You are on page 1of 16

Journal of Corporate Finance 25 (2014) 403418

Contents lists available at ScienceDirect

Journal of Corporate Finance


journal homepage: www.elsevier.com/locate/jcorpfin

The effect of banking relationships on the future of nancially


distressed rms
Claire M. Rosenfeld
Mason School of Business, College of William & Mary, P.O. Box 8795, Williamsburg, VA 231878795, United States

a r t i c l e

i n f o

Article history:
Received 19 October 2012
Received in revised form 6 January 2014
Accepted 11 January 2014
Available online 21 January 2014
JEL classification:
G30 (Corporate finance and governance)
G20 (Financial institutions and services)

a b s t r a c t
In this study I empirically examine U.S. publicly traded firms to determine the impact of banking
relationships on the future of financially distressed firms. Results demonstrate that obtaining a
relationship-backed loan in the six months prior to distress identification significantly increases
the probability of future firm emergence from distress. However, this effect decreases as the
severity of firm distress increases. These results are robust to variations in banking relationship
measures and to addressing endogeneity. This study provides evidence consistent with the value
of lending relationships stemming from the ease of transmission of soft information within the
lender's organization.
2014 Elsevier B.V. All rights reserved.

Keywords:
Corporate finance
Financial distress
Banking relationships

1. Introduction
A banking relationship develops when a lender repeatedly provides credit to a firm.1 Repeated lending can reduce information
asymmetries between the borrower and lender by providing lenders soft information on the firm's ability to repay debt. This
soft information can include information on management's ability to overcome adverse situations, the firm's internal control of
spending, and the veracity of the firm's financial statements. Information asymmetry between borrower and lender can be more
evident when a firm is in financial distress, a time when there is more uncertainty regarding the firm's viability. Further, a firm
facing distress needs funds to remain financially viable, so the firm is especially reliant upon the bank for funding while the bank
maintains its interest in its current and future revenue stream. These circumstances of heightened information asymmetries
paired with more prominent interdependence between firm and lender create a useful setting for the analysis of how reduced
information asymmetries resulting from relationship lending affect financially distressed firms.
There are two views on how lending relationships affect financially distressed firms. First, it is reasonable to expect that
relationship banks employ their informational advantage particularly when choosing to lend to distressed firms. This
discretionary issuance of loans results in a benefit to both the financially distressed firm, by providing capital, as well as the
lender, by providing the prospect of future rents upon emergence from distress. In contrast, Weinstein and Yafeh (1998) find that
although prior lenders provide credit to a financially distressed firm, they inhibit the firm's ability to generate profits through
higher interest payments and conservative investment policies, thereby weakening the firm's future prosperity and the bank's
value of the lending relationship.
This study examines the effect of banking relationships on the success probability of financially distressed firms. Specifically, I
perform probit regressions with controls for firm, loan, industry, and macroeconomic attributes to analyze the marginal effect that

Tel.: +1 757 221 4713; fax: +1 757 221 2884.


E-mail address: claire.rosenfeld@mason.wm.edu.
1
Throughout this paper bank refers to any lending institution.
0929-1199/$ see front matter 2014 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jcorpn.2014.01.003

404

C.M. Rosenfeld / Journal of Corporate Finance 25 (2014) 403418

banking relationships have on the probability of a financially distressed firm's future emergence from distress. While this is not the
first study on how banking relationships affect firms (see Ongena and Smith (2000) for a review on banking relationships), this study
contributes to the still growing body of knowledge on how lenders affect their borrowers' long-term performance.
The issue of how a banking relationship affects the performance of a financially distressed firm is an economically relevant
question. When a firm maintains its viability, its creditors remain whole. However, once a firm defaults, that default not only affects
the firm through increased cost of debt, detraction of management's attention away from daily duties, and general loss of employee
morale, thereby creating more inefficiencies, but it also directly affects its lenders through possible losses.2 Further, when the firm
defaults, its stockholders suffer from being residual claimants below a possibly already impaired group of creditors. In sum, when a
firm faces financial distress, that risk of default is born by the firm, its employees and stakeholders, the lender, and the lender's equity
holders. Thus, understanding how a lender affects a financially distressed borrower's future performance is worthwhile.
This paper makes three main contributions. First, I find that obtaining distressed funding in the six months prior to distress
identification from a relationship lenderdefined as a lead lender that was a prior lead lenderhas a positive impact on the
emergence from distress of financially distressed large U.S. firms. Since emergence from distress increases the probability of the
lender remaining whole on its loan to the distressed borrower, this positive effect on future firm performance is consistent with
lending relationships not only having value to the distressed borrower, but also to its lender. Further, I find that as the degree of
firm financial distress increases, the likelihood of future firm emergence from distress diminishes, even though the firm obtains
relationship funding. That is, relationship lenders still provide credit to their severely distressed borrowers, even when the
lenders can expect no advantageous future outcome from extending this credit. These findings are consistent with Weinstein and
Yafeh's (1998) findings that relationship lenders provide credit to their distressed firms, while the findings are inconsistent with
those of Elsas and Krahnen (1998), where German Hausbanks provide liquidity insurance to distressed firms, but only through
moderate distress.
Second, I provide empirical evidence consistent with the ease of information transmission within a lending organization providing
a mechanism through which lending relationships create value. Specifically, distressed firms seeking to replicate the benefits of
relationship banking through a non-relationship lender may do so by borrowing from a non-relationship less-complex lender,
measured by the number of entities within the lending organization, or by borrowing from a non-relationship comparatively-smaller
lender (based on assets). This finding is consistent with Stein's (2002) theory that a decentralized organization, as opposed to a
largely hierarchical organization, lends itself better to the transmission of soft information.
Finally, the main findings are robust to addressing the endogeneity of determining the lending relationships of loans to
financially distressed firms. I use bivariate probit regressions to control for endogeneity, whereby I simultaneously predict future
firm emergence from distress given an exogenous (actual) relationship and the nature of the lending relationship given an
identifying instrument. In particular, I find that the firm's prior reliance upon relationship funding significantly positively predicts
the distressed lending relationship, while rejecting the presence of endogeneity.
The remainder of this paper is organized as follows: Section 2 contains a literature review. Section 3 contains a description of
the study's sample selection and provides variable definition and descriptive statistics. I describe the study's methodology in
Section 4. In Section 5, I discuss the main results, while I discuss robustness checks in Section 6. I analyze mechanisms through
which the relationship creates value in Section 7, and in Section 8, I address endogeneity. I conclude in Section 9.
2. Literature review
The modern literature on banking relationships has a foundation in James's (1987) study of bank loans, followed by Sharpe's
(1990) and Rajan's (1992) work on the informational rents extracted by lenders, which establishes a set of testable theories for
future empirical work. More recent empirical work analyzes German firms (Elsas, 2005; Elsas and Krahnen, 1998), Japanese firms
(Weinstein and Yafeh, 1998), Belgian firms (Degryse and Ongena, 2005), Norwegian firms (Ongena and Smith, 2001), Czech firms
(Ongena et al., 2011), Indian firms (Berger et al., 2008), Asian firms in financial crisis (Jiangli et al., 2008), small American firms
(Berger and Udell, 1995; Cole, 1998; Petersen, 1999; Petersen and Rajan, 1994), and large American firms amidst formal
bankruptcy proceedings (Dahiya et al., 2003a).3 Houston and James (1996, 2001), Gonzales and James (2007), and Schenone
(2004, 2010) also study banking relationships involving large publicly traded firms.
There are several hypotheses common to papers in this literature.4 Most prominently, there is the notion that banking
relationships alleviate information asymmetry through continued contact with their customers. This reduced information
asymmetry can benefit the borrower through better loan terms (Berger and Udell, 1995; Bharath et al., 2011; Petersen and Rajan,
1994; Santos and Winton, 2008), more easily accessible capital (Cole, 1998; Cotugno et al., 2012; Jiangli et al., 2008; Petersen,
1999) and improved liquidity insurance (Elsas and Krahnen, 1998). My work adopts the information asymmetry hypothesis, but
unlike earlier papers, I focus on the potential influence of banking relationships on the ability of publicly traded financially
distressed firms to improve their financial position over an extended period of time.
My analysis evaluates the long-term effect of banking relationships on financially distressed firms. Such firms are unique both
in their reliance upon external funding, which allows them to remain in operation, and in the increased information asymmetries
created by distress. Although Dahiya et al. (2003a) investigate the impact of relationship debtor-in-possession financing, they
2
3
4

When a rm defaults, Dahiya et al. (2003b) document that its lead lender's stock suffers a signicantly negative abnormal return.
These rms are beyond mere nancial distress. They require legal protection in order to remain in operation.
For a review of banking relationship literature, please see Ongena and Smith (2000).

