The Lack of Diversification from Diversifying Mergers SAKYA SARKAR Tulane University 1 ABSTRACT Using a structural model, I estimate coinsurance the reduction in default risk from cash-flow pooling from US mergers. Diversifying mergers produce similar coinsurance as related mergers and no higher coinsurance than if firms were to merge randomly; this is because diversifying acquirers are often large and diversified, leaving minimal scope for further risk reduction. In contrast, for mergers involving small or highly levered targets, coinsurance is significant. Mergers with significant coinsurance are more likely to occur; such mergers are associated with increase in debt; and when such mergers are announced, stock price increases, which suggests that stockholders benefit from coinsurance. 1 Visiting Assistant Professor, Tulane University, A.B. Freeman School of Business, New Orleans, LA 70118. http://business.tulane.edu/faculty/facinfo.php?recordid=ssarkar1@tulane.edu. Comments welcome; please email ssarkar1@tulane.edu. I thank John Matsusaka, Kenneth Ahern, Daniel Carvalho, Derek Horstmeyer, Lee Cerling, Ted Fee, Scott Joslin, Maria Ogneva, Oguzhan Ozbas, Garrett Swanburg, and Yongxiang Wang for their valuable suggestions. I acknowledge financial support from University of Southern California and Tulane University. 1
1. Introduction Most large American corporations are diversified; diversifying mergers, between unrelated firms with minimal operational synergies, constitute a significant proportion of M&A activity. In theory, diversifying mergers should produce diversification, that is reduce risk by pooling (less than perfectly correlated) cash-flows; a reduction in the risk of distress may then benefit firms by reducing the deadweight costs of distress and by enhancing tax shields this idea is known as coinsurance. A long literature, using reduced form estimation, documents coinsurance to be a major benefit from diversifying mergers; other literature studies coinsurance in business groups. 2 In contrast to this previous reduced-form literature that relies on cash-flow correlation as a proxy for coinsurance, this paper is the first to use a structural model to estimate the coinsurance from mergers. In its estimation, the structural model uses not only correlation but also leverage, size, and volatility. Essentially a tradeoff cost model of capital structure in continuous time, the structural model estimates coinsurance using pre-merger stock prices. By relying solely on premerger stock prices in its estimation, the structural model bypasses confounding factors, such as operational synergy and change in capital structure, that also affect value gain from the merger. 3 2 Recent papers include Penas and Unal (2004), Billet, King and Mauer (2004), Leland (2007), Kuppuswamy and Vilalonga (2010), Banal-Estañol, Ottaviani and Winton (2013), Hann, Ogneva and Ozbas (2013), Duchin (2010), Fulghieri and Sevilir (2011), Martos-Vila, Rhodes-Kropf and Harford (2013), Matvos and Seru (2014), Khanna and Yafeh (2005), Fisman and Wang (2010), Jia, Shi and Wang (2013) and Luciano and Nicadano (2014). 3 The model, by dint of its structure, estimates the value gain attributable solely to coinsurance. To assess whether the structural model is reliable, I undertake several robustness exercises. First, in Appendix A, I document that the structural model predicts defaults well out-of-sample. Second, I estimate coinsurance using an alternative Merton s (1974) approach. Finally, in Appendix C, I estimate coinsurance using an accounting approach: Altman (1997) Z Scores. To the extent that these approaches differ in their assumptions, the estimated coinsurance benefits are different; yet, these approaches agree so far as the main results in this paper are concerned. 2
Using this structural approach, I arrive at two novel results: (1) diversifying mergers, in most cases, produce insignificant coinsurance, contrasting the previous literature; and (2) mergers involving small or highly levered targets produce significant coinsurance. The first result (that coinsurance is insignificant for diversifying mergers) is best conveyed by the observation that the estimated value gain from coinsurance average only 2.3% of deal value for diversifying mergers, compared to 2.2% of deal value for related mergers. For the typical merger, coinsurance is barely sufficient to offset the fees paid to investment banks that average 1.2% of the deal value (Hunter and Jagtiani (2003)). Not only is the coinsurance insignificant for diversifying mergers in an absolute sense, it is also insignificant in a relative sense. As a relative benchmark for coinsurance, I construct a sample of hypothetical random mergers, by randomly drawing firms from COMPUSTAT and pairing them and then using the structural model to compute counterfactual coinsurance benefits for those random mergers. The coinsurance from these random pairings exceeds the coinsurance from actual mergers: even if firms were to choose their partners randomly, those random mergers would produce higher coinsurance than is observed in actual (diversifying) mergers. Diversifying mergers produce insignificant coinsurance because in my sample US mergers with deal size above $10 million, for which data is available from SDC, CRSP, and COMPUSTAT most firms that undertake diversifying acquisitions are large and already diversified (compared to random firms in the COMPUSTAT or compared to certain acquirers in related mergers); even before the merger, their default risk is low, with minimal scope for further risk reduction through coinsurance. 3
