Creditor control rights and borrower protection: the role of borrower consent clause in private debt contracts

Abstract

We investigate how borrower consent clause (BCC) is used in private debt contracts as a contract design mechanism to protect borrower interests. We find that the probability of including a BCC in debt contracts increases in the intensity of creditor control rights measured by number of financial covenants. Furthermore, we document that performance covenants result in higher likelihood of BCC inclusion than capital covenants do. For robustness checks, we use alternative proxies for creditor control rights, and employ simultaneous equation and propensity score matching to address endogeneity. The baseline results still hold. Exploiting Anti-Recharacterization Law (ARL) as a quasi-natural experiment for strengthened creditor rights, we find that adoption of ARL increases the likelihood of BCC inclusion. Using credit default swap (CDS) trading as a setting of weakened creditor control rights, we document the inception of CDS trading is associated with lower likelihood of including BCC. Furthermore, we find that the association between creditor control rights and BCC is more pronounced for borrowing firms with good quality and more conservative financial reporting.

1 Introduction

Borrower consent clause (BCC, hereafter) is a contracting term in bank loan contracts that requires creditors to seek the borrowers’ consent before transferring some or all of a loan to a third party.¹ On the one hand, practitioners suggest that “borrowers are keen to maintain control over the composition of their syndicate as far as possible, so they know who they are dealing with at any given point” (Singleton 2017). On the other hand, BCC inclusion can reduce the flexibility of creditors to transfer their loans to a third party and thus impair loan trading liquidity (Loan Market Association 2017). Anecdotal evidence shows that it is common that borrowers refuse to grant consent to creditors who intend to assign their loans to hedge funds or similar investment vehicles (Kibbe et al. 2010). Overlooking this clause in loan contracts may result in nullification of loan transfer and even lawsuits (Carey 2018).

Despite the implications of BCC for both creditors and borrowers, existing literature mostly focus on creditor control rights, with little attention paid to BCC in debt contracting. Kamstra et al. (2014) and Pyles and Mullineax (2008) are two notable exceptions. Both studies use a sample of corporate loans prior to 2004, and find that loans with sale constraints (i.e., BCC) on average charge higher interest rates. However, neither study addresses a more fundamental question regarding why BCC is included, the circumstances under which lenders are willing to accept such a costly clause, and the equity market consequences of BCC inclusion for borrowing firms. We tackle these questions in this study to fill the gap in the literature, by using a sample of 5963 unique borrowing firms over the sample period of 1993–2015.

Why do borrowers use BCCs? We argue that borrowers include BCCs to reduce potential renegotiation costs due to the uncertainty in dealing with unknown creditors, especially those with strong creditor control rights. Because it is impractical to specify contracting parties’ rights and obligations in all future state contingencies, incomplete contracting theory suggests that contracting parties tend to use noise but contractible signals (i.e., debt covenants) to allocate control rights, ensuring joint surplus and optimal contracting for them (Aghion and Bolton 1992; Bolton and Scharfstein 1996). A growing literature show that debt contracts include a wide range of financial and non-financial covenants (Dichev and Skinner 2003; Bradley and Roberts 2015), and covenant violation leads to creditors gaining control rights, which influences borrowers’ corporate governance, investment, and financing decisions even in the absence of a payment default (Chava and Roberts 2008; Roberts and Sufi 2009a; Nini et al. 2012; Denis and Wang 2014).

While contingent allocation of control rights may benefit both contracting parties, frequent renegotiation and covenant violations also imply higher renegotiation costs and may become a concern for borrowers, especially when the contingent control rights can be transferred to a third party via loan assignments. Under this circumstance, borrowers face uncertainties regarding who gains the contingent control rights and with whom they will be dealing with during future renegotiations.² This issue is particularly important given that banks have increasingly adopted the “originate-to-distribute model” in their lending business in the recent decades (Bord and Santos 2012), and the likelihood of loan sales and the transfer of creditor control rights therein have increased. Furthermore, existing literature suggests that creditors gain stronger contingent control rights ex ante when information asymmetry is high (Gârleanu and Zwiebel 2009; Hollander and Verriest 2016; Prilmeier 2017) and when borrowing firms want to signal their quality (Demiroglu and James 2010). However, whether borrowers use BCCs in debt contracts to address their concerns about the uncertainty associated with ex post loan transfer remains an empirical question. Accordingly, in our first hypothesis, we predict that the intensity of creditor control rights is associated with higher likelihood of BCC inclusion in debt contracts (creditor control rights hypothesis).

BCC inclusion is costly for creditors as it affects their management of credit risk and liquidity risk of loan portfolios. To mitigate the potential costs arising from accepting BCC in loan contracts, creditors may only accept BCC for loans made to good quality borrowers for two reasons. First, as credit worthy borrowers may ex ante concede greater contingent control rights to creditors to signal their high quality (Gârleanu and Zwiebel 2009; Demiroglu and James 2010), these borrowers are more likely to negotiate a BCC successfully. Second, granting a BCC to a quality borrower is also less costly to creditors as they are less concerned about rebalancing loan portfolios consisting of quality borrowers. Therefore, in our second hypothesis, we predict that the positive association between BCC and creditor control rights is more pronounced in high quality borrowers (borrower quality hypothesis).

As accounting ratios are usually used in debt contracting to facilitate the allocation of creditor control rights (Aghion and Bolton 1992), accounting conservatism, an accounting practice that recognizes economic losses more timely than gains, can tighten creditor control rights by triggering more timely covenant violation and renegotiations (Zhang 2008; Watts 2003). If the tightened control rights due to conservative financial reports are more of a concern for borrowers conceding control rights to creditors, we expect that the probability of including BCC increases with the extent of conservative reporting. Therefore, we expect the positive association between creditor control rights and BCC is more pronounced in loan contracts to borrowers with more accounting conservatism (conservative reporting hypothesis).

To test our hypotheses, we use data on all loan facilities extended to US borrowers in 1993–2015 in the DealScan database, which includes the initial loan terms and conditions and the identities of lenders and borrowers. We identify the contract terms pertaining to creditor control rights and various loan assignment clauses including BCC which restricts creditors’ loan assignment, as well as other loan pricing and non-pricing information from DealScan. We obtain firm characteristics from Compustat and stock return data from the Center for Research in Security Prices (CRSP). Our sample includes 25,826 loan packages issued to 5963 unique firms constituting 21,977 firm-year observations from 1993 to 2015.

Our primary results show that the probability of BCC inclusion increases with the intensity of creditor control rights proxied by the number of financial covenants in loan contracts, supporting H1. The results are robust to using tightness of financial covenants as an alternative measure of creditor control rights. Christensen and Nikolaev (2012) classify financial covenants into performance and capital covenants and find that the intensity of performance covenants is positively associated with probability of debt renegotiations, while capital covenants are not. We hence conduct additional analyses on the effects of performance and capital covenants on the probability of BCC inclusion. We find that the positive association between performance covenants and BCC is greater than that between capital covenants and BCC, implying that the control rights induced by the intensity of performance covenants are of a bigger concern for borrowers than those arising from capital covenants, consistent with the findings in Christensen and Nikolaev (2012) that the intensity of performance covenants is associated with stronger contingent control rights. The results of using financial covenant tightness and performance covenants tightness are qualitatively similar.

We also use the inception of credit default swaps (CDS) trading as a unique setting of weakened contingent control rights and examine its impact on the probability of BCC inclusion. Recent studies show that purchase of CDS dilutes creditors’ incentives to monitor borrowers and results in fewer debt covenants being included in debt contracts (Ashcraft and Santos 2009; Subrahmanyam et al. 2014; Shan et al. 2019), implying that the inception of CDS trading could weaken creditors’ demand for control rights. We expect and find a decline in the usage of BCC after the inception of CDS trading.

We employ three approaches to address the potential endogeneity issues of creditor control rights. First, we use a simultaneous equation framework to jointly estimate the determination of BCC inclusion and creditor control rights proxied by the number of financial covenants. We instrument the inclusion of BCC using the borrowing firm’s industry average usage of BCC, and instrument the number of financial covenants using a bank’s loan quality prior to (current) loan initiation. The results show that the usage of BCC continues to be associated with greater creditor control rights, consistent with our primary results.

