Corporate loan securitization and the standardization of financial covenants * Zahn Bozanic The Ohio State University mailto:[email protected]Maria Loumioti USC Leventhal School of Accounting [email protected]Florin P. Vasvari London Business School [email protected]September 2014 Abstract We apply textual analysis on a large sample of financial covenant definitions to measure covenant standardization and find that securitized corporate loans include more standardized covenants. We document that financial covenant standardization increases the liquidity of securitized loans in the primary and secondary loan market. Consistent with a decrease in illiquidity premiums, covenant standardization decreases the cost of securitized loans without being associated with a lower likelihood of default. We also find that covenant standardization is associated with less disagreement between credit rating agencies, potentially contributing to the higher liquidity of securitized loans. Our findings suggest that financial covenant standardization is positively related to corporate loan securitization and has a significant impact on loan liquidity. Keywords: Securitization, Financial Covenants, Syndicated Loans, Standardization JEL Classifications: G17, G21, G32, M41 * We are grateful to Panos Patatoukas and KR Subramanyam and workshop participants at London Business School, Stockholm School of Economics, University of Southern California and University of Oulu (Finland) for their helpful comments and suggestions. We thank Blake Sainz for his excellent research assistance. Loumioti acknowledges financial support from Leventhal School of Accounting. Vasvari acknowledges funding from the London Business School RAMD Fund. All remaining errors are our own.
54
Embed
Corporate loan securitization and the standardization of financial ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Corporate loan securitization and the standardization of financial covenants*
We apply textual analysis on a large sample of financial covenant definitions tomeasure covenant standardization and find that securitized corporate loans includemore standardized covenants. We document that financial covenantstandardization increases the liquidity of securitized loans in the primary andsecondary loan market. Consistent with a decrease in illiquidity premiums,covenant standardization decreases the cost of securitized loans without beingassociated with a lower likelihood of default. We also find that covenantstandardization is associated with less disagreement between credit ratingagencies, potentially contributing to the higher liquidity of securitized loans. Ourfindings suggest that financial covenant standardization is positively related tocorporate loan securitization and has a significant impact on loan liquidity.
* We are grateful to Panos Patatoukas and KR Subramanyam and workshop participants at London Business School,Stockholm School of Economics, University of Southern California and University of Oulu (Finland) for theirhelpful comments and suggestions. We thank Blake Sainz for his excellent research assistance. Loumiotiacknowledges financial support from Leventhal School of Accounting. Vasvari acknowledges funding from theLondon Business School RAMD Fund. All remaining errors are our own.
1
1. Introduction
Although there is significant empirical evidence on the widespread use of financial
covenants in syndicated loan contracts, these covenants are typically written on a relatively small
set of accounting numbers. This is puzzling, given the large volume of accounting information in
borrowers’ financial statements and lenders’ sophistication (Skinner, 2011). In this paper, we
provide insights into this issue by exploring whether the securitization of syndicated loans
through collateralized loan obligations (“CLOs”) increases the standardization of accounting
information used in financial loan covenants. We define standardization as the process of
increasing the similarity and comparability of financial covenant definitions (e.g., De Franco,
Kothari and Verdi, 2011). In addition, we explore the real effects of financial covenant
standardization and investigate whether standardization affects the liquidity of securitized loans
by decreasing information processing costs.
Collateralized loan obligations (“CLOs”) are special purpose vehicles that are set up by an
investment bank (“CLO arranger”) and an investment management firm (“CLO manager”).1
CLOs’ investment strategy is to profit from the difference in the average interest rate on the
corporate loans they buy (“CLO collateral”) and the interest rate on the debt issued to finance the
acquisition of these loans (“CLO notes”). To achieve this interest rate differential, CLOs invest
in a large and highly diversified pool of corporate loans. Consequently, a CLO ends up holding
small tranches in more than 200 corporate loans from various borrowers covering 15 to 25
different industries. The large amount of accounting information that describes financial loan
covenants and determines creditors’ control rights associated with securitized loans can generate
1 CLOs grew to become the dominant institutional investor in the syndicated loan market reaching a 60 percentmarket share and securitizing syndicated loans with a total value of about $100 billion annually before the creditcrisis. Thereafter, by 2013, the level of annual investments in corporate loans by CLOs nearly reached pre-crisislevels (Standard and Poor’s, 2014).
2
significant transaction costs.2 Such costs increase with the extent to which financial covenant
structures become more complex and borrower-specific since assessing these covenants involves
more extensive information collection, monitoring and enforcement efforts.
However, certain mechanisms specific to the CLO operating model constrain these costs.
First, the selection of leveraged corporate loans as eligible CLO collateral relies on specific and
predetermined diversification criteria on borrowers’ industry and geography as well as loans’
maturity and rating category. These restrictions are imposed at the CLO set-up stage by credit
rating agencies that rate CLO notes to diversify away the idiosyncratic credit risk of each
individual loan investment. Thus, covenant-based metrics are largely ignored in determining the
structure of the CLO pool.3 Second, CLO managers’ performance is monitored by specific
compliance tests such as over-collateralization criteria of the CLO notes and average loan rating
thresholds for the CLO collateral. These monitoring mechanisms exclude information related to
the covenant structure of the loans in the pool, since assessing the quality of so many covenants
and the accounting information used in covenant thresholds is costly and induces subjectivity.
Third, the set of loan characteristics disclosed to CLO investors does not include details about
financial covenants, consistent with the fact that investors place less weight on this information
to monitor CLO performance or face information processing costs themselves. Thus, CLO
investors receive information only on a narrow set of loan characteristics, such as loan
maturities, spreads, ratings and default rates which simplify disclosures about CLO portfolio
quality.
2 The transaction costs are potentially high given the typical size of the marginal investment that a CLO makes in anindividual loan. In our sample, the average size of an investment in a loan is $1.5 million, while the face value of theloan is $350 million.3 After 2010, about 50 percent of the CLOs issued included restrictions on the percentage of covenant-lite loans inthe CLO portfolio. Nevertheless, the average cap on the amount of covenant-lite loans has increased from 25-30percent in 2010-2011 to 50-60 percent in 2013 (Standard & Poor’s Rating Direct – Structured Finance, 2013).
3
Because CLOs’ operating model has limited reliance on covenant information content or
covenant quality, we expect that CLOs will contribute to an increase in the standardization of
covenants. At the same time, loan underwriters are also likely to limit their use of customized
covenants. They often employ financial covenants and contractual choices from loan agreements
of prior borrowers to lower their contracting costs, further contributing to standardization across
covenant definitions (Simpson, 1973; Rajan and Winton, 1995; Choi and Triantis, 2014). When
underwriting banks prepare the documentation to launch a syndicated corporate loan, they
regularly start with their own preliminary term sheets for financial covenants. The covenant term
sheets are subsequently adjusted as underwriting banks negotiate with and receive feedback from
loan investors. Similarly, syndicated loans securitized right after their origination are more likely
to include standardized financial covenants, since loan underwriters will exert less effort to write
customized loan covenants with borrower-specific accounting information if these loans are
subsequently transferred to CLOs.
