LEGAL RISK AS A DETERMINANT OF SYNDICATE STRUCTURE IN THE PROJECT FINANCE LOAN MARKET

Benjamin C. Esty Associate Professor of Finance

Harvard Business School

and

William L. Megginson* Professor & Rainbolt Chair in Finance Michael F. Price College of Business

The University of Oklahoma

First Draft: June 27, 2000 Current Draft: November 6, 2001

Comments Welcome

* The authors thank Yiorgos Allayanis, Viral Acharya, Malcolm Baker, Amir Barnea, Charlie Hadlock, Campbell Harvey, Kose John, Leora Klapper, Stefanie Kleimeier, Larry Lang, Josh Lerner, Scott Mayfield, Ehud Ronn, Rick Ruback, Andrei Shleifer, David Smith, Jeremy Stein, and seminar participants at the World Bank, Federal Reserve Board, Texas Finance Festival, European Financial Management Association, Florida, Harvard’s Kennedy School of Government, Harvard Business School, Illinois, Kansas, Maryland, and NYU for helpful comments on earlier drafts. We also thank Fuaad Qureshi for research assistance. Finally, we are especially grateful for the financial support from the University of Oklahoma’s Michael F. Price College of Business, which allowed us to purchase the Loanware database, and the Division of Research at Harvard Business School. An earlier version of this paper was titled, “Syndicate Structure as a Response to Political Risk in the Project Finance Loan Market.” Please address all correspondence to:

Benjamin C. Esty Morgan 381 Harvard Business School Boston, MA 02163 Tel: (617) 495-6159 Fax: (617) 496-6592 e-mail: besty@hbs.edu

2

LEGAL RISK AS A DETERMINANT OF SYNDICATE STRUCTURE

IN THE PROJECT FINANCE LOAN MARKET

Abstract

This paper examines how legal risk, defined as the strength of creditor rights and legal

enforcement, affects debt ownership concentration in the project finance loan market. Using a sample of

495 project finance loan tranches from 61 countries, worth $151 billion, we document high levels of debt

ownership concentration: the largest single bank holds 20.3% while the top five banks collectively hold

61.2% of a typical project finance loan tranche. We also show that weak creditor rights and poor legal

enforcement are associated with more diffuse ownership structures, which leads us to conclude that

international project finance lenders structure syndicates to deter strategic default rather than to enhance

monitoring incentives or facilitate low-cost re-contracting in the event of default. On a more theoretical

level, the results illustrate the continuous nature of debt ownership and refute the overly simplistic

distinction between single bank creditors and atomistic public bondholders commonly described in the

literature.

Key words: bank lending, project finance, syndication, international corporate governance, creditor rights, legal rules and enforcement

JEL classification: G21, G32, F34

LEGAL RISK AS A DETERMINANT OF SYNDICATE STRUCTURE IN THE PROJECT FINANCE LOAN MARKET

Banks extended $2 trillion of syndicated loans in 2000, making the syndicated loan market the

largest source of corporate funding available today. It is also one of the fastest growing. In fact, volume

has been growing at a compound annual rate of more than 10% per year over the past decade. Despite the

market’s size and recent growth, there has been surprisingly little empirical research on either the total

market for syndicated loans or the subset of syndicated project finance (PF) loans.1 According to Project

Finance International (various years), an industry trade journal, banks extended $110 billion of PF loans

in 2000, up significantly from the $40 to $50 billion extended during the mid 1990s.

Project-financed investments differ from corporate-financed investments because the assets are

financed as stand-alone entities rather than as part of a corporate balance sheet. Although creditors may

have partial recourse for a period of time or for a fraction of the total loan amount, project debt is, by

definition, non-recourse to sponsoring organizations. Nonetheless, project companies tend to be highly

levered. Esty, et al (1999) show that recent projects have an average debt-to-total capitalization ratio of

69%, with a range from 50% to 90%. The majority of this debt, 84% in 2000 and more than 90% during

the 1990s, has come from banks. Because high leverage entails greater distress costs and bank debt entails

tighter covenants, stricter oversight, and shorter loan maturities than public bonds, there must be

countervailing benefits that justify the use of high leverage and bank finance in the context of project-

financed investments.

In an attempt to shed light on the benefits of bank finance and the role banks play in corporate

governance, we analyze debt ownership concentration in the context of syndicated project finance loans.

Finance theory assigns three economic functions to banks as financial intermediaries. Early work by

Campbell and Kracaw (1980), Diamond (1984), and Fama (1985) describes a monitoring role for banks.

More recent work by Bolton and Scharfstein (1996) and Gertner and Sharfstein (1991) notes that banks

1 Preece and Mullineaux (1996), Dennis and Mullineaux (2000), Jones, Lang, and Nigro (2000), and Simons (1993) study aspects of syndicate structure; Altman & Suggitt (2000) study default rates for syndicated loans; Megginson et al (1995) study announcement day returns; and Boehmer and Megginson (1990) study loan pricing.

2

provide low-cost re-contracting in the event of default. Finally, in the context of international lending,

banks also play a role in deterring voluntary, or strategic (i.e. borrowers who can, but choose not to, pay),

default by making the option to default more costly (Chowdry, 1991; Bolton and Scharfstein, 1996).

Empirical evidence generally supports the idea that bank finance improves firm performance and value.2

We focus on these three functions—monitoring, re-contracting, and deterring—because they

generate different empirical predictions regarding the relation between debt ownership concentration and

legal risk. Although “legal risk” can mean many things in the context of international lending, creditors

are primarily concerned with debt repayment by solvent firms and recoverability of value from bankrupt

firms. In either case, lending relationships are governed by contracts that are usually based on US (New

York) or UK commercial codes. Enforcement of loan terms, however, depends on the legal system in the

country where project is located. As a result, creditors must understand not only their legal rights in the

countries in which they lend, but also the effectiveness of enforcement mechanisms in those countries.

We measure legal risk in two ways. First, we use LaPorta, Lopez-de-Silanes, Shleifer, and

Vishny’s [hereafter, LLSV] (1998, 1999a) index of creditor rights, a measure of theoretical rights.

Because LLSV show that civil law systems provide lower levels of protection for both creditors and

shareholders, legal origin provides another measure of creditor rights. We also measure legal risk using

Berkowitz, Pistor, and Richard’s [hereafter, BPR] (2001) composite legality index, which measures the

effectiveness of legal enforcement. For the purposes of this paper, we define countries as having high

legal risk if creditors have weaker rights (i.e. civil law countries) or ineffective enforcement. We contend

that creditors are more susceptible to strategic defaults in countries with high legal risk.

We hypothesize that bankers adjust syndicate structures (e.g. the number of banks and the

concentration of holdings) to reflect legal risk. In the presence of high legal risk and a correspondingly

higher probability of default, they might prefer more concentrated lending syndicates (i.e. fewer banks

2 This list includes, among others: Datta, Iskandar-Datta, and Patel (1999), De Long (1991), Gorton and Schmid

(2000), James (1987), Lummer and McConnell (1989), and Hoshi, Kashyap, and Scharfstein (1991)

3

holding larger shares) to reduce the expected information, administrative, and other re-contracting costs

associated with default. At the same time, concentrated debt ownership should improve their ability to act

as delegated monitors. On the other hand, bankers might prefer less concentrated lending syndicates (i.e.

more banks holding smaller shares) in the presence of high legal risk as a way to discouraging strategic

defaults—by making the syndicates larger, they increase the expected cost of restructuring.

To test these competing hypotheses, we collected a sample of 495 tranches from syndicated project

finance loans made between 1986 and 2000. These tranches come from 61 different countries, which

ensures wide cross-sectional variation in our legal risk measures, have a total value of $151 billion, and

have an average value of $304 million. Using univariate analysis, we document high levels of debt

ownership concentration. At closing, the largest single debt provider holds an average (median) of 20.3%

(14.8%) of the tranche. The top five banks collectively hold an average (median) of 61.2% (57.3%) of the

tranche. Although the largest share declines with size, the largest single bank still holds almost 10% of

tranches over $500 million. The concentration of debt ownership far exceeds the concentration of equity

ownership documented in US companies and is more similar to equity ownership concentration in

countries with weak shareholder rights (LLSV, 1999b).3 We also document a significant relation between

legal risk and syndicate structure: tranches in countries with weak creditor rights or poor legal

enforcement exhibit less concentrated debt ownership structures. This finding is consistent with the

deterrence role, but not the monitoring or re-contracting roles played by banks.

This analysis differs from existing research in three ways. First, we study creditor-based

governance, in general, and the disciplinary role of banks, in particular, in an environment where bank

participation is likely to have first-order effects. Whereas bank debt accounts for approximately 10-40% of

total capital in the typical industrial firm, it accounts for 60-90% of total capital in project companies.

Second, we study the level and determinants of debt ownership concentration rather than the reasons for

3 Demsetz and Lehn (1985), Morck, Shleifer, and Vishny (1988), Holderness and Sheehan (1988), McConnell and

Servaes (1990), Gorton and Rosen (1995), and Himmelburg et al (1999) study equity concentration in US firms.

4

syndication, the determinants of debt structure (maturity, priority, public vs. private, etc.), or the impact of

bank relationships on firm performance.4 The syndicates in our sample contain from two to 62 banks,

though large syndicates can have 100 or more banks. These structures highlight the continuous nature of

debt ownership and refute the overly simplistic distinction between single bank creditors and atomistic

public bondholders commonly described in the literature. Finally, in contrast to most of the previous

research on equity ownership and shareholder rights (LLSV, 1998, 1999a; Claessens et al, 1999, 2000; and

Johnson et al, 1999, all study “anti-director” rights), we focus exclusively on debt ownership and creditor

rights. Whereas the other studies find a negative relation between equity ownership concentration and

shareholder rights (concentrated equity ownership in countries with weak shareholder rights), we find a

positive relation between debt ownership concentration and creditor rights.5

We view our results as a an important first step towards answering Shleifer and Vishny’s (1997)

call for further research on creditor governance. Rather than conclusively proving a single theory of debt

ownership, we document basic facts about the level and determinants of debt ownership concentration, and

interpret these facts based on existing theories of financial intermediation. Viewed in this light, our results

provide both a foundation for future empirical research on debt ownership as well as a set of stylized facts

for future theoretical research on creditor governance.

The paper is organized into four sections. In the first section, we review the relevant banking and

governance literature as a way to generate hypotheses regarding the relations between syndicate structure

and legal risk. Section II discusses our dataset and provides univariate analyses of project risk, legal risk,

and syndicate structure variables. Our analysis on the determinants of syndicate structure appears in

Section III. We conclude with a brief discussion of our findings and their implications in Section IV.

