CDS trading and nonrelationship lending dynamics

In a concurrent paper, Shan et al. (2016) also document the higher likelihood of loans syndicated by nonrelationship lead arrangers following CDS initiation. They attribute this finding to borrowers switching to new lead arrangers because their current borrower-lead-arranger relationship is compromised by CDSs (i.e. borrowers switch away from relationship lenders because they are concerned about these lenders’ weaker monitoring incentives). In contrast, we suggest that CDSs decrease adverse selection between incumbent and nonrelationship lenders, thus increasing the latter’s willingness to compete for new borrowers and consequently the likelihood of winning a loan deal. Our findings with respect to nonrelationship lead arrangers’ monitoring incentives further undermine Shan et al.’s (2016) motivation, as we show that relationship lead arrangers actually do not decrease their monitoring following CDS initiation, while nonrelationship lead arrangers monitor substantially less intensively than relationship lead arrangers.

Hedging channel is important, even if CDS spreads reflect private information, because this information may not fully reveal relationship lenders’ knowledge. Although CDS spreads provide information about the overall credit riskiness of the borrower, nonrelationship lenders cannot pinpoint the drivers of this riskiness. This should allow relationship lenders to draw more accurate inferences about the borrower and loan quality. Further, frictions associated with CDS trading are likely to prevent CDS spreads from being fully revealing (Blanco et al. 2005; Batta et al. 2016).

Corporate loan securitizations differ from the securitization of other loan types, such as mortgages, where originating banks indeed sell off the entire mortgage to the securitization SPV.

Consider hypothetically that a lead arranger can sell its entire share to a CLO or on the secondary market. In this case, the lead arranger relinquishes its control rights to the loan buyer and essentially forgoes the lending relationship with the borrower. Therefore, if a nonrelationship lead arranger wants to establish a new lending relationship with a borrower, it is unlikely to sell its entire share. In contrast, when a lead arranger hedges its credit exposure via CDSs, it retains control rights and its lending relationship because hedging does not require the lender to sever a loan contract with the borrower, further highlighting the benefits of the CDS setting for exploring our research question.

High riskiness of these borrowers is explained by the demand of CLOs and institutional secondary market buyers for loans with very high interest spreads (e.g., Bushman et al. 2010; Benmelech et al. 2012; Lim et al. 2014; Bozanic et al. 2018; Loumioti and Vasvari 2019; Addoum and Murfin 2020).

We acknowledge that, with respect to our second channel of the revelation of private information in the CDS spread, loan fair values as estimated and reported by CLOs and secondary loan prices may also reflect private information. However, as we discuss above, loan fair values and secondary prices are not available for the largest segment of the loan market—investment grade and lower risk non-investment grade borrowers. CDS trading also provides us with a more powerful setting to explore the reduction in adverse selection and its effect on nonrelationship lending, because it works through both the information revelation and the hedging channel, with the latter being largely unavailable for lead arrangers via securitization and secondary market sale channels.

Amiram et al. (2017) are interested in exploring the overall effect of CDSs on information asymmetry between the lead arranger and syndicate participants and therefore do not distinguish empirically between adverse selection and moral hazard. However, Sufi (2007) suggests that moral hazard is the key problem associated with information asymmetry within the syndicate and emphasizes that moral hazard, and not adverse selection, forces the lead arranger to retain a larger share of the loan and form a more concentrated syndicate. This suggests that Amiram et al. (2017) findings may be attributed primarily to the moral hazard problems within the syndicate.

