Wealth creation versus wealth redistributions in pure stock-for-stock mergers1 Carlos P. Maquieiraa, William L. Megginsonb, * and Lance Nailc a University of Chile, Santiago, Chile Terry College of Business, University of Georgia, Athens, GA 30602, U.S.A. c School of Business, University of Alabama-Birmingham, Birmingham, AL 35299, U.S.A. b Received 3 November 1995; revised 6 August 1997. Available online 27 July 1998. Abstract We examine wealth changes for all 1283 publicly traded debt and equity securities of firms involved in 260 pure stock-for-stock mergers from 1963 to 1996. We find no evidence that conglomerate stock-for-stock mergers create financial synergies or benefit bondholders at stockholders' expense. Instead, we document significant net synergistic gains in nonconglomerate mergers and generally insignificant net gains in conglomerate mergers. Conglomerate bidding-firm stockholders lose; all other securityholders at least break even. Convertible securityholders experience the largest gains, due mostly to their attached option values. Certain bond covenants are value-enhancing while leverage increases are value-reducing. Author Keywords: Conglomerate mergers; Stock-for-stock mergers; Bond valuation 1. Introduction Although finance theorists have examined corporate mergers from many different perspectives, most of these models predict one of two primary effects. Either mergers create net new wealth from operating or financial synergies, or they redistribute existing wealth between stakeholder classes. Though empirical support exists for most models, it has proven difficult to examine wealth creation and wealth transfers in a single analysis. Our study accomplishes this by examining the wealth changes of all 1283 publicly traded debt and equity securities of a matched sample 1 of 520 companies involved in 260 pure stock-for-stock mergers from January 1963 through March 1996. Pure stock-for-stock mergers offer an ideal opportunity to test for wealth creation and/or wealth redistributions because there are no cash outflows or asset changes. Thus, the sum of the market values of the merged firm's securities should equal the sum of the market values of the merging firms' securities (adjusted for overall market movements in financial asset values), unless the merger creates net wealth gains or losses. If net wealth is created by the capture of operating and/or financial synergies, there should be an increase in the summed market values of the combined firm's securities – and most or all of this ‘net synergistic gain' should accrue to stockholders, the firm's residual claimants. Ravenscraft and Scherer (1987), Bhagat et al. (1990), and Kaplan and Weisbach (1992) generally predict that operating synergies will be created only in mergers between firms in the same or related industries, and Healey et al. (1992) document particularly strong performance improvements for mergers involving firms with overlapping businesses. Models predicting the creation of financial synergies, such as those presented in Levy and Sarnat (1970), Lewellen (1971), Weston and Mansinghka (1971), Williamson (1975), Amihud and Lev (1981), Stapleton (1982), and Amihud et al. (1986) almost invariably assume that these synergies are to be found only in conglomerate mergers, or mergers between firms in different industries. Because conglomerate mergers, in general, neither reduce competition nor provide operating economies of scale, it is often assumed that these mergers do not yield any operating synergies or create product or factor market power, though they may increase a firm's debt capacity or create other types of financial benefits.2 Financial synergies can arise from: (1) reduction of default risk (and thus borrowing costs) by joining together firms with imperfectly correlated cash flow streams, (2) diversification of equity risk for stockholders, or (3) contracting efficiencies created by allowing managers to reduce their employment risk by creating larger, less risky firms. In contrast to synergistic wealth creation, wealth redistributions are usually expected to occur when a merger merely changes the relative riskiness of the cash flow streams of two or more securities. The theoretical models of Higgins and Schall (1975) and Galai and Masulis (1976) suggest that conglomerate mergers will lower equity values and raise bond values (leaving total firm value unchanged), while Shastri (1990) shows that these mergers can have many different effects, such as wealth redistributions from stockholders to bondholders (or vice versa) or within securityholder classes, depending upon the covariance between the returns of the merging firms.3 To date, however, only Eger (1983) has 2 documented wealth redistributions between bondholders and stockholders; studies by Kim and McConnell (1977), Asquith and Kim (1982), Dennis and McConnell (1986), and Travlos (1987) show no significant redistributions. Our empirical design, with its matched bidder and target firms and large sample of senior securities, allows for more direct testing of these wealth transfers. We employ a new valuation methodology that is a modification of standard event-study methodology. Our technique compares the actual post-merger values of different security classes with the values that would be predicted without a merger, and allows us to easily handle widely varying merger completion periods. Since we collect price data for all publicly traded nonconvertible preferred stock, nonconvertible bonds, and convertible security issues of the firms in our sample, we are also able to reexamine Dennis and McConnell's (1986) senior-security results using only stock-for-stock mergers (rather than a mixture of cash and stock transactions) and by using a much larger sample of matched bidder/target pairs. Also, our sample of 602 bonds and 130 preferred stocks represents the largest sample of senior securities employed in any published merger study. Though not the principal objective of this study, we also examine whether transactions that increase corporate ‘focus' create more value than transactions that decrease focus. As the term is defined and used in Comment and Jarrell (1995) and John and Ofek (1995), focus-increasing transactions are those that reduce the number of lines of business a firm operates in, and increasing focus is consistent with the popular notion of concentrating corporate activities on the firm's core competencies or principal lines of business.4 In the context of our study, conglomerate mergers tend to decrease corporate focus, while nonconglomerate mergers (i.e., those between firms in the same or related industries) tend to preserve or even increase it. By analyzing the differential wealth creation and redistribution effects of conglomerate and nonconglomerate mergers, we thus provide indirect evidence on the economic benefits of focus-preserving (or focus-increasing) versus focus-decreasing mergers. We find little evidence that conglomerate stock-for-stock mergers create financial synergies, or that bondholders in these mergers benefit more than either stockholders or bondholders in nonconglomerate mergers. Instead, we find that real operating synergies are created by stock-for-stock mergers, particularly in nonconglomerate mergers, and that the increases in market value resulting from these synergies are shared by almost all securityholder classes. Purely financial factors (i.e., leverage changes) influence merger-induced security value changes, but they are not dominant. On average, common stockholders experience 3 significant wealth increases which are greater for target than for bidder firm shareholders. However, bidding firm stockholders in nonconglomerate mergers experience wealth increases while bidding firm stockholders in conglomerate