However, the rise of corporate loan securitization may not necessarily increase financial
covenant standardization primarily for two reasons. First, only a fraction of syndicated loan
tranches is securitized while the remaining tranches are sold to banks or other investors that do
not have similar incentives to CLOs. Second, syndicate members may negotiate complex and
borrower-specific financial loan covenants to obtain pecuniary benefits and/or ex post control
rights (Li, Vasvari and Wittenberg-Moerman, 2014). For example, when loans are renegotiated
due to covenant violations, lenders obtain significant benefits such us renegotiation fees, greater
interest rates or more control over the borrower’s investing and financing activities (e.g., Roberts
and Sufi, 2009; Roberts, 2013).
4
We investigate the relationship between corporate loan securitization and financial
covenant standardization using a sample of US-based public companies which issued syndicated
loans in the period 2000–2009. We obtain data on securitized loans from Creditflux, a global
publication platform that covers detailed information on the origination and performance of
CLOs’ investment portfolios. We match these loans with LPC DealScan to obtain their
characteristics and Compustat to obtain borrowers’ financial information. For those loans with
complete Dealscan and Compustat data, we then retrieve the loan contracts from companies’
SEC filings in EDGAR. We are able to obtain a sample of 440 securitized and 703 non-
securitized institutional loan contracts. For both securitized and non-securitized loans, we hand
collect 3,303 financial loan covenant definitions. We focus our analysis on the complete
covenant definition rather than the covenant title since previous studies report that the definitions
of accounting terms vary substantially across financial covenants (e.g., Leftwich, 1983; Li,
2012).
To assess financial covenant standardization, we develop an empirical proxy by
measuring the similarity of the contracting language that is used to define individual covenants.
For each covenant, we calculate the number of words that overlap with the words in covenant
definitions of loans issued by other borrowers over the prior calendar year. Namely, we compute
the cosine textual similarity between covenant definitions using a vector space model similar to
models used in plagiarism software and search engine algorithms (e.g., Salton, Wong, and Yang,
1975).4 This approach has recently been introduced in the accounting and finance literatures
4 More specifically, cosine textual similarity is computed as follows: we take two complete definitions of similarfinancial covenant types from two loans of different borrowers. We identify and list all the words in thesedefinitions, excluding “stop-words” and keeping only the word stems. Then, we count how many times each word isused in each definition. This process creates two vectors with the number of times each word is mentioned in thetwo covenant definitions. The cosine of the angle between these vectors is our covenant similarity score. Moredetails on how the measure is computed are provided in Appendix B.
5
(e.g., Brown and Tucker, 2011; Hoberg, Phillips and Prabhala, 2012; Bozanic and Thevenot,
2014). Since covenants are set at the loan level, we estimate the covenant standardization
measure for each loan by averaging the cosine similarities of its covenants with the same-type
covenants in all loans issued by other borrowers in the prior calendar year. The covenant
similarity score is a continuous variable with values ranging from zero (if two covenant
definitions share no common word) to one (if the definitions of two same-type covenants are
identical).5 Using a multivariate regression, we show that the covenant similarity score is higher
when borrower and loan characteristics are more similar and that these similar characteristics
explain a significant proportion of the variation in the covenant similarity score, thus validating
our empirical proxy.
We show that financial covenant standardization in loan contracts increased during 2000–
2007 and drastically dropped in 2008–2009, matching the evolution of the corporate loan
securitization volume over the 2000–2009 period. In multivariate analyses, we find that
corporate loan securitization is positively associated with covenant standardization, controlling
for borrower, loan and underwriter characteristics. More specifically, we find that securitization
increases our covenant similarity score by up to 20 percent of its standard deviation. These
results are robust to a propensity score matched analysis on borrower performance and loan
characteristics, as well as to tests which address the potential for reverse causality (i.e., whether
CLOs purchase loans with more standardized financial covenants).
Next, we investigate whether the standardization of financial covenants in securitized
loans affects loan liquidity in the primary and secondary syndicated loan market. Consistent with
standardization contributing to a decrease in screening costs and costly information disclosures
5 In addition, we use an alternative covenant standardization measure which is the average of the ratio of similarcovenants across the loans issued in the prior year. Our results are robust to this measure.
6
(e.g., Amihud and Mendelson, 1988), we find that covenant standardization is negatively related
to the number of days a loan remains open after its launch date, our loan liquidity measure in the
primary loan market. An increase in the covenant similarity score by one standard deviation
decreases the time a loan remains open in the primary market by 3 trading days or 21 percent of
the standard deviation of the “time-on-the-market.” Moreover, we investigate the liquidity of
securitized loans in the secondary loan market and find strong evidence that securitized loans
with more standardized covenants trade more and are purchased by a greater number of CLOs.
This finding suggests that covenant standardization contributes to a decrease in CLOs’ and their
counterparties’ information processing costs when trading.
In complementary analyses, we investigate whether covenant standardization is
associated with a reduction in the illiquidity premiums reflected in the securitized loan’s spreads.
We document a negative relation between the covenant similarity score and the LIBOR-spreads
of securitized loans. A one standard deviation increase in covenant similarity decreases the
LIBOR-spread by 5 percent or 12 basis points. In addition, we do not find evidence that
borrowers are less likely to default on securitized loans with more standardized financial
covenants. This latter finding suggests that the lower spread is not due to securitized loans’ lower
propensity to default but could be due to a lower illiquidity premium as a result of the
expectation that these loans will be more liquid.
Finally, we investigate whether covenant standardization decreases the information
asymmetry between debt market intermediaries, thus facilitating loan trading. We find a positive
relation between covenant standardization and the agreement in securitized loans’ ratings issued
by Standard & Poor’s and Moody’s. Less disagreement in the credit assessments of these top two
7
rating agencies likely reduces the uncertainty among CLO managers and their counterparties
when trading these loans.
We add to the debt literature in several ways. First, we provide novel evidence on how
developments in the credit market affect the standardization of covenants in loan contracts. We
show that corporate loan securitization, which relies on significant secondary loan market
trading, contributes to more similar financial covenant definitions across syndicated loans. As
such, our study is related to De Franco, Vasvari, Vyas and Wittenberg-Moerman (2013) who
find that bond covenant “stickiness” is partly driven by bond market intermediaries, such as lead
arrangers and legal advisors, who prefer standard covenant definitions. Also, by identifying a
loan market mechanism that amplifies financial covenant standardization, we add to the well-
established empirical literature on the factors that drive contractual terms in corporate loans. So
far, this literature has mainly investigated to role of agency based determinants (e.g., Beatty and
Weber, 2003; Asquith, Beatty and Weber, 2005; Bharath, Sunder and Sunder, 2008; Beatty,
Weber and Yu, 2009; Ball, Bushman and Vasvari, 2008).
