Corporate Syndicated Loans as a
Source of Private Information for Interbank Markets
Stijn Ferrari, Issam Hallak, Thomas Matthys, and Rudi Vander Vennet
Non-Technical Summary
Banks lend to each other in interbank markets, but at the same time collaborate in lending
activities to non-financial firms through syndicated loans.1 So, to quote Nobel laureate
Eugene Fama, banks are “different” indeed.2 The question we raise is whether a bank would
use the private information about other banks gained in syndicated loan markets, when it
comes to interbank lending. To our knowledge this study is the first that highlights a source
of private information in interbank markets, thereby validating the assumption that observed
interbank relationship lending is informationally driven. Understanding whether and how
private information is channeled into interbank markets is essential to understanding its
functioning, and furthering financial stability.
Indeed, the 2008 financial crisis due to Lehman Brothers’ collapse proved once again the
vital function of the interbank market in the financial system. Among others, it became
evident that the lack of information about the value of other banks’ assets was a major
determinant of the sudden market disruption. Nevertheless, there is mounting empirical
evidence suggesting that not all lending relationships vanished, and that private information
plays a role in interbank lending – and did so even at the peak of the crisis. Yet, no sources
of private information have been highlighted so far, thus questioning the existence of such
informationally driven interbank relationship lending.
In this paper we investigate one source of private information. We believe that if a bank
participates to a larger share of corporate syndicated loans of another bank, then the former
holds further private information about the latter’s corporate loans. In fact, medium and
large corporate loans are subject to substantial private information so that valuable pieces
of information are observable to syndicated loan members only. As a result we expect that
1
Syndicated loans are loans funded by at least two banks. They are particularly aimed at medium and
large companies.
2
What’s different about banks? Eugene F. Fama. Journal of Monetary Economics, 1985, 15(1).
the first bank would offer different loan terms to the second bank on the interbank market,
in terms of volumes and pricing. This is consistent with the well-documented “relationship
lending” phenomenon in corporate loan markets which occurs whenever banks hold privileged
information about their customers.
We use a proprietary transaction-level data set that contains all transfers sent and received by institutions through TARGET2, between June 2008 and December 2013. The
database includes daily information on bilateral loans between TARGET2 members. Similar to previous studies we make estimates of interbank loans pricing and volumes, and include
measures of concentration of interbank relationships. We complement the interbank dataset
with information on non-financial corporate syndicated loans that we obtain from DealScan.
We construct a measure of concentration of syndicated loans sharing between two banks
which captures the degree of private information banks hold about each other.
We find that an interbank borrower whose corporate syndicated loans are further shared
with an interbank lender receive more favorable loan terms from the latter, both in terms of
pricing and volumes. Therefore, interbank lenders have an informational advantage about
interbank borrowers should both share corporate lending activities. The results thus provides
innovative evidence that interconnectedness in interbank markets is also driven by private
information.
The results bear significant implications. Among others, they show that banks interconnectedness through shared asset exposures does not necessarily pose a threat to the financial
stability, because the sharing of assets provides with valuable private information about
other banks assets; this information can be transferred to and used in the interbank market,
thereby contributing to liquidity allocation efficiency among banks and the stability of the
financial system. The work is very innovative as it highlights a potential link between asset
sharing and systemic risk mitigation.
Keywords: Interbank Market, Relationship Lending, Private Information, Financial Stability.
Corporate Syndicated Loans as a
Source of Private Information for Interbank Markets
Stijn Ferrari, Issam Hallak, Thomas Matthys, and Rudi Vander Vennet‡
PRELIMINARY AND INCOMPLETE
PLEASE DO NOT QUOTE
December 2015
Abstract
Banks are a special type of company: they contemporaneously lend to each other
(interbank loan) and “co-lend” to large corporations through syndicated loan. A syndicated loan is a loan funded by at least two banks. Recently authors showed evidence of
the existence in interbank loan markets of lending patterns that are consistent with relationship lending (e.g., Cocco, Gomes and Martins, 2009; Afonso, Kovner and Schoar,
2013). A relationship lender is a lender who holds private information about her customer. Yet, no sources of private information have been identified. We show that a
lending bank extracts private information about the value of a borrowing banks’ assets
by sharing its corporate syndicated loans. Indeed, loans to large corporations typically
bear substantial amounts of private information. We use a large proprietary dataset of
the euro area interbank market, and find that interbank borrowers who share a larger
proportion of their syndicated loans portfolio with an interbank lender obtain better
loan terms and larger amounts. In our analysis we control for relationships on the
interbank market and bank characteristics.
‡
Stijn Ferrari is Economist at the National Bank of Belgium. Issam Hallak is Economist at the European
Commission Joint Research Center, and affiliated with KU Leuven. Thomas Matthys is Researcher at Ghent
Univerity and Vlerick Business School. Rudi Vander Vennet is Professor at Ghent University. Corresponding
author: Thomas Matthys, thomas.matthys(at)vlerick.com. We are grateful to the Flemish Government
Academic Research Fund for financial support. The views expressed in this paper neither reflect those of
the National Bank of Belgium, nor those of the European Commission.
Corporate Syndicated Loans as a
Source of Private Information for Interbank Markets
Abstract
Banks are a special type of company: they contemporaneously lend to each other
(interbank loan) and “co-lend” to large corporations through syndicated loan. A syndicated loan is a loan funded by at least two banks. Recently authors showed evidence of
the existence in interbank loan markets of lending patterns that are consistent with relationship lending (e.g., Cocco, Gomes and Martins, 2009; Afonso, Kovner and Schoar,
2013). A relationship lender is a lender who holds private information about her customer. Yet, no sources of private information have been identified. We show that a
lending bank extracts private information about the value of a borrowing banks’ assets
by sharing its corporate syndicated loans. Indeed, loans to large corporations typically
bear substantial amounts of private information. We use a large proprietary dataset of
the euro area interbank market, and find that interbank borrowers who share a larger
proportion of their syndicated loans portfolio with an interbank lender obtain better
loan terms and larger amounts. In our analysis we control for relationships on the
interbank market and bank characteristics.