The finding of insignificant coinsurance from diversifying mergers contrasts with a large body of literature. However, this finding can be reconciled with the previous literature: the structural model in this paper, in accordance with the theoretical literature on coinsurance (Leland (2007)), estimates coinsurance solely by the reduction in default risk, whereas coinsurance benefits may also stem from other channels, such as the avoidance of costly external finance (Hann, Ogneva and Ozbas (2013)). To the extent that these other channels, hitherto neglected by the theoretical literature on coinsurance, are important, the structural model may underestimate coinsurance, and the coinsurance from reduction in these non-distress risks may be substantial. The other main result of the paper, that coinsurance is significant for certain mergers, is supported by four coherent findings. The first finding: for the top decile of mergers, which contains 184 mergers, coinsurance is estimated by the structural model to be significant, exceeding 5.6% of deal value. These are mergers between smaller firms or mergers that involve highly levered targets, at high risk of distress without the merger, being acquired by healthy firms. Targets benefit from coinsurance more than acquirers coinsurance average 2.4% of the pre-merger firm value for targets, compared to just 0.4% for acquirers because targets are at much higher risk of distress without the merger: mean 10 year default probability is 10.6% for targets and 4.3% for acquirers. Targets are riskier because they are significantly smaller: median target size is $140 million, compared to $1.1 billion for acquirers. And because targets are more indebted: mean target debt is 20.1% while acquirer debt is 17.8%. This substantially positive coinsurance benefit for targets coheres with the empirical regularity that announcement returns are largely positive for targets and small (often negative) 4
for acquirers (Andrade, Mitchell and Stafford (2001)). A significant proportion of the gains for targets from the merger may then be originating from coinsurance. The second finding on the significance of coinsurance: mergers with significant coinsurance are more likely to occur. If the structural model is correct in identifying high-coinsurance mergers, that is, given two firms, it can tell whether their merger will create significant value from coinsurance, then, ceteris paribus, such high-coinsurance mergers should be more likely to occur. Indeed, high-coinsurance predicts a merger incidence; a logistic regression of merger incidence for the combined sample of 1,884 mergers and 1,884 randomly paired firms on a dummy for the merger having high-coinsurance yields a positive coefficient, significant both economically and statistically. The third supporting finding: when coinsurance is significant, there is a capital structure adjustment around the merger, which is consistent with coinsurance increasing the debt capacity. In mergers with significant coinsurance, the method of payment is less likely to be all stock; a stock payment would reduce leverage, whereas in mergers with significant coinsurance, firms should increase their leverage to take advantage of their increased debt capacity. The effect is economically significant: a one standard-deviation increase in coinsurance (expressed as a fraction of the combined pre-merger firm value) reduces the probability of an all stock payment from 24% to 17%. The result survives the common controls for the method of payment (Martin (1996)). The final supporting finding: when coinsurance is significant, it benefits stockholders. Although much empirical evidence suggests that coinsurance benefits bondholders (Billet, King and Mauer (2004), Penas and Unal (2004)), there is minimal empirical evidence that coinsurance 5
benefits stockholders. Instead, the literature asserts that when coinsurance from a merger is high, it hurts stockholders because the reduction in asset volatility due to coinsurance reduces the option value of the equity, which transfers value from stockholders to bondholders (Galai and Masulis (1976), Higgins and Schall (1975), Mansi and Reeb (2002)). So ingrained is the notion that stockholders suffer from coinsurance that MBA textbooks label coinsurance as a deleterious reason to merge (Ross, Westerfield, and Jaffe (2008)). In contrast to this view, I document that when a merger is announced, the cumulative abnormal announcement return (target and acquirer combined) increases by 0.89% for every 1% value gain from coinsurance estimated by the model. The finding is statistically significant after including standard controls and robust across sub-samples. This finding suggests that when coinsurance benefits are significant, it can be a valid motive for mergers from the perspective of stockholder wealth maximization. In addition to presenting these new empirical results on coinsurance, this paper extends a series of recent theoretical papers on coinsurance. Banal-Estanol, Ottaviani and Winton (2014) model coinsurance using an optimal contracting framework, whereas Luciano and Nicodono (2014) model coinsurance within business groups using a two-period trade-off model. Leland (2007) models financial synergies from mergers using a two-period tradeoff cost framework. This paper, while retaining Leland s trade-off cost framework, adopts a continuous time approach (Leland (1994)). A theory paper, Leland (2007) calibrates the financial synergies for certain assumed values of parameters; in contrast, this paper estimates the default risk and the coinsurance for an actual sample of mergers. 6
The remainder of this paper proceeds as follows: Section 2 presents a structural model of coinsurance; Section 3 describes the data; Section 4 discusses the first main result: the (lack of) coinsurance from diversifying mergers; Section 5 discusses the second main result: when coinsurance is significant; Section 6 concludes the paper. 2. A structural model for estimating coinsurance The structural model in this paper extends Leland (2007) while retaining its broad framework; therefore, it may be worth sketching Leland s model and highlighting key differences with this paper. In Leland (2007), there are two firms, 1 and 2, with values, V 1 (t) and V 2 (t), that merge at time t. If the value of the merged firm is V 12 (t), the value gain from the merger is = V 12 (t) V 1 (t) V 2 (t). (1) To estimate, which he calls financial synergies, Leland assumes a two-period trade-off model: there is one random future cash-flow for each firm, CF 1 (t) and CF 2 (t), whose joint distribution is bivariate normal. There are tax benefits of using debt, and there is a deadweight cost of distress; firms choose their debt so as to maximize firm value. Leland assumes that there are no operational synergies; the cash flow of the merged firm is the sum of the cash flows of the individual cash flows. Leland then estimates coinsurance,, which he calls financial synergies, through equation (1); he computes firm values, V 12 (t), V 1 (t), V 2 (t) numerically using Monte-Carlo simulations for different assumed values of the parameters: cash-flow correlation, mean and variance of cash flows, tax-rates, and default costs. Although this paper retains the basic trade-off cost framework from Leland (2007) and estimates coinsurance using Equation (1), there are important differences: (i) Instead of a two- 7