Second, we provide first hand evidence on the consequences of debt contract
standardization, and in particular covenant standardization, with respect to loan liquidity in the
primary and secondary debt market. Thus, we add to the empirical literature on corporate loan
securitization (e.g., Ivashina and Sun, 2011; Nadauld and Weisbach, 2011) and secondary loan
trading (Wittenberg-Moerman, 2008) by highlighting an important determinant of loan liquidity
that affects information processing costs.
Third, we build on recent studies that investigate the important role of textual information
in corporate disclosures (e.g., Li, 2008; Hoberg and Phillips, 2010; Brown and Tucker, 2011;
Bozanic, Cheng and Zach, 2013). We assess the complexity of loan covenants’ specifications
8
relative to covenants in the loan contracts of peer firms and its effect on the marketability of
syndicated loans. We also explore how a recently developed credit market mechanism, the
corporate loan securitization process, is shaping debt contracting language (e.g., Bozanic, Cheng
and Zach, 2013). Consequently, we show that contract standardization is not only initiated by
lawmakers (e.g., Smith, 2006), but also by market participants that have incentives to induce
contractual standardization.
2. Literature review and Hypothesis development
The role of accounting-based loan covenants in mitigating adverse selection and moral
hazard has been widely explored in the accounting and finance literatures (e.g., Smith and
Warner, 1979; Berlin and Mester, 1992; Rajan and Winton, 1995). Bank lenders often structure
loan covenants based on financial statement data and use accounting adjustments to better
by selecting corporate loans based on specific and predetermined diversification criteria
regarding borrowers’ industry and geography as well as loans’ maturity and rating category.
These restrictions are imposed upon the CLO at set-up stage by credit rating agencies that rate
the CLO’s notes. Thus, covenant-based metrics are largely ignored in determining the structure
of the CLO pool.7 Second, CLOs are monitored based on certain predetermined compliance tests
6 CLO managers also trade loans often (Lou, Loumioti and Vasvari, 2014). A detailed assessment of individualloans’ level of covenant protection at the time when a loan is purchased can increase significantly the transactioncosts.7 In fact, prior to 2010, most CLOs had no constraints with respect to the acquisition of covenant-lite loans, thusencouraging an extreme form of covenant standardization.
11
that are not built on loan covenants, but on a closed set of loan characteristics (primarily on loan
ratings and maturity).8 Despite their important monitoring role, loan ratings only partially capture
the structure of financial loan covenants that facilitate lenders’ control rights due to the
complexity of the terms in syndicated loan contracts (Ayotte and Bolton, 2009). Since CLO
managers are not evaluated on covenant-based metrics, they are less likely to be interested in
loans with customized financial covenants that would allow them to monitor borrowers’
underlying business model and financial performance. Third, CLOs report to investors only a
few loan characteristics, such as maturities, ratings, and spreads, ignoring the structure, number,
or quality of the financial loan covenants.
Since financial covenants are not an important loan feature for CLOs’ business model, we
expect that CLOs will not demand customized loan covenants. As a result, CLOs will be less
likely to negotiate and provide feedback on financial loan covenants to loan underwriters. Also,
to the extent that an originating bank expects to securitize a loan immediately after its issuance
by transferring it to a CLO, it will not exert significant effort to customize the covenant terms.
On the other hand, financial covenants are set at the loan package level, and not all tranches in
the loan deal are securitized. Banks and other loan syndicates that keep these tranches on their
balance sheets may have incentives to demand more borrower-specific financial covenants that
meet strict internal risk management rules. Also, because bank lenders gain access to borrower-
specific private information via their relationships with borrowers and have low renegotiation
costs, they might favor more customized financial covenants that enhance their control rights and
limit their credit exposure to borrowers (e.g., Li, Vasvari and Wittenberg-Moerman, 2014).
8 Some of the compliance tests are the overcollateralization of senior and junior CLO securities, the averageweighted rating of the collateral pool, the percentage of loans in the risky CCC-bucket, and the percentage of loansfrom borrowers that defaulted on their payments.
12
We further expect more standardized financial covenants to bring benefits to the CLO
managers who invest in loans with such covenants. Specifically, standardized covenants may
improve the marketability of syndicated loans in both primary and secondary loan markets. This
is because covenant standardization decreases trading counterparties’ contract reading costs and
information asymmetry, thus contributing to higher loan liquidity. As such, standardized
covenants reduce the need for costly information disclosures and additional accounting due
diligence (e.g., Amihud and Mendelson, 1988).
3. Sample selection
We obtain data on securitized corporate loans from the CLO-i database provided by
Creditflux. Creditflux is a global news platform covering structured investment issuance and
performance that has been tracking data on all CLO deals since January 2008. Creditflux
retrieves its data from monthly CLO trustee reports that disclose CLOs’ activities and securitized
loans’ performance to CLO investors. CLO-i includes complete data on CLO portfolio structure,
CLO compliance tests, and CLO transactions, including borrowers’ names, loan types, ratings,
balances, maturities and default events. We retrieve loan specific data from LPC DealScan which
provides information on loan terms, loan types, lenders in the syndicate as well as the period a
loan package is marketed in the primary loan market.
We match CLO-i data with LPC DealScan and Compustat databases, a process which
yields a sample of 1,075 unique securitized corporate loans issued by 605 unique public
borrowers during the period 2000–2009. Of those, we are able to retrieve the loan contracts for
440 securitized loans from borrowers’ SEC filings via EDGAR following the search procedure
outlined by Nini, Smith, and Sufi (2009). To ensure comparability with our sample of securitized
loans, we then match the securitized loans to a sample of institutional loans identified in LPC
DealScan. We focus on institutional loans to eliminate the effect of differences between the
13
middle and highly leveraged loan market on contract design.9 In addition, unlike bank loans,
institutional loans are typically rated (Ivashina and Sun, 2011). Thus, we try to hold constant the
demand for homogenous loans from credit rating agencies that prefer these loans because they
facilitate comparisons of credit risk levels.
We classify a loan as institutional if it includes at least one term loan tranche B-H but
does not include a CLO in its primary syndication structure, as presented in LPC DealScan.10 To
improve the classification, we require that institutional tranches have LIBOR-spreads higher than
250 basis points since institutional investors typically buy into high-yielding loans. Based on
these filters, the total number of non-securitized institutional loans issued by public borrowers in
LPC Dealscan is 4,529 over the period 2000-2009. We then select institutional loans with
available data from a subsample of 1,951 loans where more than half of the tranches are
institutional; this requirement ensures that we do not select loans that are distributed mainly to
banks. From this sample, we are able to retrieve the actual loan contracts of 703 institutional
loans from the SEC filings in the EDGAR system. Our final sample therefore includes 1,143
unique loans (440 securitized and 703 non-securitized institutional loans) issued by 806 unique
borrowers.