4 The closest work is Preece and Mullineaux’s (1996) analysis of syndicate size, broadly defined (small, medium,

etc.), and loan renegotiability. Our paper goes into greater depth on both fronts and uses a larger sample. Other studies, notably Jones, Lang, and Nigro (2000) and Simons (1993), analyze loan shares retained by arranging banks as a test of adverse selection in syndicated lending, by ignore shares held by providing banks.

5 Johnson et al (1999) analyze the relation between creditor rights and market performance during the Asian crisis,

but creditor rights is not significant. Levine (1999), Wurgler (2000), and Miller and Puthenpurackal (2000) find that creditor rights significantly affect economic growth, capital allocation, and loan pricing, respectively.

5

I. Background Information and Hypotheses

The primary goals of this paper are to document levels debt ownership concentration and to test for

a relation between syndicate structure and legal risk in the project finance loan market. Before presenting

explicit hypotheses relating syndicate structure and legal risk, it is important to establish some basic facts

and terminology about syndicated lending, project finance, and bank-based corporate governance. A bank

syndicate is a collection of banks that jointly extends a loan to a specific borrower.6 Unlike a loan sale to a

third party in which no direct contract exists between the borrower and the buyer, syndication involves a

direct contract between each member bank and the borrower (Pennachi, 1988; and Gorton and Pennachi,

1995). Lending syndicates, at least in their simplest form, resemble pyramids with arranging banks

(arrangers) at the top and providing banks (providers) at the bottom. Prior to closing a loan, the arranging

(or “mandated”) banks meet with the borrower, perform a credit analysis, negotiate key terms and

conditions, and prepare an information memorandum for providing banks. Once the key terms are in

place, the arranging banks invite certain banks to participate in the deal and allocate shares to them as they

see fit.7 The syndication process allows us to assume that syndicate structures are endogenously

determined in response to project characteristics. This distinction between borrower-arranged and creditor-

arranged financing structures is a critical aspect of our paper. After closing, the arranging banks monitor

compliance with loan covenants, negotiate contingent agreements when they arise, and lead negotiations in

default situations. Because the arranging banks play a more prominent role both leading up to and after

syndication, we focus most of our attention on the arranging banks.

Syndicated loans of this type are the predominant form of funding for project-financed

investments. What defines project finance is the creation of a legally-independent project company

6 For an in-depth analysis, see Esty’s (2001a) case study on how Chase syndicated the Hong Kong Disneyland

project loan. Also see Howcroft and Solomon (1985); Terrell and Martinson (1978); or Rhodes (2000). 7 In an underwritten deal, the process works somewhat differently. The arranging banks agree to make the loan, and

later attempt to syndicate it to providing banks in a process known as general syndication.

6

financed with non-recourse debt (Esty, 2001b). In creating a project company, sponsoring firms sign

contracts with, among others, construction firms, raw material suppliers, buyers, host governments, and

capital providers. Lenders, for example, negotiate commitment letters, collateral packages, and loan

documents with project companies, and inter-creditor agreements among themselves. This nexus of

contracts is intended to ensure loan repayment when the project company is solvent and loan recoverability

when the project is in default. According to Shleifer and Vishny’s (1997, p. 737) definition—“corporate

governance deals with the ways in which suppliers of finance assure themselves of getting a return on their

investment”—contracting constitutes one form of corporate governance. Bankers’ dependence on

contracts forces them to rely on legal rights and enforcement to generate returns. Arguably, bankers could

rely more heavily on price mechanisms than contractual enforcement, but the careful attention to legal

structuring seen in most projects belies this argument.

While contracting is one way banks attempt to assure themselves of getting a return, they perform

at least three important functions to ensure repayment: they monitor borrower compliance, provide low-

cost re-contracting in the event of default, and deter strategic default. In the context of syndicated lending,

the structure of the syndicate in terms of size and concentration of holdings affects their ability to perform

each function. For example, Campbell and Kracaw (1980), Diamond (1984), and Fama (1985) show that

banks provide valuable monitoring services. Subsequent empirical research by James (1987) and Lummer

and McConnell (1989) on the returns associated with new loan agreements, and by De Long (1991),

Hoshi, Kashyap, and Scharfstein (1991), and Gorton and Schmid (2000) on the relation between firm

performance and bank finance, validates the effectiveness of bank monitoring. More recently, however,

Rajan (1992), Houston and James (1996), Kang and Stulz (1998), Weinstein and Yafeh (1998), and Morck

and Nakamura (1999) have highlighted the disadvantages of bank control, namely the danger of getting

locked into banking relationships.8 Because project finance involves a one-time transaction rather than an

on-going relationship, and involves multiple rather than single creditors, concerns regarding lock-in are

8 Ongena and Smith (2000a) present results that cast doubt on importance of lock-in as a disadvantage of bank debt.

7

less relevant in this setting.

Although this research on monitoring has yielded valuable insights, it ignores an important feature

of syndicated lending, namely the potential problem with free riding that can occur in large syndicates.

Unless a bank has a large ownership position, it will not have an incentive to monitor; it will incur all the

costs of monitoring, but receive only a fraction of the benefits. In situations that require extensive

monitoring, therefore, one should expect to see smaller syndicates with more concentrated holdings.

Working against the incentive to concentrate holdings as a way to improve monitoring incentives,

however, is the incentive to diversify one’s loan portfolios. Banks, facing higher risk of any kind,

including legal risk, might have to compromise between these conflicting objectives.

The second function banks play is that they facilitate low-cost re-contracting in the event default.

Bolton and Scharfstein (1996) and Gertner and Scharfstein (1991) present theoretical models showing that

small groups of banks are able to restructure firms faster and more cheaply than large groups of public

bondholders can. Consistent with these models, Gilson, John, and Lang (1990) find that the time and cost

of restructuring increases as the number of creditors increases, yet declines with the fraction of bank debt.

Further research by Preece and Mullineaux (1996) shows that the positive abnormal return associated with

new bank loan announcements is negatively related to syndicate size, which they assert shows that

renegotiation costs gradually offset monitoring benefits as syndicate size increases. Low-cost re-

structuring is very important in project finance because projects involve dedicated assets with going

concern, but not salvage, value. As a result, asset liquidations are rare and restructurings are the norm (see

Hoffman, 1998, p. 656). Knowing that restructuring is the likely course for a defaulted project, lenders

rationally structure syndicates to minimize restructuring costs.

Bolton and Scharfstein (1996), however, note lenders create perverse incentives for strategic

defaults when they reduce restructuring costs. In fact, lenders can reduce these incentives by intentionally

making default more costly—deterring strategic default is a third governance function played by banks.

Banks can make default more costly by increasing the number of creditors, which increases the complexity

8

and negotiating cost of re-contracting (Gertner and Scharfstein, 1991; Diamond, 1991; and Bolton and

Scharfstein,1996). A second way to make default more costly is to impose a penalty on defaulting firms.

In Chowdry’s (1991) model of sovereign lending, banks threatening to withhold future lending. He

assumes that banks can credibly commit not to lend only if they have previously lost money to a particular

borrower. Thus, as you increase syndicate size, you reduce the pool of potential creditors willing to lend to

a defaulting borrower at some point in the future. Of course, the defaulting country must depend on

external finance for this threat to be credible. Yet for projects with little on-going funding needs, this form

of deterrence may be less effective. It may, however, be more effective against sponsoring firms which

rely on external finance to fund their capital expenditures.

II. A. Hypotheses Relating Syndicate Structure and Legal Risk

We hypothesize that legal risk affects the way banks perform their monitoring, re-contracting, and

deterrence functions, and that banks adjust syndicate structure—both the size and the concentration of

holdings—to reflect legal risk. When creditors have weak rights or operate in countries with weak

enforcement, strategic default becomes more likely. In these settings, the challenge is to minimize the cost

of strategic defaults in any one of several ways. First, banks can try to detect misappropriation of cash

flows prior to default through enhanced monitoring. Alternatively, they can use high leverage as a tripwire

and resort to low-cost restructuring if and when default occurs. Or finally, they can try to deter strategic

defaults from happening in the first place. Interestingly, these strategies have different implications for

syndicate structure.

II. A.1. The monitoring hypothesis

Because completed projects tend to have low marginal costs and little need for re-investment, they

generate significant amounts of free cash flow, which can be siphoned off by corrupt management,

controlling shareholders, or host governments (see Johnson et al, 2000, on tunneling in emerging markets).

9

In fact, one of the most important things banks do is monitor the project’s cash flow “waterfall”—the

allocation of cash flows according to a strict, contractual hierarchy of claims.9 But the incentive to monitor

requires concentrated ownership. For this reason, we expect to see high levels of debt ownership

concentration. We could also logically predict there would be a positive relation between legal risk and

debt ownership concentration. In other words, when legal risk is high (creditors have weak rights or three

is weak enforcement), syndicates will contain fewer banks holding larger shares. This prediction applies

most strongly to the arranging banks, the ones responsible for setting up the syndicate and for monitoring

on-going performance.

II.A.2. The low-cost re-contracting hypothesis

If banks resort to early detection and low-cost re-contracting as their defense against strategic

default, then we should observe a positive relation between legal risk and debt ownership concentration.

As readers will note, the monitoring and low-cost re-contracting hypotheses generate similar empirical

predictions. As a result, we will have trouble distinguishing between the two in the event the analysis

documents a positive relation.

II.A.3. The deterrence hypothesis

Finally, if banks rely on prevention as their defense against strategic default, then we should

observe a negative relation between legal risk and debt ownership concentration: when legal risk is high,

syndicates will contain more banks holding smaller shares. By creating larger syndicates, banks credibly

signal that ex post re-negotiation is going to be more costly.

To summarize this section, we believe that legal risk affects the way banks carry out their

monitoring, re-contracting, and deterrence functions, and how they structure lending syndicates in terms of

the number of banks and the concentration of holdings. In the next section, we test these hypotheses using

9 Esty and Millett (1998) describe the details of cash waterfall in the context of the Petrozuata project.

10

our sample of project finance loan tranches. While the hypotheses generate opposing predictions regarding

syndicate structure, we readily admit that observed structures represent complex combinations of

objectives. What we observe in the end is the net effect of sometimes conflicting objectives mixed with

elements of randomness due to market conditions at the time the loans were made.

II. Data and Methodology

Our sample of syndicated loans comes from Capital DATA’s Loanware database, which contains

information on more than 85,000 syndicated loan tranches made between 1980 and April 2000.10 We

began with all 5,646 loan tranches with project finance listed as the loan purpose, but excluded tranches

less than $75 million because they tend to have simpler and less interesting syndicate structures.11 This

screen left 2,545 loan tranches, a yield of 88.4% by value and 45.1% by number. Because the unit of

observation is a loan tranche, multiple tranches from the same loan appear as separate observations in our

database—22 loans, containing 49 tranches, have multiple tranches in our dataset. Table 1 describes the

full population of syndicated loans and the subset of project finance loans based on loan signing date.

