Journal of Financial Economics: Taylor D. Nadauld, Michael S. Weisbach

Journal of Financial Economics 105 (2012) 332352
Contents lists available at SciVerse ScienceDirect
Journal of Financial Economics

journal homepage: www.elsevier.com/locate/jfec
Did securitization affect the cost of corporate debt?$

Taylor D. Nadauld a,n, Michael S. Weisbach b
a
b
Department of Finance, Brigham Young University, Provo, Utah, USA

Ohio State University and NBER, USA
a r t i c l e in f o
abstract
Article history:
Received 23 February 2011
Received in revised form
6 July 2011
Accepted 4 August 2011
Available online 15 March 2012
This paper investigates whether the securitization of corporate bank loan facilities had
an impact on the price of corporate debt. Our results suggest that loan facilities that are
subsequently securitized are associated with a 17 basis point lower spread than that of
facilities that are not subsequently securitized. We consider facility characteristics that
are associated with the likelihood of securitization and estimate the extent to which
these characteristics are related to spreads. We document that Term Loan B facilities,
facilities of B-rated rms, and facilities originated by banks that originate CLOs are
securitized more frequently than other facilities. Spreads on facilities estimated to be
more likely to be subsequently securitized have lower spreads than otherwise similar
facilities. The results are consistent with the view that securitization caused a reduction
in the cost of capital.
& 2012 Elsevier B.V. All rights reserved.
JEL classication:
G21
G31
Keywords:
Securitization
Collateralized loan obligations (CLO)
Corporate bank loans
1. Introduction
While much attention has focused on the role of securitization in the mortgage market, relatively little has been
paid to the securitization of corporate bank loans in the form
of collateralized loan obligations (CLOs).1 This lack of attention is surprising given that both the volume of
non-investment-grade bank loans and the number of newly
originated CLOs spiked dramatically between the years 2002
and 2007, from over $125 billion of loans and 43 CLOs in
$
We would like to thank Viral Acharya, Heitor Almeida, Murillo
Campello, Jennifer Dlugosz, Isil Erel, Andrew Karolyi, Eric Oberg, Rene
Stulz, and participants in seminars at Brigham Young University,
Erasmus University, HEC Paris, Ohio State University, and the University
of Amsterdam for helpful suggestions. We also thank our unusually
diligent referee, Victoria Ivashina, for helpful comments and
suggestions.
n
Corresponding author.
E-mail address: [email protected] (T.D. Nadauld).
1
See, for example, Mian and Su (2009), Demyanyk and Van
Hemert (2011), DellAriccia, Igan, and Laeven (2008), Keys, Mukherjee,
Seru, and Vig (2010), Loutskina and Strahan (2009), and Nadauld and
Sherlund (forthcoming).
0304-405X/$ - see front matter & 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jneco.2012.03.002
2002 to over $540 billion of loans and 213 CLOs in 2007.2

One potential explanation for this pattern was that the
popularity of CLOs created a demand for collateral that
could be used to construct them (see, for example,
Gennaioli, Shleifer, and Vishny, in press). Consequently,
banks active in the securitization business had incentives
to adjust their lending behavior to increase the quantity of
loans that could be used for collateral, affecting the cost of
capital for rms in the primary lending market.
In this paper we estimate the extent to which the
practice of securitizing bank debt inuenced the cost of
such debt for borrowers in the primary corporate debt
market. To do so, we rely on a sample of over 4,000 loan
facilities from the Dealscan database. We match these
loan facilities to a database of CLOs provided by Moodys.
The Moodys sample provides data on the characteristics
and identity of all the securities serving as collateral in
2
The estimates on the dollar volume of non-investment-grade loans
originated in 2002 and 2007 are likely understated in both years. The
reported gures are calculated using our sample of U.S.-originated
syndicated bank loans made to non-nancial borrowers with available
sales data that is described in Section 3.
T.D. Nadauld, M.S. Weisbach / Journal of Financial Economics 105 (2012) 332352
any Moodys-rated CLO as of the rst quarter 2009, which

can be used to predict the attributes of securitizable
loan facilities. We rst address the question of whether a
loan facility that was ultimately securitized had a lower
spread than an otherwise similar loan facility that was not
securitized. Our estimates indicate that the answer to this
question is a clear yes; they suggest that of the most
common type of securitized facility, a B-rated, Term Loan
B, a facility that is ultimately securitized has about a 17
basis point lower yield than an otherwise identical loan
that was not securitized.
There are a number of possible interpretations to this
result. It is possible that CLOs, for some unidentied
reason, were more likely to add relatively low-yielding
loans in their portfolio, or that an omitted variable is
driving both spreads and securitization demand in our
specication. Alternatively, there could have been a
reduction in the demand for capital from the rms whose
loan facilities tend to be securitized. These interpretations
imply that CLOs could have relatively low-yielding loan
facilities in their portfolios, but that this ownership
distribution of loan facilities is not associated with any
additional costs. More consistent with the securitizationdriven explanations of the Financial Crisis is a story in
which banks have a high demand for securitizable loan
facilities when they put together CLOs and consequently
lower the yields to attract more borrowers.3 This latter
explanation implies that demand for loans from CLOs has
a real effect, as it lowers the cost of capital for the rms
obtaining the loans.
To distinguish between these explanations, we estimate the effect on a loans spread of factors that are
correlated with the demand for bank loan facilities from
participants in the securitization market and thus potentially impact credit supply in the primary market, but are
unrelated with the demand for credit from corporate
borrowers in the origination market. First, we consider
features related to the structure of corporate loans.
Corporate loans are typically made up of a number of
tranches (also referred to as facilities), often containing an
amortizing Term Loan A portion, a bullet Term Loan B
portion, and a Revolver that can be drawn down at the
discretion of the borrower. These tranches usually have
the same seniority, but because of the amortization of
Term Loan A and the options implicit in the revolver, they
have different effective maturities and yields. An important difference between the tranches, however, is that
typically only the Term Loan B (bullet) portion is securitized, while the other tranches are typically held by the
issuing bank.4 Consequently, the difference in yields
between the tranches potentially reects differences in
demands for securitizable loans, among other factors.
3
The terms supply and demand can be ambiguous when discussing
a primary and secondary loan market. The phrase demand for securitizable loan facilities refers to demand from the secondary market,
which is manifest as increased supply in the primary loan-origination
market.
4
In Section 2.2 we discuss at least two reasons why Term Loan Bs
are preferred as collateral by arrangers of CLOs.
333
Second, the credit rating of a borrower affects the

likelihood of a loan facility being securitized. In a sample
of B- and Ba-rated loan facilities, holding other factors
constant, the lower credit quality B-rated rms obtained
facilities that were more likely to be securitized. This
pattern is consistent with Benmelech and Dlugosz (2009),
who nd that cash-ow arbitrage CLOs tend to employ
B-rated bank facilities as collateral. Demand from deal
arrangers in the securitization market for B-rated collateral rather than Ba-rated collateral represents an increase
in demand for corporate bank loans that is unrelated to
factors affecting the demand for loans from borrowers in
the primary loan market.
Third, we use the identity of the loan-arranging bank
in our identication strategy. Over the 6-year period
between 2002 and 2007, 79% of all CLOs were underwritten by just ten banks. Leveraged loans arranged by
banks that also underwrite CLOs are more likely to be
securitized because CLO underwriters can more accurately evaluate the value of a particular loan to CLO
structures. That is, CLO underwriting banks are more
likely to arrange a given bank loan for the purpose of
using it as collateral in a CLO.5 Consistent with this logic,
we document that in our sample, loan facilities arranged
by these securitization-active banks were securitized
more frequently than loan facilities issued by other banks.
We exploit these three institutional features of the CLO
market to identify the impact of securitization on the cost
of corporate credit. Since increased demand for securitizable assets occurs more frequently in Term Loan B
facilities originated by securitization-active banks, we
estimate the difference between spreads on these facilities relative to Term Loan B facilities issued by nonsecuritization-active banks. We nd that spreads on Term
Loan B facilities originated by securitization-active banks
are 11 basis points lower than spreads on facilities issued
by other banks, holding other factors constant.
If this decline occurs because of demand for securitizable loans, then we expect this decline in spreads to be
larger in B-rated debt than in Ba-rated debt because
B-rated debt is securitized more frequently. For this
reason, in a sample of Term Loan B facilities only, we
estimate the difference in spreads between B-rated facilities originated by securitization-active banks and compare those estimates to spreads on Ba-rated facilities
originated by securitization-active banks. Consistent
with the securitization demand-driven hypothesis, our
estimates indicate that the difference in the spread on
loans issued by securitization-active banks to B-rated
borrowers was almost 21 basis points lower than the
spread on loans made by non-securitization-active banks
to B-rated borrowers. In contrast, there is virtually no
difference in spreads between Ba-rated loans from securitization-active banks and non-securitization-active banks.
In addition, if the demand to securitize Term Loan B
facilities issued to B-rated borrowers was highest among
securitization-active banks, we would expect to nd
differences in the spreads of Term Loan B facilities with
We discuss more specically why this is the case in Section 2.
334
these securitization-friendly attributes relative to Term

Loan A facilities or revolving facilities. Our estimates
indicate that the incremental effect of being a securitization-friendly Term Loan B facility is 30.6 basis points
relative to Term Loan As and is 20.8 basis points relative
to revolvers. These equations imply that the drop in
spreads for securitization-friendly facilities does not
reect a bank-specic characteristic, such as an increased
risk tolerance, since this characteristic would likely be
reected in the spreads of securitization-friendly Term
Loan A facilities and revolvers as well.
Another way to do the comparison is by estimating the
factors that affect the differences across facilities within a
particular loan. This approach has the advantage of having
common underlying risks within the loan, although the data
are limited to loans that contain multiple types of facilities.
In this case, the differences in spreads between Term Loan B
and either Term Loan A or revolvers decline with characteristics associated with securitization, again consistent with
the view that demand for collateral for CLOs affects the
pricing of the facilities in the primary market.
Overall, the results are consistent with the view that
CLOs demand for collateral affected pricing in the corporate debt market. The effect of the securitization-driven
debt market on the housing market during the 20022007
period is well-known. Shivdasani and Wang (2011) and
Axelson et al. (2011) show that the debt market was an
important driver of the leveraged buyout (LBO) boom
during that period. Our results suggest that securitization
had a pricing impact on the corporate debt market more
broadly.
Perhaps the most closely related work to ours is
Ivashina and Sun (2011), who provide evidence that
demand for loans from institutional investors in the
secondary market, measured by the time a loan remains
in syndication, is negatively related to spreads on these
loans. The authors present evidence that some but not all
of their effect is due to demand from CLOs. Our analysis
extends their work in a number of directions. First, we
focus exclusively on the role of securitization in loan
pricing. Second, we provide direct evidence on the types
of loans most associated with securitization, in particular,
Term Loan B facilities, loans issued by banks that also
issue CLOs, and loans that are made to B-rated borrowers.
Third, we use these characteristics as a way to identify
the causal nature of the relation between pricing and
securitization. To the extent that Ivashina and Suns
(2011) and our identication strategies are different from
one another and our results are nonetheless similar, each
papers analysis complements the others and strengthens
the conclusion that securitization affects the demand for
collateralizable loan facilities, and ultimately the cost of
capital in the primary market.
Other related work includes Kara, Marques-Ibanez,
and Ongena (2011), who identify a relationship between
the securitization-activity of loan-originating banks and
lower spreads in the European loan market. Benmelech,
Dlugosz, and Ivashina (2011) test whether the CLO market
suffers from a lemons problem by investigating whether
loans originated by the bank that acts as the CLO underwriter underperform the rest of the CLO portfolio. Nini
(2008) examines the investment behavior of rms more

likely to benet from securitization-driven increases in
credit supply. Finally, there is a literature discussing
numerous other factors outside of securitization which
inuence the spread charged to corporate borrowers in
the bank loan market (see Guner, 2006; Su, 2007;
Ivashina, 2009).
2. Institutional background
Collateralized loan obligations (CLOs) transform speculative-grade bank loan facilities into highly rated bonds
through the process of pooling and tranching.6 The
incidence of securitizing corporate bank loans exploded
between 2002 and 2007 and has been well-documented
in the literature and popular press. Over 80% of the CLOs
originated between 1996 and 2008 were originated in the
years 20022007, with nearly 50% occurring in 2006 and
the rst half of 2007 alone.7 Also noteworthy is the fact
that between 2002 and 2007, 79% of new CLOs were
originated by just ten banks. In short, the creation of CLOs
was very active over a relatively short period of time and
concentrated in a relatively small number of originating
banks.
Theoretically, securitization can occur for a number of
reasons. First, DeMarzo (2005) suggests that securitization reduces asymmetric information problems about the
quality of individual loans by providing the purchaser
with a diversied pool of loans from different issuers,
presumably having a lower information asymmetry problem than any individual loan would have. Second, Gaur,
Seshadri, and Subrahmanyam (2011) point out that
securitization reduces market incompleteness, providing
investors who value that particular cash-ow distribution
with the ability to purchase it. Finally, Coval, Jurek, and
Stafford (2009), relying on the assumption that investors
purchase bonds based solely on credit ratings, attribute
the proliferation of securitization activity between 2002
2007 to the potential for deal arrangers to deliver the
cheapest possible set of assets that can obtain a high
quality credit rating, thereby engaging in ratings arbitrage. The observation that the bank loans of B-rated
borrowers were securitized more frequently than Barated loans is consistent with a cheapest-to-deliver
view of collateral selection in securitization.
Yet, none of these theories of securitization provides a
rationale for why securitization should affect the spreads
on loan facilities that subsequently are securitized. In
other words, given two loan facilities, why should their
spreads differ depending on whether they are subsequently securitized if the risks associated with the loans
repayment are identical? For this to be the case, the very
process of securitization likely inuences the lenders
decision about the loan at the time of origination.
6
The term tranche is used to describe different bonds arising from
a CLO deal and is also used to describe different loan facilities within a
bank loan. The term tranche in this sentence refers to the tranches
(bonds) that arise from the origination of a CLO.
7
Data on CLO originations are presented in Table 2.
2.1. Lender demand for CLO collateral and loan spreads