Next, we hand collect the accounting-based covenants of the loan contracts in our
sample. Since lenders may use different language to describe a type of covenant, we categorize
covenants into twelve types based on the LPC DealScan classifications: “Max. Capex”, “Max.
9 More specifically, middle market loans are generally issued by more financially healthy borrowers and are nottraded, while institutional loans are primarily issued by non-investment grade borrowers and are largely distributedto institutional investors that may subsequently trade these loans.10 It is likely that we misclassify some institutional loans as non-securitized. This is because institutional loans mighthave been sold after their origination to CLOs. The average CLO holding period is approximately 11 consecutivemonths, and approximately 2 years in total. To mitigate this concern, in a robustness test we rerun the analysis forloans originated after January 2005 and the results hold.
“Min. Interest Coverage”, “Min. Liquidity”, and “Min. Net Worth.” We identify 3,303 unique
financial loan covenant definitions. We find that 156 loans (55 securitized and 101 non-
securitized) have no accounting-based covenants (i.e., they are “covenant-lite” loans). While two
loans may use the same financial covenant type, the definition of accounting terms across
contracts could vary significantly. Thus, we hand collect the definition of the accounting terms
used to define the financial covenants in our sample. For example, when the “Interest Coverage
Ratio” is defined as “EBITDA to Interest Expenses”, we collect the accounting definition for
EBITDA and interest expenses described in the contract, as well as the definitions of all
accounting terms used to define EBITDA and interest expenses (e.g., net income, leases, etc.).
Appendix A provides examples of financial covenant specifications.
Table 1 provides details on loan characteristics by year and covenant structure for the 440
securitized and 703 non-securitized loans in our sample. Table 1, Panel A reports the number of
loans (securitized and covenant-lite loans) and financial covenants by year, as well as the
average number of financial covenants by loan year. Consistent with the growth in securitized
loan issuance, the number of securitized loans and covenant-lite loans in our sample increases
during the period 2000–2007 and sharply drops afterwards. Moreover, the average number of
financial loan covenants steadily drops in the period 2000–2007 and increases in 2008–2009,
consistent with lenders’ lower monitoring incentives during the credit boom. Table 1, Panel B
reports the number of financial covenants by covenant type for the 3,303 covenants (1,355
securitized and 1,948 non-securitized) in our sample. Consistent with previous studies (e.g.,
Drucker and Puri, 2008), securitized loans include more financial covenants, and especially more
interest coverage, capital expenditures and leverage covenants.
15
4. Research design and variable definition
4.1. Covenant similarity score
Developing a proxy for the similarity across financial covenants in different loans is
challenging, since lenders are likely to adjust accounting data in covenant definitions (Leftwich,
1983). For example, the way “minimum EBITDA” is defined in one contract may be completely
different from the EBITDA definition in another loan contract. Based on the underlying
assumption that standardized covenants will share more common words with other covenants in
the same covenant category, we proxy for accounting-based covenant standardization by
assessing the degree of overlap in the vector of unique words used to define covenants.
To do so, we first remove all stopwords (e.g., “and”, “a”, “the”, “of”) and pare the
remaining words down to their stems.11 Next, we calculate the pairwise cosine textual similarity
for all reduced-form financial covenant definitions based on a vector space model commonly
used in computational linguistics (e.g., Salton, Wong, and Yang, 1975), which has been recently
introduced in the accounting and finance literatures (e.g., Brown and Tucker, 2011; Hoberg,
Phillips and Prabhala, 2012; Bozanic and Thevenot, 2014). To perform the calculation, a
comparison is drawn between two N x 1 vectors, one vector representing the N unique words in a
given financial covenant definition and another vector for the same covenant type from a loan
issued by a different borrower in the prior year.12 The angle between these two vectors for each
pair of same-type covenants (e.g., minimum EBITDA compared to minimum EBITDA) is the
cosine textual similarity score.13
To compute a loan specific covenant standardization measure, we average the cosine
11 For example, “trusted” and “trusting” become “trust” for calculation purposes.12 The twelve covenant types are based on LPC Dealscan classifications. See Section 3 above.13 Appendix B provides additional detail on the computation of the measure.
16
similarities of the covenants in a loan with the same-type covenants in all other borrowers’ loans
that were issued in the prior year (Covenant Similarity Score). Thus, our proxy for covenant
similarity is a continuous variable with values ranging from zero (if two covenants share no
common word) to one (if the definitions of two same-type covenants are identical).14 By
definition, covenants classified in the category “others” will have a covenant similarity score of
zero. For covenant-lite loans, we code the covenant similarity score as one (i.e., the maximum
value). This is consistent with Ayotte and Bolton’s (2009) argument that covenant-lite loans are
perfectly comparable in terms of their covenant structure.15
Figure 1 shows the trend in covenant standardization over time. Consistent with the
growth in the securitized loan market, covenant similarity increases in the period 2000–2007.
This trend reverses in the period 2008–2009 when the securitization market froze. Figure 2
compares covenant standardization over time for institutional loans and securitized loans. While
covenant similarity for both institutional and securitized loans increases over time, the covenant
similarity score for securitized loans is consistently higher and reverses in the crisis years
tracking the trends in the securitization market. This pattern provides some preliminary evidence
with respect to the impact of securitization via CLOs on covenant standardization.
In addition, we compute an alternative covenant standardization measure which does not
rely on textual analysis. We compute the average of the ratio of similar covenants across the
loans issued in the prior year (Percentage of Same Covenants). This ratio is computed for each
loan pair as the number of common covenants between the current loan and the other loan
14 It is worth mentioning that the covenant similarity score reflects textual rather than semantic similarity. Forexample, if a net worth covenant is defined as assets minus liabilities in a loan contract and the definition of thesame-type covenant in another contract is book value of equity, these two covenants will have very low cosinesimilarity.15 In their model, lenders’ intention to completely standardize covenants in securitized loans leads to the exclusionof covenants from loan contracts. In robustness tests, we exclude covenant-lite loans and covenants classified as“others” from our tests and our results hold.
17
previously issued divided by the total number of financial covenants specified in the loan
contract.
4.2. Research design
4.2.1. Securitization and covenant standardization
Our first test explores the relation between corporate loan securitization and financial
covenant standardization at the loan level. We test our hypothesis using an OLS model, where
the dependent variable is the Covenant Similarity Score.
Asquith, P., Beatty, A., Weber, J., 2005. Performance pricing in debt contracts. The Journal ofAccounting and Economics 40 (1): 101–128.
Ayotte, K., Bolton, P., 2009. Covenant Lite Lending, Liquidity and Standardization of FinancialContracts. Working Paper.
Ball, R., Bushman, R., Vasvari F.P., 2008. The debt-contracting value of accounting informationand loan syndication structure. Journal of Accounting Research 46 (2): 247-288.
