Loan Financing Cost in Mergers and Acquisitions

Loan Financing Cost in Mergers and Acquisitions Ning Gao, Chen Hua, Arif Khurshed The Accounting and Finance Group, Alliance Manchester Business School, The University of Manchester Version: January, 2018 Abstract We investigate loan price in mergers and acquisitions (M&As), using hand-matched loan information for a sample of 330 U.S. M&A transactions. We find the loan price measured by the all-in-drawn spread (AIDS) increases significantly with the relative size of a deal and decreases with the proportion of stocks offered in the consideration. These results are robust to several specifications that address endogeneity concerns. We posit that deal size is a major concern for lenders because it involves more uncertainties, greater business complexity and greater integration difficulties. Further, the contingent pricing mechanism built in stock offers significantly relive the lenders concerns. Our study also suggest the characteristics of major corporate investment impact loan price.

Loan Financing Cost in Mergers and Acquisitions Abstract We investigate loan price in mergers and acquisitions (M&As), using hand-matched loan information for a sample of 330 U.S. M&A transactions. We find the loan price measured by the all-in-drawn spread (AIDS) increases significantly with the relative size of a deal and decreases with the proportion of stocks offered in the consideration. These results are robust to several specifications that address endogeneity concerns. We posit that deal size is a major concern for lenders because it involves more uncertainties, greater business complexity and greater integration difficulties. Further, the contingent pricing mechanism built in stock offers significantly relive the lenders concerns. Our study also suggest the characteristics of major corporate investment impact loan price.

1 Introduction Loan pricing is a central issue in the banking literature. Measured by the All-In-Drawn Spread (AIDS), loan price is underpinned by various risks, including credit risk, market risk and liquidity risk (Gatev and Strahan, 2009). In particular, extent literature has identified several factors that determine loan price, including information asymmetry induced moral hazard between the lead and participant banks in a loan syndication (Ivashina, 2009), lenderborrower previous lending relationship (Bharath, Dahiya, Saunders, and Srinivasan, 2011; Boot, 2000a), previous syndication relationship between the lead and participant banks Bharath et al. (2011), the borrower s organizational structure (Aivazian, Qiu, and Rahaman, 2015) and the borrower s accounting quality (Bharath, Sunder, and Sunder, 2008). These studies emphasize the importance of borrower characteristics and the relations among lenders and borrowers in determining loan price. The authors assume that loans are homogeneous across difference purposes within a borrower and the characteristics of individual investment projects do not matter. This is understandable as canonical text book teaching maintains that debt financing is arranged against a firm s existing portfolio of assets and the cost of debt is determined at the firm level rather than the project level (Brealey, Myers, Allen, and Mohanty, 2012). 1 Nevertheless, M&As are arguably the largest and most complex type of corporate investment. They have profound impact on both a company s growth opportunities and on the value of its existing assets. A sizable proportion of bank loans is made to finance mergers and acquisitions (M&As). For example, about 15% of the syndicated loans recorded in the Dealscan database are used for M&As during 1986 to 2003 (Bharath, Dahiya, Saunders, and Srinivasan, 2011), equivalent to an estimated total amount of 6.213 trillion dollars. 2 In the current study, we aim to examine whether certain M&A characteristics significantly impact loan price, provided M&As are large and complex business that affect a firm profoundly. Our analysis are guided by two primary lines of thoughts. First, previous literature suggests the size of an M&A deal relative to the bidder size (i.e., relative deal size) is associated with multiple merger-related risks. Alexandridis, Fuller, Terhaar, and Travlos (2013) suggest that deal size proxy for unobserved complexity of a transaction which negatively impact the acquiring firm s shareholders value. Datta (1991) postulate that post-transaction integration involve substantial costs to the merging firms. Indeed, larger deals involve more 1 In contrast, the cost of equity is measured according to the security market line specific to a project. 2 This amount is likely to be an under estimate, because, to calculate this amount, we multiply the average size of all syndicated loans by the number of syndicated loans for M&A. However, the loans for M&A are usually larger than those for other purposes. 1

problems in ex post integration and greater cost to achieve the expected synergies in the years post acquisitions. Larger deals also increases the difficulty of due diligence from both the acquirers s and the lender s perspective. It is more difficult to assess the value and risk of the incremental cash flows from a larger deal. In line with this argument, Faccio and Masulis (2005) use the relative size as a measure of information asymmetry between the acquirer and the target. To trace and measure the risks underlying the relative size of a deal can be costly for lenders. It also involves great inaccuracy in the forecasting. However, the relative size is readily available and easy to measure. Therefore, we hypothesize that banks take the relative size into consideration when negotiating loan price. A larger deal involves greater risks and has higher loan price. The second line of thought is based on the measures taken by the bidders to mitigate those merger related risks for bidder shareholders. In particular, Hansen (1987) and Fishman (1989) postulate that there is a substantial level of information asymmetry on both the bidder s and the target s value. A bidder can contingently price a deal by offering stocks to the target shareholders. Because this contingent pricing mechanism prices the target in all states of economy in the future, it forces the target shareholders to share the risk associated with the future cash flows from the combined firm and mitigates the risks accrued to the bidder shareholders. The bidder is less likely to overpay under a stock offer. A stock offer also aligns the interests of the bidder and target shareholders, making the post-transaction integration less costly. Therefore, we expect stock offers to significantly lower loan price in merger transactions. We are not aware of any study that examines loan price in the context of M&A transactions. This is probably due to the difficulty linking loan data to specific merger transactions. Indeed, we find there is no data readily usable and it has to be hand collected. We design a rigorous procedure to link loan facilities from Dealscan to merger transactions from the SDC database (more details in Section 3.2). We gathered 330 U.S. M&A transactions announced during 1993 to 2013 with loan financing information. We find the relative deal size is positively and significantly associated with higher average AIDS of the loans, ceteris paribus. A one standard deviation increase of the relative deal size increases the AIDS by 13.08% of its arithmetic mean. Moreover, the bidders that include stocks in the consideration are charged significantly lower AIDS than those only offer cash. A one standard deviation increase in the proportion of stocks offered in the consideration reduces the AIDS by 8.7 percent of its arithmetic average. To ensure our results are robust to various endogeneity issues, we further subject our analysis to alternative specifications, including the Seemingly 2