mergers suffer significant wealth losses. Virtually all classes of nonconvertible bondholders and preferred stockholders experience net wealth increases, but these are dwarfed by the positive unexpected wealth increases accruing to convertible bondholders in both merger categories as well as to convertible preferred stockholders in nonconglomerate mergers, which are driven by a relatively small fraction of in-the-money securities that experience very large unexpected wealth increases as a result of the merger. The value changes documented in this study are not caused by changes in systematic risk, since we find that the typical merged firm's actual market-model beta is almost exactly equal to its weighted-average predicted value. On the other hand, we find that the actual return variance of the typical merged firm's equity (and assets) is significantly higher than its predicted value. Regression analyses confirm that more total wealth is created in nonconglomerate than in conglomerate mergers, that the wealth gains accruing to bondholders decline after 1980, and that the wealth change to bondholders is directly related to the presence of two specific covenants in individual bond contracts. Wealth changes are also inversely related to the merger-induced change in firm leverage for most security classes. Although unsurprising – no securityholder would voluntarily exchange a less-levered claim for a more-levered one without compensation – this result has been heretofore undocumented. Our study is organized as follows. Section 2 discusses the sample selection criteria and presents summary statistics, and Section 3 presents our new valuation methodology. Section 4 presents the results of the univariate difference in means and medians tests, as well as supplemental multivariate regression tests for wealth redistributions and synergistic gain creation. Section 5 concludes. 2. Data sources and sample selection criteria Since this study's objective is to examine non-cash mergers that are self-contained financial systems (no cash outflow to investors), we select only pure stock-for-stock mergers and collect data on all outstanding public and private security classes of the matched pairs of merging firms. While we collect aggregate private debt and preferred stock measures, these are not included in our valuation measures since we lack market prices. However, our other bond and preferred stock estimates suggest that these issues also increased in value after the merger. To 4 be included in our sample, a merger must meet the following selection criteria. 1. The merger is initiated between January 1963 and December 1995 and is completed by March 1996. The announcement of the intention to merge in the Wall Street Journal Index (WSJI) is taken as the merger announcement date, while the merger effective date is obtained from the Center for Research in Security Prices (CRSP) database as the delisting date. 2. The merger is completed, so partial exchanges of stock are excluded from the sample. (Roughly 6% of the combinations examined are ‘clean-up' mergers, in which a controlling corporate parent purchases all the subsidiary shares not already owned.) 3. Both firms are listed on either the New York Stock Exchange or the American Stock Exchange, and daily stock returns are available on the CRSP tapes. 4. Only common or preferred stock is used as payment to the target firm's shareholders. We define the payment method as the one actually used, not necessarily the method mentioned in the initial merger announcement. We also use the actual exchange ratio in our computations of post-1977 acquired firm shareholder returns when this differs from the exchange ratio announced initially. Since we are not performing a merger announcement event study, our reliance on ex post data is not problematic, as the actual terms of the merger are always known two months after completion (the post-event measurement date we use). 5. There are no major contaminating events such as another merger announcement, major asset sale or purchase, or large security issue during the event period from two months before the merger announcement date through two months after the effective date. This event period also solves a measurement problem recently identified by Schwert (1996), who finds that pre-bid stock price runup can often lead to a material increase in the total cost of acquiring a given firm. This runup only begins to be material one and one-half months prior to the bid announcement date, however. 6. Capital structure, security issue, and line of business data are available from the appropriate Moody's Manual for the bidder, target, and combined firms. The final sample contains 260 mergers, involving 520 firms. In 18 cases, the same firm is the acquirer in two mergers, and there are two cases (Colgate–Palmolive and Pepsico) of a firm being the bidder in three 5 completed mergers. We also eliminate roughly ten mergers with multiple bidders. In addition to common stock, we also collect prices and issue terms (maturity, coupon rate, amount outstanding, sinking fund payments due, etc.) for all publicly traded securites of the merging and merged firms. This process yields usable information on 78 nonconvertible preferred stock, 83 convertible preferred stock (including 31 new securities issued as a means of payment), 535 nonconvertible bond, and 67 convertible bond issues. Combined with the common stocks of the merging firms, this yields a total sample of 1283 individual security issues. Concentrating on successful pure stock-for-stock mergers that are uncontaminated by contemporaneous security issues or other major transactions clearly costs us in three important ways. First, cash payments are more common than stock payments in U.S. mergers and acquisitions, and this disparity has varied over time (see Comment and Schwert, 1995). Second, research indicates that stock-for-stock mergers have systematically lower offer premiums for target firm stockholders, significantly negative abnormal returns for acquiring firm stockholders, and lower net synergistic gains created.5 Third, focusing on successful stock-for-stock mergers, which generally will be friendly as well, eliminates many contested and resisted (hostile) transactions from our study (see Huang and Walkling, 1987; Jennings and Mazzeo, 1993; Cotter and Zenner, 1994). We trade off these drawbacks for a self-contained system that encompasses all the financial effects of a specific, and important, type of merger. 2.1. Classification of conglomerate and nonconglomerate mergers Since we predict different wealth effects in conglomerate and nonconglomerate mergers, we need a consistent method of classifying our transactions into one of these two groups. Unfortunately, firms were not required to disclose the detailed line of business data needed to compute revenue-based Herfindahl measures of focus until 1977, and our sample includes mergers from as early as 1963. We therefore classify mergers based on a direct examination of the primary line of business listing for each company in the appropriate annual edition of the Moody's Manual. If the merging firms have the same primary line of business, the merger is classified as nonconglomerate; if the two firms have different primary lines, the merger is classified as conglomerate. Although there is no ‘consensus' definition of conglomerate versus nonconglomerate mergers in the finance or economics literature, we feel our classification scheme is defensible, and a very similar technique (matching by two-digit SIC codes from CRSP) has been employed by other 6 researchers, including Berger and Ofek (1995), Mann and Sicherman (1991), Sicherman and Pettway (1987), and Smith (1990). Tighter definitions of nonconglomerate mergers, or of ‘industry matches' for