period model, this paper develops a continuous-time model; each firm s assets, X 1 (t) and X 2 (t), evolve according to a correlated Geometric Brownian motion. (ii) Instead of assuming standard parameter values, the parameters of this model are estimated from pre-merger stock prices (which is one of the advantages of a continuous-time approach). (iii) Instead of assuming firms are levered optimally before and after the merger, debt in this model is obtained from the data. (iv) Instead of assuming that operational synergies are zero, this model allows for operational synergies. 2.1 Valuing the standalone (pre-merger) firms I estimate pre-merger firm values in accordance with Leland (1994). Each pre-merger firm i has assets valued X i (t) that evolves following with two correlated geometric Brownian motions: dx i (t) X i (t) = ((μ i δ)dt + σ i dw i(t)), i = 1,2 σ i > 0, δ 0. (2) The correlation coefficient of the two processes is constant: ρ [ 1,1]. Also constant are the expected growth rate of assets, μ i, its volatility, σ i, and the payout rate to equity, δ. The firm issues debt that pays a constant coupon, C i, per instant of time, as long as the firm is solvent. The firm becomes insolvent when the value of its assets falls below a default boundary, K i, at time T i = Min{t 0; X i (t) = K i }. When insolvent, the firm defaults on the debt; a fixed proportion, α, of the value of assets is lost due to the deadweight costs and the debt holders obtain the remaining (1 α)k i. This is the cost of debt. 4 4 The concept of default in Leland (1994) differs from Lewellen (1971). Leland s model is in continuous time, while Lewellen s is a one period model. In Leland s model, default occurs if the market value of assets falls below the default barrier, whereas in Lewellen s model, default occurs if cash flow falls short of the current portion of debt payable. 8
The benefit of debt stems from its tax-deductibility. When solvent, a tax benefit of τc i accrues to the firm per instant of time, with τ being the tax rate. This benefit is lost when the firm defaults. Thus, the net benefit to leverage is the tax benefit of debt minus the expected cost of distress. The levered firm s value, V(X i (t)), is the sum of all future expected cash flows, discounted at the risk free rate, r, with expectations computed under the risk neutral probability measure. V(X i (t)) = X i (t) + E [ T i τ=t τc i e rτ dτ] E [e rt i αk i ]. (3) Equation (3) demonstrates that the levered firm s value, V(X i (t)), is the sum of three components: the value of the firm s assets, X i (t), that is unaffected by the leverage; the tax shield, E [ T i τ=t τc i e rτ dτ], which is the expected value of all future tax deductions, discounted to the current time; and the cost of debt, E [e rt i αk i ], which is the expected deadweight loss on default, discounted to the current time. Computing these expectations, the value of the firm simplifies: V(X i (t)) = X i (t) + τ C i (1 r [X i (t) ] γ i ) αki [ X i (t) ] γ i, K (4) i K i where, γ i = {(r δ 1 σ 2 i 2 ) + [(r δ 1 σ 2 i 2 ) 2 + 2σ 2 2 i r] } 2 /σi. 1 The default barrier, K i, is endogenous: equity holders choose when to default, such that equity value is maximized. The endogenous default barrier is given by K i = (1 τ)c i. This value r+0.5σ 2 i of K i can be substituted in equation (4), to obtain the firm value. Equation (4) presents the value of the firm as a function of the coupon rate C i, where the coupon rate depends on leverage. If we further assume that firms choose their leverage 9
optimally, we can choose the value of C i that maximizes firm value. However, empirically, the actual leverage departs significantly from the optimal leverage predicted by Leland s model, or for that matter the optimal leverage from most tradeoff models (Fama and French (2002), Leary and Roberts (2005)). As an extreme example, in my sample of 1,884 mergers, as many as 242 targets and 118 acquirers have zero debt, which is significantly different from the optimal leverage predicted by any tradeoff model (Strebulaev and Yang (2013)). Moreover, the observed leverage differs from the optimal leverage predicted by the model over a long time; it is often stable over 20 years (Lemmon, Roberts and Zender (2006)). 5 Hence, instead of assuming that firms are optimally levered, I obtain the actual debt before the merger from COMPUSTAT. The product of the observed debt and the risk free rate provides the coupon rate C i (Elkamhi, Ericsson and Parsons (2012)). Now, to value the pre-merger firm using equation (4), we need the structural parameters: σ 1,σ 2, and ρ. These structural parameters are estimated from the past stock price information, in accordance with the procedure described in sub-section D. 2.2. Valuing the merged firm To value the merged firm, I assume the following. First, the value of the assets of the merged firm, X 12 (t), is the sum of the pre-merger asset values: X 12 (t) = X 1 (t) + X 2 (t). This assumption of additivity abstracts operational synergies (Lewellen (1971), Leland (2007)). Applying Ito s Lemma, the dynamics of X 12 (t) are given by dx 12 (t) X 12 (t) = [(μ 1 δ)s(t) + (μ 2 δ)(1 s(t))]dt + σ 1 s(t)dw 1 (t) + (1 s(t))σ 2 dw 2 (t), (5) 5 A class of tradeoff models incorporate the notion of dynamic inaction: firms optimally rebalance infrequently, only when the benefits from rebalancing offset the cost of debt issuance (Fischer, Henkel and Zechner (1989)); these models can explain the leverage profitability puzzle (Danis, Rettl and Whited (2014)). 10
