K) within approximately 90 days of each fiscal year-end.4 These annual reports provide an overview of the firm’s operations and financial statements covering the prior fiscal year. Importantly, firms are required to provide an overview of their competitive position within their industry and are expected to name their competitors under certain conditions. Specifically, Section 101 of Regulation S-K of the Securities Act of 1933 (17 C.F.R. §229.101) requires 4
Depending on location, size, and other factors, a firm may be required to file forms 10-KSB, 10-K405, 20-F, or 40-F in lieu of filing a “plain vanilla” 10-K. All references to “10-K” in the paper include these other types of forms.
10
firms to include the following information in their 10-K filings: “...competitive conditions in the business involved including, where material, the identity of the particular markets in which the registrant competes, an estimate of the number of competitors and the registrant’s competitive position, if known or reasonably available to the registrant. Separate consideration shall be given to the principal products or services or classes of products or services of the segment, if any. Generally, the names of competitors need not be disclosed. The registrant may include such names, unless in the particular case the effect of including the names would be misleading. Where, however, the registrant knows or has reason to know that one or a small number of competitors is dominant in the industry it shall be identified. The principal methods of competition (e.g., price, service, warranty or product performance) shall be identified, and positive and negative factors pertaining to the competitive position of the registrant, to the extent that they exist, shall be explained if known or reasonably available to the registrant [emphasis added].” Prior to 2002, only a small fraction of public companies chose to list their competitors in their SEC filings. However, the number of companies listing competitors sharply increased following the July 2002 passage of the Sarbanes-Oxley Act, which focused primarily on improving firms’ corporate governance and accounting practices. As such, the analysis in this paper is restricted to 10-Ks filed from 2002 onwards. Time considerations also forced me to stop collecting 10-K data for fiscal years beyond 2008. Thus, I began by generating a list of all firms in the merged CRSP/Compustat database each fiscal year from 2002-2008. This produced a total of 36,176 firm-year observations. For each firm-year observation, the 10-K form for that fiscal year was downloaded manually from the SEC’s EDGAR website. Each 10-K was then searched by hand for the phrase “compet,” and each search hit was manually examined to determine whether the firm referenced an actual competitor. If it was determined that the 10-K made reference to the name of a competing firm, that firm’s name was manually recorded. In firms with multiple segments, the same competitor was often listed multiple times, and each reference was recorded along with the name of the relevant operating segment as listed in the 10-K. Unfortunately, the material disconnect between Compustat’s segment names and the segment names reported in firms’ 10-K filings made it difficult to incorporate segment data in the definition of FSIs. Hence, segment data 11
was discarded and each competitor was only included once in a firm’s industry. This is akin to equally-weighting competitors across a firm, which likely introduces noise in the computation of industry statistics. As a result, the assignment procedure described above works against FSIs having increased explanatory power relative to other industry classification standards, which often assign firms to industries based on the largest or most profitable segment within the firm’s domain. After identifying all of a firm’s potential competitors, the names of said competitors were matched to the CRSP/Compustat database and the Compustat Global database by hand. A number of important assumptions that were made during this matching process. First, many firms’ 10-K filings listed the name of a subsidiary or operating division of a publiclytraded company as a competitor rather than listing the name of the publicly-traded (parent) company as a competitor. For example, many elevator manufacturers listed Otis Elevator as a competitor, though Otis is a fully-owned subsidiary of a publicly-traded company named United Technologies. In these cases, the parent company’s Compustat code (i.e. the GVKEY for United Technologies) was assigned to all firms that listed Otis Elevator as a competitor. In a number of other cases, however, a firm’s 10-K listed competing products without specifying the name of the company that produced such products. In these cases, the product names were discarded since a competing firm was not directly named. Finally, a small fraction of 10-Ks listed competitors in an indirect fashion – for example, many telecommunications providers listed “RBOCs” (Regional Bell Operating Companies) or “ILECs” (Incumbent Local Exchange Carriers) as competitors. While it would have been possible to identify such competitors on a case-by-case basis, these indirect references were also discarded in the name of conservatism. Hence, if anything, the FSIs in this paper are likely to understate the full list of competitors included in firms’ 10-K filings. While GVKEYs were collected for firms listed in Compustat’s foreign database, the present analysis is restricted to Compustat’s North American database, which contains accounting and financial information. Compustat data is then matched with stock returns and market capitalization data from CRSP. Other data sources are also used. Corporate governance data is sourced from Andrew Metrick’s website (http://faculty.som.yale.edu/andrewmetrick/) and the text-based network classifications designed by Hoberg and Phillips (2010a,b) are 12
sourced from the authors’ website (http://www.rhsmith.umd.edu/industrydata/). Of the 36,176 firm-year observations in the sample, 17,021 such 10-Ks, or approximately 47%, contained valid competitors that could be matched to Compustat (in total, approximately 60% of the 10-Ks in the sample contained one or more competitor names, but many of these 10-Ks only listed private or foreign companies as competitors).5 Importantly, this means that a material fraction of firms did not report a list of competitors. I consider possible ways of assigning competitors to these firms in the following section. For now, however, all of the tests involve only those firms that reported competitor lists in their public filings.
3.2
Summary Statistics
Table 1 reports summary statistics for my sample of firm-specific industries based on industry size. The average and median number of competitors listed in each 10-K and matched to a Compustat GVKEY are around 9 and 7, respectively. Hence, these industries are a lot smaller than standard classifications such as three-digit SIC codes.
4
Characteristics of FSIs
One concern with self-reported competitors is that firm managers may have an incentive to strategically choose the firms that they list as competitors in their regulatory filings. For example, despite the fact that Regulation S-K specifically prohibits firms from listing “misleading” competitors, managers may be more likely to list competitors that earned low stock returns over the fiscal year in question. Even worse, managers may simply chose not to report competitors if their firm performed particularly badly to avoid drawing comparisons to better-performing firms, particularly since reporting “misleading” competitors would be against the law. Both of these agency considerations be particularly tempting if managers’ compensation is tied to the relative performance of the firm versus its competitors, either via stock prices or through operating variables such as sales or profitability. However, an agency story could also operate in the opposite direction: firms that did particularly poorly 5
This compares favorably to CapitalIQ, which according to Rauh and Sufi (2012) can only assign approximately 30% of firms to direct competitors, and must assign the other 70% of firms to industries in an indirect fashion.
