Professional Documents
Culture Documents
Common Information Asymmetry Factors in Syndicated Loan Structures
Common Information Asymmetry Factors in Syndicated Loan Structures
a r t i c l e i n f o a b s t r a c t
Article history: This paper provides a comprehensive study of the syndicate structure and its relationship to information
Received 21 April 2011 asymmetry and loan spread by using principal component analysis on a large set of 40 structure-related
Accepted 10 December 2011 variables. A total of six structure components are identified and related to syndicate quality, syndicate
Available online 16 December 2011
members’ heterogeneity or share concentration, lead arranger’s characteristics, lead lender’s or syndi-
cate’s location, lender–borrower relationships, and lead institution type. In multivariate settings, all six
JEL classification: components are significant determinants of loan spread, either directly or indirectly through their impact
C31
on other components. Lead share retention, previous lender–borrower relationships and syndicate qual-
C38
F34
ity are shown to be bilaterally related to loan spread. Structure components differ regionally, which can
G21 provide an explanation for the European pricing discount observed in the literature. An Asian discount is
L14 observed and cannot be explained by structure differences.
Ó 2011 Elsevier B.V. All rights reserved.
Keywords:
Syndicated loan market
Principal component analysis
Syndicate structure
Information asymmetry
European pricing discount
1. Introduction at a time, typically retention by the lead arranger and the number
of lenders. Although very important, these variables do not capture
Extensive research conducted in the past 15 years on syndi- the multidimensionality of the syndicate structure that includes
cated loans concludes that one of the key differences between syn- many different characteristics that combine and interact to in-
dicated loans and bilateral loans (or sole lender loans) is the crease or decrease the diversification and information asymmetry
addition of within-syndicate information asymmetries. The way premiums of a loan.2 Judging the information asymmetries within
the syndicate is structured serves as a mechanism to address po- a syndicate solely on its size or arranger retention overlooks a
tential agency problems that can arise in such a context. However, variety of factors that can influence them. These factors range from
the syndicate structure is complex and involves many characteris- previous lender–lender or lender–borrower relationships to the dif-
tics that may mitigate or enhance within-syndicate information ference in experience between the lead and syndicate participants.
asymmetries and their associated premium. Therefore, despite The first purpose of the paper is therefore to identify the princi-
the fact that syndicated loans have, by definition, more informa- pal components of the syndicate structure. This will not only allow
tion asymmetries (between syndicate members) than bilateral us to capture all the major characteristics of a syndicate structure
loans, the overall level of this information asymmetry is hard to without generating unnecessary multicollinearity or overidentifi-
determine and can still vary from one syndicate to another.1 cation problems in multivariate settings, but also to combine
Although recent papers have examined the impact of the struc- structure metrics into a small number of significant, easily inter-
ture on loan terms or the link between structure and information pretable, components. These components are not only useful to
asymmetry, most have focused on one or two structure measures better understand the multidimensionality and complexity of a
syndicate structure through parsimonious models but also to com-
pare the structure of different syndicates along more than one
⇑ Corresponding author. Tel.: +1 819 821 8000x62976; fax: +1 819 821 7934. factor. The second objective of the paper is to use these structure
E-mail addresses: claudia.champagne@usherbrooke.ca (C. Champagne), frank. components in a multivariate setting to better understand how
coggins@usherbrooke.ca (F. Coggins).
1
As evidenced, for instance, by the high standard deviations of the structure
variables and components in the sample, available in Tables 1 and 3 and discussed
2
later. See Ivashina (2009) for a complete discussion of these two premiums.
0378-4266/$ - see front matter Ó 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.jbankfin.2011.12.009
1438 C. Champagne, F. Coggins / Journal of Banking & Finance 36 (2012) 1437–1451
they are related to one another and to information asymmetries information asymmetries and on the relationship between the
and to test their impact on loan spreads. structure and loan terms.
The contributions of this paper to the syndicated loans literature
are threefold. First, by identifying common syndicate structure fac- 2.1. Syndicate structure and information asymmetries
tors related to information asymmetries across syndicated loans,
this paper provides a new approach to characterize and quantify The structure of a loan syndicate has been extensively studied
the multidimensional structure of a syndicate. While most papers in the past 15 years. This research has generally concluded that
focus on one or two variables, usually related to syndicate size the structure is conditioned by information asymmetries. Specifi-
and lead share retention, to proxy for syndicate structure, we use cally, lender–borrower information asymmetries create informa-
a large set of variables to capture different aspects of a syndicate tion asymmetries or frictions between syndicate members that
that may increase or reduce information asymmetries. Starting may lead to agency problems. There are two types of agency prob-
with this set of 40 structure variables, we show that six components lems observed in this context: moral hazard and adverse selection
account for more than 60% of the variability in international syndi- problems. The first problem, moral hazard, occurs when the lead
cate structures and that the components can be interpreted as the arranger reduces its incentive to monitor the loan optimally once
quality of the syndicate, the heterogeneity of its members and syn- it is not responsible for the totality of it (Jensen and Meckling,
dicate concentration, the characteristics of the lead arranger, the 1976). The second problem, adverse selection, arises when the lead
geography of the syndication and its lead arranger, the average rela- arranger has private information about the borrower acquired
tions between the borrower and the syndicate members, and the through due diligence or prior relationships with the borrower. If
lender institution type. Second, this paper re-examines the impact the other members of the syndicate do not have access to this
of syndicate structure on loan spreads using principal components information, a lemons problem can occur if the lead retains a larger
that capture the multidimensionality of syndicate structure, and portion of the best-quality loans and a smaller portion of the low-
finds that all six components are significant determinants of loan er-quality loans. The two types of agency problems have been
spread, either directly or indirectly though their impact on other studied in the context of syndicated loans, as detailed in the fol-
structure components. This simultaneous estimation of different lowing paragraphs.