The origination of a CLO has two principal participants,
the CLO underwriter (bank) and a collateral manager
(asset manager). The CLO creation process can occur one
of two ways: either the investment bank can contact a
collateral manager and agree on a deal, or a collateral
manager can decide that they want to be in the business
of managing a portfolio of bank loans and set up a CLO as a
vehicle to do so. Both the asset manager and CLO underwriter have incentives to originate deals. CLO underwriters
have incentives to originate deals because they are compensated in the form of deal fees and through the sale of the
bonds arising from the CLOs, while asset managers are
compensated based on assets under management.
CLOs require loan facilities as collateral. If a bank-loan
originator has a high demand for securities that can
subsequently become part of a CLO, it has an incentive
to offer the borrower a lower interest rate than it otherwise would given current market rates and the risk on
that particular investment.8 Consequently, securitized
facilities can end up having interest rates that do not
fully reect the risks associated with the cash-ow distributions generated by their collateral. In this way, CLO
demand can impact loan facility spreads at origination. It
is possible that demand from a single loan arranger to
place a facility in a CLO could not be sufcient to induce
all syndicate members to accept a lower spread. Most
likely, syndicate members will be more prone to accepting a lower spread if more than one of them is active in
the securitization market.9
2.2. The structure of loans
An important feature that we rely on to identify the
demand for securitizable loan facilities is the structure of
the bank loans themselves.10 Bank loans usually receive
capital from a syndicate of lenders, comprised of up to three
types of investors: institutional investors, banks, and nance
companies. To cater to the preferences of each type of
investor, these loans contain multiple facilities or tranches.
Commercial banks typically invest in the revolving portion
of the loan. Revolving credit lines can be drawn down and
repaid at the discretion of borrowers, but fees are paid based
8
The underwriter of a CLO frequently warehouses loan facilities
that are to be used as collateral in a CLO during the portfolio selection
process. An underwriting bank is likely to feel more comfortable warehousing a loan facility that it originated rather than a facility on which it
has less information.
9
The data presented below provide some evidence, albeit limited, in
support of this possibility. Of the Term Loan B facilities with exactly two
lead arrangers that are B-rated and hence more likely to be securitized,
42.3% of the time both lead arrangers are securitization-active. Among
the less likely to be securitized Ba-rated loans, both lead arrangers are
securitization-active only 37.4% of the time. The 5% difference suggests
that B-rated loans tend to have a larger number of syndicate participants
that are also active in originating CLOs. A similar difference exists
between B- and Ba-rated loans when comparing the securitization
activity of the lead arranger when only one lead arranger participates
in the loan origination.
10
We rely on the institutional details provided in a primer on bank
loans published annually by Standard and Poors (2007).
335
on the entire amount committed to the lending facility

regardless of whether it has been drawn down. Term Loan
A facilities are typically senior, amortizing loans that are
generally kept by the issuing bank or purchased by other
commercial banks. In contrast, Term Loan B facilities,
which are also senior and usually pari passu to the Term
Loan A, typically mature with bullet payments, and are
usually sold to institutional investors.
Term Loan B facilities are appealing to CLOs for a number
of reasons.11 The main difference between Term Loan A and
Term Loan B concerns the payout structure, which is
amortizing for Term Loan A facilities and bullet for Term
Loan B. The amortizing payout structure leads to a shorter
duration for Term Loan A than Term Loan B facilities. This
shorter duration creates reinvestment risk, and even if there
are good places to reinvest the money, the CLO managers
are more frequently required to search for places to reinvest
the steady stream of amortized principal as it is paid back.
Collateral managers nd it easier to reinvest the bullet
payments of Term Loan Bs when they mature or prepay.12
In addition, the purchasers of AAA bonds (insurance companies, pension funds) produced by CLOs tend to prefer to
purchase bonds with longer maturity. It is easier for CLO
arrangers to create bonds with a longer expected maturity
with collateral of longer duration.
CLOs are more likely to invest in institutional tranches
(Term Loan B specically) than in the revolving facilities
or Term Loan A facilities (see Table 1 for evidence). When
demand for loans from CLOs increased dramatically in the
20022007 period, this demand likely affected the Term
Loan B portion of corporate loans substantially more than
either the Term Loan A portion or the revolver portion.
2.3. The ratings structure of CLOs
The predominant form of corporate loan securitization is
what is known as a cash-ow arbitrage CLO.13 The idea
underlying this structure is to combine a group of low-rated
loans into a structure that has a higher rating. Covenants in
CLO deals require that each deal maintain a minimum
weighted-average credit rating (WAR), which is calculated
as the average credit rating of the collateral, weighted by the
size of each piece of collateral. Not surprisingly, CLOs are
almost always structured to use the minimum quality
collateral so as to satisfy this constraint and yield a high,
usually AAA, rating for the CLO. Benmelech and Dlugosz
(2009), analyzing a comprehensive sample of cash-ow
arbitrage CLOs, show that the weighted-average credit rating
in 85% of the deals in their sample was B (B1 in Moodys
rating classication). Furthermore, they document that the
average B1 rating of the collateral was almost uniformly met
11
This paragraph is based on conversations with practitioners who
originate CLOs and summarizes the typical arguments for why CLOs
usually prefer to purchase the Term Loan B tranche of a loan rather than
the Term Loan A tranche.
12
Because of these differences in duration, we control econometrically for differences in term structure when estimating differences in
spreads between Term Loan As, Term Loan Bs, and revolvers below.
13
Other forms of CLOs include balance-sheet CLOs, whose primary
purpose is to remove loans from banks balance sheets. According to Moodys
data on CLO originations, balance-sheet CLOs represent about 1% of all CLOs.
336
with uniformly B-rated collateral (with collateral concentrated specically in the B1 rating).14
Consequently, when demand for CLO collateral
increased in the securitization-active period, we expect
that this demand will be reected in demand for B-rated
collateral more than for higher-rated debt. This increase
should be in the institutional (Term Loan B) tranche of the
loan and not for the Term Loan A or revolver tranches of
the loans. In addition, it should be higher for loan facilities
issued by banks active in securitization, that presumably
package these facilities immediately into the CLOs that
they arrange themselves. These differences in attractiveness to originators of CLOs lead to clear identication of
the effect of demand for collateral on the relative pricing
of different kinds of bank loan facilities.
3. Data
3.1. Sample construction
To study the way in which securitization affects the
pricing of loan facilities, we utilize data from two publicly
available databases: Reuters Dealscan and Moodys
Enhanced Monitoring Service (EMS) data services. We
restrict our sample to Dealscan loan facilities originated
between 2002 and September of 2007, at which time CLO
originations began to decline.15 Our sample requires that
facilities have Moodys credit ratings on the borrowing
rms senior debt at the time of loan origination.16 Our
analysis focuses specically on rms with Moodys
senior-debt credit ratings of B (B1, B2, or B3) or Ba (Ba1,
Ba2, Ba3). We further restrict the sample to include only
syndicated loans originated in the United States and
remove nancial rms with Standard Industrial Classication (SIC) codes between 6000 and 6500. We also
identify and remove any second-lien or mezzanine facilities from the sample.
The Moodys EMS sample provides data on the characteristics and identity of all the securities serving as
collateral in any Moodys-rated CLO as of the rst quarter
2009. The strength of the Moodys data is that it allows us
to identify exactly which loan facilities serve as collateral
in CLOs as of the rst quarter 2009. Unfortunately, it
14
The average CLO deal in their sample transformed the B-rated
collateral into securities in which 70% of the principal balance of the
originated bonds was rated AAA, while 90% of the principal was rated
BBB or higher. The pattern is very similar in the Moodys sample of CLOs
we analyze below.
15
Our identication strategy relies on differences in CLO demand
for securitizable loan facilities on account of facility attributes, rather
than substantial changes in CLO demand through time. As such, we
begin our sample in 2002 so that our estimates are concentrated during
the securitization boom.
16
We require the rating on the senior debt and not on the specic
loan because it is available for a much larger number of loans: there are
4,536 facilities in Dealscan that meet our sample criteria for which there
is a rating on the senior debt and only 2,351 for which there is a rating
on the specic facility. For the loans in our sample that have both
facility-specic credit ratings and senior-debt credit ratings, the ratings
are either the same or within one ratings class (i.e., B1 senior debt and
Ba3, B1, or B2 facility-specic rating) over 71% of the time.
provides only a one-time snapshot of securitized facilities,

and does not contain unique identiers that map to
Dealscan.
To illustrate how our matching of the two databases
occurs, consider the following examples from the Dealscan and Moodys EMS databases. In October 2006,
Armstrong World Industries raised $1,100M through a
bank loan; $300M in the form of a revolving line, $300M
in a Term Loan A facility, and the remaining $500M in a
Term Loan B facility. The Moodys EMS database reports a
security with the name Armstrong World Ind.Term
Loan B in the amount of $500M with expected maturity
in October 2013. It is important to note that no other
facility associated with the Armstrong World Industries
bank loan appears in the Moodys database. Consequently,
we classify the Term Loan B portion of the Armstrong debt
issuance as being securitized while the revolving and
Term Loan A facilities are not.
3.2. Potential biases introduced through snapshot
sampling
We emphasize that there are limitations inherent
in constructing the database in this manner. Because
the Moodys data represent a snapshot of CLOs in 2009,
we cannot tell whether a loan facility was initially
securitized and subsequently paid off prior to 2009 or
if it was included and then dropped from the CLO prior
to 2009.17 We also cannot be sure that the facilities we
observe as being securitized in 2009 were intended to be
securitized at origination. In addition, the imperfect
nature of matching suggests that some matches were
likely missed due to human error or uncertainty in the
naming of the facilities. Consequently, it is likely that the
matching process employed here understates the number
of facilities in our sample that are securitized. Finally, it
could be the case that better facilities, those with a lower
spread, are more likely to be outstanding as of 2009.
For these reasons, our primary empirical strategy does
not rely on an exact determination of whether a loan
facility was securitized. Instead, we rely on the EMS
database to identify attributes of loan facilities that are
simply correlated with securitization activity. This
approach avoids potential survivorship bias coming from
higher quality loans being more likely to be outstanding
in 2009.
3.3. Summary statistics
Table 1 presents summary statistics on the full sample
of facility-level loan data. Panel A breaks down these
statistics by facility type. It indicates that revolving
17
Collateral managers of CLOs actively manage the portfolio
through time selling their participation in one loan facility and replacing
it with another. They would do so because their compensation is in part
tied to the adequacy of the collateral pool. CLOs do contain covenants
that constrain the frequency of trading within the collateral. Benmelech
and Dlugosz (2009) report that typically 1015% of the par value of
assets may be traded per year in addition to defaulted securities or
credit-risk securities.
337
Table 1
Facility summary statistics by facility type, credit rating, facility purpose, and facility attribute.
In this table we report summary statistics on loan facilities identied as having been securitized and compare them against the attributes of nonsecuritized facilities. The Moodys EMS database provides a snapshot of the characteristics and identity of all the securities serving as collateral in any
Moodys-rated CLO as of the rst quarter of 2009. The securitized sample is dened as any loan facility identied as collateral in a CLO according to
Moodys EMS database as of the rst quarter 2009. We outline the matching process in the text. We report summary statistics on the number of loan
facilities and all-in-drawn spread over LIBOR charged on loan facilities at the time of loan origination. Panels AC report summary statistics by facility
type, credit rating, and facility purpose, respectively. Panel D reports summary statistics on facility attributes. The Dealscan sample consists of senior
secured and unsecured loan facilities originated between 2002 and September 2007 which have Moodys credit ratings on the senior debt of the issuing
rm available at the time of loan origination. We restrict our sample to include syndicated loan facilities originated in the United States and exclude rms
with SIC codes between 6000 and 6500. We also restrict our analysis to facilities with senior debt rated Ba1, Ba2, Ba3, B1, B2, and B3. We identify and
remove any second-lien or mezzanine facilities from the sample.
Panel A
Non-securitized sample
Facility type
Revolver
Term Loan A
Term Loan B
Securitized sample
Count
2514
278
1413
All-in-drawn spread over LIBOR