13
relative to their competitors may want to make it seem like their poor results were caused by the competitiveness of the industry in which they operate, in which case they may report a long list of competitors, including some that may not be accurate. Even if managers do not strategically report competitors, we might still expect certain types of firms to be more likely to report competitors than others. For example, firms in competitive industries with low barriers to entry may compete with so many other firms that it would be very burdensome to provide a full list of these competitors (for example, banks). Likewise, large conglomerates with many segments may compete against so many other firms in their various product lines that a full list of competitors would be difficult to compile. Furthermore, firms such as public utilities that essentially operate as regulated monopolies may not have any real competitors, and hence would not need to provide a competitor list.
4.1
What Types of Firms Report Competitors?
To test for these possibilities, I begin by examining the stock returns of “reporting” versus “non-reporting” firms during the twelve months prior to the firm’s fiscal year-end date. Reporting firms are firms that listed at least one competitor in its 10-K for the fiscal year in question, while non-reporting firms did not list any competitors in their 10-K filings. To examine the link between reporting firms and stock returns, I run a logistic regression where the dependent variable equals one if the firm is a reporting firm, and equals zero if the firm is a non-reporting firm. The main explanatory variables are prior (raw) 12-month and one-month stock returns, size, the firm’s book-to-market ratio (B/M), and operating variables (return on assets, return on equity, and sales scaled by total assets). Return on equity (ROE) is defined as net income divided by average book equity and return on assets (ROA) is defined as after-tax operating income divided by average total assets. Standard errors are clustered by year and by firm. Table 2 contains the results of these tests. The full-sample results with no fixed effects are a mixed bag: the size, B/M, stock return, ROE, and ROA variables all have negative and statistically significant coefficients, while the scaled sales variable has a positive and significant coefficient. The negative relationships between the propensity to report competitors and prior returns and operating metrics go against the hypothesis that managers are strate14
gically choosing to report their competitors. However, the scaled sales variable is positively related to reporting competitors, which may be supportive of the idea that managers are strategic in their choice to report competitors. One concern with the first column of Table 2 is that many of the variables are likely to be persistent across time and cross-sectionally across industries. Hence, in the second column of the table, I add year and industry fixed effects to the regression (the industry fixed effects are based on six-digit GICS classifications). Thus, my regression is now identifying off of variation within each industry during a given year. Hence, intra-industry differences such as differences between the competitiveness of industries will be captured by the fixed effects. Furthermore, since managers are likely to make strategic decisions based upon the firms in their “industry,” it makes sense to examine within-industry variation among firms’ decisions to report competitors. The table shows that most of the results from the first column also hold in the second column (albeit more weakly). The sole exception is the prior month’s stock return variable, which flips in sign from negative to positive. However, this result is likely a function of chance, as it seems unlikely that a single month’s stock return would cause managers to change their decision to report competitors relative to the stock return over the whole fiscal period, which seems far more relevant to their decision-making. Thus, the first two columns of the table provide at best weak evidence that managers are strategically choosing to report their competitors. However, it is possible that the explanatory variables do not do a good job of identifying firms that are prone to agency problems. To better aid in the identification of firms with agency concerns, I add the G-index variable constructed by Gompers, Ishii, and Metrick (2003) to the regression. The G-index variable is incremented in units based on corporate provisions that reduce shareholders’ rights. Hence, high values of the G-index are associated with poor corporate governance, and vice versa. Since the G-index is only available for large firms, the sample size is reduced by approximately 80% in regressions that include the G-index. However, these firms still account for over 90% of the total market capitalization within the CRSP/Compustat universe. The third column in Table 2 contains the logistic regression results with no fixed effects 15
using the smaller sample. To identify effects that are driven by the change in sample size alone, I leave out the G-index variable in this column. Reducing the sample size does have an effect on the results: in particular, a number of differences exist between the results in the third column and the results in column one. First, the loading on the B/M variable, while still negative, is no longer statistically significant. Second, the loadings on the stock return variables flip from negative to positive, with the 12-month return variable also being statistically significant. These results are consistent with an agency-based story, though such an effect only appears to exist within large firms. However, the scaled sales variable now flips sign from positive to negative and is now negative and statistically significant. Hence, all three operating variables show up as negative and statistically significant in column three. Column four adds the G-index variable to the sample. Consistent with an agency story, the G-index loading is negative and statistically significant, indicating that firms with good corporate governance are more likely than firms with poor corporate governance to provide a listing of their competitors. However, the inclusion of the G-index also has an effect on a number of the other variables in the regression. Specifically, the B/M variable returns to being negative and statistically significant, the 12-month return variable flips sign again and is now negative and statistically significant, and the scaled sales variable flips sign to now become positive (though not statistically different from zero). The fifth column of the table adds year and industry fixed effects to the regression containing the G-index variable. This is perhaps the cleanest test of the within-industry factors that may cause firms to make the (usually persistent) decision to report competitors in their public filings. As in previous columns, the loading on size is negative and statistically significant, indicating that smaller firms are more likely to report competitors than larger firms. Unlike the previous column, the loading on B/M is now statistically insignificant. Interestingly, the loading on 12-month stock returns flips sign again, this time from negative (in columns one, two and four) to positive and statistically significant. The one-month return variable is also positive, though not significant. The G-index variable also remains negative and statistically significant. Together, these three findings are consistent with an agencybased explanation: firms with high prior stock returns and good corporate governance are more likely than other firms in their industry to report a list of competitors in their public 16