structure variables has never been done, and results support a num- While the structure of the syndicate can theoretically be seen
ber of different theories purported to explain loan spreads, such as both as a consequence of or a solution to agency problems, studies
the agency, certification premium and home bias theories, and offer generally conclude that the syndicate is structured to reduce infor-
new insights into their interaction.3 Further, we show that lead ar- mation asymmetries between the participants involved. In accor-
ranger share retention is not the only structure-related endogenous dance with adverse selection predictions, the characteristics of
determinant of spread, and that syndicate quality and previous len- the lead arrangers have been shown to be significant determinants
der–borrower relationships are also bilaterally related to loan spread. of syndicate structure. For example, the proportion of the loan that
Finally, we use structure components to compare three major is retained by the lead arranger is shown to be negatively related to
syndicated loan markets: US, Europe and Asia. We find that struc- the reputation of the lead (Panyagometh and Roberts, 2010). The
ture components, as well as their impact on loan spread, differ quantity and quality of information about the borrower also have
regionally. US syndicates are, on average, of higher quality and an impact on the syndicate structure. They are negatively related
with more reputable and more-experienced lead arrangers than to the share retained by the lead lender (Simons, 1993) and posi-
European or Asian syndicates. They are also associated with the tively related to the number of lenders in the syndicate (Dennis
strongest and longest lender–borrower relationships. Further, loan and Mullineaux, 2000). Panyagometh and Roberts (2002) find that
spreads are less sensitive to most structure components in Europe lead lenders syndicate a larger portion of loans that are subse-
than in the US, while they are less sensitive to lead arranger char- quently upgraded, a sign that lead banks do not engage in exploit-
acteristics and syndicate concentration in Asia, everything else ative behavior, while Jones et al. (2001) observe a negative relation
held equal. Results also provide an explanation for the European between loan rating and lead share. However, they highlight that
pricing discount observed in the literature. Specifically, the ob- arrangers may still exploit their informational advantage and
served discount can be attributable to different syndicate struc- syndicate more of the low quality loans than the syndicate
tures for European borrowers, which lead to lower diversification members would have accepted under a symmetric-information
and certification premiums, respectively. Finally, a new Asian pric- environment.
ing discount is observed, which cannot be explained by different Consistent with moral hazard, loan syndicates can also imply a
syndicate structures. free riding problem that reduces each lender’s incentive to monitor
The remainder of the paper is organized as follows. Section 2 and renegotiate if necessary. For instance, Preece and Mullineaux
reviews the literature on determinants of syndicate structure and (1996) find that the syndicate size (i.e. the number of lenders) is
loan spreads. Section 3 presents the methodology and the results negatively related to abnormal returns following loan announce-
obtained from the principal component analysis. Section 4 tests ments because of the higher renegotiation costs. Further, Esty
the components in univariate and multivariate models of loan and Megginson (2003) conclude that fewer lenders represent best
spread and syndicate structure. Section 5 concludes the paper. practices to promote monitoring efficiency and flexibility in
restructuring and that, in countries with strong creditor rights
and reliable legal enforcement, lenders create smaller and more
2. The determinants of syndicate structure and loan spread concentrated syndicates to facilitate monitoring and low cost con-
tracting. Lee and Mullineaux (2004) observe that smaller and more
To our knowledge, no previous study has addressed the issue of concentrated syndicates can avert hold-out behavior by syndicate
the multidimensionality of the syndicate structure and the identi- members or coordination problems in renegotiations. Sufi (2007)
fication of structure components. However, there is a large body of observes that lead arrangers retain a larger share and form more
literature on the relationship between syndicate structure and concentrated syndicates when borrowers require more intense
due diligence and monitoring. Lead arranger reputation or past
3
See, for example, Dennis and Mullineaux (2000), Sufi (2007), Ivashina (2009),
lead–borrower relationships can help mitigate the information
Cook et al. (2003), and Carey and Nini (2007). The theories are discussed in more asymmetries within the syndicate. Nonetheless, because low cost
detail in Section 2. restructuring can encourage borrowers to default strategically,
C. Champagne, F. Coggins / Journal of Banking & Finance 36 (2012) 1437–1451 1439
creditors may have an incentive to increase the size of the syndi- and Megginson, 2003; Ivashina 2009) and partly based on our own
cate to make default more costly or to impose a future penalty measures. The variables are defined in Appendix A. They represent
on defaulting firms (Bolton and Scharfstein, 1996; Chowdhry, characteristics of the syndicate size (e.g. number of leads and par-
1991). ticipants), share concentration (e.g. lead retention and syndicate
concentration index), relationship between the lenders and the
2.2. Syndicate structure and loan terms borrower (e.g. average intensity and duration of the relationships),
quality of the entire syndicate (e.g. average reputation and experi-
The impact of syndicate structure on loan terms has also been ence), heterogeneity between the arrangers and the participants
studied, mostly using the syndicate size or lead arranger share as (e.g. reputation and experience range), lead type (e.g. bank or
a proxy for structure. Following agency theory, the way a syndicate investment bank) and geography (e.g. lead or syndication region).