25th
Median Mean
75th
%tile
%tile
150.0
150.0
200.0
225.0
225.0
250.0
226.8
228.0
276.7
Count
275.0
300.0
325.0
34
20
277
Panel B
Term Loan Bs only - Non-securitized sample
Moodys sr. debt credit rating
Ba1
Ba2
Ba3
B1
B2
B3
Corporate purposes
Debt repayment
Takeover
Working capital
Leveraged buyout (LBO)
Acquisition line
Other
200.0
143.8
200.0
250.0
150.0
225.0
235.3
175.6
240.4
250.0
200.0
275.0
Diff
t-Stat
8.5
52.4
36.3
(1.06)
( 3.40)
( 7.39)
Term Loan Bs only - securitized sample

Count
25th
%tile
Median
Mean
75th
%tile
Count
25th
%tile
Median
Mean
75th
%tile
Diff
t-Stat
49
138
290
458
338
140
175.0
175.0
200.0
225.0
225.0
250.0
200.0
225.0
225.0
250.0
275.0
275.0
224.5
230.7
250.0
273.9
306.6
332.3
275.0
275.0
275.0
300.0
350.0
400.0
9
22
50
101
68
27
175.0
150.0
175.0
200.0
225.0
225.0
175.0
187.5
225.0
225.0
250.0
250.0
181.9
200.0
223.0
241.1
265.4
259.3
187.5
225.0
250.0
275.0
287.5
275.0
42.5
30.7
27.0
32.8
41.2
73.1
( 2.86)
( 2.13)
( 3.24)
( 4.67)
( 3.39)
( 4.27)
Panel C
Term Loan Bs onlyNon-securitized sample
Facility purpose

25th
Median Mean
75th
%tile
%tile
Term Loan Bs onlysecuritized sample

Count
25th
%tile
Median
Mean
75th
%tile
Count
25th
%tile
Median
Mean
75th
%tile
Diff
t-Stat
445
123
183
168
209
112
173
200.0
225.0
200.0
200.0
225.0
218.8
225.0
250.0
275.0
225.0
250.0
275.0
250.0
267.5
267.9
308.5
250.9
279.7
289.1
266.3
277.4
325.0
350.0
275.0
325.0
325.0
300.0
304.2
85
30
39
19
63
12
29
200.0
175.0
200.0
175.0
225.0
175.0
200.0
225.0
200.0
200.0
225.0
250.0
225.0
219.6
241.2
214.2
227.2
226.3
256.7
256.3
243.1
250.0
250.0
250.0
275.0
300.0
337.5
264.3
26.8
94.3
23.6
53.4
32.4
10.0
( 2.65)
( 6.56)
( 2.14)
( 3.18)
( 4.04)
( 0.39)
Diff
t-Stat
Panel D
Term Loan Bs onlyNon-securitized sample
Loan attribute
Count
25th
%tile
Facility amount ($ million)

Sales size ($ million)
Tranche maturity (months)
Syndicate size (#of
participants)
% With covenants
1432
1432
1355
1431
$125.0
$272.7
60.0
3.0
1432
Term Loan Bs Onlysecuritized sample

Median
Mean
$219.5 $388.0
$616.5 $1,757.3
72.0
71.4
5.0
8.5
54.2%
facilities are the most common facility, followed by Term

Loan Bs and Term Loan As. Panel B focuses exclusively on
the Term Loan B sample and provides statistics on the
number and spread over London InterBank Offered Rate
(LIBOR) for securitized and un-securitized loan facilities of
75th
%tile
Count
25th
%tile
$400.0
$1,500.7
84.0
9.0
278
278
276
277
$180.0
$347.3
72.0
3.0
278
Median
Mean
$315.0 $519.8
$704.1 $3,055.0
84.0
76.3
5.0
7.0
62.6%
75th
%tile
$600.0
$1,510.7
84.0
7.0
$131.8
(2.91)
$1,297.6 (1.58)
5.0
(5.74)
1.4
( 2.37)
8.40%
(2.63)
different credit qualities. As expected, facility spreads

increase monotonically as credit quality declines. Panel
C reports the same statistics broken down by the reported
purpose of the facility. The somewhat vague description
of corporate purposes appears most frequently in the
338
Table 2
The creation of CLOs through time and by originator.
This table reports the frequency and timing of CLO originations by CLO underwriter. Data include all CLOs originated in the United States and Europe as
reported in Moodys EMS database. Of the 884 reported CLOs, less than 1% are reported as balance-sheet CLOs. The remaining are primarily cash-ow
arbitrage CLOs. Over 80% of all deals were created between the 2002 and 2007 period, with almost 50% of the deals being created in 2006 and 2007. Over
78% of the deals were originated by the top-ten originating banks. The top-ten underwriting banks over our 2002September 2007 sample period are
classied as securitization-active.
CLO originating bank
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
Total
JPMorgan
Citigroup
Bear, Stearns & Co. Inc.
Credit Suisse
Lehman Brothers
Deutsche Bank
Goldman Sachs & Co.
Morgan Stanley
Wachovia Securities Inc.
Merrill Lynch & Co.
Bank of America Securities LLC
UBS Securities LLC
IXIS Securities North America Inc.
Barclays Capital
CIBC World Markets Inc.
Royal Bank of Scotland plc
Salomon Smith Barney
BNP Paribas
Others (28 other banks)
0
1
0
0
1
0
0
0
0
1
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
1
0
2
0
0
0
0
2
0
0
0
0
0
0
0
1
2
1
1
2
1
4
0
0
0
0
2
0
0
0
2
3
1
4
2
2
1
2
3
2
1
1
0
0
0
3
0
0
0
4
2
0
2
2
1
0
5
3
2
3
1
0
0
0
1
0
2
0
7
6
0
3
2
1
4
5
3
0
1
1
1
0
0
0
0
1
0
1
6
0
2
7
3
3
4
3
3
2
1
0
0
1
2
0
3
0
3
7
2
3
10
3
8
5
2
3
2
2
1
2
1
0
0
2
1
2
7
1
12
5
6
5
3
3
5
4
2
3
3
0
0
0
0
0
5
13
17
13
12
11
2
6
7
9
10
5
5
2
0
0
0
0
1
10
26
28
21
13
10
16
12
16
18
7
14
5
9
2
0
1
0
2
20
18
24
17
14
19
18
17
18
18
10
7
5
1
4
0
7
0
3
13
5
11
1
1
5
5
3
2
2
3
3
1
0
4
0
0
0
0
5
94
85
78
69
64
63
63
63
63
50
37
21
17
12
10
8
8
7
72
Total
16
29
31
29
43
56
64
123
220
213
51
884
data, followed by debt repayments and corporate takeovers. Facilities supporting leveraged buyouts represent
16% of the Term Loan B facilities we study, and these
loans are securitized at a higher rate than other loan
facilities. Panel D reports statistics on facility attributes
such as facility amounts, annual sales of the borrowing
rms, tranche maturity, syndicate size, and the percentage of facilities with nancial covenants.
Table 1 also shows the differences in the attributes of
loan facilities identied as having been securitized, relative to their non-securitized counterparts. The table
indicates that securitized loan facilities have lower
spreads, with the largest differences in spreads concentrated in lower credit-quality facilities. In addition, securitized facilities have larger facility sizes, slightly longer
maturities, and were arranged by smaller syndicates. The
differences between securitized and non-securitized facilities highlight the importance of controlling for differences in facility size, sales size, tranche maturity, and
covenants, in drawing inferences on the incremental
effect of securitization.
The time-series pattern of CLO origination in our
sample is illustrated in Table 2. The sample is predominately made up of CLOs originated between 2002 and
2007. In addition, the CLOs are highly concentrated
among a relatively small number of originating banks,
with the top ten banks underwriting 692 of the 884 CLOs
(78%) over the entire sample and 569 of the 719 deals
(79%) during the 20022007 time period. Although
Table 2 focuses on the CLO underwriting activity of 18
unique banks, at least 38 different banks were involved in
the underwriting of at least one CLO between the years
1996 and 2008.
For each loan facility in our sample, we identify the

bank which served as the lead arranger of the bank loan, a
role often lled by more than one bank (45% of originations in the full sample). We classify banks as being a lead
arranger if they are designated by Dealscan as being a
top-tier arranger, or lead arranger in the deal summary.18 We classify facilities as being originated by a
securitization-active bank if any of the lead arranging
banks is among the top ten CLO underwriting banks. Of
the 272 securitized loan facilities that were arranged by
the top ten securitization-active banks, 117 of the
facilities (43%) had the same bank involved in both the
arranging and securitization of the facility. In contrast, of
the 59 securitized facilities that we identify as being
arranged by non-securitization-active banks, only eight
(13.6%) of them had the same bank involved in the
arranging and securitization process. These calculations
suggest that there was a connection between the arrangement and securitization of the facilities from the securitization-active banks, and that these banks are
particularly likely to securitize loan facilities that they
arrange themselves.
Fig. 1 plots the quantity of originations of Term Loan B
facilities over time. Panel A plots the 3-month moving
average of the number of Term Loan B facilities originated
by securitization-active and non-securitization-active
banks, where a securitization-active bank is dened as
one of the top ten banks ranked by the number of
18
Being lead arranger is typically much more involved than being a
member of the lending syndicate, since lead arrangers usually serve as
the primary underwriter of loan quality, develop the banking relationship with borrowers, and arrange the lending syndicate.
339
Full sample
No. of Term B facilities
(3-mth mov. avg.)
35
Securitization-active
30
Non-securitization active
25
20
15
10
5
2005 Q1
2005 Q3
2006 Q1
2006 Q3
2007 Q1
2007 Q3
2005 Q1
2005 Q3
2006 Q1
2006 Q3
2007 Q1
2007 Q3
2005 Q1
2005 Q3
2006 Q1
2006 Q3
2007 Q1
2007 Q3
2004 Q3
2004 Q1
2003 Q3
2003 Q1
2002 Q3
2002 Q1
2001 Q3
2001 Q1
2000 Q3
2000 Q1
1999 Q3
1999 Q1
1998 Q3
1998 Q1
Securitization-active sample
B-rated
Ba-rated
20
15
10
2004 Q3
2004 Q1
2003 Q3
2003 Q1
2002 Q3
2002 Q1
2001 Q3
2001 Q1
2000 Q3
2000 Q1
1999 Q3
1999 Q1
1998 Q3
5
1998 Q1

(3-mth mov. avg.)
25
Non-securitization-active sample
B-rated
10
Ba-rated
8
6
4
2004 Q3
2004 Q1
2003Q3
2003 Q1
2002 Q3
2002 Q1
2001 Q3
2001 Q1
2000 Q3
2000 Q1
1999 Q3
1999 Q1
1998 Q3
2
1998 Q1

(3-mth mov. avg.)
12
Fig. 1. Plots of Term Loan B origination volume.