filings. However, the loadings on two of the three operating variables (ROE and ROA) remain negative and significantly different from zero, which is not consistent with an agency story. Of course, it is possible that the agency story only exists through stock returns, and hence managers do not examine operating performance when choosing to report competitors. However, the converse is also true. In short, the results from column five are difficult to interpret: on the one hand, firms may report competitors strategically when they have done well (stock returns) or when they have done poorly (operating performance), while on the other hand, these results may simply be due to chance. To further isolate the potential for strategic reporting, column six of the table introduces firm fixed effects to the regression. Hence, the coefficients in this column will be driven by variation in the explanatory variables among firms that chose to report competitors in one year, but subsequently stopped reporting competitors, or vice versa. Since the choice to report competitors tends to be quite persistent, this significantly reduces the number of observations for which there is measurable variation. Furthermore, some of the explanatory variables tend to be persistent (i.e. the G-index), and hence are unlikely to have much explanatory power. However, if reporting decisions are driven by performance-related agency considerations, this column should be able to isolate such considerations. The column shows that most of the variables that were previously statistically significant in other specifications are now no longer significant. The only two variables that are statistically different from zero in column six of the table are the size and scaled sales variables. The size variable is now positive and large, indicating that firms are more likely to begin reporting competitors during the sample period after an increase in asset size. In contrast, the scaled sales variable is large and negative, indicating that firms are more likely to begin reporting competitors after a decrease in this variable. However, given that assets are in the denominator of the scaled sales ratio, this effect could also be driven by an increase in assets rather than a decrease in sales. To examine this possibility in further detail, I ran a separate regression replacing the scaled sales measure with year-over-year sales growth. In untabulated results, I find that the sales growth variable is not statistically significant. Hence, firm size appears to be the only variable that shows promise at predicting whether firms will begin or stop listing competitors in their public filings. [To do: look at mergers.] 17
In summary, evidence that managers report competitors strategically appears to be at best mixed. Within industries, there is evidence that poorly governed firms with low prior stock returns are less likely to report competitors than better-governed firms with higher prior stock returns. However, there is also evidence that firms with poor operating performance are more likely to report competitors than their peers. Furthermore, when the identification is limited to firms that changed their reporting habits, none of the agency-related variables are statistically significant. Hence, it is difficult to disentangle an agency-related explanation for firms’ decisions to report competitors. The only persistent effect seems to be size, but even this variable produces puzzling results: on the one hand, smaller firms are more likely to report competitors than larger firms within the same industry, but on the other hand, firms that begin reporting competitors after previously failing to report competitors tend to have experienced an increase in assets. Thus, Table 2 seems to suggest that the decision to report competitors is more or less random, at least within (traditional) industries.
4.2
Are Firms’ Competitor Lists Biased?
Table 2 examines the characteristics of firms that choose to report competitors in their public filings (versus firms that choose not to report competitors). However, conditional on the choice to report competitors, firms may also be biased in the choice of which companies to list as competitors. For example, managers may strategically report competitors that have underperformed during the past twelve months in order to make the firm look like it is beating its peers, even if this is not the case in practice. To examine this possibility, I run a set of tests in this section that compare the performance of firms that choose to list competitors with the performance of the firms selected as competitors. Hence, the sample considered in this section only includes firms that report competitors. Since Table 2 finds a relationship between firms’ G-index and their decision to report competitors, I also restrict the sample in this section to firms with a valid G-index value (the results are qualitatively unchanged when the broader sample is used). If managers are strategically selecting poorly performing competitors, firms that report competitors should earn higher returns and/or have better operating performance than their competitors. To examine this possibility, Table 3 lists the average differences in stock re18
turn/operating performance variables between a firm and its listed competitors. Since differences are paired, no time effects should exist within the data. The first column of the table lists the average differences across the entire panel. Paired t-statistics are used to determine whether firms’ variables differ from those of their competitors. The first column shows that firms do not earn higher (or lower) cumulative stock returns than the firm’s competitors in the 12 months and one month prior to the firm’s fiscal year-end date.6 Hence, on average, firms do not appear to select competitors that have earned low stock returns. In addition, firms appear to list competitors that have better operating performance than the firm itself – the average differences in ROE and ROA between firms and their competitors are negative and statistically significant. Furthermore, firms appear to list competitors that are both larger and have higher B/M ratios than the firm itself. Hence, under the assumption that strategic reporting would tend to produce weaker competitors, the data appear to show that firms do not strategically select competitors to report in their public filings. To gain comfort in this result, I also split the sample according to firms’ G-index values following the conventions outlined in Gompers, Ishii, and Metrick (2003). Firms with a G-index of five or less are categorized as “Democracies,” while firms with a G-index of 14 or more are categorized as “Dictatorships.” I next examine whether the difference between firms and their listed competitors is on average identical across Democracies and Dictatorships. If Dictatorships are more likely than Democracies to list competitors with low stock returns or poor operating performance, this could be an indication that at least a subset of firms are strategically listing their competitors. However, the next three columns of Table 3 show that this is largely not the case. Democracies and Dictatorships differ little with the exception of the scaled sales, B/M, and size variables. The table shows that Dictatorships are more likely than Democracies to list competitors with low relative sales, which may be indicative of strategic selection. However, in the absence of other variables that might indicate strategic selection, this evidence alone seems relatively weak. This is particularly true given that Dictatorships are more likely than Democracies to list firms with low B/M 6
I do not adjust for risk, under the presumptions that (i) risk is appropriately reflected in returns, and hence on average, not adjusting should not introduce any bias in the results, (ii) any executive compensation contracts that are tied to industry peers are likely based on risk-unadjusted returns, and (iii) managers most likely believe that investors care about raw returns rather than risk-adjusted returns.
19
values as competitors. Hence, it does not appear that much evidence exists to support the hypothesis that firms are strategically selecting their competitors.