**** Insert Table I about here ****

Although comprehensive in many ways, the Loanware database has two shortcomings for our

purposes. First, it provides detailed project and limited sponsor information only after 1995, and even then

on an apparently random basis. As a result, we are unable to collect potentially interesting data items such

as project leverage or sponsor characteristics. Nor are we able to collect information about the lenders

other than name and participation levels. Second, the loan data is incomplete, which reduces our sample

size because most observations lack syndicate share information as well as other key variables such as loan

10 The Loanware database contains information on approximately 18,000 bilateral loans (loans between a single bank and a borrower), but we exclude them from our analysis. 11 Altman and Suggitt (2000) use a $100 million threshold in their analysis of syndicated loan default rates.

11

spreads. Interestingly, Table I shows that the presence of syndicate share information is not related to

either tranche size or signing date.

Of the 2,545 PF tranches greater than or equal to $75 million, 495 tranches worth $150.5 billion

have complete syndicated share information, for a yield of 18.3% by value and 8.8% by number.

Loanware reports syndicate structure as of the closing date, and so we were unable to track what happens

to ownership after the loan has closed.12 As a test for sample selection bias, we compared data that is

available for both the larger set of 2,545 loan trances with data from the smaller subset of 495 tranches

with syndicate share information. For example, the average size and the number of total banks are not

significantly different between the two samples, which provides some evidence against a selection bias.

Given the reported dollar investment made by each bank at closing, we manually calculated several

concentration measures for each tranche, including the Herfindahl-Hirschman Index (HHI)13, the share of

the five largest participating banks (five-bank concentration ratio, CR5), the largest single bank share

(CR1), and the total share held by arranging banks. We defined arrangers to include any bank that is listed

as a mandated arranger, arranger, or co-arranger in the database. As part of this process, we calculated

syndicate size in terms of the number of total banks, number of arranging banks, and number of providing

banks. In addition, we collected information on loan fees and pricing spreads, loan characteristics (tranche

size, maturity, whether the loan was secured or guaranteed, and whether the loan was a refinancing), and

project characteristics (signing date, industrial sector, and location). After reviewing the project

description, we assigned each tranche to one of eight industrial sectors (Industrial, Mining, Oil & Gas,

Petrochemical, Power, Telecom, Transportation, and Leisure & Property). Loanware recently added a

12 Although arranging banks can and do sell down their positions after closing, they usually retain meaningful

positions. In personal interviews with bankers and traders, they are very aware and concerned about signaling effects associated with selling down loans they originated. Borrowers, for similar reasons, dislike when banks sell their loans.

13 The HHI is given by the formula ∑==

N

iiSHHI

1

2 where Si is the dollar share of the ith bank. Other studies of

corporate governance use the Herfindahl index to measure the concentration of control rights (see, for example, Demsetz and Lehn, 1985; or Gorton and Schmid, 1999).

12

sector variable to the database and we verified our classification scheme was consistent with their recent

assignments. Given project location, and our desire to control for sovereign risk independent of project

risk, we collected the most recent Institutional Investor country credit rating (II RATING) prior to

closing—the scale runs from zero (high risk) to 100 (low risk). Institutional Investor publishes ratings

twice per year based on a survey of international bankers.14 The ratings are forward-looking estimates of

sovereign debt capacity and repayment probabilities. As a word of caution, it is important to remember

that the II Rating is an inverse scale so that country risk decreases as the II Rating increases.

Table II presents the geographic distributions for various sub-samples of syndicated loan tranches.

The full sample of syndicated loans and the subset of project finance loans are concentrated in the US and

UK. When countries are sorted by the dollar value of all project loans (column 4), the project sub-sample

exhibits greater dispersion than the full sample: 61.4% of PF loans are in the Top 15 countries compared

to 82.4% of all syndicated loans. The sub-samples of PF loan tranches greater than $75 million (columns 6

and 7) and of PF loans with syndicate data (columns 8 and 9) have geographic distributions similar to

those of the full sample of PF tranches. Although our sample corresponds to the full database of PF loans,

we are somewhat surprised at the relative absence of projects from South American countries.

**** Insert Table II about here ****

With this subset of loan tranches, we then gathered data on the legal rules and level of enforcement in the

countries where the projects are located. We measure creditor and shareholder (anti-director) rights using

LLSV’s (1998, 1999a) indices. The creditor rights index runs from 0 (weak protection) to 4 (strong),15 but

suffers from two problems. First, the index is based on a single point in time and, therefore, do not reflect

changing political or legal conditions over our 15-year sample period. Slightly more problematic is the fact

14 This rating is based assessments provided by 75 to 100 leading international banks. The responses are weighted

by a formula that gives greater weight to banks with more global exposure and better country analysis systems. 15 LLSV (1998, p. 1135) award one point if there is no automatic stay on assets, secured creditors get paid first,

there are restrictions on reorganizations, and if management does not stay in reorganizations.

13

that the index yields some counter-intuitive results. For example, the US, Canada, and Australia are

classified as having weak creditor rights while South Korea, Indonesia, and Egypt are classified as having

the strongest creditor rights. A more general classification scheme based on legal origin—common vs.

civil law—yields more intuitive results: the US, Canada, and Australia are all common law countries

while South Korea, Indonesia and Egypt are civil law countries. LLSV (1998) also provide a classification

by legal origin, and show that common law countries provide stronger legal protection than civil law

countries do, yet provide little explanation for why this is the case. One can argue that common law

systems provide greater flexibility to address new or heretofore unseen situations. In contrast, civil law

systems are restricted to the current body of laws and, therefore, have limited ability to address new or

changing situations. Hoffman (1998, pp. 76-77), in his book The Law and Business of International

Project Finance, notes another reason why common law countries provide stronger creditor rights: they

provide greater leeway in the types of collateral that can be seized in default situations and the types of

liens that can be placed on assets. In contrast, civil law countries forbid “floating liens.”

Creditor rights are of limited use if they cannot be enforced. And so we measure the strength of a

country’s legal system using BPR’s (2001) legality index, which is a summary statistic from a principal

components analysis on five measures of legal enforcement from the PRS Group16: effectiveness of the

judiciary, rule of law, the absence of corruption, risk of contract repudiation, and risk of government

expropriation. The index runs from a low of 8.51 for the Philippines to a high of 21.91 for Switzerland,

has an inter-quartile range of 11.84 (Turkey) to 20.44 (Australia), and covers the same 49 countries in the

LLSV (1998) analysis. As a testament to the importance of legal rights and enforcement, rating agencies

explicitly analyze the validity of contracts, the rights of property owners, the ability to take security over

project assets, and the ability to enforce judgments when they rate project finance transactions (Fitch

IBCA, 2001, p. 10). Appendix 1 provides a breakdown of our sample by legal origin (civil vs. common

law) and by the BPR Legality index (weak vs. strong enforcement).

16 The PRS Group has produced the International Country Risk Guide containing these ratings since 1982.

14

Having described our dataset and defined key variables, we turn to our univariate analysis of the

data. Table III presents four panels describing project characteristics, legal environment, syndicate

structure, and loan pricing variables. To illustrate the importance of size effects, Table III is broken into

two sections, one for all 495 loan tranches greater than $75 million in size, and the other for the 74 loan

tranches greater than $500 million in size.

**** Insert Table III about here ****

II.A. Project and Tranche Variables

Panel A provides general information about the projects and loan tranches in our sample. Because

project data is available for only a limited number of observations, the sample sizes drop from 495 to as

low as 45 observations for some variables. For the full sample, the average (median) tranche size is $304

($180) million and is part of a project costing $820 ($586) million. Panel A also confirms our earlier

assertion that projects are highly leveraged transactions. Projects have a debt-to-total capitalization ratio of

69.4% (70.7%), and the tranches provide 47.8% (43.8%) of total capital. The average tranche matures in

just over nine years, which is long for bank debt but consistent with Kleimeier and Megginson’s (2000)

analysis of syndicated loans more generally. The average tranche is in a country with an II Rating 68.5.

For purposes of comparison, New Zealand, Iceland, and the United Arab Emirates had 1999 II Ratings of

74.0, 67.8, and 63.2, respectively. Yet the projects, as indicated by the large standard deviation, exhibit

significant heterogeneity with respect to sovereign risk. For instance, 10% of our tranches are in countries

with risk ratings below 44.0; Egypt (45.4), India (44.3) and Argentina (42.4) had 1999 II Ratings at this

level. In terms of size effects, the tranches over $500 million are from projects with slightly lower

leverage, but longer maturities (neither is statistically significant).

II.B. Legal Risk Variables

Panel B presents descriptive statistics for our legal rights and enforcement indices. The LLSV

15

(1998) creditor rights index shows that the average score is 2.4 on a scale from 0 to 4. Larger projects tend

to be in countries stronger creditor rights: the median index is 3.0 compared to 2.0 for the full sample.

When it comes to the shareholder rights, both the mean and medians are approximately equal. Finally, the

BPR legality index, has an average score of 17.6 and a standard deviation of 4.0, which indicates there is

significant heterogeneity across our sample in terms of enforcement. Larger projects tend to be located in

countries with stronger legal enforcement.

II.C. Syndicate Structure Variables

Panel C presents a description of syndicate structure, something we consider to be one of the major

contributions of our paper. We find that debt ownership is highly concentrated. On average, the single

largest bank holds 20.3% of the tranche (median of 14.8%), the five largest banks hold 61.2% (57.3

%median), and the average Herfindahl-Hirschman Index is 14.9% (10.25%). It is not necessarily

surprising that the top 5 banks cluster near 60% because waivers of loan provisions typically require

approval by banks controlling at least 60% to 66.67% of the principal outstanding—some changes require

100% of principal. When you consider that tranches represent on average 47.8% of total project capital

(tranche size/total project size from Panel A), the largest single bank is providing almost 10% of total

capital (=20.3% share * 47.8% of total capital; 6.5% at the median). In dollar terms, the largest single

bank holds an average of $61.7 million while the top five banks hold a total of $186.0 million in tranches

greater than $75 million. Thus project debt ownership is significantly more concentrated than US equity

ownership (Morck, Shleifer, and Vishny 1998; McConnell and Servaes, 1990; and Himmelberg, Hubbard,

and Palia, 1999), and more closely resembles equity ownership structures in countries with weak

shareholder protection (La Porta et al, 1999). Even compared to equity blockholdings, debt ownership is

highly concentrated. Holderness and Sheehan (1988) find that 20% (15%) of publicly traded firms in 1984

had at least one non-officer (officer) that owned more than 10% of the firm. We find that the largest single

bank provides 10% or more of the total tranche in 72% of our tranches.