Second, to address the endogeneity concern that creditor control rights may be determined by (observable) borrower and loan characteristics, we employ propensity score matching (PSM) method to match on firm and loan characteristics and obtain a matched sample. The diagnostic test results indicate that the matching procedure successfully removes all the observable differences between the treatment and control sample. The regression results continue to show that creditor control right is positively associated with BCC inclusion using the PSM matched sample.

Third, we exploit the staggered adoption of Anti-Recharacterization Law (ARL) in Texas and Louisiana in 1997 and then Alabama in 2001 as a quasi-natural experiment and an exogeneous shock to creditor rights and examine how ARL might affect BCC. ARL enables creditors to seize borrowers’ collateral without any delay by facilitating the transfer of borrower’s assets to a special purpose vehicle available to the creditors when the borrowing firm files for bankruptcy (Li et al. 2016), strengthening creditor rights during bankruptcy process. We show that loans syndicated in states after the enactment of ARL have much higher probability of BCC inclusion in loan contracts than loans in neighboring non-ARL states with similar firm- and loan-specific characteristics, suggesting that borrowers are more likely to demand the inclusion of BCC in response to strengthened creditor rights associated with ARL.

In our second hypothesis (H₂), we predict a stronger association between the presence of a BCC and creditor control rights when loans are extended to “good type” borrowers. Consistent with this prediction, we document that the positive relationship between BCC inclusion and creditor control rights is more pronounced for borrowing firms that have a higher market to book ratio, are larger, not financially constrained, and have rated bonds. In our third hypothesis (H₃), we predict that the positive association between creditor control rights and BCC is more pronounced in loan contracts to borrowers with more conservative financial reporting. Following Basu (1997) and Ball and Shivakumar (2005), we use the return-based and non-return-based accrual and earnings change measure, respectively, to gauge the degree of accounting conservatism. We find that the likelihood of BCC inclusion is much higher when borrowers report more conservatively, consistent with H₃.

Next, we conduct additional analyses to further assess the economic consequences of BCC on creditors and borrowers. In particular, we explore the relationship between BCC inclusion and the ex-post loan trading liquidity, as well as the abnormal equity returns around loan announcement and/or loan initiation of borrowers. Consistent with the notion that a BCC imposes restriction on loan trading, we find that loans with a BCC are less likely to be traded in the secondary loan market compared to loans without such clause. Consistent with our prediction that borrower consent clause is granted to “high quality” borrowers, our empirical results show that loans with a BCC have a greater positive abnormal return in the equity market surrounding loan announcement or initiation, compared to similar loans without such clause.

Our study makes two contributions to the existing literature. First, we focus on the BCC, a relatively unexplored area of credit agreements, as an important mechanism in contract design to reduce future renegotiation costs for borrowers due to dealing with unknown creditors with strong creditor rights. We document that the probability of BCC inclusion increases with the intensity of creditor control rights. Complementing extant studies in financial contracting that focus on how creditor control rights improve contracting efficiency (Aghion and Bolton 1992; Bolton and Scharfstein 1996; Demiroglu and James 2010), we show that BCC is also an important contracting mechanism to address contract incompleteness and improve contracting efficiency in the originate-to-distribute banking era.

Second, our study sheds light on the circumstances under which BCCs are included in debt contracts. We show that BCCs are more likely to be included in loan contracts when creditor control rights are strong, when borrowers are of “good quality”, and when borrowers issue more conservative financial reports.

The rest of the paper proceeds as follows. Section 2 presents the literature review and hypotheses development. Section 3 discusses data and models. Section 4 discusses the empirical results on creditor control rights and BCC, including robustness checks. Section 5 presents the results on the impact of BCC on lenders and borrowing firms. Section 6 concludes.

2 Literature review and hypotheses development

2.1 Incomplete contracting, creditor control rights, and debt covenants

Incomplete contracting theory suggests that financial contracts are incomplete because contracting parties cannot stipulate all future states or actions, resulting in contracting parties’ ex post opportunistic behavior (e.g., Grossman and Hart 1986; Hart and Moore 1988; Aghion et al. 1994; Dewatripont and Tirole 1994). For example, borrowers may undertake risky projects after they secure loans from creditors, and lenders may terminate funding for a positive net present value project (Sharpe 1990; Rajan 1992). To reduce the ex post opportunism in incomplete contracting, allocation of decision rights is the key to maximize the joint surplus of the contracting parties in that borrowers with high credit risk should delegate control rights to creditors, while borrowers with low default risk should retain control rights (Aghion and Bolton 1992; Bolton and Scharfstein 1996).

Demiroglu and James (2010) show that borrowers concede more control rights to creditors to signal their quality. However, allocating more control rights to creditors could affect the probability and outcome of future renegotiations (Dewatripont and Tirole 1994), which may concern borrowers especially when creditor control rights can be transferred to third parties via loan assignment. Under this circumstance, borrowers, rather than creditors, face uncertainty regarding who gains the control rights and with whom they negotiate during future renegotiations.

Aghion and Bolton (1992) further indicate that as future states of nature may not be contractible, the allocation of control rights can be based on a contractible signal, although noisy, such as change in accounting ratios that reflect the uncontactable state of nature. Specifically, accounting-based covenants in debt contracts are used to facilitate the allocation of control rights, because failure to maintain the threshold of accounting ratios specified in debt contracts may result in creditors taking over the control rights and debt renegotiations. Recent empirical studies provide a wide range of evidence regarding the presence of covenants in debt contracts and the role of these covenants in facilitating the transfer of control rights. For example, Dichev and Skinner (2003) suggest that financial covenants in debt contracts serve as “trip wire” to facilitate the transfer of control rights. Bradley and Roberts (2015) show that financial and nonfinancial covenants are widely used in debt contracts. Recent studies also show that most covenant violations lead to deal renegotiations and allow creditors to undertake a wide range of actions such as modifying loans’ maturity, interest rate, credit availability, and the structure of covenants and collaterals (e.g., Gopalakrishnan and Parkash 1995; Chen and Wei 1993; Roberts and Sufi 2009b; Chava and Roberts 2008; Nini et al. 2009; Roberts 2015). Upon covenant violation, creditors may also exercise control rights to influence the investment and financing activities of the borrowing firms and, in turn, corporate governance and firm value (e.g., Roberts and Sufi 2009a; Nini et al. 2012; Demiroglu and James 2010).³ In addition, existing studies document that upon covenant violation creditors push for operational changes and within firm resource allocation for the most productive use, shape executive bonus plans, reduce CEO compensation and risk-taking incentives in compensation package, and limit corporate acquisition activity (Ersahin et al. 2021; Armstrong et al. 2022; Balsam et al. 2018; Becher et al. 2022).

2.2 Hypotheses development

As discussed in the previous section, in an incomplete contracting setting with high uncertainty about unverifiable future states, borrowers may ex ante concede control rights to creditors to signal their quality or reduce creditors’ concern over borrowers’ opportunistic behavior. However, conceding control rights to creditors may not be efficient for the borrowing firms because such concession may increase borrowing firms’ uncertainty in dealing with creditors, especially when the originating lenders can transfer loans to third parties via loan assignment or outright loan sale. We argue that to balance the control rights granted to creditors, borrowers are likely to demand BCC to self-protect against the possibilities that their loans with tight creditor control rights be assigned to unknown creditors. Furthermore, tighter creditor control rights may trigger covenant violations and debt renegotiation more frequently (e.g., Roberts and Sufi 2009b; Nikolaev 2018), which may prompt borrowers to demand BCC inclusion in debt contracts to reduce the potential renegotiation costs arising from dealing with unknown creditors in renegotiations. Therefore, we propose the following hypothesis:

Hypothesis 1

The probability of BCC inclusion is positively associated with the intensity of creditor control rights (creditor control rights hypothesis).

From creditors’ perspective, BCC inclusion is costly because it restricts their flexibility to transfer loans to a third party, which makes it more difficult for them to manage the credit risk and liquidity risk of their loan portfolios. Hence, we predict that creditors are more likely to grant a BCC to “good types” borrowers especially when the latter accept tighter creditor control rights to signal their quality. Based on the above arguments, we propose the following hypothesis:

Hypothesis 2

The positive association between creditor control rights and BCC is more pronounced for “good types” borrowers (borrower quality hypothesis).