4.3
Fixes
Tables 2 and 3 examine whether biases exist in the manner in which firms (i) choose to report a list of competitors versus not reporting a list at all; and (ii) select competitors to report conditional on choosing to provide a competitor list. The tables contain a number of results. First, I find mixed evidence in support of a selection effect in firms’ decision to report a list of competitors. Conditional upon reporting competitors, however, I find little evidence that firms are strategically reporting competitors that underperformed their peers along standard operating dimensions and/or in the stock market. This latter finding is particularly important given the structure of the dataset. If firms do not report competitors for strategic reasons, there are a number of ways to assign competitors to these firms in a reasonably unbiased fashion. In contrast, if firms report competitors in a biased fashion, the entire dataset may be tainted, and fixing this problem would be far more complicated. Thankfully, the data shows that to correct potential selection biases, we need only to find a way to assign competitors in an unbiased fashion to firms that did not report any competitors. The first solution I consider is to assign competitors to firm i based on the firms that list firm i as a competitor (in contrast, firms that listed a competitor are left unchanged). I refer to this combination as F SI − I industries, since they only include “first-level” competitors. However, these industries are also quite small. This is particularly troubling for stock return tests, as noted by Lewellen (2012a), since industry returns computed using a small number of stock are likely to be clouded by idiosyncratic noise. Thus, to increase the number of firms in each industry, I next include competitors’ competitors in each firm’s industry definition. This combination is referred to as F SI − II industries, since they include “second-level” competitors. Obviously, this additional level of competition eliminates many (but not all) of the benefits associated with FSIs – for example, the Coca-Cola and PepsiCo analogy is no longer valid. However, tiny industries are unlikely to be very useful, so the introduction of second-level competitors represents a tradeoff between industry size and accuracy.7 7
I could also impose a limit on the number of second-level competitors that could be added to each
20
It is important to note that while these assignment procedures are more reflexive than the original 10-K competitor lists, the FSI-I and FSI-II industries still need not be transitive or disjoint. For example, consider the following scenario: (i) firm A reports firms B and C as competitors; (ii) firm B does not report competitors; (iii) firm C reports firms A, B, D, and E as competitors; and (iv) firm D reports firms B and F as competitors. In the FSI-I classification, firm A’s competitors will be B and C, while firm B’s competitors will be A, C, and D. This latter result stems from firms A, C and D reporting firm B as a competitor. Next, under the FSI-II classification, firm A’s competitors will be firms B, C, D and E, while firm B’s competitors will be firms A, C, D, E, and F. Hence, even under the FSI-II classifications, industry definitions need not be transitive or disjoint. Given that FSI-I industries tend to be small, I thus use FSI-II industries for the remainder of my tests.
5
Asset Pricing Application: Stock Returns
The most obvious asset pricing application of FSIs is to examine whether FSIs do a better job than traditional classifications at explaining the variation in stock returns. In this section, I examine whether FSIs explain more variation in stock returns than traditional industry classifications. Other asset pricing applications not covered in this paper include using FSIs to determine whether industries can explain the value premium and momentum, which may potentially be intra-industry and inter-industry phenomena, respectively (see, e.g., NovyMarx (2009), Novy-Marx (2011) (value premium); Moskowitz and Grinblatt (1999), Grundy and Martin (2001) (momentum)). As stated in the previous section, my tests involve FSI-II industries. Following the literature, I exclude all financial firms and utilities from the sample (in untabulated results, I also find that these firms are far less likely to report competitors than firms in other industries). I also exclude all firms with a market capitalization below the 10% NYSE size breakpoint each month, as well as all firms with a CRSP share code that is not 10 or 11. Finally, I exclude all industries that contain less than five firms, as these industry returns industry. In untabulated results, doing so improves the explanatory power across all tests listed in the paper. However, for the purposes of conservatism, I stick with the entire list of second-level competitors.
21
are likely to contain a significant amount of idiosyncratic noise (Lewellen (2012a)). I begin by comparing the performance of equally-weighted FSIs against traditional industry classifications.8 To parsimoniously incorporate a variety of commonly-used standards of differing granularities, I choose to examine the following classification standards: three-digit SIC classifications, six-digit GICS classifications, and FF48 classifications. I also include the TNICs of Hoberg and Phillips (2010a,b). By definition, the FSI assigned to firm i does not contain firm i (all such references are deleted), but this is not true within standard industry classifications. Thus, to create a level playing field, I exclude firm i when computing the industry return that is matched to firm i. To construct my stock return tests, I begin by computing firm and industry returns for each of the various industry classification standards. To avoid look-ahead bias, industry definitions from fiscal year t are matched with stock returns from July of year t + 1 to June of year t + 2. Hence, a firm that reports its competitor list in its 10-K filed in December of year t will be assigned to its corresponding FSI from July of year t + 1 to June of year t + 2, and the industry returns on this FSI will be computed for these dates as well. Since the sample of 10-Ks ranges from 2002-2008, the sample of stock returns ranges from July 2003 to June 2010. Following Rauh and Sufi (2012), I run panel regressions of the form: ri,t − rf,t = α + β1 (rm,t − rf,t ) + β2 (rind,t − rf,t ) + εi ,
(1)
where ri,t − rf,t represents the excess return on firm i in month t, rm,t − rf,t represents the excess return on the market portfolio, and rind,t − rf,t represents the return on firm i’s industry (excluding firm i). Standard errors are clustered by time. Following Rauh and Sufi (2012), industry returns are computed on an equal-weighted basis for my initial tests. Table 4 presents the basic stock return results. The table shows that on an equal-weighted basis, FSIs explain a larger fraction of stock returns than their traditional counterparts as measured by adjusted R2 . FSIs also explain a greater fraction of stock returns than Hoberg 8
The decision to focus on equally-weighted industries stems in part from Regulation S-K, which only requires firms to list competitors if the competitor is “dominant.” It also stems in part from the decision to use FSI-II industries, which are more likely to include one or more massive firms than FSI-I industries. For similar reasons, Rauh and Sufi (2012) also focus on equally-weighted industries.