control samples are employed by Eckbo (1983) and Eckbo (1985), Demsetz and Lehn (1985), Morck et al. (1990) and Lang and Stulz (1994). Since, for our purposes, it best to use a definition of conglomerate merger that embodies a combination between truly dissimilar firms, we opt for a definition that involves firms in different industrial groups (two-digit codes) rather than firms that simply produce different product lines (three- or four-digit codes). While this classification seems very imprecise, Megginson et al. (1997) document that over 85% of the post-1977 mergers in their sample that are classified as conglomerate (focus-decreasing) or nonconglomerate (focus-preserving or focus-increasing) using this simple SIC code/line of business screen would have been classified the same way using the more sophisticated revenue-based Herfindahl measure. Kahle and Walkling (1996) find substantial differences between the SIC codes reported in CRSP and Compustat for the same firm. To examine the potential impact of this industrial code ambiguity on our study, we compare our merger classifications with those that would have been assigned based solely on CRSP, solely on Compustat, and on the use of both databases. We find that our classifications would have resulted in exact matches with CRSP 81.2% of the time and 79.7% of the time with Compustat. In 68.0% of the cases, all three classification schemes yield the same classification. Our method agrees with CRSP in classifying mergers as conglomerate 92.6% of the time, with Compustat 82.5% of the time; and all three methods agree 74.6% of the time. Naturally, the nonconglomerate classification agreement percentages are much lower, but this simply adds a conservative bias to our analyses. The time period distribution of the full sample, as well as of the conglomerate and nonconglomerate merger subsamples, is presented in Table 1. Of the 260 stock exchange mergers in our sample, 135 are classified as conglomerate and 125 as nonconglomerate. Consistent with Comment and Jarrell (1995), the fraction of conglomerate mergers declines monotonically over time. Table 1. Distribution of sample mergers by type and by year of announcement.Distribution of pure stock-for-stock mergers in the sample, by year of initial announcement, over the period 1963–1995. The sample mergers are grouped as conglomerate or nonconglomerate mergers based on whether the combining firms have the same primary lines of business 7 listings in the appropriate annual Moody's Industrial, Financial, or Utilities Manual. Table 2. Relative size and leverage ratios for sample companies involved in stock-for-stock mergers over the period 1963–1996 (announcement period ends in 1995, but several mergers are completed in 1996). The relative sizes of bidder and target firms are computed as the book value of the target firm's total assets divided by the book value of the bidding firm's total assets in the reporting period immediately prior to the announcement date of the merger. The leverage 8 ratios of the bidder, target, and combined firms are computed as the book value of total debt divided by the book value of total assets. The first section of Table 2describes the relative size of the bidder and target firms in the full sample. On average (median), the target firm's pre-merger equity market value is equal to 23.1% (18.2%) of that of the bidder. The relative size of targets to bidders is much higher in nonconglomerate mergers than in conglomerate combinations (27.1% mean and 22.8% median versus 19.4% and 15.6% mean and median). These differences may reflect the ‘portfolio' acquisition strategy of conglomerate bidders, as well as the practical fact that there may be only a limited number of (relatively large) firms within an industry for a nonconglomerate acquirer to bid for. The second part of Table 2 shows that the firms in the full sample have mean and median book value leverage ratios of 41.2%, with leverage defined as the ratio of total debt to total assets. There is remarkably little variation in leverage, either by type of merger or by whether a firm is a bidder or a target. Target firms in nonconglomerate mergers have an average (median) leverage ratio of 47.2% (44.0%), and acquiring firms in both types of mergers have average leverage ratios of 41.2% (39.8% and 41.6% for conglomerate and nonconglomerate mergers, respectively), but average leverage ratios for all other subsamples fall in the range of 41.8% to 44.4%. 3. Methodology for computing security valuation changes Our objectives in this study require us to determine the wealth effect of mergers on all of the publicly traded securities of the combining firms. Unfortunately, the length of time it takes a merger to be completed ranges from a low of 11 months to a high of 31 months for the mergers in our sample. We thus develop a method of computing valuation changes that allows adjustment both for overall market movements and for changes in the number of securities outstanding over any appropriate holding period. Our principal methodology for determining merger-related value changes is to generate predicted post-merger valuations for all of the securities of the merged firm. These are based on: 1. The pre-merger valuations of each outstanding security issue of the merging firms, 9 2. overall market movements in matching asset prices during the event period (t−2 months before merger announcement to t+2 months after the effective date) between the merger announcement and effective dates, 3. any cash distributions made to securityholders during the event period, and 4. any changes in the number of outstanding securities in a class due to conversions, calls, sinking fund payments, or open-market repurchases. Once predicted values are computed for each security issue, we compute a valuation prediction error (VPE) for that security by subtracting the predicted post-merger value from the actual post-merger value. While this technique is just a variant of event-study methodology (which other researchers have used to measure event-related senior security valuation changes), we are unaware of any other study that uses precisely the same procedure. The VPE is computed in essentially the same manner for all security classes, and can thus easily be aggregated by class, firm, or merger. The prediction methodology used for each security is described below. 3.1. Generating predicted values for merged-firm common and preferred stock Of all the security classes for which predictions are generated, merged-firm common stock is unique in that at least one (and sometimes both) of the pre-merger common stocks outstanding invariably ceases to exist after the merger effective date. Depending upon the accounting treatment of the combination, either the acquiring firm's common stock continues to trade after the merger or a new class of common stock of the merged firm begins trading. In either case, the single common stock class resulting from the merger should be equal to the market-adjusted summed value of the merging firms' common stocks, plus any wealth transfers from other securityholders or net synergistic gains. (From 1977 on, we can reliably obtain exchange ratios for all of our sample mergers, allowing us to compute separate VPEs for the common stock of both the bidder and target firms, as well as for the merged firm's equity; we report these individual results in Table 3 as well.) 10 Table 3. Valuation prediction errors for conglomerate and nonconglomerate stock-for-stock mergers over the period 1963–1996 (announcement period ends in 1995, but several mergers are completed in 1996). Valuation prediction errors (VPEs) are computed as the percentage difference from predicted market value for equity and debt securities from two months before the merger