Financial contracts include accounting-based covenants to facilitate transfer of control rights (e.g., Aghion and Bolton 1992). Accounting conservatism that recognizes economic losses more timely than economic gains can speed up the transfer of control rights and therefore strengthen creditor control rights (Zhang 2008; Watts 2003). Accordingly, we predict that to alleviate their concerns over heightened creditor control rights associated with conservative financial reporting, borrowers that report conservatively are more likely to request BCC inclusion in debt contracts with more financial covenants.

Hypothesis 3

The positive association between creditor control rights and BCC is more pronounced for borrowers with more conservative financial reporting (conservative reporting hypothesis).

3 Data and sample

3.1 Sample selection

Our primary sources of data are DealScan and LSTA/LPC mark-to-market loan pricing database, both provided by Thomson Reuters’s Loan Pricing Corporation (LPC). DealScan provides detailed information on loan contract characteristics at the time of loan initiation, including the identities of both borrowers and creditors, price and nonprice terms of loans. The LSTA/LPC mark-to-market loan pricing database contains secondary loan market information, including the identity of the borrowing firm as well as various price quotes and dealer information. We construct borrowing firms’ characteristics by acquiring financial data from Compustat and stock market data from CRSP.

We assemble our sample in the following steps. First, we merge DealScan with the Compustat database using the DealScan-Compustat link file compiled by Chava and Roberts (2008). Each deal (package) reported in DealScan may contain more than one facility (tranche) with different loan types (i.e., revolver or term loan), loan amounts, spreads, maturities, compositions of lenders, and loan purposes, but with the same collateral, covenants, and loan assignment clauses governing the whole loan package. As both the BCC (and other assignment clauses) and covenants information are provided at the package level, we conduct our BCC analyses at the package level and take the maximum value of the facility-level variables as the package-level variable. For example, we use the longest maturity of all loan facilities in a package as the package maturity and the largest number of lenders of all facilities as the number of lenders for the package, and so forth.⁴ Our final sample includes 25,826 loan packages belonging to 5963 unique borrowers over the sample period of 1993–2015.

We complement our data with the covenant violation data provided by Professor Michael Roberts, which contains all covenant violations reported in the Securities and Exchange Commission (SEC) filings during the period 1996–2012 for publicly traded firms in the United States.⁵ To identify loan trading data, we merge DealScan data with the LSTA/LPC mark-to-market loan pricing data by matching LIN numbers and DealScan facility numbers.

3.2 Identification of BCC

When loans are transferred to loan buyers, the creditor control rights associated with them along with cash flow rights are also transferred to loan buyer. A credit agreement may also contain a clause that prohibits or restricts the assignment of a portion or the entire loan to a third party. Also, an assignment clause can specify the eligibility requirements that a third party must meet in order to gain ownership in a loan. These requirements can be related to the type of financial institution that is allowed to buy a loan—for example, only commercial banks are allowed to buy participation of a certain loan—or they can be related to the minimum resources, net worth or capital, that the buyer is required to have—for example, the buyer needs to have a minimum capital and reserves of more than $1 billion to be an assignee in a particular loan. In particular, we examine the BCC, a clause that requires borrower consent for all loan transfers.

More specifically, we define six general types of assignment clauses: (1) whether the agreement includes a clause that requires borrower consent for all assignments (BCC); (2) whether the consent of lead lender is required for a lending syndicate member to transfer loan (lead lender consent clause); (3) whether the agreement has eligibility requirements for loan buyers: eligible loan buyers may be restricted to certain types of financial institutions, such as commercial banks, funds, insurance companies or may be defined based on minimum size of assets, capital, net worth, surplus, and so forth (buyer eligibility clause); (4) whether there are minimum holding requirements. This group of assignment clauses put restrictions on original lenders selling all their participations in the secondary market (minimum holding clause). It is very common that the agent lenders are required to retain all or a significant portion of their participation in a loan; (5) whether there are restrictions on the portion of the loan that can be assigned (minimum assignment clause); (6) whether the consent of institutional investors is required for a lending syndicate member to transfer loan (institutional investor OK clause). Of these clauses, the BCC and lead lender consent clause can be directly inferred from, respectively, the Company Consent and the Agent Consent variables in DealScan’s Package table. To identify the other four clauses, we read the comment section of each credit agreement provided in DealScan’s Package Assignment Comment table and infer whether the clause exists in the agreement.

Table 1, Panel A shows the yearly distribution of deals with BCC. The percentage of deals with a BCC range from 44% in 1993 to 69% in 2008. The portion of deals with a BCC is relatively larger in 2002–2008 than that in other periods, corresponding to the growth in the secondary loan trading market that facilitates loan trading. Panel B shows the distribution of different loan assignment clauses. Among our sample of 25,826 loan packages, 14,198 (55%) packages have a BCC, 14,756 (57.1%) have lead lender consent, 2347 (9.1%) have buyer restriction, 14,057 (54.4%) have minimum assignment, 1850 (7.2%) have minimum holding clause, and 935 (3.6%) have institutional investor OK clause.

Table 1 Sample distribution

3.3 Sample descriptive statistics

Table 2 presents the descriptive statistics. Out of 25,826 packages in our sample, 55% have a BCC. On average, each package has 1.422 financial covenants; among loan packages with financial covenants, the average number of performance (capital) covenants is 1.628 (0.694). Each package on average has 8.35 lenders, 90.8% are commercial banks, and 9.2% are nonbank institutional investors. An average borrowing firm has a total assets size of $7383.375 million, a market-to-book ratio of 1.701, a tangibility ratio of 34.8%, a profitability ratio of 2.3%, and a leverage ratio is 25.9%. On average, 23.2% of the borrowing firms have issued leveraged loans, and 16.6% of borrowing firms are financially constrained.

Table 2 Descriptive statistics

4 Empirical results

4.1 Creditor control rights and BCC

4.1.1 Baseline analysis

Our first hypothesis states that borrowers of loans with tighter creditor control rights are more likely to demand BCC inclusion in debt contracts to protect themselves. To test this conjecture, we employ the following firm-fixed effects linear probability model:⁶

$$BCC_{i,j,t} = \alpha_{i} + \delta_{t} + \beta_{1} FinCov_{i,j,t} + \gamma X_{i,t - 1} + \theta C_{i,j,t} + \varepsilon_{i,j,t}$$

(1)

where i, j, and t denote firm, loan packages, and time, respectively. BCC is an indicator that equals one if a loan package contains BCC and zero otherwise. $\alpha_{i}$ and $\delta_{t}$ represent firm and year fixed effects, respectively. FinCov_i,j,t is the number of financial covenants, which is a measure of the intensity of creditor control rights, following Demiroglu and James (2010). $X_{i,t - 1}$ is a set of firm-specific control variables, and $C_{i,j,t}$ is a set of loan package–level control variables to be discussed below.

We control for firm characteristics that proxy for information asymmetry and agency costs of debt that might affect creditor control rights. Among the firm characteristics, we include firm size, market-to-book ratio, tangibility, profitability, leverage, and an indicator for financial constraint. All variables are described in Appendix 1. Larger firms, more profitable firms, firms with better growth prospect, and firms less financially constrained are assumed to be quality firms. Hence, we expect that they are more likely to be granted a BCC. Prior studies also suggest that firms with a larger proportion of tangible assets have higher liquidation costs (Bolton and Oehmke 2011; Favara et al. 2012) and creditors’ bargaining position is stronger in firms with a larger proportion of tangible assets (Feldhütter et al. 2016). Therefore, we expect tangibility is positively associated with the existence of a BCC. Among loan-specific variables, we include the loan-to-asset ratio, loan maturity, revolver loan indicator, leveraged loan indicator, and security (an indicator for loan collateral). Larger loans and revolvers loans are more likely to have a BCC because borrowers want to screen syndicated members to ensure that they are capable of meeting future funding obligations.