Fig. 1 plots the quantity of originations of Term Loan B facilities over time. Panel A plots the 3-month moving average of the number of Term Loan B
facilities arranged by securitization-active and non-securitization-active banks. We classify lead arranging banks as being securitization-active if any lead
arranging bank involved in the origination of a loan facility is among the top ten banks ranked by the number of securitization deals underwritten. Panel
B splits the sample into facilities with B-rated senior debt and Ba-rated senior debt and plots the origination activity of each ratings class by
securitization-active banks. Panel C plots differences in securitization activity by ratings class for non-securitization-active banks. The sample includes
syndicated loan facilities originated between 1998 and 2007, as reported in Dealscan, with Moodys senior-debt credit ratings of B (B1, B2, or B3) or Ba
(Ba1, Ba2, Ba3). We restrict the sample to include only syndicated loans originated in the United States and removed nancial rms with SIC codes
between 6000 and 6500. Section 3 of the paper more fully describes the data screens employed in arriving at our nal sample.
securitization deals underwritten during our sample period. The plot indicates that origination patterns were very
similar until 2002, when securitization-active banks began
to increase their origination of Term Loan B facilities
aggressively. Panel B plots the origination activity of
securitization-active banks by credit rating. The plot indicates that much of the boom in Term Loan B originations
was concentrated in facilities of a lower credit-quality.
Finally, Panel C plots the origination activity of non-
securitization-active banks by credit quality. While nonsecuritization active banks also increased their origination
of lower credit-quality facilities, the plots suggest that the
difference was not as dramatic as the difference observed
in the origination activity of securitization-active banks.
Overall, these plots suggest that there was a large increase
in securitization activity after 2002, and that this increase
was concentrated in the securitization-active banks, and in
lower credit-quality issues.
340
4. Results
4.1. Do securitized loans have lower spreads?
4.1.1. Univariate tests
Panel B of Table 1 presents the average difference in
spreads between Term Loan B facilities that were securitized and those that were not securitized according to the
Moodys data. As a rough control for facility attributes, we
present the comparison by rating class for the most
commonly issued rating levels: Ba1, Ba2, Ba3, B1, B2,
and B3.
This panel indicates that in each rating class, spreads
are lower for the securitized sample than for the nonsecuritized sample, with the difference ranging from 27
basis points for the Ba3 rating category to 73.1 basis
points for the B3 category. The average difference in
spreads among the B-rated facilities is 49 basis points,
as opposed to the 33.4 basis point difference among Barated facilities. The lower spreads for securitized facilities
are consistent with the notion that the ability to securitize
a facility leads to a lower spread, and consequently a
lower cost of capital for the rm borrowing the money.
However, it is not a conclusive estimate of the incremental effect of securitization on spreads for at least three
reasons. First, we have an imperfect matching of loan
facilities to CLOs that likely understates the fraction of
loans that are securitized. Second, these differences do not
control for other factors that are known to affect facility
pricing. Finally, it is possible that some other factor could
affect both spreads and securitization decisions simultaneously, leading the observed relation between the two to
occur but not to be causal.
The rst issue, involving the imperfect matching of
loan facilities to CLOs, results from both the fact that we
have only a snapshot of CLOs as of 2009, and also from the
by hand nature of the matching process. As a result, it is
likely that most or all of the facilities classied as
securitized really were securitized, but that a number of
facilities that we classied as non-securitized actually
were securitized. Such a misclassication would imply
that the differences documented in Table 1 would understate the true differences in spreads between securitized
and non-securitized loan facilities.
4.1.2. Multivariate tests
The second issue, that factors in addition to whether a
particular loan facility is securitized are likely to affect
spreads, is addressed using a multivariate regression
framework. We create a dummy variable set equal to
one if the facility was part of a CLO in the Moodys
database, and estimate a regression of the following form:
Spreadi,t a0 b Securitizedi,t g Borrower f actorsi,t
sales, size of the lending syndicate, tranche maturity,

industry, year, and lead arranger xed effects.
We control for macroeconomic factors such as the
term structure of interest rates and changes in the market
price of risk. Our term structure control is calculated as
the difference between the 5-year and 3-month T-bill rate
in the month of deal origination. To control for changes
in the price of credit risk, within each month, we calculate
the spread of a 10-year Ba- and B-rated bond index
over the 10-year T-bond rate. Loan facilities whose
borrowers have B-rated (Ba-rated) senior debt are
assigned the value of the B-spread (Ba-spread) in the
month of origination.
Table 3 reports estimates of this equation using the
sample of Term Loan B facilities with a Moodys senior
debt rating of B or Ba. Column 1 suggests that the impact
of securitization on spreads is negative for all Term Loan B
facilities, but not statistically different from zero. The
estimates in column 3 show that the impact of securitization on yield is concentrated in the lower-rated B facilities, which are associated with a statistically signicant
17.7 basis point drop in spread. These results are consistent with the notion that banks lower the spreads on loan
facilities that will be securitized for the more commonly
securitized B-rated facilities.
4.2. Empirical approach
The results in Table 3 suggest that securitized loan
facilities are associated with lower spreads, holding other
factors constant, at least to the extent that the facilities
risk is measured by Moodys ratings. But what is the
appropriate interpretation of this nding? The pattern is
certainly consistent with the explanation in which high
demand for securitizable assets from CLOs led to a lower
cost of capital. However, the negative association between
securitization and spreads does not by itself rule out an
alternative explanation in which the demand for funds in
the borrowing market shifted differentially, with a relatively smaller increase in demand for loans that could
easily be securitized than for loans that are difcult to
securitize.19
The usual way to address this type of endogeneity is
through an instrumental variables approach. Unfortunately, to be a valid instrument for whether a facility is
securitized would require a variable to be related to the
securitization decision but unrelated to the spread
charged in the facility. Since in practice, virtually all
publicly available variables relating to the facility potentially are at least arguably related to its spread, the use of
instrumental variables is not likely to be a practical
solution.
To distinguish between alternative explanations, we
instead rely on the fact that the securitization process by
j Macro f actorst ei,t ,

where the subscripts refers to loan facility i at time t. The
list of borrower controls includes credit ratings along with
dummy variables for secured loan facilities, loan facilities
with covenants, and the existence of a pricing grid. We
also control for facility amount, facility type, borrowers
19
In considering a primary-market demand explanation for the
observed lower spreads on securitized facilities, it is not clear why CLOs
would choose to purchase loan facilities with lower spreads than
average. One possibility would be if spreads were not priced completely
by the market for some reason, leading banks to sell the lower spread
facilities and keep the higher spread ones for any particular quality level.
341
Table 3
Do securitized loans have lower spreads?
In this table we estimate an OLS regression of facility attributes and other controls on the all-in-drawn spread of loan facilities originated between 2002
and September 2007. We create an indicator variable, Securitized indicator, for loans identied as collateral in a CLO as of 2009 using Moodys EMS
database. Term structure is calculated as the difference between the ve-year and three-month T-bill rate in the month of deal origination. Credit spread is
the spread of a 10-year Ba- and B-rated bond index over the 10-year T-bond rate, respectively. Covenants indicator is equal to one if a facility has nancial
covenants. Pricing grid indicator is equal to one if the spread is adjustable based on accounting performance. Size of syndicate is the number of lenders
involved in the lending syndicate. Tranche maturity is the number of months until the facility matures. Log facility amount and Log sales size represent the
dollar amount of the loan facility and total sales of the borrowing rm, measured in U.S. $ million. Secured indicator is an indicator equal to one if the loan
facility is secured. In various specications, as indicated in the table, we control for tranche-type xed-effects (Revolver, Term Loan A, or Term Loan B),
credit-rating xed-effects (Ba1, Ba2, Ba3, B1, B2, B3), tranche-purpose xed-effects (corporate purposes, LBO, debt repayment, etc.), year xedeffects, lead arranger xed-effects, and industry xed-effects (two-digit SIC code). We restrict our sample to include syndicated loan facilities originated
in the United States and exclude rms with SIC codes between 6000 and 6500. We also restrict our analysis to loan facilities with senior debt rated Ba1,
Ba2, Ba3, B1, B2, and B3. We identify and remove any second-lien or mezzanine facilities from the sample. Standard errors are clustered by year and by
borrowing rm. The symbols ***, **, and * indicate signicance at the 1%, 5%, and 10% levels, respectively.
Dependent var.: All-in-drawn spread
Term B only, Ba and B-rated sr. debt

(1)
Term B only, Ba-rated sr. debt

(2)
Term B only, B-rated sr. debt

(3)
6.872
(1.46)
3.214
(0.24)
32.857nnn
(3.15)
9.372
(1.35)
8.665
(1.15)
0.325nn
(2.38)
1.133nnn
(2.69)
7.133nn
(2.20)
3.828n
(1.75)
2.743
(0.48)
334.458nnn
(6.49)
16.476nnn
(3.00)
0.931
(0.06)
44.273nnn
(3.94)
19.570nnn
(2.98)
3.492
(0.43)
0.212
(1.12)
1.083nnn
(4.96)
0.743
(0.29)
5.024n
(1.74)
10.052
(0.92)
302.778nnn
(4.00)
17.698nn
(2.07)
7.714
(0.80)
21.228n
(1.83)
8.264
(1.04)
12.308
(1.52)
0.660nnn
(5.39)
1.257n
(1.94)
11.715nnn
(3.15)
1.422
(0.47)
13.352
(1.04)
621.844nnn
(9.28)
Tranche-type xed effects

Moodys sr. debt xed effects
Tranche-purpose xed effects
Year xed effects
Lead arranger xed effects
Industry xed effects
No
Yes
Yes
Yes
Yes
Yes
No
No
Yes
Yes
Yes
Yes
No
No
Yes
Yes
Yes
Yes
Standard errors clustered by:

Borrower
Year
Adjusted R2
Observations
Yes
Yes
0.386
1613
Yes
Yes
0.482
540
Yes
Yes
0.312
1073
Securitized indicator
Term structure
Credit spread
Covenants indicator
Pricing grid indicator
Size of syndicate
Tranche maturity
Log facility amount
Log sales size
Secured indicator
Constant
its very nature leads to a preference for certain types of

loan facilities over others. For example, it is commonly
believed that it is easier to construct CLOs from Term Loan
Bs, which have bullet payments that are set in advance,
than Term Loan As, which have amortizing payments, or
revolvers, whose payments depend on the extent to
which the borrower chooses to draw them down. In
addition, Table 2 suggests that CLOs are highly concentrated in a small number of issuing banks, so loan
facilities initiated by those banks are more likely to be
securitized than facilities by other non-issuing banks.
Finally, Benmelech and Dlugosz (2009) suggest that
B-rated loan facilities are the most commonly used type
of facility in CLOs, a nding conrmed in the summary
statistics of our sample of securitized facilities (see