announcement date through two months after the effective date of merger for individual securities. Predicted market values are computed based on overall market movements in the same classes of securities over the measurement period ( t−2 months to t+2 months), and t-statistics (mean and percent positive) and Wilcoxon statistics (median) are presented in parentheses. This table presents mean and median VPEs, and percent positive statistics for 260 pure stock-for-stock mergers, classified according to whether the combination is a conglomerate or a nonconglomerate merger. This classification is made based on whether the combining firms have the same primary lines of business listings in the appropriate annual Moody's Industrial, Financial, or Utilities Manual. To adjust for market movements over our study period, we compute an index value of the CRSP value-weighted return (including all distributions) over the period beginning two months before the merger announcement through two months after the merger effective date. We then calculate the predicted merged-firm equity value, (Pred MVCS Comb)i, as the sum of the beta-adjusted product of this index number and the equity value of each the merging firms: where (MVCS Bidder)i is the pre-merger market value of bidder firm common stock in merger i, (MVCS Target)i the pre-merger market value of target firm common stock in merger i, (CSIndex)i the cum-dividend geometric return on the CRSP value-weighted index from two months before merger announcement to two months after merger completion, B the market-model beta for bidder firm, estimated over months t−62 to t−2, and T the market-model beta for target firm, estimated over months t−62 to t−2. The actual market value of the combined firm's common stock in merger i, (MVCS Comb)i, is computed as the price per share of the merged firm times the number of shares outstanding, plus the dividends per share paid on the merging firms' stocks during the security holding period. We then compute a valuation prediction error for the common stock in merger i, (VPE CS Comb)i. This is our measure of the synergistic gain (or loss) and/or wealth redistribution accruing to the common stockholders of the merging firm as a result of the merger. 11 Beginning with preferred stocks, we find that multiple issues of a given security class (preferred stock or debt) will often be traded before and after the merger, though the number of securities outstanding might well change due to redemption or partial conversion. We therefore compute predicted valuations and VPEs for each individual security issue as well as for all of the issues of that class involved in a merger. The basic methodology for generating these results for preferred stock is very similar to that for common stock. Instead of the CRSP index, we adjust for market movements using the S&P Preferred Stock Index. We also use this process to compute predicted values for the convertible preferred stocks in our sample. In effect, we are correcting only for changes in the value of the fixed claim portion of the convertible preferred shares. We will assume that any dramatic change in the common equity portion (the conversion privilege) of the convertible preferred share is caused by the merger itself. A similar assumption is made for convertible bonds. To avoid double-counting equity gains, we subtract from the combined firm's stock market value the value of new common stock created by the conversion of previously issued convertible preferred stock or convertible bonds. We also document that 31 mergers involve the distribution of newly created convertible preferred stock to target firm shareholders as payment for their shares. In some cases, target stockholders are given their choice of common or preferred stock, while in others only preferred stock is used as payment. We combine the market value of this newly created stock with the outstanding stock of the merged firm to calculate a measure of the actual market value of common equity that is truly comparable to the predicted value. 3.2. Generating predicted values for convertible and nonconvertible bonds For each individual bond issue k in the sample, we find the U.S. Treasury bond outstanding at that time that most closely matches bond k in maturity and coupon rate. We then subtract the matched T-bond's yield to maturity (YTM) from bond k's YTM to compute a pre-merger yield spread for bond k, (Pre-Merger Spread)k. To operationalize this measure, we assume that the pre-merger yield spread will remain unchanged over time, even though the level and shape of the yield curve can change. This assumption allows the computation of an expected YTM value for each bond k (Exp YTM Corp Bond)k, after the merger effective date based on the observed value of its matching Treasury bond at that time. With an expected YTM, the expected post-merger price of bond k can be computed using the terms of the issue (remaining maturity, coupon rate) and a standard internal rate of return bond pricing 12 formula. The valuation prediction error for each bond k (VPE Corp Bond)k, is then computed as before, after adjusting for bond redemptions, conversions, and sinking fund payments. A full description of the VPE methodology for each security class is presented in Nail (1996). 4. Empirical results Table 3 presents VPE results for the securities in our sample categorized by bidder and target and, more importantly, by whether the merger is classified as conglomerate or nonconglomerate. Panel A presents commmon and preferred stock results, while Panel B examines convertible and nonconvertible bonds as well as our measure of net synergistic gains. The first significant result is that the common stockholders in nonconglomerate mergers experience average VPEs that are economically and statistically significantly higher (8.58% versus 3.28%) than do shareholders in conglomerate mergers. The median difference (8.55% versus 1.98%) is even larger. Further, a significant 66.4% of the nonconglomerate mergers result in positive overall VPEs, while an insignificant 56.3% of the conglomerate mergers yield positive results. These findings are consistent with theoretical models predicting that new wealth will be created only in nonconglomerate mergers through the creation of operating synergies, and that most of these wealth increases will accrue to common stockholders. For comparison purposes, Bradley et al. (1988) document that successful tender offers increase the combined value of the target and acquiring firm by an average of 7.4% over the period 1963–1984, and similar figures are provided by Lang et al. (1991), Eckbo (1992), and Berkovitch and Narayanan (1993). For the 102 mergers occurring after 1976, VPEs can be computed for common stockholders of bidder and target firms individually. These results strongly suggest that bidding firm stockholders benefit in nonconglomerate mergers, while bidding firm stockholders in conglomerate mergers are harmed. The differences in mean (6.14% versus −4.79%) and median (4.64% versus −7.36%) VPEs, and in the fraction of positive VPEs (61.8% versus 36.2%), between the 55 nonconglomerate acquiring firms and the 47 conglomerate acquiring firms are highly significant, both statistically and economically. We also find that the difference between combined stockholder VPEs becomes much more pronounced after 1977. Stockholders in the 55 nonconglomerate mergers in this subsample experience sigificantly positive mean (9.86%), median (9.84%), and percent positive (70.9%) VPEs, while stockholders in the 47 conglomerate mergers from this period earn insignificant 0.12% and −2.37% mean and median VPEs, and only 46.8% of these mergers are value-increasing. Since 13 these mergers can be classified quite precisely as either focus-preserving or