The results of estimating Equation (1) are presented in Table 3. Column (1) of Table 3 provides the results of a probit model, Column (2) presents the results with linear probability firm fixed effects, and Column (3) shows the results when we restrict the sample to loans with at least one financial covenant in a linear probability model. The results in Columns (1)–(3) indicate that the coefficients on the number of financial covenants are positive and significant at the 1% level, suggesting that loans with a larger number of financial covenants are more likely to have a BCC. In particular, in the linear probability model in Column (2), the coefficient on FinCov is 0.189 and significant at the 1% level, suggesting that the addition of one financial covenant leads to about 19% increase in the probability of including BCC in the contract.

Table 3 Creditor control rights and borrower consent clause

Demerjian and Owens (2016) develop an aggregate measure of covenant tightness and show that their measure of covenant tightness is associated with higher probability of covenant violations and renegotiations.⁷ We follow Demerjian and Owens (2016)and use tightness of financial covenants (FinCov_Tightness) as an alternative measure of the intensity of creditor control rights. Results reported in Column (4) of Table 3 show that the loans with tighter financial covenants are more likely to have BCC, consistent with the argument that contracts with stronger creditor control rights are more likely to include a BCC. Overall, these results suggest that BCC serves as an important contractual mechanism to improve contracting efficiency.

The coefficients on the control variables indicate that among borrower characteristics, larger firms and firms with less financial constraints are more likely to have a BCC. Among the loan-specific variables, larger loans are more likely to have a BCC.

4.1.2 The effect of performance covenants and capital covenants on BCC

Our baseline analysis shows that when debt contracts have tighter financial covenants, borrowers are more likely to demand BCC as tighter financial covenants may trigger more frequent renegotiations and covenant violations. Christensen and Nikolaev (2012) further classify financial covenants into performance covenants and capital covenants and show these two types of covenants have different implications for debt contracting.⁸ In particular, they argue that while performance covenants are used as trip wires to reduce borrowers’ ex post opportunistic behavior via the transfer of contingent control rights to creditors, capital covenants align the interests of shareholders and creditors ex ante by requiring that shareholders maintain sufficient capital within the firm to increase the sensitivity of the shareholder stake to managerial actions, thereby reducing agency costs of debt. They find that performance covenants, as trip wires to facilitate allocation of the contingent control rights, are associated with more frequent contract renegotiations; in contrast, capital covenants that are used to address agency problem ex ante do not lead to more frequent contract renegotiations. If BCC is used as a strategic response to the increased frequency of future renegotiations associated with the intensity of performance covenants rather than the ex ante agency problem, we expect that the positive association between performance covenants and BCC is stronger than that between capital covenants and BCC.

We report the results of these two types of financial covenants in Table 4, with columns (1)–(3) on the effect of the number of performance (PerfCov) and capital covenants (CapCov) and columns (4)–(6) on the effect of the tightness of these two covenants, respectively. Columns (1) and (2) show that the coefficient on PerfCov is positive and significant at the 1% level, while that on CapCov is negative yet insignificant. When we include both PerfCov and CapCov in the same regression (Column (3)), the coefficients on PerfCov and CapCov remain qualitatively similar to those in columns (1) and (2). The results suggest that performance covenants are associated with higher likelihood of BCC inclusion, while capital covenants are not. The results using the tightness of performance and capital covenants (columns (4)-(6)) are qualitatively similar to those of using the number of PerfCov and CapCov. Overall, the results in Table 4 are consistent with our argument that because performance covenants are more likely to trigger debt renegotiations than capital covenants, the intensity of performance covenants results in higher likelihood of BCC inclusion in debt contracts than that of capital covenants.⁹

Table 4 Performance, capital covenants and BCC

4.1.3 The effect of the inception of credit default swap (CDS) trading on BCC

To corroborate the findings that borrowers demand BCC as a strategic response to tight contingent control rights, we examine how the inclusion of BCC varies with the inception of CDS trading which weakens creditors’ demand for control rights. Once banks purchase CDS to offload credit risk and gain credit protection, they have less concern about borrower default (Ashcraft and Santos 2009), and become less flexible in subsequent debt renegotiations (Subrahmanyam et al. 2014). Shan et al. (2019) find that the inception of CDS trading reduces banks’ incentives to monitor borrowers and leads to loosening of debt covenants, suggesting CDS trading substitutes for covenants on loan contracts as an alternative discipline mechanism on borrowers. If the inclusion of BCC is a strategic response to tight creditor control rights via contractual design, we would expect a decline in the usage of BCC after the inception of CDS trading.

We obtain CDS trading data between 2000 and 2012 from Markit CDS Composites Pricing database and identify the first and the last day a borrowing firm’s CDS trading price appear in the database. We construct an indicator variable Post_CDS that equals one if a syndicated loan is originated after the first day of CDS trading but before the last day of trading, and zero otherwise. We replace CCR in Eq. (1) with Post_CDS and conduct the analyses with and without the measure of creditor control rights (i.e., number of financial covenants).¹⁰ As Shan et al. (2019) focus on the substitution effect of CDS trading for covenants as alternative monitoring device, we use the sample with at least one financial covenant for this analysis. The results are provided in Table 5.

Table 5 The inception of CDS trading and BCC

Columns (1) to (3) of Table 5 report the results analyzing the relation between CDS inception and the inclusion of BCC using our full sample (loans belong to borrowers with and without CDS market). As we control for firm and year fixed effects in these estimations, we expect the coefficients on Post_CDS to capture the time-variant response in contracting terms to the inception of CDS trading. The negative coefficient on Post_CDS (column (1): coefficient = − 0.088; t-statistics = − 5.056) suggests that the inception of CDS trading is associated with a lower likelihood of including a BCC in debt contracts. The results are qualitatively similar when controlling for the number of financial covenants (Columns (2)). We then interact Post_CDS with FinCov to check whether the inception of CDS trading attenuates the impact of financial covenants on the inclusion of BCC. Results reported in Column (3) of Table 5 show that the coefficient on the interaction term Post_CDS × FinCov is positive yet insignificant, suggesting the inception of CDS trading does not mitigate the effect of financial covenants on the likelihood of BCC inclusion.

We also conduct the analysis using a subsample of loans to borrowers with outstanding CDS only. The results reported in Columns (4) to (6) show that the inception of CDS trading is still negatively associated with the likelihood of BCC inclusion with and without controlling of FinCov, consistent with the results using the full sample. The coefficient on the interaction term Post_CDS × FinCov remains positive yet insignificant.

4.2 Addressing endogeneity issues

4.2.1 Joint determination of BCC and creditor control rights

Our analyses so far show that in response to strong creditor control rights, borrowing firms demand a BCC clause to reduce the potential renegotiation costs arising from dealing with unknown creditors in renegotiations. However, BCC also reduces the marketability of loans for the creditor. Therefore, creditors and borrowers might negotiate and tradeoff between the inclusion of BCC and imposing financial covenants in debt contracts. In other words, the inclusion of BCC and use of financial covenants may be jointly determined.

To address this simultaneity concern, we follow Bharath et al. (2011) and model a setting in which the inclusion of BCC and the intensity of financial covenants are jointly determined. Using this framework, we need instruments for the two endogenous variables BCC and creditor contingent control rights. Murfin (2012) finds that banks that experienced loan defaults in the past write tighter debt contracts on subsequent loans than their peers do because recent defaults update the banks’ perception about their own poor screening ability. As a result, banks impose tighter covenants to make up for the weaker ex ante screening. Following this logic, we expect banks that had high credit risk exposures in the past would tighten creditor control right on the loans made subsequently (relevance restriction). On the other hand, banks’ credit risk exposure should not directly affect the probability of BCC inclusion other than through banks’ demand for tighter credit control rights (exclusion restriction). Therefore, we use banks’ ex-ante loan portfolio quality, proxied by the number of lead lenders with an increase in net loan charge-off in the quarter prior to loan initiation (Increase_Chargeoff), as an instrumental variable for FinCov. We employ the average percentage of BCCs used by the borrowing firm’s peers in the same industry based on two-digit SIC code (SIC2_BCC), to instrument for the usage of BCC by the borrowing firm. Given the peer effect, the industry usage of BCC can affect individual firm’s demand for BCC (relevance restriction), but unlikely affect the number of financial covenants included in debt contracts (exclusion restriction).