Table 1).
Our approach is to compare the relative yields on
facilities with securitization-friendly characteristics to
those with securitization-unfriendly characteristics
during the 20022007 securitization boom. The securitization affecting supply of capital argument predicts
that the demand from CLOs for facilities having these
attributes increased relative to other types of loan facilities, leading to lower yields. However, there is no
particular reason why demand for capital should be
reected differentially between these types of facility
offerings. Therefore, we interpret the difference in yields
for loan facilities originated by securitization-active banks
342
relative to non-securitization-active banks and differences between B-rated facilities and Ba-rated facilities
as reecting the effect of securitization on the cost of
capital.
4.3. Which factors affect the likelihood that a loan is
securitized?
A key assumption in our analysis is that the factors we
posit as being related to the securitization decision
actually do affect the likelihood a particular loan facility
is securitized. To ensure that these assumptions hold in
our sample, we estimate equations predicting whether a
particular loan facility from the Dealscan sample is
securitized according to the Moodys classication. We
estimate the following econometric specication:
PrSecuritizedi,t a0 b Facility typei,t g Borrower f actorsi,t
j Macro f actorst ei,t :
The Facility type variables are dummy variables indicating the type of facility. The equation also includes the
borrower controls used in Table 3, with the exception of
tranche maturity, which is extremely highly correlated
within tranche type.20 Because the dependent variable is
dichotomous, we estimate the equation by probit. We
constrain the sample to include years 20022009, since
loan facilities originated prior to 2002 likely would have
matured prior to 2009, making it impossible for such
facilities to serve as collateral in CLOs as of 2009.21
Table 4 reports estimates of this equation. The results
in column 1 suggest that Term Loan B facilities are more
likely to be securitized than Term Loan A facilities (the
omitted category). The estimated coefcient implies,
holding other factors constant, that Term Loan Bs are 3
percentage points more likely to be securitized than Term
Loan As (the omitted group in the probit estimation).22
Revolvers are securitized at a signicantly lower rate than
either type of Term Loan (unconditionally, with less than
1% frequency). The estimated coefcient on the securitization-active indicator suggests that loan facilities
initiated by the banks we consider to be Securitizationactive are more likely to be securitized than other
facilities. We estimate whether B-rated facilities are
securitized at a different rate than Ba-rated facilities
through the inclusion of a B-rated dummy variable. The
resulting estimates conrm those of Benmelech and
Dlugosz (2009) and are consistent with the unconditional
results presented in Table 1.
In column 2 we include tranche purpose xed-effects,
year xed-effects, and industry xed-effects, and reestimate the model. These controls do not materially change
the inference on securitization likelihood for Term Loan B
20
Revolvers have a very short maturity, while Terms B and A
facilities have longer maturities. Including tranche maturity in the probit
specication is essentially redundant to controlling for facility type
when revolvers are included in the sample.
21
Over 90% of the loan facilities in our sample have an expected
tranche maturity of less than 7 years.
22
Unconditionally, 12.8% of Term Loan Bs are securitized compared
to a 4.3% frequency for Term Loan As.
or securitization-active facilities. However, in this specication we cannot reject the hypothesis that Ba and B
facilities are securitized at the same rate at conventional
signicance levels since the t-statistic drops to 1.44. The tstatistic is sensitive to the choice of clustering standard
errors by borrower; when they are clustered instead by
credit rating, the t-statistic on the B-rated dummy variable rises to 3.20, which is statistically signicant at
conventional levels.
In column 3 we test whether B-rated Term Loan B
facilities are securitized more frequently than their Ba-rated
Term Loan B counterparts. The test requires an interaction
term, the interpretation of which can be problematic when
estimated via probit (see Ai and Norton, 2003), so we
estimate this equation using a linear probability model.23
The coefcient on the interaction of Term Loan Bs with Brated senior debt is 0.018, and is marginally statistically
signicant (t-statistic1.62).24 Since Ba-rated Term Loan Bs
were securitized about 10% of the time according to the
Moodys data (which substantially undercounts the fraction
that were securitized in practice), an estimated 1.8 percentage point increase in securitization likelihood for B-rated
Term Loan Bs relative to Ba-rated Term Loan Bs represents
about an 18% increase in securitization likelihood. Column 3
also tests whether Securitization-active banks securitize
Term Loan B facilities more frequently than Term Loan B
facilities originated by non-securitization-active banks. The
positive estimated interaction term suggests that the effect of
being from a securitization active bank on securitization
likelihood is most pronounced in Term Loan B facilities.
Overall, the results in Table 4, together with the evidence
in Benmelech and Dlugosz (2009), suggest that B-rated loan
facilities are more likely to be securitized than Ba-rated ones,
that facilities from banks designated as Securitizationactive are more likely to be securitized than facilities from
other banks, and that these differences are most pronounced
in Term Loan B facilities.
4.4. The impact of CLO demand on loan pricing: differences

in spreads by type of loan
The hypothesis that demand from CLOs affected pricing
on loan facilities predicts that when CLO demand is high,
then facilities that are most attractive to CLOs should be
favored by banks issuing loans. Consequently, these facilities
should be priced more aggressively than facilities that are
less attractive to CLOs. Since loan facilities originated by
securitization-active banks appear to be more likely to be
securitized than facilities originated by non-securitizationactive banks and because CLO demand appears to be more
concentrated in B-rated debt, we hypothesize that spreads
will reect differences in this demand.
23
Although the mean of the dependent variable in this equation is
0.071, Angrist and Pischke (2009) suggest that even when the mean of
the dependent variable is close to either zero or one, a linear probability
model will nonetheless produce accurate estimates of marginal effects
(see Angrist and Pischke, 2009, p. 109).
24
These estimates are obtained when standard errors are clustered
both by borrower and by credit rating. The t-statistic falls to 0.97 when
standard errors are clustered by borrower and year.
343
Table 4
What type of loans get securitized?
In columns 1 and 2 we estimate a probit model where the dependent variable is equal to one if the loan facility is identied as collateral in a CLO as of 2009, and
zero otherwise. Column 3 employs the same dependent variable but the coefcients are estimated using an OLS linear probability model. Term Loan B is equal to
one if the facility is a Term Loan B facility. Securitization active bank is equal to one if any of the lead arrangers of the loan facility are among the top-ten CLO
underwriters (see Table 2). B-rated senior debt is equal to one if the senior debt associated with the loan facility is rated B1, B2, or B3 at loan origination. Revolver is
equal to one if the loan facility is a revolving facility (Term Loan As serve as the omitted loan facility type). Covenants indicator is equal to one if a facility has
nancial covenants. Pricing grid indicator is equal to one if the spread is adjustable based on accounting performance. Syndicate size is the number of lenders
involved in the lending syndicate. Log facility amount and Log sales size represent the dollar amount of the loan facility and total sales of the borrowing rm,
measured in U.S.$ million. Secured indicator is an indicator equal to one if the loan facility is secured. In various specications, as indicated in the table, we control
for tranche-purpose xed-effects (corporate purposes, LBO, debt repayment, etc.), year xed-effects, and industry xed-effects (two-digit SIC code). We
restrict our sample to include syndicated loan facilities originated in the United States and exclude rms with SIC codes between 6000 and 6500. We also restrict
our analysis to loan facilities with senior debt rated Ba1, Ba2, Ba3, B1, B2, and B3. We identify and remove any second-lien or mezzanine facilities from the sample.
The sample size in column 2 drops slightly because the inclusion of xed effects perfectly predicts outcomes in the left-hand side variable. These observations are
dropped from the estimation. Standard errors are clustered by year. The symbols ***, **, and * indicate signicance at the 1%, 5%, and 10% levels, respectively.
Estimated using probit
Dependent var. 1 if loan is identied as having been

securitized
Term B, Term A, and

Revolvers
(1)
Estimated using OLS (Mean dep.

var. 0.071)
Term B, Term A, and

Revolvers
(2)
Term B, Term A, and Revolvers

(3)
Term Loan BnB-rated senior debt
0.018n
(1.62)
Term Loan BnSecuritization-active bank
0.038n
(1.94)
Term Loan B
0.341nnn
(3.128)
0.484nnn
(3.966)
0.030nnn
(2.92)
B-rated senior debt
0.156nn
(1.987)
0.126
(1.440)
0.004
(0.80)
Securitization-active bank
0.218nnn
(2.691)
0.301nnn
(3.387)
0.006
(0.93)
Revolver
0.874nnn
( 7.327)
0.907nnn
( 7.207)
0.065nnn
(8.55)
Covenants indicator
0.069
(0.796)
0.076
(0.910)
0.011nnn
( 2.636)
0.158nnn
(4.096)
0.008
(0.256)
0.522nnn
(5.139)
5.222nnn
( 7.229)
0.300nnn
(3.157)
0.070
(0.771)
0.005
( 1.418)
0.067
(1.553)
0.010
(0.270)
0.452nnn
(4.333)
4.835nnn
( 4.909)
0.036nnn
(3.29)
0.003
(0.17)
0.001n
(1.95)
0.015nnn
(6.79)
0.002
(0.24)
0.037nnn
(5.40)
0.402nnn
(3.92)
No
No
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
0.201
Yes
No
0.297
Yes
Yes
5101
4755

Syndicate size
Log facility amount
Log sales size
Secured indicator
Constant
Year xed effects
Borrower
Credit rating
Psuedo-R2
Adjusted R2
Observations
We test whether increased demand by CLOs lowered

spreads by estimating the following specication:
Spreadi,t a0 b1 Sec:Activej,t *B ratedi,t b2 Sec:Activej,t
b3 B ratedi,t dX i,t ei,t ,
0.124
5070
where the sample is constrained to Term Loan Bs only.

The estimate on the interaction of the securitizationactive and B-rated dummy variables represents the
difference in the spread relative to non-securitization-active
B-rated facilities or securitization-active Ba-rated facilities,
holding other factors constant. The other variables in the
344
equation, represented by X i,t, are the term structure and

credit spread variables discussed above, along with the
remaining set of the controls described in our discussion
of Table 3. We include tranche purpose xed-effects, year
xed-effects, industry xed-effects, and credit-rating xedeffects where appropriate. As in Table 3, we cluster standard
errors by year and borrower.
Table 5 reports estimates of this equation over the
2002September 2007 sample period. Column 1 reports
results controlling only for the impact of securitizationactive banks on spreads, column 2 adds the B-rated
dummy, and column 3 includes both of those dummy
variables plus the interaction term. The estimate of the

coefcient on the interaction term is 20.9, which
implies Term Loan B facilities from securitization-active
banks had 20.9 basis point lower spreads than Term Loan
B facilities from other banks, holding other attributes of
the facility constant. When the interaction term is
included in Table 5, the coefcient on the securitizationactive bank dummy, which measures the impact on
Ba-rated debt (since there is a dummy interacting the
B-rated dummy with the securitization-active bank
dummy), is actually positive, albeit not signicantly different from zero. These estimates suggest that the pricing
Table 5
Do securitization-friendly Term Loan B facilities have lower spreads?
and September 2007. Securitization active bank is equal to one if any of the lead arrangers of the loan facility are among the top-ten CLO underwriters (see
Table 2). B-rated senior debt is equal to one if the senior debt associated with the loan facility is rated B1, B2, or B3 at loan origination. Term structure is
calculated as the difference between the 5-year and 3-month T-bill rate in the month of deal origination. Credit spread is the spread of a 10-year Ba- and
B-rated bond index over the 10-year T-bond rate, respectively. Covenants indicator is equal to one if a facility has nancial covenants. Pricing grid indicator
is equal to one if the spread is adjustable based on accounting performance. Size of syndicate is the number of lenders involved in the lending syndicate.
Tranche maturity is the number of months until the facility matures. Log facility amount and Log sales size represent the dollar amount of the loan facility
and total sales of the borrowing rm, measured in U.S. $ million. Secured indicator is an indicator equal to one if the loan facility is secured. In various
specications, as indicated in the table, we control for credit-rating xed-effects (Ba1, Ba2, Ba3, B1, B2, B3), tranche-purpose xed-effects (corporate
purposes, LBO, debt repayment, etc.), year xed-effects, and industry xed-effects (two-digit SIC code). We restrict our sample to include syndicated
loan facilities originated in the United States and exclude rms with SIC codes between 6000 and 6500. We also restrict our analysis to loan facilities with
senior debt rated Ba1, Ba2, Ba3, B1, B2, and B3. We identify and remove any second-lien or mezzanine facilities from the sample. Standard errors are
clustered by year and by borrowing rm. The symbols ***, **, and * indicate signicance at the 1%, 5%, and 10% levels, respectively.
Term Loan B sample

(1)
Term Loan B sample

(2)
Term Loan B sample

(3)
11.473n
(1.65)
11.337
(1.55)
20.924nn
(2.39)
20.918nnn
(2.65)
2.801
(0.34)
36.366nnn
(5.37)
4.870
(0.35)
33.965nnn
(2.90)
11.133
(1.56)
9.730
(1.32)
0.351nn
(2.34)
1.194nnn
(2.92)
8.142nnn
(2.86)
5.272nn
(2.21)
1.863
(0.30)
421.483nnn
(7.20)
5.412
(0.43)
32.985nnn
(3.00)
15.189nn
(1.97)
10.454
(1.24)
0.415nnn
(2.69)
1.294nnn
(3.05)
9.119nnn
(2.91)
4.045
(1.43)
4.697
(0.59)
440.514nnn
(5.56)
5.406
(0.44)
33.239nnn
(3.02)
15.153nn
(2.00)
10.246
(1.22)
0.414nnn
(2.80)
1.280nnn
(3.04)
9.153nnn
(2.93)
4.082
(1.46)
4.784
(0.60)
655.358nnn
(6.68)

Year xed effects
Lead arranger xed effects
Yes
Yes
Yes
Yes
No
No
Yes
Yes
Yes
No
No
Yes
Yes
Yes
No
Standard errors clustered by

Borrower
Year
Adjusted R2
Observations
Yes
Yes
0.372
1613
Yes
Yes
0.343
1613
Yes
Yes
0.344
1613
Securitization-active banknB-rated sr. debt