C.M. Rosenfeld / Journal of Corporate Finance 25 (2014) 403418

405

evaluate firm performance strictly within bankruptcy proceedings. Thus, their sample consists solely of firms in severe financial
distress, and within that sample, they evaluate the impact of relationship funding on the probability of emergence and time to
bankruptcy resolution. My work differs from theirs in that I examine firms under varying levels of distress ranging from firms
experiencing modest financial distress to severely financially distressed firms, and I analyze general firm financial performance
over a three-year window of time, regardless of whether a firm enters and/or exits bankruptcy. Additional work on distressed
firm performance includes a study by Fan et al. (2013), who evaluate the impact of government quality and corporate ownership
structure on distressed firm resolution in China.
In related work, Gilson et al. (1990) find that stockholders of financially distressed firms are better off when debt is
restructured privately, rather than in formal bankruptcy proceedings. Andrade and Kaplan (1998) study financially distressed
firms that took part in highly leveraged transactions, finding that [t]o the extent that they do occur, the costs of distress are
highly concentrated in the period after firms become distressed, but before they enter Chapter 11 (p. 1487).
3. Sample selection
3.1. Initial universe
My initial sample universe consists of the intersection of COMPUSTAT's North America Industrial Annual database and Reuters
Loan Pricing Corporation's DealScan database, and spans from 1982 to 2009. DealScan is a loan database that details commercial
loan transactions by providing, among other information, loan date, tranche amount, lender information, fee structure, loan
terms, and type of loan.5 Most of the loans listed in DealScan are syndicated, many by multiple lead lenders.6 I match COMPUSTAT
to DealScan borrowers by company name, and when necessary, city and state of firm headquarters. I restrict the sample to firms
that have at least two loans listed in DealScan so that there is a basis for determining the nature of a loan's lending relationship.
To isolate the issue of financial distress from the common financial instability of start-up firms, I require all firms to be publicly
traded for at least three years. I also eliminate firms in the financial sector (SIC codes 60006799). By restricting the sample to the first
observation for each firm that is identified as distressed and receives a loan in the required timeframe, I simplify the analysis to not be
conditional upon surviving prior distress and eliminate controls for borrowerlender pairs learning about the distress process.
This approach results in a sample with 79,504 tranches7 from 52,449 loans to 13,014 distinct firms. Since lending relationships
are at issue, the first loan for each firm (or 13,014 loans) is ineligible for the sample since it only provides a basis on which to
determine the nature of future loan relationships. Since I am evaluating firm performance over the three years following distress
identification, my sample ends in August of 2007, when this calibration of the Moody's KMV EDF series ends.
3.2. Sample denition
There are a number of ways to identify a firm as financially distressed. Gilson et al. (1990) use a combination of a low
cumulated stock return with news of a defaulted payment, while Andrade and Kaplan (1998), among other methods, examine
firms with low interest coverage ratios. Shumway (2001) points out that these static methods incorporate less information than
in a hazard model setting. He argues that hazard models adjust for a firm's period at risk, incorporate explanatory variables that
change with time, and they may produce more efficient out-of-sample forecasts by utilizing much more data than static
models (p. 102103). In this paper I use expected default frequencies (EDFs) from Moody's KMV to identify financially distressed
firms. Thus, I further limit the sample to firms for which Moody's KMV provides EDFs, which begins in mid-1993. 8
The Moody's KMV model applies Merton's (1974) framework wherein the equity of the firm is a call option on the underlying
value of the firm with a strike price equal to the face value of the firm's debt. Moody's KMV solves two nonlinear simultaneous
equations for the value of the firm and the firm's volatility. After inferring these values, Moody's KMV determines the default
point based on contractual obligations and then calculates the distance to default as the number of standard deviations from
default (Moody's KMV (2004)). They scale the distance to default into EDFs using actual historical default rates.
Moody's KMV calculates their forward-looking EDFs on a monthly basis with only the data available as of the EDF date and
promptly distributes the data.9 EDFs are based on probability of firm default, rather than on firm extinction or filing for federal
bankruptcy protection under Chapters 7 or 11. This focus matches the ideology for identifying financial distress, which also occurs
without necessitating firm extinction or formal bankruptcy proceedings.
Moody's KMV EDFs range from .02% to 20% and are referred to without the percentage sign. A firm with an EDF of, say, 10 will
default within the next year with ten percent probability. Randomly sampling 100 such firms will result in 10 firms defaulting
within one year, on average. A firm with an EDF of 10 is ten times as likely to default as a firm with an EDF of 1. These referential
characteristics of the default measure are unique to the Moody's KMV EDFs and cannot be readily simulated.
5

Please see Carey et al. (1998) or Dichev and Skinner (2002) for a thorough database description.
Details on how I address multiple lead lenders follow with the discussion of the denition of a relationship lender.
Loan deals can be comprised of single or multiple tranches, or slices, of funding. Each tranche can have distinct terms including dollar amount, maturity, fees,
and covenants. DealScan delineates the aggregate deal amount and each tranche amount. Due to the lengthy nature of the loan syndication process, I count all
deals activating within 30 days as the same loan deal.
8
Though Moody's KMV have recalibrated their model in 2007 to span from 0.01% to 35%, they have not backlled the historical EDFs far enough with the
recalibrated values to be used in this study. Thus, I use their former model, which spans from 0.02% to 20% and dates back to 1993.
9
In this study, I use a compilation of the original EDF data where each monthly observation was received early in the month following the EDF date.
6
7

406

C.M. Rosenfeld / Journal of Corporate Finance 25 (2014) 403418

Table 1
Descriptive statistics.
Panel A contains summary statistics on all Expected Default Frequencies (EDFs) for the entire sample universe of the intersection of COMPUSTAT, CRSP, and
DealScan. Panels B and C contain summary statistics on the sample omitting distressed firms with an EDF of 20. Panel B contains historical summary statistics for
the sample firms. Year-over-year changes are calculated as the annual percentage change in value between the most recent quarter end prior to the distressed
loan and one year prior to that. Panel C contains sample observations and firm emergence from distress by calendar year of the distressed loan date. Obs is the
number of observations identified in each year. Percentages are determined in proportion to Obs. Relationship is the percentage of sample observations that
borrow from a lead lender that was a lead lender on a loan in the five years prior to the firm's distressed loan, as listed in DealScan. Merged (liquidated/delisted) is
the proportion of observations that merge (are liquidated or permanently delisted) in the three years following the identification of distress.
PANEL A: SUMMARY STATISTICS: SAMPLE UNIVERSE
Variable

Min

Max

Median

Mean

Std. Dev.

p75

p90

EDF (sample universe)

485,069

0.02

20

0.79

3.5260

5.9264

3.26

14.78

PANEL B: HISTORICAL SUMMARY STATISTICS: EDF b 20


Variable

Median

Market cap prior quarter year-over-year change


Sales prior quarter year-over-year change
EPS prior quarter year-over-year change
Leverage prior quarter year-over-year change
Operating profit margin prior qtr year-over-year chg
Prior year EDF mean
Prior year EDF std. dev.

649
640
628
621
542
659
659

23.51%
5.48%
54.39%
1.12%
13.02%
2.6033
0.8701

PANEL C: OBSERVATIONS BY CALENDAR YEAR


Year

Obs

Relationship

Emerge from distress

Merged

Liquidated/delisted

1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Total

43
31
51
64
108
104
108
60
71
13
6
659

39.53%
48.39%
45.10%
48.44%
56.48%
65.38%
63.89%
58.33%
54.93%
46.15%
66.67%
55.84%

55.81%
54.84%
37.25%
28.13%
25.93%
36.54%
62.04%
65.00%
74.65%
84.62%
50.00%
48.10%

16.28%
9.68%
25.49%
15.63%
13.89%
11.54%
11.11%
6.67%
4.23%
0.00%
16.67%
12.14%

9.30%
0.00%
3.92%
10.94%
8.33%
6.73%
4.63%
5.00%
7.04%
7.69%
16.67%
6.68%

To be categorized as financially distressed, I require that a firm meet two constraints: (1) its EDF is high compared to other
firms and (2) its EDF is high compared to its own history. Specifically, a financially distressed firm has an EDF above the mean EDF
for the entire history of all sample universe firms (3.526). Note that Panel A of Table 1 shows that this financial distress threshold
is greater than the initial sample universe median (0.79) and the 75th percentile (3.26) of EDFs. To ensure that the firm is
experiencing increased distress compared to its recent history, I require that the EDF exceeds one standard deviation above its
prior year mean EDF.
Table 1 Panel B shows other measures that support that the sample firms are in distress.10 Year-over-year median changes for
market capitalization, earnings per share, leverage, and operating profit reflect that the sample firms have experienced adverse
financial changes over the year prior to distress identification.
I narrow the sample by requiring all firms to have a loan activated in the six months prior to distress identification. This
restriction allows the distressed firm's management to act in anticipation of the upcoming distress diagnosis while it requires the
bank to grant a loan before distress is assessed, and it reduces lender selection bias. All loans are new deals, as opposed to
amendments to existing loans, as presented by DealScan.11 Table 1 Panel C lists the number of final sample firm observations by
loan year. The final sample spans distress identification from 1994 through 2004, includes 659 firms and has an increased number
of firms identified as financially distressed between 1998 and 2000 ranging between 104 and 108 firms in these years.12 In the
other sample years, 71 firms or less are identified as distressed.
10
As explained later in this section, most of the analysis is performed on a sample of rms that omits the most severely distressed rms (those with an EDF of
20). Table 1 Panels B and C reect this sample modication.
11
Although all loans are listed as new loans by DealScan, they could be augmented existing credit agreements that have been re-drafted, therefore constituting a
pseudo-new loan. Nevertheless, such loans are still a consensual credit agreement when the rm is nearing distress.
12
Although using KMV Moody's EDFs to identify distressed rms limits the sample period to start in 1994, this still allows for the sample to include both
economic expansions and contractions. The sample clearly omits the 19901991 recession, which is a recession that was closely tied to the instability of the
banking system. Thus, the 19942004 sample period has a fairly constant stability within the banking system. This allows for analysis of rm distress without
integration of widespread lender distress into the data. Further, many consider that DealScan's best coverage began in the mid 1990's, which means that limiting
the sample to beginning in 1994 omits some bias in loan reporting.