We estimate the following simultaneous equations to account for the joint determination of inclusion of BCC and the intensity of financial covenants:

$$Prob(BCC_{i,j,t} = 1) = \delta_{t} + \beta_{1} FinCov_{i,j,t} + \emptyset_{1} SIC2\_BCC_{k,t} + \gamma X_{i,t - 1} + \theta C_{i,j,t} + \varepsilon_{i,j,t}$$

(2)

$$FinCov_{i,j,t} = \delta_{t} + \beta_{2} BCC_{i,j,t} + \emptyset_{2} Increase\_Chargeoff_{i,j,t} + \gamma X_{i,t - 1} + \theta C_{i,j,t} + \varepsilon_{i,j,t}$$

(3)

In the 1st stage we estimate the reduced forms (excluding the two endogenous variables, FinCov and BCC, on the right- hand side) of equation (2) and (3) to obtain the predicted values for each endogenous variable (Predicted_BCC and Predicted_FinCov, respectively). Then we substitute the reduced form fitted values (Predicted_FinCov and Predicted_BCC) for the endogenous variables (BCC and FinCov) in equations (2) and (3), which are the second stage models.

Columns (1) and (2) of Table 6 present the results of the IV estimation of the effect of FinCov on BCC. The first stage results in Column (1) show that the coefficient on Increase_Chargeoff is positive and significant at the 1% level, indicating the increase in bank net charge off ratio is associated with a greater number of financial covenants. The second stage results in Column (2) show that the coefficient on Predicted_FinCov is positive and significant at the 1% level, suggesting FinCov is associated with higher likelihood of BCC inclusion. Columns (3) and (4) report the results for the IV estimation of the effect of BCC on FinCov. The first stage results in Column (3) show that the coefficient on industry usage of BCC (SIC2_BCC) is positive and significant at the 1% level, indicating the industry usage of BCC is associated with higher likelihood of BCC inclusion by individual firm in the same industry. The second stage results in Column (4) show that the coefficient on Predicted_BCC is positive and significant at the 1% level, suggesting the usage of BCC is associated with a greater number of financial covenants. We also test the relevance and validity of our instrument variables. The under-identification test statistics (Kleibergen-Paap rk LM statistic) for both BCC and FinCov regressions are significant, rejecting the null hypothesis that the excluded instruments are not correlated with the endogenous regressors. The weak identification test statistics (Cragg-Donald Wald F-statistics) are all higher than the Stock and Yogo (2005) critical value of 16.38, suggesting that Increase_Chargeoff is a strong instrument for FinCov and that SIC2_BCC is a strong instrument for BCC.

Table 6 Robustness check: simultaneous equations

4.2.2 Propensity score matching

Another endogeneity issue is that creditor control rights may be determined by (observable) borrower and loan characteristics. To address such endogeneity concern, we employ propensity score matching (PSM) method to match on firm and loan characteristics by estimating the following probit model and obtain a matched control sample.

$$Probit(Hcov_{i,jt} = 1) = i_{i} + \delta_{t} + \gamma X_{i,t - 1} + \theta C_{i,j,t} + \varepsilon_{i,j,t}$$

(4)

where Hcov is an indicator variable that equals 1 if a loan contract contains at least one financial covenant and 0 otherwise.¹¹$i_{i}$ and $\delta_{t}$ represent industry and year fixed effects, respectively. $X_{i,t - 1}$ is a set of firm-specific control variables including firm size, market-to-book, tangibility, profitability, leverage, and financial constraint indicator, as defined in Sect. 4.1.1. $C_{i,j,t}$ is a set of loan package–level control variables including loan-to-asset, maturity, revolver loan indicator, leveraged loan indicator, and collateral indicator, as discussed in Sect. 4.1.1. Specifically, we first match each loan with at least one financial covenant (Hcov = 1), without replacement, with a loan without any financial covenant (Hcov = 0), requiring the difference in the probability of having at least one financial covenant between the two observations not exceeding 1% (i.e., clipper = 0.01).

We report the results of the probit model in Panel A of Table 7. Column (1) shows that several firm and loan characteristics are significantly related to the probability of having at least one financial covenant. More specifically, larger firms and financially constrained firms are less likely to have financial covenants, while firms with higher market to book ratio and loftier profitability are more likely to have financial covenants.

Table 7 Robustness check: propensity score matching

We then conduct a diagnostic test to verify whether firm- and loan-specific characteristics in treatment (Hcov=1) and control (Hcov=0) groups are comparable, and report the results in Panel B of Table 7. Before the PSM matching, all firm- and loan-specific characteristics are significantly different except for ROA. After the PSM matching, none of the differences in observable firm- and loan-specific characteristics between the two groups is statistically significant. The diagnostic test results suggest that our matching procedure successfully removes almost all the observable differences between the treatment and control sample. Columns (2) and (3) of Panel A report the results using the dichotomous variable (Hcov) and continuous variable of financial covenants (FinCov) as the independent variables, respectively, based on the PSM matched sample. The coefficients on both Hcov and FinCov are significant at the 1% level, suggesting that our baseline results are robust to using PSM method to address the endogeneity issue of creditor control rights due to observable borrowing firm- and loan-specific characteristics.

4.2.3 Quasi-natural experiment: the anti-recharacterization law (ARL)

In this section, we exploit ARL as a quasi-natural experiment and an exogeneous shock to creditor rights and assess its impact on the likelihood of BCC inclusion. Several U.S. states enacted ARL in late 1990s and early 2000s to ensure that creditors can seize the collateral pledged by borrowing firms timely during bankruptcy process. Firms conducting secured borrowing can transfer collateral to a special purpose vehicle (SPV) and minimize the risk exposure of SPV. In theory, an SPV allows creditors to possess the collateral without any delay during bankruptcy process. However, in practice whether creditors can timely claim the collateral in an SPV depends on a bankruptcy court’s ruling on the characterization of the asset transferred to an SPV. If bankruptcy courts rule that the assets under an SPV is considered as a loan rather than sales, creditors cannot recover their claim in an SPV until the borrowing firm is liquidated or restructured, which introduces significant uncertainty and delays in creditors’ access to collateral. Consequently, the enactment of ARL significantly strengthened creditor rights in seizing collateral by stipulating that collateral held under an SPV be characterized as sales.

This law was introduced by Texas and Louisiana in 1997, by Alabama in 2001, Delaware in 2002, South Dakota in 2003, Virginia in 2004, and Nevada in 2005. However, the Reaves Brokerage Company, Inc. v. Sunbelt Fruit & Vegetable Company Inc. case in 2003 significantly reduces the influence of ARL because the ruling of this case completely relies on a federal standard to determine the nature of assets in SPV and override the influence of ARL of Texas. As a result, the influence of the ARL is weakened after 2003.

We argue that while the inclusion of more covenants in debt contracts strengthens creditor control rights by triggering covenant violation and renegotiation, ARL strengthens creditor rights by ensuring creditor’s claim of borrowers’ secured assets during liquidation process and may result in borrowers demanding for BCCs to avoid deal with unknown creditors during the liquidation process. Accordingly, we use ARL as a shock to creditor rights and employ a difference-in-differences approach to analyze the effect of ARL on borrowers’ demand for BCC during the period in which the ARL has stronger impact, i.e., years before 2003.¹² As only Texas, Louisiana, and Alabama enacted this law before 2003, we focus on the loans made in 1995–2003. We use the loans made to firms located in the neighboring states of these three states as the control sample, i.e., New Mexico, Oklahoma, Arkansas as neighbors of both Texas and Louisiana, and Georgia, Florida, and Tennessee as neighbors of Alabama.¹³ We estimate the following empirical model:

$$Prob(BCC_{i,j,t} = 1) = \alpha_{i} + \beta_{1} ARL + \beta_{2} Post + \beta_{3} Post \times ARL + \beta_{4} FinCov + \gamma X_{i,t - 1} + \theta C_{i,j,t} + \varepsilon_{i,j,t}$$

(5)

where ARL is an indicator that equals 1 if a loan deal is syndicated for a borrower headquartered in one of the ARL states during the sample year, i.e., in 1995–1996 or 1998–1999 in Texas or Louisiana, or in 1999–2000 or 2002–2003 in Alabama; and 0 if a loan is syndicated for a borrower headquartered in one of the non-ARL neighboring states. Post is an indicator that equals 1 if a loan is borrowed by a firm in Texas, Louisiana or their neighboring states (Alabama or its neighboring states) in 1998–1999 (2002–2003) and 0 otherwise. All other variables are defined as in Eq. (1).

