Product market competition and choice of debt financing: evidence from mergers and acquisitions

Product market competition and choice of debt financing: evidence from mergers and acquisitions Haekwon Lee University at Buffalo School of Management (haekwonl@buffalo.edu) Current draft: August 10, 2017 Abstract I investigate how product market competition affects a firm s choice between private and public debt in financing mergers and acquisitions. I find that acquirers facing intense product market competition prefer private debt, especially for firms in industries with more potential bidders and for firms with lower market share. I also show that higher product market competition increases the cost of private debt, particularly for firms that rely heavily on their relationship lenders, while it has no impact on the cost of public debt. Collectively, these findings suggest that disclosure costs associated with product market competition increase acquirers reliance on private over public debt, despite it being relatively more expensive. Keywords: Choice of debt financing; Product market competition; Cost of debt Please address correspondence to: Haekwon Lee Department of Finance State University of New York at Buffalo 255 Jacobs Management Center, Buffalo, NY, 14260 Email: haekwonl@buffalo.edu

1. Introduction This paper examines how product market competition affects a firm s choice of debt financing for mergers and acquisitions (M&A) and contract terms for both private and public debt. Prior studies in corporate capital structure document significant institutional differences between private and public debt markets and provide explanations on the benefits and costs of private versus public debt financing. 1 Extant literature further shows that corporate debt structure is determined by credit quality (Denis and Mihov, 2003), accounting quality (Bharath, Sunder, and Sunder, 2008), and corporate ownership structure (Lin et al, 2013). While most of these studies focus on the firm-level determinants of debt structure, this paper focuses on the link between industry-level competition dynamics and the choice between private and public debt. In particular, this study investigates how the risk of information leakages to rivals in a competitive product market affects its choice of debt financing. Research questions in the present study are motivated by prior literature focusing on differential disclosure costs between private and public financing. For instance, studies of initial public offerings explore the tradeoff between private and public financing; although private financing is more expensive than public financing, the former enables firms to keep proprietary information hidden from competitors (Bhattacharya and Ritter, 1983; Maksimovic and Pichler, 2001; Spiegel and Tookes, 2007; Chemmanur, He, and Nandy, 2010). This implies that firms in a highly competitive product market are less likely to finance their operations using public financing (Chemmanur, He, and Nandy, 2010). Similarly, previous studies compare disclosure cost between private and public debt financing. Relative to private debt, public debt has higher 1 Past studies show that private debt lenders provide better monitoring (Diamond, 1984; Boyed and Prescott, 1986; Berlin and Loyes, 1988), have more access to proprietary information (Fama, 1985), and have more re-contracting flexibility (Chemmanur and Fulghieri, 1994; Gertner and Scharfstein, 1991). 2

risk of information leakage to rivals (Bhattacharya, Boot, and Thakor, 2004; Yosha, 1995). The essence of the argument is that detailed firm information disclosed to public debt market can be observed by a firm s competitors, which can jeopardize its competitive position and negatively affect future profitability (Yosha, 1995; Bhatacharya and Cheisa, 1995). Consequently, I expect that product market competition would have an impact on its debt structure because of the cost associated with publicly revealing information. There are a number of other potential ways that a firm s choice between private and public debt financing could be related to product market competition. First, firms facing high product market competition might prefer to obfuscate in order to avoid sharing proprietary information with their rivals which would be reflected in poor accounting quality. If this is the case, due to superior information gathering and processing abilities of the private lenders, firms with poor accounting quality would face lower adverse selection costs in private debt market compared to public debt market and rely more on private debt to mitigate the adverse selection costs (Bharath, Sunder, and Sunder, 2008). Second, a firm s competitive pressure might act as an external disciplinary tool and enhance corporate governance (Giroud and Muller 2010, 2011). This view predicts that firms facing high product market competition would rely less on private debt because of their less need for bank s efficient monitoring. Finally, high product market competition can increase a firm s default risk (Bolton and Scharfstein, 1990; Valta, 2012; Morellec, Valta, and Zhdanov, 2015). Since private debt contract is easier to be renegotiated relative to public debt contract (Bharath, Sunder, and Sunder, 2008; Morellec, Valta, and Zhdanov, 2015), firms might rely more on private debt to avoid liquidation in case of default. 3