is structured affects the loan spread either through a diversification Table 1 shows the descriptive statistics for the 40 structure vari-
or information asymmetry premium. Ivashina (2009) argues that ables studied, and their correlation coefficient with loan spread.4
in equilibrium the information asymmetry premium required by The structure variables that are most correlated with spread,
the participants is offset by the diversification premium required which includes a syndicate information asymmetry premium, nota-
by the lead arranger, which increases with the lead share. An in- bly comprise concentration measures (HH-INDEX: 0.33 and TOP3:
crease in the lead share therefore increases loan spreads. Syndicate 0.34), the number of lenders (LENDERS: 0.195), the institution type
structure can also be related to non-price loan terms. For instance, of the lead arranger (LEAD-BANK: 0.22 and LEAD-INVEST: 0.15) and
Coleman et al. (2006) find that larger banking syndicates lend for the number of countries or industries involved in the syndicate
longer maturities, but due to a decline in contractual flexibility (COUNTRIES: 0.247 and TYPES: 0.19). Because of data availability,
and monitoring, lend at lower yield spreads. Consistent with the principal component analysis is conducted on two sets of structure
presence of lender–borrower information asymmetries, Cook variables: (i) the subset of 36 items (on the total sample of 20,336
et al. (2003) conclude that lenders can extract a certification pre- observations) that excludes the four concentration-related
mium that is related to their reputation. Further studies find that variables (i.e. LEAD-EXPOSURE, HH-INDEX, LEAD-SHARE AND TOP3),
the location of the lead arranger or the syndication market can and (ii) the entire set of 40 structure variables (on a restricted sample
have an impact on the price of the loan and often observe a home of 7718 observations). We perform the following orthogonal linear
bias (e.g. Carey and Nini, 2007; Houston et al., 2007). Finally, the transformation of the data matrix X containing the structure
institution type of the lender can influence loan spreads. For in- variables:
stance, Harjoto et al. (2006) and Nandy and Shao (2010) observe
that bank-led syndicates are associated with lower spreads and Y T ¼ XT W
lower-risk borrowers. P
T
Table 1
Descriptive statistics for the syndicate structure variables. This table presents summary statistics for the 40 different variables described in Appendix A and used in the principal
component analysis in Section 3 of the paper, as well as their correlation with loan spreads. N is the sample size.
significantly loading on any other component, we can interpret the the heterogeneity of the syndicate, the lower the cohesion
six components as follows7: and concentration of the syndicate and the higher the informa-
tion asymmetries.
1. Syndicate quality component (QUALITY): measures the quality 3. Lead arranger component (LEAD): measures the quality of the
of syndicate members and their relationships with one another syndicate’s lead arranger in terms of reputation, market share,
as indicated by the length and the intensity (measured by the importance and experience. Given that the syndicated loan
average number of common deals between lenders) of their market is one of repeat transactions, the higher the quality of
past connections, as well as the average reputation (based on the lead arranger, the lower the information asymmetries
the lenders’ league table ranking), market share, importance within the syndicate.
in the loan market (based on volume ranking) and experience 4. Syndicate geography component (GEOGRAPHY): measures the
(based on the number of previous transactions) of the syndicate geographical location of the lead arranger and the region of syn-
members. The stronger the past relationships between lenders, dication. Everything else held equal, a foreign lead arranger
the higher the trust and cohesion within the syndicate, which increases information asymmetries.
can reduce coordination problems and lower information fric- 5. Relationship component (RELATIONSHIP): measures the inten-
tions between members. sity and duration of the relationship between the syndicate
2. Syndicate heterogeneity component (HETEROGENEITY): mea- members (leads and participants) and the borrower. Everything
sures the heterogeneity within the syndicate in terms of the else held equal, the stronger the relationship, the lower the
asymmetry of the intensity and duration of past alliances information asymmetry.
between its members. It also captures the heterogeneity that 6. Institution type component (TYPE): indicates the type of lead
is due to the number of lenders and participants. The higher arranger in terms of its financial sector (bank or investment
firm). A bank typically has more experience in selecting and
monitoring borrowers, which will reduce information frictions.