B-rated sr. debt
Term structure
Credit spread
Covenants indicator
Size of syndicate
Tranche maturity
Log facility amount
Log sales size
Secured indicator
Constant
impact of being active in securitization is concentrated in

the B-rated debt, which is the most commonly securitized
type of loan facility.
This 20.9 basis point spread for B-rated debt from
securitization-active banks is a substantial magnitude; a
20.9 basis point discount on an average-sized Term Loan B
facility ($400 million U.S. dollars) represents annual savings of close to US$ 1 million over the life of the loan. This
estimate implies that the impact of the issuing banks
identity is approximately the same as the impact of
covenants on spreads. In addition, increased CLO demand
also potentially affects non-pricing loan characteristics
such as the existence and tightness of loan covenants. To
the extent that the banks belonging to the Securitization-active group reects the likelihood of securitization,
it implies that banks are willing to discount loan facilities
to be able to have them as collateral for CLOs.
4.5. Comparisons of Term Loan B facilities to Term Loan A
facilities and revolvers
The results from Table 5 suggest that loan facilities
that are sold into CLOs are issued at a lower spread than
facilities that are not sold into a CLO. In addition, the
difference in spreads between loan facilities originated by
securitization-active banks and non-securitization-active
banks is most pronounced in facilities of lower credit
quality, which are the facilities most commonly securitized. While we have interpreted this nding as a
consequence of heightened demand for collateral from
CLOs, an alternative explanation is that the results occur
because of some unmeasurable bank-specic factor, such
as an increase in banks risk tolerance over this period for
the banks we classify as securitization-active. Such an
increase in risk tolerance for the securitization-active
banks could lead to both an increase in securitization
activity as well as an increase in lending to lower creditquality borrowers, consistent with the empirical evidence
presented thus far.
To distinguish between the demand for collateral and
increased risk tolerance explanations for the lower
spreads on Term Loan B facilities from securitizationactive banks, we consider how other facilities issued by
these same banks are priced. We do so because the other
facilities are not as attractive as collateral for CLOs as
Term Loan B facilities but have similar seniority as Term
Loan B facilities and consequently similar risk. In particular, we estimate whether the pricing of Term Loan B
facilities originated by securitization-active banks, especially for lower-rated facilities, is different from the
pricing of Term Loan A facilities with similar securitization-friendly attributes. To perform this test, we use a
sample made up of both Term Loan A facilities and Term
Loan B facilities, and estimate whether the relation
between securitization activity, ratings, and pricing documented above for Term Loan B facilities also holds for
Term Loan A facilities.
In doing so, it is important to control for other
differences between Term Loan A and Term Loan B
facilities. Both are typically senior bank debt with equal
priority, but have a different payoff structure because
345
Term Loan A facilities are normally amortizing and Term

Loan B bullet. Given that these loans have no prepayment
penalties, Term Loan B facilities will always be issued at a
higher spread than Term Loan A. However, if the demand
for collateral is largest at securitization-active banks and
for B-rated loans, we expect the differences to be largest
for these types of loans. To evaluate this hypothesis, we
include interaction terms into the equation that allow for
separate effects based on type of facility (Term Loan A or
Term Loan B), the securitization activity of the bank, and
the rating of the borrowing rms senior debt. In addition,
our estimates include the same controls as in earlier tests,
as well as year, tranche-purpose, and industry xedeffects.
Column 1 of Table 6 reports estimates of this equation.
The effect of demand for collateral can be seen in the
coefcient on the Term Loan B*Securitization-active
bank*B-Rated senior debt, which picks up the incremental effect on pricing of a loan facility having these
attributes, which are those most associated with being
desirable as collateral for a CLO. The results in this column
indicate that securitization-friendly Term Loan B spreads
are 30.6 basis points lower than spreads on Term Loan A
facilities with securitization-friendly attributes. This substantial coefcient suggests that the incremental effect of
having characteristics that make a facility particularly
attractive as collateral for CLOs implies that these facilities are sold at a substantial discount to otherwise similar
facilities.
In addition to Term Loan A facilities, most loans also
contain revolvers, which are lines of credit that can be
drawn down at the borrowers discretion. These facilities
also have the same seniority as Term Loan B facilities but
have a shorter expected maturity. Because the options
implicit in these revolvers make them less attractive as
collateral for CLOs, we expect to observe the spread
between revolvers and Term Loan Bs to decline for those
facilities for which the Term Loan Bs are most attractive
as collateral for CLOs.
Column 2 of Table 6 presents estimates of an equation
comparing the spreads of Term Loan Bs and revolvers,
again using interaction terms to allow for separate effects
of type of facility, the rating of the senior debt associated
with the facility, and the securitization activity of the
issuing bank. The results from this equation are similar to
the results for Term Loan As in the rst column. In
particular, the incremental effect of having a securitization-active originating bank and a B rating is to lower the
facilities spread by about 20.8 basis points, and this drop
is statistically signicantly different from zero.
To evaluate the magnitude of the results presented in
Table 6, recall that results presented in Table 5 indicate
that in a Term Loan B sample, B-rated debt is priced 20.9
basis points lower in facilities originated by securitization-active banks than in B-rated facilities originated by
non-securitization-active banks. The results in Table 6
indicate that a similar discount for B-rated facilities
originated by securitization-active banks does not exist
in a Term A or revolver sample. In fact, the results in
Table 6 indicate that B-rated, Term Loan A facilities
originated by securitization-active banks are priced about
346
Table 6
Is the securitization-friendly decline in spreads specic to Term Bs only?
and September 2007. Term Loan B is equal to one if the facility is a Term Loan B facility. Securitization active bank is equal to one if any of the lead
arrangers of the loan facility are among the top-ten CLO underwriters (see Table 2). B-rated senior debt is equal to one if the senior debt associated with
the loan facility is rated B1, B2, or B3 at loan origination. Term structure is calculated as the difference between the 5-year and 3-month T-bill rate in the
month of deal origination. Credit spread is the spread of a 10-year Ba- and B-rated bond index over the 10-year T-bond rate, respectively. Covenants
indicator is equal to one if a facility has nancial covenants. Pricing grid indicator is equal to one if the spread is adjustable based on accounting
performance. Size of syndicate is the number of lenders involved in the lending syndicate. Tranche maturity is the number of months until the facility
matures. Log facility amount and Log sales size represent the dollar amount of the loan facility and total sales of the borrowing rm, measured in U.S. $
million. Secured indicator is an indicator equal to one if the loan facility is secured. We control for tranche-purpose xed-effects (corporate purposes,
LBO, debt repayment, etc.), year xed-effects, and industry xed-effects (two-digit SIC code). We restrict our sample to include syndicated loan
facilities originated in the United States and exclude rms with SIC codes between 6000 and 6500. We also restrict our analysis to loan facilities with
senior debt rated Ba1, Ba2, Ba3, B1, B2, and B3. We identify and remove any second-lien or mezzanine facilities from the sample. Standard errors are
clustered by year and by borrowing rm. The symbols ***, **, and * indicate signicance at the 1%, 5%, and 10% levels, respectively.
Term Loan B and Term Loan A sample

(1)
Term Loan B and Revolver sample

(2)
Term Loan BnSecuritization-active banknB-rated sr. debt
30.606nn
(2.03)
20.761nn
(2.27)
Term Loan BnSecuritization-active bank
17.350
(1.14)
12.008
(0.95)
9.027
(0.44)
40.230nnn
(2.76)
13.681
(0.77)
23.407n
(1.70)
7.571
(1.31)
9.085
(0.92)
7.284n
(1.70)
52.713nnn
(11.41)
14.893nnn
(2.60)
33.749nnn
(4.18)
4.965
(0.40)
34.684nnn
(2.76)
14.973nn
(2.11)
10.560n
(1.65)
0.410nnn
(3.07)
1.252nnn
(3.16)
8.700nnn
(3.22)
4.587
(1.64)
4.440
(0.70)
548.103nnn
(5.12)
7.719
(1.13)
28.180nnn
(4.52)
12.736n
(1.89)
17.318nnn
(3.21)
0.536nnn
(3.91)

Year xed effects
Yes
Yes
Yes
Yes
Yes
Yes

Borrower
Year
Adjusted R2
Observations
Yes
Yes
0.408
1904
Yes
Yes
0.428
4238
Term Loan BnB-rated sr. debt

Term Loan B
B-rated sr. debt
Term structure
Credit spread
Covenants indicator
Size of syndicate
Tranche maturity
Log facility amount
Log sales size
Secured indicator
Constant
9.0 basis points higher than their non-securitization

active counterparts. Similarly, a B-rated, securitizationactive discount does not exist in the revolver sample.
Taken together, the results presented in Tables 5 and 6
indicate that the discount associated with securitizationfriendly attributes is concentrated in the Term Loan B
sample, consistent with a CLO-demand-driven hypothesis
20.518nnn
(7.67)
0.246
(0.10)
16.044nnn
(3.41)
538.217nnn
(12.37)
as opposed to an alternative increased risk tolerance

explanation.
4.6. Within-loan comparisons of facility spreads
Thus far our analysis has been conducted at the loanfacility level. In doing so, our goal was to measure the
incremental effect that demand for collateral from CLOs

has on the pricing of facilities, so we have controlled for
factors that affect spreads but are unrelated to demand for
collateral. The results suggest that, holding these other
factors constant, the effect of characteristics associated
with the likelihood of securitization, in particular being
issued by a securitization-active bank and having a B
rating, lowers the spread difference for a Term Loan B
facility relative to an otherwise similar Term Loan A or
revolving facility. We interpret this nding as consistent
with the demand for collateral explanation. However, the
econometric controls we use are imperfect and it is
possible that unobservable differences across types of
loans and not demand for collateral from CLOs could
explain this result.
Because of this issue, we use an approach introduced
by Ivashina and Sun (2011) and compare the spreads
across facilities in the same loan. Since the facilities are
from the same loan, they necessarily have the same risk.
Using this specication, the number of observations in the
sample declines, since not all loans containing Term Loan
B facilities also contain a Term Loan A or revolver. To
perform this comparison, we estimate equations that
predict the difference in spreads between Term Loan B
facilities and either Term Loan A facilities or revolvers
within the same loan as a function of factors likely to
affect the demand for collateral from CLOs.25 In doing so,
we also include into the equation other factors that
potentially affect the differences in spreads.
We present estimates of the within-loan spread differences in Table 7 for differences between Term Loan B and
Term Loan A facilities, and in Table 8 for differences
between Term Loan B and revolvers. In each case, the
coefcient on the securitization-active bank dummy variable interacted with the B-rated dummy variable is
negative and signicant (albeit at the 10% level for the
revolver comparison in Table 8). The magnitudes of the
securitization discount presented in Tables 7 and 8 are
roughly consistent with those presented in Tables 5 and 6.
The 47.8 estimate on the interaction term reported in
column 3 of Table 7 indicates that the difference in
spreads between Term Loan A and Term Loan B facilities
within the same deal is 47.8 basis points lower if the loan
is originated by securitization-active banks and has
B-rated senior debt, not far from the estimated 30.6
across-deal difference between securitization-friendly
Term Loan As and Term Loan Bs reported in Table 6. The
estimated within-deal coefcient of 14.2 for Term Loan
Bs compared to revolvers reported in column 3 of Table 8
is also fairly close to the across-deal estimate of 20.8
basis points reported in Table 6.
These ndings suggest that even comparing within
loans, facilities that are most likely to be securitized are
priced at a discount relative to other facilities. It does not
appear that the measured pricing discount of facilities
25
In the event where a bank loan includes multiple facilities of the
same type, we use the facility with the lowest spread, so as to be
conservative in our estimate of differences in spread. Of the 2,878 deals
in our sample, 52 deals have multiple Term Loan A or Term Loan B
facilities.
347
likely to be securitized occurs because of unobserved