Beatty, A., Weber, J., 2003. The Effects of Debt Contracting on Voluntary Accounting MethodChanges. Accounting Review 78 (1): 119–142.
Beatty, A., Weber, J., Yu, J., 2009. Conservatism and debt. Journal of Accounting & Economics45 (2-3): 154–174.
Berlin, M., Mester, L. J., 1992. Debt covenants and renegotiation. Journal of FinancialIntermediation 2 (2): 95–133.
Bozanic, Z., Cheng, L., Zach, T., 2013. Private Debt Contract Complexity: What the NumbersDon’t Tell You. Working Paper.
Bozanic, Z., Thevenot, M., 2014. Qualitative Disclosure and Changes in Sell-Side FinancialAnalysts’ Information Environment. Contemporary Accounting Research, Forthcoming.
Brown, S. V., Tucker, J.W., 2011. Large-Sample Evidence on Firms’ Year-over-Year MD&AModifications. Journal of Accounting Research 49 (2): 309–346.
Choi, A. H., Triantis, G. G., 2014. Market Conditions and Contract Design: Variations in DebtContracting. New York University Law Review, Forthcoming.
Christensen, H. B., Nikolaev, V.V., 2010. Accounting-based covenants and credit market access.Working Paper.
De Franco, G., Kothari, S., Verdi, R. S., 2011. The Benefits of Financial StatementComparability. Journal of Accounting Research, 49: 895–931.
De Franco, G., Vasvari, F. P., Vyas, D., Wittenberg-Moerman, R., 2013. Sticky Covenants.Working Paper.
Dichev, I. D., Skinner, D. J., 2002. Large–Sample Evidence on the Debt Covenant Hypothesis.Journal of Accounting Research 40 (4): 1091–1123.
32
Drucker, S., Puri, M., 2008. On Loan Sales, Loan Contracting, and Lending Relationships.Review of Financial Studies 22 (7): 2835–2872.
Hoberg, G., Phillips, G., Prabhala, N., 2012. Product Market Threats, Payouts, and FinancialFlexibility. Forthcoming, Journal of Finance.
Ivashina, V., Sun, Z., 2011. Institutional Demand Pressure and the Cost of Corporate Loans.Journal of Financial Economics 99 (3): 500–522.
Jensen, M. C., Meckling, W. H., 1976. Theory of the firm: Managerial behavior, agency costsand ownership structure. Journal of Financial Economics 3 (4): 305–360.
Leftwich, R., 1983. Accounting information in private markets: Evidence from private lendingagreements. Accounting Review 58 (1):23–42.
Li, N., 2012. Performance Measures in Earnings-Based Financial Covenants in Debt Contracts.Working Paper, University of Texas at Dallas.
Li, N., Vasvari, F. P., and Wittenberg-Moerman, R., 2014. Dynamic Threshold Values inEarnings-Based Covenants. Working Paper.
Lou, Y., Loumioti, M., Vasvari, F. P., 2014. CLOs’ Trading Activity and Performance. WorkingPaper.
Morgan, D., 2002. Rating banks: risk and uncertainty in an opaque industry. The AmericanEconomic Review 92 (4): 874–888
Nadauld, T. D., Weisbach, M. S., 2011. Did securitization affect the cost of corporate debt?Journal of Financial Economics 105 (2): 332–352.
Nini, G., Smith, D. C., & Sufi, A., 2009. Creditor control rights and firm investment policy.Journal of Financial Economics, 92(3): 400–420.
Rajan, R., Winton, A., 1995. Covenants and Collateral as Incentives to Monitor. Journal ofFinance 50 (4):1113–1146.
Roberts, M., 2013. The Role of Dynamic Renegotiation and Asymmetric Information inFinancial Contracting. Working Paper, University of Pennsylvania.
Roberts, M. R., Sufi, A., 2009. Renegotiation of financial contracts: Evidence from private creditagreements. Journal of Financial Economics 93 (2):159–184.
Salton, G., Wong, A., Yang, C. S., 1975. A Vector Space Model for Automatic Indexing.Communications of the ACM 18: 613–620.
Simpson, D. B., 1973. The Drafting of Loan Agreements: A Borrower's Viewpoint. The BusinessLawyer 28 (4): 1161–1196.
33
Skinner, D.J., 2011. Discussion of “Accounting standards and debt covenants: Has the ‘BalanceSheet Approach’ led to a decline in the use of balance sheet covenants?”. Working Paper.
Smith, H. E., 2006. Modularity in Contracts: Boilerplate and Information Flow. Michigan LawReview 104: 1175–1221.
Smith, C. W., Warner, J. B., 1979. On financial contracting: An analysis of bond covenants.Journal of Financial Economics 7 (2): 117–161.
Stiglitz, J. E., Weiss, A., 1981. Credit Rationing in Markets with Imperfect Information.American Economic Review 71 (3): 393–410.
Wittenberg-Moerman, R., 2008. The role of information asymmetry and financial reportingquality in debt trading: Evidence from the secondary loan market. Journal of AccountingResearch 46 (2-3): 240–260.
34
Appendix A
Examples of Financial Covenant Definitions
Example 1: Consolidated Interest Coverage Ratio
Consolidated Interest Coverage Ratio is defined as Consolidated EBITDA to Consolidated Interest Charges
“Consolidated EBITDA” means, for any period, for the Borrower and its Restricted Subsidiaries on a consolidatedbasis, an amount equal to Consolidated Net Income for such period plus (a) the following to the extent deducted incalculating such Consolidated Net Income: (i) Consolidated Interest Charges for such period, (ii) the provision forFederal, state, local and foreign income taxes payable by the Borrower and its Restricted Subsidiaries for suchperiod, (iii) depreciation and amortization expense, and (iv) other expenses of the Borrower and its RestrictedSubsidiaries reducing such Consolidated Net Income which do not represent a cash item in such period or anyfuture period and minus (b) the following to the extent included in calculating such Consolidated Net Income: (i)Federal, state, local and foreign income tax credits of the Borrower and its Restricted Subsidiaries for such period,and (ii) all non-cash items increasing Consolidated Net Income for such period; provided that for the purposes ofSection 7.20, if the Borrower or any Restricted Subsidiary shall acquire or dispose of any material property or aSubsidiary shall be redesignated as either an Unrestricted Subsidiary or a Restricted Subsidiary, in any case, duringthe period of four fiscal quarters ending on the last day of the fiscal quarter immediately preceding the date ofdetermination for which financial statements are available and up to and including the date of the consummation ofsuch acquisition, disposition or redesignation, then Consolidated EBITDA shall be calculated, in a mannersatisfactory to the Administrative Agent in its reasonable discretion, after giving pro forma effect to such acquisition(including the revenues of the properties acquired), merger, disposition or redesignation, as if such acquisition,merger, disposition or redesignation had occurred on the first day of such period.
