Managed vs. Unmanaged Changes in the Capital Structures of Firms with Extreme Leverage

Transcription

1 BAYLOR UNIVERSITY Managed vs. Unmanaged Changes in the Capital Structures of Firms with Extreme Leverage Is the extreme use of financial leverage a choice? Steve Green: steve_green@baylor.edu John Martin: j_martin@baylor.edu * Steve Rich: steve_rich@baylor.edu * Contact Author 3/26/2011 In this paper we study the capital structures of highly levered firms. Specifically, we differentiate between intentional or managed changes in the firm s use of financial leverage versus unmanaged changes that are beyond the direct control of management. We find that the financial leverage (book and market) of extreme leverage firms rises rapidly in the years immediately prior to their identification as extreme-leverage firms, and the increase in leverage is not the result of managerial decisions to issue debt or repurchase common stock. Instead, firms become extremely levered primarily because of circumstances that are beyond the direct control of management. We conclude that firms do not become highly levered because it is their optimal capital structure. Consequently, any attempt to explain their use of financial leverage using capital structure theory cannot succeed. Green, Martin, and Rich

2 TABLE OF CONTENTS 1 Introduction Identifying Extreme Leverage Firms... 1 Measuring Financial Leverage... 2 Sample Selection... 3 Who are the Extreme Leverage Firms?... 3 Book versus Market Leverage... 4 Extreme Leverage and Industry Group Sources of Change in the Leverage for Extreme Leverage Firms... 8 Evolution of Leverage Ratios for Extreme Leverage Firms... 8 Managed versus Unmanaged Changes in Book Leverage Isolating Managed versus Unmanaged Changes in Book Leverage Analysis of Managed versus Unmanaged Changes in the Market Leverage Ratio Concluding Remarks References Appendix A How Recent Capital Structure Studies Measure Financial Leverage Appendix B Firm Classification with SIC and Historical SIC Codes Appendix C Impact of Sample Restrictions on Sample Composition Appendix D An Example Firm: Goodyear Tire & Rubber Green, Martin, and Rich

3 MANAGED AND UNMANAGED CHANGES IN THE CAPITAL STRUCTURES OF FIRMS WITH EXTREME FINANCIAL LEVERAGE 1 INTRODUCTION Capital structure theory seeks to explain the financial choices that firms make regarding their use of debt and equity financing sources. The unstated assumption underlying empirical tests of these theories is that corporate managers control the firm s capital structure. Recent work by Denis and McKeon (2010) questions the degree to which management does control the firm s capital structure suggesting that we must differentiate between intentional and unintentional changes in a firm s use of leverage if we are to successfully identify the determinants of corporate financing decisions and capital structures. In this paper we study a group of firms that have extremely high financial leverage and analyze how they came to be so highly levered. Specifically, we investigate whether their extreme-leverage capital structures are a consequence of the exercise of managerial discretion (i.e., managed changes in the firm s capital structure) or the result of forces outside the direct control of the firm s management (i.e., unmanaged changes). We find that most firms with extreme levels of financial leverage become highly levered due to factors that are beyond the direct control of management. That is, managerial decisions do not directly cause the firm s use of extremely high financial leverage. Consequently, our findings suggest that we should not expect capital structure theories directed at explaining managerial financing decisions to be of much help in explaining the use of financial leverage by these extreme-leverage firms. The paper is organized as follows: In Section 2, we define financial leverage using both book- and market-based metrics and use each to identify extreme leverage firms. This involves examining the cross-section of capital structures for all industrial firms in the Compustat database for the period In Section 3, we investigate whether extreme-leverage firms became extremely levered as a result of capital structure choices or due to factors management could not control. Finally, in section 4 we offer some summary comments regarding the implications of our study of extreme-leverage firms for future capital structure research. 2 IDENTIFYING EXTREME LEVERAGE FIRMS To identify extreme leverage firms requires that we first identify an appropriate measure of financial leverage. Although this may seem like a simple task, there is little agreement as to how to best measure leverage. In studies of the level of financial leverage or of changes in financial leverage published in the Journal of Finance, the Journal of Financial Economics, or the Review of Financial Studies from 2008 through 2010, about a third of the studies present results for both market and book leverage ratios, about a third present results for only book leverage ratios, and about a third present results for only market leverage ratios (See Appendix A). 1 Although book and 1 The seven studies that report results for both market and book leverage include: Almazan, De Motta, Titman, and Uysal (2010), Byoun (2008), Leary (2009), Lemmon, Roberts, and Zender (2008), Matsa (2010), Ovtchinnikov (2010), and Shivdasani and Steffanescu (2010). The seven studies that report results for only book leverage include: Brav (2009), Chang and Dasgupta (2009), Desai, Foley, and Hines (2008), Huizinga, Laeven, and Nicodeme (2009), Leary and Roberts (2010), Rauh and Sufi (2010), and Roberts and Sufi (2009). The six studies that report results for only market leverage include: Bharath, Pasquariello, and Wu (2009), Banerjee, Dasgupta, and Kim (2008), Carlson and Lazark (2010), Harford, Klasa, and Walcott (2009), Korteweg (2010), and Morellec and Zhandov (2008). Several studies that present results for only one of the measures of leverage state they find similar results with the other measure of leverage. Note also that several studies that report only results for book leverage examine private firms where market values are unavailable. Green, Martin, and Rich Page 1