focus-increasing (nonconglomerate) and focus-decreasing (conglomerate), these results provide strong – and as yet unique – evidence supporting the corporate focus hypothesis. The wealth transfers from acquiring firm to target firm stockholders in conglomerate mergers represent the only true wealth redistributions we document. Furthermore, only nonconglomerate mergers create significant net wealth for the combined firm's stockholders. Although most of this new wealth accrues to target firm stockholders, bidder firm stockholders also experience net wealth increases. This result is consistent with a gains-sharing explanation for nonconglomerate mergers. Based on the findings in Lang et al. (1989), we also examine whether market-to-book ratios are different for bidders in conglomerate versus nonconglomerate mergers in our sample. We find that the 1.32 mean (0.98 median) market-to-book ratio (defined as the market value of equity plus the book value of debt and preferred stock divided by the book value of total assets) of our nonconglomerate acquirers is significantly greater than the 1.16 mean (0.89 median) market-to-book ratio of our conglomerate acquirers at the 1% significance level (t=2.94). We also find that acquiring-firm senior securityholders in nonconglomerate mergers almost always experience net wealth increases that are not only significant in their own right but are also generally greater than for their counterparts in conglomerate mergers. The sole exception to this pattern is observed in convertible bonds, where the holders of these securities experience extremely large, positive VPEs, which are actually slightly larger for conglomerate than for nonconglomerate mergers. The mean (median) VPE for convertible bondholders in conglomerate mergers is 20.92% (9.48%) versus 17.51% (9.04%) for nonconglomerate mergers. More generally, we find that convertible securityholders experience surprisingly large, positive returns in stock-for-stock mergers. With the exception of convertible preferred stockholders in conglomerate mergers, who experience insignificant mean (median) returns of 2.15% (−0.20%), the mean (median) VPEs for holders of convertible securities range from 12.45% (4.00%) for bonds to 24.30% (14.18%) for preferred stocks. The final two rows of Table 3 present the summed VPEs for all of the securities of the merged firms, which is our measure of the net synergistic gains created by stock-for-stock mergers. Given the individual security results, it is not surprising that we document significant mean (6.91%) and median (6.79%) net synergistic gains for nonconglomerate mergers which are significantly larger than the insignificant mean (3.91%) and 14 median (1.25%) gains for conglomerate mergers. Furthermore, fewer than half (48.2%) of conglomerate mergers create net synergistic gains for securityholders, while almost two-thirds (64.0%) of nonconglomerate mergers are wealth-creating.6 4.1. Why are convertible securityholder returns so large? In an attempt to determine the causes of the surprisingly large, positive wealth effects for both convertible bonds and preferred stock, we perform three supplemental analyses. First, we examine what actually happens to these senior securities after the merger to see if they are called or if their holders voluntarily redeem most of the securities outstanding. Second, we compare the conversion terms of bond and preferred stock issues to determine whether the high returns to convertible bondholders are the result of more favorable average conversion ratios. Finally, we analyze the two convertible security classes to determine which variables are driving these unusually large returns. We find that 91.3% of the convertible preferred stock outstanding at the time of the merger remains outstanding six months after the merger completion date (8.7% has been either called or voluntarily converted) and 79.1% remains outstanding a year after the merger has been completed. Far less of the convertible debt remains outstanding six months (77.4%) and 12 months (58.8%) after the merger effective date. These differences could result from more favorable conversion terms enjoyed by convertible bondholders relative to convertible preferred stockholders, and we examine this possibility. We compare the ratio of the common stock values of the convertible bonds to their market values (the value of shares represented by full bond conversion divided by the current value of the bonds themselves) to the same ratio computed for convertible preferred stocks. Securities that have a ratio value greater than or equal to one are classified as in the money (since their conversion options have positive intrinsic value), and those that have a ratio value of less than one are classified as out of the money. We find that the average ratio is 1.13 (median=0.88) for bonds and 0.79 (median=0.92) for preferred stock, and that 38.2% of the bonds are in the money prior to the merger but only 22.6% of the preferred stocks are in the money. The average VPE is 24.6% for in the money bonds and 18.1% for in the money preferred stocks. Thus, the large difference between convertible security VPEs is driven by a greater proportion of outlier bonds that are in the money due to more favorable conversion terms relative to preferred stockholders. 15 Using our conversion option framework also allows us to help explain the seemingly anomalous convertible bond VPEs, as well as to present an interesting result worthy of further research. While all other security classes clearly exhibit superiority in wealth creation for nonconglomerate mergers, convertible bond VPEs are actually higher in conglomerate mergers. Given the previously reported results that common stock VPEs of nonconglomerate mergers are significantly higher than those for conglomerate mergers, one would expect the conversion option values and VPEs of nonconglomerate convertible bonds to also be significantly higher. However, our subsample of the common stock VPEs associated with convertible securities is not representative of our full sample. While the average common stock VPE is 8.58% and 3.28% for nonconglomerate and conglomerate mergers, respectively, in the full sample of mergers, these averages are reversed in the subsample of mergers with convertible bonds, with values of 2.4% and 11.6% for nonconglomerate and conglomerate mergers, respectively. 4.2. Merger-induced beta and variance changes The VPE results documented above suggest that stock-for-stock mergers create value through the anticipated creation of operating synergies, rather than through the creation of financial synergies or for some other purely ‘financial' reason. Before accepting this analysis, however, we test two other possible explanations for the documented value increases: changes in equity betas and changes in equity (or asset) return variances. While there is little reason to expect that the act of merging will cause the systematic or idiosyncratic risk of the resulting firm to differ fundamentally from a weighted average of the combining firms' betas and return variances (and covariances), we concede that there is much about how assets are priced in real capital markets that we do not yet understand. If stock-for-stock mergers result in reduced systematic risk for the combined firm, this could translate into a valuation increase for both stockholders and bondholders. We examine whether the actual market-model beta of the resulting firm, m, is significantly different from its expected value, M, computed as a weighted average of the betas of the bidder, B, and target firm, T. As in Bhagat et al. (1987) and Kaplan and Stein (1990), we use the equity beta calculated from a market model regression to proxy for a stock's systematic risk. We find little evidence that changes in