“Consolidated Interest Charges” means, for any period, for the Borrower and its Restricted Subsidiaries on aconsolidated basis, the sum of (a) all interest, premium payments, debt discount, fees, charges and related expensesof the Borrower and its Restricted Subsidiaries in connection with borrowed money (including capitalized interest)or in connection with the deferred purchase price of assets, in each case to the extent treated as interest inaccordance with GAAP, excluding one-time charges in respect of loan origination or similar fees and non-cashamortized amounts with respect thereto, and (b) the portion of rent expense of the Borrower and its RestrictedSubsidiaries with respect to such period under capital leases that is treated as interest in accordance with GAAP.
“Consolidated Net Income” means, for any period, for the Borrower and its Restricted Subsidiaries’ gross revenuesfor such period, including any cash dividends or distributions actually received from any other Person during suchperiod, minus the Borrower’s and its Restricted Subsidiaries’ expenses and other proper charges against income(including taxes on income to the extent imposed), determined on a consolidated basis in accordance with GAAPconsistently applied after eliminating earnings or losses attributable to outstanding minority interests and excludingthe net earnings of any Person other than a Restricted Subsidiary in which the Borrower or any of its Subsidiarieshas an ownership interest. Consolidated Net Income shall not include (i) any gain or loss from the Disposition ofassets, (ii) any extraordinary gains or losses, or (iii) any non-cash gains or losses resulting from mark to marketactivity as a result of the implementation of Statement of Financial Accounting Standards 133, “Accounting forDerivative Instruments and Hedging Activities” (“SFAS 133”).
Example 2: Total Leverage Ratio
The ratio of Indebtedness to EBITDA
“Indebtedness” means of any Person (without duplication): (a) indebtedness created, issued or incurred by suchPerson for borrowed money (whether by loan or the issuance and sale of debt securities or the sale of property toanother Person subject to an understanding or agreement, contingent or otherwise, to repurchase such property fromsuch Person); (b) obligations of such Person to pay the deferred purchase or acquisition price of property or services,other than trade accounts payable (other than for borrowed money) arising, and accrued expenses incurred, in theordinary course of business so long as such trade accounts payable are payable within 90 days of the date the
35
respective goods are delivered or the respective services are rendered; (c) Indebtedness of others secured by a Lienon the property of such Person, whether or not the respective Indebtedness so secured has been assumed by suchPerson; (d) obligations of such Person in respect of letters of credit or similar instruments issued or accepted bybanks and other financial institutions for account of such Person; (e) Capital Lease Obligations of such Person; (f)Indebtedness of others guaranteed by such Person; (g) if the aggregate consideration payable by such Person toextend and exercise any option acquired in connection with any Acquisition (an “Extension and Exercise Price”)exceeds 20% of the aggregate consideration payable in connection with such Acquisition, such Extension andExercise Price; (h) any put obligations, but only to the extent that such Put Obligations (other than the PutObligations in existence on the Fourth Restatement Effective Date relating to WNAB-TV (Nashville, Tennessee)),whether arising under the same or different agreements, exceeding $25,000,000 in the aggregate shall not have beenapproved by the Administrative Agent (such approval not to be unreasonably withheld) prior to the incurrencethereof; and (i) obligations of such Person in respect of surety and appeals bonds or performance bonds or othersimilar obligations; provided that the term “Indebtedness” shall not include (i) obligations of such Person, (ii)obligations of such Person under any Program Services Agreement, Outsourcing Agreement or other similaragreement, (iii) any liability shown on such Person’s balance sheet in respect of the fair value of Interest RateProtection Agreements, (iv) any put obligations, and (v) any liability shown on the balance sheet of such Personsolely as a result of the application of FIN 46 and for which such Person is not primarily or contingently liable forpayment.
“Capital Lease Obligations” of any Person means the obligations of such Person to pay rent or other amountsunder any lease of (or other arrangement conveying the right to use) real or personal property, or a combinationthereof, which obligations are required to be classified and accounted for as capital leases on a balance sheet of suchPerson under GAAP, and the amount of such obligations shall be the capitalized amount thereof determined inaccordance with GAAP.
“EBITDA” means, for any period, the sum, for the Borrower and its Subsidiaries (determined on a consolidatedbasis without duplication in accordance with GAAP), of the following for such period (subject to Section 1.05(d)):(a) net income for such period; plus (b) the sum of, to the extent deducted in determining net income for such period,(i) provision for taxes, (ii) depreciation and amortization (including film amortization), (iii) Interest Expense, (iv)Permitted Termination Payments (or to the extent the same shall be included in determining corporate expensespursuant to clause (c)(ii) below for such period), (v) extraordinary losses (including non-cash losses on sales ofproperty outside the ordinary course of business of the Borrower and its Subsidiaries), (vi) all other non-cashcharges (including non-cash losses on derivative transactions and non-cash interest expenses), (vii) all transactioncosts paid or incurred by the Borrower in connection with the Fourth Restatement Effective Date Transactions andthe Tender Offer Transactions, and (viii) all amounts paid in cash by the Borrower and its Subsidiaries toCunningham and its Subsidiaries pursuant to the transactions contemplated by the Cunningham MOU that are inrespect of, or credited toward, the purchase price of any Stations to be acquired by the Borrower or any of itsSubsidiaries from Cunningham or are in respect of local marketing agreement fees, but not exceeding $11,000,000in the aggregate for any twelve month period; minus (c) the sum of, to the extent included in net income for suchperiod, (i) non-cash revenues, (ii) corporate expense (but only to the extent already not deducted in determining netincome for such period), (iii) interest and other income, (iv) extraordinary gains (including non-cash gains on salesof assets outside the ordinary course of business), (v) benefit from taxes, (vi) non-cash gains on derivativetransactions, and (vii) cash payments made during such period in respect of items under clause (b)(vi) abovesubsequent to the fiscal quarter in which the relevant non-cash charge was reflected as a charge in the statement ofnet income; minus (d) Film Cash Payments made or scheduled to be made during such period.
“Interest Expense” means, for any period, the sum, for the Borrower and its Subsidiaries (determined on aconsolidated basis without duplication in accordance with GAAP), of (a) all cash interest expense in respect ofIndebtedness during such period, (b) the net amounts payable (or minus the net amounts receivable) under InterestRate Protection Agreements accrued during such period (whether or not actually paid or received during suchperiod) and (c) restricted payments made during such period pursuant to Section 7.08(a) in respect of interestpayments on the Holding Company convertible debentures (including any such interest payments thereon madepursuant to Section 7.08 of the Existing Credit Agreement prior to the Fourth Restatement Effective Date during anyfiscal quarter that is included in such period). Any reference herein to calculating Interest Expense for any period ona “pro forma” basis means that, for purposes of the clause (a) above, (i) the Indebtedness on the basis of which
36
Interest Expense is so calculated shall mean Indebtedness outstanding as of the relevant date of calculation aftergiving effect to any repayments and any incurrence of Indebtedness on such date and (ii) such calculation shall bemade applying the respective rates of interest in effect for such Indebtedness on such date.