The overall effect of product market competition on choice between private and public debt is an empirical issue. To investigate this, the present study analyzes the impact of product market competition on the acquirer s choice of debt financing for M&A. This study focuses on M&A financing for several reasons. First, corporate takeovers are among the largest investments that a company will ever undertake (Betton, Eckbo, and Thorburn, 2008) in which private and public debts are vital sources of external financing. 2 Second, prior research shows various links between product market competition and M&A. For instance, Akdogu (2011) shows that acquirers overpay for a target when the costs of losing the target to industry rivals are higher. Aktas, de Bodt, and Roll (2013) suggest that M&A announcement is, in general, bad news for bidders industry competitors because potential synergy or potential market share increase for bidders will put competitive pressure on them. This implies that industry rivals have incentive to act in advance in case of information leakage about a possible M&A deal. Consequently, I conjecture that acquirers in a competitive product market would be more sensitive to possible information leakage to their rival firms, which could affect their preference of private debt over public debt in financing M&A deals. To shed light on the issue, I obtain a sample of M&A transactions from the Thomson Reuters Securities Data Company (SDC) Mergers and Acquisitions Database that are partially or fully funded by either private debt (syndicated loans) or public debt (corporate bonds) from 1997 to 2012. Using Thomson Reuters Loan Pricing Corporation Deal Scan database (DealScan), I identify merger related loans extended to acquirers from one year before the deal announcement to one month after the announcement as being used to finance M&A deals. 2 In 2016, syndicated lending (private debt) for acquisition financing accounted for 17% of the US syndicated loan market with a volume of $391 billion. The largest corporate bond issuance in 2016 was Anheuser Bush s $46 billion deal to acquire SABmiller PLC. 4

Using Thomson Reuters Securities Data Company (SDC) Global New Issues database, I also identify bonds issued by the acquirers from one year prior to the deal announcement to one month after the announcement as being used to finance the M&A deal. 3 The final sample includes 1,390 M&A deals in which 569 deals are classified as bank financed, and the remaining 821 deals are classified as bond financed. 4 The paper s main findings are as follows. Using the Herfindahl index for text-based network industry classification (TNIC HHI) provided by Hoberg and Phillips (2016), I find that acquirers that face higher product market competition are more likely to finance their M&As with private rather than public debt. This result holds for alternative measures of product market competition: firm level aggregate product similarity with rivals based on text-based industry classification (TNIC similarity) provided by Hoberg and Phillips (2016) and industry-wide net profit margin (Lerner index). 5 To test the robustness of the results, I conduct a battery of tests using alternative empirical specifications. First, to mitigate concerns that the results may be driven by firms with limited access bond markets, I exclude acquirers with no prior bond issuance and show that the results are qualitatively similar. Second, I show that the results are qualitatively identical regardless of whether firm-level bond ratings or a binary variable that classifies investment 3 Average time between M&A announcement and deal completion date is around 39 days. Hence, I use one month after the deal announcement as the end of my identification window. The results in the paper are robust to the use of alternative windows, e.g., one year before the announcement to the deal announcement. 4 Because this paper investigates acquirers choice of financing between private and public debt, the 139 deals with both bond and loan financing are excluded from the sample. 5 The industry-wide net profit margin (or Lerner index), widely used in industrial organization literature, measures the extent to which prices are set above marginal cost. Prior studies that use Lerner index as a proxy for market competition include Aghion et al. (2005), Gaspar and Massa (2006), Giroud and Mueller (2010), and Aghion, Reenen, and Zingales (2013). 5

grade and speculative grade firms are included in the regressions. Third, I show that the results hold after controlling for the acquirer s monitoring environment proxied by the level of institutional ownership. Fourth, because firms with poor accounting quality are more likely to rely on private debt (Bharath, Sunder, and Sunder, 2008; Dhaliwal, Khurana, and Pereira, 2010), I control for firm level accounting quality measure suggested by Dechow and Dichev (2002) and show that the results hold. Finally, because supply of credit affects firm s capital structure (Leary, 2009), I show that the results are robust to the inclusion of supply factors: industry-level relative cost of loan and bond and KZ index (Kaplan and Zingales, 1997), which proxies for firm-level financial constraint. Uysal (2011) and Harford and Uysal (2014) argue that the probability of firms making acquisitions is higher in industries with high volume of M&A transactions. Firms with lower market shares are considered to have lower market power (Nickell et al., 1992; Nickell, 1996) and are vulnerable to industry leaders competitive pressure (Katz and Shapiro, 1987). Thus, I conjecture that the acquirers would be more sensitive to possible information leakage to industry rivals when M&A transaction volume in an industry is high or when they are industry followers. Using subsamples based on M&A liquidity and market share, I find that the impact of product market competition on likelihood of private debt financing is stronger for acquirers in an industry with greater M&A volume and for acquirers that are industry followers. Numerous studies document that banks can extract information rents by charging higher interest rates (Schenone, 2010; Hale and Santos, 2009) by acquiring firm specific information during lending process (Rajan, 1992; Schenone, 2010). Rajan (1992) and Diamond (1991) argue that borrowers option to issue public debt lowers lenders incentives to extract information rents. Thus, I conjecture that acquirers avoiding public debt financing because of 6