beta are systematically influencing the study results, though they do appear to add noise to our estimates. The 0.024 unit difference between the predicted beta of the merged firm, M=1.004, and its actual beta, m=0.980, is statistically insignificant. An examination of the distribution of 16 actual versus expected betas, however, indicates that the insignificant average beta change is at least partly caused by a ‘washing out' of extreme positive and negative changes. While 46% of actual merged-firm betas are within 0.15 units of their expected value, almost 20% are more than 0.35 units from their predicted values. We also examine whether the actual stock return variance of the merged firm, , is equal to its predicted value, , calculated as the sum of the weighted variances and covariances between bidder firm return variance, , and target firm return variance, . The methodology described in Ohlson and Penman (1985) and Cox and Rubinstein (1985) is used to calculate these variances using historical (daily) stock price data. We find clear evidence that the actual (daily) stock return variance of the merged firm is significantly higher than predicted for the combined sample of all mergers and for the conglomerate and nonconglomerate subsamples. The actual daily stock return variance (standard deviation, M) of the merged firms in the full sample is 0.00042 ( M=0.0203 or 2.03% per day), while the predicted variance is 0.00034 ( M=1.83%). The difference between the predicted and actual values is significant at the 0.001% level (t=−4.91) and fully 67.0% of the mergers have actual variances higher than their predicted values. Since a higher-than-expected stock return variance might explain our surprisingly high equity and convertible security wealth changes, this variable is included in our regression analyses of VPEs, which are discussed below. When we repeat these analyses using asset betas, rather than equity betas, our results are qualitatively identical. An asset variance is derived from an equity variance by assuming that all of a firm's outstanding debt is riskless and that all of the variability of the return on that firm's assets loads on the stock price. In effect, the equity variance is multiplied by the equity-to-capital ratio. This obviously yields a variance estimate that is lower than the ‘true' variance of returns on the firm's assets, but since the computed stock return variance yields the highest possible variance estimate, computing both measures provides polar estimates of actual variance. 4.3. Regression analysis of VPE results So far we have documented far fewer true wealth redistributions than predicted by the conglomerate merger/financial synergy literature, as well as a surprising number of wealth gains and anomalously large VPEs 17 for convertible securities. We now examine these results further using regression analysis, and provide two sets of analyses for each security class. First, the combined sample of all common stocks is examined, then bidder and target stocks are examined separately. For preferred stocks and corporate bonds, we examine the full sample of each class together, with a dummy variable indicating when an issue is a convertible security, and then examine nonconvertible and convertible preferred stocks and bonds separately. We first describe three factors common to all four regressions, and then describe factors that are expected to affect only individual security issues. The first variable is a dummy variable (Nonconglom) used to proxy for the type of merger, where one stands for nonconglomerate and zero stands for conglomerate merger. The second (Pre-Williams) and third (Post-1980) variables are temporal dummy variables, motivated by the findings in Bradley, et al. (1988), that proxy for whether the merger occurred prior to the implementation date of the Williams Act in July 1968 or after the Reagan Administration came to power in January 1981. Variables four through six are used in the regression analyses of individual common stock, preferred stock, and bond issues. Variable four (Target) is a dummy variable taking a value of one if the security in question is issued by a target firm and zero if it is a bidding firm security issue. As before, we also perform separate regressions on bidder and target firm common stock VPEs (but not preferred stock or bond), since there is overwhelming evidence that target firm stockholders earn significantly higher merger-related returns than do bidder firm stockholders. The fifth variable, change in leverage (ΔLeverage), measures the absolute percentage point change in market value leverage that the securityholders of a given company are expected to experience. Since no rational investor would willingly trade a less-heavily-indebted financial claim for a more-indebted one, a merger-induced increase in leverage should decrease VPEs for all security classes. To see how this is computed,consider a bidding firm with a 50% debt-to-total capital ratio (with all values in market value terms) and total capital of $1 billion, that acquires a firm with total capital of $500 million and a debt-to-capital ratio of 20%. The resulting combined firm will have a leverage ratio of 40%, so the securities of the bidding firm will experience a ten percentage point decrease (from 50% to 40%) in leverage, while the target firm's securities will experience a 20 percentage point increase. The sixth variable is the predicted change in asset return variance (Δvariance) computed as discussed in Section 4.2. Since the asset return measure is the one suggested by Shastri (1990), we use it rather than the 18 stock return variance featured earlier. Shastri predicts that an increase in variance will cause common stock to increase in value, but the same variance increase will cause nonconvertible preferred stock and bondholders to suffer a decline in the value of their securities. Predictions regarding convertible securities are ambiguous since they combine elements of both equity and debt. The seventh and last variable refers only to preferred stock and bond issues. It is a dummy variable (Convertible) that takes on a value of one if the bond or preferred issue in question is convertible into common stock, and takes a value of zero if the issue is nonconvertible. As mentioned, we also perform separate regression analyses on convertible and nonconvertible preferred stocks and bonds, since our VPE results are so dramatically different (higher) for convertible securities. The combined-security regressions we use to estimate the cross-sectional determinants of the VPEs for common stock, preferred stock, corporate bonds, and net synergistic gain are summarized in Eq. 2, Eq. 3 and Eq. 4 below. Common stock issue j, in merger i: (2) Preferred stock or corporate bond issue j, in merger i: (3) Net synergistic gains (summed values of all merged firm securities) for merger i: (4) The supplemental regressions are performed separately for bidder and target firm common stocks and for convertible and nonconvertible preferred stocks and bonds, and are identical to Eq. 2 and (3) except that the appropriate dummy variables are not included. Note that both preferred stock and bond issues are examined using Eq. 3. Note also that the dummy variables in these equations are designed so that the base case, where all the dummy variables have zero values, corresponds to a conglomerate merger from the period July 1968 through December 1980 involving a nonconvertible security issued by a bidding firm. Table 4 presents the results of these regression analyses. 