Unrelated Regressions (SUR), the simultaneous equation system and the Heckman two-stage self-selection-robust procedure. Our results persist in all these alternative specifications. We contribute to two strands of literature. First, we examine the loan financing cost of merger transactions. Financing is a prime issue in M&A transaction, as important as valuation. The cost of loan financing directly impact the potential of merger gains. Previous literature, however, mainly focuses on the determinants or consequences of the means of payment (e.g., Faccio and Masulis (2005), among others). There are only a few papers that broach the financing issues of M&A transactions. Schlingemann (2004) finds the sources of finance significantly impact the cross section of bidder gains; Bharadwaj and Shivdasani (2003); Martynova and Renneboog (2009) find bidders using more debt to finance cash offers obtain higher gains; Martynova and Renneboog (2009) also examine the determinants of debt financing in merger transactions. Differing from these few previous studies, we investigate how the cost of loan financing is determined by the characteristics of M&A deals. We find the relative deal size and the contingent pricing mechanism built in stock offers are two factors that persistently impact loan price. Second, we contribute to the literature on loan pricing by demonstrating that the features of major corporate investment determines the loan price. Extent studies have emphasized the relation between lenders and borrowers (Bharath et al., 2011; Boot, 2000b), the relation among syndication partners (Ivashina, 2009), the borrower s organizational structure (Aivazian et al., 2015) and the accounting quality of the borrowing firms (Bharath et al., 2008). These studies treat loans for different purposes within a borrower as homogeneous. We find that, for major corporate investments like M&As, the transaction characteristics also impact loan prices after controlling all the determinants highlighted in the previous literature. In merger transactions, both the risks associated with the scale of the transaction and the bidder s effort to counter act those risk matter. The remainder of the paper is organized as follows. Section 2 describes the variable selections and the specification of econometric models. Section 3 describes the data sampling procedures, and section 4 reports the results of analysis. Section 5 is the conclusive remarks. 2 Variables and Econometric Specification 2.1 Variables A The loan financing cost We use All-in-drawn spread (henceforth, AIDS) provided by Dealscan as the measurement of the loan financing cost. Dealscan defines AIDS as the amount the borrower pays in basis 3

points over LIBOR for each dollar drawn down. It adds the spread of the loan with any annual (or facility) fee paid to the bank group. In other words, AIDS includes all monetary cost of a loan facility. One M&A can be funded by multiple loan facilities, thus we use the average AIDS in subsequent analysis. Specifically, we use the arithmetic average AIDS or the weighted average AIDS (weighted by the amount of each facility), however, the results are similar. We only report the results based on the arithmetic average AIDS for the brevity. The results based on the weighted average is available at request. B Relative size of the target firm Following Alexandridis et al. (2013) and Vermaelen and Xu (2014), we use the relative size of the transaction to the acquiring firm to investigate how the size of target firm relative to the acquirer affects the loan financing cost. But our sample covers the M&As with both public and private target firms, therefore, we calculate the relative size as the fraction of the transaction value to the acquiring firms market value at the end of last fiscal year. In addition, we also use the fraction of the transaction value to the acquirer s book value of total assets at the end of last fiscal year, and we get qualitatively same results. C Contingent pricing variables We use two variables to capture the effect of contingent pricing. One is a dichotomous variable, which equals 1 if the acquiring firm pays by the combination of cash and stock, 0 otherwise. Another one is the fraction of the stock s value in the consideration to the transaction value. The reason of selecting two variables is that we find the disclosed means of payment of some transactions are different from what is recorded in SDC M&A database, when we manually check the M&A information in acquiring firms 10-Qs and 10-Ks filings. So, we encode the means of payment binary variable depending on the disclosed information. 2.2 Econometric specifications We use seemingly unrelated regressions (henceforth, SUR) model to estimate the main resutls. Prior literature argues that price term and non-price terms (maturity and collateral) are simultaneously determined within a loan facility (e.g. Bharath et al. (2011), Aivazian et al. (2015). Thus, they use loan facility level data and structural equations model (henceforth, SEM) to estimate the consistent mutual impacts among loan terms. However, in our M&As sample, two-third of the transactions are funded by more than one facility, which is presented in descriptive statistic results latter. Since we investigate the loan financing 4