heterogeneity; rather the evidence is consistent with the
view that issuing banks price these loan facilities more
aggressively because they can be easily sold to CLOs.
5. To what extent do the measured differences in
spreads come from securitization?
5.1. Are the differences in spreads driven solely by LBO
deals?
The securitization boom between 2002 and 2007
coincided with a high volume in the LBO market.
Shivdasani and Wang (2011) suggest that this pattern
was not a coincidence, and that structured nance was an
important determinant of the 20022007 LBO boom by
increasing the availability and pricing of leverage. These
authors present evidence of a pricing effect in aggregate
and in a smaller loan-level sample of Term Loan B
facilities supporting public targets of LBO transactions.
Since some of the loans in our sample are used to nance
LBOs, it is possible that our results could merely reect
the 20022007 LBO boom.
However, only 16% of the facilities in our Term Loan B
sample are used to provide nancing to LBOs; the vast
majority of them are used for general corporate purposes.
To evaluate whether our results are driven by the LBO
loan facilities in our sample, or whether the demand for
collateral affected the cost of capital more generally in
this period, we exclude from our sample of facilities those
used to nance LBOs and re-estimate the equations
reported above.26 The results are similar to those for the
full sample. Term Loan B facilities with securitizationfriendly attributes exhibit lower spreads than Term Loan
Bs without securitization-friendly attributes, and Term
Loan Bs exhibit lower spreads than Term Loan As and
revolvers with securitization-friendly attributes. Withindeal results are also quantitatively similar, although
differences between Term Loan Bs and revolvers within
the same deal are not statistically signicant at the
conventional levels when LBOs are excluded. Overall, it
appears that the pattern of securitization affecting the
cost of debt nancing affected corporate borrowing more
generally than just for buyouts during this period.
5.2. Do the results reect the inuence of CLOs, or of
institutional demand more broadly?
We have documented that the spreads on loan facilities that tend to be favored by CLOs are lower than
otherwise similar spreads on facilities that are less likely
to be held by CLOs. We have attributed this difference to
CLO demand; however, it is possible that this decrease
reects a more general demand for these facilities by
institutions.
Since we do not know the identity of every investor, it
is impossible to draw denitive conclusions about the
26
Estimates of the main specications from Tables 58 on the nonLBO subsample are available upon request.
348
Table 7
Within-deal differences in loan spreads: Term Bs vs. Term As.
In this table we estimate an OLS regression of facility attributes and other controls on differences in the all-in-drawn spread of Term Loan B facilities
compared to Term Loan A facilities within the same deal. The sample size of 186 represents the number of deals that have both Term Loan A and Term
Loan B facilities during our 2002September 2007 sample period as well as data on the required control variables. Securitization active bank is equal to
one if any of the lead arrangers of the loan facility are among the top-ten CLO underwriters (see Table 2). B-rated senior debt is equal to one if the senior
debt associated with the loan facility is rated B1, B2, or B3 at loan origination. Term structure is calculated as the difference between the 5-year and 3month T-bill rate in the month of deal origination. Credit spread is the spread of a 10-year Ba- and B-rated bond index over the ten-year T-bond rate,
respectively. Covenants indicator is equal to one if a facility has nancial covenants. Pricing grid indicator and Term A pricing grid indicator are equal to one
if the spread is adjustable based on accounting performance. Size of syndicate is the number of lenders involved in the lending syndicate. Tranche maturity
and Term A maturity represent the number of months until the Term Loan B and Term Loan A facilities mature, respectively. Log facility amount and Term A
facility amount represent the dollar amount ($ million) of the Term B and Term A loan facilities, respectively. Log sales size represents the sales of the
borrowing rm, measured in U.S. $ million. Secured indicator is an indicator equal to one if the loan is secured. We control for tranche-purpose xedeffects (corporate purposes, LBO, debt repayment, etc.), year xed-effects, and industry xed-effects in all specications and credit-rating xedeffects in column 1. We restrict our sample to include syndicated loan facilities originated in the United States and exclude rms with SIC codes between
6000 and 6500. We also restrict our analysis to loan facilities with senior debt rated Ba1, Ba2, Ba3, B1, B2, and B3. We identify and remove any secondlien or mezzanine facilities from the sample. Standard errors are clustered by year and by borrowing rm. The symbols ***, **, and * indicate signicance
at the 1%, 5%, and 10% levels, respectively.
Dependent var.: Within
deal spread
Term Loan B spread vs. Term Loan A

spread in the same deal
(1)
Securitization-active
banknB-rated sr. debt

(2)

(3)
7.996
(0.43)
4.146
(0.20)
6.160
(0.40)
47.764nnn
(2.82)
23.347
(0.69)
44.489nnn
(2.83)
0.201
(0.03)
3.371
(0.25)
1.096
(0.04)
5.602
(0.42)
7.710
0.879
(0.21)
0.451
(0.04)
0.589
(0.02)
5.732
(0.42)
6.054
1.296
(0.25)
1.741
(0.13)
3.091
(0.08)
5.220
(0.40)
5.214
(0.39)
0.081
(0.77)
0.649
(0.62)
1.065
(0.92)
1.624
(0.46)
0.178
(0.06)
3.941
(0.41)
3.380
(0.47)
21.431
(0.10)
(0.29)
0.156
(1.22)
0.915
(0.91)
1.245
(1.17)
5.669
(1.19)
0.995
(0.55)
4.741
(0.46)
0.212
(0.03)
82.070
(0.38)
(0.23)
0.099
(0.97)
0.795
(0.89)
1.111
(1.19)
5.170
(0.87)
0.822
(0.84)
4.533
(0.52)
5.637
(0.57)
59.662
(0.29)
Yes
No
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
0.513
186
Yes
Yes
0.488
186
Yes
Yes
0.502
186
B-rated sr. debt

Term structure
Credit spread
Covenants indicator
Term A pricing grid
indicator
Size of syndicate
Tranche maturity
Term A maturity
Log facility amount
Term A facility amount
Log sales size
Secured indicator
Constant
Moodys sr. debt xed
effects
Tranche-purpose xed
effects
Year xed effects
Borrower
Year
Adjusted R2
Observations
349
Table 8
Within-deal differences in loan spreads: Term Bs vs. revolvers.
In this table we estimate an OLS regression of facility attributes and other controls on differences in the all-in-drawn spread of Term Loan B facilities
compared to revolving facilities within the same deal. The sample size of 1,150 represents the number of deals that have both Revolvers and Term Loan B
facilities during our 2002September 2007 sample period as well as data on the required control variables. Securitization active bank is equal to one if any
of the lead arrangers of the loan facility are among the top-ten CLO underwriters (see Table 2). B-rated senior debt is equal to one if the senior debt
associated with the loan facility is rated B1, B2, or B3 at loan origination. Term structure is calculated as the difference between the 5-year and 3-month
T-bill rate in the month of deal origination. Credit spread is the spread of a ten-year Ba- and B-rated bond index over the ten-year T-bond rate,
respectively. Covenants indicator is equal to one if a facility has nancial covenants. Pricing grid indicator and Revolver pricing grid indicator are equal to one
if the spread is adjustable based on accounting performance. Size of syndicate is the number of lenders involved in the lending syndicate. Tranche maturity
and Revolver maturity represent the number of months until the Term Loan B and revolving facilities mature, respectively. Log facility amount and Revolver
facility amount represent the dollar amount ($ million) of the Term B and revolving loan facilities, respectively. Log sales size represents the sales of the
borrowing rm, measured in U.S. $ million. Secured indicator is an indicator equal to one if the loan facility is secured. We control for tranche-purpose
xed-effects (corporate purposes, LBO, debt repayment, etc.), year xed-effects, and industry xed-effects (two-digit SIC code) in all specications
and credit-rating xed-effects in column 1. We restrict our sample to include syndicated loan facilities originated in the United States and exclude rms
with SIC codes between 6000 and 6500. We also restrict our analysis to loan facilities with senior debt rated Ba1, Ba2, Ba3, B1, B2, and B3. We identify
and remove any second-lien or mezzanine facilities from the sample. Standard errors are clustered by year and by borrowing rm. The symbols ***, **,
and * indicate signicance at the 1%, 5%, and 10% levels, respectively.
Dependent var.: Within deal spread
Term Loan B spread vs. Revolver

(1)

(2)

(3)
6.887nn
(2.55)
6.798nn
(2.29)
8.014
(1.50)
14.171n
(1.90)
2.971
(0.64)
2.373
(0.47)
3.823
(0.68)
13.278nn
(2.08)
4.358
(0.90)
6.076nn
(2.10)
1.180
(0.34)
0.028
(0.14)
0.115
(0.75)
0.254
(1.46)
9.286nnn
(2.92)
7.627nnn
(3.69)
1.771
(1.51)
2.798
(0.39)
19.993
(0.38)
4.369
(0.79)
12.087n
(1.79)
5.040
(0.99)
6.106nn
(2.06)
3.138
(0.75)
0.030
(0.15)
0.190
(1.14)
0.288
(1.56)
10.289nnn
(2.93)
8.770nnn
(4.46)
1.300
(1.18)
0.990
(0.13)
41.140
(1.08)
4.406
(0.80)
12.084n
(1.76)
4.587
(0.93)
5.991nn
(2.02)
2.624
(0.65)
0.034
(0.17)
0.174
(1.10)
0.282
(1.54)
10.292nnn
(2.94)
8.763nnn
(4.50)
1.326
(1.20)
0.649
(0.09)
27.786
(0.76)

Year xed effects
Yes
Yes
Yes
Yes
No
Yes
Yes
Yes
No
Yes
Yes
Yes

Borrower
Year
Adjusted R2
Observations
Yes
Yes
0.130
1150
Yes
Yes
0.111
1150
Yes
Yes
0.113
1150

B-rated sr. debt
Term structure
Credit spread
Covenants indicator
Revolver pricing grid indicator
Size of syndicate
Tranche maturity
Revolver maturity
Log facility amount
Revolver facility amount
Log sales size
Secured indicator
Constant
sources of the demand increase for these facilities.

However, the fact that the difference in spreads is largest
among B-rated loans is strongly suggestive of the fact that
CLOs are an important driver of these differences for two
reasons. First, as argued by Coval, Jurek, and Stafford
(2009), there are reasons to suspect CLOs have a preference for lower-rated debt, since one reason for their
existence is to arbitrage ratings. Consistent with this argument, Benmelech and Dlugosz (2009) show that CLO
collateral is indeed more highly concentrated in B-rated
350
Table 9
Differences in CLO ownership by credit rating.
This table reports data on CLO ownership of securitized Term Loan B facilities. A portion of the same loan facility can be owned by multiple CLOs. Panel
A reports the sum total of CLO ownership as a percent of the total loan facility, sorted by Moodys senior debt credit rating at the time of loan origination.
Panel B reports the average CLO ownership percentage in a given facility across all CLOs by credit rating. The upper portion of Panel C calculates the
difference in total CLO ownership percentages between facilities with B-rated senior debt and facilities with Ba-rated senior debt. The lower portion of
Panel C calculates the difference in total CLO ownership percentages between facilities with B2- and B3-rated senior debt compared to Ba1- and Ba2rated senior debt. The upper portion of Panel D calculates the difference in average ownership percentages between facilities with B-rated senior debt
compared to facilities with Ba-rated senior debt. The lower portion of Panel D calculates the difference in average ownership percentages between
facilities with B2- and B3-rated senior debt compared to facilities with Ba1- and Ba2-rated senior debt. The ownership percentages reported in Moodys
EMS database are as of the rst quarter 2009.
Panel A: Total CLO ownership % of Term Loan Bs
Panel B: Average CLO ownership % of Term Loan Bs
Moodys senior
debt rating
Mean
Std.
dev.
Ba1
Ba2
Ba3
B1
B2
B3
10
24
52
93
61
28
3.08%
9.94%
17.34%
14.21%
15.49%
15.83%
6.98%
11.88%
23.01%
20.74%
20.35%
26.93%
Avg. fac. amt. Moodys senior debt

($M)
rating
744.4
693.5
667.2
409.1
569.1
766.4
Panel C: Differences in total CLO ownership % of Term Loan Bs
Mean
N
Mean
N
Ba1, Ba2, Ba3

rated
Diff.
t-Stat.
14.89%
182
13.62%
86
1.27%
(0.48)
Diff.
t-Stat.
7.67%
(2.52)
15.59%
89
7.92%
34
Mean
Std.
dev
Avg. fac. amt.

