Strategic syndication: is bad news shared in loan syndicates?

Highlighting the lack of disclosure rules for government investigations, Blackburne et al. (2021) and Solomon and Soltes (2021) show that many firms do not disclose SEC investigations.

Our arguments also imply that lead arrangers will withhold not only negative but also clean inspection news. Although lead arrangers may benefit by disclosing clean inspection outcomes to syndicate participants, these disclosures are likely to increase participants’ awareness of FDA inspections and the associated risk for borrowers. This greater awareness will hinder lead arrangers’ ability to withhold negative inspection outcomes.

In supplementary analyses, we do not find that loans issued following bad inspection outcomes in the pre-OGI period have a higher loan pricing or more restrictive non-price contractual terms. These findings indicate that syndicate participants are not compensated or better protected for the greater risk associated with borrowers’ worse performance following bad inspection outcomes.

Some of the participants in a loan syndicated by J.P. Morgan just before Enron’s bankruptcy filing accused the lead arranger of deliberately concealing Enron’s financial condition and of using part of the loan proceeds to lower its own exposure to the company. Similar allegations of withholding information on Enron’s financial problems have been made against Citigroup, which has also served as a lead arranger on some of Enron’s syndicated loans.

Panyagometh and Roberts (2010) find that the larger percentage of a loan syndicated (i.e., the lower loan share retained by the lead arranger) is not associated with borrowers’ worse financial performance subsequent to the loan origination, as measured by changes in borrowers’ Z-scores. Although these findings suggest that lead arrangers do not exploit syndicate participants, this paper’s research design is unlikely to be strong enough to uncover lead arrangers’ misbehavior. The strength of our study, which enables us to document lead arrangers’ opportunistic behavior, is that the FDA inspection is a well-defined event, providing us with a unique opportunity to capture private information that is likely available to lead arrangers but not to syndicate participants in the pre-OGI period. Our research design also incorporates an exogenous disclosure regime change with the passage of the OGI and allows us to directly identify, through FDA FOIA logs, specific instances where the lead arranger is likely to possess information about inspection outcomes but participants are unlikely to be aware of it.

The FDA is responsible for enforcing the Federal Food, Drug, and Cosmetic Act, as well as any subsequent amendments and related Acts. Manufacturing firms must operate in accordance with these laws.

A warning letter is the final notice the firm will receive before the FDA imposes more severe enforcement actions such as fines, injunctions, recalls, seizures, or criminal prosecution. In recent years, firms with insufficient responses to a Form 483 have received a warning letter within 10 to 11 months of the triggering inspection (Unger 2018). Warning letters occur throughout both the pre-OGI and post-OGI periods. Only two of our sample firms received a warning letter prior to loan issuance, with two additional firms receiving them after the loan issuance.

For example, the January 2020 FDA Closed FOIA Log – which reports all of the requests completed in that month – reveals that the FDA only recently fulfilled requests from 2014 to 2016, along with a number of more recent ones. A review of closed FOIA logs from other months highlights that these types of time lags are not abnormal.

When interpreting disclaimers in participation agreements, a multitude of courts have dismissed claims of fraud and misrepresentation in favor of lead banks (e.g., Jimerson 2017). For example, in the Banco Espanol de Credito v. Security Pac. Nat’l Bank, 973 F. 2d 51 (2d Cir. 1992), the lead bank refused to extend credit to a borrower when it became aware of private information that the borrower was in severe financial difficulty. However, it sold participation in the loan, and the borrower ultimately defaulted. The participating banks sued Security Pacific on the basis that it had withheld material information. On both the initial judgment and appeal, the courts dismissed the case because the Participant Agreement included the disclaimer that the participant “acknowledged that it has independently and without reliance upon Security [Pacific] and based upon such documents and information as the participant has deemed appropriate, made its own credit analysis.” The appeals court also stated that the waiver provisions “specifically absolved [lead bank] of any responsibility to disclose information relating to [the borrower’s] financial condition. Moreover, as an arm’s-length transaction between sophisticated financial institutions, the law imposed no independent duty on [lead bank] to disclose information that the plaintiffs could have discovered through their own efforts.” In addition, many cases cite the Office of Comptroller of Currency (OCC) Banking Circular 181, which requires participant lenders to conduct an independent analysis of credit quality as they would make a loan directly to the borrower. Thus, even in the absence of disclaimers in the Participation Agreement, OCC guidelines may preclude participants’ claims of misrepresentation and fraud because banks should comply with sound banking practices promulgated by the OCC.