7
Analysis of the principal components on the entire set of 40
As a robustness test, a principal component analysis is also conducted on the
quantiles of the structure variables instead of their actual measures (in Table 1) to
structure variables, including the four concentration variables,
obtain uniform units across the variables. Untabulated resulting components and shows that the second component is now related also to concen-
rotated factor optimal weights are relatively similar. tration variables (HH-INDEX and TOP3), while the GEOGRAPHY
Table 2
Rotated factor patterns for the six structure components. This table summarizes the rotated factor patterns for the six structure components retained in Section 3 of the paper. Rotation is performed with varimax. Panel A presents the
rotated factor patterns on the subset of 36 structure variables using the variables as defined in Appendix A and measured in Table 1. Panel B presents the rotated factor patterns for the entire set of 40 structure variables, including
concentration variables, using the variables as they are defined in Appendix A and measured in Table 1. In Panels A and B, variables with a loading greater than |0.40| for a specific component are shaded in gray. In Panel A, variables that
meaningfully load on a particular component and not on any other are bolded. Number of observations is 20,336 for Panel A and 7718 for Panel B.
and the LEAD components simply change position (in terms of the We recalculate the scores for the six components, including CON-
percentage of variance explained), as shown in Panel B of Table 2. CENTRATION, using weights from Panel B of Table 2. Again, to sim-
The remaining components are relatively similar. Interestingly, the plify the interpretation of the resulting factor-based score easier,
proportion of the variance explained by the second component is we recode HH-INDEX and TOP3 such that all the loadings are posi-
similar in both cases (with or without concentration variables), tive. A higher factor-based score therefore implies a lower concen-
which may indicate that the two components, CONCENTRATION tration index or top3 share, everything else held equal.
and HETEROGENEITY, can act as substitutes.
4. Analysis of structure components
3.3. Weighted factor-based scores
4.1. Univariate analysis of the structure components
To use the components in subsequent analysis, we calculate
weighted factor-based scores. A factor-based score is a linear com- As discussed in Section 2, the loan syndicate is structured to re-
posite of the variables that demonstrate meaningful loadings for duce within-syndicate information asymmetry that can be related
the component in question. Because they are not true principal to (lead-) lender–borrower information asymmetries. Panels A and
components, they can demonstrate nonzero correlation with one B of Table 3 show weighted factor-based scores for the six struc-
another.8 However, two of the advantages of factor-based scores ture components (including the CONCENTRATION component or
are their tractability and interpretability given that they are based not) identified in Section 3, conditional on two borrower character-
on fewer distinct variables. Based on the meaningful loading vari- istics commonly associated with the level of borrower information
ables for each component, which are arbitrarily defined as variables asymmetries: borrower’s opaqueness (measured by the absence of
that load with an optimal weight greater than |0.4| for a particular a credit rating on the loan) and borrower’s home country’s state of
component (identified in gray shading in Panel A of Table 2) while economic development (based on the IMF’s classification of emerg-
not significantly loading on any other component (identified in bold ing and developed countries). Opaque or emerging-country bor-
in Panel A of Table 2), variables are weighted by their loading factor rowers are associated with lower quality syndicates and leads,
and added together to obtain factor-based scores.9 The following based on their QUALITY factor-based scores, but with syndicates
artificial variables are thus created: that are more homogeneous. Opaque and emerging-country
where Wi is the weight of the variable in the component, as given in borrowers have weaker RELATIONS components and are associated
Panel A of Table 2, with two exceptions. For the GEOGRAPHY and with more concentrated syndicates (i.e. lower CONCENTRATION
TYPE components, positive and negative loadings make the interpre- components), which is consistent with Sufi (2007). These results
tation of the factor-based score difficult. To address this issue, we re- suggest that the structure components are indeed related to infor-
code LEAD-US, SYND-US-CA and LEAD-INVEST to vary in the opposite mation asymmetries.
direction, while the remaining loadings are multiplied by 1 to ob- Panel C of Table 3 shows the values for the 6 structure compo-
tain positive weights everywhere.10 A high score for TYPE therefore nents, conditional on borrowers’ regions.12 US borrowers are asso-
implies that the lead arranger is a bank and not an investment firm. ciated with syndicates of higher quality (456.33) than European
Similarly, the highest score for GEOGRAPHY would be for a structure (396.43) or Asian (218.93) borrowers. They are also associated with
in which the lead is not from the US but from the UK and that is not higher-quality leads. Syndicates for European borrowers are based
syndicated in the US or Canada but in Europe or Asia. For share-re- on weaker relationships between the borrower and lenders, and
lated variables, the HETEROGENEITY component is changed to a con- are the most heterogeneous and diffuse. Asian borrowers are associ-
centration component11: ated with the most homogeneous and concentrated syndicates and
with the lowest-quality syndicates and leads.
CONCENTRATION ¼ W 1 LENDERS þ W 2 PARTICIPANTS þ W 3
HH INDEX þ W 4 TOP3 þ W 5
4.2. Multivariate analysis of loan spread
ASYMMETRY DURATION
Studies of the impact of the syndicate structure on loan spreads
typically consider one or two structure measures at a time, which
does not capture the multidimensionality of syndicate structure.
8
A component score, by contrast, is a linear composite of the optimally-weighted Because structure variables are correlated, adding all or most rele-
observed variables and includes all the variables. As an untabulated robustness check, vant variables induces important multicollinearity or overidentifi-
all the estimations and analyses are also conducted using component scores with
cation problems in any multivariate analysis. To address these
qualitatively similar results.
9
Weighted factor-based scores based solely on the first condition (i.e. weight
12
greater than |0.4|) yield very similar results. Samples are divided according to borrower region and not lead arranger region
10
Because rotated factor weights are based on the correlation matrix, the recoding because the latter is included in the structure components. However, all estimations
simply changes the signs of the weights and not their value. and analyses are also conducted using samples based on lead arranger region, with
11
Results are qualitatively similar if the two meaningful concentration variables are very similar results (in the majority of cases, the region of the borrower is the same as
added to the five original HETEROGENEITY components. the lead arranger’s). Results are available upon request.