cost at the M&A transaction level, we take the averaged values of AIDS and maturity of all facilities for one M&A transaction if the M&A is funded by multiple facilities. The loan facility level variables are converted to the M&A level. Thus, the loan terms variables in our data perhaps do not influence others structurally but statistically, assuming the equilibrium in loan terms contracting generated only within the loan facility. Accounting for our specific M&A level data, the SEM may not fit it. Moreover, a loan facility can be used for multiple purposes, but Dealscan database just reports two purposes of a loan facility. But both price and non-price loan terms might be affected by the unobservable purposes, which cannot be controlled for in the regression models. The correlations among loan terms cannot be addressed in SEM. Therefore, we use SUR model, because this model allows error terms of the equations in a system to be statistically correlated, capturing the correlations among loan terms only through error term in each equation. For example, in the SUR model, the dependent variables of three equations are average loan spread, average maturity and collateral dummy variable, respectively. The X i (i = 1, 2, 3) is the vector of independent variables for each equation. All dependent variables do not appear on right-hand side of the equations, but ɛ i can be correlated. Spread = X 1 α + ɛ 1 Maturity = X 2 β + ɛ 2 (1) Collateral = X 3 γ + ɛ 3 The independent variables of three equations cannot be totally same, otherwise it will generate same estimates as OLS does. Therefore, we use several instrumental variables proposed by Bharath et al. (2011) for each equation. In loan spread equation, we include acquiring firm s ratio of EBITDA to sales, current assets ratio, the natural logarithm of interest coverage and market-wide default spread before the M&A announcement. EBITDA to sales ratio and current assets ratio are related to future cash flow, with higher ratios suggesting more cash inflow in future, thus decreasing the likelihood of default. Interest coverage directly reflects the acquiring firm s ability to repay the loan. Market-wide default spread could affect the cost of loans in the whole market. All of them are likely to affect loan spread, but they are unlikely to affect the non-price terms in the loan contract. Consequently, we include them as the independent variables in the loan spread equation. We also control for the characteristics of the acquiring firm, the M&A, and the loan contract as well as some additional variables. The detail description of variables can be found in the appendix. In the maturity equation, the additional variable is the natural logarithm of acquiring 5

firm s asset maturity. Hart and Moore (1994) argue that the firm attempts to match its debt maturity to the economic life of the assets, therefore, the firm s assets maturity may affect the choice of maturity of new debts, which is unlikely to influence other loan terms. For collateral equation, we add loan concentration and industry-level tangibility ratio to distinguish the independent variables vector from another two equations. The former one is based on the finding in Berger and Udell (1995) that larger new borrowing relative to the total debt is positively associated with the likelihood of being asked for collateral. The latter one is based on argument in Bharath et al. (2011) that companies in the industries with more tangible assets are more likely to provide collateral. Beside the estimates of SUR model, we also estimate OLS and SEM for robustness., using transaction or loan facility level data. 3 Data and Sample Selection 3.1 Data sources The M&A sample is obtained from SDC M&A database. We choose completed M&As announced during 1993 to 2013 in US. The reason of choosing 1993 as the start is that online SEC filings of listed companies begins at 1994, which allows us to manually verify whether the M&A is funded by the loans or not in 10-Qs and 10-Ks. In our sample, the acquiring firms are public, but the target firms can be public or private. Subsidiary targets transactions are not included. Additionally, the M&A should have non-missing values with respect to the means of payment, announcement and effective dates, and transaction value. To get the sample of loan financed M&As, we filter the sample with following rules. First, we drop M&As deals without acquiring firms GVKEY, otherwise we cannot match the M&A to the control variables from Compustat and CRSP. Second, we drop M&As without cash in the consideration, since acquiring firms do not need borrow cash to fund the transactions. Third, we keep M&As indicated as funded by the borrowing, bridge loan or line of credit in SDC M&A database. So far, we get candidates of loan financed M&As for subsequent analysis. Next, we identify the sample of loan facilities that can be used for funding M&As. The data of loans are from Thomson Reuters DealScan 3. Basically, a loan contract, called as loan package in Dealscan, consists of one or multiple loan facilities, however, loan facilities within a loan package can contain different price and non-price loan terms for different 3 DealScan collects worldwide loan contract information and it has a qualified coverage on US market. Main information sources of DealScan are SEC filings and other public documents, including 10-Ks, 10-Qs, 8-Ks and registration statements, with supplementary information gathered from lenders and other internal sources. 6

purposes. Leaded by the leader lenders, the participant lenders can also be different for each facility within a loan package. Dealscan provides information about loan contract in terms of a variety of loan contract terms, the role of the lenders and basic information about the borrowers, however, it does not provide widely used identifier code (e.g. GVKEY) for the borrowing firms, which makes loan facilities difficult to match variables of borrowing firms in other databases (e.g. CompuStat or CRSP). We describe data sets matching procedures explicitly in subsequent section. A loan facility can be used for several purposes, but only first and second purpose are listed in DealScan. To find out the loan facilities that are used for funding M&As, we first keep loan facilities which are tagged as Acquis. Line, Merger or Takeover in either first or second purposes 4. Moreover, the loan facilities for Corp. Purpose are also possible for funding the M&As. For example, Sykes Enterprises Inc. acquiring ICT Group, Inc. is announced on October 6, 2009 and completed on February 10, 2010, with per issued and outstanding share of ICTG will be converted into 7.69 in cash and 7.69 in Sykes per stock. A loan facility recorded in DealScan which is lent to Sykes Enterprises Inc. and begins on December 11, 2009 is tagged as for corporate purpose or working capital. Arguably, this loan facility is very likely to be used paying for the transaction. Thus, we also keep loan facilities for Corp. Purpose. Then, we verify whether an acquiring firm finances the M&A transaction with loans by manually reviewing 10-Qs and 10-Ks in SEC online filings. More detail procedures of manual verification are described below. 3.2 Matching procedures and verifying the quality of matching We match the M&As from SDC M&A with loan facilities from Dealscan using a Compustat- Dealscan linkage file provided by Chava and Roberts (2008). Since the linkage file covers the loan facilities those of which begins between January 1983 and August 2012, we extent the sample from September 2012 to December 2014 5. Next, we match the M&A transactions funded by loans to corresponding loan facilities. Specifically, we separate the loan facilities into two subsamples: one is facilities with target name in DealScan (2280 facilities), and another one is facilities without target name in DealScan (10057 facilities). For the former subsample, we generate all possible pairwise linkages by GVKEY of the borrowing firm. Then, we drop incorrect linkages by restricting a rule of dates: beginning date of the facility should not be later than the effective date of the M&A, and not earlier than 7 days before 4 For the detail description of the tags of loan purposes, please see the appendix 5 We apply fuzzy matching technique via creating linkages and spelling distance measurement to help manually check for merging. 7