where s(t) = X 1 (t) is stochastic. Equation (5) demonstrates that in contrast to the pre- X 1 (t)+x 2 (t) merger standalone firms, the dynamics of the asset value are not a geometric Brownian motion: obtaining a closed form solution for firm value, as in Leland (1994), is no longer feasible. In Appendix C, I relax this additivity assumption and consider operational synergies; for a broad range of operational synergies, the effect of operational synergies on coinsurance is minimal. Second, in accordance with Lewellen (1971), I assume that the debt of the merged firm is given by the sum of the pre-merger debts: in the context of the model, the merged firm s coupon rate, C 12 = C 1 + C 2. An alternative to this assumption is to compute the optimal leverage for the merged firm, to trade off the cost of debt against its tax benefit (Leland (2007)). However, in the previous subsection, I obtained the pre-merger leverage from COMPUSTAT, instead of assuming that the pre-merger firm is optimally levered. Therefore, assuming the post-merger firm to be optimally levered, while the pre-merger firms are from their optimal leverage, will be comparing apples with oranges. 6 Empirically, for most firms, the post-merger leverage is comparable to the pre-merger combined leverage; Appendix F documents that the debt/asset one year after a merger (mean, 22%) is close to the pre-merger leverage (mean 26%). Third, I assume the post-merger default boundary, K 12, to be the sum of the pre-merger default boundaries: K 12 = K 1 + K 2. In contrast to the pre-merger firms, the equity holders do not choose the default boundary optimally; instead, they inherit the default barrier prior to the 6 In theory, firms should increase their leverage following the merger, taking advantage of the cheaper cost of debt due to coinsurance; the extent to which leverage should increase depends on the cost of adjustment. 11
merger. Finally, I assume that the proportion of firm value lost in distress, α, the tax rate, τ, and the payout rate, δ, are the same for both firms and that they are unchanged by the merger. 7 The value of the merged firm, V(X 12 (t)), is then the sum of all expected future cash flows, discounted to the present at the risk free rate, computed under the risk neutral measure. V(X 12 (t)) = X 12 (t) + E [ T 12 τ=t where T 12 is time to default for the merged firm. τc 12 e rτ dτ] E [e rt i αk 12 ]. (6) Once the values of the firms before and after the merger are estimated, I estimate the coinsurance benefits,, by inserting these values from equations (3) and (6) into equation (1): = ( τc 1 r αk 1)E (e rt 1 e rt 12) + ( τc 2 r αk 2)E (e rt 1 e rt 12). (7) The term, ( τc 1 r αk 1)E (e rt 1 e rt 12), represents the change in net benefit from leverage due to the merger for the first firm; similarly the second term is the benefit for the second firm. 2.4. Parameter Estimation I estimate the structural parameters volatilities, σ 1, σ 2, and the correlation between the two Brownian motions, ρ from pre-merger stock prices. The closed-form expression of the equity price for the pre-merger firms, E i (t), is provided by Leland (1994): E i (t) = X i (t) + (1 τ) C i r ((1 τ) C i r K i) [ X i (t) K i ] γ i, (8) 7 In general, the cost of distress, the tax rate, or the payout rate is neither homogenous nor unchanged by the merger. Glover (2013) demonstrates that there is significant cross-sectional variation in cost of distress. Zhu and Singhal (2011) document that the proportion of value lost in bankruptcy is higher for diversified firms. Tax rates too differ across firms. Although it is straight-forward to obtain pre-merger tax rates, computing post-merger tax rate is less straightforward. Similarly, there is significant cross-sectional variation in payout rates. However, when two firms with different payout rates merge, it is not clear ex-ante what the post-merger payout rate will be. 12
where γ i = {(r δ 1 σ 2 i 2 ) + [(r δ 1 σ 2 i 2 ) 2 + 2σ 2 2 i r] } 2 C /σi, and, K i = (1 τ) i r+ 1 2 σ 2. i 1 In equation (8), the only unknown are the asset value, X i (t) and the volatility of asset returns, σ i. The risk free rate, r, is known; the stock price, E i (t), can be easily obtained for firms with traded common stock; the tax rate, τ, is assumed to be 35%, and the payout rate, δ, is assumed to be 1%, in accordance with Leland (1994). The unknowns in equation (8), volatility and asset values, are obtained by solving the equation numerically. To solve the equation, as an initial estimate of the volatility, σ i, I use the standard deviation of the daily stock returns, computed over a 252 day window that ends 42 days before the merger announcement. This end date is selected because previous literature has found a run-up in target prices prior to a merger announcement, beginning approximately 42 days before the merger announcement (Schwert (1996)). The initial estimate of volatility, σ i 1, is inserted into equation (8), (the superscript 1 denoting that this is the first iteration); and then, equation (8) is solved numerically to estimate the asset value X i 1 (t), for each of these 252 days. From these asset values, X i 1 (t), the asset returns are computed: R i 1 (t) = X i 1 (t) X i 1 (t 1), 1. The standard deviation of these asset returns over the 252-day window provides an updated estimate for the volatility, σ i 2. The process is repeated and new estimates of volatility are obtained until the absolute difference between the estimates of volatility from two successive iterations is less than 10 4. That is, σ i N σ i N 1 < 10 4. Usually, convergence occurs within five iterations. When convergence occurs, the estimate from the last iteration, σ i N, is used as the final estimate for volatility: σ i = σ i N. 13