Table 3
Descriptive statistics for structure components, borrower-specific and loan-specific variables, conditional on borrower characteristics. This table summarizes the summary statistics for the structure component scores and factor-based
scores, borrower-specific and loan-specific variables used in the analyses throughout the paper. Component scores and weighted factor-based scores are defined in Section 3 of the paper. The remaining variables are defined in
Appendix B. N is the number of observations. Borrower industry, loan purpose, loan type, and year dummy variables are not reported to save valuable journal space.
Panel A: Structure components conditional on borrower transparence Panel B: Structure components conditional on borrower country ’s economic development
Equality of Developed-country borrowers Emerging-country borrowers Equality of
Rated borrowers Opaque borrower Means Variances Means Variances
N Mean Std. dev. N Mean Std. dev. t-Value F-value N Mean Std. dev. N Mean Std. dev. t-Value F-value
Components based on 36 structure variables
QUALITY 7089 477.575 281.11 8817 356.423 272.62 27.38 1.06 14,818 423.170 283.63 1088 236.746 205.22 28.06 1.91
HETEROGENEITY 7089 705.890 425.83 8817 468.279 393.92 36.16 1.17 14,818 587.805 426.50 1088 388.582 358.44 17.45 1.42
LEAD 7089 475.183 407.71 8817 325.268 352.48 22.47 1.34 14,818 404.421 390.00 1088 224.041 262.97 20.99 2.20
GEOGRAPHY 7089 1.189 1.31 8817 2.580 2.35 47.18 3.25 14,818 1.770 1.94 1088 4.538 2.18 40.76 1.26
RELATIONS 7089 37.882 42.90 8817 18.601 29.43 32.23 2.13 14,818 28.270 37.71 1088 12.540 27.50 17.69 1.88
TYPE 7089 1.535 0.45 8817 1.566 0.37 4.67 1.45 14,818 1.553 0.41 1088 1.547 0.35 0.52 1.33
1443
(continued on next page)
1444 C. Champagne, F. Coggins / Journal of Banking & Finance 36 (2012) 1437–1451
3.28
3.84
0.12
0.47
0.46
spread. As mentioned above, syndicate structure can affect loan
spread either through a diversification or information asymmetry
premium. With the exception of HETEROGENEITY and CONCENTRA-
Non-US borrowers
21.700
5.554
2.439
0.987
0.319
0.294
2.076
4.18
4.94
0.47
0.45
2.589
3.237
0.993
7.507
0.319
0.284
Asian borrowers
25.01
X X
6.27
1.64
1.73
0.15
0.47
0.47
6 N
SPREADl ¼ b0 þ bi STRUCTUREi;l þ bi X i;l þ el ð1Þ
i¼1 i¼7
European borrowers
In model (1), SPREADl is the all-in loan spread over LIBOR for loan l,
18.832
Mean
1.512
0.976
3.870
0.332
0.317
1.430
variables for loan l. To control for the most potential risk factors,
including loan type and purpose, observations are taken at the
N
2.22
0.34
0.29
2.03
0.04
3.189
2.459
0.999
0.132
2.802
0.094
COUNTRY; BORROWER
INDUSTRY; TYPE; PURPOSE; YEAR
10,781
10,781
10,781
10,781
10,781
10,781
10,781
2.51
2.85
0.39
0.37
0.07
3.951
2.336
2.685
0.995
0.193
0.158
INFO-SYND
FIRST-ALL
INFO-ALL
13
Untabulated robustness tests show that results are similar when done on deal
SENIOR
level observations.
14
To formally detect multicollinearity in all the models used in the study, Variable
Inflation Factors (VIF) are calculated. None of the VIFs are problematic.
Table 4
Multivariate regressions for loan spreads for US and European borrowers. This table summarizes the results for regression model (1) when loan spreads are regressed against the six structure components measured with weighted factor-
based scores, controlling for loan-specific and borrower-specific variables. The structure components are defined on the subset of 36 structure variables in Panels A, C, D and F and on the subset of 40 structure variables in Panels B, E and
H. Multivariate regressions are estimated using OLS and t-values are corrected for heteroskedasticity.
N is the number of observations used in each model specification. Borrower industry, loan purpose, loan type, loan term and year dummy variables are not reported to save valuable journal space.
*
Indicates significance at the 10% levels.
**
Indicates significance at the 5% levels.
***
Indicates significance at the 1% levels.