C.M. Rosenfeld / Journal of Corporate Finance 25 (2014) 403418

407

The sample allows for testing the impact of obtaining a relationship or non-relationship loan. Thus, the sample is conditioned
upon distressed firms obtaining a bank loan. Any selection bias due to firms successfully acquiring bank debt is inherent in the
sample, and is difficult to mitigate, especially when addressing the endogeneity of banking relationships. It is worth noting,
however, that the distress-identifying EDF measure will incorporate that the firm obtained a loan, along with whether the loan
was from a relationship lender. The sample should still reflect distressed firms, since after accounting for any benefit or detriment
from obtaining a relationship- or non-relationship-backed loan, the EDF is still absolutely (compared to the sample universe) and
relatively (compared to the same firm) large. That is, according to this paper's results, a relationship-backed loan should decrease
the firm's EDF. Thus, the sample might omit borderline-distressed relationship-backed firms but include borderline-distressed
non-relationship-backed firms. I find that this is not the case; instead there are more borderline-financially-distressed
relationship-backed firms than non-relationship-backed firms, though the relationship-backed firms have a lower mean EDF. 13
If the first time a firm is identified as being absolutely and relatively financially distressed and obtains distressed funding is
also when the firm experiences the highest possible EDF (of 20), it is reasonable to expect that firm to have a different set of
possible outcomes following a cash infusion than a firm that is first identified as distressed by obtaining a lower EDF. Thus, I
partition the sample into those firms with an EDF of 20 at the time of distress identification and those firms with an EDF less than
20.

4. Methodology
4.1. Emergence denition
I measure future firm performance from the perspective of whether the distressed firm's lenders face reduced exposure to
losses three years following the identification of the firm as financially distressed. The distressed firm's lenders are better off
risk-wise if the firm decreases its probability of default. This is observed through (1) the firm decreasing its EDF to a level that is
less than its average EDF calculated over the year prior to distress identification14 or (2) the firm becomes part of another entity
that is better off financially. The second method of emerging from distress is observed through the distressed firm being acquired
(the acquirer should be in a better financial position to procure the funds and stakeholder approval to make such an acquisition)
or from permanently delisting at a better EDF than the firm's prior year average EDF.15
Table 1 Panel C lists firm emergence from distress as a proportion of firm observations by year of loan activation and type of
exit. Of the 659 sample firms, 48.10% (317 firms) emerge from distress, 12.14% (80 firms) merge in the three years following the
firm's identification of financial distress, and 6.68% (44 firms) liquidate or permanently delist. Note that not all firms that delist
fail to emerge from distress.

4.2. Relationship denitions


The loan that is closest to the identification of distress and that falls within six months prior to the identification of distress is
termed the distressed loan. For the main findings shown here, I use an indicator (as in Schenone (2004)) to denote that any lead
lender on a distressed tranche was a lead lender on any DealScan-listed facility within the five years prior to the distressed loan.16 17
I take the maximum relationship indicator among all lead lenders across all tranches within a loan deal to arrive at the loan
relationship.18 Relationships are tracked through bank mergers and acquisitions, as in Ljungqvist et al. (2006). To identify a tranche's
lead lenders, I use all the lenders in the first non-blank DealScan category in the following order: Lead Arranger, Lead Manager or
Co-Lead Manager, Manager or Co-Manager, and Lead Role. The idea behind this measure is that any prior lead lender is privy to
regular financial updates and so has an informational advantage over a non-prior lender. Further, through the initial loan processing,
any prior lead lender presumably has gained some firm-specific knowledge that outside lenders do not have. Table 1 Panel C shows
the proportion of sample firms obtaining relationship-backed distressed funding by year. In most years, 45% or more of the sample
obtains relationship funding.
13
Of the borderline-nancially-distressed sample rms, dened as an EDF less than 5, there are 116 non-relationship backed rms versus 192 relationshipbacked rms. The sample mean EDF for non-relationship-backed borderline-nancially distressed rms is 4.16, while it is 4.06 for relationship-backed rms. This
difference is signicant at the 5% level.
14
Recall the basis for identifying nancially distressed rms is that the rm experiences an EDF greater than one standard deviation above the average EDF,
calculated over the rolling prior year.
15
To determine emergence, I use the monthly EDF from the relevant event date (three years following identication of distress, acquisition, or permanent
delisting).
16
Since 364-day facilities and revolvers with a maturity less than one year are not susceptible to the same capital requirements as facilities with a maturity
longer than one year, I conservatively exclude them from the distressed rms' prior loans when determining lending relationships. This eliminates any systematic
differences present in such funding. I thank an anonymous referee for identifying this issue.
17
While this measure captures whether the lead lender was a prior lead lender, there is no guarantee that a non-relationship lender truly has no relationship
with the rm. DealScan offers the best available data on private debt deals over $100,000. Nearly 70% of DealScan's data comes from SEC nancial lings, with the
remainder of the data stemming from lender disclosure. While DealScan offers the most comprehensive data available, it does not capture other services rendered
by lenders to clients. Thus, the results in this study should be biased towards zero.
18
The vast majority of deals with multiple tranches have the same lending relationship indicator across all tranches.

408

C.M. Rosenfeld / Journal of Corporate Finance 25 (2014) 403418

4.3. Control variables


I control for five categories of characteristics that either affect the future outcome of the firm or the likelihood of the firm to
obtain relationship funding. These five categories include degree of distress, firm, loan, industry, and macroeconomic attributes.
Variable definitions are listed on Table 2. Table 3 provides summary statistics of the moderately financially distressed sample's
control variables.
4.3.1. Degree of distress
I control for the degree of distress by including the firm's expected default frequency at the time of identification of financial
distress. Table 3 shows that the minimum sample EDF is 3.53, and the maximum value is 19.65. The median value is 5.19, so the
median sample firm has a 5.19% chance of defaulting in the following year. One goal of this study is to determine how
relationships affect future firm performance as the degree of distress increases. I also include the interaction of the firm's
relationship indicator and EDF to determine if the effect of the lending relationship is consistent through varying levels of distress.
As Ai and Norton (2003) find, an interaction effect between two variables cannot be evaluated simply by looking at the sign,
magnitude, or statistical significance of the coefficient on the interaction term when the model is nonlinear (p.129). Whenever
analyzing a system with only one interaction term, I also perform the adjustment for this interaction term (inteff), as created for
Stata by Norton, Wang & Ai.
The final firm distress control I include is the standard deviation of the firm's EDF calculated over the year prior to the
identification of the firm as distressed. I include this control to capture whether the firm has stagnated in a nearly distressed state
with little variation or whether the firm's degree of distress has varied greatly in the time leading up to its inclusion in the sample.
Table 2
Variable definition.
Variables

Definition

Emergence

Indicator denoting future firm emergence from distress, which occurs if three years following its identification
as financially distressed the firm has recovered to an EDF lower than its prior year's average EDF. Mergers are
considered emergences, and a firm that experiences a higher EDF upon liquidation or permanently delisting
than its prior year's average EDF is considered a failure.
Indicator that any lead lender on the distressed loan was a lead lender on a prior loan within the last five years
The proportion of prior funds (number of loans) the lead lender on the distressed loan lent as a lead lender
throughout the firm's prior five years as covered by DealScan
Relationship indicator from the most recent loan prior to the distressed loan
Moody's KMV measure at the time of distress identification
Calculated over year prior to identification of firm as distressed

Relationship indicator
Proportion of dollars (number) of loans
Lagged relationship
Expected default frequency
EDF standard deviation
Firm controls1
Age
Leverage
Operating profit margin
Fixed assets/total assets
Net sales/assets
Assets
Operating cash/interest paid
Solvent indicator
% Change in market value
Loan attributes
Months from loan to distress
Purpose indicators
Refinancing

Industry controls
Industry indicators

Number of years between IPO and distress identification date2


As in Frank and Goyal (2007), (Debt in current liabilities + long-term debt)/(Debt in current
liabilities + long-term debt + market value of equity + preferred stock liquidating value deferred taxes)
Operating profit/sales
Proxies for ability to pledge security
Captures firm's reliance upon assets to generate revenues
Measured in billions
Interest coverage ratio. Dichev and Skinner (2002) find it the second-most prevalent accounting-based loan
covenant in DealScan. Net operating cash flow (including interest paid)/interest paid
Indicator that as of distress identification, Moody's KMV's asset value exceeds its liabilities value
Year-over-year percent change in market value

Number of 30-day months between loan activation date and distress identification
Indicators for top three specific loan purposes as categorized by DealScan: working capital, corporate purposes
and debt repayment
Indicator that either (a) DealScan denotes any tranche in the loan as a refinancing or (b) any new debt
incurred by the firm in the same quarter as the loan is less than 25% of the loan's value.

Indicators of major industry SIC classification. To avoid multi-collinearity, only include controls for
manufacturing, transportation, retail and services

Macroeconomic controls
CFNAI
Year indicators

Chicago Fed National Activity indicator as of month-end prior to loan activation date.
Indicator of calendar year in which loan was activated

Mechanisms
NumEntities 5
Small bank

Indicator denotes that there are five or fewer entities within the lender's organization
Indicator denotes that the lender is in the bottom 10% of the lenders in this sample, as measured by assets

1
Firm controls are as of the fiscal yearend prior to the identification of the firm as distressed. All firm controls except for age are Winsorized at 1st and 99th
percentile.
2
Ln(Age) produces qualitatively similar results.

C.M. Rosenfeld / Journal of Corporate Finance 25 (2014) 403418

409

Table 3
Summary statistics.
This table contains summary statistics on variables included in the regression analysis. Average maturity is the dollar-weighted average maturity across tranches
within the distressed loan.
SUMMARY STATISTICS: EDF b 20
Variable

Sum

Min

Max

Median

Mean

Std. dev.