22
and Phillips (2010a,b)’s TNIC classifications. The improvements in explanatory power are material: on an equal-weighted basis, adjusted R2 s improve by about 13% relative to threedigit SIC, FF48, and TNIC classifications, and by about 4% relative to six-digit GICS classifications. The panel also includes a “horserace” test between the various industry classifications. Consistent with FSIs being “more correct” definitions of industries, column 6 of the panel shows that the factor loading on FSIs is more than 50% larger than its nearest competitors (six-digit GICS and TNICs). Hence, FSIs offer a material improvement over other industry classification standards at explaining the variation in stock returns. Columns 7-9 of the panel examine how FSIs compare with the static competitor definitions from CapitalIQ used by Rauh and Sufi (2012). Rauh and Sufi (2012) also examine the relative performance of their static CapitalIQ competitors versus three-digit SIC classifications in panel regressions using equal-weighted industry returns with standard errors clustered by time. While differences in the testing methodologies unfortunately cloud a direct comparison between FSIs and the CapitalIQ data, the results show that FSIs appear to be at least as good as the static classifications used by Rauh and Sufi (2012) at explaining variation in stock returns. While FSIs improve over traditional classifications on an equal-weighted basis, the “network” mapping of firms generated by their FSI links allows for alternative weighting schemes. These alternative schemes can either be theoretically driven (i.e. network concepts such as “betweenness,” “closeness,” and “degree centrality”) or empirically driven (i.e. determine the weighting scheme that maximizes adjusted R2 s). In the current draft, I focus on the latter concept by attempting to construct the FSI weighting scheme that maximizes adjusted R2 s. To prevent look-ahead bias, I use the following algorithm. First, I construct the firm return and FSI return for each firm during the 12 months prior to the firm’s fiscal year-end date. For each fiscal year, I then increment the weights of each class of firms (directly reported competitors; indirect competitors; competitors’ competitors) from 0 to 1 in units of 0.05 and run a panel regression similar to (1) on all firms for that fiscal year.9 I then select the weighting scheme that delivers the highest adjusted R2 value. Next, this scheme 9
This range of weights spans any possible weighting scheme that sums to one under the restriction that weights cannot be negative. For example, a weight of {1, 1, 1} is equivalent to equal-weighting all three firm classes, while a weight of {0, 1, 0} is equivalent to only including indirect competitors in the FSI return.
23
is used to weight industry returns on each FSI from July of year t + 1 to June of year t + 2, mirroring the industry assignment period for that fiscal year. Since the data used to generate the weights is backward-looking, this procedure ensures that there is no look-ahead bias in the weighting scheme. Table 5 shows the optimal weighting scheme for each year according to this algorithm. The table also reports the adjusted R2 values from the first-stage regressions that were used to determine the optimal weights (the reported R2 values are based on the optimal weighting scheme). The table shows that on average, while the optimal weight on direct competitors is always one, the average optimal weight on indirect competitors is only 0.35, while the optimal weight on competitors’ competitors is only about 0.20. Hence, while the indirect competitors and competitors’ competitors add significant value to the industry definitions, the optimal weighting scheme still places most of the weight on directly-listed competitors. The next three columns of Table 5 normalize the weights to sum to one. These columns show that of a total weight of 100%, approximately two-thirds of the weight is placed on directly-listed competitors. The last column of Table 4 examines the performance of these “optimal” FSIs relative to other classification standards. The adjusted R2 value from the regression in (1) improves to 0.28, which is a 17% improvement over three-digit SIC, FF48, and TNIC classifications, and an 8% improvement over six-digit GICS classifications. Hence, while “optimal” FSIs do not offer that large of an improvement over equal-weighted FSIs, the difference between optimal FSIs and other industry classifications is quite large. A number of other possibilities exist to construct more “optimal” industries using FSIs. These include theoretical possibilities such as using network theory, as well as empirical possibilities such as combining FSIs with TNICs or with traditional classifications to improve the power of such classifications to explain stock returns.
6
Corporate Finance Application: Leverage
Table 6 examines whether FSIs do a better job of explaining the variation in corporate capital structure than other types of industry classifications. I begin by restricting the sample to all 24
firms with assets over $10 million that contain at least five industry competitors. Financial firms and utilities are also excluded from the sample. Following Rauh and Sufi (2012) and Hoberg and Phillips (2010a,b), I then regress the book leverage ratio of each firm against the equal-weighted leverage ratio of its competitors. Thus, all of the industry leverage measures (including FSIs) are equally weighted. Book leverage is defined as total debt divided by total assets (Compustat variables dltt+dlc / at). Standard errors are clustered by firm and time, and time fixed effects are included in the regression. The table shows that FSIs offer a significant improvement over traditional classification standards at explaining the variation in leverage. The regression with FSIs has an adjusted R2 of 0.16, which is 45% higher than the R2 s from regressions involving three-digit SIC and FF48 classifications (0.11) and 33% higher than the adjusted R2 of the regression involving six-digit GICS classifications (0.12). Thus, like stock returns, FSIs do a much better job of explaining the variation in leverage than traditional industry classifications. Furthermore, like stock returns, I can use both theoretically-motivated and empirically-motivated methods to increase the explanatory power of FSIs in future tests.
7
Conclusions
This paper introduces firm-specific industries, or FSIs. Unlike traditional “fixed” industry classifications, FSI definitions are neither transitive nor disjoint. As such, FSIs should in theory do a better job of capturing industry dynamics and measuring industry effects. I examine the potential benefits of FSIs through an asset pricing application and a corporate finance application. On the asset pricing side, I show that FSIs offer a material improvement over existing classifications in their ability to explain the variation in stock returns. On the corporate finance side, I show that FSIs offer an improvement over traditional classifications in their ability to explain corporate capital structure.