Keywords: Interbank Lending, Relationship Lending, Private Information, Financial
Stability.
JEL Classification: G21.
1
1
Introduction
The 2008 financial crisis proved once again the vital function of the interbank market in
the financial system and the need for understanding its mechanisms. Following the Lehman
Brothers collapse and the resulting interbank market disruptions, banks were unable to
refinance their liquidity needs until confidence was restored thanks to the intervention of the
official authorities. Being a key tool for the liquidity management of banks, the interbank
market is a major factor of interconnectedness and sytemic risk of the financial sector; a well
functioning interbank market reflects the overall health of the financial system (e.g., Allen
and Gale, 2000; Ho and Saunders, 1985). Information plays a major role in the functioning
of the interbank market. In fact, the main determinant of the breakdown in 2008 was the
presence of informational asymmetries about the value of borrower’s assets (Afonso, Kovner,
and Schoar, 2011).
There is growing empirical evidence of the existence of relationship lending patterns in
interbank markets (e.g., Cocco et al., 2009; Affinito, 2012; and Afonso, Kovner, and Schoar,
2013; Braüning and Fecht, 2013). Relationship lenders are lenders characterized by the
holding of information non-observable to outsiders, so-called private information. There is
extensive knowledge of relationship lending in corporate debt markets. Relationship lenders
are banks which produce private information about borrowers thanks to a close contact
with borrowing firms, as well as a longer and repeated lending relationship (e.g., Diamond,
1984; Fama, 1985; James, 1987, Boot, 2000). Typically, relationship lenders provide their
customers with larger funding amounts and better borrowing terms, and support their cutomers in bad times (Petersen and Rajan, 1995; Sette and Gobbi, 2015). The behaviour of
relationship lenders is justified by the soft information they hold but also their attempt to
keep customers for which they hold an informational advantage.
While relationship lending patterns are found in the interbank market, it remains unclear
how private information enters interbank markets. Producing private information requires
time while interbank loans typically are overnight loans that are swiftly agreed on a daily
basis. In fact, corporate loan duration on average is in years so that creditors have sufficient
time to observe the borrower and decide whether or not to repeat lending. Further, banks
conduct time-consuming due diligence about loan applicants whereby they gather and analyze relevant information. Last, the extremely short life of interbank loans prevents any
kind of renegotiations, while renegotiation is a major source of private information (Boot,
2
2000). Thus, the characteristics of interbank loans hardly suit the private information production process. Already Cocco et al. (2009) argue that it is conceivable that banks collect
information about interbank market counterparties outside the interbank market since banks
undertake many kinds of transactions together.
This paper aims to highlight one channel through which private information enters the
interbank market that would validate the existence of relationship lending in this market.
For this we use a peculiar feature of banks: banks lend to each other in interbank markets
and at the same time co-lend to non-financial firms through syndicated loans. A syndicated
loan is a loan funded by at least two banks. The syndicated loan market has evolved over
the last decades into a key lending vehicle through which banks lend to large corporations
(e.g., Ivashina and Scharfstein, 2010). The syndicated loans thus constitute a large share of
banks assets. In our sample, we find that the total amount of syndicated loans represents
nearly 20% of loans contracted by the top 50 most active banks in this market (43% of top
10 lenders), and 10% of total assets (16% for top 10 banks). About 42% of top 10 banks
corporate loans are syndicated loans. Importantly, large corporate loans are shown to be
amongst banking assets bearing largest amounts of private information (Lee and Mullineaux,
2004). As a result, the more a bank participates to another bank’s syndicated loans, the
more the former holds private information about the latter’s assets. We show that banks
build their interbank lending decisions upon this private information similarly to relationship
lenders in corporate debt markets.
Our analysis is based on the TARGET 2 data set of interbank loans. Nearly all large
interbank loans in euro are processed through TARGET 2. We construct lending amounts
and interest rates of interbank loans following the algorithms described by Furfine (2001) and
Arciero et al. (2013). Then, consistently with Cocco et al. (2009) we construct the monthly
Borrowing Preference Index for each pair of banks as well as the Lending Preference Index.
The borrowing preference index of a bank, say Bank 1, with respect to another bank, say
Bank 2, is the ratio of the amount of loans borrowed by Bank 1 from Bank 2 over the total
amount of loans borrowed by Bank 1. The lending preference index is constructed in a
similar fashon using lent instead of borrowed amounts.
We construct a similar pairwise Syndicated Loans Share index in the syndicated loan
market. The syndicated loans share of a bank, say Bank 1, with respect to another bank,
say Bank 2, is the ratio of loan amounts to which Bank 1 and Bank 2 participated divided
3
by the total amount of loans to which Bank 2 participated. This measures the share of Bank
2’s corporate syndicated loans for which Bank 1 holds the same private information as Bank
2. It thus reflects the amount of private information the interbank lending bank holds about
the value of corporate loans of the interbank borrowing bank. Holding private information
about a borrowing bank provides the lending bank informational advantages akin to those
of relationship lenders highlighted in corporate loan markets.
The main result is that the larger is the share of syndicated loans a lending bank share
with a borrowing bank, the lower is the interest rate. In the full sample and using Afonso et
al. (2011) model, we find that a unit of standard deviation in the share contributes to 15% of
the standard deviation of the interest rate of the loan. The impact is robust to a various set
of models that includes bank characteristics. We also look at borrowing amounts. A bank
is more likely to lend to another bank as the former (lender) participated to a larger share
of syndicated loans of the latter (borrower). The impact on the borrowed amounts is robust
to a various set of model estimates. By looking at the impact of banks sharing syndicated
loans on interbank relationship lending, we provide innovative evidence that banks extract
private information about each other in markets they collaborate, which they subsequently
use when dealing in the interbank market. We also validate the existence of information
based interbank relationship lending.
Our paper primarily contributes to the growing literature on informational frictions and
relationship lending in the interbank market. Among others Cocco et al. (2009) and Afonso
et al. (2014) showed the existence of relationship lending patterns in the interbank market
without analyzing how banks were enabled to produce private information about each other.