Recover
Relationship indicator
Proportion of dollars of loans
Proportion of numbers of loans
EDF
EDF*relationship indicator
EDF standard deviation
Age
Leverage
Operating profit margin
Fixed assets/assets
Sales/assets
Assets (billions)
Operating cash/interest paid
Solvent indicator
% Change in market value
Distress date loan date (30-day months)
Loan purpose: working capital
Loan purpose: corporate purposes
Loan purpose: debt repayment
Refinancing indicator
Agriculture indicator
Mining indicator
Manufacturing indicator
Financial indicator
Wholesale indicator
Retail indicator
Services indicator
Public sector indicator
Construction indicator
Transportation indicator
Chicago Fed Nat'l Activity indicator
Lagged relationship indicator
Rship*Num entities 5
Nonrship*Num entities 5
Rship*Small bank
Nonrship*Small bank
Average maturity (days)

659
659
659
659
659
659
659
659
659
659
659
659
659
659
659
659
659
659
659
659
656
659
659
659
659
659
659
659
659
659
659
659
528
503
503
440
440
606

317
368

0
0
0
0
3.53
0
0.0202
3.5428
0
1.0444
0.0083
0.1324
0.0078
81.28
0
0.9092
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1.47
0
0
0
0
0
32

1
1
0.9913
1
19.65
19.65
6.3189
35.5291
0.9043
0.6757
0.8971
4.7540
31.0330
184.58
1
3.8792
6.1000
1
1
1
1
1
1
1
0
1
1
1
1
1
1
1.47
1
1
1
1
1
4410

0
1
0.1795
0.1667
5.19
3.69
0.8701
8.2163
0.3960
0.0787
0.2475
1.1695
0.2504
1.4771
1
0.1643
1.8000
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0.11
1
0
0
0
0
1096

0.4810
0.5584
0.2746
0.3151
6.7076
3.5466
1.2147
12.7792
0.4070
0.0852
0.3125
1.3065
1.3319
4.5809
0.9514
0.0024
2.2881
0.1533
0.2671
0.3991
0.8674
0.0030
0.0577
0.4476
0.0000
0.0865
0.1062
0.1684
0.0015
0.0273
0.0986
0.0734
0.5739
0.0159
0.0219
0.0477
0.0545
1342.13

0.5000
0.4970
0.3038
0.3643
3.6631
4.1081
0.9895
9.4763
0.2499
0.1840
0.2351
0.8210
4.3435
28.2704
0.2151
0.7066
1.8886
0.3605
0.4428
0.4901
0.3394
0.0550
0.2333
0.4976
0.0000
0.2813
0.3084
0.3745
0.0390
0.1631
0.2984
0.5942
0.4950
0.1252
0.1464
0.2134
0.2273
806.61

627

101
176
263
569
2
38
295
0
57
70
111
1
18
65
303
8
11
21
24

4.4. Additional summary statistics


The last line in Table 3 lists the average loan maturity calculated as the dollar-weighted average over the tranches in the
distressed deal. Since the loan maturity is determined at the same time as or after the nature of the lending relationship, I omit
this trait from regressions to maintain econometric integrity. That the median maturity is 3 years reinforces the applicability of
the three-year timeline for evaluating long-term firm performance.
4.4.1. Loan terms
During the syndication process, the loan spread helps determine from which lender the borrower will obtain funds. Most other
loan terms are decided after the borrower commits to the lender (Standard and Poor's, 2007). Thus, the lender, and, therefore, the
lending relationship will affect a variety of loan terms. Table 4 contains three panels, each providing a comparison of loan terms
amongst relationship and non-relationship loans and distressed loans and loans prior to distress.
First, Panel A shows that for loans issued prior to distress, firms pay a significantly lower spread, provide seniority more frequently,
and abide by more financial covenants when borrowing from relationship rather than non-relationship lenders. Second, Panel B
shows that on distressed loans, firms pay a lower spread and have more financial covenants with relationship lenders than with
non-relationship lenders. However, when firms obtain non-relationship distressed loans, they provide security more often.19 Firms

19

This is consistent with Berger et al. (2011), who nd evidence consistent with collateral mitigating information gaps between borrower and lender.

410

C.M. Rosenfeld / Journal of Corporate Finance 25 (2014) 403418

Table 4
Loan characteristic t-tests.
Panel A contains results of t-tests comparing firms' relationship loan traits to their non-relationship loan traits for all loans prior to the onset of financial distress.
Panel B contains results of t-tests comparing distressed loan traits between relationship-backed (R) firms and non-relationship-backed (NR) firms. Panel C
contains results of t-tests comparing firms' distressed loan traits to their prior loan traits. All traits are tranche-dollar-weighted averages, except for indicator
measures, which take the maximum value across tranches within a loan. Spread is the all-in-spread drawn. Maturity is measured in days. Seniority and Security
are indicator variables. Covenants include only financial covenants. Refinancing is an indicator that either (a) DealScan denotes any tranche in the loan as a
refinancing or (b) any new debt incurred by the firm in the same quarter as the loan is less than 25% of the loan's value. Significance at the 1%, 5% and 10% level are
denoted by ***, **, and *, respectively.
PANEL A: PRIOR LOANS: RELATIONSHIP vs. NON-RELATIONSHIP
Variable

Relationship mean

Non-relationship mean

Difference

T-statistic

Spread
Maturity
Seniority
Secured
Number of covenants

226
243
254
179
233

166.4578
1417.97
0.9542
0.7201
1.9091

204.5602
1478.18
0.9082
0.7288
0.6247

38.1024
60.2100
0.0459
0.0087
1.2844

4.9808***
0.9425
2.5926**
0.271
11.3941***

PANEL B: DISTRESSED LOANS: RELATIONSHIP vs. NON-RELATIONSHIP


Variable

Relationship mean

Non-relationship mean

Difference

T-statistic

Spread
Maturity
Seniority
Secured
Number of covenants
Refinancing
Loan to asset book value

325 R 234 NR
341 R 265 NR
361 R 282 NR
273 R 225 NR
325 R 260 NR
367 R 289 NR
368 R 291 NR

222.5499
1315.7550
0.9669
0.8208
2.7138
0.8965
0.8462

263.3873
1376.0630
0.9562
0.9251
1.9962
0.8304
0.6619

40.8374
60.3080
0.0107
0.1043
0.7177
0.0660
0.1843

3.9073***
0.9129
0.7721
3.8173***
4.2859***
2.4825**
1.8992***

Difference

T-statistic

PANEL C: DISTRESSED LOANS vs. PRIOR LOANS


Variable

Distressed loan mean

Prior loans mean

Spread
Maturity
Seniority
Secured
Number of covenants

543
595
634
466
570

238.8959
1343.4930
0.9617
0.8753
2.4333

206.7112
1336.0310
0.9459
0.7871
1.2930

32.1847
7.4620
0.0158
0.0883
1.1403

5.3946***
0.2110
1.814*
4.7272***
13.1504***

obtain distressed refinancing from relationship lenders more often than non-relationship lenders.20 Further, Panel B shows that firms
with relationship loans have a higher loan-to-asset-book-value ratio (LTV), perhaps because more relationship distressed loans are
refinancings. So although relationship loans are issued at a higher LTV, it may be due to the greater proportion of loans that largely
refinance existing debt rather than predominantly create new debt.
Third, Panel C shows, not surprisingly, that firms pay a higher loan spread and more frequently provide seniority and security
to their lenders when they receive distressed funding than in prior loans. Further, distressed firms are subject to more financial
covenants than prior loans. Thus, the terms on these distressed loans reflect that the lenders are facing increased risk. One
element that does not vary between relationship and non-relationship loans, or between distressed and prior loans is loan
maturity, while every other element has a significant difference depending on firm financial circumstances or the source of the
funds. Consequently, it is evident that lending relationships affect loan terms, which may, in turn, affect firm performance. All in
all, firms face both costs and benefits when obtaining relationship funding.
4.5. Estimation procedure
I use the following probit model to determine the effect that each of the above defined variables has on the future of financially
distressed firms:
PrEmergence rship Xrship dis Xdis firm Xfirm loan Xloan industry Xindustry econ Xecon

where the subscripts refer to relationship, degree of distress, firm, loan, industry, and macroeconomic attributes, respectively.
Results from probit regressions on an indicator of future firm emergence from distress are listed in Table 5. The figures adjacent to
the control variable are coefficients, with standard deviations below in parentheses. I report marginal effects in the column
adjacent to the coefficient. All regressions are performed with robust errors at the firm level.
20
Renancing is identied in two ways. First, a loan is deemed a renancing if DealScan categorizes any of the distressed tranches as a renancing. The second
way a loan is identied as a renancing is through comparison of the rm's quarterly debt levels. If the loan creates new debt of less than 25% of the loan value,
according to quarterly changes in long-term debt, the loan is considered a renancing.