“Film Cash Payments” means, for any period, the sum (determined on a consolidated basis and withoutduplication) of all payments by the Borrower and its Subsidiaries made or scheduled to be made during such periodin respect of film obligations; provided that amounts applied to the prepayment of film obligations owing under anycontract evidencing a film obligation under which the amount owed by the Borrower or any of its Subsidiariesexceeds the remaining value of such contract to the Borrower or such Subsidiary, as reasonably determined by theBorrower, shall not be deemed to be Film Cash Payments.
Example 3: Limitation on Capital Expenditures
“Capital Expenditures” shall mean with respect to any Person for any period, the sum of (i) the aggregate of allexpenditures by such Person and its Subsidiaries during such period that in accordance with GAAP are or should beincluded in “property, plant and equipment” or in a similar fixed asset account on its balance sheet, whether suchexpenditures are paid in cash or financed, and (ii) to the extent not covered by clause (i) above, the aggregate of allexpenditures by such Person and its Subsidiaries during such period to acquire by purchase or otherwise the businessor fixed assets of, or the capital stock of, any other Person; provided that there shall be excluded from CapitalExpenditures the purchase price paid in any Permitted Acquisition; provided, further, that any rolling stock which isinitially accounted for as a Capital Expenditure at the time of acquisition thereof but which is transferred to a thirdparty and becomes subject to an operating lease within 60 days after the date of acquisition thereof which leasewould not be required to be treated as an addition to “property, plant and equipment” or in a similar fixed assetaccount on a consolidated balance sheet of Parent and its Subsidiaries prepared in accordance with GAAP, shall beexcluded from Capital Expenditures.
Example 4: Net Worth
“Net Worth” means, as of any date of determination, the total consolidated stockholders’ equity (determinedwithout duplication) of the Borrower and its Subsidiaries at such date.
37
Appendix B
Cosine Textual Similarity
We measure covenant standardization by assessing the degree of overlap in the vector of uniquewords used to define covenants. To do so, we first remove from the covenant definition allstopwords (e.g., “and”, “a”, “the”, “of”) and pare the remaining words down to their stems. Forexample, “trusted” and “trusting” become “trust” for calculation purposes.
Next, we estimate the extent to which two covenant definitions are similar by calculating thepairwise cosine textual similarity for all pairs of reduced-form financial covenant definitionsbased on a vector space model used in plagiarism software and search engine algorithms (seeSalton, Wong, and Yang, 1975) as follows:
We count how many times each word is used in each covenant definition. This processcreates two vectors with the number of times each word is mentioned in the two covenants.To illustrate, assume we have two covenant texts, T1 and T2, with three words (W1, W2, W3)each. W1 occurs in T1 2 times, W1 occurs in T2 3 times, and so forth:
T1 = (2W1, 3W2, 5W3)
T2 = (3W1, 7W2, W3)
The cosine similarity of the two vectors above is a mathematical measure of how similar thetwo vectors are on a scale of [0, 1] with 1 being the outcome if the vectors are either identicalor their values differ by a constant factor. For cosine similarities resulting in a value of 0, thecovenant definitions do not share any attributes (or words) because the angle between theword vectors is 90 degrees. The cosine similarity is computed as:
cos Ɵ = T1·T2 / ||T1||*||T2|| = 0.6758
where the vector product is T1·T2 = 2*3 + 3*7+ 5*1 and the normalized vectors are computedas ||T1|| = sqrt(22 + 32 + 52) and ||T2|| = sqrt(32 + 72 + 12).
To obtain a loan specific covenant standardization measure, we average the cosine similarities ofthe covenants in a loan with the same-type covenants in all other borrowers’ loans that wereissued in the prior year. Thus, the Covenant Similarity Score is a continuous variable with valuesranging from zero (if two covenants share no common word) to one (if the definitions of twosame-type covenants are identical).
38
Appendix C
Variable definitions
Variable Definition
Cash Flow VolatilityThe standard deviation of borrower’s operating cash flows over the last five years,deflated by total assets.
Covenant Similarity ScoreThe average similarity score at the loan level of covenant i in loan k with covenant jin loan m only if i and j are of the same covenant type. See Appendix B for furtherdetail.
Lending RelationshipThe ratio of total loan size a borrower took from the lead arranger in the past to totalsize of loans the borrower took in the past.
Leverage Total liabilities to total assets.
LIBOR-spread The natural logarithm of all-in-drawn LIBOR-spread of the Term B tranche.
Liquidity Current assets to current liabilities.Loan Amount The natural logarithm of the loan amount.
Loan DefaultBinary variable that equals one if the borrower defaulted on a securitized loan, andzero if the borrower did not default on a securitized loan.
Loan DistributionAnnual change in the number of CLOs holding at least one tranche of a securitizedloan, divided by the average number of CLOs holding a tranche of a securitized loanin the same year.
Loan TradesLoan sales and purchases less purchases where both transacting parties are CLOs fora securitized loan in a year, divided by the average trading activity of a securitizedloan in the same period.
Loan Rating DifferenceThe average difference between Moody's and Standard & Poor’s loan rating overour sample period.
Loan Maturity The natural logarithm of loan maturity (in months).
No covenantsBinary variable that equals one if a loan contract does not include financialcovenants, and zero otherwise.
Number of Covenants The number of financial loan covenants, including net worth covenants.
Pct. Same CovenantsAverage number of the same financial covenants with other loans originated in thelast year to the number of financial loan covenants.
Revolving TrancheBinary variable that equals one if the loan includes a revolving tranche, and zerootherwise.
ROA Operating income to total assets.
Same Loan RatingThe number of quarters a securitized loan's S&P and Moody's rating are the same,divided by the number of quarters the loan is held by CLOs. S&P (Moody's) loanrating is a scale variable, where 1=AAA,…, 25=D.
Securitized LoanBinary variable that equals one if the loan includes at least one securitized trancheand zero otherwise.
Size The natural logarithm of total assets.
Syndicates The natural logarithm of the number of co-syndicates in the loan.
Time-on-MarketThe number of days the loan remains outstanding in the primary loan market (Closedate- Launch date).
39
Figure 1: Covenant similarity score
Figure 1 reports the average covenant similarity score for our sample of 703 institutional non-securitized and 440 securitized loans in2000–2009 (primary axis), including covenant-lite loans (covenant similarity=1). The pattern looks similar for the sub-sample of 608institutional non-securitized and 385 securitized loans in 2000–2009, excluding covenant-lite loans (primary axis). Using our sampleof 1,143 corporate loans, the percentage of securitized loans is estimated as total number of securitized loans issued in a year dividedby annual total loan issuance (secondary axis). The percentage of same covenants is the ratio of the same covenants a loan shares to allother loans in the sample of 1,143 loans, divided by the number of covenants in the loan.