By highlighting one source of private information in this market we validate the existence of
information based relationship lending. We show that banks “co-lending” on the syndicated
loan market reduces information asymmetry between banks and explains the relationship
lending patterns observed in previous studies.
Our paper also contributes to the literature on financial stability. Recently authors
emphasized the systemic risks associated with the syndicated loan markets (e.g., Cai et al.,
2014; Nirei, Caballero, and Sushko, 2015). By holding common exposure on the same asset
two banks are interconnected and general failure of a sector may lead to a contagious effect
on the whole financial system. We add to this literature by showing that sharing the same
asset enables banks to produce additional valuable information about other banks which is
4
beneficial to the stability of the interbank market. We thus add to the growing literature on
syndicated interconnectedness by showing that by working closely with one another, banks
transfer information from the syndicated loan market to the interbank market.
Last, our paper contributes to the literature on syndicated loans. Previous research
focused on the relationship between borrowers and syndicate members. Emphasis has been
put on the role of lead banks and the ability of banks to keep informational advantages over
long-arm’s lending instruments (Lee and Mullineaux, 2004; Sufi, 2007; Bharath et al., 2011).
We took an innovative perspective of syndicated loans. By showing that syndicated loans
promote the efficient allocation of funds between banks, we provide evidence of their value
in the financial system. In fact syndicated loans imply that banks share assssssets with large
amounts of private information and therefore allow banks to extract information about each
other that they use in interbank lending activities. The size of the syndicated loan market
has boomed in the last decades showing further benefits.
2
Data and Empirical Strategy
2.1
Data
In order to analyse the impact of syndicated loans sharing on interbank lending, we combine
the information from two different data sets. Interbank market data is obtained from a
proprietary transaction-level data set that contains all transfers sent and received by institutions through TARGET2, covering the period June 2008 to December 2013. The TARGET2
system (Trans-European Automated Real-Time Gross settlement Express Transfer system)
is the main large value payment system of the Eurosystem1 . The participants in this system
are predominantly euro area financial institutions and several large non-euro area banks.
From this data set we identify unsecured overnight loans using a computer algorithm that
builds on the algorithm proposed by Furfine (1999). In its original version Furfine (1999)
identifies an interbank loan as a transaction between two banks that has a value greater
than $1 million, with a payment on the following business day in the opposite direction that
has a value that can reasonably be assumed to be the initial transaction plus interest. As
interest rates vary over transactions, a plausibility corridor is set from 50 basis points below
the minimum to 50 basis points above the maximum of a day’s federal funds rate. Recently,
1
In 2012 TARGET2 settled 92% of the total large value payments traffic in euro.
5
Arciero et al. (2013) have considerably refined the Furfine algorithm to be applied to TARGET2, and are able to identify interbank loans with reduced uncertainty. The algorithm is
set to identify loans with a minimum value of e1 million, with variable increments depending on the loan size. The plausible interest rate corridor is set around EONIA of 25 basis
points. We use the Arciero algorithm to identify interbank loans for two reasons. First, the
algorithm includes a procedure to efficiently select the correct loan in case of multiple plausible matches, significantly reducing the probability of error. Second, contrary to previous
extensions of the Furfine algorithm, the data set obtained using the Arciero algorithm has
been comprehensively validated against real data.
Our interbank loan data set includes date, amount, interest rate, as well as the unique
identity of each institution. The identities of the institutions are aggregated to the parent
level, and we drop transactions between institutions belonging to the same parent (no liquidity transfers within a group). The data set thus collected contains transactions between
794 financial institutions from 52 countries.
The second data set is a sample of syndicated loans obtained from Loan Pricing Corporation’s DealScan, which contains detailed information on syndicated loan contract terms and
syndicate members. The final sample includes 139,915 syndicated loan tranches to 46,288
firms issued between January 1987 and January 2014. The DealScan database contains
224,796 syndicated loan tranches to firms for these years. We exclude tranches without information on lead arranger (4,155) or tranche amount (858). Further, we exclude tranches
that mature before 2008 (79,868) because our interbank market sample starts in 2008. A
tranche-level analysis as opposed to a deal-level analysis is appropriate in our case as we
have noticed that lenders may vary for different tranches within a deal. Moreover, different
tranches within a deal may have varying maturities. Focusing on the tranche level hence
provides more granular information about banks participations.
Finally, we obtain bank financials from Bankscope. The database, compiled by Fitch /
Bureau Van Dijk from publicly available data, provides yearly information on bank balance
sheets and income statements, including credit risk variables, and financial and profitability
ratios. The sample we collect covers the period 2008 – 2013. We obtain information on 704
banks that are active on the interbank market. Consistently with the previous literature
we consolidated banks for mergers and acquisitions and allocated the loan portfolio of the
acquired bank to the acquiring bank starting from the effective date of the acquisition.
6
2.2
Measuring the Syndicated Loans Share
In order to measure the amount of information, say Bank 1 holds about the syndicated
loans given by Bank 2, we construct a pair-level index Syndicated Loans Share1,2 using
Dealscan syndicated loan data. Its construction is as follows. First, we identify the volume of a loan held by a lender, using the loan amount and the percentage held whenever
available in the dataset. Whenever the latter is missing, we allocate 31.84% of the loan to
the lead-bank2 and distribute the remaining amount evenly among non-lead banks. Second,
we calculate the volume of syndicated loans outstanding for each interbank borrower on a
d
). A tranche is considered to be outstanding at day
daily basis (V olume OutstandingBorr
d when the active date is before d, and the maturity date is after d. Third, we calculate
a bank-pair level amount of outstanding syndicated loans held by an interbank borrower
Borr and for which an interbank lender Lend participated to the same syndicated loans
d
(V olume OutstandingLend,Borr
).