1445
1446 C. Champagne, F. Coggins / Journal of Banking & Finance 36 (2012) 1437–1451
diminish the information asymmetry premium. A higher HETERO- American market. Depending on the distribution of the sample,
GENEITY component in the syndicate is also related to a lower one should therefore be careful when extending observations or
spread. Although higher heterogeneity implies lower concentra- conclusions based on US samples to other countries.
tion and cohesion, which are associated with adverse selection These results are interesting in the context of the European
and moral hazard problems, it can also imply that the loan is dif- pricing discount observed on the syndicated loan market. Carey
fuse among many lenders. The negative sign indicates that the re- and Nini (2007) observe a discount of 30 bps in Europe and cannot
lated information asymmetry premium is more than offset by the explain it with differences in borrower, loan and lender character-
reduction in the diversification premium, which is consistent with istics. The authors use syndicate size as a proxy for the syndicate
Ivashina (2009). Moreover, the CONCENTRATION coefficient in Pa- structure along with the fraction of lenders with each nationality.
nel B is negative, indicating that larger and more dispersed (less Houston et al. (2007) conclude that North American borrowers
concentrated) syndicates are related to lower spreads. The positive with European lead arrangers can get a 40 bps discount. The
LEAD component coefficients of 0.014 in Panel A and 0.01 in Panel authors include the number of leads and domestic lead indicators
B show that more reputable and more-experienced lead arrangers as syndicate structure proxies. Although these syndicate character-
are associated with higher spreads, which is consistent with the istics are certainly important factors in determining loan spreads,
certification premium observed by Cook et al. (2003). US borrowers they may be a poor proxy for the multidimensional syndicate
pay higher spreads when their lead arranger is foreign or when structure. Given the large differences in at least five of the six
their loan is syndicated in Europe or Asia as opposed to the US structure components between European and US borrowers in pa-
(captured by the GEOGRAPHY component), which is evidence of a nel C of Table 3, different syndicate structures could help explain
domestic bias. However, this component is not significant in Panel the European discount. This is formally verified in panels G and
B when concentration variables are added to the components. H, in which we combine the two sub-samples of US and European
Finally, the TYPE component indicates that bank-led syndicates borrowers and include a dummy variable to capture a European
(investment firm-led) are associated with lower (higher) spreads, effect in the model.15 In Panel G, when no structure component is
which is consistent with Harjoto et al. (2006) and Nandy and Shao included, European borrowers are associated with lower-spread
(2010), discussed earlier. The remaining coefficients are in line loans (28.97 bps lower). In Panel H, with the inclusion of structure
with the literature; larger, more profitable and less indebted components, the European pricing effect disappears. Further, it is
(measured with DEBTA) borrowers are related to lower spreads. not captured by the GEOGRAPHY component of the syndicate struc-
To further test whether the components are related to informa- ture.16 The pricing discount observed in previous papers is therefore
tion asymmetries, we include interactive variables combining each potentially attributable to different syndicate structures in Europe,
of the components with OPAQUE, a measure of borrower opaque- such as more diffuse or lower quality syndicates and lead arrangers
ness that is typically related to lender–borrower information (see Table 3). These differences can lead to lower concentration and
asymmetries, in model (1). Results are available in Panel C and certification premiums.
show that borrower information asymmetries are related to all The multivariate analysis of loan spreads for Asian borrowers is
the components except LEAD. The quality of the syndicate has a presented in Table 5. The sensitivity of loan spreads to structure
stronger impact on loan spreads when the borrower is opaque components is also different from that of US borrowers. Loan
(0.027) than for transparent borrowers (0.016). A similar phe- spreads in Asia are related to the heterogeneity or concentration
nomenon occurs for RELATIONS and TYPE. The heterogeneity of of the syndicate only weakly or not at all, and are not influenced
the syndicate is positively (negatively) related to the loan premium by the reputation of the lead arranger, which may be related to
when borrowers are opaque (transparent), which indicates that a the low LEADS component in Asia (see Table 3). Further, Asian
more homogeneous (concentrated) syndicate is beneficial when borrowers pay lower spreads when they deal with a non-US lead
the borrower needs more due diligence and monitoring. LEAD is or have their loan arranged in Europe or Asia. From the coefficients
only a significant (and positive) determinant of the spread for in panel C, we see that opaque Asian borrowers benefit much more
transparent borrowers, which may indicate that the certification from a higher quality syndicate or lead arranger than US or Euro-
premium is offset by a reduction in syndicate information asym- pean borrowers, while transparent borrowers, who do not benefit
metries for opaque borrowers only. These results support the the- as much from the reduction in syndicate information asymmetries,
ories and studies on syndicate structure discussed in Section 2. are charged a higher certification premium.17 Like their US counter-
They show that syndicate-related premiums are explained not only parts, opaque Asian borrowers pay a higher premium with a more
by lead arranger concentration but also, simultaneously, by the heterogeneous syndicate. Nonetheless, contrary to US borrowers,
quality and cohesion of the entire syndicate, the characteristics Asian opaque (transparent) borrowers pay a higher (lower) spread
of the lead arranger, the location of the lead and the syndication when they have more and longer past relationships with syndicate
market, lender–borrower relationships and the type of lender. Fur- members.
ther, borrower information asymmetries are related not only to the Using a similar methodology as for the European discount, we
concentration of the syndicate, but also to the other structure test for an Asian discount. Results are available in panels D and
components. E. Results in panel D shows that an Asian effect is observed and
is even larger than the European pricing effect observed previously,
4.3. European and Asian markets with a lower spread of more than 107 bps for Asian borrowers as
opposed to US borrowers. The inclusion of structure components
Panels D–F present a similar analysis on a sub-sample of Euro- in panel E reduces the effect to 88.86 bps but does not remove
pean borrowers. According to panel D (Panel E), only LEAD (QUAL- its significance. Based on our results, it therefore appears that a
ITY) is a significant determinant of the spread for European
borrowers. When conditioning on the opaqueness of the borrower
in panel F, the quality and cohesion of the syndicate (QUALITY) and 15
Because spread models are region-dependent, results using an OLS regression
its lead arranger’s reputation (LEAD) are negatively related to the may be biased. However, because the European discount was previously inferred with
such a model on international samples, we wish to keep the same methodology here.