Figure 2: Covenant similarity score, securitized vs. institutional loans
Figure 2 reports the average covenant similarity score for our sample of 703 institutional non-securitized and 440 securitized loans in2000–2009 (primary axis), including covenant-lite loans (covenant similarity=1).
Variables are described in Appendix C. Standard deviations reported in parentheses. All values of the continuous variables arewinsorized at 1% and 99% level. ***Significant at 1%, ** 5% and * 10% level.
45
Table 4
Covenant Similarity Score – Validation test
Covenant Similarity Score
Variable Coeff. t-stat.
D(Number of Covenants) -0.003 *** -4.22
D(LIBOR-spread) -0.002 *** -5.61
D(Loan Amount) -0.001 -1.42
D(Maturity) -0.023 *** -29.95
Same Lender 0.003 *** 2.15
D(Lending Relationship) 0.001 0.86
Same Loan Purpose 0.001 1.02
D(Syndicates) -0.009 *** -14.04
Same Industry 0.006 *** 2.90
D(Liquidity) -0.002 ** -1.96
D(ROA) -0.132 *** -11.68
D(Leverage) -0.011 *** -4.20
D(Cash Flow Volatility) -0.092 *** -4.22
D(Size) 0.000 -0.62
Constant 0.351 *** 12.81
N= 79,134R2= 0.28
The dependent variable is the Covenant Similarity Score defined as the average similarity score at the loan level of covenant i in loan kwith covenant j in loan m only if i and j are of the same covenant type. Same Lender equals one if the loans are issued by the same leadlender, and zero otherwise. Same Loan Purpose equals one if the loans have the same purpose, and zero otherwise. Same Industry equalsone if borrowers are from the same industry (12-industry FF), and zero otherwise. All other independent variables are the absolute valuesof the differences in loan and borrower characteristics where the financial covenants refer to. Variables are defined in Appendix C.Covenant type, lead lender, industry (12 industry portfolios), year of loan origination and loan purpose fixed effects included. The valuesof the continuous variables are winsorized at 1% and 99%. Standard errors are corrected for heteroskedasticity; cluster is at the loan level.*** Significant at 1%, ** 5% and * 10% level, two-tailed tests.
46
Table 5
Securitization and Covenant Standardization
Panel A: Securitization and Covenant Standardization
All loansPct. of Same Covenants Covenant Similarity Score
I II IIIVariable Coeff. t-stat. Coeff. t-stat. Coeff. t-stat.
The table reports the tests for the relation between loan securitization and financial covenant standardization. Panel A reports the baseline OLS regression results. Panel B reportscross-sectional tests. The dependent variable in the first column is the Pct of Same Covenants, defined as the ratio of same covenants a loan has compared to all other loans in thesample originated in the last year to the total number of covenants in the loan. The dependent variable in all other specifications is the Covenant Similarity Score, defined as theaverage textual cosine similarity of the financial covenants in a loan compared to covenants in loans to different borrowers originated in the last year. In specification (IV), thesample includes non-securitized loans and securitized loans with more than 80 percent of their size being securitized. In the next two specifications, the sample includes non-securitized loans and loans securitized upon origination (V) or sold subsequently to CLOs (VI). In specification (VIII), we eliminate our sample to companies that issued bothsecuritized and non-securitized loans in our sample period. Panel C presents the diagnostic results for the propensity score matching tests. The treatment is whether a loan issecuritized, and the outcome variable is the Covenant Similarity Score, defined as the average textual cosine similarity of the financial covenants in a loan compared to covenantsin loans to different borrowers originated in the last year. The one-to-one matching of treated loans is done in random order and without replacement. Matching loans are within adistance (“caliper”) of 0.01 of the propensity score of the loans in the treatment group. The average treatment effect, t-statistic and balance statistics for the matching procedure arereported. All variables are defined in Appendix C. Lead lender, industry (12 industry portfolios), year of loan origination and loan purpose fixed effects included. The values of thecontinuous variables are winsorized at 1% and 99%. Standard errors are corrected for heteroskedasticity; cluster is at the borrower level (except specification (IV)). *** Significantat 1%, ** 5% and * 10% level, two-tailed tests.
50
Table 6
Covenant Standardization and Loan liquidity
Panel A: Covenant standardization and liquidity in the primary market
The table reports the tests for the relation between financial covenant standardization and loan liquidity in the primary loanmarket. The dependent variable is the number of days a loan remains open in the primary market (Time-on-Market). The sampleincludes 343 loans issued in 2000-2007. All variables are defined in Appendix C. Industry (12 industry portfolios), year of loanorigination and loan purpose fixed effects included. The values of the continuous variables are winsorized at 1% and 99%.Standard errors are corrected for heteroskedasticity; cluster is at the borrower level. ***Significant at 1%, ** 5% and * 10%level, two-tailed tests.
51
Panel B: Covenant standardization and liquidity in the secondary loan market
The table reports the tests for the relation between financial covenant standardization and loan liquidity in the secondary loan market. All variables are defined in Appendix C.Industry (12 industry portfolios), year of loan origination and loan purpose fixed effects included. The values of the continuous variables are winsorized at 1% and 99%. Standarderrors are corrected for heteroskedasticity; cluster is at the borrower level. ***Significant at 1%, ** 5% and * 10% level, two-tailed tests.
The table reports the tests for the relation between financial covenant standardization, LIBOR-spread and the probability of aborrower’s defaulting on a securitized loan. LIBOR-spread is the all-in-drawn LIBOR-spread of the term B loan tranche. Defaultequals one if a borrower defaulted on a securitized loan in the period 2008-2013, and zero otherwise. In the second column, weuse a probit model, and marginal effects are reported. All variables are defined in Appendix C. Lead lender (only in specificationI), industry (12 industry portfolios), year of loan origination and loan purpose fixed effects included. The values of the continuousvariables are winsorized at 1% and 99%. Standard errors are corrected for heteroskedasticity; cluster is at the borrower level.***Significant at 1%, ** 5% and * 10% level, two-tailed tests.
The table reports the tests for the relation between financial covenant standardization and S&P and Moody’s loan ratingagreement. The dependent variable in specification I is the number of quarters that S&P and Moody’s issued the same loan ratingfor a securitized loan divided by total number of quarters the loan was held by CLOs (Same Rating). The dependent variable inspecification II is the absolute value of the average difference in quarterly loan ratings issued by S&P and Moody’s in 2008-2011(Loan Rating Difference). All variables are defined in Appendix C. Industry (12 industry portfolios), year of loan origination andloan purpose fixed effects included. The values of the continuous variables are winsorized at 1% and 99%. Standard errors arecorrected for heteroskedasticity; cluster is at the borrower level. ***Significant at 1%, ** 5% and * 10% level, two-tailed tests.