The Syndicated Loans ShareLend,Borr,d measures the amount of private information Bank
Lend (interbank lender ) holds about Bank Borr (interbank borrower ) through syndicated loan
markets at day d and is expressed as:
Syndicated Loans ShareLend,Borr,d =
d
V olume OutstandingLend,Borr
d
V olume OutstandingBorr
A high Syndicated Loans Share ratio signifies that an interbank lender carries relatively
more information about an interbank borrower. Indeed, the higher the total volume of
shared syndicated loans between an interbank lender and interbank borrower relative to the
interbank borrower’s portfolio of syndicated loans, the more information the interbank lender
has about the interbank borrower. A high syndicated loans share thus reduces uncertainty
about the asset quality assessment of the interbank borrower, and vice versa.
2.3
Constructing the Interbank Loan Interest Rate
We construct the normalized Interest rate (i) as follows. First, we calculate the difference
between the interest rate on a given transaction between banks (iLend,Borr,d ) and the (marketwide) interest rate on overnight transactions on that day (īd ). We calculate a daily volume
2
We calculated the average loan percentage held by lead banks in the entire Dealscan universe for reported
values, and obtained 31.84%.
7
based weighted average of the interest rates. Second, we divide by the standard deviation of
overnight interest rates for that day (σdi ). Third, we average the interest rate measure to a
monthly level for all loans from bank Lend to bank Borr:
iLend,Borr,t =
1 X iLend,Borr,d − īd
Tt d∈t
σdi
where Tt denotes the number of trading days in period t. The construction of the normalized
interest rate follows Cocco et al. (2009) and is justified by the GARCH effect documented
in interbank market interest rates (Hamilton, 1996). This measure allows us to investigate
whether interbank borrowing banks that share more assets with their lender are able to
obtain lower interest rates compared to the average market-wide interest rate, conditioning
on the average level of syndicated loans share between banks in the market.
2.4
Control Variables
In this section we describe other variables that are likely to have an impact on interbank loan
terms. The first set of variables relates to public information available to interbank lending
banks. These include bank Size (measured by the natural logarithm of total assets), Return
on Assets (ROA), and the Loan Loss Provision ratio (LLP ratio). The latter is defined as loan
loss provisions divided by net interest revenue. Contrary to Cocco et al. (2009), who include
the proportion of non-performing loans (NPLs) to the total of outstanding loans, we use the
LLP ratio because our data set covers 52 different countries. While definitions of NPLs vary
distinctly over the countries in our sample, the loan loss provisions ratio is a standardized
measure. We thereby follow among others De Jonghe and Öztekin (2015) who include the
loan loss provision ratio in their cross-country analysis of bank capital. The second set of
control variables relates to previous interbank relationships and interbank bargaining power,
both of which have been found to be important determinants of the interest rate. Similar
to Cocco et al. (2009), Affinito (2012), and Braüning and Fecht (2013) we consider the
intensity of interbank relationship lending, that is, how important a particular counterparty
is for a bank relative to all other banks in the interbank market. We compute a Lender
Preference Index (LPI) equal to the amount AmountLend→Borr
lending bank Lend granted
t
to borrowing bank Borr at time t, relative to the total amount lending bank Lend granted
to all interbank market participants over period t:
8
LP ILend,Borr,t
P
AmountLend→Borr
t
= Pd∈t
Lend→All
d∈t Amountt
Similarly, we compute Borrower Preference Index (BPI) which equals the ratio of the
amount AmounttLend→Borr borrower Borr borrowed from lender Lend, relative to the total
amount borrowed by borrower Borr over the same period t. The ratios are constructed
monthly. Lastly, we calculate the relative importance of banks active in the interbank
market. Market Share represents the total amount that the bank has lent (borrowed) in
the market during one month relative to the total amount that was lent (borrowed) in the
market during that month.
2.5
Summary Statistics
Table 1 provides summary statistics for our main variables. The first panel depicts statistics
for the euro area unsecured overnight interbank market. Since our Interest rate measure
is the difference between the interest rate on the loan and the market-wide interest rate
for that period, it is zero on average. The Borrower Preference Index (BPI) and Lender
Preference Index (LPI) have a mean (median) of 0.73 (0) per cent and 0.93 (0) per cent,
respectively. As the mean is significantly larger than the median for the two ratios, we
observe that banks lend (borrow) little to (from) most banks, but large amounts to (from) a
few of them. We should note that these characteristics are based on a sample which includes
banks that interact with each other on the syndicated loan market but not necessarily do so
on the interbank market. If the bank is not active on the interbank market in the period,
the preference indices are set to zero. Indeed, when we compare these statistics to previous
literature where samples only include banks active on the interbank market, we find lower
concentration of lending and borrowing activity, on average3 .
Lender and borrower characteristics are reported in the second and third panel of Table
1. The data in these panels are winsorized at the 1 and 99% level. On average, borrowing
banks are larger, have similar return on assets, and a lower loan loss provisions ratio than
lending banks. The market share for a lending (borrowing) bank equals 0.17% (0.13%)
on average, and the median value equals zero. While there is a positive correlation with
size as measured by total assets, the correlation coefficient is limited for lending (0.12) and
3
Cocco et al. (2009) report a mean (median) of 8.39 (4.09) per cent for the LPI and 7.94 (3.07) per cent
for the BPI. Affinito (2012) reports a mean (median) of 13 (27) for the LPI and 19 (33) for the BPI.
9
borrowing (0.11) banks. The last panel of Table 1 depicts statistics on the Syndicated Loans
Share variable and our primary sample of syndicated loans. On average, a bank borrowing on
the interbank market shares about 12.5% of its syndicated loan portfolio with its interbank
lender. The average loan size is 200 million Euro, while the average loan has 6.6 lenders,
2.7 lead arrangers, and 3.9 participants. For the loans that have this information available,
the amount kept by the lead arranger is on average 31.84%, with a standard deviation of
22.39%.
3
Empirical Strategy and Results
Our empirical analysis is composed of two parts where we investigate the determinants of
interbank loan terms. In the first set of estimations we control for bank heterogeneity using
bank characteristics obtained from Bankscope. In the second set of estimations we control
for bank heterogeneity using fixed effects variables constructed from the interbank market
data set. The latter set of estimations tests the robustness of our findings for the inclusion
of a more encompassing set of regressors.