The results are reported in Table 8. Columns (1) and (2) report the results for a sample of syndicated loans in ARL states. The coefficient on Post is positive and significant at the 1% level. The results of the fully specified model (Eq. (5)) are reported in columns (3) and (4). The coefficient on ARL is negative yet insignificant, suggesting that before the enactment of ARL, the probability of including BCC in loans made to firms in the adopting states is not significantly different from that in the non-adopting states. The coefficient on Post is insignificant, suggesting that there is no significant change in the probability of including BCC in the non-ARL states in the post-period. Finally, the positive and statistically significant coefficient on ARL × Post (coefficient = 0.369, t-statistics = 2.412) indicates that the probability of BCC inclusion is higher for loans borrowed by firms in the ARL states in the post ARL period, compared to the loans borrowed by firms in the non-ARL states in the same period. The marginal effect reported in Column (4) suggests that all else equal, the probability of including BCC in the post ARL period increases by 14% for firms in ARL states.

Table 8 Robustness check: Anti-recharacterization law (ARL) as a quasi-natural experiment

In the difference-in-differences analysis reported in Table 8, we select a sample of loans made to firms in the neighboring non-ARL states as the control sample and find that ARL strengthens creditor rights, resulting in higher likelihood of borrowers demanding inclusion of BCC in debt contracts. However, as selecting the control sample based on only the location of borrowers may omit firm and loan characteristics that systematically affect the control and treatment sample, we conduct further analysis by using propensity score matching to match loans in ARL states and their neighboring non-ARL states based on loan- and firm- characteristics and then re-estimate Eq. (5) using the matched sample.¹⁴

We report the results in Table 9. The diagnosis test results reported in Panel A show that while the loan- and borrower-specific characteristics are statistically different in most dimensions before the matching, none of the firm- and loan-specific variables is statistically significant after matching, suggesting the PSM matching procedure successfully removes the observable differences and hence the control group (i.e., loans in non-ARL states) is comparable to the treatment firms (i.e., loans in ARL states). Regression results based on the matched sample are presented in Panel B of Table 9. Column (1) shows the coefficient on ARL is negative and significant (coefficient = − 0.567; t-statistics = − 2.457), suggesting that before the enactment of ARL, the probability of including BCC in loans made to firms in ARL states is smaller than that in non-ARL states. The coefficient on Post is positive yet insignificant (coefficient = 0.025; t-statistics = 0.067), suggesting that there is no significant change in the probability of BCC inclusion in non-ARL states in the post-period. On the other hand, the coefficient on ARL × Post is positive and significant (coefficient = 0.459, t-statistics = 2.829), suggesting that firms in ARL states are more likely to demand BCC in the post-ARL period, compared to their matched firms in bordering non-ARL states. The results are consistent with those in Table 8 where firms from neighboring states are selected as the control group.

Table 9 ARL and BCC inclusion using PSM matched sample

As the passage of ARL and intensity of financial covenants strengthens different creditor rights, we also use the matched sample to analyze the interactive effect of ARL and FinCov by including the three-way interaction term ARL × Post × FinCov and relevant two-way interaction terms in Eq. (5).¹⁵ We report the corresponding results in Column (2). We find that the coefficient on $ARL \times Post$ remains positive and significant at the 5% level (coefficient = 0.399), while the coefficient on ARL becomes insignificant. Furthermore, the coefficient on the three-way interaction ARL × Post × FinCov is positive yet insignificant, suggesting that strengthening creditor rights via ARL and tightening creditor control rights imposed by financial covenants do not lead to greater demand for BCC. The insignificant result may be because the creditor rights arising from ARL and the intensity of financial covenants are different. Specifically, ARL is an exogenous shock to creditor rights by ensuring creditor’s claim of borrowers’ secured assets during liquidation process, while imposing financial covenants in debt contracts strengthens creditor control rights by triggering covenant violation and renegotiation.

4.3 Test of H₂: creditor control rights, borrower quality, and BCC

In our second hypothesis, we argue that high quality borrowers who concede the control rights to creditors are more likely to negotiate a BCC protection successfully. Furthermore, it is less costly for creditors to grant BCC to good quality borrowers. In this section, we investigate whether the positive association between BCC and creditor control rights is more pronounced for quality borrowers. We classify large borrowers, borrowers with access to public debt market, and borrowers with high market to book ratio and low financial constraint as high quality borrowers. We then estimate the following equations:

$$BCC_{i,j,t} = \alpha_{i} + \delta_{t} + \beta_{0} HQ_{i,t - 1} + \beta_{1} FinCov_{i,j,t} + \beta_{2} HQ_{i,t - 1} \times FinCov_{i,j,t} + \gamma X_{i,t - 1} + \theta C_{i,j,t} + \varepsilon_{i,j,t}$$

(6)

HQ is a placeholder for a group of indicator variables for high market to book (High Mkbk), rated (Rated), large firms (Large), and not financially constrained firms (Unconstrained). High Mkbk (Large) equals 1 if the market to book ratio (firm assets) is greater than the sample median and 0 otherwise. Rated equals 1 if a borrower has S&P long term debt rating and 0 otherwise. Unconstrained equals 1 if a borrower’s Whited-Wu index (WW index) (Whited and Wu 2006)is greater than the sample median and 0 otherwise.¹⁶

The results reported in Table 10 indicate that the coefficients on $HQ_{i,t - 1} \times FinCov_{i,j,t}$ are all positive and significant at the 1% level, with the magnitude of coefficients ranging from 0.015 ($Hmkbk_{i,t - 1} \times FinCov_{i,j,t} )$ to 0.101 $\left( {Large_{i,t - 1} \times FinCov_{i,j,t} } \right)$, suggesting that given the same level of creditor control rights, the likelihood of high quality borrowers obtaining BCC increases by 1.5% to 10.1%,compared to low quality borrowers.

Table 10 Borrower quality, creditor control rights, and BCC

4.4 Test of H₃: creditor control rights, accounting conservatism, and BCC

In the third hypothesis, we predict that borrowers that report conservatively are more likely to request the inclusion of a BCC in contracts with more financial covenants. To test this prediction, we construct three alternative accounting conservatism measures, AT1, AT2, and AT3. First, we follow Basu (1997) and estimate the following equation to obtain a return-based measure of accounting conservatism, asymmetric timeliness (AT1):

$$\frac{{E_{it} }}{{P_{it - 1} }} = \alpha_{0} + \alpha_{1} DR_{it} + \alpha_{2} R_{it} + \alpha_{3} DR_{it} \times R_{it} + \varepsilon_{it}$$

(7)

where E_it is the earnings per share before extraordinary items of firm i in year t; P_it−1 is the stock price of firm i at the end of year t—1; R_it is the buy-and-hold return for firm i over the 12-month period ending three months after the end of fiscal year t; DR_it is an indicator variable that equals 1 if R_it is negative, and 0 otherwise.$\alpha_{3}$ captures the asymmetric timeliness (AT1) of economic losses recognition in earnings relative to economic gains in earnings.

The send measure, AT2, is based on the following Ball and Shivakumar (2005) model:

$$TACC_{it} = \beta_{0} + \beta_{1} DCFO_{it} + \beta_{2} CFO_{it} + \beta_{3} DCFO_{it} \times CFO_{it} + \varepsilon_{it}$$

(8)

where TACC is the total accruals (ib-oancf) scaled by total assets at the beginning of the year, CFO is cash flow from operating activities (oancf) scaled by total assets at the beginning of the year, DCFO is an indicator that equals 1 if CFO is negative and 0 otherwise. The coefficient $\beta_{3}$ on DCFO × CFO represents the relative speed of economic losses are recognized into accruals. The higher the $\beta_{3}$, the greater the asymmetric timeliness of accruals to recognize economic losses relative gains (AT2), and the more conservative the financial reporting.