In the course of estimating the volatility, I also estimated a series of asset values, X N i (t). These estimated asset values, from the final iteration step, are used to generate the asset returns for each firm. The sample correlation between the asset returns of firms 1 and 2 provides the estimate of the correlation coefficient between the two Brownian motions, ρ. Once the structural parameters are estimated, the coinsurance benefits are estimated using equation (7), by evaluating the expectations through Monte-Carlo simulations. 2.5 Monte Carlo Simulations For each merger in the sample, I simulate the asset values for the target and the acquirer recursively: X i (t + t) = X i (t)ex p{(r 1 σ 2 i 2 ) Δt + σ i tw i (t)}. (9) The initial asset value for the simulations, X i (t), is the asset value 42 days before the merger announcement. I simulate daily asset values, so that Δt = 1, and upto t = 252,000. The shocks for the target and acquirer, W i (t) are drawn from a bivariate normal distribution, i.i.d. across time: W 1(t) W 2 (t)) ~ N ( 0 0, [1 ρ ρ 1 ] ). I simulate 1,000 paths for each firm. To ensure that the 1,000 paths are sufficient, I compute the difference between the standalone firm values computed through simulations and the firm values computed theoretically (using equation (7)). The absolute value of the difference is less than 0.5%, which suggests that 1,000 paths are sufficient. Once these asset values are estimated, the coinsurance,, is computed as the average coinsurance benefit across the 1,000 paths: = ( τc 1 αk r 2) 1 j j N N (e rt 1 e rt 12 j=1 ) + ( τc 2 αk r 2) 1 j j N N (e rt 2 e rt 12 j=1 ), (10) 14
where T j i is the time to default of firm i, for the simulation path j. The estimated coinsurance benefit depends on the choice of α, which is the proportion of value lost on default. In this paper, I choose α = 16.5%, the midpoint of the estimates of Andrade and Kaplan (1998). I check the sensitivity of the coinsurance estimates to this assumption by choosing other values of α and by considering the fixed costs of distress. 3. Data I compute coinsurance benefits for two different samples. One is composed of mergers that occurred in the United States between 1981 and 2013. The other is a sample of randomly paired firms, also corresponding to the same period. The data on mergers are obtained from the Securities Data Corporation s (SDC) U.S. Mergers and Acquisitions Database. The sample consists of mergers in the United States that satisfy the following criteria: (1) the announcement date was between January 1, 1981, and December 31, 2013; (2) the deal size was above $10 million (2013 dollars); (3) the transaction was completed; (4) the acquirer did not have more than a 5% stake in the target before the merger; (5) after the merger, the acquirer owns more than 99% of the target; (6) at least one of the target and the acquirer have debt outstanding; (7) the acquirer or the target is not a financial firm, as indicated by their primary SIC code; and (8) the target and the acquirer are both public firms, with daily stock market return data available on Center for Research in Security Prices (CRSP) and fundamentals data available from Compustat North America. The sample contains 1,884 mergers. The sample size of 1,884 mergers, between 1981 and 2013, appears reasonable. In comparison, Akbulut and Matsusaka (2010) study 4,764 mergers from 1950 to 2006, including mergers between financial firms. 15
The sample of randomly paired firms is constructed as follows. First, I take all firm years in the Fundamentals Annual file from the Compustat North America Database that cover the same time period as the merger sample: 1981 to 2013. For comparability with the merger sample, I retain those observations with market capitalization above $10 million (2013 Dollars) and eliminate all financial firms, as indicated by their primary SIC codes. Let us call this the population of comparable firms. From this population, I randomly draw 1,884 firm years without replacement. This constitutes the sample of hypothetical acquirers. For each observation, the year provides the year of the hypothetical merger. Next, for each hypothetical acquirer, I randomly draw, without replacement, one firm from the population of comparable firms, such that the firm belongs to the same year cohort as the acquirer; this firm is the hypothetical target. Finally, from that year, I draw a random date, which is the hypothetical announcement date. This completes the sample of randomly paired firms. For both of these samples, the mergers as well as the randomly paired firms, I collect data on the stock price, return, and shares outstanding for both the target and the acquirer from the CRSP Daily Stock File. Data are collected for each day over a 252-day window; the data collection ends 42 days prior to the merger announcement. This end date is chosen because the previous literature has found a run-up in target prices prior to a merger announcement, beginning approximately 42 days before the announcement date (Schwert (1996)). When data are not available for all 252 days, I solely retain those firms with at least 90 days of data. I also collect accounting data on short-term and long-term debt from the COMPUSTAT North America database. The fiscal year for these data corresponds to a year before the merger announcement. 16
I compute the size, leverage and cumulative abnormal return on the merger announcement, for both the samples. Detailed variable descriptions are available in Appendix A. For both the target and the acquirer, the size is computed as the market capitalization 42 days before the merger announcement. Market leverage is computed as the ratio of book value of debt to the sum of the book value of debt and the market capitalization. The cumulative abnormal returns (CAR) are computed using the Fama-French three-factor model, over a three-day (-1, 1) window around the announcement date. The betas for the model are computed using daily returns, over a 252-day window, which ends 42 days before the merger announcement. Data on Fama-French factors are collected from Kenneth French s website. 8 The combined CAR for the merger is the weighted average of the target and acquirer CARs, where weights correspond to pre-merger market capitalizations. Table 1 presents the summary statistics: Panel A pertains to the mergers and Panel B to the randomly paired firms. For mergers, the median size of the acquirers is $ 1.2 billion, whereas the median target size is only $100 million: the median acquirer is 12 times larger than the median target. For the random pairings, the median acquirer is worth $160 million, and the median target is worth $140 million; the acquirer and target are similar in size for random pairings, in contrast with mergers. For the merger sample, the median target leverage is 21.5%, and the acquirer leverage is 17.2%. Thus, the targets have slightly more debt than the acquirers. For the randomly paired firms, the median target leverage is 31.5%, and that for acquirers is 32.4%. Thus, both the target and the acquirer for the random pairings have significantly more debt than merger participants. 8 http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html 17