announcement date of the M&A 6. This is step makes the loan facilities matched to the closest announced M&As, reducing the likelihood of mismatching between the loan and the M&A. Then, we identify the correct linkage of M&As and facilities by manually comparing the target firms names from Dealscan and SDC. For loan facilities without target names in DealScan, we use GVKEY as the identifier to match and apply the 7-day criterion to exclude improper linkages. So far, 433 M&As are matched to loan facilities. In order to guarantee that the acquiring firm indeed finances the M&A by loans, we manually search for SEC 10-Ks and 10-Qs filings of acquiring firms and check whether loan financing in M&As is disclosed. Specifically, using the acquiring firm s CIK code 7, we search for 10-Qs and 10-Ks of acquiring firms that filing dates are just subsequent to the completion dates of the M&As. Next, we search for target firm s name in 10-Qs and 10-Ks. Since our loan facility sample includes those are for general corporate purpose, which might be not for funding the M&A but mismatched to the M&A, this step ensures the matched loan facility is most likely to fund the M&A. Finally, we get 330 out of 433 M&As verified as loan financed M&As. This sample size is comparable to prior empirical studies on debt financing in M&As (e.g. 155 cash tender offers in Bharadwaj and Shivdasani (2003) and 607 M&As with pre-merger debts financing (measured by the change in assets net of the book value of equity divided by assets) in Schlingemann (2004)). 4 Empirical Analysis 4.1 Sample distribution and summary statistics Table 1 presents the distribution of our sample along four dimensions. Panel A presents the distribution of M&As across announcement years. It shows that the trend of the number of transactions announced fluctuates with the economy. For instance, more M&As are announced during economy boom over 1997 to 1999, taking approximately 30% of the sample, while less transactions are announced during the recession from 2008 to 2011. Panel B reports that 91% of the acquiring firm have one M&A observation in our sample, and 9% of the acquiring firms have multiple M&As. Panel D shows that 94% of loan facilities with the inception date between the announcement date and effective date of M&As. 6 To guarantee the plausibility of date criteria and the quality of matching, we manually search borrowing firms names and targets firms names in Factiva and review the news in Reuters and Wall Street Journal around the facility starting date. Finally, for M&As that are found in Factiva, no facility starting date is later than effective date of the M&A and the earliest facility starting date is 6 days before announcement date of the M&A. So this 7-day criterion is proper. 7 As SDC M&A database does not provide CIK but CUSIP, we add CIK from Compustat to the M&As sample, using CUSIP as the identifier. For several unsuccessful CIK-CUSIP matching, we search for the names of acquiring firms on EDGAR website to get the correct CIK code. 8

[Insert Table 1 Here] Table 2 reports the summary statistics of variables. In terms of loan financing cost, arithmetic average AIDS and weighted average AIDS have very similar distribution in our sample. Both means are approximate 184 basis points higher than LIBOR. The natural logarithm of these two variables have almost same mean and median. The mean and median of relative size ratio are 0.676 and 0.454, respectively. For the means of payment, 31.5% of transactions in our sample use mixed means of payment, while 68.5% are paid by cash only. The mean of the proportion of stock used in the consideration is 10.4%, which shows that more cash is used than stock in the consideration for loan funded M&As 8. 97.6% of M&As are funded by lenders who have previous lending business with the acquiring firm. 69.4% of M&As in our sample are offered performance pricing terms in loan contracts. 36% of transactions take place between two firms from different Fama-French 48 industries, and 72.4% of target firms are publicly listed. 9.7% are cross-border M&As. 5.2% are tagged as hostile deals, and tender offers takes 37.6% of M&As in the sample. Moreover, Table 2 also presents the characteristics of acquiring firms and loan contracts. Acquiring firms in the sample are large firms, with the mean value of total assets reaching 812 million dollars. Acquiring firms mean of leverage is 0.238. The average Altman Z-score is larger than 4, reflecting low bankruptcy risks. On average, market-to-book ratio is 2.32, and EBITDA is 18.4% of total sales. 28.9% of total assets of acquiring firms are tangible assets, and current assets are about twice as many as current liabilities. The mean of credit rating score are 5.82. Additionally, the mean of natural logarithm of 1 plus EBITDA to interest expenses is 2.56. In terms of loan contract characteristics, the average loan facility amount is 411 million and average maturity is more than 36 months. 60% of acquiring firms are required to offer collateral to secure loans. 77% of M&As are funded by loan facilities with financial covenants restrictions in loan contracts. Notably, about 95% of M&As are funded by syndicated loans. [Insert Table 2 Here] 4.2 Univariate analysis Table 3 presents univariate analysis on arithmetic average AIDS against variables of interests. Acquiring firms which acquire relatively larger targets are charged statistically 8 We should point out that the means of payment disclosed in 10-Qs and 10-Ks are slightly different from that provided by SDC M&A database. When we review 10-Ks and 10-Qs, we find that the number of transactions recorded as pure cash payment transactions in SDC M&A database is a little smaller than that disclosed in the filings. Previous literature also points out potential incorrect information in SDC M&A database (Faccio and Masulis, 2005), therefore we cannot judge conclusively which is correct. 9