16

Second, there are noticeable size effects across the two samples: the single largest (top five) bank

share declines by more than half, from 20.3% (61.2%) for all tranches greater than $75 million to 9.6%

(36.7%) for the tranches greater than $500 million. Thus, even within our size-biased sample—the

tranches less than $75 million have, presumably, more concentrated ownership structures—we observe

concentrated ownership structures. With regard to the average arranger share, it falls from 16.7% to 7.0%

as you move from the full sample to the largest tranches. What is more surprising is the fact that the total

arranger share does not decline much as size increases: the average total arranger share falls from 39.2%

to 34.2%. The reason total arranger share is relatively invariant to size is that the number of arrangers

increases from 3.6 banks in the average tranche to 5.8 banks in largest tranches.

II.D. Loan Pricing Variables

Panel D describes loan spreads and fees for the loans used in this study—we use the loan spread as

a dependent variable to create a “Loan Pricing Residual” to measure project risk (see below). In terms of

loan fees, the median commitment fee (the fee charged for making funds available) is 30 basis points, the

minimum participation (upfront or closing) fee is 30 basis points for the smallest providers, and the

maximum participation fee is 50 basis points for the largest providers. The variable Undrawn Return is the

sum of all fees paid assuming the borrower does not draw down any of the loan proceeds; it is

approximately equal to the commitment fee. In terms of loan spreads, the mean and median spreads over

LIBOR are 122.8 and 102.5 basis points, respectively (107.9 and 97.5 for the larger tranches). We also

calculate the mean and median spreads over other base rates such as HIBOR (Honk Kong) and SIBOR

(Singapore); this variable is known as the Loan Spread. Finally, we collect the Drawn Return, which

equals the sum of fees and spreads assuming the loan is fully drawn. The mean and median drawn return

is 132.2 and 122.2 basis points. The fees and spreads are slightly smaller for the larger tranches. As a

caveat, note that loan pricing data is reported far less frequently than other data items, a fact that limits our

sample size when we use pricing data to create a proxy for project risk.

17

In summary, this analysis shows that projects are highly leveraged transactions, that project debt is

highly concentrated, and that size has a major effect on debt ownership concentration. We now attempt to

explain the determinants of syndicated structure using regression analysis, with particular attention on the

role of legal rights and enforcement.

III. The Determinants of Syndicate Structure

In this section, we examine the relation between syndicate structure, creditor rights, and legal

enforcement using two sets of Tobit regressions. In the first set of regressions, we use syndicate

concentration as the dependent variable. We measure concentration in six ways: Herfindahl Index, largest

single bank share, combined share of the top five banks, total arranging bank share, average arranging

bank share, and average providing bank share (we do not include total providing bank share because it is

the complement of total arranging bank share). In the second set of regressions, we use syndicate size

(number of total banks, number of arrangers, and number of providers) as the dependent variable. While

syndicate concentration and size are clearly related, they differ in some important ways. The concentration

variables document bank specific shares that distinguish individual banks . In contrast, the size variables

do not distinguish between banks other than by type (arranger vs. provider)—all banks count as a single

observation. Moreover, the concentration measure allows us to analyze monitoring intensity as a

motivation for syndicate structure under the assumption that incentives are correlated with fractional

holdings while the size measure allows us to analyze the re-contracting and deterrence functions under the

assumption that restructuring costs are increase with syndicate size. In both sets of regressions, we use a

Tobit specification because the data are censored: loan shares run between 0% and 100% while syndicate

size has a minimum value of two banks.

The key independent variables are our measures of legal risk. We use a CIVIL LAW dummy

18

variable to measure creditor rights—it equals one for civil law countries, indicating weaker creditor rights,

and zero for common law countries. The BPR LEGALITY INDEX measures legal enforcement. We also

include LLSV’s SHAREHOLDER (anti-director) RIGHTS as an independent variable for two reasons.

On the one hand, stronger shareholder rights could be a measure of shareholder monitoring, which, in turn,

could be a substitute for creditor monitoring. On the other hand, stronger shareholder rights could be used

against creditors. With limited power against shareholders, creditors might resort to high-cost re-

contracting (diffuse ownership) as the only way to curb abusive use of powers. Both explanations predict a

negative relation between debt concentration and shareholder rights.

The independent variables control for tranche characteristics, project characteristics, and sovereign

risk. In particular, the tranche variables are: SIZE (the inverse of tranche size, in millions of US dollars);

MATURITY (tranche maturity, in years); dummy variables equal to one for REFINANCED,

GUARANTEED, and SECURED loans. The project variables include J.P. Morgan’s Emerging Markets

Bond Spread (a weighted index of spreads used to measure lending conditions in emerging markets) and

industry sector dummy variables. We do not include other information on the sponsors—the financing is,

after all, on a standalone basis—or measures of project risk. Without a doubt, other project characteristics

are important. For example, whether a project contains a long-term, off-take (purchase) contract or a

fixed-price, turnkey construction contract has a major effect on the overall level of risk. Yet our database

does not include this information nor can we get it from the proprietary loan documents supporting each

deal. The fact that most project companies are private firms severely hinders data collection.

To address this problem, we create a new variable to measure project risk using the loan spread.

We first regress the loan spread on all of the other independent variables using an OLS specification

(results not shown), and then calculate a LOAN PRICING RESIDUAL for use in the Tobit regressions on

syndicate structure. The idea is that the regression residual will be a proxy for unobserved project risk:

high positive residuals indicate high, unmitigated project risk. Of course, our dependent variable, loan

spread, will contain a premium for sovereign risk. In an attempt to distinguish between sovereign and

19

project risk, we include the II RATING as a measure of sovereign risk in both the loan pricing regressions

and the syndicate structure regressions. We also include a dummy variable for tranches in the United

States because they account for 15.2% of our sample (75/495, Table 1), and because the US has a large

bond market available for borrowers—similar dummy variables for tranches in the UK and Australia, the

countries with the second and third largest number of tranches, were not significant.

III.A. Syndicate Concentration and Legal Risk

Table IV presents the results on the relation between loan concentration, legal rights, and

enforcement. Based on the Chi-square statistics, all of which are significant at the 1% level, the

regressions explain a significant amount of variation in syndicate structure. Unfortunately, data restrictions

reduce the sample from 495 tranches to approximately 300 tranches. Analysis of our key variables

between the included and excluded observations does not reveal significant differences.

**** Insert Table IV about here ****

In almost all of the regressions, the creditor rights and enforcement variables are significant. The CIVIL

LAW (Weak Creditor Rights) dummy variable is negative and significant in all regressions, which implies

that tranches in civil law countries are less concentrated. Importantly, the coefficients are economically

meaningful, as well. For example, the coefficient is negative 0.035 for the largest single bank share.

When you consider that the largest single bank share holds on average 20.3% of the total tranche (see

Table III), the 3.5% decline for tranches in countries with weak creditor rights equals a 17.5% reduction.

Judged against the median value for the largest single bank share of 14.8%, the decline is 23.6%. Similar

calculations for the average Herfindahl Index and total arranger share show declines of 28% and 45%,

respectively, in civil law countries. The existence of a positive relation between creditor rights and debt

ownership concentration stands in contrast to the negative relation between shareholder rights and equity

20

ownership concentration found by LLSV (1998).17

The BPR Legality Index is positive and significant in five out of six regressions. The positive sign

indicates that as enforcement mechanisms strengthen, debt becomes more concentrated and individual

banks are willing to hold larger shares. Moving from a country like Turkey with an index rating of 11.84

to a country like Australia with a rating of 20.44 (the inter-quartile range), increases the largest single bank

share by 8.6% [= (20.44 – 11.84)*0.010]. This change represents a 42% increase over the average share

held by the largest single bank (8.6% divided by 20.3% from Table III).

Thus, we observe less concentrated debt ownership in countries with either weak creditor rights

(civil law countries) or weak enforcement (low BPR index), a finding that is consistent with only the

deterrence motive for structuring debt syndicates. Both the monitoring and the low-cost re-contracting

hypotheses predict more concentrated ownership structures. It appears that creditors view costly

restructuring as an effective way to deter strategic default in countries with high legal risk. An alternative

interpretation, discussed below, is that banks choose to hold smaller portions of riskier loans because of

portfolio diversification concerns.

The other legal rights variable, SHAREHOLDER RIGHTS, has a negative coefficient and is

significant in three out of six regressions. This finding is consistent with either the idea that shareholder

and creditor monitoring are substitutes or the idea that creditors protect themselves from stronger

shareholders by making strategic default more costly.

Of the project, tranche, and location variables, size, sector dummy variables, and sovereign risk are

all significant in at least four regressions. The positive coefficient on the size variable (inverse of size)

implies that larger tranches are less concentrated (consistent with Table III.) The project sector variables

are also jointly significant in five out of the six regressions. Although not significant, the LOAN

PRICING RESIDUAL is positive in five out of size regressions, indicating a tendency towards higher

17 Ongena and Smith (2000b) find that firms maintain fewer and, presumably, larger banking relationships in countries with strong creditor rights and effective legal systems.

21

concentration in projects with some kind of “residual” risk. Along the same lines, the II RATING variable

is negative and highly significant in all but one regression (Regression #4). Given the scale (high ratings

mean low sovereign risk), this means that ownership concentration increases in high-risk countries, and

that the lead banks, the arrangers, hold larger shares of riskier loans; providing banks also hold larger

shares of loans in riskier countries. While limited lending capacity in high-risk markets could explain this

finding, our result is consistent with other research on syndication that attributes the finding to agency

concerns about adverse selection.18 In the case of syndicated bank loans, Simons (1993), Dennis and

Mullineaux (2000), and Jones, Lang, and Nigro (2000) find that arranging banks retain larger shares of

riskier loans. Gorton and Pennachi (1995) find the same thing in the case of loan sales. Apparently,

reputation alone is not sufficient to mitigate concerns about adverse selection. Increased ownership shares

also resolves adverse selection problems in equity transactions. Leland and Pyle (1977), Admati and

Pfleiderer (1994) and Lerner (1994) show that entrepreneurs and venture capitalists signal quality by

holding larger equity positions. Our finding of greater concentration in high-risk countries is not consistent

with a diversification motive.

III.B. Syndicate Size and Legal Risk

Table V presents the results on the relation between syndicate size and legal risk, which is

intended to shed light on the deterrence and re-contracting roles played by member banks. Using a similar

Tobit specification, we find that the regressions have a high degree of explanatory power: the chi-square

statistics are all significant at the 1% level.