spread for opaque borrowers, and positively related to spread for 16
Another specification in which the GEOGRAPHY component is removed from the
transparent borrowers. These results show that that spread models model generates similar results.
differ according to the region and that loan spreads on the Euro- 17
Interactive variables with GEOGRAPHY and LEADS are associated with high
pean market are less sensitive to syndicate structure than on the multicollinearity and are removed from the model to ensure valid inference.
C. Champagne, F. Coggins / Journal of Banking & Finance 36 (2012) 1437–1451 1447
Table 5
Multivariate regressions for loan spreads for Asian borrowers. This table summarizes the results for regression model (1) when loan spreads are regressed against the six structure
components measured with weighted factor-based scores, controlling for loan-specific and borrower-specific variables. The structure components are defined on the subset of 36
structure variables in Panels A and C and on the subset of 40 structure variables in Panels B and E. Multivariate regressions are estimated using OLS and t-values are corrected for
heteroskedasticity.
N is the number of observations used in each model specification. Borrower industry, loan purpose, loan type, loan term and year dummy variables are not reported to save
valuable journal space.
*
Indicates significance at the 10% levels.
**
Indicates significance at the 5% levels.
***
Indicates significance at the 1% levels.
pricing discount, which cannot be explained by different syndicate determined with loan spread: heterogeneity (or concentration),
structures, persists for Asian borrowers.18 relations and quality. Results also show that structure components
can also be related to other components. For instance, a higher
RELATIONS component is associated with higher-quality and more
4.4. Complementary results heterogeneous syndicates, while the presence of a bank as lead
arranger increases the heterogeneity and lowers the concentration
We re-examine the results by taking into account bilateral rela- of the syndicate. It is also associated with stronger relationships.19
tions not only between structure and spread but also between
structure components. Untabulated results using endogenous
5. Conclusion
models show that the signs of the relationships between the
components and the spread are similar to those using an OLS esti-
This paper is the first comprehensive study of all dimensions of
mation of model (1), with the exception of the geography compo-
a syndicate structure. We start with a very large set of 40 struc-
nent, which is no longer a direct significant determinant of spread,
ture-related variables and use principal component analysis to
but rather an indirect determinant through its influence on HETER-
identify six principal components of the syndicate structure, there-
OGENEITY, CONCENTRATION, or RELATIONS. Overall, the effect of the
by providing a new approach to characterize and quantify the mul-
structure components on loan spreads is generally stronger with
tidimensional structure of a syndicate. We can interpret the six
the endogenous model. Three components are endogenously
components as the quality of the syndicate, the heterogeneity of
its members or share concentration, the characteristics of the lead
18
The discount appears to be present for most countries in Asia, including Japan, arranger, the geography of the syndication and its lead arranger,
although some countries have very few observations from which strong inferences
can be drawn. Dividing the sample into Asian crisis (1997) and post-crisis periods or
19
winsorizing the data (at the 1% and 5% percentiles) to remove outliers do not change Untabulated results using endogenous models for European and Asian borrowers
the results. are also consistent with previous results.
1448 C. Champagne, F. Coggins / Journal of Banking & Finance 36 (2012) 1437–1451
the average relations between the borrower and the syndicate Structure components also appear to differ regionally. US bor-
members, and the lenders’ institution types. rowers are associated with syndicates and leads of higher quality
In univariate settings, we can associate opaque or emerging- than European or Asian borrowers. Syndicates for European bor-
country borrowers with lower quality syndicates and leads, weaker rowers are based on weaker lender–borrower relationships and
previous relationships with the lenders and more homogeneous or are the most heterogeneous and diffuse. Asian borrowers are asso-
concentrated syndicates. We re-examine the impact of syndicate ciated with the most homogeneous and concentrated syndicates,
structure on loan spreads using multivariate regressions and find and with the lowest-quality syndicates and leads. Multivariate
that the six structure components are significant determinants of spread models also differ according to borrower region. Loan
loan spreads, either directly or indirectly through their impact on spreads are less sensitive to syndicate structure on the European
other structure components. Consistent with agency theory, high- market than on the American market. However, opaque Asian bor-
er-quality syndicates with stronger cohesion can diminish the infor- rowers benefit much more from a higher quality syndicate or lead
mation asymmetry premium. Heterogeneous or less concentrated arranger than their US or European counterparts. Finally, we can
syndicates are related to lower spreads, which is consistent with a explain the European pricing discount observed in the literature
reduction of the diversification premium. In keeping with certifica- by different syndicate structures for European borrowers than
tion premium theory, reputable and more-experienced lead arrang- for US borrowers, such as more diffuse syndicates and lower-
ers are associated with higher spreads. However, results show that quality syndicates and lead arrangers. These differences can in
lead quality is a significant (and positive) determinant of spread turn lead to lower concentration and certification premiums,
only for transparent borrowers. For opaque borrowers, the benefits respectively.