C.M. Rosenfeld / Journal of Corporate Finance 25 (2014) 403418

411

Table 5
Probit regression results.
Predicting the probability of future emergence from distress. The table below shows probit regression coefficients with their standard errors below in parentheses
and marginal effects beside the coefficients. All errors are robust at the firm level. Adjusted Mean Z-Statistic is the mean z-statistic for Norton, Wang & Ai's
adjusted interaction term within logit and probit regressions. Significance at the 1%, 5% and 10% level are denoted by ***, **, and *, respectively.
PROBIT REGRESSIONS WITH MARGINAL EFFECTS
1

Refinancing
All distressed
Relationship indicator
Expected default frequency
EDF*Rs hip indicator
Adjusted Mean Z-Statistic
EDF standard deviation
Firm
Age
Leverage
Operating profit margin
Fixed assets/total assets
Net sales/assets
Assets (billions)
Operating cash/interest paid
Solvent indicator
% Change in market value
Loan attributes
Months from loan to distress
Purpose: working capital
Purpose: Corporate purposes
Purpose: debt repayment
Industry
Manufacturing indicator
Transportation indicator
Retail indicator
Services indicator
Macroeconomy
CFNAI
Year indicators
Refinancing indicator
Number of observations
Pseudo R-squared
Log pseudolikelihood

0.4328**
(0.1747)
0.0341**
(0.0153)
0.0338**
(0.0158)
2.0274
0.1222***
(0.0473)
0.0004
(0.0057)
0.1179
(0.2553)
0.3056
(0.2831)
0.1638
(0.2380)
0.0148
(0.0679)
0.0094
(0.0150)
0.0005
(0.0018)
0.3732**
(0.1791)
0.1045
(0.0774)

Marginal effect
0.1706

0.3356
(0.2789)

Marginal effect
0.1295

0.0136
0.0135

0.0486

0.0002
0.0469
0.1216
0.0652
0.0059
0.0038
0.0002
0.1447
0.0416

0.0054
(0.0280)
0.0076
(0.1686)
0.0301
(0.1439)
0.0747
(0.1309)

0.0021

0.3380**
(0.1372)
0.6414***
(0.2069)
0.1993
(0.1864)
0.3645**
(0.1662)

0.1336

0.2690**
(0.1175)
Yes

0.1070

806
0.1293
485.28

EDF = 20

0.0030
0.0120
0.0297

0.2381
0.0785
0.1419

0.2842***
(0.0828)
0.0074
(0.0172)
1.1339*
(0.6876)
1.3255
(0.8942)
1.1965
(0.8436)
0.0629
(0.1779)
0.0127
(0.0420)
0.0010
(0.0076)
0.3019
(0.3128)
0.1987
(0.1813)
0.0497
(0.0948)
0.1223
(0.4642)
0.2426
(0.4132)
0.5633
(0.4022)

0.1094

0.0028
0.4363
0.5101
0.4604
0.0242
0.0049
0.0004
0.1155
0.0765

0.0191
0.0465
0.0915
0.2178

0.4331
(0.3981)
0.3517
(0.7414)
0.0361
(0.4705)
0.1818
(0.4886)

0.1636

1.6822***
(0.4100)
Yes

0.6473

147
0.3703
63.21

0.1277
0.0139
0.0709

EDF b 20
0.5776**
(0.2274)
0.0190
(0.0276)
0.0636**
(0.0292)
2.0883
0.0883
(0.0911)
0.0004
(0.0063)
0.1665
(0.3004)
0.1485
(0.2918)
0.1977
(0.2562)
0.0204
(0.0770)
0.0042
(0.0155)
0.0001
(0.0020)
0.4527*
(0.2724)
0.1352
(0.0850)
0.0153
(0.0302)
0.0272
(0.1891)
0.0249
(0.1586)
0.0410
(0.1449)

Marginal effect EDF b 20


0.2265
0.0076
0.0253

0.0352

0.0002
0.0663
0.0592
0.0788
0.0081
0.0017
0.0000
0.1734
0.0539

0.0061
0.0109
0.0099
0.0163

0.2737*
(0.1532)
0.6581***
(0.2157)
0.1939
(0.2124)
0.4106**
(0.1831)

0.1086

0.1507
(0.1264)
Yes

0.0600

659
0.1265
398.57

0.2458
0.0766
0.1599

0.5840**
(0.2283)
0.0225
(0.0277)
0.0643**
(0.0292)
2.1005
0.0750
(0.0915)
0.0002
(0.0063)
0.1477
(0.3008)
0.1462
(0.2954)
0.2521
(0.2592)
0.0113
(0.0774)
0.0053
(0.0158)
0.0002
(0.0020)
0.4604*
(0.2735)
0.1293
(0.0856)
0.0120
(0.0303)
0.0293
(0.1912)
0.0423
(0.1596)
0.0320
(0.1523)

Marginal effect
0.2288
0.0090
0.0256

0.0299

0.0001
0.0588
0.0583
0.1004
0.0045
0.0021
0.0001
0.1760
0.0515

0.0048
0.0117
0.0169
0.0127

0.2748*
(0.1538)
0.6852***
(0.2186)
0.2247
(0.2131)
0.4082**
(0.1833)

0.1090

0.1429
(0.1278)
Yes
0.1289
(0.4856)
656
0.1286
395.79

0.0569

0.2544
0.0886
0.1589

0.0511

5. Results
Columns 1 through 6 of Table 5 show one main result: obtaining funds from a relationship lender positively affects the future
performance of all distressed firms, except when analyzing only the most severely distressed firms. When eliminating the most

412

C.M. Rosenfeld / Journal of Corporate Finance 25 (2014) 403418

severely distressed firms from the sample (those with an EDF of 20), obtaining a relationship-backed loan still significantly
positively affects the future performance of financially distressed firms. Specifically, when firms obtain distressed funding from a
prior lender rather than a new lender, it increases the probability of future firm emergence from distress by approximately
22.65%. At the mean, this decreases the probability of sustained or worsened distress by about 43.6% (0.2265/(10.4810) 43.6%).
This is consistent with relationship lenders using their informational advantages while extending credit to financially distressed
firms.
Further, for both the entire sample as well as the sample omitting the most severely distressed firms, the interaction term of
the lending relationship and degree of distress is significantly negative. Thus, this positive effect of procuring distressed funding
from a relationship lender diminishes as the degree of firm distress increases. This finding is consistent with the results in
Columns 3 and 4, which analyze only the most severely distressed firms, where there is an insignificant effect of relationships on
the future performance of such firms.21 Using Norton, Wang & Ai's inteff correction, we see that for the entire sample and sample
omitting the most severely distressed firms, the mean z-statistics of this interaction variable is below 2, so is significant at the
5% level. Economically this can be interpreted as a one-standard deviation increase in EDF (of 3.66) will decrease the positive
effect of obtaining distressed funding from a relationship lender on the probability of emerging from distress by approximately
2.53%.22
These findings also support that the comparative advantage experienced by relationship lenders does not span the entire
distress spectrum. Rather, when dealing with severely distressed borrowers, relationship lenders still extend credit even though
they cannot expect to gain from their informational advantages. This is consistent with Weinstein and Yafeh's (1998) findings
that Japanese relationship lenders provide credit to their distressed borrowers and is inconsistent with Elsas and Krahnen's
(1998) findings that German Hausbanks provide liquidity insurance to firms, but only through moderate distress.
As shown in Panel B of Table 4, a significantly higher proportion of relationship-backed distressed loans are refinancings than
for non-relationship-backed distressed loans. Columns 7 and 8 of Table 5 report results for the same specification as in Columns 5
and 6 with an additional variable: the refinancing indicator. Note that this indicator is insignificant, and the other coefficients and
their significance remain largely unchanged. Specifically, there is no evidence of a significant change in the relationship coefficient
when including the refinancing indicator. Thus, I omit this variable in the remaining analysis.
6. Robustness
Since Table 5 shows that lending relationships significantly affect future firm performance only for the sample omitting the
most severely distressed firms, I will report robustness checks for only that sample.
6.1. Alternative relationship measures
To this point I have used an indicator to measure banking relationships. Now I introduce two alternative approaches to
measuring banking relationships: the proportion of (1) dollars of loans and (2) number of tranches the lead lender lent to the firm
over the five years prior to the distressed loan. Both continuous measures are similar to those in Bharath et al. (2007) who use the
proportion of prior loans to the distressed firm in terms of (1) dollars and (2) number of tranches issued by any lead lender on the
distressed tranche.23 In particular, I use the maximum value across lead lenders and distressed tranches as the relationship
measure.24 Columns 1 and 2 (3 and 4) of Table 6 report probit results using the continuous measures of the proportion of prior
loan dollars (number of loans) to measure relationships. Again, I find that there is a significantly positive effect on the future
performance of financially distressed firms that obtain loans from relationship lenders. Further, the interaction of each continuous
relationship measure and the firm's EDF remains significantly negative.
6.2. Lender frequency
Some might claim that there's a distinction between repeated lending and relationship lending, and that the difference
depends on the lender's loan market activity. To address this issue, I first identify the most active DealScan lenders. I follow the
methodology of Bharath et al. (2007) to proxy for dollars lent by each lead lender: I divide the loan dollar amount by the number
of lead lenders.25 I find that four banks dominate the syndicated loan market: Bank of America, Citi, Chase Manhattan, and JP

21

This result is qualitatively similar when omitting the year controls, as may be merited by the small sample size.
Recall, this effect accumulates as degree of distress distances itself further from the mean value of 6.71.
23
Bharath et al. (2007) also limit their rm history to the prior ve years.
24
There is a high correlation of 0.8667 between the two continuous measures, which weigh the proportion of prior loans by dollars and number of tranches, and
there is a strongly positive correlation of 0.8063 (0.7708) between the indicator relationship measure and the proportion of prior dollars (number) of loans
measure. The correlation coefcients between the continuous measures and the indicator measure are lower than Bharath et al. (2007) nd. Further, they nd a
slightly stronger correlation between their continuous measures. These differences may be attributed to the different nature of lending relationships between
lenders and nancially distressed rms or because Bharath et al. evaluate only DealScan's top 20 most-active lenders.
25
Though this methodology is an inaccurate determinant for the actual amount of the loan each lender retained, this measure captures the lender's activity in
the syndicated loan market and identies the most frequent participants.
22