**** Insert Table V about here ****

18 While the evidence is consistent with agency explanations, we cannot distinguish between adverse selection or moral hazard as a motivating force. If we had post-closing ownership data, e.g. after sell-down had occurred, then we could distinguish between the two. Adverse selection would predict high original ownership while moral hazard would predict high on-going ownership.

22

Consistent with the previous results on syndicate concentration, all of the coefficients on the

creditor rights and legal enforcement variables are significant. The CIVIL LAW (weak creditor rights)

dummy variable is positively related to the number of total banks and providing banks, but negatively

related to the number of arranging banks. On average, tranches in civil law countries contain 3.0

additional banks, a 21% increase over the average number of banks in a tranche from Table III (3.0 / 14.4

banks). The positive coefficient for providing banks, and the corresponding negative coefficient on

arranging banks, indicates that syndicate size increases because there are more providing banks—this

finding is an example of something we could not detect using share/concentration information only. With

an average of 11.0 providing banks in a syndicate (Table III, Panel C), the coefficient of 4.45 in

Regression #3 indicates that the number of providing banks increases by 41% in countries with weak

creditor rights; the number increases by almost 50% when compared to the median number of providing

banks (9.0 from Table III). Similarly, weak enforcement results in larger syndicates. Moving from

Australia with a rating of 20.44 to Turkey with an index rating of 11.84 increases the total number of

banks by 8.2 banks [ = (11.84 - 20.44)* - 0.956], a 57% increase in syndicate size (8.2 divided by 14.4

from Table III).

Again this evidence, especially in conjunction with the earlier results on syndicate concentration,

is consistent with only the deterrence function. In countries with weak creditor rights or weak

enforcement, syndicates are larger and banks hold smaller shares. As a result, re-contracting is more

complex with more parties involved. At the very least, several additional banks would have to approve

covenant waivers or other exceptions, assuming a 60% or so threshold existed. Clearly this increase in

syndicate size runs contrary to a desire to facilitate low-cost re-contracting. The results also appear to

refute the monitoring hypothesis. At first, the finding that there are fewer arranging banks in situations

with high legal risk would appear to be consistent with a desire for increased monitoring—having fewer

lead banks (arrangers) could reduce free-riding and increase monitoring. The arrangers, however, hold

smaller shares in these settings (Table IV shows the average arranger share falls by 5.3% off an average

23

holding of 16.7%, see Table III), which would reduce their incentives to monitor.

As mentioned in the discussion of hypotheses in Section II, one could interpret the positive

(negative) relation between legal risk and syndicate size (concentration) as evidence of diversification

rather than deterrence. Because the results reflect a net, not a gross, benefit of a give syndicate structure,

both factors probably play a role. We believe, however, that deterrence is more important than

diversification. In civil law countries, syndicates contain up to 50% more providing banks. This dramatic

increase in size not only complicates restructuring attempts, but also makes them more expensive. In

contrast, the average providing bank cuts its share by 1.8% (Table IV, Regression #6) compared to an

average holding of 9.3% (Table III, Panel C). While this decline may seem large, it is not that large in

dollar terms. In fact, the decline of 1.8% represents a decline in the average tranche share of $5.5 million

(1.8%*$304 million average tranche size) compared to an average total holding of $28.3 million

(9.3%*$304 million average tranche size). For the large banks that participate in the syndicated loan

market, the diversification benefits resulting from a relatively small decline in tranche share, especially

compared to a multi-billion dollar loan portfolio, seem small compared to the potential deterrence benefits

created by increasing the syndicate size by 40-50%.

Of the remaining variables, the coefficients on SIZE are negative and significant (larger tranches

include more banks), on MATURITY are negative and significant (longer maturity tranches have more

banks), and on the US dummy variable are negative and significant (US tranches are smaller). Consistent

with the negative coefficient on the II RATING variable in the concentration regressions—banks hold

smaller shares in safer (i.e. low sovereign risk) countries, we observe a positive coefficient on II RATING

in the size regressions, at least in Regressions #1 and #3. Holding creditor rights and enforcement

constant, this finding implies smaller syndicates in countries with higher levels of sovereign risk. One

interpretation of this finding is that projects exposed to higher levels of sovereign risk may be more subject

to liquidity defaults—temporary imbalances between cash inflows and outflows. As a result, bankers want

to ensure rapid approval of covenant waivers in the event of minor problems or low-cost restructuring in

24

the event of more serious problems.

III.C. Sensitivity Analysis

We ran sensitivity analyses to make sure the results in Tables IV and V were robust to alternative

independent variables and regression specifications. With regard to the independent variables, we replaced

the II RATING with ICRG’s composite rating, the inverse of tranche size with the natural logarithm of size

in millions, and year dummy variables instead of the JP Morgan Emerging Markets Bond Index without

changing the results. Instead of the BPR legality index, we tried the various components (legality,

enforcement, and corruption variables) from LLSV (1998) both individually and in combination. The

problem with this approach is that the variables are highly correlated, which is exactly why BPR did the

principal components analysis in the first place. For this reason, the composite legality index provided

more meaningful results. We also changed the sample period by including only tranches from 1990 to

1999 and the regressions specification by running a fixed effects specification to control for inclusion of

loans with multiple tranches, but the results did not change. Based on this analysis, we conclude that our

primary finding, that syndicates are larger and less concentrated in countries characterized by high legal

risk (weak creditor rights or weak enforcement), is robust.

Although we believe our secondary finding, that syndicates are smaller and more concentrated in

countries with greater sovereign risk, is due to agency and re-contracting concerns, an alternative

explanation based on restricted lending capacity in high-risk countries cause this finding. Because lending

to projects in high-risk countries is a complex activity requiring specialized underwriting skills, only a

limited number of banks participate in this market. By necessity, syndicates are smaller and more

concentrated in these markets. If true, this “capacity” hypothesis should be apparent for arranging banks,

the ones responsible for originating new loans.

We test the capacity hypothesis by analyzing lender concentration across various sovereign risk

categories for syndicated project finance loans and for all kinds of syndicated loans (see Panels A and B of

25

Table VI, respectively.) If there is limited lending capacity available, then lender concentration should

increase as sovereign risk increases. We use the collective market share for of the Top 5, 10, and 20

arranging banks in a market as proxies for lender concentration. Panel A, however, shows that the Top 10

Bank Share actually declines from 56.9% to 31.0% as you move from low- to medium-risk countries

(country risk ratings from 100 to 50 in deciles 1 to 5)—similar results exist for the Top 5 and Top 20 Bank

Shares. Contrary to the capacity hypothesis, this decline reflects participation of an increasing number of

arranging banks. For deciles 5 through 10 (medium- to high-risk countries), however, lender concentration

increases from 31.0% to 100.0%, which is consistent with the capacity hypothesis. Especially in deciles 9

and 10, where fewer than 20 different arranging banks are in the market, the capacity hypothesis seems like

a reasonable explanation for the increase in syndicate ownership concentration. Nevertheless, even in

decile 8, there are 45 different arranging banks willing to lend to projects in these countries.19 When you

consider that the typical syndicate has an average (median) of 3.6 (2.0) arrangers (see Table III), these

numbers imply that at least 90% of the available banks do not participate in any given syndicate. A similar

analysis of total participating banks (not shown) indicates there are 211 different banks in decile 8, yet an

average of 11 providing banks appear in a typical syndicate. Again, more than 90% of the available banks

do not participate. For this reason, we believe the capacity hypothesis is a factor in only the most risky

countries (deciles 9 and 10).

**** Insert Table VI about here ****

We conducted two additional tests of the capacity hypothesis using the project finance data and

one other test using data on other kinds of syndicated loans. First, we reran the regression analysis in

Tables IV and V after removing the observations from countries with the highest sovereign risk.

19 The number 45 is an understatement of the actual number of arrangers because the Loanware database treats merged banks as a single bank rather than two banks, and there have been numerous bank mergers during the 1990s. If you run this analysis without correcting for mergers, there are 63 different arranging banks in decile 8, a 40% increase in number. Unfortunately, if you fail to correct for mergers, the database does not consolidate subsidiaries into bank holding companies either. As a result, Chase New York and Chase Hong Kong appear as separate arrangers when, in fact, they are not.

26

Regardless of whether we remove deciles 9 and 10, 8 to 10, or even 7 to 10, the results hold: both

arranging banks and providing banks hold larger shares in riskier countries. Second, we reran the

regression analysis with a proxy for willingness to lend in high-risk countries under the assumption that

fewer banks should be willing to take non US dollar exposures. In other words, fewer banks should be

willing to arrange Thai baht loans compared to US dollar loans for the same Thai projects. To test this

hypothesis, we created an interaction dummy variable equal to one for all non-US dollar tranches in

countries with risk ratings less than 50 (deciles 6 to 10)—we tested alternative thresholds with no change

in results. A positive coefficient in the concentration regressions (larger shares) and a negative coefficient

in the size regressions (fewer banks) would be consistent with the capacity hypothesis. When we reran the

regressions in Tables IV and V (not shown), however, the variable was not significant in any of the

regressions.

As a final test of the capacity hypothesis, we repeated the arranger concentration analysis in Panel

A using the entire set syndicated loans in the Loanware database. Panel B reveals a similar pattern with

decreasing concentration through decile 5 (monotonically until decile 5), which means the finding is not

restricted to the project finance loan market only. The fact that the Top 10 Bank Share is 76.95% in decile

1 indicates a high degree of concentration in a market where capacity cannot be the explanation—these

loans are in low-risk countries and there are 459 different arranging banks present. Instead of optimal

contracting or restricted capacity, we believe that arranger reputation, much like the market for

underwriting IPOs (see Carter and Manaster, 1990), may explain this finding.

The apparent robustness of the positive relation between sovereign risk and syndicate

concentration raises an interesting question. Why would banks respond to greater legal risk with larger,

less concentrated syndicates, but respond to greater sovereign risk with smaller, more concentrated

syndicates. We believe the difference exists because banks are tying to address different kinds of

problems. Legal risk exposes banks to strategic defaults, which they try to deter by increasing syndicate

size. The presence of sovereign risk, on the other hand, creates two different problems. First, there are

27

agency concerns about adverse selection in the syndication market. As seen in other syndication markets,

the leading institutions must hold larger positions to signal credibly loan quality. Second, sovereign risk

may be a proxy for the probability of liquidity defaults. For example, a project may default for sovereign

reasons (e.g. currency inconvertibility) rather than project reasons (e.g. low revenue). Under these

circumstances, lenders want smaller syndicates to ensure low-cost re-contracting. Structure therefore

depends on an assessment of the potential problems in a given market. In summary, we believe that

syndicate structure reflects a complex interaction of capacity considerations, reputation effects, portfolio

concerns, agency conflicts, and optimal contracting. Yet we maintain that the net positive relation between

syndicate concentration and sovereign risk is the result of concerns about adverse selection and low-cost

re-contracting.