high disclosure cost might have less bargaining power over banks and suffer from high loan spread while there is no significant effect on bond spread. On the other hand, if product market competition affects mainly firms accounting quality or default risk, I should observe a positive relation between spreads of both type of debt and intensity of product market competition. 6 Using 550 M&A loans, I find that product market competition is positively related to the cost of loans. 7 The results suggest that the acquirers in a highly competitive environment suffer from significantly higher loan costs than do those in a less competitive environment, with a difference of about 25% of the average loan cost. This constitutes approximately a 55bp higher loan spread for the acquirers in a competitive environment compared to others. In addition, I show that the results are robust after accounting for endogeneity in loan contract terms by employing a methodology suggested by Bharath et al. (2011). In contrast, I find no significant relation between product market competition and the cost of bonds from a sample of 1,053 M&A bonds. I interpret these results as evidence that acquirers facing high product market competition suffer from higher loan costs due to their loss of bargaining power over lenders because of their reluctance to issue public debt. Extant studies find that banks are likely to extract rents from informationally captured (or locked-in) borrowers who rely heavily on the relationship banks (Sharpe 1990; Greenbaum, Kanata, and Venezia, 1989; Schenone, 2010). The results on debt choice and debt cost collectively show that acquirers avoiding public debt financing because of disclosure cost have 6 Bharath, Sunder, and Sunder (2008) find that accounting quality increases both loan spread and bond spread negatively. Valta (2012) shows that default risk arising from product market competition is priced in loan spread. Driessen (2004) provides evidence that default risk is priced in bond spread. 7 Some loans are included in multiple M&A deals. Hence, I pick the M&A deal in which announcement date is closest to the loan contract initiation date. The result is robust to keeping all loans and treating them independently. Same methodology applies to M&A bond sample. 7

less bargaining power over private debt lenders in a competitive product market. Hence, I conjecture that the locked-in borrowers are likely to suffer more from bank s rent extraction than the ones with banking alternatives in a competitive environment. In support of this conjecture, I find that the locked-in acquirers suffer from substantially higher loan costs in a competitive environments than those relying less on relationship banking. The difference is estimated as 46% of the average cost of loans. Overall results suggest that the acquirers in a highly competitive product market are more likely to rely on private debt to avoid possible information leakages, although it is an expensive way of financing because of the bank s information rent extraction. To ascertain whether private debt is better way of financing than public debt at mitigating potential information leakages in M&A, I compare bid premiums between private debt financed and public debt financed M&A deals. Aktas, de Bodt, and Roll (2010) show that the presence of potential rival bidders can increase bid premiums paid in M&A deals. Also, Akdogu (2011) provides theoretical evidence that cost of losing the target to industry rivals leads an acquirer to rationally overpay for the target. Hence, I conjecture that if private debt is better than public debt at mitigating information leakages, bid premiums should be lower for loan financed deals than bond financed deals. Furthermore, I expect that the overpayment in bond financed deals will be more pronounced in a more competitive product market in which acquirer s cost of losing the target would be higher. Using Officer s (2007) proxy premium, I find consistent results that premiums are lower for loan financed deals relative to bond financed deals and that the difference is more pronounced in a relatively more competitive product market. The findings in this paper contribute to the capital structure literature in several ways. First, much of the empirical research on debt structure focuses on firm level determinants of 8

the choice between private and public debt. In contrast to prior studies, this paper shows the link between industry dynamics and corporate debt structure. Second, the results suggest that the effect of competition on a borrower s choice between private and public debt accounts for the disclosure cost of public financing by showing that the borrowers with more competitive pressure prefer private debt financing. By analyzing debt contracts, I show that loan costs are higher when product market competition is high, especially for the locked-in borrowers, whereas bond costs are not affected. In summary, overall results suggest that acquirers in a competitive product market face a tradeoff between lower disclosure cost of private debt and higher financing cost in its choice between private and public debt. The remainder of the paper is organized as follows. Section 2 describes the data sources and sample selection process. Section 3 discusses results on empirical analysis. Conclusion is provided in section 4. 2. Data and variables 2.a. Syndicated loans Data for syndicated loans are from the Thomson Reuters Loan Pricing Corporation Deal Scan database (DealScan). DealScan provides information on loan pricing, contract details, and terms and conditions, in addition to detailed information on borrower and lender identities. The loan information is organized by package that includes multiple loan facilities; on average, each package contains approximately 1.5 facilities. All loan facilities in the same package share the same borrower firm, but the composition of the lending syndicate, identities and roles of each lenders, loan type and purpose, contract initiation and maturity date, and other deal 9

characteristics may vary. 8 For each loan facility, I obtain information on loan pricing, recorded as all-in-drawnspread, defined as the total annual cost including a set of fees and fixed spread over LIBOR rate. I further record loan maturity (recording at initiation, in months), facility amount (in US$), number of lenders, and indicator variables identifying collateralized loans and loans with performance pricing. The sample includes indicator variables based on the database fields identifying loan type. I exclude loans whose status is Cancelled or Rumor. Further, I exclude from the sample all loans for which data on the composition of the lending syndicate is missing and loans with conflicting information (for example, loans marked as single-lender loans for which multiple lenders are listed). Finally, I keep loan facilities whose primary purpose is Takeover, Acquisition Line, or Merger. 2.b. Corporate bonds I obtain sample of newly issued bonds by U.S. companies from Thomson Reuters Securities Data Company (SDC) Global New Issues database. The database provides detailed information on the newly issued bonds, including the spread, issuance date, yield-to-maturity (YTM), maturity date, net proceed, and letter rating. For each newly issued bond, I obtain information on the spread, i.e., the difference between the yield-to-maturity of the corporate bond of interest and the yield-to-maturity of a U.S. Treasury bond with comparable maturity, measured in basis points. I also obtain information on bond maturity (in months), amount of proceeds, and indicator variables identifying callable, puttable, and security type provided by 8 A detailed description of the data is provided in Chava and Roberts (2008). 10