3.1
Interbank Loans Pricing
We investigate the determinants of the interest rate on interbank markets. We do so using a
regression analysis because it allows us to estimate the impact of different bank characteristics
on the interbank loan terms within a single regression, as opposed to a matching methodology.
We first estimate the impact of the Syndicated Loans Share variable on the Interest rate
measure, conditioning only on previous interbank lending relationships:
iLend,Borr,t = α + β1 Syndicated Loans ShareLend,Borr,t
(1)
+ β2 LP ILend,Borr,t + β3 BP ILend,Borr,t + βt Dt
where t indexes time, Dt are time dummies, the subscripts Lend and Borr refer to lenders
and borrowers, respectively. We expect borrowers who share more of their syndicated loans
portfolio with interbank lenders to pay lower interest rates because of reduced information
asymmetry. Hence, we expect a negative sign on the Syndicated Loans Share variable.
Further, we expect the LPI to have a positive sign, and the BPI to carry a negative sign.
10
Establishing interbank relationships should have a favorable impact on the rate, whichever
side of the market the bank is on.
Column (1) of Table 2 shows the estimation results for Model (1). The results show that
interbank borrowers who share a higher proportion of their syndicated loan portfolio with
their interbank lender pay a lower interest rate. This suggests there is lower information
asymmetry between interbank counterparties when they work more closely together on the
syndicated loan market. Moreover, the LPI has a statistically significant positive sign, suggesting that lenders who have a relationship with their borrowers receive a better (higher)
interest rate. In a counterintuitive manner, the BPI has a positive coefficient as well, implying that borrowers who build a relationship with their lenders are charged a higher interest
rate. Since the interest rate is an endogenous decision based on bank characteristics, we
allow for bank heterogeneity by including bank Size, Return on Assets, and the LLP Ratio
as explanatory variables. The model that we estimate is:
iL,B,t = α + β1 Syndicated Loans ShareLend,Borr,t
(2)
+ β2 SizeBorr,t + β3 ROABorr,t + β4 LLP RatioBorr,t
+ β5 SizeLend,t + β6 ROALend,t + β7 LLP RatioLend,t
+ β8 LP ILend,Borr,t + β9BP ILend,Borr,t + βt Dt + Lend,Borr,t
Estimation results for Model [2] are shown in columns (2) and (3) of Table 2. We find
a negative and statistically significant impact of syndicated loans share on the interest rate
charged, even after controlling for bank characteristics. Further, larger lenders receive lower
interest rates, while larger borrowers seem to attract liquidity at a lower cost. The interbank
borrower’s profitability, as measured by return on assets, has a positive impact on the pricing
of the loan, while a lower asset quality (higher LLP Ratio) increases the interest rate on the
loan. Lenders receive higher interest rates on loans to borrowers with which they have a
relationship, while again we do not find evidence of a beneficial effect of having a borrower
relationship.
Results in columns (4) and (5) of Table 2 include market shares of banks to investigate
whether borrowing banks receive better rates because of their better bargaining power (Osborne and Rubinstein, 1994). Including market shares also enables us to insulate the impact
of the reduced information asymmetry that follows from increased syndicated loans share.
11
Market shares are positively related with size, with correlation coefficients equal to 0.12 for
lenders, and 0.11 for borrowers. Column (4) of Table 2 shows that borrowers with larger
market shares receive more favorable interest rates. Conversely, lenders’ relative importance
in the interbank market has a negative effect on the interest rate. We find that the effect of
syndicated loans share on loan interest rates is robust to the inclusion of bank characteristics
and market shares. These results are also economically significant. For the estimations in
column (4), a one standard deviation increase in the borrower’s shared assets with its lender
decreases the interest rate on a loan by 7 basis points. We include both market share and
size in column (5) of Table 2. The borrower’s market share no longer has an impact on
the loan interest rate, suggesting that borrower’s lower borrowing cost is not attributable to
higher bargaining power. Coefficients on control variables remain largely consistent over the
5 models.
Overall, we find convincing evidence of syndicated loans share as a source to reduce
information asymmetry, illustrated by lower interest rates on interbank loans. The results
confirm that banks use public information to monitor their counterparties in the interbank
market (Furfine, 2001), and that lending relationships are a determining element in the
pricing of interbank loans (Braüning and Fecht, 2013). Larger borrowers, and borrowers
with higher asset quality obtain a lower cost of liquidity. Contrary to previous literature,
establishing borrower relationships does not seem to reduce the cost of liquidity.
3.2
Interbank Loan Terms and Unobserved Banks Characteristics
In order to address a wider heterogeneity across banks than merely size, profitability and
loan quality as proxied in Model [2], we subject our results to a regression using proxies for
unobserved characteristics of banks as suggested by Afonso et al. (2013). Doing so allows
us to verify whether the robustness of the effect of syndicated loans share on interbank loan
terms. Also we extend the investigation to the impact of syndicated loans share on both
pricing and amounts of interbank loans.
In order to capture unobserved bank characterisics we compute a monthly average spread
of the borrower. Average Spread Borrower captures unobserved characteristics of the borrowing bank in a given month.4 Similarly, Average Spread Lender is the average spread a
4
The number being constant within a month for a borrower, another interpretation of this variables is
that it is a bank-month level fixed effects. In fact Afonso et al. (2013) labeled this variable bank-month level
fixed effect.
12
lender receives on nterbank loans over the course of one month. As a final control variable we
include the one month lagged borrower preference index BP I −1 . We estimate the following
model:
iLend,Borr,t = α + β1 Syndicated Loans ShareLend,Borr,t
(3)
+ β2 Average SpreadBorr,t + β3 Average SpreadLend,t
−1
+ β4 BP ILend,Borr,t
where t, Borr and Lend respectively index time, borrower, and lender.
In a second model, we investigate the amount of an interbank loan, controlling for lender
and borrower’s unobserved characteristics. The dependent variable, AmountLend,Borr,t , is
the average amount of a transaction between a borrower and a lender in a given month.