The third measure AT3 is based on the following Ball and Shivakumar (2005) model:

$$\Delta E_{it} = \gamma _{0} + \gamma _{1} D \Delta E_{it - 1} + \gamma _{2} \Delta E_{it - 1} + \gamma _{3} D\Delta E_{it - 1} \times \Delta E_{it - 1} + \varepsilon_{it}$$

(9)

where $\Delta E_{it}$ is the change in net income in year t, $\Delta E_{it - 1}$ is the change in net income in year t−1, and $D\Delta E_{it - 1}$ is an indicator variable if change in net income $\Delta E_{it - 1}$ is negative and 0 otherwise. A more negative coefficient $\Delta_{3}$ on $D\Delta E_{it - 1} \times \Delta E_{it - 1}$ implies more timely recognition of the losses that are transitory and reverse sooner.

The firm-year accounting conservatism measure is obtained by estimating the above three regressions for each two-digit NAICS industry over a five-year period prior to the loan syndication year. Each industry with the number of firm-years less than 200 and the number of firms less than 20 is removed from the estimation to avoid small sample bias. We interact creditor control rights proxied by FinCov with each of the three measures of accounting conservatism AT1, AT2, and AT3, and estimate the following linear probability model with firm fixed effects to evaluate the incremental impact of accounting conservatism on the probability of BCC in response to tighter creditor control rights.

$$BCC_{i,j,t} = \alpha_{i} + \delta_{t} + \beta_{0} Hat_{i,t - 1} + \beta_{1} FinCov_{i,j,t} + \beta_{2} Hat_{i,t - 1} \times FinCov_{i,j,t} + \gamma X_{i,t - 1} + \theta C_{i,j,t} + \varepsilon_{i,j,t}$$

(10)

where Hat is a placeholder for Hat1, Hat2, and Hat3. Hat1(Hat2) equals 1 if AT1 (AT2) is greater than the sample median, and 0 otherwise. Hat3 equals 1 if AT3 is negative, and 0 otherwise. All other control variables are defined in equation (1).

Table 11 reports the results. Consistent with our expectation, the results show that the coefficients on all the interaction terms between Hat1, Hat2, and Hat3 and FinCov are positive and significant at the 1% level (Column (1): coefficient =0.027; t-statistics =5.349; Column (2): coefficient =0.035; t-statistics = 7.214; Column (3): coefficient =0.023; t-statistics =4.689). These results suggest that borrowers with conservative financial reporting are more likely to require BCC in the debt contracts to alleviate the concern of strong creditor control. The negative and significant coefficients on Hat1 and Hat2 suggest that in firms with loan contracts containing fewer financial covenants, their conservative financial reporting has a negative association with the inclusion of BCC.

Table 11 Accounting conservatism, creditor control rights, and BCC

5 Additional analyses

To further examine the impact of the inclusion of a BCC on lenders and borrowing firms, we explore the association between BCC inclusion and the probability of loan trading and the equity market response surrounding the announcement and initiation of loans with BCC.

5.1 BCC and the probability of loan trading

A BCC requires lenders to obtain borrower permission before they can transfer loans, imposing a constraint on creditors’ ability to transfer loans. We expect that loans with a BCC are less likely to be traded in the secondary market. To test this conjecture, we employ the following model:

$$\Pr \left( {Trade = 1} \right)_{i,qt} = \alpha_{1} + \beta_{1} BCC + \gamma X_{i,qt} + \theta C_{i,j} + \mu_{2ind} + \delta_{qt} + \varepsilon_{i,qt}$$

(11)

where Trade is an indicator variable if a loan is traded in quarter t and 0 otherwise. BCC is an indicator variable that equals one if a firm loan contains a BCC and zero otherwise. $\mu_{2ind}$ and $\delta_{qt}$ denote two-digit industry and quarter fixed effects, respectively. We follow Drucker and Puri (2009) and include firm and loan characteristics to control for factors affecting the probability of loan trading. $X_{i,qt}$ is a vector of firm characteristics measured at the beginning of each quarter, including natural logarithm of total assets (Log(Assetsq))), market-to-book ratio (Mkbkq), return on assets (Roaq), leverage ratio (Levgq), and investment grade and noninvestment grade indicators. $C_{i,j}$ denotes a vector of loan characteristics including loan size (Lloansize), loan maturity (Log(Maturity)), collateral indicator (Secured), percentage of nonbanks (Nonbank Prc), total number of financial covenants (FinCov), the number of days between loan syndication and the first trading date (Log(Loan age)) and indicator for revolver loan (Revolver) and institutional loans (Institutional). We also control for loan purpose fixed effects and credit rating fixed effects.

Table 12 reports the results. Consistent with our conjecture, we find that loans with a BCC are less likely to be traded in the secondary loan market (Column (1) coefficient on $BCC$ = − 0.154; t-statistics = − 3.819), suggesting that BCC does restrict creditors’ flexibility in transferring their loans in the secondary loan market. In Column (2), we show the results with other loan assignment clauses, including buyer restriction, minimum assignment, minimum holding, and institutional investor OK clause. The coefficient on BCC is still negative and significant at the 1% level, which is qualitatively similar to that reported in Column (1). The coefficients on both Min Holding and Restriction on Buyer Type are negative and significant at the 5% level, but smaller in magnitude than the coefficient on BCC; the coefficient on Institutional Investor OK is positive and significant at the 1% level; and the coefficient on Min Assignment is positive yet insignificant. These results are generally consistent with our expectation that strict assignment clause reduces loan trading liquidity, while relaxed clause increases loan trading liquidity.

Table 12 BCC and the probability of trading

5.2 Equity market reactions to the announcement of loans with BCC

Our main premise for creditors accepting BCC is that borrowers who are granted with such clause are of good quality and the costs of lower loan liquidity is mitigated by the benefits of lending to such quality borrowers. If this is the case, then granting the clause could convey positive private information about borrower quality to public investors. We examine whether the announcement and initiation of loans with BCC is accompanied with a higher abnormal return of the borrowing firm in the equity market.

The results are reported in Table 13. First, we compare the cumulative market-adjusted abnormal returns over a three-trading-day window (centered on the loan initiation date) between the loans with and without BCC. The results in Column (1) show that the abnormal returns around loan initiation are higher for loans with a BCC than for loans without such a clause. Because loan initiation date may lag announcement date, we conduct an additional test by drawing a random sample of 7000 loans from our original sample. Then for each loan, we search Factiva archives for news report on loan announcements to identify the actual loan announcement date. We find announcement dates for 2665 loans and compare cumulative market-adjusted abnormal returns around the actual loan announcement date for loans with and without a BCC. We still find that loans with BCC experience more positive abnormal return following loan announcement, compared to loans without such a clause, and the difference is positive and significant at the 5% level (Column (2)). More specifically, we find that the abnormal return of loans with BCC is 20–50 basis points higher than that of loans without such a clause. The difference is significant both economically and statistically.

Table 13 Abnormal returns around loan announcement and loan initiation date

6 Conclusion

We examine the inclusion of BCC as a contractual mechanism to reduce renegotiation costs and protect borrowers from negotiating with unknown creditors. Using the number of financial covenants to proxy for creditor control rights, we document that borrowers demand BCC as a strategic response to strong creditor control rights, suggesting that BCC serves as an important contractual mechanism to improve contracting efficiency. The baseline results are robust to using covenant tightness as an alternative proxy for creditor control rights. Furthermore, we find that performance covenants result in higher likelihood of BCC inclusion than capital covenants do. Using the inception of CDS trading as a unique setting of weakened creditor control rights, we find that the inception of CDS trading is associated with a lower likelihood of BCC inclusion.

To address the potential endogenous issues, we employ three approaches including a simultaneous equation framework to jointly estimate BCC inclusion and creditor control rights, propensity score matching to address endogeneity due to observable firm and loan characteristics, and exploiting the adoption of ARL as a quasi-natural experiment and an exogenous shock to creditor rights. We find that the baseline results are robust to addressing endogeneity issues using all these approaches. Moreover, cross sectional analyses reveal that the positive association between BCC and creditor control rights is stronger for quality borrowers and borrowers with more conservative financial reporting. Our study contributes to the extant literature by documenting that BCC serves as an important contracting mechanism to address contract incompleteness and improve contracting efficiency in the originate-to-distribute banking era.