4 market based measures of financial leverage often yield similar results for the broad cross-section of firms, it is not clear whether this holds for the extreme leverage firms we study. Therefore, we use both book- and marketbased leverage metrics in our analysis. MEASURING FINANCIAL LEVERAGE We measure book-value financial leverage as the Book Debt to Capital Ratio (BL) ratio which we define as follows: 2 (1) The acronyms in parentheses are the Compustat variable names. Note that the numerator of the ratio only contains interest bearing debt comprised of short-term debt plus long-term debt where debt in current liabilities (DLC) includes both short-term notes as well as the current portion of long-term debt. The denominator equals the sum of the firm s invested capital which excludes consideration for non-interest bearing liabilities. Welch (2010) discusses the problems that can arise where the treatment of non-interest bearing liabilities is not treated consistently when calculating a firm s financial leverage. Specifically, if these liabilities are considered debt in the numerator of the debt ratio it is important that they be also be included in the denominator, and vice versa. We exclude them from both numerator and denominator of both our book- and market-value based debt ratios. Note that we did not use net debt (total debt less cash and short-term investments) as our measure of financial leverage since it is not commonly used. Only two of the studies that we surveyed used a net debt measure. However, Bates, Kahle, and Stulz (2009) report that the average net debt ratio has been declining since 1980 and became negative in 2004 and remained below zero through the end of their sample period in We define market leverage using a quasi-market value leverage debt ratio as follows: 3 (2) This is a quasi-market value ratio in that debt and preferred equity are measured using their book values and only the firm s common equity is measured using its market value. The use of book values for debt and preferred 2 In the capital structure studies published in the Journal of Finance, the Journal of Financial Economics, or the Review of Financial Studies from 2008 through 2010, the most common method for measuring book leverage is (DLC + DLTT)/AT where AT equals the total assets of the firm. However, this measure only appears in nine of the 16 studies (while only 14 studies present results for book leverage, two discuss how they calculate book leverage in unpublished results) and is the only measure of book leverage in five of the 16 studies. In panel a. of Appendix A we show the measures of book leverage for all 16 studies. 3 In capital structure studies published in the Journal of Finance, the Journal of Financial Economics, or the Review of Financial Studies from 2008 through 2010, the most common method for measuring market leverage is (DLC + DLTT)/(DLC + DLTT + CSHO*PRCC_F). However, this measure appears in only 5 of 14 studies (while only 13 studies present results for market leverage, one discusses how they calculate market leverage in unpublished results). In panel b. of Appendix A we show the measures of market leverage for all 14 studies. Green, Martin, and Rich Page 2

5 equity reflects the difficulty involved in obtaining market values for these two classes of securities and is common practice in previous capital structure studies. SAMPLE SELECTION Our sample includes North American industrial firms for the years 1987 through We exclude utilities (SIC ), financials (SIC ), firms with missing historical SIC codes (SICH), and firms lacking sufficient Compustat data to calculate Equations (1) or (2). We begin with 1987 because we use historical SIC codes to eliminate utilities and financial firms from our historical samples, and Compustat reports historical SIC codes for only a handful of firms prior to Appendix B shows how differences between current and historical SIC codes change over time. We impose two further restrictions on the firms we examine. First, when examining book leverage we drop firms with book debt to capital ratios that do not fall between 0 and 1. Second, when examining market leverage we drop firms with either outstanding shares (CSHO) or a market price at the end of the firm s fiscal year (PRCC_F) equal to zero. Since these restrictions are not the same for book and market leverage measures, the samples used for each measure are not identical. 4 WHO ARE THE EXTREME LEVERAGE FIRMS? For our purposes we will define a firm as having extreme financial leverage if its leverage is in the 95 th percentile of all firms using both book and market leverage ratios. To get an idea as to just how much financial leverage firms use at different times over our sample period we prepared Figure 1. Panel a. depicts the time series of various percentiles of leverage ratios for the period Specifically we show the 95 th, 85 th, 50 th, and the 15 th percentiles for each year. 5 In addition, Panel b. of Figure 1 indicates the number of firms for which we are able to calculate book and market leverage ratios for each year We can make several observations about debt ratios from this figure: First, the capital structures of high leverage firms (85 th, 95 th, and 99 th percentiles) exhibit distinct trends over time. The market and book leverage of high leverage industrial firms generally rose prior to 1990, fell between 1990 and 1993, rose from 1993 to 2002, fell from 2002 to 2006, and rose again through Second, the median firm s leverage generally fell from 1987 through 2007 but temporarily rose in 2008 at the beginning of the financial crisis. Third, low debt firms (15 th percentile) reduced their debt towards zero through the year 2000 and remained unlevered thereafter. 4 Differences in the two samples stem from differences in the data needed to calculate each of the leverage measures and differences in the data available for each firm. In Appendix D we show the impact of each of the data restrictions on the size of our samples. Differences between the samples do not affect the results since we find the same basic results when we used a pooled sample of all firms for which we can calculate both book and market debt ratios. We show the sizes of both samples in Panel b of Figure 1. In 1987 through 1999, the book-leverage sample is larger and in 2000 through 2009, the market-leverage sample is larger. 5 We also examine the 5 th percentile, but these firms have no debt in any of the years we examine. 6 These general trends do not hold exactly for the 85 th, 95 th, and 99 th percentiles for both measures of leverage. The 85 th percentile for book leverage reaches a local maximum in 1989, a local minimum in 1994, and a local maximum in The 95 th percentile of book leverage reaches a local maximum in 1989, a local minimum in 1994, and a local maximum in And the 99 th percentile of book leverage reaches a local maximum in And the 99 th percentile of market leverage reaches a local minimum in Green, Martin, and Rich Page 3

