Financial institution network and the certification value of bank loans
1. Financial institution network and
the certification value of bank loans
Christophe J. Godlewski
UHA & EM Strasbourg
Bulat Sanditov
Telecom EM
AFFI Conference 2015, Cergy-Pontoise
2. Take away
2
• Financial institutions network and reputation
• Certification value of bank loans
• European syndicated loans (2001-11)
• Social network analysis + event study methodologies
• Presence of central and reputable lenders increase
borrower’s stock market reaction to a loan announcement
• Stronger effect when informational frictions are important
• Effect vanishes away during severe distruption in the
functioning of financial markets
3. Background & motivations
3
• Banks produce private information on borrowers (Diamond
1984…)
• Bank loans bear a certification value => AR > 0 for
borrower’s stock around the date of bank loan
announcement (James 1987…)
• Maintaining reputation for diligent screening & monitoring
=> mitigate informational frictions & agency problems
• Syndicated loans market (4.7 trln $, 2014): lead bank
reputation is crucial (Ross 2010…)
• Lead bank = structure deal, negotiate loan terms, organize
syndicate
• Reputable leader => enhance monitoring, attract
participants, signal quality, reduce agency costs…
4. Background & motivations (cont.)
4
• Lender reputation trust & reciprocity = critical forms of
social capital (Song 2009) driven by social networks (Cagno
& Sciubba 2010)
• Social network features of syndicated lending market =
information & capital networks (Baum et al. 2003, 2004)
• Repeated interactions => trust & reciprocity => solve
informational frictions => mitigate agency problems
• => important for firms seeking external financing
(Brander et al. 202, Wang & Wang 2012)
• => affect pricing and structure of bank loan agreements
(Cai 2009, Godlewski et al. 2012, Gatti et al. 2013)
5. Aim & contributions
5
• Do banks’ network/reputation affect certification value of
bank loans?
1. Impact of bank network/reputation on certification value
of bank loan => borrower AR / event study methodology
2. Social network metrics (Centrality centrality) to proxy
reputation => richer / comprehensive measure
3. European focus => bank private debt = main source of
external financing for companies
6. Empirical design | Data
6
• Loan and syndicate characteristics : Bloomberg
• Amount, spread, maturity, announcement date…
• Number of lenders, roles (titles)…
• Borrower characteristics : Factset
• Balance sheet & stock market information
• Country characteristics : GFDD (WB) + Djankov et al. (2007)
• European non-financial companies (24 countries)
• January 2001 – June 2011
• 254 companies / 465 loans / 906 lenders
7. Empirical design | SNA methodology
7
• Network = collection of nodes & links
• Banks’ participation in syndicated loans = affiliation
network
• => bipartite network with 2 types of nodes = actors
(banks) linked with events (deals)
• Projection of bipartite network
• => links between lead and participant banks
• => overlapping moving 3 years windows (Baum et al.
2003…)
• 3 classifications of leaders:
• Mandated arranger or Lead arranger (1)
• + Lead manager, Book runner, Book manager… (2)
• + Co / Joint, Managers… (3)
8. Empirical design | SNA methodology (cont.)
8
(a)
(b)
1 2 3 4 5 6 7 8 9 10
A B C
Lenders
Loans
11
D
1
2
3
4
5
6
7
8
9
10
11
9. Empirical design | SNA methodology (cont.)
9
• Leaders social network metrics => focus on Centrality
centrality
• => how well leader is positioned within a network
• => control over the flow of information/capital
• => interaction, reciprocity, trust => social capital =>
proxy of reputation
• Formally = number of the shortest paths between all pairs
of lenders in a network, which pass through a lender,
deflated by the number of alternative shortest paths
• Compute average, median and interquartile of Centrality
centrality by syndicate
• => 3 measures of centrality + 3 classifications of leaders
= 9 measures of network/reputation
10. Empirical design | Event study methodology
10
• Multi-event and multi-country setting
• Modified market model 𝐴𝑅𝑖 = 𝑅𝑖 − 𝑅 𝑚 (Fuller et al. 2002)
• Use local-currency national market indexes (Campbell et al.
2010)
• Bank loan announcement date = event date (day 0)
• Excluding contaminated events
• Compute three-day period CAR (-1,1)
• Multivariate analysis relies on OLS (robust s.e. clustered at
loan level) :
𝐶𝐴𝑅 −1, 1 = 𝛼 + 𝛽 × 𝐿𝑒𝑛𝑑𝑒𝑟𝑠 𝑐𝑒𝑛𝑡𝑟𝑎𝑙𝑖𝑡𝑦
+ 𝛾 × 𝐿𝑜𝑎𝑛 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠 + 𝜃 × 𝐵𝑜𝑟𝑟𝑜𝑤𝑒𝑟 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠 + 𝜗
× 𝐶𝑜𝑢𝑛𝑡𝑟𝑦 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠 + 𝜀
15. Results | Interaction terms
15
Variable
Small
loan
Short
maturity
Secured Covenants
Small
syndicate
League table
avg Centrality (1) 0.6244 2.1705 2.0629 3.8361 -3.9506 3.5214
avg. Centrality (1) x
Variable
2.2794 -0.4267 -0.0232 -9.8643 6.5792 -3.6920
Variable
Low
sales
Low debt
Low
profit
avg Centrality (1) -0.2916 0.4310 1.0274
avg. Centrality (1) x
Variable
1.8076 1.0551 0.0530
Variable
Low
stock
market
Low private
credit
Low bank
z score
High bank
concentration
Weak
creditor
rights
Crisis
avg Centrality (1) 2.0932 4.5575 10.0746 3.3854 0.7350 3.2628
avg. Centrality (1) x
Variable
1.9213 -3.4674 -12.8608 -2.0184 5.1910 -4.3345
Ibid.
Interaction variable = dummy (use of sample median for cont. Variables)
16. Conclusion
16
• Syndicate centrality / reputation matter for certification
value of bank loans in Europe
• Presence of central / reputable leaders increase stock
market reaction (AR) to a loan announcement
• Impact on AR reinforced when informational frictions are
important but effect vanishes away during financial crisis of
2008
• Contribution to recent literature on the role of reputation
and networks in financial intermediation
• Important for the development of credit markets, especially
in Europe
• Limits = potential endogeneity in matching of borrowers
and lenders
Editor's Notes
An illustration of how a bipartite network can be projected to a one-mode network is displayed in Figure.
A path between a pair of lenders i and j is a sequence of lenders beginning with lender i and ending with lender j such that each lender in this sequence is unique and has ties with lenders preceding and following him in the sequence.
Two lenders are connected if there is a path between them.
The length of a path is the number of steps (‘edges’ ) separating one from the other.
Distance between two lenders is defined as the length of the shortest path (called ‘geodesic’ ) connecting them.
Further, a connected component is a subset of nodes (lenders) such that any two nodes from this subset are connected.
An isolate is a component which consists of a single node.
For instance, lenders 1 and 10 are connected because there are several paths between them, e.g., through lenders 2, 4, 5, 7 and 8.
The corresponding geodesic, or shortest path from 1 to 10, is (1 – 2 – 4 – 8 – 10) which has length 4.
This network has two components {1÷10} and {11}. Lender 11 is an isolate as it is disconnected from the rest of the network.
Robustness checks:
Including loan spread as explanatory var. (sample size falls to 283 loans, i.e. reduction of 40%)
Centrality results robust / loan spread N.S.
Alternative specifications w/r loan variables (endogeneity issues)
Stepwise inclusion of different loan variables (with and without loan spread): amount, maturity, secured
Centrality results robust