IV. Summary and Conclusions

This paper examines the relationship between legal rules and syndicate structure as a way to

improve our understanding of the governance role played by banks in the project finance loan market. We

find that debt ownership is highly concentrated, and significantly more concentrated than equity ownership

in most US industrial firms. While it is true that debt does not have the control rights associated with

equity, except in default scenarios, the structure of debt ownership affects monitoring effectiveness, re-

contracting costs, and deterrence, three benefits traditionally associated with bank debt. Second, we show

that debt ownership concentration is positively related and syndicate size is negatively related to creditor

rights and legal enforcement after controlling for loan size, sovereign risk, and project risk. These findings

provide support for the hypothesis that arranging banks structure loan syndicates to discourage strategic

default by making re-contracting more costly.

We view these results as an initial foray into three, largely unexplored, realms of finance—

28

syndicated lending, project finance, and creditor governance—and believe more research is needed on all

three fronts. For example, we have shown that syndicated loans represent an intermediate form of debt

financing. The fact that PF syndicates range in size from two banks for the smallest loans to 130 banks for

the largest loans such as Eurotunnel’s $13.2 billion loan, has not been adequately recognized in financial

theory. Reality is far more complex than most of the simple financing models admit, and the range

between private debt and public bonds is far more continuous than discrete. Our hope is that this paper

will help inform future models of debt choice by clarifying actual debt ownership structures and

highlighting some the key determinants of syndicate structure. Additional research is needed, however, to

clarify the relation between syndicate structure and sovereign risk, and between syndicate structure and

credit risk. With regard to project finance, we have shown that project-financed investments have

significantly different capital and governance structures than traditional, corporate-financed investments.

Why firms use project finance and how they select particular capital and ownership structures still needs to

be addressed—Esty (2001b) analyzes these issues in a related paper. Finally, with regard to the

governance role played by banks, we have shown that bankers structure syndicates in ways that are

consistent with attempts to deter strategic default. Whether less concentrated syndicates do, indeed, deter

strategic default and whether syndicate structure affects loan pricing are two unanswered questions.

29

Bibliography

Admati, Anat R. and Paul Pfleiderer, Robust financial contracting and the role of venture capitalists,

Journal of Finance 49, 371-402. Altman, Edward I. and Heather J. Suggitt, 2000, Default rates in the syndicated bank loan market: A

mortality analysis, Journal of Banking and Finance 24, 229-253. Berkowitz, D., K. Pistor, and J.F. Richard, 1999, Economic development, legality, and the transplant

effect, University of Pittsburgh working paper, November. Boehmer, Ekkehart and William L. Megginson, 1990, Determinants of secondary market prices for

developing country syndicated loans, Journal of Finance 45, 1517-1540. Bolton, P. and D. Scharfstein, 1996, Optimal debt contracts and the number of creditors, Journal of

Political Economy 104, 1-25. Campbell, Tim S. and William A. Kracaw, 1980, Information production, market signaling, and the theory

of financial intermediation, Journal of Finance 35, 863-882. Carter, R., and S. Manaster, 1990, Initial public offerings and underwriter reputation, Journal of Finance

45, p. 1045-1067. Chowdry, Bhagwan, 1991, What is different about international lending? Review of Financial Studies 4,

121-148. Claessens, Stijn, Simeon Djankov, Larry H.P. Lang, 199, The separation of ownership and control in East

Asian corporations, World Bank working paper. Claessens, Stijn, Simeon Djankov, Joseph P.H. Fan, and Larry H.P. Lang, 2000, Expropriation of minority

shareholders: Evidence from East Asia, Journal of Financial Economics (forthcoming). Datta, Sudip, Mai Iskandar-Datta, and Ajay Patel, 1999, Bank monitoring and the pricing of corporate

public debt, Journal of Financial Economics 51, 435-449. Dell, J.C., J.W. Kunkle, W. Streeter, D. Champeau, and R. Hunter, 2001, Rating Approach to Project

Finance, Fitch IBCA, Duff & Phelps Special Report, April 26. De Long, J. B., 1991, Did Morgan’s men add value? An economist’s perspective on financial capitalism,

in Peter Temin, Ed.: Inside the Business Enterprise: Historical Perspectives on the Use of Information (University Press, Chicago).

Demsetz, H., and K. Lehn, 1985, the structure of corporate ownership: Causes and Consequences, Journal

of Political Economy 93, 1155-117. Dennis, S.A., and D.J. Mullineaux, 2000, Syndicated loans, Gatton College of Business and Economics,

University of Kentucky, mimeo. Diamond, D.W., 1984, Financial intermediation and delegated monitoring, Review of Economic Studies,

pp. 393-414.

30

Diamond, Douglas W., 1991, Monitoring and reputation: The choice between bank loans and directly

placed debt, Journal of Political Economy 99, 689-721. Esty, B.C., 2001a, Structuring loan syndicates: A case study of the Hong Kong Disneyland project loan,

Journal of Applied Corporate Finance, forthcoming Esty, B. C., 2001b, The structure and governance of project-financed investments, Harvard Business

School mimeo, August. Esty, B.C., S. Harris, and K. Krueger, 1999, An overview of the project finance market, Harvard Business

School case study 9-200-028. Esty, B.C., and M. Millett, 1998, Petrolera Zuata, Petrozuata C.A., Harvard Business School case #299-

012. Fama, Eugene F., 1985, What’s different about banks? Journal of Monetary Economics 15, 29-39. Gertner, Robert and David Scharfstein, 1991, A theory of workouts and the effects of reorganization law,

Journal of Finance 46, 1189-1222. Gilson, S.C., K. John, and L.H.P. Lang, 1990, Troubled debt restructurings: An empirical study of private

reorganization of firms in default, Journal of Financial Economics 27, 315-353. Gorton, Gary and Frank Schmid, 2000, Universal banking and the performance of German firms, Journal

of Financial Economics (forthcoming). Gorton, Gary B. and George G. Pennachi, 1995, Banks and loan sales: Marketing non-marketable assets,

Journal of Monetary Economics 35, 389-411. Himmelberg, C., R.G. Hubbard, and D. Palia, 1999, Understanding the determinants of managerial

ownership and the link between ownership and performance, Journal of Financial Economics 53, 353-384.

Hoffman, S.L., 1998, The Law and Business of International Project Finance, Kluwer Law International

(The Hague, The Netherlands). Holderness, C., and D. Sheehan, 1988, The role of majority shareholders in publicly held corporations: An

exploratory analysis, Journal of Financial Economics 20, 317-346. Hoshi, T. A. Kashyap, and D. Scharfstein, 1991, Corporate structure, liquidity, and investment: Evidence

from Japanese industrial groups, Quarterly Journal of Economics 106, 33-50. Houston, Joel and Christopher James, 1996, Bank Information monopolies and the mix of private and

public debt claims, Journal of Finance 51, 1863-1889. Howcroft, B. and C. Solomon, 1985, Syndicated lending by banks, Banger Occasional Papers in

Economics No. 22, J. Revell, ed. (University of Wales Press). James, C., 1987, Some evidence on the uniqueness of bank loans, Journal of Financial Economics,

31

19:217-235. Johnson, Simon, Rafael LaPorta, Florencio Lopez-de-Silanes, and Andrei Shleifer, 2000, Tunneling,

American Economic Review (forthcoming). Johnson, S., P. Boone, A. Breach, and E. Friedman, 1999, Corporate governance in the Asian financial

crisis, MIT working paper. Jones, J., and W.W. Lang, and P. Nigro, 2000, Recent trends in bank loan syndications: Evidence from

1995 to 1999, Economic and Policy Analysis Working Paper #2000-10, Office of the Comptroller of the Currency, Washington, DC, December.

Kang, Jun-Koo and René M. Stulz, 1998, Do banking shocks affect borrowing firm performance? An

analysis of the Japanese experience, working paper, Ohio State University. Kaplan, Steven N. and Bernadette A. Minton, 1994, Appointments of outsiders to Japanese boards:

Determinants and implications for managers, Journal of Financial Economics 36, 225-257. Kleimeier, S. and W.L. Megginson, 2000, Are project finance loans different from other syndicated

credits?, Journal of Applied Corporate Finance 12. La Porta, Rafael, Florencio López-de-Silanes, Andrei Shleifer, and Robert W. Vishny, 1998, Law and

finance, Journal of Political Economy 106, 1113-1150. La Porta, Rafael, Florencio López-de-Silanes, Andrei Shleifer, and Robert W. Vishny, 1999a, The quality

of government, Journal of Law, Economics, and Organization 15, 222-279. La Porta, Rafael, Florencio López-de-Silanes, Andrei Shleifer, and Robert W. Vishny, 1999b, Investor

protection: Origins, consequences, reform, working paper (MIT, Cambridge, MA). Leland, H.E. and D. H. Pyle, 1977, Informational asymmetries, financial structure, and financial

intermediation, Journal of Finance 32, 371-387. Lerner, Josh, 1994, The syndication of venture capital investments, Financial Management 23, 16-27. Levine, R., 1999, Law, finance, and economic growth, Journal of Financial Intermediation 8, pp. 8-35. Lummer, S. L., and J.J. McConnell, 1989, Further evidence on the bank lending process and the capital-

market response to bank loan agreements, Journal of Financial Economics, 22:99-122. McConnell, J. and H. Servaes, 1990, Additional evidence on equity ownership and corporate value,

Journal of Financial Economics 27, 595-612. Megginson, W.L., A.B. Poulsen, and J. F. Sinkey, Jr., 1995, Syndicated loan announcements and the

market value of the banking firms, Journal of Money, Credit, and Banking 27, pp. 457-475. Miller, D.P., and J. Puthenpurackal, 2000, The costs, wealth effects, and determinants of international

capital raising: Evidence from public Yankee bonds, Texas A&M working paper, October. Morck, R. and M. Nakamura, 1999, Banks and corporate control in Japan, Journal of Finance 54, 319-

32

339. Morck, R., A. Shleifer, and R. Vishny, 1988, Management ownership and market valuation, Journal of

Financial Economics 20, 193-215. Ongena, S., and D.C. Smith, 2000a, The duration of bank relationships, Journal of Financial Economics

61, pp. 449-475. Ongena, S., and D. C. Smith, 2000b, What determines the number of bank relationships? Cross-country

evidence, Journal of Financial Intermediation 9, pp. 26-56. Preece, D., and D. Mullineaux, 1996, Monitoring, loan renegotiability, and firm value: The role lending

syndicates, Journal of Banking and Finance 20, 577-593. Project Finance International, IFR Publishing (London, UK), Issues #209 (1/24/01), #185 (1/26/00), #161

(1/27/99), #137 (1/28/98), #113 (1/29/97), #91 (2/28/96), and #67 (3/2/95). Rajan, R.G., 1992. Insiders and outsiders: The choice between informed and arm’s-length debt. Journal of

Finance 47, 1367-1400. Rhodes, T., 2000, Syndicated Lending: Practice and Documentation, 3rd edition, Euromoney Books

(London, UK). Shleifer, A. and R. Vishny, 1997, A survey of corporate governance, Journal of Finance 52, 737-783. Simons, Katerina, 1993, Why do banks syndicate loans?, New England Economic Review

(January/February), 45-52. Terrell, H.S. and H.G. Martinson, 1978, Arranging and marketing syndicated Eurocurrency loans, The

Bankers Magazine (November/December), 35-40. Weinstein, D.E., and Y. Yafeh, 1998, On the costs of a bank-centered financial system: Evidence from the

changing main bank relations in Japan, Journal of Finance 43, 635-672. White, H., 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct test for

heteroskedasticity. Econometrica 48, 817-838. Wurgler, J., 2000, Financial markets and the allocation of capital, Journal of Financial Economics 58, pp.