the SDC database. Finally, I exclude non-debt fixed claims (e.g. lease financing) and issuances classified as employee stock option plans. 2.c. Debt financed M&A deals I obtain acquisition sample from the Thomson Reuters Securities Data Company (SDC) Mergers and Acquisitions Database. I select domestic (U.S.) M&A deals announced between January 1, 1997 and December 31, 2012. The sample requires that (1) the acquirer is not a financial firm (SIC code between 6000 and 6999) or utilities (SIC code between 4900 and 4999); (2) the acquirer is U.S. public firm with annual financial statement information available from COMPUSTAT and with stock return data from the Center for Research in Security Prices (CRSP); (3) the deal is not in the form of Repurchase, Recapitalization, Spinoff, Selftender, Leveraged Buyout, or Privatization, ; (4) the acquirer owns less than 50% of the target share before the announcement and seeks to hold more than 50% of the shares after the deal completion; (5) the deal paid partly or entirely with cash; and (6) the deal value reported in SDC exceeds $1 million. Because the average number of trading days between the merger announcement date and the effective date is approximately one month, I define a merger as a loan (or bond) financed deal if any loan (or bond) is taken by the bidder firm from one year before the merger announcement date and one month after the announcement date. Using the Dealscan-Compustat link file provided by Professor Roberts, I identify 569 M&A deals as loan financed deals and 821 M&A deals as bond financed deals after merging the bond issuance data with the M&A 11

dataset. 9 Because this article focuses on the acquirer s choice of financing, my sample excludes the M&A deals with both loan and bond financing. 2.d. Competition measures As measures of competition, I mainly rely on text-based measures of product market competition provided by Hoberg and Phillips (2016). 10 Construction of competition measures in Hoberg and Phillips s database begins with identifying related firms using information on all firms mandatory product description in 10-K. After estimating firm-by-firm pairwise product similarity, rival firms are grouped by their similarities, a method defined as Text-based industry classification (TNIC). Hoberg and Phillips (2016) use these groups to provide aggregate Herfindahl index (TNIC) and aggregate similarity measures (TNIC) for each firm in the Compustat universe. The lower the Herfindahl index is for a firm, the more competition the firm faces. The higher the similarity score is, the more competition the firm faces. As a main measure of product market competition, I construct a binary variable TNIC HHI that equals one if the Herfindahl index (TNIC) are in the lowest quartile of the yearly distribution of the Compustat universe and zero otherwise. As an alternative measure of product market competition, I construct a binary variable TNIC Similarity that equals one if the Similarity (TNIC) are in the highest quartile of the yearly distribution of the Compustat universe 9 I would like to thank Professor Michael Roberts for sharing the Dealscan-Compustat link file for the period between 1983 and 2012. For more details on the link file, please refer to Chava and Roberts (2008). The file is available at http://finance.wharton.upenn.edu/~mrrobert. 10 I would like to thank Professor Gerard Hoberg and Professor Gordon Phillips for sharing the competition measures database for the period between 1996 and 2015. For more detailed description on the competition measures, please refer to Hoberg and Phillips (2016). The file is available at http://hobergphillips.usc.edu/industryconcen.htm. 12

and zero otherwise. Finally, I include Lerner index (or industry net profit market) used in industrial organization literature. To measure the Lerner index, I calculate the yearly industry median (two digit SIC code) operating profits before depreciation, interest, special items, and taxes over sales. The operating profit is calculated as sales (Compustat item 12) subtracted by cost of goods sold (Compustat item 41) and selling, general, and administrative expenses (Compustat item 189). When the operating profit information is unavailable, I use operating income (Compustat item 178). I construct the Inverse Lerner index as (1-Lerner index) multiplied by 100 and use the percentage points. 2.e. Additional data Accounting information is from the Compustat database. I estimate size (log of total assets), valuation (Tobin s q), profitability (ROA), tangibility (net fixed assets to total asset), leverage (debt to total asset), bankruptcy risk (Altman Z-score), interest coverage (EBITDA to interest expenses), cash holding (cash holding to total asset), firm age (log of firm age) based on accounting information and included as firm-level control variables in all empirical models. Using M&A deal specific information from SDC Mergers and Acquisitions Database, I construct M&A deal level controls: indicator variables identifying horizontal merger based on two digit SIC code, deals for public target, deals with multiple bidder, hostile takeover, tender offers, and acquirer s toehold position, relative deal size (i.e., transaction value over acquirer s market capitalization 11 day before the announcement date), and M&A liquidity index (i.e., aggregate transaction value to aggregate book value of assets in the same two-digit SIC in a year). Analyst information is from the Thomson Reuters I/B/E/S database. I construct the number of analysts covering a firm by counting the number of analysts issuing earnings 13

