Uncertainty and debt covenants

Gleick (1987), referring to analytical attempts to predict future economic events, notes, “In practice, econometric models proved dismally blind to what the future would bring” (pg. 20).

I discuss the differences in borrower- and creditor-initiated renegotiations in Section 3.3.

This perspective on Knightian uncertainty is pervasive in the literature; however, there is disagreement as to whether it is consistent with Knight’s original intent (LeRoy and Singell 1987).

This assumption is not meant to suggest that moral hazard problems (the main agency conflict to which covenant use is attributed) do not exist or are not important in contracts; rather, in my analysis, agency conflicts related to moral hazard are not the focus. I expect that moral hazard problems are relevant in debt contracts but are addressed through some means other than financial covenants (for example, negative covenants, such as dividend restrictions, or some other aspect of contract design or capital structure). More broadly, the different provisions in debt contracts are likely in place to address different types of contracting conflicts.

Asymmetric information between the borrower and lender plays a clear role in contracting theory (Demiroglu and James 2010). To focus on symmetric uncertainty between the borrower and creditor, I assume that their information endowment is similar.

As an example, consider a scenario in which there are two possible ranges for an outcome but which range the outcome belongs to is uncertain ex ante: Range 1 is [10, 20], and Range 2 is [25, 35]. If a signal reveals a low outcome, it will be clear that the outcome will draw from Range 1, but the signal will not provide information on where the outcome will be in Range 1.

More precisely, p must be sufficiently close to \(\hat {p}\) that the cost of changing the contract is higher than the benefit to either party of changing it.

Uncertainty is more severe when there are more unknown parameters or when there is relatively little information available about a parameter (or parameters.)

In this study’s framework, the primary lever for the creditor in setting loan terms is the interest rate; this is due to the assumption of a single investment project with a single cash flow. In reality, the creditor will be pricing the borrower as a set of investments. These investments will have multiple cash flows, and the amount and the timing of these flows will be uncertain. As such, the creditor could vary multiple loan terms, such as maturity, collateral, and restrictive provisions, when setting the contract.

A more general interpretation is that the creditor wants a covenant that is triggered when some parameter of the loan (for example, interest rate, loan amount, or maturity) is revealed to be suboptimal. The creditor will attempt to renegotiate these terms following (or in anticipation of) technical default. This illustrates an important difference between a pre-negotiated loan parameter changes (for example, performance pricing) and the more general transfer of control rights; if it is unknown which parameter(s) will be revealed as suboptimal, it is difficult to specify a menu of changes ex ante. I examine performance pricing in greater detail in Section 6.2.1.

Drucker and Puri (2009) find that some Dealscan loan-observations that report no financial covenants are data errors and that covenants are in fact included in these loans. To include these cov-lite loans, but avoid data errors, I collect a subsample of loans that have SEC filings available but no covenants reported on Dealscan. I hand-collect loan agreements from SEC filings (from 10-K, 10-Q, and 8-K filings) to verify which have no covenants; I include 429 cov-lite loans in the sample. The results from the analysis are not sensitive to inclusion of these observations.

I thank Michael Roberts for making these data public. I use the link file available in November 2014, which includes links for packages up to August 2012. For loans with an active date after August 2012, I hand-match loans to Compustat by borrower name.

Murfin (2012) notes that financial covenant intensity captures only one dimension of the package’s covenant portfolio; three other aspects—the covenants’ initial slack (initial value less threshold value of the covenant accounting measure), the variance of the underlying accounting measure, and the correlation between accounting measures—also contribute to the frequency of technical default. To the extent that initial slack and correlation vary between borrowers and packages, this confounds covenant intensity as a proxy for overall covenant protection. However, covenant intensity is commonly used in the literature (Bradley and Roberts 2004; Billett et al. 2007; Christensen and Nikolaev 2012). I examine alternative measures of covenant use, incorporating slack and correlation between measures, in a robustness test.

The BLS sorts firms by sectors based on two-, three-, and four-digit NAICS codes; the classification can be found at http://​www.​bls.​gov/​mfp/​mprdload.​htm.

https://​www.​philadelphiafed.​org/​research-and-data/​real-time-center/​livingston-survey

The controls include the number of lenders in the loan’s syndicate, an indicator for whether the loan has a covenant that restricts capital expenditures, an indicator for performance pricing, an indicator for collateral, and the loan’s maturity.

I exclude Rating from the regression for Unrated, as they are highly correlated.

Consistent with this idea, Nikolaev (2016) finds that performance covenants are associated with debt contract renegotiation, while capital covenants are not (although the paper does not distinguish between borrower- and creditor-initiated renegotiations).

I include the Market-to-Book ratio as a control variable in the main tests. This ratio has been used as a proxy for growth opportunities and thus agency conflicts (Skinner 1993; Smith and Watts 1992). The negative coefficient in the main tests is inconsistent with the predictions of agency (wherein higher levels of the ratio mean greater agency conflicts.) There are, however, a variety of alternative interpretations of the market-to-book ratio, including accounting conservatism (Roychowdhury and Watts 2007) and deficiencies in GAAP (Lev and Sougiannis 1999); thus I do not focus on this result as providing evidence related to agency conflicts.

After the attacks, the NYSE and NASDAQ did not reopen until September 17. Although the private loan market was not directly affected, I exclude loan packages initiated from September 11 through 16, due to overall disruption in the financial markets. There are only six packages recorded for this period, and their inclusion in the test subsample does not affect inferences.

There are two threats to the validity of the Sept. 11 attacks as an exogenous shock to uncertainty. First, the Enron accounting scandal was revealed in October 2001, which falls within the six month post-period. However, the VIX did not reflect any increase in uncertainty when the scandal was initially disclosed. Specifically, from a high in the mid 40s immediately following Sept. 11, the VIX trended downward and into the low 20s and high teens through May 2002. Uncertainty only increased again in June 2002, coinciding with revelations about accounting problems at Worldcom. It is likely that the market viewed Enron as idiosyncratic at the time that the scandal was made public, and only when another major scandal broke, did concern grow that accounting issues were widespread. As such, I do not expect that the Enron scandal contaminates the post-shock period. Second, the U.S. economy was in recession from March to November 2001. If a recession influences the use of financial covenants, for example, with increased macroeconomic risk leading to greater demand for covenants, this could limit the inferences that can be drawn from this sample. This recession, however, was short and shallow and, more importantly, spanned both the pre- and post-periods. As such, it should not damage inferences from this test.

It is also notable that Market-to-Book is positive and significant in regressions for Dividends and Collateral. To the extent this variable captures some aspects of agency conflicts, this supports the idea that dividend restrictions and collateral requirements are associated with agency conflicts.

Asquith et al. (2005) present evidence that performance pricing addresses adverse selection costs.