187-214.

33

Table I Distribution of Syndicated Loans Over Time

This table shows the distribution of syndicated loans from the Loanware database over time (excluding bilateral loans). The columns include all tranches, all project finance tranches, project finance tranches greater than $75 million, and project finance tranches greater than $75 million with syndicate data (our sample).

All Loan Types Project Finance Loans

All Tranches Project Finance Loans Tranches > $75 Million

Project Finance Loans Tranches > $75 million

with Syndicate Data Percent by Signing

Date Value ($B) Number Value ($B) Number Value ($B) Number Value ($B) Number Value ( 8/6) Number (9/7) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) 1980 $91.6 1,096 $10.1 105 $7.7 30 $0.0 0 0.0% 0.0% 1981 181.6 1,535 14.1 146 11.3 46 0.0 0 0.0 0.0 1982 163.5 1,638 9.7 113 7.5 34 0.0 0 0.0 0.0 1983 101.9 1,189 10.7 122 8.4 39 0.0 0 0.0 0.0 1984 201.3 1,704 11.8 148 8.6 40 0.0 0 0.0 0.0 1985 233.7 1,485 6.6 61 5.4 20 0.0 0 0.0 0.0 1986 228.7 1,447 9.8 97 8.2 33 0.8 2 9.8 6.1 1987 363.7 2,041 21.7 143 18.9 58 0.7 4 3.7 6.9 1988 578.0 2,836 19.1 215 14.4 72 0.7 3 4.9 4.2 1989 676.5 3,218 28.5 215 24.6 99 4.4 13 17.9 13.1 1990 574.6 3,626 45.4 232 41.7 109 5.2 25 12.5 22.9 1991 554.2 3,765 49.9 331 44.7 148 6.3 21 14.1 14.2 1992 625.9 4,880 48.9 381 42.6 177 6.9 28 16.2 15.8 1993 788.6 5,031 53.0 398 45.8 174 9.8 31 21.4 17.8 1994 1,073.3 5,732 60.8 386 54.3 183 11.1 39 20.4 21.3 1995 1,396.9 7,019 72.5 493 65.1 241 19.6 73 30.1 30.3 1996 1,609.3 8,317 58.5 455 49.7 187 16.4 66 33.0 35.3 1997 2,056.8 10,016 99.7 513 92.2 279 27.2 77 29.5 27.6 1998 1,698.7 8,703 75.5 459 68.3 243 18.4 61 26.9 25.1 1999 1,947.3 8,028 85.1 544 77.1 282 11.4 46 14.8 16.3 2000 542.2 1,760 31.2 89 30.1 51 11.7 6 38.9 11.8 Total $15,688.4 85,066 $822.5 5,646 $726.6 2,545 $150.5 495 % of Total 5.2% 6.6 4.6 3.0 1.0 0.6 % of Project Finance 88.4 45.1 18.3 8.8 Source: Capital DATA Loanware

34

Table II Distribution of Syndicated Loans by Country

(Sorted by the dollar value of all Project Finance Loans)

This table shows the geographic distribution of syndicated loans from the Loanware database. The columns include all tranches, all project finance tranches, project finance tranches greater than $75 million, and project finance tranches greater than $75 million with syndicate data (our sample).

All Loan Types

Project Finance Loans

All Tranches

Project Finance Loan Tranches >$75

Million

Project Finance Loan Tranches >$75

Million with Syndicate Data Percent by

No. Country Value ($B) Number

Value ($B) Number

Value ($B) Number

Value ($B) Number

Value (8/6)

Number (9/7)

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)

1 United States 9,483.8 43,879 124.64 489 124.6 489 $20.2 75 15.8% 14.7% 2 United Kingdom 1,390.2 5,914 101.05 207 101.1 207 17.1 47 16.2 21.3 3 Australia 370.9 1,833 40.34 140 40.3 140 10.9 33 26.5 23.1 4 Indonesia 108.4 1,492 40.09 120 40.1 120 8.9 30 21.4 23.6 5 China 96.1 1,528 24.55 116 24.6 116 7.3 38 29.4 32.2 6 Taiwan 67.7 548 23.58 48 23.6 48 17.5 25 74.2 52.1 7 Malaysia 78.6 938 22.40 83 22.4 83 5.6 24 24.2 27.3 8 Canada 533.8 2,312 22.27 78 22.3 78 2.3 7 10.3 8.8 9 Hong Kong 176.2 577 22.00 77 22.0 77 13.9 38 62.6 48.1

10 Thailand 74.3 1,197 20.92 77 20.9 77 5.1 18 21.6 37.5 11 Italy 258.4 1,897 13.72 31 13.7 31 0.8 5 5.8 15.6 12 Saudi Arabia 65.6 302 13.56 39 13.6 39 3.6 8 25.2 20.0 13 Turkey 65.4 891 12.61 79 12.6 79 1.6 7 12.3 8.5 14 Mexico 138.1 742 11.64 40 11.6 40 0.5 4 3.7 8.0 15 Qatar 15.3 50 11.58 28 11.6 28 0.6 2 5.2 7.1

Average 22.6% 22.2% Totals: Top 15 Countries $12,922.7 65,166 $505.0 1,652 505.0 1,652 $115.9 361 Full Database $15,688.4 85,066 $822.5 5,646 $726.6 2,545 $150.5 495 Top 15 /Full Database 82.37% 76.61% 61.39% 29.26% 69.50% 64.91% 77.01% 72.93%

Source: Capital DATA Loanware

Table III Univariate Analysis of Syndicate Structure

This table provides a univariate description of the main variables used in the study. The variables are broken into four groups: project variables (Panel A), legal variables (Panel B), syndicate structure variables (Panel C), and loan pricing variables (Panel D). The table shows the number of loan tranches with data available, and the mean, median, and standard deviation for each variable.

Project Finance Loan Tranches >$75m Project Finance Loan Tranches >$500mVariable Number Mean Median Std. Dev. Number Mean Median Std. Dev. Panel A: Project and Loan Variables

Project Size ($ millions) 113 820.0 586.0 1302.1 20 1927.9 1304.7 2683.8Tranche Size ($ millions) 495 304.0 180.0 540.9 74 948.2 702.3 1194.3Leverage: Debt/Total Project Size (%) 45 69.4% 70.7% 13.7% 9 65.6% 68.0% 13.4%Tranche/Total Debt (%) 45 66.3% 67.6% 28.2% 9 60.5% 59.5% 34.9%Tranche/Total Project Size (%) 113 47.8% 43.8% 26.2% 20 42.5% 43.6% 23.2%Maturity (Years) 474 9.4 9.0 4.8 74 10.2 10.0 5.6Instit. Investor Rating [0-100 low risk] 493 68.5 68.5 18.2 74 70.4 72.5 16.7

Panel B: Legal Risk Variables

LLSV (1998) Creditor Rights [0-4 strong] 406 2.4 2.0 1.4 57 2.5 3.0 1.5LLSV (1998) Shareholder Rights [0-6 strong] 408 3.8 4.0 1.3 57 4.1 4.1 1.2BPR (1999) Legality Index 408 17.5 19.1 4.0 57 18.5 20.4 3.6

Panel C: Syndicate Structure Variables

Concentration Ratios Herfindahl-Hirschman Index (HHI) 495 14.9% 10.25% 13.5% 74 5.7% 4.9% 3.4%Largest Single Share (CR1) 495 20.3% 14.8% 16.2% 74 9.6% 8.0% 5.3%Top Five Shares (CR5) 495 61.2% 57.3% 25.6% 74 36.7% 33.4% 15.6%

Arranging Banks Total Arranger Share (%) 495 39.2% 33.0% 28.5% 74 34.2% 30.2% 25.9%Number of Arrangers 495 3.6 2.0 3.9 74 5.8 4.0 5.8Average Arranger Share (%) 490 16.7% 11.4% 16.1% 73 7.0% 5.9% 5.0%

Providing (Non-arranging) Banks Total Non-Arranger Provider Share 495 60.8% 67.0% 28.5% 74 65.8% 69.8% 11.6%Number of Non-Arranger Providers 495 11.0 9.0 9.3 74 22.2 21.0 11.6Average Non-Arranger Provider 457 9.3% 6.5% 9.9% 73 3.6% 3.1% 2.5%

Total Number of Banks 495 14.4 12.0 10.3 74 28.0 25.5 11.7Syndicate Shape (Providers/Arrangers) 490 5.7 3.5 6.5 73 8.4 5.2 9.4

Panel D: Loan Pricing Variables

Fees (bp) Commitment 262 31.9 30.0 20.3 47 29.1 25.0 15.1Participation (Minimum) 266 36.9 30.0 29.3 56 29.8 27.5 18.5Participation (Maximum) 266 53.1 50.0 36.1 56 47.9 42.5 29.5

Spreads (bp) LIBOR Spread 292 130.7 120.0 83.0 46 107.9 97.5 59.2Loan Spread 404 122.8 102.5 79.1 68 105.0 95.0 53.1Drawn Return 287 132.2 122.2 77.9 46 112.3 100.8 60.3Undrawn Return 349 30.4 28.4 23.5 65 28.5 28.6 19.8

Source: Capital DATA Loanware; La Porta, Lopez-de-Silanes, Shleifer, Vishny (LLSV, 1998).