C.M. Rosenfeld / Journal of Corporate Finance 25 (2014) 403418

413

Table 6
Robustness checks: Probit regressions with marginal effects.
Predicting the probability of future emergence from distress. The table below shows probit regression coefficients with their standard errors below in parentheses
and marginal effects beside the coefficients. Only observations with an EDF less than 20 are included in the underlying samples. All errors are robust at the firm
level. Columns 1 and 2 (Columns 3 and 4) show results from measuring the lending relationship with the proportion of prior funds (number of loans) the lead
lender on the distressed loan lent as a lead lender throughout the five years prior to distress, as covered by DealScan. Columns 5 & 6 exclude relationships with the
top four lenders throughout the sample history (Chase, Morgan, Bank of America, and Citi), and the sample excludes loans where only those four most active
lenders were the lead lenders. Columns 7 & 8 omit from the sample those firms that merge in the three years following the identification of the firm as financially
distressed. Adjusted Mean Z-Statistic is the mean z-statistic for Norton, Wang & Ai's adjusted interaction term within logit and probit regressions. Significance at
the 1%, 5% and 10% level are denoted by ***, **, and *, respectively.
PROBIT REGRESSIONS WITH MARGINAL EFFECTS

Relationship measure
Expected default frequency
EDF*Rship measure
Adjusted mean Z-statistic
EDF standard deviation
Firm
Age
Leverage
Operating profit margin
Fixed assets/total assets
Net sales/assets
Assets (billions)
Operating cash/interest paid
Solvent indicator
% Change in market value
Loan attributes
Months from loan to distress
Purpose: Working capital
Purpose: corporate purposes
Purpose: debt repayment
Industry
Manufacturing indicator
Transportation Indicator
Retail indicator
Services indicator
Macroeconomy
CFNAI
Year indicators
Number of observations
Pseudo R-squared
Log pseudolikelihood

% $ Rship
loans

Marginal
effect

% # Rship
loans

Marginal
effect

Non-dominant

Marginal
effect

No mergers

Marginal
effect

1.0146***
(0.3652)
0.0139
(0.0254)
0.1153***
(0.0448)
2.3784
0.1003
(0.0905)
0.0015
(0.0063)
0.1628
(0.3017)
0.1560
(0.2907)
0.1856
(0.2572)
0.0275
(0.0771)
0.0028
(0.0156)
0.0000
(0.0020)
0.4553*
(0.2728)
0.1364
(0.0846)
0.0144
(0.0302)
0.0415
(0.1896)
0.0410
(0.1589)
0.0641
(0.1454)
0.2836*
(0.1541)
0.6623***
(0.2171)
0.1877
(0.2130)
0.4193**
(0.1846)
0.1477
(0.1260)
Yes
659
0.1285
397.66

0.4043
0.0055
0.0459

0.0400

0.0006
0.0649
0.0622
0.0739
0.0109
0.0011
0.0000
0.1744
0.0544

0.0057
0.0165
0.0163
0.0255

0.1125
0.2472
0.0742
0.1631

0.0588

0.8886***
(0.3161)
0.0114
(0.0253)
0.0962**
(0.0399)
2.2439
0.1000
(0.0898)
0.0017
(0.0063)
0.1860
(0.2984)
0.1644
(0.2920)
0.1903
(0.2570)
0.0336
(0.0772)
0.0047
(0.0155)
0.0001
(0.0021)
0.4623*
(0.2703)
0.14*
(0.0849)
0.0124
(0.0302)
0.0362
(0.1900)
0.0413
(0.1592)
0.0758
(0.1449)
0.2737*
(0.1542)
0.6664***
(0.2160)
0.1770
(0.2130)
0.4214**
(0.1849)
0.1482
(0.1265)
Yes
659
0.1289
397.50

0.3541
0.0045
0.0383

0.0399

0.0007
0.0741
0.0655
0.0758
0.0134
0.0019
0.0001
0.1769
0.0558

0.0049
0.0144
0.0165
0.0302

0.1086
0.2486
0.0700
0.1639

0.0590

0.7317***
(0.2808)
0.0276
(0.0314)
0.0655*
(0.0350)
1.7185
0.0758
(0.1039)
0.0070
(0.0079)
0.1389
(0.3926)
0.1220
(0.4085)
0.1893
(0.3227)
0.0097
(0.0953)
0.0042
(0.0289)
0.0016
(0.0025)
0.4914
(0.3257)
0.1669
(0.1070)
0.0238
(0.0376)
0.1371
(0.2310)
0.2319
(0.2085)
0.0850
(0.1897)
0.3816**
(0.1868)
1.0706***
(0.2833)
0.6837***
(0.2587)
0.3638
(0.2257)
0.3954**
(0.1623)
Yes
442
0.1730
252.52

0.2850
0.0110
0.0260

0.0301

0.0028
0.0552
0.0485
0.0753
0.0038
0.0017
0.0006
0.1856
0.0664

0.0095
0.0542
0.0913
0.0338

0.1504
0.3604
0.2521
0.1414

0.1572

0.6022**
(0.2503)
0.0349
(0.0307)
0.0711**
(0.0320)
2.0359
0.1011
(0.1019)

0.2278
0.0135
0.0275

0.0391

0.0020
(0.0069)
0.4617
(0.3490)
0.0391
(0.3019)
0.0984
(0.2840)
0.0380
(0.0848)
0.0095
(0.0156)
0.0007
(0.0021)
0.5104*
(0.2849)
0.1069
(0.0907)

0.0008

0.0380
(0.0331)
0.1000
(0.2040)
0.0201
(0.1724)
0.0574
(0.1563)

0.0147

0.2825*
(0.1664)
0.6809***
(0.2341)
0.1848
(0.2280)
0.4367**
(0.2020)
0.0683
(0.1457)
Yes
579
0.1823
322.03

0.1787
0.0151
0.0381
0.0147
0.0037
0.0003
0.1804
0.0414

0.0390
0.0078
0.0222

0.1086
0.2345
0.0700
0.1603

0.0264

414

C.M. Rosenfeld / Journal of Corporate Finance 25 (2014) 403418

Morgan Chase.26 Using this measure, over the sample's time span of January 1994 through August 2007, each of these four lenders
was allocated over $1 trillion in loans, while the next most active lender was allocated less than $400 billion.
With such market dominance, it is reasonable to question whether these four lenders can have relationships with borrowers,
as opposed to merely being the repeated source of funds. Further, since these lenders are so dominant in the sample, it could be
that the results so far are largely indicative of only these dominant banks, a term used by Ross (2010). To determine if the
positive effect of lending relationships on the future of moderately financially distressed firms holds for the non-dominant banks,
I remove from the initial sample universe all loans that are led solely by the dominant banks. Since McCahery and Schwienbacher
(2010) claim that these very active, or top tier, banks self-select the best borrowers, eliminating these lenders from the sample
may also reduce sample endogeneity issues. For the remaining loans, I only allow the lending history with non-dominant banks
to be captured by the relationship indicator. The results from probit analysis using this measure of relationships are in Columns 5
and 6 of Table 6. This modified relationship indicator still has a significantly positive effect on the future performance of
financially distressed firms and a negative effect of the interaction of the lending relationship and firm EDF.
6.3. Mergers
Some people consider whole firm acquisition by way of a merger a distinct event from a firm remaining its own entity or
permanently delisting, both of which have a more observable outcome. For this reason, I omit from the sample all firms that
merge in the three years following identification of distress and repeat the probit analysis on this subsample. Results from this
analysis are reported in Columns 7 and 8 of Table 6. The main result still holds: obtaining distressed funding from a relationship
lender has a significantly positive effect on future firm performance, and this effect decreases as the degree of distress increases.
7. Mechanism
To this point, the analysis has shown that obtaining relationship-backed distressed funding has a positive effect on the future
performance of financially distressed firms and that this effect decreases as degree of firm distress increases. Now I search for
evidence of the attribute of lending relationships that creates this value for financially distressed firms.
7.1. Methodology
A mechanism is a conduit through which the lending relationship creates value. When present in non-relationship lenders, the
mechanism creates the same effect on future firm emergence from distress as obtaining funds from a relationship lender. When
present in a relationship lender, this attribute should have no further significant effect on future emergence from distress, as the
benefit of this attribute should be embedded in the relationship measure. With this in mind, I create two indicator variables
denoting that the loan is relationship- or non-relationship-backed. I then interact these variables with the mechanism variable. I
will estimate the model:27
PrEmergence rship Xrship dis Xdis firm Xfirm loan Xloan industry Xindustry econ Xecon
rshipmechanism Xrshipmechanism nonrshipmechanism Xnonrshipmechanism

where the first six betas and set of controls are the same as in the probit analysis, and the new variables represent the interaction
of the relationship (non-relationship) indicator and the mechanism. A viable mechanism should have a significantly positive
nonrship*mechanism and an insignificant rship*mechanism.
7.2. Mechanism
The literature commonly hypothesizes that relationship lending establishes its value through reductions of information
asymmetry. One specific facet of reduction in information asymmetry is the ability to transmit qualitative information from the
loan manager to the lender's syndicated lending group, which typically operates at the highest level within the lending
organization. Stein (2002) finds that A decentralized approachwith small, single-manager firmsis most likely to be attractive
when information about projects is soft and cannot be credibly transmitted. In contrast, large hierarchies perform better when
information can be costlessly hardened and passed along inside the firm (p. 1891).
To test the influence of the ease of information transmission within the lending organization, I use two different measures. The
first controls for the complexity of the lending organization as measured by the number of entities within the lender's
highest-level holding company. The idea is simple: the more entities there are within an organization, the more likely it is to have
impediments with conveying soft information to higher levels of the hierarchy.28 Columns 1 and 2 of Table 7 show regression
26
The lending decision on syndicated loan is usually made across banks within a holding company. Thus, all lender analysis and relationship formation takes
place at the bank's highest holder position. Chase Manhattan was an incredibly active lender prior to merging with JP Morgan in 2000.
27
Due to the multiple interaction terms in this regression, it is not possible to perform the Wang, Norton & Ai correction for interaction terms. Results hold when
using a linear probability model.
28
Liberti and Mian (2009) nd that greater hierarchical/geographical distance between the information collecting agent and the loan approving ofcer leads to
less reliance on subjective information and more on objective information (p. 4057).