6 Fourth, while book leverage appears far less volatile over time than market leverage, book and market leverage seem to generally move together over time. We identify the extreme-leverage firms in 1993, 2002, and 2006 and explore how these firms came to be so highly levered by examining the time series of their capital structure ratios leading up to the sample year. We chose these years since these are the years where the leverage of high-debt firms reached local minimums and maximums within the sample period we studied. Note that we get similar results when we examine firms with extremely high leverage in mid-trend years such as In addition, by starting with 1993 we had up to six years of historical data for each extreme leverage firm. BOOK VERSUS MARKET LEVERAGE Even though the 95 th percentile of book and market leverage metrics tend to move together over time, firms with extremely high market leverage do not necessarily have an extreme level of book leverage and vice versa. As a result, firms with extreme book leverage are not generally the same firms with extreme market leverage. Figure 2 provides an indication of the lack of consistency between book and market leverage ratios for firms classified as extremely levered. Specifically this figure contains the crosssectional distribution of book value leverage for firms classified as extremely highly leveraged using the market leverage metric, and vice versa. For example, in 2006 the group of firms ranked in the highest 5% of firms using book leverage contained 8 firms that had market leverage ratios that were less than 10%. Similarly, for the firms in the top 5% of firms based on their market leverage, there were 2 firms with 20% to 30% financial leverage measured using book leverage. 7 7 These differences occur even if we use a common sample of all firms for which we can calculate both book- and marketleverage ratios. With this common sample, we still end up with two sets of extreme leverage firms. For example, in 2006, we identify 246 firms with extreme book leverage and 246 firms with extreme market leverage. However, only 107 (43.5%) of the firms have both extreme book and extreme market leverage. We find similar disparities in 1993 and For the firms with extreme book leverage, the Spearman rank correlation between firms ranked on the basis of book and market leverage ranges between (2006) and (1993) and for firms with extreme market leverage, the Spearman rank correlation between firms ranked on the basis of their book and market leverage ranges between (2006) and (1993). Green, Martin, and Rich Page 4

7 Figure 1. Distribution of Financial Leverage between 1987 and 2009 This figure presents the 95 th, 85 th, 50 th, and 15 th leverage percentiles for 1987 through 2009 using both book and market leverage ratios calculated using equations (1) and (2). For example, the line labeled 95 th shows the level of leverage for a firm to be in the 95 th percentile of all firms in terms of their leverage. Panel a. Average Financial Leverage Percentiles Book Leverage th 95th 85th 50th 15th Market Leverage th 95th 85th 50th 15th Panel b. Number of firms for which book and market leverage can be calculated Year Book Market Year Book Market Green, Martin, and Rich Page 5