($ M)
10
25
53
95
63
28
0.29%
0.53%
0.63%
0.99%
0.80%
0.50%
0.22%
0.73%
1.02%
2.28%
0.98%
0.48%
744.4
717.2
654.9
402.1
563.4
766.4
Panel D: Differences in average CLO ownership % of Term Loan Bs
B1, B2, B3
rated
B2, B3 rated Ba1, Ba2 rated
Ba1
Ba2
Ba3
B1
B2
B3
debt than Ba-rated debt. In contrast, there is no reason to

suspect that other institutional investors have a preference
for B-rated debt, and many have rating requirements that
lead to a preference for higher-rated debt.
Second, using the Moodys data, we can calculate the
quantities of debt of alternative ratings classes held by
CLOs, and it appears that they hold higher quantities of Brated debt than of higher-rated debt. Table 9 reports
summary statistics on CLO ownership of Term Loan B
facilities. Panel A reports total CLO ownership percentages
of Term Loan B facilities by ratings class. Panel B reports
average CLO ownership percentages by ratings class.
Panels A and B demonstrate that total ownership percentages are larger, on average, for facilities associated with
B-rated senior debt compared to facilities associated with
Ba-rated senior debt. The same pattern is true for average
ownership percentages; on average, CLOs exhibit larger
average ownership fractions in lower-rated debt.
In Panel C, we test whether the difference in total
ownership percentages is statistically different between
the B-rating and Ba-ratings classes, on average. The upper
portion of Panel C shows that the difference in total
ownership is a statistically insignicant 1.27%, because
total ownership in facilities associated with B1-rated
senior debt is similar to total ownership in the Ba3 ratings
category. The lower portion of Panel C compares total
ownership in the B2 and B3 ratings categories against
total ownership in Ba1 and Ba2 ratings categories, omitting B1 and Ba3 ratings to generate a larger difference in
credit quality. The difference in total ownership across
Mean
N
B1, B2, B3
rated
Ba1, Ba2, Ba3

rated
Diff.
t-Stat.
0.85%
186
0.56%
88
0.29%
(1.84)
Diff.
t-Stat.
0.25%
(1.76)
B2, B3 rated Ba1, Ba2 rated

Mean
N
0.71%
91
0.46%
35
these categories demonstrates that the lower-rated B1

and B2 ratings hold a statistically signicant 7.67% larger
total ownership percentage compared with the Ba1 and
Ba2 ratings categories. Panel D performs similar tests with
average ownership percentages and nds similar results.
While we cannot conclusively say that institutional
demand more broadly than CLOs has some impact on
the differences in spreads documented above, these
calculations suggest that CLO demand is an important
part of the story.
6. Conclusion
Understanding the debt buildup from 2002 through
2007 followed by the Financial Crisis of 2008 is an
important area of research for nancial economists. A
key element of both the debt buildup and the Crisis is
often thought to be the increase in securitization of loans
over this period. By packaging loan facilities together in
CLOs, the market created a highly rated security as a
combination of low-rated (usually B-rated) ones. While
the volumes of both loans and CLOs both increased
substantially during the securitization-active period, the
link between these phenomena is unclear. It is possible
that this high volume of debt issuances could reect a
high demand for capital by rms. Alternatively, as has
been alleged by theories of the Financial Crisis (see
Gennaioli, Shleifer, and Vishny, in press), the demand
from CLOs for securitizable loan facilities could have led
banks to make loans they otherwise would not have
made. This paper evaluates the pricing implications of this

hypothesis, and nds support for the view that demand
for securitizable loans lowers the cost of capital for rms.
Controlling for other factors, including the risk of the
loan facility, facilities that are securitized are issued at
about a 17 basis point lower spread than an otherwise
identical facility that was not securitized. This nding is
consistent with the notion that the demand for securitizable loan facilities led banks wishing to securitize their
facilities to discount them to attract more borrowers.
However, there are other interpretations to this nding;
it is possible that rms demand for capital is such that
rms borrowing patterns led loan facilities that get
securitized to have lower spreads.
As a way of evaluating explanations based on demand for
capital, we consider how the relative spreads between loan
facilities with characteristics that are considered to be
securitization-friendly and loan facilities with characteristics considered to be securitization-unfriendly are different
during the Securitization Boom. In particular, we consider
the payoff structure of the debt, with Term Loan B facilities
whose bullet payoff structures are considered more securitization-friendly than amortizing Term Loan A facilities or
revolvers, whose payoff and amount borrowed is at the
discretion of the borrower. Second, we consider the issuer
of the loan facility, with banks who are top arrangers of CLOs
to be more likely to securitize their own facilities than
facilities from other banks. Finally, we consider the rating
of the facility, with evidence that loan facilities made to Brated borrowers are more likely to be securitized than
facilities with other ratings.
Our ndings suggest that spreads on facilities with
characteristics making the facility desirable to a CLO
decline relative to other facilities with less securitization-friendly characteristics during the securitization
boom. The spread on Term Loan B facilities issued by
securitization-active banks to borrowers with B-rated
senior debt dropped 20.9 basis points relative to Term
Loan B facilities with less securitization-friendly attributes from 2002 through the fall of 2007. Consistent with
the CLO demand hypothesis, we provide evidence that the
decline in spreads is concentrated in Term Loan B facilities
with securitization-friendly characteristics rather than
Term Loan A or revolving facilities with otherwise similar
securitization-friendly characteristics. This result suggests
that the decline in spreads cannot be attributed to a factor
associated with the particular banks originating the CLOs,
such as an increase in their risk tolerance. This general
pattern of a decline in securitization-friendly facilities
relative to non-securitization-friendly facilities over the
20022007 period is true even for facilities that are
part of the same loan package, and is true for a sample
excluding LBO loans.
Taken together, we interpret the evidence as being
consistent with the hypothesis that securitization had a
causal impact on the cost of corporate capital. Were an
omitted variable to be driving our results, it would have to
be uniquely correlated with Term Loan Bs originated only
by securitization-active banks to borrowers with B-rated
senior debt but uncorrelated with Term Loan As and
revolving facilities with the same characteristics. While
351
overall institutional demand is one such potential variable, the concentration of spread differences in B-rated
rather than Ba-rated debt suggests that securitization
plays an important role.
The results in this paper raise a number of questions as
well. For example, what was the cause of increased
demand for securitized debt? In this paper, we take
increased CLO activity as given, yet we know little about
the ultimate drivers of nal demand for bonds collateralized by corporate debt. Second, does the securitization
process, with its complicated cash-ow rules and multiple
market participants, lead to information destruction
resulting in lenders not charging spreads commensurate
with risks? Fundamentally, the question posed by the
theoretical literature on securitization remains, namely,
how does the restructuring of cash ows from a pool of
assets into a securitization structure create real value?
Finally, while we do provide evidence that the advent of
CLOs led to cheaper capital for corporate borrowers, we
do not provide evidence suggesting that CLOs led to
inefcient lending outcomes. Whether demand from CLOs
caused banks to lend too much, or whether cheaper, CLOfueled capital led to overinvestment in the real economy,
remain important and unanswered questions.
References
Ai, C., Norton, E., 2003. Interaction terms in logit and probit models.
Economic Letters 80, 123129.
Angrist, J., Pischke, J., 2009. Mostly Harmless Econometrics: An Empiricists Companion. University Press, Princeton, NJ.
Axelson, U., Jenkinson, T., Stromberg, P., Weisbach, M., 2011. Borrow
cheap, buy high? The determinants of leverage and pricing
in buyouts. Unpublished. London School of Economics, University
of Oxford, Stockholm School of Economics, and Ohio State
University.
Benmelech, E., Dlugosz, J., 2009. The alchemy of CDO credit ratings.
Journal of Monetary Economics, Carnegie-Rochester Conference
Volume 56, 617634.
Benmelech, E., Dlugosz, J., Ivashina, V., Securitization without adverse
selection: the case of CLOs. Journal of Financial Economics,
forthcoming.
Coval, J., Jurek, J., Stafford, E., 2009. The economics of structured nance.
Journal of Economic Perspectives 23, 325.
DellAriccia, G., Igan, D., Laeven, L., 2008. Credit booms and lending
standards: evidence from the subprime mortgage market. IMF
unpublished.
DeMarzo, P., 2005. The pooling and tranching of securities: a model of
informed intermediation. Review of Financial Studies 18, 135.
Demyanyk, Y., Van Hemert, O., 2011. Understanding the subprime
mortgage crisis. Review of Financial Studies 24, 18481880.
Gaur, V., Seshadri, S., Subrahmanyam, M., 2011. Securitization and real
investment in incomplete markets. Management Science 57,
21802196.
Gennaioli, N., Shleifer, A., Vishny, R., Neglected risks, nancial innovation
and nancial fragility. Journal of Financial Economics, http://dx.doi.
org/10.1016/j.jneco.2011.05.005, in press.
Guner, B., 2006. Loan sales and the cost of corporate borrowing. Review
of Financial Studies 19, 687715.
Ivashina, V., 2009. Asymmetric information effects on loan spreads.
Journal of Financial Economics 92, 300319.
Ivashina, V., Sun, Z., 2011. Institutional demand pressure and the cost of
corporate loans. Journal of Financial Economics 99, 500522.
Kara, A., Marques-Ibanez, D., Ongena, S., 2011. Securitization and lending
standards: evidence from the wholesale loan market. Unpublished,
Loughborough University, European Central Bank, and Tilburg
University.
Keys, B., Mukherjee, T., Seru, A., Vig, V., 2010. Did securitization screening lead to lax? Evidence from subprime loans. Quarterly Journal of
Economics 125, 307362.
352
Loutskina, E., Strahan, P., 2009. Securitization and the declining impact
of bank nancial condition on loan supply: evidence from mortgage
originations. Journal of Finance 64, 861922.
Mian, A., Su, A., 2009. The consequences of mortgage credit expansion:
evidence from the U.S. mortgage default crisis. Quarterly Journal of
Economics 124, 14491496.
Nadauld, T., Sherlund, S., The impact of securitization on the expansion
of subprime credit. Journal of Financial Economics, forthcoming.
Nini, G., 2008. How non-banks increased the supply of bank loans: evidence
from institutional term loans. Unpublished, Wharton Business School.
Shivdasani, A., Wang, Y., 2011. Did structured credit fuel the LBO boom?
Journal of Finance 66, 12911328.
Standard & Poors, 2007. A Guide to the Loan Market. Standard & Poors,
New York.
Su, A., 2007. Information asymmetry and nancing arrangements:
evidence from syndicated loans. Journal of Finance 62, 629668.

Journal of Financial Economics: Taylor D. Nadauld, Michael S. Weisbach

Uploaded by

Copyright:

Available Formats

Journal of Financial Economics: Taylor D. Nadauld, Michael S. Weisbach

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Journal of Financial Economics: Taylor D. Nadauld, Michael S. Weisbach

Uploaded by

Copyright:

Available Formats

Journal of Financial Economics 105 (2012) 332352

Contents lists available at SciVerse ScienceDirect

Journal of Financial Economics

Did securitization affect the cost of corporate debt?$

Department of Finance, Brigham Young University, Provo, Utah, USA

2002 to over $540 billion of loans and 213 CLOs in 2007.2

any Moodys-rated CLO as of the rst quarter 2009, which

Second, the credit rating of a borrower affects the

We discuss more specically why this is the case in Section 2.

these securitization-friendly attributes relative to Term

(2008) examines the investment behavior of rms more

2.1. Lender demand for CLO collateral and loan spreads

on the entire amount committed to the lending facility

provides only a one-time snapshot of securitized facilities,

All-in-drawn spread over LIBOR

Term Loan Bs only - securitized sample

All-in-drawn spread over LIBOR

Term Loan Bs onlysecuritized sample

Facility amount ($ million)

Term Loan Bs Onlysecuritized sample

facilities are the most common facility, followed by Term

different credit qualities. As expected, facility spreads

For each loan facility in our sample, we identify the

No. of Term B facilities

No. of Term B facilities

Fig. 1. Plots of Term Loan B origination volume.

sales, size of the lending syndicate, tranche maturity,

j Macro f actorst ei,t ,

Term B only, Ba and B-rated sr. debt

Term B only, Ba-rated sr. debt

Term B only, B-rated sr. debt

Tranche-type xed effects

Standard errors clustered by:

its very nature leads to a preference for certain types of

statistics of our sample of securitized facilities (see

4.4. The impact of CLO demand on loan pricing: differences

Dependent var. 1 if loan is identied as having been

Term B, Term A, and

Estimated using OLS (Mean dep.

Term B, Term A, and

Term B, Term A, and Revolvers

Term Loan BnB-rated senior debt

Term Loan BnSecuritization-active bank

B-rated senior debt

Pricing grid indicator

We test whether increased demand by CLOs lowered

where the sample is constrained to Term Loan Bs only.

equation, represented by X i,t, are the term structure and

variables plus the interaction term. The estimate of the

Term Loan B sample

Term Loan B sample

Term Loan B sample

Moodys sr. debt xed effects

Standard errors clustered by

Securitization-active banknB-rated sr. debt

impact of being active in securitization is concentrated in

Term Loan A facilities are normally amortizing and Term

Term Loan B and Term Loan A sample