Our institutional setting where lead arrangers obtain information that is not shared with syndicate participants is not uncommon. During the loan due diligence process and over the life of the loan, borrowers often disclose sensitive information to the lead arranger which is not made available to syndicate participants (Sansone and Taylor 2007; LSTA 2017). This information is specifically referred to as “borrower restricted information,” in contrast to the “syndicate information” that is available to all the members of the syndicate.

Similar arguments and inferences have been made by the court in ACito v. IMCERA Group, Inc (1995). and Matrixx Initiatives, Inc. v. Siracusano (2011), among other cases.

The court perception changed only in 2012 (i.e., in the post-OGI period) with the new precedent of Public Pension Fund Group v. KV Pharmaceutical Co. (2012). The court ruled that the issuance of Form 483 may be material depending on a number of factors, including the number, severity, and pervasiveness of objectionable conditions noted, as well as whether a company has failed to address or correct deficiencies noted by the FDA. Thus, during the post-OGI period when inspection outcomes are disclosed by the FDA and are easily observable to participants, a bad inspection outcome might qualify as a material adverse effect and in turn trigger a violation of compliance covenants. But that would depend on deficiencies noted by inspectors as well as the borrower’s efforts in remediating the problems.

Although these cases are in the context of securities law, securities law should provide us with an appropriate benchmark to assess how to interpret, from a legal perspective, a material adverse effect in the lending setting.

A potential concern is that a lead arranger may retain a larger loan share at loan origination, but partially sell it to collateralized loan obligations (CLOs), which will subsequently securitize it, or on the secondary loan market. However, CLOs expect lead arrangers to retain most of their loan share to incentivize them to continue monitoring a borrower (e.g., Benmelech et al. 2012; Bozanic et al. 2017). Similarly, due to monitoring concerns, secondary loan market participants expect lead arrangers to retain their loan share (e.g., LSTA 2007). Further, the vast majority of borrowers whose loans are securitized or sold on the secondary loan market have higher risk non-investment grade ratings (i.e., credit ratings of BB- and below), as explained by the demand of CLOs and secondary market buyers for loans with very high interest spreads. Only 7.8% of our sample loans are issued to borrowers with BB- or below ratings; our primary findings are unchanged when these loans are excluded from the sample.

Imposing loan share data restriction results in a similar sample attrition in other studies. For example, Sufi (2007) loses 65% of sample observations when he requires data on the lead arranger share (see his Table 2). Ivashina (2009) explicitly says the data on a loan share retained by the lead arranger is available for only 30% of sample observations. Similarly, Ball et al. (2008) show in their Table 1 that this data exists for approximately 24% of their sample loans.

Since its inception in 2010, FDAzilla has built its database by routinely submitting a variety of FOIA requests to the FDA and compiling inspection data going back to 2000.We also consult borrowers’ 10-K filings, particularly Exhibit 21s, which list the names of major subsidiaries. This procedure helps us link inspections to the correct parent firm.

If a given SIC code has more than one FDA Center match, we assign the borrower without an inspection to the FDA Center with the highest number of matches. Also, please note that not all firms on the SRO list are inspected over our sample period. Therefore, if we cannot match the non-inspected borrower to an inspected firm in the same four-digit SIC industry, we match it to an inspected firm in the same three- or two-digit SIC industry.

A relatively high percentage of bad inspection outcomes in our sample should not be interpreted as an indication that inspections are mostly negative. Our sample firms tend to be large with a high number of sites. Given that inspections are performed on the site level, a large firm with several sites often experiences multiple inspections in a year. We only require one of those inspections to have a bad outcome for Bad Inspection to equal one, which likely explains the high bad inspection frequency at the firm level. For comparison, Down (2020) reports that bad inspections occur in around one-third of inspections for a more general sample of firms not restricted by Dealscan data.

Please note that we do not include, in the model, an interaction term between Inspection and Bad Inspection (Post* Inspection*Bad Inspection) because this term is identical to Bad Inspection (Post*Bad Inspection), as it takes the value of one (zero) when Bad Inspection is equal to one (zero). The same applies to Models (2)–(4).

Because a significant proportion of borrowers in our sample (32%) have only one loan, we do not incorporate firm fixed effects in the model. Furthermore, we cluster standard errors by industry because we believe that independence issues are likely to occur at this level. For example, because borrowers in the same industry tend to share similar risks, syndicate structure and loan contractual terms are likely to vary by industry. As a large fraction of sample borrowers have only one loan, firm-level clusters are unlikely to adequately address within-group correlations and will not correct for the potential independence issues. Nevertheless, we replicate our main specifications in column 1 of Table 3 by using firm-level clusters and find similar results.