significant higher spread than those acquiring relatively smaller targets, which is consistent with our hypothesis that the larger target firm relative to the acquiring firm in the M&A is associated with higher loan financing costs. In terms of contingent pricing effect of paying by stock in the consideration, the table shows that the average AIDS for transactions paid by only cash is 23 basis points smaller than those paid by the combination of cash and stock and the difference is significant at 5% level. Similarly, transactions paid by more stock in the consideration are charged higher spread than those paid by less stock, but the difference is insignificant. These results contradict our hypothesis that using stock can decrease the loan financing cost, however, we argue that those are likely to be driven by the fact that more stock are used to purchase larger target. The correlation coefficients of mixed payment dummy and the proportion of stock in the payment are positive and significant at 1% level. Therefore, we present the marginal effect of mixed payment dummy and the proportion of stock in the consideration in regression analysis after controlling for other variables. Table 3 also reports the differences of average AIDS against a group of M&As characteristics. No statistically significant different on average AIDS between diversifying deals and focusing deals, but the M&As of acquiring public targets are charged less spread than acquiring private targets, and the difference is statistically significant at 1% level, which is coincident with our hypothesis that because of less information asymmetry and the likelihood of overpayment, the loan financing cost of public target M&As is lower than private target M&As. Cross-border and hostile deals get lower average AIDS, however, the statistical significance might be a problem because of few observations in each group (32 and 17 respectively). Additionally, the result shows that tender offer transactions get lower spread and the difference is significant at 1% level, which is likely to imply that, for lenders, the effect of speed of completion outweighs the effect of the likelihood of overpayment, decreasing the probability of acquiring firm s default. Next, we presents the empirical results in multivariate regression analysis. [Insert Table 3 Here] 4.3 Multivariate analysis Column (1) of Table 4 reports the results of seemingly uncorrelated regression model using M&A transaction level data. The coefficient of relative size ratio is 0.218 and significant at 1% level, with one standard deviation increase of relative size ratio (0.6) enhancing 13.08% arithmetic average AIDS increase. This estimated coefficient indicates that loan financing cost is higher for acquiring firms if the size of target firm is relatively larger to the size of 10

the acquiring firm. In terms of contingent pricing effect, the coefficient of mixed payment dummy variable is 0.142 and significant at 1% level, which means, on average, acquiring firms which pay by the combination of cash and stock are charged 13.2% less spread than those pay by only cash in the consideration, after controlling for other factors. This result is consistent with our hypothesis that using stock in the consideration can lower the loan financing cost in the M&A. Alternatively, we use a continuous variable, the proportion of stock used in the consideration, to estimate the equations system, which is presented in column (2). The coefficient of the proportion of stock used in the consideration is also negatively significant ( 0.408) at 1% level, with one standard deviation increase decreasing 8.7% arithmetic average AIDS, which means more stock used in the consideration decrease more loan financing cost in the M&A. Additionally, acquiring firms characteristics and loan contract characteristics may also affect the loan financing cost. Larger acquiring firms are charged statistically significant less spread, while higher leveraged acquiring firms are charged higher spread. Higher market-tobook ratio reflects better investment opportunities, associating with loan spread negatively and significantly. If acquiring firms have more tangible assets, the loan spread will be significantly lower. We do not document a significant impact of acquiring firms bankruptcy risks on loan financing cost, which is likely to be attributed to a fact that acquiring firms are usually large and financially healthy firms with small risks in terms of bankruptcy. The positive and significant coefficient of credit rating score shows the worse credit ratings are associated with higher loan financing cost. For characteristics of M&As, we do not get their statistically significant impacts on the average AIDS after controlling for other variables. The coefficient of Ln(Facility amount) is statistically significant, which is contradict to prior evidence, however, we argue that this positive coefficient is likely to reflect that larger facility amount might associated with larger transactions. The correlation between relative size ratio and Ln(Facility amount) is 0.141 and significant at 5% level, supporting this argument to some extent. The coefficient of financial covenants dummy variables are significantly positive at 10% level, which is likely to reflect that lenders use financial covenants to monitor acquiring firms behavior to decrease the agency cost. The coefficient of syndicated loan dummy is not significant, which might be attributed to insufficient variation of non-syndicated loan funded M&As in our sample. Moreover, higher interest coverage and lower market-wide default spread are also significantly associated with spread negatively, but profitability (EBITDA/Sales) and current ratio do not show significant impact on spread in our sample. For relationship lending, the coefficient is significantly negative ( 0.364) at 5% 11