For the merger sample, the cumulative abnormal return ranges from -38.5% to 57.8%; the mean is 1.5%; there is much dispersion. In contrast, the CARs are closer to zero for the random pairings: the mean is -0.0%, and the median is -0.1%. This reflects the fact that the CARs for the random pairings do not correspond to any actual merger event, whereas CARs in mergers correspond to the announcement of the merger. Overall, the summary statistics for the sample of mergers, presented in Table 1, are comparable to the previous literature (Akbulut and Matsusaka (2010), Andrade, Mitchell and Stafford (2001)). In Table 2, I present the inputs to the structural model. A crucial input to the structural model is the correlation of asset returns, ρ. Row 1 of the table (Panel A) shows that the correlation ranges from -0.18 to 0.99, with mean 0.23 and median 0.18. Another input is the volatility of the asset return, σ i. As rows 2 and 3 show, the targets (median 47.0%) are generally more volatile than the acquirers (median 34.7%). This is expected, given that targets are usually smaller. Certain targets, and to a lesser extent certain acquirers, are very volatile. The coupon rates, C i, which depend on the leverage, are presented in rows 4 and 5. The X i (0) median annualized coupon rate is 1.1% of the firm value for the targets and 0.8% for the acquirers. Finally, the default barrier, is 6.2% of the firm value for the targets and 5.9% for the acquirers. D i, is presented in rows 6 and 7. The mean default barrier X i (0) In contrast, as panel B demonstrates, the input parameters for the randomly paired firms are very different. These firms are more volatile. The median volatility for acquirers is 35.8%, and that for targets is 35.9%. These firms also pay a higher coupon: the median is 1.6% for acquirers 18
and 1.5% for targets. These differences in input parameters reflect the fact that the randomly paired firms are smaller and more levered compared to merger participants. 4. Results: (the lack of) coinsurance from diversifying mergers 4.1 Coinsurance from diversifying and related mergers Since Lewellen (1971) conceptualized coinsurance as an explanation for diversifying mergers, several papers have documented coinsurance to be a major benefit of diversification. Hahn, Ogneva and Ozbas (2013) report that coinsurance benefits diversified firms by reducing their cost of capital. Duchin (2010) finds that diversified firms hold less cash; he argues that coinsurance enables diversified firms to manage with less precautionary cash holdings. Kuppuswamy and Villalonga (2010) document that diversified firms increased in value during the recent financial crisis; they attribute this value gain, in part, to coinsurance. In contrast, the literature is largely silent regarding coinsurance in the context of related mergers, even though there is no theoretical reason for limiting coinsurance to diversifying mergers. Motivated by this literature, I test whether diversifying mergers produce higher coinsurance. I classify a merger as diversifying when none of the divisions of the target or the acquirer have any three-digit SIC codes in common; otherwise, it is classified related (Kaplan and Weisbach (1992), Akbulut and Matsusaka (2010)). 9 Of the 1,884 mergers in the sample, 265 are classified as diversifying and the remaining 1,669 as related. In Table 3, I present the coinsurance benefits, as a percentage of the deal value, for related and diversifying mergers. 9 The previous literature uses several other approaches for classification: defining industries at the two-digit level (Matsusaka (1993) or using text-based measures instead of SIC codes (Hoberg and Philliips (2010)). 19
Row 1 presents the results for the diversifying mergers: the mean value gain from coinsurance is 2.3% of the deal value and the median 0.9%; the 75 th percentile is 2.5%; the 90 th percentile is 9.2%, whereas the 99 th percentile is 26.2%. For related mergers: from row 2, the mean is 2.1%; the median is 0.7%; the 75 th percentile is 2.4%; the 90 th percentile is 8.3%, whereas the 99 th percentile is 19.2%. These numbers suggest that coinsurance is small for diversifying mergers (as well as for related mergers). In fact, for a majority of mergers, both diversifying and related, coinsurance is smaller than the fees paid to investment bankers that average 1.22% of the transaction value (Hunter and Jagtiani (2003)). The other takeaway from Table 3 is that the coinsurance from related and diversifying mergers is similar. This similarity is echoed by Figure 1, which is a plot of the distributions of coinsurance for related and diversifying mergers. To test whether the distribution of coinsurance is the same for both related and diversifying mergers, I evaluate the Kolmogorov-Smirnov (KS) statistic. The KS statistic is 0.053, which corresponds to a P-value of less than 0.54, suggests that the null hypothesis that the distribution of coinsurance for both related and diversifying mergers is the same cannot be rejected, even at the 10% level of significance. To summarize, coinsurance is small for (most) diversifying mergers, and it is as small as that from related mergers. This finding contrasts with the literature that documents coinsurance to be an important benefit from corporate diversification. 4.2. Coinsurance estimates and cost of distress Are coinsurance estimates small because the structural model assumes the cost of distress to be small? To answer this, I introduce fixed costs of distress to the structural model: when a 20