8 Figure 2. Distribution of Leverage Measures for Extreme Leverage Firms In Figure 2 we examine the consistency of book- and market-based leverage ratios in identifying extreme leverage firms. We show histograms of market leverage for firms with extreme book leverage and histograms of book leverage for firms with extreme market leverage. Some extreme market leverage firms have book leverage ratios that exceed 1 which can occur when a firm s negative book equity pushes book capital below the firm s debt. In extreme cases, book equity can become so negative that book capital becomes negative. Since the resulting negative book leverage ratios indicate high rather than low leverage, we show these firms to the right of > 100% book leverage. Book Leverage Market Leverage Distribution of Market Leverage for Firms with Highest (highest 5%) Book Leverage in 2006 Distribution of Book Leverage for Firms with Highest (highest 5%) Market Leverage in to 10% to 20% to 30% to 40% to 50% 50 to 60% to 70% to 80% 80 to 90% 0 90 to 100% to 10% 10 to 20% 20 to 30% 30 to 40% 6 40 to 50% to 60% to 70% to 80% to 90% to 100% > < 0% 100% Distribution of Market Leverage for Firms with Highest (highest 5%) Book Leverage in 2002 Distribution of Book Leverage for Firms with Highest (highest 5%) Market Leverage in to 10% 10 to 20% 20 to 30% 30 to 40% 40 to 50% 50 to 60% 60 to 70% 70 to 80% 80 to 90% 90 to 100% 0 to 10% 10 to 20% 20 to 30% 30 to 40% 40 to 50% 50 to 60% 60 to 70% 70 to 80% 80 to 90% 90 to 100% > < 0% 100% Distribution of Market Leverage for Firms with Highest (highest 5%) Book Leverage in Distribution of Book Leverage for Firms with Highest (highest 5%) Market Leverage in to 10% 10 to 20% 20 to 30% 30 to 40% 40 to 50% 50 to 60% 60 to 70% 70 to 80% 80 to 90% 90 to 100% 0 to 10% 10 to 20% 20 to 30% 30 to 40% 40 to 50% 50 to 60% 60 to 70% 70 to 80% 80 to 90% 90 to 100% > < 0% 100% Green, Martin, and Rich Page 6

9 EXTREME LEVERAGE AND INDUSTRY GROUP In Table 1, we find significant clustering of extreme leverage firms by industry group for all three sample years: 1993, 2002, and In the table, we group the sample into 49 industry portfolios based on the Fama-French industry definitions. 9 We then calculate the percentage of firms in each industry that have extreme leverage and rank the industries using this percentage. We find large differences across industries in the percent of firms that are extremely levered. In 2006, for example, the percent of firms with extreme book leverage runs from 0% to 33.3% and the percent of firms with extreme market leverage falls within the range of 0% to 15.4%. We can reject the independence of industry group and the percent extreme leverage firms at the 1% level for all years and for both measures of leverage using Pearson s Chi-squared statistic. 10 As a result, we conclude that extreme-leverage firms tend to cluster in certain industries. Next we examine whether the same industries have a high/low percentage of firms with extreme leverage in 1993, 2002, and When we rank industry leverage using book leverage, we can reject independence between 2006 and 2002 and between 2006 and 1993 at the 1% level and between 2002 and 1993 at the 5% level using Kendall s Tau rank correlation between the industry ranks across time (found in Panel b. of Table 1). If we compare ranks between years based on market values, we reject independence between ranks in 2006 versus 2002 and between 2002 versus 1993 at the 1% level, and between 2006 versus 1993 at the 10% level. These results suggest that firms with extremely high leverage tend to cluster in the same industries across time. Despite the relationship across time in the industry rank based on percent of extreme leverage firms, it is difficult to identify specific industries with a high percentage of extremely leveraged firms in all three of the years that we examine. For example, when we rank industries according to the percent of firms with extreme book leverage, only four (Fun, Coal, Telcm, and Trans) are ranked in the top third (ranks 1 through 14) in 1993, 2002, and When we rank industries according to the percent of firms with extreme market leverage, only six (Agric, Cnstr, Steel, Autos, PerSv, and Trans) are ranked in the top third in all three years. Similarly, it is difficult to identify specific industries with a low percentage of extremely levered firms in all three years. When we rank industries according to the percent of firms with extreme book leverage, only six (Clths, Drugs, Ships, Hardw, Chips, and 8 We selected the years 1993, 2002, and 2006 for further study based on observed patterns in the use of financial leverage by the entire cross-section of firms. Both 1993 and 2006 represent local maximums in average leverage while 2002 represents a local minimum in average leverage. Because the number of firms in our sample varies over time, the number of firms identified as extremely levered also varies over time. For example, in 1993 there were 333 extreme book leverage firms and 287 extreme market leverage firms; in 2002 there were 307 extreme book leverage firms and 324 extreme market leverage firms; and 2006 there were 265 extreme book leverage firms and 277 extreme market leverage firms. 9 Fama and French s industry listings can be found at: Note that several firms are not included in Table 1 because their SIC codes do not fall into any of the SIC ranges in Fama and French s definitions of 49 (or 48) industry portfolios. The statistical relationship between industry and extreme leverage is essentially unchanged if we include these firms from excluded industries. The excluded SIC codes include 900: Fishing, Hunting, and Trapping; 3990: Misc Manufacturing Industries; 9995: Nonoperating Establishments; and 9997: Industrial Conglomerates. This final group includes firms such as GE (2006 and 2002) and Berkshire Hathaway (2006). The number of firms excluded by year and SIC code follow: 2006: 900 = 1; 3990 = 17; 9995 = 94; 9997 = 9; 2002: 900 = 2; 3990 = 15; 9995 = 72; 9997 = : 900 = 0; 3990 = 19; 9995 = 11; 9997 = Only 43 of Fama and French s 49 industries show up in Table 2 since we have excluded utilities and financial firms from our analysis. 11 Since there are numerous ties in the ranks, we report only Kendall s tau-b to correct for these ties. Green, Martin, and Rich Page 7