19 Table 4. Regression analyses of valuation prediction errors for individual security classes and for net synergistic gains of the merged firm securities for sample firms involved in stock-for-stock mergers over the period 1963–1996. Ordinary least-squares regressions in which the dependent variable is the valuation prediction error computed for each security of the merging firms and the independent variables are a measure of the expected variance of the merged firm's stock return–computed using the value-weighted average of the pre-merger variances and return correlation coefficients of the merging firms' stocks ( Variance); the expected change in firm leverage (ΔLeverage), measured as the absolute percentage change in market value leverage that the securityholders of a given firm are expected to experience; and a series of dummy variables proxying for whether the merger is a nonconglomerate combination (Nonconglomerate), whether the merger occurred prior to the implementation date of the Williams Act in July 1968 (Pre-Williams), whether the merger occurred after the Reagan Administration took power in January 1981 (Post-1980), whether the security being examined is from a target firm (Target), and whether the security being examined is convertible into common stock (Convertible). Coefficient t-statistics from the regressions are presented in parentheses. 4.3.1. Common stock regression results The first line of Table 4 analyzes VPEs for the combined bidder and target common stocks for the entire sample of 260 mergers. The only significant variable is the nonconglomerate merger dummy (t=2.41), and the coefficient value implies that common stockholders earn an average 6.2 percentage points extra return if they are involved in a nonconglomerate rather than a conglomerate merger. The second set of common stock results, analyzing individual VPEs for bidder and target firm stocks for 102 mergers occurring after 1976, indicates that the only significant variable in the combined sample is the target dummy variable. Being the recipient of a takeover bid is associated with a 38.1% VPE, and this coefficient is highly significant (t=4.26). The overall F-value (11.58) of the regression is also significant, and the adjusted R2 indicates that this equation explains almost 21% of the variation in stock VPEs. When bidder and target stockholder VPEs are examined separately, the explanatory power and overall significance of each regression equation is much reduced, and the only significant coefficient in the target firm 20 regression is the intercept ( =0.436, t=3.42). However, we document both a significant negative (t=−1.74) relationship between the change in leverage and acquiring firm stock VPE and a significant positive (t=2.01) relationship between acquirer VPE and merger type. On average, a one percentage point increase in leverage for the acquiring firm's stockholders (in the merged firm) is associated with a 0.51 percentage point reduction in VPE. Being involved in a nonconglomerate merger yields a VPE that is 7.98 percentage points higher than that earned by acquiring firm stockholders in conglomerate mergers. These results suggest that nonconglomerate mergers create valuable operating synergies and that increasing leverage decreases financial asset values. On the other hand, the fact that the variance change measure is also insignificant in both bidder and target regressions suggests that while stock return variance is larger than expected, this increase is not priced. The post-1980 variable is also insignificant for both targets and bidders, and neither of the time period dummy variables is significant for the combined-stock sample, so no significant difference in equity returns is detected based on the regulatory environment within which the merger occurs. This contrasts with Bradley et al. (1988) finding that the division of gains between bidder and target changed in favor of the target firm after the Reagan Administration came to power in January 1981. Although the exact reasons for the differences between our results and theirs cannot be determined, it is possible that tender offers (and the inherently competitive nature of open-market cash takeover bids) were affected much more fundamentally by the regulatory changes they study than were (inherently friendly) stock-for-stock mergers. 4.3.2. Preferred stock regression results When all 135 convertible and nonconvertible preferred stock issues are analyzed together, two variables are significant. First, the nonconglomerate variable again implies that preferred stockholders in these mergers earn a significant 18.1 percentage point premium (t=2.23) over their counterparts in conglomerate combinations. Second, being fortunate enough to be a preferred stockholder in a target (rather than a bidder) provides a 24.2 percentage point premium (t=1.90). Examining nonconvertible and convertible preferred stock issues separately yields a striking result. While none of the nonconvertible regression variables are significant (nor is the overall F-value), the nonconglomerate variable is both large and highly significant in the convertible preferred stock regression. Being involved in a nonconglomerate versus a conglomerate merger earns a convertible preferred shareholder a 20.6 percentage point premium (t=2.91), which is over three times the premium common stockholders receive and over fifteen times what nonconvertible 21 bondholders receive for being involved in nonconglomerate mergers. The absolute and relative size of this premium for convertible preferred stockholders (but not bondholders) in nonconglomerate mergers remains a puzzle. 4.3.3. Corporate bond regression results The overall pattern of bond regression results is similar to that for common stock, in that when both convertible and nonconvertible securities are analyzed together an impressively large adjusted R2 (0.1871) and F-value (18.35) is obtained, while the separate regressions of convertible and nonconvertible bonds yield lower overall explanatory power but a greater number of (and more interesting) significant explanatory variables. The combined-security regression indicates, first, that convertible bondholders receive a VPE that is 18.42 percentage points higher (t=10.44) than that earned by the average nonconvertible bondholder and, second, that each one percentage point increase in leverage is associated with a 0.37% reduction in the VPE that bondholders receive (t=−4.06).7 Analyzing nonconvertible bonds separately confirms that increasing leverage reduces bondholder VPEs (coefficient=−0.122, t=−2.11), but also yields the intriguing result that bond returns are 1.86 percentage points higher (t=2.59) in nonconglomerate than in conglomerate mergers and that bondholder VPEs declined by 1.92 percentage points (t=−2.06) after 1980. The nonconglomerate and leverage results strengthen the conclusion that nonconglomerate mergers create more value than do conglomerate combinations and that merger-induced leverage increases harm all classes of securityholders. The significant negative coefficient (−0.019, t=−2.08) on the post-1980 dummy variable suggests that the takeover market changed during the 1980s in a way that was harmful to nonconvertible bondholders. As with the equity results, it is unclear why nonconvertible bondholders in stock-for-stock mergers experienced the same post-1980 diminution in merger-related returns that Bradley et al. (1988) document for stockholders in firms making cash tender offers, particularly since we cannot identify who might have benefited at bondholders' expense during this period (all of the other post-1980 coefficients in our security return regressions are insignificant). As a sensitivity check, we also examine whether some of these bondholder returns can be explained by differences in bond covenants, and to our surprise find some evidence that this is occurring. Table 5 presents a description of the frequency with which nine types of bond covenants are 22 observed in a randomly