forecasts for the firm during the previous year. Table 1 includes a full description of the definitions and sources for the variables. *** Insert Table 1 about here *** 3. Empirical results 3.a. Descriptive statistics Panel A of Table 2 provides summary statistics on competition measures, acquirer characteristics, M&A deal characteristics, and macroeconomic controls. The sample in this paper includes 1,390 M&A deals. Of those, 15% are identified as deals in which acquirers face higher competition. Acquirers are mostly large firms with median assets equal to $1.8 billion. Average firm age is about 24 years. M&A deal size is, on average, about $525.51 million. Panel B of Table 2 presents summary statistics on terms of loans and bonds. For loan characteristics, average loan spread is 222.45 bp over LIBOR. The average maturity is 54.78 months, or approximately 4.5 years. The loan size is on average $225.43 million. About 86% of the loans are collateralized and about 72% includes performance pricing in the loan deal. By looking at historical borrowing information of the acquirers, I measure how informationally captured the acquirer is by the lead arranger of the loan deal. Using a relationship intensity measure as a proxy for the captive borrowing suggested in Schenone (2010), I identify that lead arranger, on average, contributes about 37% of the total number of borrowing. For bond characteristics, average bond spread is about 205.87 bp. Maturity and bond size is 135.07 months and $446.66 million, respectively. Bonds with callable option accounts for 53% of the bond sample whereas 19% of the bond sample are puttable bonds. *** Insert Table 2 about here *** 14

3.b. The impact of product market competition on choice between private and public debt in financing M&A In this section, I test how product market competition affects the acquirer s choice between private and public debt financing. I conjecture that acquirers that face high product market competition would be more likely to rely on private debt financing because of their disclosure cost. Following Dhaliwal, Khurana, and Pereira (2011), I model the choice of debt issuance (public vs. private) in a logit framework. 11 The response variable is a binary variable that is equal to one for private debt (loan) and zero for public debt (bond). The main explanatory variable is TNIC HHI. I control for Bond market access, a binary variable that equals one if the firm has prior bond issuance and zero otherwise. I further control for firm characteristics by including Total assets (Log), Tobin s q, Profitability, Tangibility, Leverage, Altman Z, Coverage, Cash holding, Firm age (Log), Analysts following (Log), and Speculative rating, as defined in Table 1. Motivated by prior literature, I control for M&A deal characteristics by including Cross industry, Competing deal, Public deal, Hostile, Tender offer, Toehold, Relative size, and M&A liquidity, as defined in Table 1. Finally, the model includes fixed effects for years and industries (2-digit SIC codes). Standard errors are adjusted for heteroskedasticity and clustered by acquirer. The sample includes M&A deals partially or entirely financed by either private or public debt. This test includes both completed and incomplete deals. 12 11 This is similar to Hadlock and James (2002) and Bharath, Sunder, Sunder (2008). Hadlock and James (2002) use logistic regression and modeled choice between private (loan) financing and other types of financing. Bhrarath, Sunder, Sunder (2008) use probit model to test the choice between private (loan) and public debt (bond). The results are similar using probit model. 12 This purpose of this test is to see how acquirers make their financing choice between private and public debt before the deal completion. Hence, incomplete deals are included in the sample. The results are similar when only completed deals are included in the sample. 15

The results (see Table 3) show that larger firms and firms with greater tangible assets are more likely to rely on public debt financing, which is consistent with findings from prior literature. The coefficient on Bond market access is highly significant and negative, which implies that firms that have access to the bond market are more likely to rely on public debt financing. Consistent with the prediction, acquirers with higher product market competition are more likely to finance M&A with private debt. The results are robust, using alternative measures of competition: TNIC similarity and Inverse Lerner index, as reported in columns 2 and 3. *** Insert Table 3 about here *** 3.c. Robustness tests In this section, I perform robustness tests using various model specifications. First, to ensure that the results in the previous section are not driven by acquirers without access to the public debt market, I re-estimate the baseline model using the subsample of acquirers who have issued bonds in prior years. The results (see column 1 of Table 5) are robust to excluding the sample of acquirers without access to public debt market. Second, numerous studies find that product market competition increases firm s default risk (Bolton and Scharfstein, 1990; Valta, 2012; Morellec, Valta, and Zhdanov, 2015). Since banks have better renegotiating abilities compared to public lenders (Bharath, Sunder, and Sunder, 2008; Morellec, Valta, and Zhdanov, 2015), firms facing high default risk resulting from competitive pressure might rely more on private debt to avoid liquidation. Accordingly, I recognize that the positive relationship between product market competition and the likelihood of private debt financing might be due not to the acquirer s disclosure cost, but to high default 16