36

Table IV Determinants of Syndicate Structure—Concentration and Share Measures

This table shows the results of Tobit regressions on syndicate structure variables. The dependent variables are various debt ownership concentration measures: Herfindahl Index, largest single bank share, share of the five largest banks, arranging banks, total and average share for arranging banks, and average share for providing banks. The table shows the coefficient and t-statistic, corrected for heteroscedasticity, in parentheses.

Concentration and Share Measures

Herfindahl Index

Largest Single Share

Top 5 Banks Share

Total Arranger

Share

Average Arranger

Share

Average Provider

Share Reg. #1 Reg. #2 Reg. #3 Reg. #4 Reg. #5 Reg. #6 Constant 0.106 *

(2.15) 0.145 *

(2.43) 0.502 **

(4.34) 0.776 **

(5.14) 0.120 *

(1.75) 0.037

(1.17) Weak Creditor Rights (Civil

Law) Dummy Variable -0.042 *

(-2.52) -0.035 *

(-1.70) -0.111 **

(-2.86) -0.177 **

(-3.48) -0.053 *

(-2.30) -0.018 *

(-1.67) LLSV Shareholder Rights

-0.010 (-1.40)

-0.014 (-1.59)

-0.036 * (-2.22)

-0.020 (-0.96)

-0.021 * (-2.21)

-0.001 * (-2.24)

BPR Legality Index

0.008 ** (2.61)

0.010 ** (2.67)

0.031 ** (4.23)

-0.008 (-0.85)

0.012 ** (2.77)

0.010 ** (4.68)

Loan Pricing Residual

0.0001 (1.02)

0.0001 (0.80)

0.0003 (1.51)

0.0002 (0.87)

0.0001 (0.91)

-9.0e-06 (-0.20)

Inverse of Tranche Size ($m) 13.329 ** (8.33)

15.812 ** (7.88)

39.592 ** (10.38)

6.872 (1.40)

14.464 ** (6.50)

8.458 ** (8.10)

Institutional Investor Rating -0.002 ** (-3.09)

-0.003 ** (-3.04)

-0.008 ** (-4.49)

-0.002 (-0.87)

-0.003 * (-2.53)

-0.002 ** (-4.88)

Maturity (Years) 0.002 (1.53)

0.002 (1.12)

0.008 ** (2.68)

-0.002 (-0.55)

0.002 (1.36)

0.002 ** (3.14)

Refinanced Loan Dummy Variable

-0.012 (-0.65)

-0.017 (-0.76)

-0.013 (-0.31)

-0.009 (-0.16)

0.0005 (0.02)

0.005 (0.43)

Guaranteed Loan Dummy Variable

0.004 (0.34)

0.007 (0.43)

0.020 (0.67)

-0.020 (-0.51)

-0.0001 (-0.00)

0.003 (0.39)

Secured Loan Dummy Variable

0.006 (0.58)

0.015 (1.06)

-0.005 (-0.18)

0.046 (1.38)

0.019 (1.22)

-0.003 (-0.41)

JP Morgan Emerging Market Bond Spread (bps)

6.6e-06 (0.37)

0.00002 (1.00)

0.00001 (0.31)

0.0001 (1.52)

0.00002 (0.73)

-6.1e-06 (-0.52)

US Dummy Variable 0.039 * (2.22)

0.060 ** (2.74)

0.156 ** (3.77)

-0.033 (-0.62)

0.035 (1.43)

0.048 ** (4.25)

Sector Dummy Variables Included * Included * Included * Included * Included * Included Number of Observations 304 304 304 304 303 290 Likelihood Ratio 98.26 96.56 151.55 43.06 68.75 114.05 Prob. > Chi-Square 0.000 0.000 0.000 0.001 0.000 0.000

Source: Capital DATA Loanware database; La Porta, Lopez-de-Silanes, Shleifer, Vishny (LLSV, 1998). Note: * and * * denote significance at the 10% and 1% level in a one-tailed test, respectively.

NM denotes not meaningful.

37

Table V Determinants of Syndicate Structure—Syndicate Size

This table shows the results of Tobit regressions on the number of banks included in the syndicate. The dependent variables are the number of total banks, arranging banks, and providing banks.

Syndicate Size (Number of Banks)

Total Banks

Arranging Banks

Providing Banks Reg. #1 Reg. #2 Reg. #3 Constant 29.256 **

(6.52) 17.298 ** (5.55)

15.192 ** (3.63)

Weak Creditor Rights (Civil Law) Dummy Variable

3.028 * (2.01)

-2.465 * (-2.34)

4.450 ** (3.16)

LLSV Shareholder Rights 0.670 (1.08)

-0.162 (-0.37)

0.786 (1.36)

BPR Legality Index

-0.956 ** (-3.35)

-0.442 * (-2.26)

-0.685 * (-2.57)

Loan Pricing Residual

-0.013 * (-1.96)

-0.001 (-0.21)

-0.009 (-1.51)

Inverse of Tranche Size ($m) -1707.247 ** (-11.68)

-529.457 ** (-5.03)

-1382.522 ** (-10.12)

Institutional Investor Rating 0.192 ** (2.86)

-0.015 (-0.31)

0.205 ** (3.27)

Maturity (Years) -0.326 ** (-2.93)

-0.092 (-1.21)

-0.245 * (-2.37)

Refinanced Loan Dummy Variable 1.329 (0.82)

-0.167 (-0.15)

1.350 (0.89)

Guaranteed Loan Dummy Variable -0.174 (-0.15)

-1.263 (-1.54)

0.629 (0.58)

Secured Loan Dummy Variable 0.022 (0.00)

-0.146 (-0.21)

0.086 (0.09)

JP Morgan Emerging Market Bond Spread (bps)

-0.001 (-0.90)

0.001 (0.79)

-0.001 (-0.97)

US Dummy Variable -2.921 * (-1.83)

-0.983 (-0.86)

-2.546 * (-1.71)

Sector Dummy Variables Included * Included Included * Number of Observations 304 304 304 Likelihood Ratio 154.34 62.27 128.37 Prob. > Chi-Squared 0.000 0.000 0.000

Source: Capital DATA Loanware database; La Porta, Lopez-de-Silanes, Shleifer, Vishny (LLSV, 1998). Note: * and * * denote significance at the 10% and 1% level in a one-tailed test, respectively.

NM denotes not meaningful.

38

Table VI Mandated arranger bank shares in different types of syndicated lending

This table details the average top 5, 10 and 15 bank share for mandated arrangers of project finance and all syndicated loans, ranked by Institutional Investor (II) country risk ratings. The II Ratings are broken into deciles from 1 (low risk, high II Rating) to 10 (high risk, low II Rating) for the period 1990 to 2000. The samples exclude bilateral loans.

Panel A: Project finance loans grouped by II risk decile

Decile #

II Range

Total number of loans

Value of loans $US

Bn

Number of arranging

banks

Top 5 bank share

Top 10 bank share

Top 20 bank share

1 (low) (90≤II<100) 472 $97.9 97 37.81% 56.91% 75.60% 2 (80≤II<90) 380 112.2 83 37.91 59.32 82.87 3 (70≤II<80) 361 65.9 123 29.59 43.29 65.83 4 (60≤II<70) 446 66.5 156 23.66 37.17 56.98 5 (50≤II<60) 623 73.5 215 17.29 31.04 49.86 6 (40≤II<50) 430 45.7 145 35.02 52.79 71.34 7 (30≤II<40) 224 23.8 97 37.38 55.03 73.15 8 (20≤II<30) 95 14.0 45 46.59 71.18 91.24 9 (10≤II<20) 19 2.3 14 71.53 96.39 (14 banks) 10 (high) (0≤II<10) 2 1.1 7 87.32 100.00 --

All loans (0≤II<100) 3,052 $502.9 462 23.11% 37.96% 56.53%

Panel B: All syndicated loans, grouped by II risk decile

Decile #

II Range

Total number of loans

Value of loans $US

Bn

Number of arranging

banks

Top 5 bank share

Top 10 bank share

Top 20 bank share

1 (low) (90≤II<100) 27,898 $6,814.8 459 62.55% 76.95% 87.18% 2 (80≤II<90) 5,807 1,643.1 264 34.37 54.15 78.26 3 (70≤II<80) 4,136 739.6 369 29.34 43.50 62.71 4 (60≤II<70) 2,855 287.9 323 24.32 36.24 54.74 5 (50≤II<60) 2,559 225.4 335 25.76 38.66 56.32 6 (40≤II<50) 1,997 213.9 275 33.44 48.77 68.04 7 (30≤II<40) 947 120.1 192 47.43 65.03 82.20 8 (20≤II<30) 527 57.6 139 36.56 55.96 73.53 9 (10≤II<20) 110 8.7 32 60.74 78.70 93.96 10 (high) (0≤II<10) 20 2.8 21 58.77 83.32 99.55

All loans (0≤II<100) 47,703 $10,237.0 1,269 49.50% 62.58% 76.12

39

Appendix 1 Matrix of Countries Based on Political Risk

This table shows a distribution of sample countries based on their level of legality rating (Weak vs. Strong Enforcement on the BPR Legality Index) and their legal origin (Civil vs. Common Law from LLSV, 1998). The numbers in parentheses following the country indicate the Institutional Investor country credit rating as of Septermber 1999, and the Berkowitz, Pistor, and Richard (1999) legality ratings, respectively. Legal origin is a proxy for creditor rights: on average, common law countries provide strong creditor protection than civil law countries do. The countries in bold have the most number of projects in our database.

BPR Legality Index Weak Enforcement

(Low Legality Index) Strong Enforcement (High Legality Index)

Common

Law (stronger)

Malaysia (51.7, 16.67) Thailand (48.3, 12.94) India (44.2, 12.80) Kenya (24.8, 12.00) Zimbabwe (25.1, 11.59) Nigeria (17.9, 9.39)

New Zealand (74.0, 21.55) Canada (83.5, 21.13) United States (90.9, 20.85) Australia (75.8, 20.44) United Kingdom (90.2, 20.41) Singapore (81.9, 19.53)

LLSV Legal Origin

Civil Law (weaker)

Brazil (36.5, 14.09) Mexico (48.2, 12.82) Argentina (42.4, 12.34) Turkey (38.9, 11.84) Columbia (44.1, 11.58) Peru (37.0, 10.10) Indonesia (27.1, 9.16) Philippines (45.9, 8.51)

Switzerland (93.0, 21.91) Denmark (85.1, 21.55) Austria (89.4, 20.76) Germany (92.0, 20.44) France (91.4, 19.67) Spain (87.2, 17.13) Taiwan (75.3, 17.62)