firm defaults, it incurs a fixed cost of distress, φ = 1.32 million, over and above the previously assumed 16.5% proportional cost. Table 4 presents the coinsurance benefits for the entire sample of mergers (related and diversifying are pooled together; in any case, the two distributions are similar statistically). Row 1 presents the results for the solely proportional costs case, whereas row 2 presents the case with proportional and fixed costs. On adding fixed costs, the mean coinsurance benefit is 2.25%, which is slightly higher than the case with proportional costs alone (2.11%) but still very small. Similarly, the median increases to 0.76% from 0.67%. The 90 th percentile is 5.95%, versus 5.57% previously. Introducing fixed costs increases the estimated coinsurance benefits marginally. Are estimated benefits small because the structural model assumes a small proportional cost parameter, α? Instead of assuming α = 16.5%, the midpoint of Andrade and Kaplan s (1998) estimates, I assume α = 45%, in accordance with Glover (2013), while retaining the 1.32 million fixed cost of distress. Row 3 demonstrates that the mean coinsurance benefit is 2.58% and the median is 0.84%. The coinsurance benefits increase marginally. Table 4 indicates that even under high distress cost assumptions, in which firms in default lose nearly half their value to deadweight costs, the avoidance of such distress costs through mergers creates small value gains. Regardless of the parameterization of the cost of distress, for most mergers, coinsurance is small. 4.3. Counterfactual coinsurance benefits for randomly paired firms Although coinsurance benefits, as estimated by the structural model, are small in absolute magnitude, this does not imply that coinsurance is economically insignificant; to attain that conclusion, we need a benchmark. A natural benchmark is the counterfactual coinsurance that 21
would be produced if firms were to choose their merger partners randomly. The structural model makes it possible to estimate these counterfactual coinsurance because it relies solely on premerger stock price information in its estimation. The benchmark is created by drawing firms randomly from COMPUSTAT, pairing them, and computing their counterfactual coinsurance benefits. Table 5 (Row 1) presents the counterfactual coinsurance benefits, which are expressed as a percentage of the pre-merger target value, instead of deal value because deal value is not available for random pairings for the sample of randomly paired firms. For comparison, Row 2 presents the coinsurance benefits from mergers, scaled by the pre-merger target value. The estimates maintain baseline assumptions: the proportion of firm value lost in distress is α = 16.5%; there are no fixed costs. For random pairings, the mean estimated counterfactual value gain from coinsurance is 3.00%, which is comparable to the sample of mergers (2.93%). The median for random pairings is 0.71%, again comparable to the sample of mergers (1.18%). Finally, the 90 th percentile is 14.00%, compared with 11.51% for the mergers. A t-test of the difference in means yields a t- statistic of 0.42: the null hypothesis of equality of means cannot be rejected at the 1% level of statistical significance. The coinsurance from mergers is similar to counterfactual coinsurance from random pairings. When compared to a relative benchmark, coinsurance is small. 4.4. Why is coinsurance insignificant for most mergers? Why is coinsurance small for most mergers? The reason is not because the costs of distress are small: even assuming a high cost of distress, as we observed in Table 5, coinsurance is small. If the reason is not the cost of distress, is it the probability of distress? If firms have a low probability of distress even before the merger, merging such safe firms will not reduce the default 22
risk much it will not produce much coinsurance. To explore this possibility, using the structural model, I estimate ten year default probabilities before and after the merger. Table 6 presents the results. Row 1 presents the distribution of default probability for the targets: the mean is 10.6%; the median is 2.1%; the 75 th percentile is 14.7%; the 90 th percentile is 33.7%; and the 99 th percentile is 83.1%. Row 2 presents the distribution of default probability for the acquirers: the mean is 4.3%; the median is 0.2%; the 75 th percentile is 3.1%; the 90 th percentile is 12.8%; and the 99 th percentile is 65.7%. The majority of the targets and acquirers (nine of ten acquirers and three of four targets) are at low risk of distress before the merger, which leaves minimal scope for further risk-reduction from the merger. Thus, it is not surprising that coinsurance is small for most mergers. However, one-quarter of the targets are at substantial risk of default, as are 10% of acquirers. In general, targets are at higher risk of default compared with acquirers. Potentially, these targets can benefit from coinsurance. Row 3 presents the distribution of default probability for the merged firms: the mean is 3.0%; the median is 0.1%; the 75 th percentile is 1.5%; the 90 th percentile is 8.2%; and the 99 th percentile is 47.4%. Consistent with coinsurance reducing default risk, the merged firm s default probability is lower than that for both the acquirer and target. However, the default probability is closer to the acquirer s, which suggests that targets benefit more from coinsurance than acquirers. Rows 4 and 5 present the reduction in the default probability (default probability before the merger - default probability after) for the target and acquirer. The mean reduction for the target is 7.6%; the median is 1.1%; the 75 th percentile is 10.7%; the 90 th percentile is 26.1%, while the 23
99 th percentile is 65%. For the acquirer: the mean reduction is 1.4%; the median is 0; the 75 th percentile is 1.6%; and the 90 th percentile is 4.9%, while the 99 th percentile is 22.1%. For one-quarter of targets, the default risk is reduced substantially; these are potentially high-coinsurance mergers. However, for the vast majority of acquirers and for a majority of targets, the default risk does not decrease much; this again illustrates why coinsurance is small for the majority of mergers. In view of the multi-faceted, coherent and robust evidence presented in Section 4.1 through 4.4, I conclude that coinsurance, for most diversifying mergers, is insignificant. Does this necessarily contradict the empirical, reduced-form literature that documents coinsurance to be important for diversifying mergers? This paper estimates coinsurance benefits stemming from the reduction in default risk, in accordance with the theoretical literature on coinsurance (Leland (2007)); however, coinsurance benefits may also stem from other channels, such as avoidance of costly external finance (Hann, Ogneva and Ozbas (2013)). To the extent that these other channels, which were hitherto neglected by the theoretical literature, are important, the structural model may not adequately capture all coinsurance benefits and coinsurance from diversifying mergers may actually be substantial. 5. Results: when coinsurance is significant Although coinsurance is small for most mergers, as Table 3 demonstrates, for certain mergers, the structural model estimates coinsurance to be significant; for the top decile of mergers (when sorted by coinsurance), which includes as many as 186 mergers, coinsurance exceeds 5.57%. This result is also reflected in Table V; approximately one-quarter of targets experience a substantial risk reduction from the merger due to coinsurance. In this section, I 24