One may argue that hedging is likely to be more expensive for nonrelationship lead arrangers, because CDS sellers should price their weaker monitoring incentives, as moral hazard problems increase the costs of insurance protection (Arrow 1963; Pauly 1968; Zeckhauser 1970; Shavell 1979; Arnott and Stiglitz 1991). However, the positive relation between the cost of insurance and the moral hazard hinges on two primary assumptions: 1) insurers can identify protection buyers with moral hazard problems, and 2) insurers bear higher expenses, due to the moral hazard problem of the insured. These assumptions are unlikely to hold in the CDS setting. First, because of the very high volume of speculative trades in the CDS market (e.g., Brunnermeier 2009; Oehmke and Zawadowski 2016), CDS sellers are mostly unable to determine whether the buyer’s intent is hedging or speculative trading. Thus, if CDS sellers cannot separate hedgers from speculators, they cannot require higher CDS spreads from nonrelationship lenders with a greater moral hazard problem. Second, CDS sellers do not necessarily bear the costs of credit events because they mostly maintain matched positions (D’Errico et al. 2018). Specifically, when CDS sellers write a CDS contract, they typically resell it to other market participants by employing opposite trades (i.e., back-to-back trades), which minimizes their credit risk. Thus, even if CDS sellers can recognize weaker monitoring incentives of nonrelationship lead arrangers, they may not bear the related costs and thus do not incorporate them into the CDS spreads. Consistently, in untabulated analyses, we find that borrowers that issue loans with nonrelationship lead arrangers do not have higher CDS spreads, relative to borrowers of relationship lead arrangers.

The vast majority of the CDS initiations are in years 2001–2007, ranging from 23.72% in 2001 to 4.39% in 2006. Around 58% of the CDS firm observations are post-CDS initiation.

This variable takes the value of 0 for all loans to non-CDS firms.

Our findings are unchanged when we control for the weighted-average loan characteristics of all loans in the package, where weights are based on loan (facility) size (untabulated).

We do not control for the number of covenants and performance pricing provisions, as these characteristics are typically determined during the loan negotiation and therefore cannot affect lenders’ choice of whether to start a new lending relationship. In untabulated analyses, we find that our results are robust to the inclusion of these variables.

The number of observations in these analyses is smaller than that in the full sample because a logit estimation drops observations if the binary dependent variable takes the same value within the same firm or year, while an OLS estimation drops observations if there is only one observation for each firm or year-month (i.e., within each fixed effect group).

Note that we measure economic significance based on the OLS specification for all tests.

With respect to controls, the negative and significant coefficient on Leverage suggests that higher leverage deters nonrelationship lead arrangers. The positive and significant coefficient on Tangibility implies that higher asset tangibility attracts nonrelationship lead arrangers. When borrowers issue longer maturity loans and those with guarantors, we find that these loans are more likely to be arranged by nonrelationship lenders.

We measure reporting conservatism following Khan and Watts (2009) and accrual quality based on the modified Jones model, as adjusted by Hribar and Nichols (2007).

To classify covenants as capital covenants, we follow Christensen and Nikolaev (2012). For loans with zero covenants, the Performance covenants variable takes the value of zero. In untabulated robustness tests, we find consistent results when we restrict our sample to loan packages with available covenant data.

If a loan package has more than one lead arranger (6.5% of sample packages), all lead arrangers are accounted for in estimating model (3). Our results are unchanged when we exclude these deals from the analyses.

We exclude #Covenants (PP) from the model when we examine performance covenants (interest increasing provisions), as it is highly correlated with the dependent variable.

In addition, while nonrelationship lead arrangers retain higher loan share in the pre-CDS initiation period, as reflected by the positive and significant coefficients on No Relationship Lead Arranger, the sum of the coefficients on No Relationship Lead Arranger and POST × No Relationship Lead Arranger indicates that the loan share nonrelationship lead arrangers retain in the post-period resembles that of relationship lead arrangers. This evidence further highlights the weakening of their monitoring incentives.

Consistently, Asquith et al. (2005) show that the interest rate at loan origination decreases with the inclusion of the interest-rate-increasing but not with the interest-rate-decreasing performance pricing provision.

Because Amount is in natural logarithm, we take the exponential of the coefficient on POST × No Relationship Lead Arranger (i.e., exp.(−0.084)-1) to calculate the economic significance of −8.1%, which translates into $93 million. The amounts represent 9.5% of our sample mean, given the average facility amount of $983 million in our sample.