selected subsample of 125 bonds (18 convertible, 107 nonconvertible). This table also presents the frequency of call provisions, sinking funds, and specific pledges of security as well as the results of a regression analysis of the impact the presence of these covenants and other features on bondholder VPEs.8 Table 5. The effect of bond covenants and other features on bondholder valuation prediction errors for sample firms involved in stock-for-stock mergers over the period 1963–1996. For a randomly selected subsample of 107 nonconvertible and 18 convertible bonds, this table provides the frequency with which certain covenants and other features are observed in the bond descriptions presented in the appropriate annual Moody's Manuals, and also provides a summary of the regression coefficients and p-values from a regression in which bondholder valuation prediction errors are the dependent variable and the independent variables include dummy variables indicating the presence of these covenants and other features. The following categories of bond covenants are documented: restrictions on the ability of firms to pay dividends ( Dividend restrictions); covenants specifying minimum asset coverage ratios (Asset coverage); restrictions on indebted-firm management's ability to execute asset sale and leaseback agreements ( Sale and leaseback restrictions); covenants specifying maximum levels of additional debt that can be assumed (Total debt restrictions); covenants mandating minimum acceptable levels of financial liquidity (Liquidity); covenants mandating that any newly issued debt be of the same or junior status (Seniority restrictions); restrictions on the firm's ability to engage in mergers or other corporate control activities ( Merger restrictions); covenants specifying minimum or maximum capital investment spending ( Capital investment restrictions); and restrictions on management's ability to sell firm assets (Asset sale restrictions). The other features examined are the presence of a call option held by the issuing firm (Call provision), the presence of a mandated sinking fund (Sinking fund provision), and whether the bond is secured by specific pledged assets (Secured debt). Only three types of bond covenants are observed in at least 20% of the subsample of bonds: dividend restrictions (46.4%), minimum asset coverage ratios (20.0%), and restrictions on asset sale and leaseback agreements (20.0%). On the other hand, call provisions and sinking funds are pervasive, showing up in 88.8% and 58.4% of all bond issues, respectively. Only two of the covenants (asset coverage and merger restrictions) significantly affect bondholder VPEs. On average, the presence of a 23 minimum asset coverage ratio covenant increases bondholder returns by 5.1 percentage points, and the presence of a covenant restricting mergers and other corporate control events increases bondholder VPEs by 6.8 percentage points. The presence of call features, sinking funds, or specific collateral pledges does not significantly affect the returns earned by bondholders in stock-for-stock mergers. These results are actually quite satisfying, because they suggest that the two covenants that arguably offer the best protection against merger-induced wealth redistributions from bondholders to stockholders are significantly positively related to bondholder returns. This conclusion is strengthened by the results of a supplemental bondholder VPE regression that includes a dummy variable proxying for the presence of two or more covenants, rather than dummy variables for the individual covenants. This dummy variable is insignificant, indicating that it is the protection offered by the asset coverage and merger restriction variables themselves that is valued, rather than the mere presence of multiple covenants. 4.3.4. Net synergistic gains regression results The final line of Table 4 details regression results for the combined change in value for all of the securities of the 520 firms involved in the full sample of 260 stock-for-stock mergers over the period 1963–96. Two key findings emerge from this analysis. First, the time period dummy variables are neither statistically nor economically significant. Total gain creation does not vary according to whether a merger occurred prior to the Williams Act of 1968, after the Reagan Administration came to power in January 1981, or during the 1969–1980 period, mirroring the results documented in Bradley, Desai, and Kim. Second, the significant coefficient on the nonconglomerate merger dummy variable indicates that these mergers typically yield a 7.1 percentage point higher net wealth creation than do conglomerate mergers. 5. Summary and conclusions We examine wealth changes for 1283 publicly traded debt and equity securities of firms involved in 260 pure stock-for-stock mergers from 1963 to 1996. Using a new methodology for computing market-adjusted security valuation changes, we find no evidence that conglomerate stock-for-stock mergers create financial synergies or that these mergers benefit bondholders more than stockholders. Instead, we find economically and statistically significant net synergistic gains for the securityholders of firms involved in nonconglomerate mergers, but generally insignificant net gains for securityholders in conglomerate mergers. Apart from bidding firm stockholders in conglomerate mergers, all major classes of debt and 24 equity securityholders of both bidders and targets either break even or experience significant wealth gains. Target firm shareholders always experience net wealth gains, as do bidding firm stockholders in nonconglomerate mergers, clearly suggesting that, on average, acquiring firm managers who execute nonconglomerate mergers are acting in their shareholders' best interests, while those who launch conglomerate mergers most definitely are not. Further, almost all subsamples of nonconvertible preferred stockholders and bondholders experience significant wealth increases. Surprisingly, we find that holders of convertible securities earn extremely large and significant wealth increases as a result of stock-for-stock mergers. This result is not driven solely by the gains accruing to target firm convertible securityholders (who could capture the higher returns earned by target firm common shareholders by converting their securities), but is instead often higher for bidding firm convertible securityholders. The high average convertible securityholder returns result from a relatively small fraction of in-the-money convertibles that experience very large (greater than 30%) wealth increases as a result of the merger. The larger average wealth gains for convertible bonds (19.65%) than preferred stocks (9.59%) is driven by a greater proprtion of outlier bonds that are in the money due to more favorable conversion terms relative to preferred stockholders. The overall and security-specific wealth changes we document cannot be fully explained either by changes in systematic risk (since beta appears to add linearly) or by merger-induced changes in asset return variance. The actual post-merger variance is indeed higher than predicted, but this higher variance is not reflected in a higher VPE for the merging firms' securityholders. Instead, regression analysis documents that stock-for-stock mergers benefit securityholders of the more-levered firm at the expense of the less-levered firm's securityholders, and also confirms that nonconglomerate mergers create more total wealth than do conglomerate mergers. Clearly, nonconglomerate stock-for-stock mergers create significant net new wealth for securityholders, conglomerate mergers at best do not destroy value, and there appear to be fewer actual inter-securityholder wealth redistributions than our theories would suggest. 25