risk resulting from product market competition. To address this concern, I replace Speculative rating (Binary) with firm-level ratings. The results (see column 2 of Table 5) continue to hold after replacing Speculative rating (Binary) with actual ratings. Third, prior literature shows product market competition to be closely related with firm governance (e.g., Shleifer and Vishny, 1997; Tian and Twite, 2011; Giroud and Muller, 2010, 2011). To ensure that the results are not partly driven by firm level governance, I control for institutional ownership as an additional proxy for firm level governance. Using information from the CDA/spectrum, I construct Institutional ownership calculated as the ratio of the number of shares held by institutional investors over the number of shares outstanding. The results (see column 3 of Table 5) persist after controlling for institutional ownership. Fourth, firms in a competitive product market might have poor accounting quality to avoid sharing their proprietary information with competitors. Bharath, Sunder, and Sunder (2008) and Dhaliwal, Khurana, and Pereira (2010) document that firms quality of accounting information affects their choice between private and public debt financing. The argument is that firms with poor accounting quality would face lower adverse selection costs in private debt market because of superior information gathering and processing abilities of banks; as a result, they might rely more on private debt to resolve their information problem. To ensure that the results are not driven by the acquirers quality of accounting information, I include acquirers accounting quality as additional control variable. Following Dechow and Dichev (2002), I estimate the regression model that maps current accruals into prior period, current period, and next period cash flows using quarterly Compustat data over a 3 year rolling window. I construct 17

Accounting quality measured as the standard deviations of the residuals multiplied by -1. 13 Sample size drops to 1,366 due to data availability, but the results (see column 4 of Table 5) continue to hold. The negative coefficient on Accounting quality implies that firms with poor accounting quality are more likely to rely on private debt which is consistent with prior literature. However, the effect is not statistically significant. Finally, one concern with the results for the choice of debt financing is that a firm s debt structure is not solely determined by its demand. For instance, Leary (2009) argues that because supply of capital is not infinitely elastic, a firm s debt level is determined by both demand and supply of capital. To alleviate the concern that the results are driven by supply of credit, I include two variables as additional controls. Following Almeida, Campello, and Weisbach (2004), I include Financial constraint, a binary variable that equals one if the KZ index of a firm is in the bottom three deciles of the Compustat universe and zero otherwise in column 5. 14 In column 6, I include Cost of loan vs. bond measured as the ratio of average cost of loans over average cost of bonds using a 3 year window for each industry. The cost of both types of debt is estimated by multiplying amount of proceeds, spread, and maturity. The negative coefficient on Cost of loan vs. bond implies that higher relative loan cost leads to more preference for public debt by the firms. However, the effect is not statistically significant. The effect of competition on choice of debt continues to hold in both model specification. *** Insert Table 4 about here *** 13 Prior studies that use same proxy include Francis et al. (2004) and Dhaliwal, Khurana, and Pereira (2010). 14 KZ index is measured as -1.002ΧCash flow+0.283χtobin s q+3.139χleverage-39.368χdividends- 1.315ΧCash holdings. 18

3.d. Using subsamples to determine the impact of product market competition on choice between private and public debt The results in the previous section show that acquirers are more likely to finance their M&A deals with private debt rather than public debt, which is consistent with the idea that bidders are concerned about possible information leakage to their rivals. In this section, I investigate whether the choice between private and public debt is uniform across firms with differing competitive pressure by re-estimating the baseline model using subsamples. First, Uysal (2011) and Harford and Uysal (2014) argue that high M&A transaction volume within an industry implies high probability of firms in the industry making acquisitions. Hence, I expect that acquirers would be more sensitive to possible information leakages to rival firms when there are more potential acquirers among their competitors. Second, Nickell et al. (1992) and Nickell (1996) document that firms with lower market shares are considered to have lower market power. Furthermore, Katz and Shapiro (1987) show that industry followers are vulnerable to competitive pressure from industry leaders. Thus, I expect that acquirers with lower market shares will face greater disclosure cost compared to industry leaders. Accordingly, I conjecture that the effect of product market competition on the choice of debt financing will be more pronounced for acquirers in industries with high M&A transaction volume and for those who have lower market share. To test the conjecture, I bifurcate the sample with M&A liquidity index and Market share by using sample median values. Then I run the logit regressions of the choice between private and public debt for these subsamples. Consistent with the conjecture, the results (see Table 4) show that the impact of product market competition on the choice between private and public debt financing is more pronounced for firms in industries with more M&A transaction volume 19

and for firms that have less market shares. Meanwhile, I do not find statistically significant impact for the firms in industries with less M&A transaction volume and for the firms with greater market shares. *** Insert Table 5 about here *** 3.e. The impact of product market competition on loan spreads Overall results in choice of debt financing show that acquirers prefer private debt over public debt in financing M&A deals when facing high product market competition. This supports the argument that firms in a competitive product market rely more on private debt financing because of the disclosure cost in public debt financing. Furthermore, the evidence rules out the idea that competitive pressure plays an external governance role. The governance channel of product market competition posits that firms in competitive environment would rely more on public debt because of less need for efficient monitoring by private lenders. However, one concern is that I cannot fully rule out other effects of product market competition. To address this concern, I investigate how product market competition affects loan and bond spreads. Prior literature suggests that banks can have great bargaining power over borrowers through an information monopoly (Rajan, 1992; Schenone, 2010). Further, Diamond (1991) and Rajan (1992) document that the bank s incentive to extract information rents will be lower when borrowers have an option to issue public debt. Based on these arguments, I conjecture that the acquirers that are avoiding public debt market to mitigate risk of information leakages will suffer from high loan spread due to having less bargaining power over private debt lenders. 20