10 LabEq) are ranked in the bottom third (ranks 30 through 43) in 1993, 2002, and When we rank industries according to the percent of firms with extreme market leverage, only four (MedEq, Drugs, Ships, and Hardw) rank in the bottom third in 1993, 2002, and These results suggest that the firms in certain industries tend to drift in and out of being extremely levered. 3 SOURCES OF CHANGE IN THE LEVERAGE FOR EXTREME LEVERAGE FIRMS In this section we address the basic question as to whether the observed changes in a firm s use of financial leverage over the years prior to our classifying it as having extreme leverage are due to managerial actions (managed) or circumstances outside of managerial control (unmanaged). First we document how the leverage of these firms changed in the years prior to our identifying them as extremely levered. Next we rule out changes in the sample over time as the primary driver of the rising leverage for the extreme leverage firms. We then isolate the managed and unmanaged components of changes in the financial leverage for the extreme-leverage firms and compare them to the sources of change in financial leverage for the remaining firms in our samples. EVOLUTION OF LEVERAGE RATIOS FOR EXTREME LEVERAGE FIRMS Figure 3 shows the time series of median and average leverage ratios for the portfolios of extreme leverage firms identified in 1993, 2002, and Several observations are readily apparent from the time series of debt ratios. First, the average and median leverage ratios (book and market) tend to rise prior to the year in which the firms were identified as extremely levered with much of the increased leverage occurring in the final few years before the firm is identified as extremely levered. Second, the patterns of rising leverage are similar in all three years despite a general decrease in critical leverage values prior to 1993 and 2006 and a general increase in critical leverage values prior to 2002 (see Figure 1). Thus the leverage patterns of firms with extreme leverage in 2002 mimicked the general trend in the use of financial leverage before 2002 but ran counter to the general trend prior to 1993 and Figure 4 shows the number of extreme-leverage firms for which we can calculate the book and market leverage ratios for each year prior to the firm s identification as extremely levered. Note that this number rises over time in a pattern that roughly mirrors the average and median leverage ratios. That is, the sample size increases in the years prior to the year in which the extreme leverage portfolios are formed. This pattern shows up in all years for both book and market measures of leverage. For example, for firms with extreme book leverage in 2002, the sample size grows from 85 in 1987 to 307 in Over the same period, the sample size for firms with extreme market leverage grows from 89 in 1987 to 324 in This suggests the possibility that the increasing leverage ratios may be driven by changes in sample composition over time. Green, Martin, and Rich Page 8

11 Table 1. Extreme Leverage and Industry Group Panel a. Percent of Extremely Levered Firms by Fama and French Industry Groups In Panel a. we calculate the percent of firms that exhibit extreme leverage for each of the 49 Fama-French industry groups for 1993, 2002, and We select these years for further study based on the observed patterns in the use of financial leverage by the entire cross-section of firms. Both 1993 and 2006 represent local minimums while 2002 represents a local maximum in average leverage ratios. When we classify firms by book leverage in 2006, 8.33% of the firms in the Agriculture (Agric) industry were classified as having extreme leverage. In addition, we rank the industries by the percentages of firms having the highest book and market measures of financial leverage. In 2006, for example, Agriculture had the 13 th highest percentage of firms in its industry group classified as extremely levered using book value. Note that in 2006, three industries had no firms with extreme book leverage, so all three are given the rank of 41. And in 2006, five industries had no firms with extreme market leverage and all five industries were given a rank of 39. At the bottom of the table we show Pearson s Chi-squared statistics and p- values for independence between the percent of extreme leverage firms across industry groups. Note that only 43 of the 49 industries show up in the table since we excluded utilities and financial firms from our sample. Book Market No. Name % Rank % Rank % Rank % Rank % Rank % Rank 1 Agric 8.33% % % % % % 12 2 Food 3.53% % % % % % 18 3 Soda 5.00% % % % % % 22 4 Beer 0.00% % % % % % 38 5 Smoke 33.33% % % % % % 1 6 Toys 2.33% % % % % % 38 7 Fun 11.11% % % % % % 15 8 Books 15.56% % % % % % 35 9 Hs hld 5.63% % % % % % Clths 1.32% % % % % % Hlth 3.23% % % % % % 9 12 MedEq 0.49% % % % % % Drugs 2.19% % % % % % Chems 8.80% % % % % % Rubbr 10.81% % % % % % Txtls 13.33% % % % % % 6 17 BldMt 2.38% % % % % % 8 18 Cns tr 4.69% % % % % % 4 19 Steel 5.48% % % % % % FabPr 0.00% % % % % % Mach 3.43% % % % % % ElcEq 4.26% % % % % % Autos 7.79% % % % % % Aero 9.09% % % % % % 2 25 Ships 0.00% % % % % % 38 Green, Martin, and Rich Page 9