Explanatory variables are defined as follows. Average Amount Borrower is the average
amount a borrower borrows in a transaction over one month. Average Amount Lender is
the average amount a lender lends in the market over one month. We estimate this model
as follows:
AmountL,B,t = α + β1 Syndicated Loans ShareLend,Borr,t
(4)
+ β2 Average AmountBorr,t + β3 Average AmountLend,t
−1
+ β4 BP ILend,Borr,t
where t, Borr and Lend respectively index time, borrower, and lender.
We expect a higher syndicated loans share to be associated with larger loan amounts,
as interbank lenders have more information about the creditworthiness of the interbank
borrower. Results for Models (3) and (4) are provided in Table 3. All estimations have time
fixed effects.
Columns (1) and (2) of Table 3 depict results for our specifications investigating the
determinants of the interest rate. Borrowers obtain a lower cost of liquidity from lenders
with which they share a larger proportion of their syndicated loan portfolio, confirming our
estimations from the previous section. Lagged borrower preference index BP I −1 has no
significant impact on pricing. These results are also economically significant: a one standard
13
deviation increase in syndicated loans share decreases the interest rate on the transaction
by 4 basis points. Results in column (3) and (4) of Table 3 explain the amount of the loan.
These confirm the expected positive effect of syndicated loans share on access to liquidity.
Interestingly, while the one month lagged borrower preference index has no impact on the
interest rate, there is a significantly positive effect on the access to liquidity.
4
Conclusions
Interbank markets play a vital role in the functioning of the entire financial system. Yet,
informational frictions have sprung up in recent years, leading to large disruptions. In this
paper we study the impact on interbank loan terms of corporate loans co-funding through
syndication. So-called syndicated loans are loans funded by at least two banks. By sharing
these assets, banks hold valuable private information about each other’s outstanding loans.
The objective is to investigate whether banks use the private information they gain about
other banks in syndicated loans markets when it comes to interbank lending.
For our study we construct a measure of syndicated loans share between two banks, based
on a large syndicated loans sample from 1987 to 2014. We relate the index to interbank loan
terms, namely pricing and volumes. The interbank loan data is obtained from TARGET2
which is the major platform of interbank lending in the euro area, and covers the period
July 2008 – December 2013.
We find that interbank borrowing banks that share more syndicated loans with an interbank lender pay lower rates and obtain larger interbank loan amounts. Therfore, lending
banks have an informational advantage when lending to syndicate partners, in that they
reduce uncertainty about the asset quality of the interbank borrower. Our results are robust
to the inclusion of banks characteristics.
Among others, the results support the findings of Cocco et al. (2009) and Affinito et al.
(2012, 2013) that provide evidence of the existence of relationship lending in the interbank
markets. Nevertheless our study is closer to Cocco et al. (2009) and Braüning and Fecht
(2013) who provide hints of informationally driven relationship lending between banks.
All-in-all, the study contributes to answering a fundamental question about interbank
markets: do banks use private information when lending to each other? Also the study
contributes to the broadening of our knowlege of the determinants of interbank lending
interconnectedness which has implications for the monitoring of the financial stability.
14
Appendix A
Construction of Syndicated Loans Share
Corporate loans are subject to substantial private information. Banks that participate to
the same corporate syndicated loan hold the same private information about the value of the
loan. As a result, the more a bank participates to the syndicatad loans of another bank, the
more the former holds private information about the latter’s corporate loans value. For each
pair of banks, we construct Syndicated Loans Share, an index that measures the amount of
private information a bank holds about another bank’s corporate loans through syndicated
loans.
Here is an example of how we construct Syndicated Loans Share. Suppose we want to
compute Syndicated Loans ShareA,B,31.01.2010 , i.e. the share of Bank A’s corporate loans
contracted from syndicated loans and outstanding on 31 January 2010 for which Bank B
holds private information. Table A1 describes Bank A’s syndicated loan participations,
retained amounts, and syndicate members.
The first step consists in calculating the volume of syndicated loans recorded on the
balance sheet of Bank A as of January 2010. We sum up all loan amounts held by Bank A
of syndicated loans that were on Bank A’s balance sheet on January 2010. Table A1 shows
that this amounts to e20m + e15m = e35m.
The second step consists in calculating the outstanding amount of syndicated loans held
by Bank A as of January 2010, and in which Bank B participated as well. Loans 1, 2, and 4
are the loans that Banks A and B co-funded, but of these three loans, only Loan 1 is active
on 31/01/10. Therefore on January 31, 2010, Bank B held private information about e20m
of Bank A’s loan assets contracted through syndicated loans.
Table A1: Bank A’s syndicated loan participations and Bank B’s information, 31/01/2010.
Loan
Loan
Loan
Loan
Loan
1
2
3
4
Active
date
Maturity
date
Amount
held by A
Active on
31/01/10
Outstanding
31/01/2010
Syndicate
members
Bank B holds
information
19/07/08
29/10/08
09/12/09
09/02/10
19/07/10
29/10/09
09/12/11
09/02/12
e20m
e40m
e15m
e10m
Yes
No
Yes
No
e20m
e0m
e15m
e0m
A,
A,
A,
A,
e20m
e0m
e0m
e0m
e35m
Total:
B, C
B, D
C
B, E
e20m
Finally, Syndicated Loans ShareA,B,31.01.2010 is equal to the ratio of the outstanding
amount that is shared between bank A and Bank B on the total amount outstanding at
Bank A, namely 20/35=57.1%.
15
References
[1] Acharya, V. V., Merrouche, O., 2012. Precautionary Hoarding of Liquidity and Interbank Markets: Evidence from the Subprime Crisis. Review of Finance, 28(4), 1103-1152.
[2] Affinito, M., 2012. Do interbank customer relationships exist? And how did they function in the crisis? Learning from Italy. Journal of Banking and Finance, 36(12), 31633184.
[3] Afonso, G., Kovner, A., Schoar, A., 2011. Stressed, Not Frozen: The Federal Funds
Market in the Financial Crisis. The Journal of Finance, 66(4), 1109-1139.