25
References Bhojraj, Sanjeev, Charles M. C. Lee, and Derek K. Oler, 2003, What’s My Line? A Comparison of Industry Classification Schemes for Capital Market Research, Journal of Accounting Research 41, 745–774. Bolton, Patrick, and David A. Scharfstein, 1990, A Theory of Predation Based on Agency Problems in Financial Contracting, American Economic Review 76, 956–970. Brander, James A., and Tracy R. Lewis, 1986, Oligopoly and Finacial Structure: The Limited Liability Effect, American Economic Review 76, 956–970. Chan, Louis K. C., Josef Lakonishok, and Bhaskaran Swaminathan, 2007, Industry Classifications and Return Comovement, Financial Analysts Journal 63, 56–70. Chevalier, Judith A., 1995a, Capital Structure and Product Market Competition: Evidence from the Supermarket Industry, American Economic Review 85, 206–256. , 1995b, Do LBO Supermarkets Charge More? An Empirical Analysis of the Effects of LBOs on Supermarket Pricing, Journal of Finance 50, 1095–1112. , 2004, What Do We Know About Cross-subsidization? Evidence from Merging Firms, B.E. Journal of Economics and Policy 4. Clarke, Richard N., 1989, SICs as Delineators of Economic Markets, Journal of Business 62, 17–31. Dixit, Avinash, 1989, Entry and Exit Decisions under Uncertainty, Journal of Political Economy 97, 620–638. Fama, Eugene F., and Kenneth R. French, 1997, Industry Costs of Equity, Journal of Financial Economics 43, 153–193. Ferriera, Miguel A., and Paul A. Laux, 2007, Corporate Governance, Idiosyncratic Risk, and Information Flow, Journal of Finance 62, 951–990.
26
Giroud, Xavier, and Holger M. Mueller, 2011, Corporate Governance, Product Market Competition, and Equity Prices, Journal of Finance 66, 563–600. Gompers, Paul, Joy Ishii, and Andrew Metrick, 2003, Corporate Governance and Equity Prices, Quarterly Journal of Economics 118, 107–155. Graham, John R., and Mark T. Leary, 2011, A Review of Empirical Capital Structure Research and Directions for the Future, Annual Review of Financial Economics 3, 309– 345. Grundy, Bruce D., and J. Spencer Martin, 2001, Understanding the Nature of the Risks and the Source of the Rewards to Momentum Investing, Review of Financial Studies 14, 29–78. Guenther, David A., and Andrew J. Rosman, 1994, Differences between COMPUSTAT and CRSP SIC Codes and Related Effects on Research, Journal of Accounting and Economics 18, 115–128. Hoberg, Gerard, and Gordon Phillips, 2010a, Product Market Synergies and Competition in Mergers and Acquisitions: A Text-Based Analysis, Review of Financial Studies 23, 3773–3811. , 2010b, Text-Based Network Industries and Endogenous Product Differentiation, Working Paper. Hopenhayn, Hugo, 1992, Entry, Exit, and Firm Dynamics in Long Run Equilibrium, Econometrica 60, 1127–1150. Hou, Kewei, and David T. Robinson, 2006, Industry Concentration and Average Stock Returns, Journal of Finance 61, 1927–1956. Kahle, Kathleen M., and Ralph A. Walking, 1996, The Impact of Industry Classifications on Financial Research, Journal of Financial and Quantitative Analysis 31, 309–335. Leary, Mark T., and Michael R. Roberts, 2011, Do Peer Firms Affect Corporate Financial Policy?, Working Paper. 27
Lewellen, Stefan, 2012a, Corporate Governance and Equity Prices: Are Results Robust to Industry Adjustments?, Working Paper. , 2012b, Executive compensation and peer effects, Working Paper. MacKay, Peter, and Gordon Phillips, 2005, How Does Industry Affect Firm Financial Structure?, Review of Financial Studies 18, 1434–1466. Maksimovic, Vojislav, 1988, Capital Structure in Repeated Oligopolies, Rand Journal of Economics 19, 389–407. , and Josef Zechner, 1991, Debt, Agency Costs, and Industry Equilibrium, Journal of Finance 46, 1619–1643. Miao, Jianjun, 2005, Optimal Capital Structure and Industry Dynamics, Journal of Finance 60, 2621–2659. Moskowitz, Tobias J., and Mark Grinblatt, 1999, Do Industries Explain Momentum?, Journal of Finance 54, 1249–1290. Novy-Marx, Robert, 2009, Competition, Productivity, Organization and the Cross-Section of Expected Returns, Working Paper. , 2011, Operating Leverage, Review of Finance 15, 103–134. Phillips, Gordon M., 1995, Increased Debt and Industry Product Markets: An Empirical Analysis, Journal of Financial Economics 37, 189–238. Rauh, Joshua, and Amir Sufi, 2012, Explaining Corporate Capital Structure: Product Markets, Leases, and Asset Similarity, Review of Finance 16, 115–155. Spiegel, Matthew, and Heather Tookes, forthcoming, Dynamic Competition, Valuation, and Merger Activity, Journal of Finance. Stein, Jeremy C., 2003, Agency, Information and Corporate Investment, Handbook of the Economics of Finance pp. 109–163.
28
Weiner, Christian, 2005, The Impact of Industry Classification Schemes on Financial Research, SFB 649 Discussion Paper 2005-062. Williams, Joseph T., 1995, Financial and Industrial Structure with Agency, Review of Financial Studies 8, 431–474.
29
Figure 1 Traditional Industry Definitions Industry 1:
C
B B
A
D
C D
A
BF
G
Industry 2:
E
F
G H
H
E
This figure shows the structure of traditional industry classification definitions, in which industries are both transitive and disjoint. The first set of figures on the left shows two industries formed using traditional definitions, each of which contains four firms. In the figures, an arrow emanating from firm X to firm Y indicates that firm Y is in firm X’s industry classification. As the figure shows, each of the firms in an industry is linked to all of the other firms within that industry definition. Hence, if firm X is in firm Y’s industry, firm Y is also in firm X’s industry. The second set of figures on the right displays all of the competitors of a single firm (in this case, firms A and E).