12 Table 1 (continued). Extreme Leverage and Industry Group Book Market % Rank % Rank % Rank % Rank % Rank % Rank 26 Guns 20.00% % % % % % Gold 4.71% % % % % % Mines 1.10% % % % % % Coal 8.33% % % % % % Oil 4.40% % % % % % Telcm 12.50% % % % % % PerSv 8.06% % % % % % 7 34 Bus Sv 6.52% % % % % % Hardw 1.56% % % % % % Softw 2.17% % % % % % Chips 1.71% % % % % % LabEq 0.79% % % % % % Paper 13.85% % % % % % Boxes 26.67% % % % % % 3 41 Trans 9.09% % % % % % 5 42 Whls l 4.15% % % % % % Rtail 6.87% % % % % % Meals 7.45% % % % % % 10 Chi-s quared (42 d.f.) p-value <.001 <.001 <.001 <.001 <.001 <.001 Panel b. Industry Leverage Over Time We calculate Kendall s tau-b rank correlation to test the independence of industry ranks based on the results found in Panel a. across time. Book Market 2006 v v v v v v Kendall's tau-b Kendall's score SE of score p-value Green, Martin, and Rich Page 10

13 Debt to Market Capital Debt to Market Capital Debt to Book Capital Debt to Book Capital Figure 3. Evolution of Debt Ratios for Extreme Leverage over Time This figure shows how the mean and median book and market leverage ratios change over time for the samples of extreme leverage firms identified in 1993, 2002, and For each graph, year 0 is the year in which the firm is identified as extremely levered. Median Book Leverage Average Book Leverage Years Relative to Identification as Extremely Levered Years Relative to Identification as Extremely Levered Median Market Leverage Average Market Leverage Years Relative to Identification as Extremely Levered Years Relative to Identification as Extremely Levered Green, Martin, and Rich Page 11

14 Figure 4. Evolution of Sample Sizes for Extreme Leverage Firms over Time This figure shows the number of extreme leverage firms identified in 1993, 2002, and 2006 for which we can calculate book and market leverage in the years before we identify them as extremely levered. For each graph, year 0 is the year in which the firm is identified as extremely levered. Number of Firms in Sample: Book Leverage Number of Firms in Sample: Market Leverage Years Relative to Identification as Extremely Levered Years Relative to Identification as Extremely Levered Green, Martin, and Rich Page 12

15 To examine whether the patterns we have observed in the level of financial leverage found in Figure 3 stem from changes in the financial leverage of the sample firms or from changes in the sample composition, we decompose the change in the average leverage ratios as follows ( ) ( ) ( ) Total change Change in average leverage Change in average leverage Change in average leverage in average leverage due to firms dropping for firms in the sample due to new firms being from year t -1 to year t out of the sample in both year t added to the sample (CHANGE) after year t-1 and t-1 in year t (DROP) (SAMPLE) (ADD) ALL includes all firms in the sample and SAMPLE includes only firms found in both year t and t-1 samples. Thus DROP measures changes in the average leverage ratio due to firms dropping out of the sample after year t-1, SAMPLE measures the change in average leverage due to changes in the leverage of firms included in the sample in both years t-1 and t, and ADD measures the change in average leverage due to firms being added to the sample in year t. Note that firms may be added or dropped from the sample because of missing data in certain years or because Compustat adds a firm to its database. We report the results of this decomposition for book and market measures of leverage in Table 2. Green, Martin, and Rich Page 13

16 Table 2. Impact of Changing Sample Size on Leverage In this table we examine whether the increase in average leverage of extreme leverage firms prior to the year they are identified as extremely levered stems from changes in the sample or changes within firms in the sample. CHANGE refers to the total change in leverage (book or market) between the current and prior year. DROP measures the change in the average leverage ratio due to firms dropping out of the sample. SAMPLE is the change in the average leverage ratio due to changing leverage within firms included in the sample in both the current and prior year. And ADD measures the change in the average leverage ratio due to firms being added to the sample. For all variables, a positive number indicates that the variable is contributing to an increase in the average leverage ratio and a negative number indicates the variable is contributing to a decrease in the average leverage ratio. Firms with Extreme Leverage in 2006 Book Leverage Ratios Market Leverage Ratios Fiscal Year Average CHANGE DROP SAMPLE ADD Average CHANGE DROP SAMPLE ADD Firms with Extreme Leverage in 2002 Book Leverage Ratios Market Leverage Ratios Fiscal Year Average CHANGE DROP SAMPLE ADD Average CHANGE DROP SAMPLE ADD Green, Martin, and Rich Page 14