[4] Afonso, G., Kovner, A., Schoar, A., 2013. Trading partners in the interbank lending
market. Federal Reserve Bank of New York Staff Reports, 620.
[5] Allen, F., Gale, D., 2000. Financial Contagion. Journal of Political Economy, 108(1),
1-33.
[6] Arciero, L., Heijmans, R., Heuver, R., Massarenti, M., Picillo, C., Vacirca, F., 2013.
How to measure the unsecured money market? The Eurosystem’s implementation and
validation using TARGET2 data. DNB Working Paper, 369(January 2013).
[7] Bharath, Sreedhar, Sandeep Dahiya, Anthony Saunders and Anand Srinivasan, 2011.
Lending Relationships and Loan Contract Terms. The Review of Financial Studies 24(2),
1141-1203.
[8] Boot, A. W. A., 2000. Relationship Banking: What Do We Know? Journal of Financial
Intermediation, 9(1), 7-25.
[9] Braüning, F., Fecht, F., 2013. Relationship lending in the interbank market and the
price of liquidity. Working Paper.
[10] Cai, J., Saunders, A., Steffen, S., 2014. Syndication, interconnectedness, and systemic
risk. Working Paper.
[11] Cocco, J. F., Gomes, F. J., Martins, N. C., 2009. Lending relationships in the interbank
market. Journal of Financial Intermediation, 18(1), 24-48.
[12] De Jonghe, O., Öztekin, Ö., 2015. Bank capital management: International evidence.
Journal of Financial Intermediation, 24(2), 154-177.
[13] Dennis, S. A., Mullineaux, D. J., 2000. Syndicated Loans. Journal of Financial Intermediation, 9(4), 404-426.
[14] Diamond, D. W., 1984. Financial Intermediation and Delegated Monitoring. The Review
of Economic Studies, 51(3), 393-414.
16
[15] Fama, E. F., 1985. What’s different about banks? Journal of Monetary Economics,
15(1), 29-39.
[16] Freixas, X., Holthausen, C., 2005. Interbank Market Integration under Asymmetric
Information. Review of Financial Studies, 18(2), 459-490.
[17] Freixas, X., Jorge, J., 2008. The Role of Interbank Markets in Monetary Policy: A
Model with Rationing. Journal of Money, Credit and Banking, 40(6), 1151-1176.
[18] Furfine, C., 1999. The Microstructure of the Federal Funds Market. Financial Markets,
Institutions and Instruments, 8(5), 24-44.
[19] Furfine, C., 2001. Banks as Monitors of Other Banks: Evidence from the Overnight
Federal Funds Market. The Journal of Business, 74(1), 33-57.
[20] Hamilton, James D., 1996. The daily market for Federal Funds. Journal of Political
Economy 104, 26-56.
[21] Heider, F., Hoerova, M., Holthausen, C., 2009. Liquidity hoarding and interbank market
spreads: The role of counterparty risk. ECB Working Paper Series, 1126.
[22] Ho, T. S. Y., Saunders, A., 1985. A Micro Model of the Federal Funds Market. The
Journal of Finance, 40(3), 977-988.
[23] Ivashina, V., Scharfstein, D., 2010. Bank lending during the financial crisis of 2008.
Journal of Financial Economics, 97(3), 319-338.
[24] James, C., 1987. Some evidence on the uniqueness of bank loans. Journal of Financial
Economics, 19(2), 217-235.
[25] Lee, S. W., Mullineaux, D. J., 2004. Monitoring, Financial Distress, and the Structure
of Commercial Lending Syndicates. Financial Management, 33(3), 107-130.
[26] Nirei, M., Caballero, J., Sushko, V., 2015. Bank capital shock propagatioon via syndicated interconnectedness. BIS Working Papers, 484.
[27] Osborne, M. J., Rubinstein, A., 1994. A course in game theory. MIT Press, Cambridge.
[28] Petersen, M. A., Rajan, R. G., 1995. The Effect of Credit Market Competition on
Lending Relationships. The Quarterly Journal of Economics, 110(2), 407-443.
[29] Rochet, J.-C., Tirole, J., 1996. Interbank Lending and Systemic Risk. Journal of Money,
Credit and Banking, 28(4), 733-762.
[30] Sette, Enrico and Giorgio Gobbi, 2015. Relationship Lending during a Financial Crisis.
Journal of the European Economic Association, 13(3).
17
[31] Sufi, A., 2007. Information Asymmetry and Financing Arrangements: Evidence from
Syndicated Loans. The Journal of Finance, 62(2), 629-668.
18
TABLE 1. Summary Statistics for Key Variables
This table reports mean, standard deviation, and quartiles for unsecured overnight loans, main characteristics of borrowers and
lenders in the euro area interbank market, and syndicated loans. The sample covers the period January 2008 to January 2014.
Variable
Mean
St. Dev.
25%
50%
75%
Normalized Interest Rate
Syndicated Loans Share
0.00
0.13
1.00
0.19
-0.55
0.00
-0.16
0.03
0.40
0.18
INTERBANK RELATIONSHIPS
Borrower Preference Index (percent)
Lender Preference Index (percent)
0.73
0.93
5.83
6.56
0.00
0.00
0.00
0.00
0.00
0.00
BORROWER
Assets (million Euro)
Return on Assets (percent)
Loan Loss Provisions Ratio (percent)
Market Share (percent)
190,072
20.70
35.68
0.13
5
1.07
44.10
0.75
70,951
4.00
13.34
0.00
243,134
28.00
25.04
0.00
731,666
59.00
43.16
0.00
LENDER
Assets (million Euro)
Return on Assets (percent)
Loan Loss Provisions Ratio (percent)
Market Share (percent)
162,530
20.64
38.54
0.17
6
1.10
50.07
1.42
44,861
6
14.21
0.00
218,887
29
26.76
0.00
708,305
60
42.50
0.00
200
6.58
2.70
3.88
31.84
542
6.12
4.20
4.22
22.39
22.8
3
1
0
14.29
64.8
5
2
1
25.71
181
8
5
4
45
SYNDICATED LOANS
Size of tranche (million Euro)
Total number of lenders
Total number of lead arrangers
Total number of participants
%kept by lead arranger
19
TABLE 2. Determinants of Interbank Loan Pricing
This table provides the estimate of the Models (1):
iL,B,t = α + β1 Syndicated Loans ShareL,B,t + β2 LP IL,B,t + β3BP IL,B,t + βt Dt + L,B,t
and (2):
iL,B,t
=
α + β1 Syndicated Loans ShareL,B,t + β2 SizeB,t + β3 ROAB,t + β4 LLP RatioB,t
+
β5 SizeL,t + β6 ROAL,t + β7 LLP RatioL,t
+
β8 LP IL,B,t + β9BP IL,B,t + βt Dt + L,B,t
The dependent variable is the normalized interest rate, defined as follows. First, we calculate the difference between the interest
rate on a given transaction between banks and the (market-wide) interest rate on overnight transactions on that day. Second,
we divide by the standard deviation of overnight interest rates for that day. Third, we average the interest rate measure thus
obtained to a monthly level for all loans between two banks. Size is the natural logarithm of assets. The construction of the
Lender Preference Index and the Borrower Preference Index follow Cocco et al. (2009).