To test the conjecture, I employ regression analysis to gauge the impact of product market competition on the loan spread. In the model, the response variable is logged Loan spread, measured as the total annual cost, including various fees and fixed spread, over LIBOR rate for each dollar used under the loan commitment reported in the DealScan database. The variable of interest is TNIC HHI. I also control for loan characteristics by including Loan size (Log), Maturity (Log), Collateral, a binary variable that equals to one if the loan is secured and zero otherwise, and Performance pricing, a binary variable that equals to one if the loan includes performance pricing feature and zero otherwise. 15 I also control for firm characteristics by including Total assets (Log), Tobin s q, Profitability, Tangibility, Leverage, Altman Z, Coverage, Cash holding, Firm age (Log), Analysts following (Log), and Speculative rating, as defined in Table 1. I further control for M&A deal characteristics by including Cross industry, Competing deal, Public deal, Hostile, Tender offer, Toehold, Relative value, and M&A liquidity, as defined in Table 1. Macroeconomic factors are controlled by including term spread (i.e., the yield spread between a ten-year Treasury bond and three month Treasury bond) and default spread (i.e., the yield spread between BAA and AAA corporate bond indices). The model also includes fixed effects for years, industries (2-digit SIC codes), and loan type. Standard errors are adjusted for heteroskedasticity and clustered by acquirer. Table 6 reports estimated coefficients and levels of significance. The coefficient on the main variable interest (TNIC HHI) shows that loan cost is higher for the bidders with higher product market competition than for others, with a difference of about 25% of the average spread. Considering that the average Loan spread is about 220 bp, this constitutes roughly a 55bp higher cost of loans for the acquirers who face high product market competition. 15 As in Graham, Li, and Qiu (2008) and Engelberg, Gao, and Parsons (2012). 21

Replacing TNIC HHI with alternative measures yields similar results: TNIC similarity and Inverse Lerner index. Overall, results show a positive relationship between product market competition and loan cost. *** Insert Table 6 about here *** 3.f. The impact of backdating on bond spreads in regression analysis The loan cost results show that private debt is an expensive way of financing when acquirers face high product market competition. I question whether this effect is mainly due to information rent extraction by lenders. Bharath, Sunder, and Sunder (2008) show that cost of both types of debt increase with poor accounting quality. Valta (2012) provides empirical evidence that the default risk from product market competition is priced in loan spread. Driessen (2004) show that default risk is also priced in bond spread. If the loan spread increase reflects firm s accounting quality or default risk, I expect that the bond spread should also capture the effect. However, if the effect mainly captures lenders rent extraction from the acquirers that are facing high disclosure cost, I expect to see no significant result in bond pricing. To explore this issue, I test the impact of product market competition on bond spread in regressions analysis. In the model, the response variable is logged Bond spread, measured as the difference between yield to maturity of a newly issued bond and yield to maturity of a riskfree bond reported in Thomson Reuters Securities Data Company (SDC) Global New Issues database. The main variable of interest is TNIC HHI. I control for bond characteristics, including Bond size, Bond maturity, Callable bond (Binary), Puttable bond (Binary), and Bond rating, as defined in Table 1. I further control for firm characteristics, M&A deal characteristics, and macroeconomic factors as in Table 6. I also include fixed effects for years, industries (2-22

digit SIC codes), and bond type. Standard errors are adjusted for heteroskecdasticity and clustered by acquirer. Table 7 presents the results. I find no statistically significant relationship between cost of public debt and product market competition. The result is qualitatively similar when I replace TNIC HHI with alternative measures of product market competition: TNIC similarity and Inverse Lerner index. Overall results on cost of private and public debt provide evidence that acquirers avoiding public debt financing because of high disclosure cost in a competitive environment suffer from private lenders rent extraction. *** Insert Table 7 about here *** 3.g. Joint determination of price and non-price terms of loans I recognize that price and non-price terms of loans are possibly co-determined during the negotiation process between the borrowers and the lenders. As suggested in Bharath et al. (2011), I employ an instrumental variable (IV) approach to address this concern. I find that the results on loan spread continues to hold in the IV framework. The methodology and related results are reported in the Appendix. 3.h. The impact of product market competition on loan spreads for relationship lenders Extant literature provides negative view about relationship lending by documenting that borrowers might be locked in to relationships with lenders, which might not necessary lead to lower loan costs (Sharpe, 1990; Rajan, 1992). 16 The literature also shows that because of 16 A conflicting view posits that prior lending relationship between a lender and a borrower can mitigate adverse selection and moral hazard costs by producing durable and reusable information that leads to better terms of loans for the borrowers (Boot, 2000; Bharath et al., 2011). 23