30
Figure 2 Firm-Specific Industry Definitions
B
C B
A
D
C
A
D
D
A
B
BF
F
H
E
G
D
H
H
E
F
This figure shows the structure of firm-specific industry definitions. Unlike traditional (disjoint) industry definitions, firm-specific definitions allow for an accurate mapping of the product market space. For example, firms A and B now have different sets of competitors: firm A does not compete with firm C, while firm B does compete with firm C. This would not be allowed under the traditional system of defining industries.
31
Table 1 Summary Statistics Panel A: Number of Firms per Industry with Valid GVKEY
Year 2002 2003 2004 2005 2006 2007 2008
Avg. # of firms per industry 8.6 9.0 9.3 9.6 9.8 9.8 9.6
Median # of firms per industry 7 7 7 7 7 7 7
Std. Dev. 8.2 8.6 8.7 8.6 8.6 8.6 9.0
Min 1 1 1 1 1 1 1
25% 3 4 4 4 4 4 4
75% 11 12 12 13 13 13 13
Max 153 158 163 139 127 129 135
75% 9 10 10 11 11 11 11
Max 129 125 123 119 109 112 120
Panel B: Number of Firms per Industry matched to CRSP/Compustat
Year 2002 2003 2004 2005 2006 2007 2008
Avg. # of firms per industry 7.2 7.5 7.7 8.0 8.1 8.1 8.1
Median # of firms per industry 6 6 6 6 6 6 6
Std. Dev. 6.5 6.9 6.9 6.9 6.7 6.9 7.2
Min 1 1 1 1 1 1 1
32
25% 3 3 3 4 4 3 3
Table 2 Characteristics of Firms Reporting Competitors This table contains the results from a logistic regression of a number of predictive variables on a dummy variable taking the value of one if a firm self-reported competitors in a given year and a value of zero otherwise. Explanatory variables include the natural log of total assets, the book-to-market ratio (of assets), the 12-month cumulative stock return prior to each firm's fiscal year-end date, the firm's stock return in the month prior to the fiscal year-end date, the firm's return on equity (net income / average assets), return on assets (EBIT / average assets), a scaled sales measure (sales / average assets), and the Gompers, Ishii and Metrick (2003) G -index value for each firm as of its fiscal year-end date. The sample period is 2002-2008. The first two columns include the full sample, while the next four columns only include firms with a valid G -index value, which reduces the number of firms in the sample by a factor of five. The ROE, ROA, scaled sales and B/M variables are Winsorized at the 0.5% and 99.5% levels in each fiscal year. G-index values are sourced from Andrew Metrick's website; accounting data is sourced from Compustat; and stock return data comes from CRSP. Industry controls are based on four-digit GICS industry classifications. Statistical significance at the 1, 5, and 10% levels is denoted by the symbols ***, **, and *, respectively. DV = 1 if firm i reported competitors in its 10-K in year t Full sample Variable Log Assets B/M 12-month return 1-month return ROE ROA Sales / Assets
(1)
G -index sample (2)
(3)
(4)
(6)
-0.0370***
-0.1051***
-0.2272*** -0.0924***
(0.0060)
(0.0069)
(0.0183)
(0.0159)
(0.0183)
-1.4452***
-1.0594***
-0.1849
-2.1399***
-0.1625
-1.2856
(0.0340)
(0.0346)
(0.1509)
(0.1257)
(0.1510)
(0.8481)
-0.4242***
-0.1433***
(0.0185)
(0.0197)
-0.1711**
0.1361** (0.0669)
(0.0505)
(0.2717)
0.1383**
-0.1104
(0.0672)
(0.2407)
0.1737
0.0626
0.1884
1.1391
(0.0803)
(0.3013)
(0.2590)
(0.3019)
(0.8791)
-0.1378***
-0.0925***
-0.1479*
-0.3870***
-0.1446*
0.5925
(0.0277)
(0.0284)
(0.0803)
(0.1045)
(0.0793)
(0.3894)
-0.9641***
-0.7381***
-1.8641***
-4.3756*** -1.8727***
-3.2830
(0.0904)
(0.1028)
(0.4099)
(0.4034)
(0.4090)
(2.2833)
0.0271
-0.1743***
(0.0326)
(0.0443)
0.3048***
0.1765**
-0.1303***
1.3357***
(0.0714)
(0.0142)
0.0473***
-0.1822***
(0.0182)
(0.0442)
G -index
-0.0616*** -0.0513*** (0.0092)
Constant
(5)
-0.1197***
1.0291***
3.9955***
0.0128 (0.1420)
1.3794***
8.7605
(9.5807)
(0.1880)
(0.1631)
(0.2014)
(64.3605)
Year fixed effects
No
Yes
Yes
No
Yes
No
Industry fixed effects
No
Yes
Yes
No
Yes
No
Firm fixed effects
No
No
No
No
No
Yes
Firm-year observations
-1.5551
(0.4404)
(0.0478)
Pseudo R-squared
1.1307***
(0.0104)
-0.9982**
0.11
0.29
0.12
0.27
0.27
0.71
41,475
41,361
8,666
8,658
8,658
8,658
33
Table 3 Do Firms Report Competitors Strategically? This table compares the stock returns and operating variables of firms that report competitors against the equally-weighted average returns and operating variables of such competitors. Coeffiecient estimates stem from an unbalanced panel of data that is restricted to firms that reported competitors in a given year and also have a valid G -index value. The sample period is 2002-2008. The subscripts "firm" and "comp" represent the returns/operating performance of the firm and its listed competitors, respectively. For example, the first variable measures the difference in cumulative 12-month returns prior to the fiscal year-end date between a firm and the competitors it lists in its public filings. The first column of the table shows the average difference between a firm and its competitors across the sample. The next three columns break the sample into "Democracies", governance-neutral firms, and "Dictatorships" based on the methodology outlined in Gompers, Ishii and Metrick (2003). Dictatorship firms have poor governance, while Democracies have better governance. The "Difference" column represents the difference in coefficient estimates between Dictatorships and Democracies. Standard errors and statistical significance are based on two-sided paired t -tests. Statistical significance at the 1, 5, and 10% levels is denoted by the symbols ***, **, and *, respectively. All