level, which means the loan financing cost will be lower for acquiring firms who borrow cash from lenders with previous lending business. However, we should be careful to explain its statistical and economic significance because only 8 M&As in our sample funded by nonrelationship lending, casting doubts on statistical power on the significance. The coefficient of performance pricing dummy variable is 0.173 and significant at 1% level, which means that performance pricing terms could reduce the agency cost and thus decrease the loan financing cost. In column (1) and (2), we control for industry and year effects by adding dummy variables, however, it is possible that some specific industries have impacts on loan financing cost in specific years. Therefore, we control for industry and year effects by adding industry times year dummy variables in the regressions. The results of spread equations are presented in column (3) and (4), and we get very similar estimated results as those in column (1) and (2) in terms of both magnitude and statistical significance for each variable. Furthermore, Breusch-Pagan χ 2 tests at the bottom of the table are all significant at 1 % level, rejecting the null hypothesis that error term of each equation is uncorrelated, which justifies the proper econometric specification for correlated dependent variables. [Insert Table 4 Here] 4.4 Robustness tests To address the concerns of simultaneity of price and non-price terms in loan contracts, we follow Bharath et al. (2011) and Aivazian et al. (2015) to estimate a simultaneous equations model. In the equations system, loan spread is affected by maturity and collateral, while maturity and collateral influence each other but not affected by loan spread. This econometric setting is in light of the decision procedures of loan terms that non-price terms are determined before price terms in loan contract (Ivashina, Nair, Saunders, Massoud, and Stover, 2009). To satisfy the order condition to estimate each equation, we use instrument variables (IVs) proposed in (Bharath et al., 2011). IV of maturity 9 is the natural logarithm of acquiring firms asset maturity 10. For endogenous collateral dummy variable, we use loan concentration and industry level average tangibility ratio as IVs. The procedures of estimating SEM are same as Bharath et al. (2011). We estimate each equation individually using 2-stage least square (2SLS) method. Because collateral dummy 9 Bharath et al. (2011) and Aivazian et al. (2015) also use regulated industry as an IV for loan maturity, however, regulated industry acquiring firms are excluded in our sample. Thus, we cannot use this IV in our model. 10 Asset maturity is the weighted average of current assets divided by the cost of goods sold, and net PPE divided by depreciation and amortization. 12

variable is a binary endogenous variable, we estimate a probit model in first stage and using fitted value as an IV of endogenous collateral dummy to get consistent coefficients in spread and maturity equations. Table 5 reports the second stage results of 2SLS estimate of spread equation. However, the tests of endogeneity of endogenous variables do not support maturity and collateral are endogenous in our sample. Specifically, the p values of Wooldridge robust score and robust regression F are approximate 0.3, which means the null hypothesis that price and non-price terms are exogenous cannot be rejected. [Insert Table 5 Here] We conduct some additional tests in the following part for robustness of the results. First, we use loan facility level data to estimate and results are reported in table 6. Column (1) and (2) are OLS estimates, assuming that maturity and collateral dummy are predetermined. In terms of the impacts of non-pricing loan terms on spread, maturity has no significant impacts on spread but collateral dummy variable is positively associated with higher loan spread. The significance of collateral dummy variable implies that acquiring firms who are required to offer collateral are likely to be perceived by lenders with higher default probability. The more important is that the coefficient of relative size ratio is still positive and statistically significant, and the magnitude is very close to previous SUR results, which means acquiring relatively larger target firms in the M&A is associated with higher loan financing cost. The coefficients of mixed payment and proportion of stock in the payment are negative and significant at 1% level, and the magnitudes are also very close to previous SUR results, supporting our hypothesis that using stock in the consideration can decrease loan financing cost by sharing the risks of acquisitions with target shareholders. The coefficient of proportion of stock used in the payment might be overestimated because we pool pure cash payment (0 stock used in the payment) transactions and mixed payment transactions together. To address this concern, in column (3) we use only mixed payment transactions to estimate. The coefficient is still negative but insignificant. The loss of statistical significance is likely to be attributed to a considerable drop in loan facility observations. [Insert Table 6 Here] We also use loan facility level data to estimate SUR as a robustness test. Column (4) and (5) show the results of spread equation in SUR model. Again, all variables get qualitatively same results as results immediately above. Moreover, we employ 2SLS method to estimate 13

SEM using loan facility level data, with the results presenting in table 7. Again, tests of endogeneity of endogenous variables cannot reject null hypothesis that maturity and collateral dummy are endogenous. [Insert Table 7 Here] Since only loan financed M&A transactions are included in the sample, it leads to an potential self-selection issue that the decision of choosing loans may affect the loan financing cost in the M&A. Therefore, we estimate Heckman self-selection model using M&A level data to test the robustness of coefficients for variables of interest. The results are presented in table 8. We require no missing values in all regressors of two steps of Heckman model, therefore, the number of M&As funded by loans drops from 330 to 230. In the first stage, the determinants of choosing loans to fund the M&A are based on Martynova and Renneboog (2009). Specifically, in first step, the dependent variable is binary, denoting whether a transaction is funded by loans or not, and the independent variables include cash flow to transaction value, long-term debt to total assets, returns run-up, daily abnormal returns, announcement year dummies and institutional ownership Herfindahl-Hirschman Index 11, Tobin s Q, equity beta, firm age, natural logarithm of transaction value. We also control for means of payment dummy variables and intra-industry deal dummy variable at 2-digit SIC level, however, they cannot get point estimates due to collinearity in our sample. The dependent variable of second stage is the natural logarithm of arithmetic average AIDS. All other variables are identical to the definitions in the baseline model, which can be found in variable definition table. Wald tests for Rho equaling 0 cannot be rejected at 10% level, indicating that in our sample, self-selection issue is unlikely to affect the estimated results, and the results are similar to the baseline results in terms of coefficients magnitude and statistical significance. [Insert Table 8 Here] 5 Conclusive Remarks In this paper, we empirically investigate how the size of the target firm relative to the size of the acquiring firm and the usage of stock in the consideration affect the loan financing cost in the M&A. After controlling for other variables, we find that the relatively larger 11 For the impact of control rights, unlike European sample in prior studies, ownership structure of US companies is not as concentrated as European counterparts, therefore, we use institutional ownership Herfindahl-Hirschman Index (Institutional ownership HHI) in Thomson Reuters Database to capture the effects of control rights in our sample. 14