C.M. Rosenfeld / Journal of Corporate Finance 25 (2014) 403418

415

Table 7
Mechanism regression results.
This table shows probit regression coefficients with their standard errors below in parentheses. All errors are robust at the firm level. The dependent variable is
future firm emergence from distress. Rship*Mechanism (Nonrship*Mechanism) is the interaction of an indicator that the distressed loan is from a relationship
(non-relationship) lead lender and the mechanism named in the column header. The Wald Test chi-squared values and p-values reflect the difference between
the coefficients on Rship*Mechanism and Nonrship*Mechanism. Significance at the 1%, 5% and 10% level are denoted by ***, **, and *, respectively.
PROBIT REGRESSIONS WITH MARGINAL EFFECTS

Relationship indicator
Expected default frequency
EDF*Rship indicator
EDF standard deviation
Mechanism
Rship*Mechanism
Nonrship*Mechanism

Wald Test chi-squared


Wald Test p-value
Firm controls
Loan attributes
Industry indicators
CFNAI
Year indicators
Number of observations
Pseudo R-squared
Log pseudolikelihood

Num entities 5

Marginal effect

Small bank

Marginal effect

0.6811**
(0.2695)
0.0254
(0.0345)
0.0700**
(0.0349)
0.0894
(0.1125)

0.2646

0.8239***
(0.2939)
0.0280
(0.0368)
0.0788**
(0.0373)
0.0877
(0.1158)

0.3174

0.6995
(0.6130)
0.7819*
(0.4429)
3.64
0.0563
Yes
Yes
Yes
Yes
Yes
503
0.1377
300.57

0.0101
0.0279
0.0357

0.2570
0.2867

0.1846
(0.3289)
0.7060**
(0.3135)

0.0112
0.0314
0.0350

0.0734
0.2622

3.91
0.0479
Yes
Yes
Yes
Yes
Yes
440
0.1346
263.94

results, including marginal effects, when using an indicator to denote that the greater lending organization is comprised of five or
fewer entities. Results are similar when using an indicator for 10 or fewer entities.29 Column 1 of Table 7 shows that there is a
significantly positive effect on firm emergence from distress when obtaining distressed funding from a non-relationship lender
with five or fewer entities within its lending organization. Further, this effect is significantly different from the effect of obtaining
funds from a lower-complexity relationship lender. Note that the effect of obtaining distressed funds from a relationship lender
is still significantly positive, and this effect diminishes as the degree of firm distress increases. Further, the marginal effects on
these variables are similar in magnitude to those reported in the original specification in Column 6 of Table 5. In Wald test
results not reported, there is no significant difference in the coefficients of the relationship indicator and the interaction of the
non-relationship indicator and mechanism. Thus, this evidence is consistent with the value of lending relationships between
financially distressed firms and their lenders being attributable to information asymmetry and the ability to transmit the
informational advantages within the lending organization.
The second way I measure complexity of the lending organization is through asset size of the lending organization. Even
if there are few entities within a lending organization, the mere size of the institution may prevent ease of information
transmission. Berger et al. (2005) study loans to small American firms to find that larger banks are more likely to communicate
impersonally with their borrowing firms, which is consistent with larger lenders having impediments for transmitting
information easily. To test this hypothesis in this sample, I rank all the distressed lenders' organizations' assets and denote with an
indicator whether the lending organization is in the smallest ten percent. Results using this measure are reported in Columns 3
and 4 of Table 7. Again, I find evidence consistent with the value of lending relationships to distressed firms being the ability to
convey information within the lending organization. The effect of the relationship indicator is still significantly positive, and
this effect significantly decreases as the severity of firm distress increases. In results not reported here, Wald tests show
no significant difference between the relationship coefficient and the coefficient on the interaction of the non-relationship
indicator and small lender indicator. Combined, this set of results supports that obtaining distressed funds from a small lender
that is not a relationship lender can provide a distressed firm statistically similar effect on future firm performance as obtaining
relationship-backed funding.

29
There are only nine single-entity organizations in the sample, which is too small to draw conclusions on and, therefore, is the reason why an indicator of
single-entity organizations is not analyzed here.

416

C.M. Rosenfeld / Journal of Corporate Finance 25 (2014) 403418

8. Endogeneity
The main result from the analysis thus far indicates that when it obtains funding from a relationship lender, a financially
distressed firm is more likely to emerge from distress. An important issue to address is whether it is banking relationships that
help firms perform better or that better firms obtain relationship funding. To further investigate this matter, I use bivariate probit
estimation on the indicator relationship.
Adapting Greene's (2008 p. 817) model, the specification for this two-equation model is:


emerge x1 1 1 ;

emerge 1 if emerge N 0; 0 otherwise;

relationship x2 2 2 ; relationship 1 if relationship N 0; 0 otherwise;

E 1 jx1 ; x2  E 2 jx1 ; x2  0;
Var 1 jx1 ; x2  Var 2 jx1 ; x2  1;
Cov1 ; 2 jx1 ; x2  :

Bivariate probit regressions simultaneously estimate two equations: Eq. (3) predicts future performance given an exogenous
(actual) relationship indicator and the remaining previous controls while Eq. (4) predicts the nature of the banking relationship
using the same controls as the first equation as well as an instrument to identify the relationship. Because of inability to
disentangle the interaction of the potentially endogenous variable (lending relationship) and the degree of financial distress
(EDF) while controlling for the endogeneity of the lending relationship, I limit the sample to only those distressed firms where the
effect of obtaining a relationship loan should be positiveit is on this subset where endogeneity issues should be strongest. Since
the findings to this point show an overall positive effect of lending relationships that decreases as the severity of distress
increases, the distressed firms that benefit most from relationship-backed lending are those with lower EDFs. Thus, for studying
the presence of endogeneity, I limit the sample to those observations with an EDF less than the sample mean EDF value of 6.70
and omit the interaction term of the lending relationship and EDF.
One test statistic of interest in bivariate probit analysis is rho (), which is the correlation coefficient between the residuals
from the two simultaneous probit estimations. When rho () is significantly different from zero, it is prudent to use this
simultaneous estimation because there is evidence of correlation between the error terms. When rho () is insignificant, random
shocks to the second equation, which predicts lending relationship, have no effect on the simultaneously predicted future firm
success, so lending relationships are considered exogenous.

Table 8
Bivariate probit regression results.
The table below shows regression results from bivariate probit regressions. Standard errors are in parentheses beneath the coefficients. All errors are robust at
the firm level. There are two simultaneous probit stages: the first predicts future firm emergence from distress with an exogenous relationship indicator, and the
second predicts the relationship with the same controls as well as an identifying instrument. All loan attribute controls are included in this regression with the
exception of the working capital loan purpose indicator, which was omitted to avoid quasi-complete separation. Significance at the 1%, 5% and 10% level are
denoted by ***, **, and *, respectively.
BIVARIATE PROBIT REGRESSION

Relationship indicator
Expected default frequency
EDF standard deviation

Emerge

Relationship

1.6651***
(0.1032)
0.1294
(0.0854)
0.1653
(0.1949)

Lagged relationship indicator


Constant
Firm controls
Loan attributes
Industry indicators
CFNAI
Year indicators
Number of observations
Log pseudolikelihood
Rho
Rho chi-squared p-value