G -index Category
Variable
Data
Dem
Dict
Difference
Rfirm,12 - Rcomp,12
0.016
-0.029
0.001
0.030
(0.010)
(0.028)
(0.029)
(0.029)
-0.003
-0.007
0.002
0.009
(0.002)
(0.006)
(0.006)
(0.006)
ROEfirm - ROEcomp
-0.121***
-0.103**
-0.086***
(0.012)
(0.054)
(0.023)
ROAfirm - ROAcomp
-0.034***
-0.023***
-0.016**
(0.002)
(0.008)
(0.007)
Rfirm,1 - Rcomp,1
Sales firm - Sales comp BMfirm - BMcomp Sizefirm - Sizecomp
0.017 (0.042) 0.007 (0.008)
0.009
-0.034*
0.093***
0.127***
(0.008)
(0.018)
(0.047)
(0.036)
0.041***
-0.005
0.114***
0.119***
(0.004)
(0.016)
(0.020)
(0.018)
-1.800***
-1.699***
-1.216***
0.483***
(0.028)
(0.093)
(0.133)
(0.115)
34
Table 4 Stock return tests Industries are constructed on a firm-specific basis using (i) companies that the firm listed as a competitor in its 10-K filing; (ii) other companies that point to the firm as a competitor in their own 10-K filings; and (iii) the firm's competitors' competitors. The sample period for 10-Ks is 2002-2008. Industry data from year t is matched to CRSP stock returns from June of year t+1 to July of year t+2. Thus, the sample period for stock returns is July 2003 - June 2010. All other data comes from Compustat. Industries are equally-weighted as in Rauh and Sufi (2012), and firm i is always excluded from its industry return. The dependent variable in all regressions is monthly excess firm stock returns. "Market return" represents the return on the CRSP value-weighted portfolio in month t. "FSI return" represents the return on the firm's industry as calculated using the method described above. The other variables represent equal-weighted industry returns computed using other industry classification standards. To eliminate the effects of small firms, firms smaller than the 10% NYSE size breakpoint are dropped, as are firms matched to industries that contain less than five firms. Financial firms and utilities are also excluded from the sample, as are firms with CRSP share codes other than 10 and 11. Statistical significance at the 1, 5, and 10% levels is denoted by the symbols ***, **, and *, respectively. Max DV: excess return of firm i in month t Excess returns:
(1)
(2)
(3)
(4)
(5)
Rauh and Sufi (2012) (6)
(8)
(9)
(10)
0.2267*** 0.4594*** 0.6667*** 0.3279*** 0.3844***
(0.0308)
(0.092)
(0.102)
(0.0156)
FSI returni,t
0.8390***
0.5002***
0.547***
0.520***
0.8235***
(0.0132)
(0.0309)
(0.025)
(0.033)
(0.0092)
TNIC returni,t
(0.0537)
(0.0461)
(0.0433)
(0.0469)
0.6093***
(0.0249) 0.4585***
0.1001***
0.119
0.058
(0.0407)
(0.0076)
(0.081)
(0.043)
6-digit GICS return i,t
0.7134***
0.2294***
(0.0403)
(0.0310)
FF48 returni,t
0.6672*** -0.0656 (0.0432)
Constant N Adjusted R-squared
(0.192)
0.2512***
0.2167***
(0.0483) 3-digit SIC return i,t
0.635*** 1.166*** 0.592***
R-squared
Market returnt
(0.0188)
0.0053
(7)
(0.0213)
0.0001
-0.0013
-0.0001
-0.0011
-0.0010
-0.0011
0.0019
0.0028
0.0017
0.0001
(0.0005)
(0.0016)
(0.0015)
(0.0021)
(0.0020)
(0.0013)
(0.002)
(0.003)
(0.002)
(0.0005)
117,159
117,159
117,159
117,159
117,159
117,159
144,588
144,588
144,588
117,159
0.27
0.24
0.24
0.26
0.24
0.28
0.12
0.08
0.12
0.28
35
Table 5 Optimal Weights Assigned to Competitors Optimal weight assigned to: 10-K Direct Indirect Competitors' Year Competitors Competitors Competitors 2002 1 0.55 0.20 2003 1 0.20 0.15 2004 1 0.35 0.20 2005 1 0.35 0.20 2006 1 0.30 0.20 2007 1 0.40 0.20 2008 1 0.30 0.20 Average: 1 0.35 0.19
Adjusted weight Maximum Direct Indirect Competitors' Adjusted Competitors Competitors Competitors R-squared 0.57 0.31 0.11 0.274 0.74 0.15 0.11 0.248 0.65 0.23 0.13 0.204 0.65 0.23 0.13 0.224 0.67 0.20 0.13 0.187 0.63 0.25 0.13 0.128 0.67 0.20 0.13 0.354 0.65 0.22 0.13 0.231
36
Table 6 Capital Structure Tests Industries are constructed as before. Leverage is defined as total debt divided by total assets (i.e. [dltt+dlc]/at). Financial firms and utilities are excluded from the sample. Standard errors are clustered by year and by firm. Statistical significance at the 1, 5, and 10% levels is denoted by the symbols ***, **, and *, respectively.
Variable Constant FSI
(1) -0.007 (0.004) 1.016*** (0.019)
3-digit SIC
DV: Leverage of firm i in year t (2) (3) (4) 0.032*** 0.018*** 0.026*** (0.008) (0.006) (0.009)
0.837*** (0.032)
6-digit GICS
0.890*** (0.026)
FF48 Year fixed effects Adjusted R-squared
Yes 0.16
Yes 0.11
37
Yes 0.12
0.846*** (0.030) Yes 0.11
(5) -0.023** (0.009) 0.623*** (0.023) 0.117*** (0.045) 0.308*** (0.038) 0.042 (0.040) Yes 0.20