target to the acquiring firm is positively and significantly associated with higher loan spread. This result implies that the large target is more likely to be selected by the overconfident manager in the acquiring firm, which is more likely to be overpaid. Also, it could be due to complexity and unexpected integration cost in the large transaction. All these makes the loan for funding the M&A riskier, thus charged higher loan rate. We also document a statistically significant negative relationship between the loan spread and the usage of stock in the payment. The contingent pricing effect of using stock in the consideration reduces the likelihood of overpayment. It also aligns the interest of target firm s shareholders with the acquiring firm, mitigating the integration cost after the mergers. Therefore, contingent pricing is helpful to reduce the risk of the loan repayment. This paper complements to the extant literature by providing evidence in terms of the financing cost of the M&A, especially providing evidence regarding the creditor s decision in the M&A financing. In addition, we also show that the specific purposes of the loan can affect contracting of loan terms. Our findings imply that, in spite of the positive announcement return of the loan funded M&A documented by previous literature, the managers and shareholders of the acquiring firm should not ignore the loan financing cost, especially for the M&As with relatively large target firms. The certification and monitoring from lenders create value for shareholders of the acquiring firm on the stock market (Bharadwaj and Shivdasani, 2003), but the repayment of the loan is likely to enhance the future cash volatility in the acquiring firm, which may decrease shareholders future returns. In other words, the the loan financing, to some extent, is the cost of positive returns. Although using stock to pay can lower the cost, it is a trade-off between the benefits of control right of the firm and the loan financing cost. References Aivazian, V. A., J. Qiu, and M. M. Rahaman (2015). Bank loan contracting and corporate diversification: Does organizational structure matter to lenders? Journal of Financial Intermediation 24 (2), 252 282. Alexandridis, G., K. P. Fuller, L. Terhaar, and N. G. Travlos (2013). Deal size, acquisition premia and shareholder gains. Journal of Corporate Finance 20, 1 13. Berger, A. N. and G. F. Udell (1995). Relationship lending and lines of credit in small firm finance. The Journal of Business 68 (3), 351 381. Bharadwaj, A. and A. Shivdasani (2003). Valuation effects of bank financing in acquisitions. Journal of Financial Economics 67 (1), 113 148. 15

Bharath, S. T., S. Dahiya, A. Saunders, and A. Srinivasan (2011). Lending relationships and loan contract terms. Review of Financial Studies. Bharath, S. T., J. Sunder, and S. V. Sunder (2008). Accounting quality and debt contracting. The Accounting Review 83 (1), 1 28. Boot, A. W. (2000a). Relationship banking: What do we know? Journal of Financial Intermediation 9 (1), 7 25. Boot, A. W. A. (2000b). Relationship banking: What do we know? Journal of Financial Intermediation 9 (1), 7 25. Brealey, R. A., S. C. Myers, F. Allen, and P. Mohanty (2012). Principles of corporate finance. Tata McGraw- Hill Education. Chava, S. and M. R. Roberts (2008). How does financing impact investment? the role of debt covenants. The Journal of Finance 63 (5), 2085 2121. Datta, D. K. (1991). Organizational fit and acquisition performance: Effects of post-acquisition integration. Strategic Management Journal 12 (4), 281 297. Faccio, M. and R. W. Masulis (2005). The choice of payment method in european mergers and acquisitions. The Journal of Finance 60 (3), 1345 1388. Fishman, M. J. (1989). Preemptive bidding and the role of the medium of exchange in acquisitions. The Journal of Finance 44 (1), 41 57. Gatev, E. and P. E. Strahan (2009). Liquidity risk and syndicate structure. Journal of Financial Economics 93 (3), 490 504. Hansen, R. G. (1987). A theory for the choice of exchange medium in mergers and acquisitions. The Journal of Business 60 (1), 75 95. Hart, O. and J. Moore (1994). A theory of debt based on the inalienability of human capital. The Quarterly Journal of Economics 109 (4), 841 879. Ivashina, V. (2009). Asymmetric information effects on loan spreads. Journal of Financial Economics 92 (2), 300 319. Ivashina, V., V. B. Nair, A. Saunders, N. Massoud, and R. Stover (2009). Bank debt and corporate governance. Review of Financial Studies 22 (1), 41 77. Martynova, M. and L. Renneboog (2009). What determines the financing decision in corporate takeovers: Cost of capital, agency problems, or the means of payment? Journal of Corporate Finance 15 (3), 290 315. Schlingemann, F. P. (2004). Financing decisions and bidder gains. Journal of Corporate Finance 10 (5), 683 701. Vermaelen, T. and M. Xu (2014). Acquisition finance and market timing. Journal of Corporate Finance 25, 73 91.