17 Table 2 (cont). Impact of Changing Sample Size on Leverage Firms with Extreme Leverage in 1993 Book Leverage Ratios Market Leverage Ratios Fiscal Year Average CHANGE DROP SAMPLE ADD Average CHANGE DROP SAMPLE ADD From Table 2 we conclude that changes in average leverage ratios stem primarily from leverage changes within firms in the sample rather than from changes in the makeup of the sample. For example, in the seven years prior to firms being identified as extremely levered in 2006, changes in the sample plays an important role only in the year 2000 and then only for book leverage (when high book leverage firms are added to the sample). In the seven years prior to being identified as extremely levered in 2002, changes in the sample of firms only plays a role in the change in average leverage in 1996 and then only for book leverage (when high book leverage firms are added to the sample and low book leverage firms drop out of the sample). And in the six years prior to firms being identified as extremely levered in 1993, changes in leverage plays a large role in changes in average leverage only in 1991 and only for book leverage. We conclude that when we examine the sources of changes in leverage in our extreme-leverage samples, we can focus on changes in leverage by firms within the sample and can ignore the impact of changes to the firms included in the sample. MANAGED VERSUS UNMANAGED CHANGES IN BOOK LEVERAGE We find that in most years, management actions drive changes in the book leverage for 65-75% of firms and changes in market leverage for 55%-65% of firms. However, we also find that as firms become extremely levered, managed changes in leverage play a diminishing role in explaining changes in a firm s financial leverage. The declining importance of managed changes in financial leverage occurs for both market and book leverage for firms identified as extremely levered in 1993, 2002, and And when we examine the relative importance of issuances/repurchases of debt and preferred stock and common stock, we find that debt issues/repurchases plays the dominant role in managed changes in capital structure. Common stock issues/repurchases plays a relatively small role while preferred stock issues/repurchases play almost no role. However, as firms become extremely levered, the role of debt issues/repurchases drops dramatically. This result holds for both measures of leverage for the 1993, 2002, and 2006 samples. ISOLATING MANAGED VERSUS UNMANAGED CHANGES IN BOOK LEVERAGE The focus of our analysis is changes in financial leverage over time where financial leverage is measured using book and market measures defined earlier by Equations (1) and (2). For each of these measures of financial leverage there is a corresponding definition of the change in financial leverage over time. The change in the Book Debt-to- Capital ratio for year t equals: (3) Green, Martin, and Rich Page 15

18 If the managers of a firm intentionally change the firm s book leverage by issuing or retiring securities, the changes in the firm s Book Capital will show up as changes in its short- and long-term debt (DLC + DLTT), preferred stock (PSTK), or in one or more common equity accounts that change when a firm issues or repurchases common stock (par value of common stock (CSTK t ), capital surplus/share premium reserve (CAPS t ), and/or treasury stock (TSTK t )). And changes in leverage due to profits or losses will show up as changes in retained earnings (RE). As a result, we can restate equation (3) as follows: ( ) (4) where: = change in debt in current liabilities + change in long-term debt from t to t-1. = change in book value of preferred stock from t-1 to t. = the change in par value of outstanding common shares from t-1 to t. = the change in paid-in-capital to the common stock account from t-1 to t. = change in retained earnings from t-1 to t. =the change in the firm s treasury stock account from t-1 to t. To isolate managed changes in book leverage, we measure how each firm s leverage ratio would have changed if the only changes in book leverage in equation (4) stemmed from the issuance or repurchase of securities. Since the only part of equation (4) unrelated to issuing or repurchasing securities is the change in retained earnings, we define managed changes in book leverage as the change in book leverage had the change in retained earnings been equal to zero. We then define unmanaged changes in book leverage as the difference between the total change in book leverage and the managed changes in book leverage. In Table 3 we classify each firm according to whether the firm s change in book leverage was primarily from managed (issuing or retiring securities) or unmanaged (unrelated to issuing or repurchasing securities) sources. We then report the proportion of firms where managed sources were the primary driver of the change in leverage. In Table 3 we see that the proportion of firms where managed sources drive leverage changes is almost always smaller for firms with extreme leverage than for firms that are not extremely levered. The only exception is in 1993 when we examine firms with extreme book leverage in However, these differences are not consistently significant at the 1% level until the last five (1993 and 2002 samples) or six (2006 sample) years prior to the year we identify the firms as extremely levered. 12 During this window, the percentage of firms where managed sources (issuing or repurchasing securities) drives changes in the firm s capital structure drops dramatically for extremely levered firms but remains essentially unchanged for firms that are not extremely levered. This time frame also matches up with the time during which the leverage of the extreme leverage firms rises dramatically. We therefore conclude that extreme book leverage stems primarily from factors unrelated to management decisions to issue/repurchase securities and stem instead from factors beyond the direct control of management. In Table 4 we classify each firm according to whether the largest driver of the change in the firm s book leverage was debt issues/repurchases, or preferred stock issues/repurchases, or common stock issues/repurchases, or due to unmanaged sources of changes in leverage. As in Table 3, we calculate the change in leverage due to 12 In the 2006 and 2002 extreme leverage samples, differences in the proportion of firms with extreme leverage are significant at the 5% and 10% level in several prior years but there does not appear to be any particular pattern to these differences. Green, Martin, and Rich Page 16