of a higher-quality lead offset the certification premium. Results
from the endogenous estimation of structure components show Acknowledgements
that three structure components have bilateral relationships. We
demonstrate that components interact with each other to ulti- Special thanks to an anonymous referee for many constructive
mately affect loan spreads. For instance, in accordance with home comments and suggestions, which helped us improve the paper.
bias theory, US borrowers pay higher spreads with a foreign lead We also thank Stefanie Kleimeier and other participants at the
or syndication region. However, when structure components are 2011 FMA European Conference and PBFEAM Conference. Finally,
modeled endogenously with the spread, the geography component we would like to acknowledge the Research Group in Applied Fi-
is no longer a direct determinant of loan spreads but rather an nance (GReFA), the Fonds de recherche sur la société et la culture
indirect determinant through its influence on other structure (FQRSC) and the Institut de Finance Mathématique de Montréal
components. (IFM2) for their financial support. The usual disclaimer applies.
This appendix describes the different syndicate structure variables used in the principal component analysis. Unless otherwise men-
tioned, variables are either from Dealscan (loan terms, syndicate structure, and lenders) or from Compustat Global (borrower
characteristics).
Variable Definition
Appendix A (continued)
Variable Definition
(c) Variables related to the relationship between the lenders and the borrower
REL-PARTICIPANTS Average number of past loans in the 5-year period prior to the deal active date with each
lender in the syndicate
REL-LEAD Average number of past loans in the 5-year period prior to the deal active date with the
lead arranger(s).
DURATION-PARTICIPANTS Average length of relationship between the borrower and each lender in the syndicate,
measured in number of months between the first deal and current deal active date,
DURATION-LEAD Length of relationship between borrower and lead arranger, measured in number of
months since first deal.
This appendix describes the different structure variables used in the multivariate models. The variables are divided into two categories:
(i) borrower-specific variables and (ii) loan-specific variables. Unless otherwise mentioned, variables are either from Dealscan (loan terms,
syndicate structure, and lenders) or from Compustat Global (borrower characteristics).
Variable Definition
References Carey, M., Nini, C., 2007. Is the corporate loan market globally integrated? A pricing
puzzle. Journal of Finance 62 (6), 2969–3007.
Chowdhry, B., 1991. What is different about international lending? Review of
Angbazo, L., Mei, J., Saunders, A., 1998. Credit spreads in the market for highly
Financial Studies 4 (1), 121–148.
leveraged transaction loans. Journal of Banking and Finance 22, 1249–1282.
Coleman, A., Esho, N., Sharpe, I., 2006. Does bank monitoring influence loan contract
Bolton, P., Scharfstein, D., 1996. Optimal debt structure and the number of creditors.
terms? Journal of Financial Services Research 30 (2), 177–198.
The Journal of Political Economy 104 (1), 1–25.
C. Champagne, F. Coggins / Journal of Banking & Finance 36 (2012) 1437–1451 1451
Cook, D., Schellhorn, C., Spellman, L., 2003. Lender certification premiums. Journal of Jones, J., Lang, W., Nigro, P., 2001. Recent trends in bank loan syndications: Evidence
Banking and Finance 27 (8), 1561–1579. for 1995–1999. Federal Reserve Bank of Chicago Proceedings, May, pp. 328–352.
Dennis, S., Mullineaux, D., 2000. Syndicated loans. Journal of Financial Lee, S., Mullineaux, D., 2004. Monitoring, financial distress, and the structure of
Intermediation 9, 404–426. commercial lending syndicates. Financial Management 33 (3), 107–130.
Esty, B., Megginson, W., 2003. Creditor rights, enforcement, and debt ownership Nandy, D., Shao, P., 2010. Institutional investment in syndicated loans. Unpublished
structure: evidence from the global syndicated loan market. Journal of Financial manuscript.
and Quantitative Analysis 38 (1), 37–58. Panyagometh, K., Roberts, G., 2002. Private information, incentive conflicts, and
Harjoto, M., Mullineaux, D., Yi, H., 2006. A comparison of syndicated loan pricing at determinants of loan syndications. Unpublished manuscript.
investment and commercial banks. Financial Management 35 (4), 49–70. Panyagometh, K., Roberts, G., 2010. Do lead banks exploit syndicate participants?
Houston, J., Itzkowitz, J., Naranjo, A., 2007. Borrower beyond borders: The Evidence from ex post risk. Financial Management 39 (1), 273–299.
geography and pricing of syndicated bank loans. Unpublished manuscript. Preece, D., Mullineaux, D., 1996. Monitoring, loan renegotiability, and firm value:
Ivashina, V., 2009. Asymmetric information effects on loan spreads. Journal of the role of lending syndicates. Journal of Banking and Finance 20, 557–593.
Financial Economics 92, 300–319. Simons, K., 1993. Why do banks syndicate loans? New England Economic Review,
Jensen, M., Meckling, W., 1976. Theory of the firm: Managerial behavior, Federal Reserve Bank Boston, 45–52.
agency costs and ownership structure. Journal of Financial Economics 3 (4), Sufi, A., 2007. Information asymmetry and financing arrangements: evidence from
305–360. syndicated loans. The Journal of Finance 62 (2), 629–668.