identify these high-coinsurance mergers and analyze what causes them to be high in coinsurance. Furthermore, I present other empirical evidence that supports the conclusion from the structural model that these mergers are indeed high-coinsurance. 5.1. High-coinsurance mergers I refer to the top decile of mergers, for which coinsurance exceeds 5.57% of the deal size, as high-coinsurance mergers. To identify these high-coinsurance mergers, I compare their characteristics with the remainder of the sample. Table 7 presents the results. High-coinsurance mergers involve targets that are significantly smaller (median $34.3 million) compared with the remainder of the sample (median $121.3 million). Similarly, acquirers in high-coinsurance mergers are significantly smaller (median $680 million) compared with the remainder of the sample (median $1.14 billion). In both cases, the differences in size are statistically significant at the 1% level. For high-coinsurance mergers, the acquirer/target size ratio (median 16.9) is higher compared with the remainder of the sample (median 7), although the difference is not statistically significant. Targets in high-coinsurance mergers are heavily indebted, employing a median leverage as high as 48.9%, compared with the modest 10.7% median leverage employed by targets in the remainder of the sample. Acquirers in high-coinsurance mergers also employ higher leverage, with a median of 30.0%, compared with the median leverage of 15.6%, which is employed by acquirers in the remainder of the sample; however, the difference is not as stark as for targets. In both cases, the differences are statistically significant at the 1% level. 25
As expected from the literature on coinsurance, the correlation between target and acquirer returns is lower for high-coinsurance mergers (median 0.10) compared with the remainder of the sample (median 0.17). What is perhaps unexpected is the small magnitude of the difference. In sum, coinsurance from mergers is high when it involves the following: (i) highly levered targets; (ii) levered acquirers; (iii) small firms; or (iv) merger participants have low correlation. To determine which of these effects is more important, I perform a logistic regression. The dependent variable is 1 when the merger is high-coinsurance. The explanatory variables are target leverage, acquirer leverage, target size, ratio of acquirer to target size, and correlation. Each of the variables is standardized with a zero mean and a standard deviation of one. The effect is strongest for target leverage: a one-standard deviation change in target leverage increases the probability of a merger being a high-coinsurance from 5% to 8%. The next strongest effect is for correlation: a one-standard deviation decrease in correlation increases the probability of a merger being a high-coinsurance from 5% to 7%. The other significant determinant is acquirer leverage: a one-standard deviation change in acquirer leverage increases the probability of a merger being a high-coinsurance from 5% to 7%. 5.2. Coinsurance and merger likelihood If two firms exist for which merging would create value, other things equal, those mergers should be likely to occur (Hoberg and Phillips (2010)). 10 If the structural model is correct in identifying mergers with a high potential to be coinsurance mergers, such high-coinsurance 10 Hoberg and Phillips (2010) find that when firms are randomly paired, asset complementarities from such random pairings are less than that from mergers. The researchers conclude that high asset complementarities motivate mergers. 26
mergers, as determined by the structural model, should be likely to occur; high coinsurance should predict the merger incidence. To test this hypothesis, I utilize the randomly paired sample discussed in Section IV.C (recall that these firms are drawn from Compustat at random, paired, and their counterfactual coinsurance is computed). I perform a logistic regression for the combined sample of 1,884 mergers and 1,884 randomly paired firms. The dependent variable is a dummy for merger incidence: one for a merger and zero otherwise. The main explanatory variable is a dummy that is 1 if the deal is estimated to produce high coinsurance; the coinsurance exceeds 5.57% of deal size. I use the following controls: target size, acquirer size, target leverage, acquirer leverage, acquirer market-to-book, target market-to-book, correlation of stock return, and a dummy that is one when the three-digit primary SIC codes are different. I also use year fixed effects to control for merger waves. The results are presented in Table 8. Column (1) shows the coefficient on high-coinsurance dummy to be 0.76, which is significantly different from zero at the 1% level. The likelihood of two firms merging increases when the potential for coinsurance from such a merger, as estimated by the structural model, is high. This result suggests that for certain mergers that involve smaller firms or highly levered targets, coinsurance can be the rationale for the merger. 5.3 Value gain from coinsurance and gain in stock price The structural model estimates that for certain mergers, there is a significant increase in firm value firm value in the structural model is the sum of debt and equity values due to coinsurance. If coinsurance increases firm value in certain mergers, stock prices should increase 27