increases in borrowers lender switching costs, relationship lenders have greater opportunities to extract rents from their locked-in borrowers (Sharpe 1990; Greenbaum, Kanata, and Venezia, 1989; Schenone, 2010). The results from previous section show that acquirers in a competitive product market where disclosure cost is high have less bargaining power over private lenders because of their limited financing alternatives. Collectively, I conjecture that the acquirers facing high product market competition would suffer more from bank s rent extraction when they heavily rely on relationship banks. To explore this issue, I follow Schenone (2010) and measure the intensity of the lending relationship between the lead arranger of a current loan deal and the borrower firm. By looking back at the acquirers borrowing history, I construct Captive borrowing (count), measured as the number of prior lending by the lead arranger scaled by the number with aggregate number of lending. This measure proxies for the intensity of relationship between the lead arranger and the borrowing firm and also implies the strength of the firm s reliance on the bank. 17 Table 8 presents the results. The variable of interest in the model is the interaction term TNIC HHI Captive borrowing (count). If the increase in loan costs resulting from high produce market competition is partly driven by the relationship lenders rent extractions, I would expect to see a positive coefficient associated with the interaction term. In column 2 of Table 9, I replace Captive borrowing (count) with Captive borrowing (amount) as an alternative measure of relationship intensity as in Schenone (2010). Captive borrowing (amount) is measured as the amount of prior lending by the lead arranger up to date scaled by the total amount of borrowing in US dollar. I control for loan characteristics, firm characteristics, M&A deal characteristics, 17 As in Schenone (2010), I calculate the intensity of lending relationship at parent bank level. 24

and macroeconomic factors as in Table 6. I also include fixed effects for years, industries (2- digit SIC codes), and loan type. Standard errors are adjusted for heteroskecdasticity and clustered by acquirer. Consistent with the conjecture, the results (see Table 8) show that the locked-in acquirers in a competitive product market suffer from higher loan costs by being charged with additional 46% of the average spread compared to those that rely less on relationship lenders. In contrast, I find that the borrowers that rely less on relationship banking do not face high loan costs in a competitive environment. The results are qualitatively similar when Captive borrowing (count) is replaced with Captive borrowing (amount). *** Insert Table 8 about here *** 3.i. Premiums Aktas, de Bodt, and Roll (2010) argue that competitive pressure from threat of rival bids can increase bid premium paid for target firms. I expect that if private debt is better at mitigating information leakages to potential bidders than public debt, bid premiums paid in M&A deals should be less for loan financed deals than for bond financed deals. Furthermore, Akdogu (2011) argues that an acquirer is more likely to overpay for targets when cost of losing its target to industry rivals is high. Based on the theoretical evidence, I further expect that the difference in bid premiums between loan and bond financed deals is more pronounced in a more competitive product market in which cost of losing the target to industry rivals is higher. To examine the difference in bid premiums between loan and bond financed deals, I test premium paid in M&A deals in a regression framework. The response variable is logged Proxy premium. Similar to the methodology used in Officer (2007) and Harford, Humphery-Jenner, and Powell (2012), I employ a portfolio approach and group M&A deals with public targets in 25

a 3 year window (one year before and one year after the announcement). I keep those M&A deals that have deal value excluding assumed liabilities lower or equal to 20% of the deal value excluding assumed liabilities of unlisted targets. I calculate the average of the 4-week premium (PREM4WK in SDC) and use it as the premium paid in the M&A deal for the observations with missing premium information. The variable of interest is Private which equals one if the M&A deals is loan financed and zero if the deal is bond financed. I also control for firm characteristics and M&A deal characteristics as in Table 3. Standard errors are adjusted for heteroskedasticity and clustered by acquirer. In column 2 and 3 of Table 9, I subset the sample based on intensity of competition using sample median TNIC Herfindahl index. The results (see column 1 of Table 9) show that Proxy premium is less for loan financed M&A deals than for bond financed deals and that the difference is about 57% of the average proxy premium paid for the target. The results for subsamples (see column 2 and 3 of Table 9) show that the difference in proxy premium between the two types of deals is more pronounced in a more competitive environment. Overall, the results on the M&A bid premiums are consistent with the conjecture that loan financing contains less risk of information spillover. *** Insert Table 9 about here *** 4. Conclusion This paper examines the impact of product market competition on debt financing in M&A. While prior literature on debt contracting mostly focuses on firm level characteristics, I provide a new insight into the link between industry dynamics and debt structure. In addition, I extend the researches that focus on disclosure cost of public financing (Bhattacharya and Ritter, 1983; Rajan, 1992; Bhattacharya and Chisea 1995; Maksimovic and Pichler, 2001; 26