The Lender Preference Index is constructed by dividing the total amount of loans given by the lender in the previous month
over the total amount of interbank loans it gave in te previous month. The Borrower Preference Index is constructed by dividing
the total amount of loans the borrower the lender in the previous month over the total amount of interbank loans the borrower
received in the previous month. Market share is the market share of the bank (borrower or lender). Return on assets is the
assets return in the previous year of the bank (borrower or lender). Loan loss provisions report the share of loan loss provisions
of the bank in the previous year (borrower or lender). All models include fixed year effects. We report robust errors. The sample
covers the period July 2008 to December 2013. T-statistics are in parentheses. ***, **, and * indicate statistical significance
at the 1%, 5%, and 10% level, respectively.
20
Interest Rate
Syndicated Loans Share
INTERBANK RELATIONSHIP
Lender Preference Index
Borrower Preference Index
Interest Rate
Interest Rate
Interest Rate
Interest Rate
-0.115∗∗∗
(-3.02)
-0.370∗∗∗
(-10.33)
-0.113∗∗∗
(-2.96)
0.003∗∗∗
(7.44)
-0.000
(-1.07)
0.003∗∗∗
(7.51)
-0.000
(-0.98)
0.004∗∗∗
(9.65)
0.002∗∗∗
(3.88)
0.003∗∗∗
(6.47)
0.000
(0.83)
-0.135∗∗∗
(-24.54)
-0.133∗∗∗
(-24.20)
-3.231∗∗∗
(-10.30)
0.015
(1.34)
0.001∗∗∗
(5.74)
-0.133∗∗∗
(-22.50)
0.018
(0.06)
0.018∗∗
(1.72)
0.001∗∗∗
(4.61)
-0.047∗∗∗
(-9.33)
-0.960∗∗∗
(-6.46)
-0.012∗∗∗
(-2.34)
0.002∗∗∗
(8.07)
Yes
24,581
0.06
-0.788∗∗∗
(-26.90)
0.003∗∗∗
(20.29)
0.003∗∗∗
(22.39)
BORROWER
Size
Market Share
Return on Assets
Loan Loss Provisions ratio
LENDER
Size
0.020∗∗
(1.88)
0.001∗∗∗
(4.69)
0.019∗∗
(1.80)
0.001∗∗∗
(4.68)
-0.062∗∗∗
(-13.88)
-0.057∗∗∗
(-11.81)
-0.013
(-1.59)
0.002∗∗∗
(9.63)
-0.012
(-1.54)
0.002∗∗∗
(9.71)
-1.572∗∗∗
(-10.95)
-0.022∗∗∗
(-2.69)
0.002∗∗∗
(8.23)
Yes
24,581
0.06
Yes
24,581
0.06
Yes
24,581
0.03
Market Share
Return on Assets
Loan Loss Provisions ratio
FIXED EFFECTS
Year
Observations
R2
Yes
119,148
0.01
21
TABLE 3. Fixed Effects Model for Interbank Loan Terms
This table provides the estimate of the Models (3):
iL,B,t
=
α + β1 Syndicated Loans ShareL,B,t
+
β2 Average SpreadB,t + β3 Average SpreadL,t
+
β4 L BP IL,B,t + L,B,t
and (4):
AmountL,B,t
=
α + β1 Syndicated Loans ShareL,B,t
+
β2 Average AmountB,t + β3 Average AmountL,t
+
β4 L BP IL,B,t + L,B,t
The dependent variable in columns (1) and (2) is the normalized interest rate, defined as follows. First, we calculate the
difference between the interest rate on a given transaction between banks and the (market-wide) interest rate on overnight
transactions on that day. Second, we divide by the standard deviation of overnight interest rates for that day. Third, we
average the interest rate measure thus obtained to a monthly level for all loans between two banks. The dependent variable in
columns (3) and (4) is the average loan size between a borrower and a lender in a given month. The sample covers the period
July 2008 to December 2013. T-statistics are in parentheses. ***, **, and * indicate statistical significance at the 1
Interest Rate
Interest Rate
-0.000
(-0.94)
3.311∗∗∗
(130.70)
2.587∗∗∗
(91.44)
-0.190∗∗∗
(-7.71)
-0.000
(-0.66)
3.300∗∗∗
(130.03)
2.577∗∗
(91.01)
Syndicated Loans Share
BP I −1
Average Spread Borrower
Average Spread Lender
Average Amount Borrower
Average Amount Lender
Year Fixed Effect
Observations
R2
Yes
115,674
0.25
Yes
115,674
0.25
22
Loan Size
Loan Size
13,400,000∗∗∗
(60.73)
48,300,000∗∗∗
(14.13)
13,400,000∗∗∗
(60.98)
0.792∗∗∗
(112.80)
0.776∗∗∗
(131.67)
0.786∗∗∗
(111.75)
0.759∗∗∗
(126.68)
Yes
115,674
0.29
Yes
115,674
0.29