19 unmanaged sources as the difference between the actual change in leverage and the change in leverage had RE equaled zero in equation (4). Using the same basic idea, we calculate the change in leverage due to changes in debt as the difference between the actual change in the firm s leverage and the change in the firm s leverage had Debt equaled zero in equation (4). And we calculate the change in leverage from preferred stock issues/repurchases as the difference between the actual change in the firm s leverage and the change in the firm s leverage had PSTK equaled zero. Finally, we calculate the change in leverage from common stock issues/repurchases as the difference between the actual change in the firm s leverage and the change in the firm s leverage had ( CSTK + CAPS TSTK) equaled zero in equation (4). We then classify each firm according to whether the change in leverage from debt or the change in leverage from preferred or the change in leverage from common stock or the change in leverage due to unmanaged sources was the largest contributor to the actual change in the firm s leverage. In Appendix D we demonstrate our methodology with Goodyear Tire & Rubber. In the columns of Table 4 we show the proportion of firms where the primary driver of leverage changes was changes in debt or changes in preferred stock or changes in common equity, and we compare these proportions for firms that are extremely levered to the proportions for firms that are not extremely levered. In the table, we see that debt was the primary driver of changes in leverage for a smaller proportion of extremely levered firms than for non-extreme leverage firms in every year prior to the years we separated extreme from non-extreme leverage firms. This result holds when we seprate out extreme-leverage firms in 1993, 2002, and However, the differences were not consistently significant at the 1% level until five (1993 sample), six (2002 sample), or seven (2006 sample) years prior to our formation of extreme and non-extreme leverage portfolios. We also find that while the proportion of firms where debt is the primary driver of leverage changes was fairly stable across time for non-extreme leverage firms, the proportion drops dramatically for extreme leverage firms in the last few years prior to their identification as extremely levered. The proportion for extreme leverage firms drops from approximately 50% to 20.2% for the 2006 sample, from approximately 55% to 10.6% for the 2002 sample, and from approximately 50% to 18.1% for the 1993 sample. Similar to our findings in Table 3, the proportion of extreme leverage firms where debt was the primary driver collapses over the same period that the leverage of these firms rises dramatically. In Table 4 we also see that issues and repurchases of preferred and common stock are the primary drivers of leverage changes for relatively few firms regardless of leverage and regardless of time period examined. In addition, the differences are not consistently significant at the 10% level for any of the samples. We therefore conclude that issues and repurchases of equity play a relatively small role in capital structure changes and that the role of equity issues in capital structure changes does not differ between extreme and non-extreme leverage firms. Green, Martin, and Rich Page 17

20 Table 3. Managed or Unmanaged Changes in Book Leverage In Table 3 we classify each firm according to whether managed (issuing or retiring securities) or unmanaged (changes unrelated to issuing or repurchasing securities) sources were the primary driver of the change in the firm s book leverage each year. We then report the proportion of firms where managed sources were the primary driver of the change in leverage. We report first the proportion for firms that were extremely levered (the 5% of firms that were included in the extreme-leverage sample) and then for firms that were not extreme levered (the 95% of firms that were not included in the extreme-leverage sample). We test for equality of the two proportions using a two-group proportions test. For example, when we classify firms according to whether they were extremely levered in 2006, we find that managed sources were the largest driver of changes in financial leverage from 1991 to 1992 for 46.9% of the firms classified as extremely levered but for 72% of the firms that were not classified as extremely levered. This difference was significant at the 1% level Book Sample 2002 Book Sample 1993 Book Sample Average Average Average Fiscal Extreme Not Fiscal Extreme Not Fiscal Extreme Not Year Leverage Extreme Extreme Diff z-value Year Leverage Extreme Extreme Diff z-value Year Leverage Extreme Extreme Diff z-value *** ** *** *** *** ** *** *** * *** ** *** *** *** *** *** *** *** *** *** *** Green, Martin, and Rich Page 18