Table 1: Sample distribution This table presents the distribution of loan funded M&As sample. Panel A shows the distribution of the sample by year of announcements. Panel B tabulates the number of transactions taken by an acquiring firm. Panel C presents the sample distribution by the number of loan facilities used in a transaction. Panel D shows the distribution of timing of loan facilities in terms of beginning dates. Panel A: Distribution of M&A across years Announcement Year of M&As N of M&As Percent % 1993 3 0.90% 1994 14 4.20% 1995 24 7.30% 1996 20 6.10% 1997 38 11.50% 1998 29 8.80% 1999 29 8.80% 2000 11 3.30% 2001 12 3.60% 2002 7 2.10% 2003 10 3.00% 2004 15 4.50% 2005 20 6.10% 2006 17 5.20% 2007 23 7.00% 2008 16 4.80% 2009 1 0.30% 2010 3 0.90% 2011 17 5.20% 2012 12 3.60% 2013 9 2.70% Total 330 100.00% Panel B: Distribution of acquiring firms N of M&As taken by the acquirer N of Acquirers Percent % 1 274 91.30% 2 22 7.30% 3 4 1.30% Total 300 100.00% Panel C: Loan facilities within one M&A deal N of loan facilities used within a M&A N of M&As Percent % 1 111 33.60% 2 115 34.80% 3 64 19.40% 4 28 8.50% More than 5 12 3.60% Total 330 100.00% Panel D: Timing of loan facilities Timing of loan facility starting date N of Facilities Percent % (Ann. Date 7, Ann. Date) 30 4.20% Ann. Date 13 1.80% (Ann. Date, Eff. Date) 671 94.00% Total 714 100.00% 17

Table 2: Summary Statistics This is a summary statistics table for variables at M&A transaction level. See Table 9 for detail variable definitions. N Mean Median Std. Dev. Min Max Arithmetic average AIDS (basis points) 330 184.326 175 96.798 40 381.25 Weighted average AIDS (basis points) 330 183.839 175 96.177 40 371.429 Ln(Arithmetic average AIDS) 330 5.049 5.165 0.624 3.689 5.943 Ln(Weighted average AIDS) 330 5.046 5.165 0.625 3.689 5.917 Relative size ratio 330 0.676 0.454 0.600 0.081 2.332 Mixed payment dummy 330 0.315 0 0.465 0 1 Proportion of stock in the payment 330 0.104 0 0.198 0 0.838 Relationship lending dummy 330 0.976 1 0.154 0 1 Performance pricing dummy 330 0.694 1 0.462 0 1 Diversify deal dummy 330 0.364 0 0.482 0 1 Public target dummy 330 0.724 1 0.448 0 1 Cross-border deal dummy 330 0.097 0 0.296 0 1 Hostile deal dummy 330 0.052 0 0.221 0 1 Tender offer dummy 330 0.376 0 0.485 0 1 Ln(Total assets) 330 6.695 6.572 1.657 3.963 9.780 Leverage 330 0.238 0.218 0.181 0 0.624 Market-to-book ratio 330 2.317 1.876 1.302 1.021 5.879 Altman Z-score 330 4.185 3.647 2.573 1.022 11.272 Credit rating score 330 5.824 7 1.427 2 7 EBITDA/Sales 330 0.184 0.145 0.126 0.038 0.513 Tangibility ratio 330 0.289 0.202 0.236 0.035 0.831 Current ratio 330 2.099 1.874 1.072 0.735 4.904 Ln(Interest coverage) 330 2.557 2.471 1.112 0 6.758 Ln(Maturity) 330 3.844 4.094 0.543 2.565 4.431 Collateral dummy 330 0.603 1 0.490 0 1 Ln(Facility amount) 330 19.833 19.807 1.383 17.504 22.313 Financial covenants dummy 330 0.770 1 0.422 0 1 Syndicated loan dummy 330 0.945 1 0.227 0 1 Market-wide default spread 330 0.022 0.019 0.007 0.013 0.052 Industry level mean of tangibility ratio 330 0.602 0.495 0.436 0.232 6.624

Table 3: Univariate analysis This table presents univariate analysis results on arithmetic average AIDS for variables of interest. T statistics is for mean difference test, and χ 2 is for median equality test. ***, ** and * indicate the significance at 0.01, 0.05 and 0.1, respectively. Dummy=0 / Lower than sample mean Dummy=1 / Larger than sample mean Variable N Mean Median N Mean Median T χ 2 Size of the target Relative size ratio 210 165.366 150.000 120 217.506 215.938 4.87*** 22.72*** Contingent pricing Mixed payment dummy 226 176.970 163.673 104 200.310 200.000 2.04** 4.13** Proportion of stock in the payment 247 179.436 175.000 83 198.878 200.000 1.59 2.14 Characteristics of M&As Diversify deal dummy 210 182.282 175.000 120 187.902 175.000 0.51 0.04 Public target dummy 91 219.173 204.690 239 171.058 150.000 4.13*** 15.32*** Cross-border deal dummy 298 187.956 179.375 32 150.518 150.000 2.09** 4.21** Hostile deal dummy 313 185.447 175.000 17 163.674 150.000 0.90 0.38 Tender offer dummy 206 204.187 200.000 124 151.331 129.694 4.97*** 18.91***