One may question why lead arrangers are willing to syndicate loans following negative inspection news. It is important to note that, in contrast to syndicate participants that primarily earn the interest spread on the loan share they retain, lead arrangers have two additional income sources. Lead arrangers earn substantial origination fees that typically range from 25 to 175 basis points of the total loan amount. Lead arrangers also often gain lucrative cross-selling opportunities with the borrower, such as bond and equity underwriting and M&A advisory services (Drucker and Puri 2005; Ivashina and Kovner 2011; Mora 2015; Kang et al. 2021b). These additional revenues are likely to compensate for the increase in a borrower’s riskiness due to bad inspection outcomes.

We also verify that our findings are unaffected by the financial crisis. Loans issued during most of the crisis period are excluded from our sample. This is because, to achieve clean private and public disclosure periods, we exclude loans issued during the transitional period between January 21, 2009 (i.e., the date on which the OGI was announced), and May 26, 2011 (i.e., the date on which the FDA started disclosing inspection outcomes). In robustness analyses (untabulated), we further exclude loans issued between September 14, 2008, and January 21, 2009, such that all loans issued during the crisis period are excluded from the sample. We find that our results continue to hold.

Please note that the number of observations is smaller in Table 12 than in Table 3 because for HHI analyses we require loan share data to be available to all syndicate lenders.

We use the following Ravenpack press release topic classifications: “product-recall,” “regulatory-product-review,” “regulatory-product-warning,” “regulatory-investigation,” “sanctions,” “clinical trials,” “product-discontinued,” and “product-outage.”

The sum of the coefficients on Inspection and Bad Inspection is statistically different from zero in the low disclosure partition (p-value = 0.082) but not in the high disclosure partition.

If a loan has more than one lead arranger, we require both arrangers to submit a FOIA request to be assigned to the FOIA partition. In addition, because we have FDA FOIA logs starting in 1999 only, we may misclassify some lead arrangers as not submitting requests to the FDA if they submitted them prior to 1999. However, only a very small number of sample banks submitted FOIA requests over the 1999–2001 period, suggesting that this number is likely to be even smaller prior to 1999. Further, our findings are robust if we eliminate loans issued in the first two years of our sample period (2000–2001), so we have at least a three-year FOIA request history for all lead arrangers.

The sum of the coefficients on Inspection and Bad Inspection is negative and statistically different from zero in the Lead FOIA partition (p-value = 0.032) but not in the No Lead FOIA partition.

The sum of the coefficients on Inspection and Bad Inspection is negative and statistically different from zero in the lead arranger FOIA/less informed participant specification (p-value = 0.001), but not in the lead arranger FOIA/more informed participant specification.

The sum of the coefficients on Inspection and Bad Inspection is negative and statistically different from zero in the low lead participant partition (p-value = 0.049), but not in the high lead participant partition.

For example, a bad outcome could be very costly for a drug company trying to get a new medication to market, whereas a deficiency of an employee not washing hands is unlikely to cause significant harm to a food company.

The sum of the coefficients on Inspection and Bad Inspection is negative and statistically different from zero in the high materiality partition (p-value = 0.017) but not in the low materiality partition.

We acknowledge that lower lead arrangers’ loan share in the high materiality partition can reflect not only the higher likelihood of lead arrangers being aware of these inspections but also more severe effects of these inspections on borrowers’ performance. However, this explanation also requires the lead arranger to know about the occurrence of inspections and their outcomes, further reinforcing our main inferences.

Relative to financial statement audits, FDA inspections are shorter in duration (approximately 5–6 days) and much more focused. In particular, FDA staff do not work on multiple inspections simultaneously; in contrast, they are on-site during the entire duration of the inspection. Given that the FDA is resource-constrained, it is unlikely that the agency would allow employees to spend longer periods of time at a site that is not important, further suggesting that inspection length is a reasonable proxy for its materiality.

The sum of the coefficients on Inspection and Bad Inspection is negative and statistically different from zero in the high duration partition (p-value = 0.060), but not in the low duration partition.

Our findings are similar if we estimate Logit models when the dependent variable is default, delisting, and bankruptcy.

The sum of the coefficients on Inspection and Bad Inspection is statistically different from zero in the change in ROA, change in Z-score, and bankruptcy specifications, with p-values of 0.037, 0.097, and 0.096, respectively (the sum of the coefficients is statistically different from zero in the default and delist specifications at a one-sided level, with p-values of 0.184 and 0.171, respectively).

If the loan has more than one lead arranger, we require the proportion of reputable lead arrangers in the syndicate to be above the sample median (i.e., an above-median percentage of lead arrangers must be ranked in the top five).

Our findings are similar if we estimate OLS models when the dependent variable is #Covenants or PP Increasing.