Yes
Yes
Yes
Yes
Yes
No
349
420.52
0.5550
0.4563

0.3071***
(0.0971)
0.3528
(0.2244)
0.2313**
(0.1163)
Yes
Yes
Yes
Yes
Yes
No

C.M. Rosenfeld / Journal of Corporate Finance 25 (2014) 403418

417

8.1. Instrument
I use a lagged lending relationship indicator to identify the distressed lending relationship. Seeking funding from a new lender
is costly to firms, since it requires creating and marketing a wide array of financial documents. Thus, it is rational for firms to
consistently seek funds from a limited selection of lenders. This repeated borrowing can lead to efficiency gains in firmlender
contract creation, making obtaining new funds less of an ordeal from both the borrower and lender's perspective. Another reason
firms may continually seek funds from relationship lenders is to borrow at the lower spreads that relationship lenders offer, as
documented in Table 4. Further, it is possible that firms are aware of the benefits of lending relationships if the firm encounters
distress (i.e. lower loan spreads, less frequent security requirement and higher LTV ratios) and need the services of a
comparatively larger bank (nullifying the option to replicate the value of lending relationships through seeking non-relationship
funding through a comparatively smaller bank), so they may continually rely upon relationship lenders to provide their funds. In
these scenarios, a history of obtaining funds from a prior lender may indicate a firm's general tendency to obtain funds from prior
lenders. At the same time, the historical reliance upon prior lenders as a source of funds does not reflect the ability of a firm to
emerge from financial distress.
To test this hypothesis, I use as an instrument the relationship indicator from the most recent loan prior to the distressed loan.
It is important to note that the lagged relationship indicator does not reflect that the loan prior to the distressed loan is from one
of the lead lenders on the distressed loan, but rather that one of the lead lenders on the loan prior to the distressed loan had been a
lead lender to the firm on a loan within the five years prior to that. I anticipate that a firm with a history of borrowing from a
relationship lender will be more likely to find distressed funding from a prior lender.
8.2. Results
Table 8 shows coefficients and standard errors from the bivariate probit regression. The lagged relationship indicator is a
significantly positive predictor of lending relationships. That is, a borrower with a history of obtaining relationship-backed funds
is more likely to obtain distressed funds from a relationship lender. It is also evident that rho (), the correlation of unpredicted
error terms, is insignificant. Combining the findings, when using this conceptually acceptable instrument that is a significant
predictor of lending relationships, there is no evidence of endogeneity. Thus, it is appropriate to revert to the simple probit
analysis for conclusions on the effect that lending relationships have on the future performance of financially distressed firms.
9. Conclusion
In this study I empirically examine the effect that banking relationships have on the probability of future emergence from
distress for publicly traded, financially distressed U.S. firms. Results demonstrate that obtaining a relationship-backed, rather than
non-relationship-backed, loan in the six months prior to distress identification adds value to financially distressed firms by
significantly increasing the probability of future firm emergence from distress. Additional evidence shows that as firm distress
increases, obtaining a relationship loan, rather than a non-relationship loan, decreases the probability of future firm emergence
from distress. That is, relationship lenders still provide credit to severely financially distressed firms, even though they cannot
expect to gain from their informational advantage.
These results persist after several robustness checks including the changing of the measure of lending relationships, omitting
dominant lenders from the sample, omitting the firms that merge from the sample, and controlling for the endogeneity of
determining the nature of the banking relationships backing loans. Further, I provide evidence consistent with the value of
banking relationships stemming from the ease of transmission of information within the lending organization.
Acknowledgments
I am grateful to Luca Benzoni, Bob DeYoung, Murray Frank, Ross Levine, David Smith, Haluk Unal and Andy Winton for their
invaluable input, and I thank Gjergji Cici, Jack Kareken, Zining Li, Huiyan Qiu and seminar participants at FDIC, Federal Reserve
Board of Governors, OEA at SEC, University of Minnesota, FDIC's Bank Research Conference (2007), FMA (2006), FMA Doctoral
Student Seminar (2005), Mid-Atlantic Research Conference (2006), Bocconi University's Conference on Banking Regulation and
Supervision (2008), and anonymous referees for helpful comments. All errors are mine.
References
Ai, Chunrong, Norton, Edward C., 2003. Interaction terms in logit and probit models. Econ. Lett. 80, 123129.
Andrade, Gregor, Kaplan, Steven, 1998. How costly is financial (not economic) distress? Evidence from highly leveraged transactions that became distressed.
Journal of Finance 14431494.
Berger, Allen N., Udell, Gregory F., 1995. Relationship lending and lines of credit in small firm finance. J. Bus. 68, 351381.
Berger, Allen N., Miller, Nathan H., Petersen, Mitchell A., Rajan, Raghuram G., Stein, Jeremy C., 2005. Does function follow organizational form? Evidence from the
lending practices of large and small banks. J. Financ. Econ. 76, 237269.
Berger, Allen N., Klapper, Leora F., Peria, Maria Soledad Martinez, Zaidi, Rida, 2008. Bank ownership type and banking relationships. J. Financ. Intermed. 17, 3762.
Berger, Allen N., Espinosa-Vega, Marco A., Frame, W. Scott, Miller, Nathan H., 2011. Why do borrowers pledge collateral? New empirical evidence on the role of
asymmetric information. J. Financ. Intermed. 20, 5570.

418

C.M. Rosenfeld / Journal of Corporate Finance 25 (2014) 403418

Bharath, Sreedhar, Dahiya, Sandeep, Saunders, Anthony, Srinivasan, Anand, 2007. So what do I get? The bank's view of lending relationships. J. Financ. Econ. 85,
368419.
Bharath, Sreedhar, Dahiya, Sandeep, Saunders, Anthony, Srinivasan, Anand, 2011. Lending relationships and loan contract terms. Rev. Financ. Stud. 24,
11411203.
Carey, Mark, Post, Mitch, Sharpe, Steven A., 1998. Does corporate lending by banks and finance companies differ? Evidence on specialization in private debt
contracting. Journal of Finance 845878.
Cole, Rebel A., 1998. The importance of relationships to the availability of credit. J. Bank. Financ. 22, 959977.
Cotugno, Matteo, Monferra, Stefano, Sampagnaro, Gabriele, 2012. Relationship lending, hierarchical distance and credit tightening: evidence from the financial
crisis. J. Bank. Financ. 37, 13721385.
Dahiya, Sandeep, John, Kose, Puri, Manju, Ramrez, Gabriel, 2003a. Debtor-in-possession financing and bankruptcy resolution: empirical evidence. J. Financ. Econ.
69, 259280.
Dahiya, Sandeep, Saunders, Anthony, Srinivasan, Anand, 2003b. Financial distress and bank lending relationships. J. Financ. 58, 375399.
Degryse, Hans, Ongena, Steven, 2005. Distance, lending relationships, and competition. Journal of Finance 60, 231266.
Dichev, Ilia D., Skinner, Douglas J., 2002. Large-sample evidence on the debt covenant hypothesis. J. Account. Res. 40, 10911123.
Elsas, Ralf, 2005. Empirical determinants of relationship lending. J. Financ. Intermed. 14, 3257.
Elsas, Ralf, Krahnen, Jan Pieter, 1998. Is relationship lending special? Evidence from credit-file data in Germany. J. Bank. Financ. 28, 12831316.
Fan, Joseph P.H., Huang, Jun, Zhu, Ning, 2013. Institutions, ownership structure, and distress resolution in China. J. Corp. Financ. 23, 7187.
Frank, Murray Z., Goyal, Vidhan K., 2007. Capital structure decisions: which factors are reliably important? October 10 Working Paper (Available at SSRN: http://ssrn.
com/abstract=567650).
Gilson, Stuart C., John, Kose, Lang, Larry H.P., 1990. Troubled debt restructurings. J. Financ. Econ. 28, 325353.
Gonzales, Laura, James, Christopher, 2007. Banks and bubbles: how good are bankers at spotting winners? J. Financ. Econ. 86, 4070.
Greene, William H., 2008. Econometric Analysis, Sixth edition. Pearson Prentice Hall, New Jersey.
Houston, Joel F., James, Christopher M., 1996. Bank information monopolies and the mix of private and public debt claims. Journal of Finance 51, 18631889.
Houston, Joel F., James, Christopher M., 2001. Do relationships have limits? Banking relationships, financial constraints, and investment. J. Bus. 74, 347374.
James, Christopher, 1987. Some evidence on the uniqueness of bank loans. J. Financ. Econ. 19, 217235.
Jiangli, Wenying, Unal, Haluk, Yom, Chiwon, 2008. Relationship lending, accounting disclosure, and credit availability during the Asian financial crisis. J. Money
Credit Bank. 40, 2555.
Liberti, Jose M., Mian, Atif R., 2009. Estimating the effect of hierarchies on information use. Rev. Financ. Stud. 22, 40574090.
Ljungqvist, Alexander, Marston, Felicia, Wilhelm Jr., William J., 2006. Competing for securities underwriting mandates: banking relationships and analyst
recommendations. J. Financ. 61, 301340.
McCahery, Joseph, Schwienbacher, Armin, 2010. Bank reputation in the private debt market. J. Corp. Financ. 16, 498515.
Merton, Robert C., 1974. On the pricing of corporate debt: the risk structure of interest rates. J. Financ. 449470.
Moody's KMV, 2004. Measuring & Managing Credit Risk: Measuring the EDFTM Credit Measure for Public Firms. PowerPoint presentation, www.moodyskmv.
com.
Ongena, Steven R.G., Smith, David C., 2000. Bank relationships: a review. In: Harker, Patrick T., Zenios, Stavros A. (Eds.), Performance of Financial Institutions:
Efficiency, Innovation, Regulation. Cambridge University Press, Cambridge, pp. 221258.
Ongena, Steven, Smith, David, 2001. The duration of bank relationships. J. Financ. Econ. 61, 449475.
Ongena, Steven, Tumer-Alkan, Gunseli, Vermeer, Bram, 2011. Corporate choice of banks: decision factors, decision maker, and decision processfirst evidence.
J. Corp. Financ. 17, 326351.
Petersen, Mitchell A., 1999. Banks and the role of lending relationships: evidence from the U.S. experience. Rass. Econ. 63, 3762.
Petersen, Mitchell A., Rajan, Raghuram G., 1994. The benefits of lending relationships: evidence from small business data. J. Financ. 49, 337.
Rajan, Raghuram G., 1992. Insiders and outsiders: the choice between informed and arm's-length debt. J. Financ. 47, 13671400.
Ross, David Gaddis, 2010. The Dominant Bank Effect: how high lender reputation affects the information content and terms of bank loans. Rev. Financ. Stud. 23,
27305756.
Santos, Joo A.C., Winton, Andrew, 2008. Bank loans, bonds, and information monopolies across the business cycle. J. Financ. 53, 13151359.
Schenone, Carola, 2004. The effect of banking relationships on the firm's IPO underpricing. J. Financ. 59, 29032958.
Schenone, Carola, 2010. Lending relationships and information rents: do banks exploit their information advantages? Rev. Financ. Stud. 23, 11491199.
Sharpe, Steven A., 1990. Asymmetric information, bank lending and implicit contracts: a stylized model of customer relationships. J. Financ. 45, 10651087.
Shumway, Tyler, 2001. Forecasting bankruptcy more accurately: a simple hazard model. J. Bus. 74, 101124.
Standard & Poor's, 2007. A Guide to the Loan Market. Standard & Poor's.
Stein, Jeremy C., 2002. Information production and capital allocation: decentralized versus hierarchical firms. J. Financ. 57, 18911921.
Weinstein, David E., Yafeh, Yishay, 1998. On the costs of a bank-centered financial system: evidence from the changing main bank relations in Japan. J. Financ. 53,
635672.