Appendix 1: Variable definitions

Variables	Definitions
BCC	An indicator variable that equals one if a loan deal contains a borrower consent clause and zero otherwise
Min Assi	An indicator variable that equals one if a loan deal contains a minimum assignment clause and zero otherwise
Lead lender consent	An indicator variable that equals one if a loan deal contains a lead lender consent clause and zero otherwise
Restriction on buyer type	An indicator variable that equals one if a loan deal contains a restriction on buyer type clause and zero otherwise
Institutional Investor OK	An indicator variable that equals one if a loan deal contains an institutional investor OK clause and zero otherwise
Min Holding	An indicator variable that equals one if a loan deal contains a minimum holding clause and zero otherwise
FinCov	The total number of fin covenants contained in a loan deal
Increase_Chargeoff	The number of lead lenders with an increase in the net chargeoff in the quarter prior to loan syndication
Predicted_FinCov	The number of predicted financial covenants obtained from the first-stage regression model (Eq. (2)) in the instrumental variable estimation. The instrument is Increase_Chargeoff
SIC2_BCC	The percentage of loans with BCCs syndicated in the borrowing firm’s SIC 2-digit industry and in the year prior to the firm’s loan year
Predicted_BCC	The predicted probability of including BCC in debt contracts from the first stage regression model (Eq. (2) in the instrumental variable estimation. The instrument is the percentage of loans with BCC syndicated in a firm's SIC two-digit industry and the year prior to the firm's loan year (SIC2_BCC)
PerfCov	The number of performance covenants in a loan deal based on Christensen and Nikolaev’s (2012) classification, including (1) Cash interest coverage ratio, (2) Debt service coverage ratio, (3) Level of EBITDA, (4) Fixed charge coverage ratio, (5) Interest coverage ratio, (6) Ratio of debt to EBITDA, and (7) Ratio of senior debt to EBITDA
CapCov	The number of capital covenants in a loan deal based on Christensen and Nikolaev’s (2012) classification: (1) Quick ratio, (2) Current ratio, (3) Debt-to-equity ratio, (4) Loan-to-value ratio, (5) Ratio of debt to tangible net worth, (6) Leverage ratio, (7) Senior leverage ratio, and (8) Net worth requirement
FinCov_Tightness	The aggregated measure of covenant tightness based on all financial covenants (Demerjian and Owens 2016)
PerfCov_Tightness	The aggregated measure of covenant tightness based on all performance covenants (Demerjian and Owens 2016)
CapCov_Tightness	The aggregated measure of covenant tightness based on capital covenants (Demerjian and Owens 2016)
Post_CDS	An indicator variable that equals one if a loan deal is initiated after the inception of CDS trading and zero otherwise
Log(Assets)	Natural logarithm of total assets
Mkbk	Market-to-book ratio
Tang	Tangible assets (ppent/at)
Roa	Return on assets (ni/at)
Inv Grade	An indicator variable that equals one if a borrower has an S&P long-term debt rating of BBB or above, and zero otherwise
Non-Inv Grade	An indicator variable that equals one if a borrower has an S&P long-term debt rating of below BBB or does not have a credit rating, and zero otherwise
Levg	Leverage (dltt/at)
Fc	An indicator variable for financial constraint that equals one if a firm’s Whited and Wu (2006) financial constraint index is greater than the sample median, and zero otherwise. Whited and Wu (2006) index = − 0.091 * cash – 0.062 * dvt + 0.021 * dltt/at – 0.044 * log(at) + 0.102 * industry_sale_growth – 0.035 * sale_growth
Not_Rated	An indicator variable that equals one if a firm has S&P long-term credit rating and zero otherwise
Secured	An indicator variable that equals one if a loan deal contains collateral requirement and zero otherwise
Deal/Assets	deal mount/total assets
Log(Maturity)	Natural logarithm of loan maturity. The longest maturity of all tranches in a loan deal is used for the loan deal
Leveraged loans	An indicator variable that equals one if a firm’s credit rating is below BBB or a firm does not have a credit rating, and zero otherwise
Revolver	An indicator variable that equals one if a deal contains at least one revolver loan and zero otherwise
ARL	An indicator variable that equals 1 if a loan deal is syndicated in 1995–1996 or 1998–1999 for a borrower headquartered in Texas or Louisiana or syndicated in 1999–2000 or 2002–2003 for a borrower headquartered in Alabama; and 0 if a loan is syndicated for a borrower headquartered in one of the non-adopting neighboring states
Post	An indicator variable that equals 1 if a loan is borrowed by a firm in Texas, Louisiana or their neighboring states (Alabama or its neighboring states) in 1998–1999 (2002–2003) and 0 otherwise
AT1 and Hat1	A measure of accounting conservatism based on the Basu (1997) model in which accounting earnings (epsx/pricet−1) is regressed on stock returns (Rit), an indicator variable for negative Rit (D), and an interaction term D × Rit. The coefficient on D × Rit represents the relative speed of economic losses are recognized into earnings (AT1). The higher the coefficient on D × Rit, the more timely the losses are recognized into earnings. The firm-year accounting conservatism measure is obtained by estimating the regression for each two-digit NAICS industry over a five-year period prior to the loan syndication year. Each industry with the number of firm-years less than 200 and the number of firms less than 20 is removed from the estimation to avoid small sample bias. Hat1 is an indicator variable that equals 1 if AT1 is greater than the sample median, and 0 otherwise
AT2 and Hat2	A measure of accounting conservatism based on the Ball and Shivakumar (1997) model in which accounting total accruals ((ib-oancf)/att−1) is regressed on cash flow (CFO: oancf/att−1), an indicator variable for negative cash flow (DCFO), and an interaction term DCFO × CFO. The coefficient on DCFO × CFO represents the relative speed of economic losses are recognized into accruals (AT2). The higher the coefficient on DCFO × CFO, the more timely the losses are recognized into accruals. The firm-year accounting conservatism measure is obtained by estimating the regression for each two-digit NAICS industry over a five-year period prior to the loan syndication year. Each industry with the number of firm-years less than 200 and the number of firms less than 20 is removed from the estimation to avoid small sample bias. Hat2 is an indicator variable that equals 1 if AT2 is greater than the sample median, and 0 otherwise
AT3 and Hat3	A measure of accounting conservatism based on the Ball and Shivakumar (1997) model in which accounting change in net income (∆Eit) is regressed on change in net income in year t−1(∆Et−1), an indicator variable for the negative change in net income in year t−1 (D∆Et−1), and an interaction term D∆Et−1 × ∆Et−1. A more negative coefficient on D∆Et−1 × ∆Et−1 implies more timely recognition of the losses that are transitory and reverse sooner (AT3). The firm-year accounting conservatism is estimated same as the AT1 for each 2-digit NAICS industry. Hat3 is an indicator variable that equals 1 if AT3 is negative, and 0 otherwise
High_Mkbk	An indicator variable that equals 1 if firms' market to book ratio is greater than the sample median and 0 otherwise
Rated	An indicator variable that equals 1 if a firm has long-term S&P debt rating and 0 otherwise
Large	An indicator variable that equals 1 if firms' assets is greater than the sample median and 0 otherwise
Unconstrained	An indicator variable that equals 1 if a firm's WW financial constraint index smaller than the sample median and 0 otherwise
Institutional	An indicator variable set to one if the loan is designed to be sold to institutional investors, and zero otherwise
CAR	Cumulative abnormal returns adjusted for value weighted market returns around loan announcement or loan initiation
Log(Maturity)	The natural logarithm of the number of days between the quarter end date to the loan maturity date
Mkbkq	Market-to-book ratio at quarter end
Roaq	Quarterly Return on Assets
Log(assetsq)	log(at) at quarter end
Traded	An indicator variable that equals one if a loan is traded in a quarter and zero otherwise
Log(loan age)	Natural logarithm of the days between loan initiation and the quarter end date