n?u=RePEc:red:sed016:701&r=sog

**Multiple** **Lenders**, **Strategic** **Default** **and** **Covenants** ∗

Andrea Attar † Catherine Casamatta ‡ Arnold Chassagnon § Jean Paul Décamps

Abstract

We study capital markets subject to moral hazard when investors cannot prevent side trading,

thereby facing an externality if firms raise funds from multiple sources. We analyze whether

investors’ ability to design financial covenants that may include exclusivity clauses mitigates

this externality. Following covenant violations, investors can accelerate the repayment of their

loan, adjust its size, or increase interest rates. Enlarging contracting opportunities generates

a severe market failure: with covenants, equilibria are indeterminate **and** Pareto ranked. We

show that an investors-financed subsidy scheme to entrepreneurs alleviates the incentive to

overborrow **and** sustains the competitive allocation as the unique equilibrium one.

Keywords: Side Trading, Financial **Covenants**, Nonexclusive Competition, **Strategic** **Default**.

JEL Classification: D43, D82, G33.

∗ We thank B. Biais, A. Bisin, M. Bruche, F. Castiglionesi, P.A. Chiappori, H. Degryse, P. Gottardi, D. Gromb, A.

Gümbel, M. Hellwig, B. Jullien, V. Larocca, T. Mariotti, U. Rajan, R. Townsend, P. Woolley **and** audiences at CSEF,

EM Lyon, Erasmus University, EUI, HEC-Paris, Luiss University, Max Planck Institute (Bonn), Tilburg University,

Tinbergen Institute, Università di Roma Tor Vergata, Université de Caen, Université Paris-Dauphine, University of

Cologne, the 2011 SAET meeting, the 2012 Robert Townsend Toulouse Conference, ESEM 2012, the 2012 CESIFO

conference on Applied Microeconomics, the 2013 UECE Lisbon meeting **and** the 2015 European Summer Symposium

in Financial Markets (Gerzensee) for valuable feedback. This research was conducted within **and** supported by the

Paul Woolley Research Initiative on Capital Market Dysfunctionalities at IDEI-R, Toulouse. Financial support from

the Agence Nationale de la Recherche ANR-09-BLAN-0358-01 is gratefully acknowledged.

† Università di Roma Tor Vergata, **and** Toulouse School of Economics (CRM, IDEI)

‡ Toulouse School of Economics (CRM, IDEI)

§ Université de Tours **and** Paris School of Economics

Toulouse School of Economics (CRM, IDEI)

1

1 Introduction

When they provide capital to firms, investors do not always control all the financial transactions

firms enter into. To the extent that outside financing affects firms’ behavior **and** their ability to

meet contractual obligations, the possibility of side trading creates an externality for investors: if

an entrepreneur borrows from multiple sources, she may have less incentives to improve financial

performance, which in turn affects each lender’s profit. Whether this externality can be internalized

by letting agents design appropriate financial contracts is a central question for the regulation of

modern capital markets. Yet, it remains largely unanswered. As a matter of fact, the financial

literature has either evaded the issue by postulating that agents enforce exclusive relationships from

the outset, or magnified the problem by considering an extreme form of nonexclusivity whereby

few, if any, contractual instruments can limit side trading. Capital markets probably fall between

these two extremes. On the one h**and**, no legal system imposes that entrepreneurs must raise funds

from a single investor. On the other h**and**, financial transactions are seldom secret **and** investors can

use the available information to design financial covenants in an attempt to curb firms’ financing

policy.

This paper combines these two elements to study the functioning of capital markets subject

to moral hazard. Our fundamental insight is that competition over sophisticated contracts, that

potentially include exclusivity clauses, leads to a severe market failure. The result challenges

the conventional view that a system of contracts fully contingent on additional financing is able

to enforce exclusive trading **and** foster market efficiency. As a consequence, to circumvent the

externality induced by side trading, a mere increase in the transparency of trades may be more of

a problem than a solution. Financial regulation should explicitly take into account the investors’

strategic use of information.

Investors’ contract design under the threat of competition is a relevant issue in corporate finance.

In practice, firms can enter into multiple financial relationships **and** investors extensively

use financial covenants in their contracts to define thresholds of financial ratios that firms must

comply with. 1 Since violating such covenants triggers penalties, this is a way for investors to con-

1 Rauh **and** Sufi (2010) document that most large rated firms simultaneously use different types **and** sources of

corporate debt. According to Roberts **and** Sufi (2009), almost 97% of credit agreements contain financial covenants,

90% of which limit implicitly or explicitly firms’ ability to raise additional debt (see also Demiroglu **and** James (2010)).

1

trol firms’ financing policy. Intuitively, the ability to design covenants should enhance competition

by limiting firms’ opportunity to raise additional funds, thereby eliminating the source of the externality.

In a strategic context, however, covenants can act as a barrier to entry to protect incumbents’

rents against the threat of new offers. 2

These effects are analyzed in the following model. A representative entrepreneur seeks funds

to invest in a project subject to moral hazard: she has to forgo a private benefit **and** exert effort to

increase the project’s profitability. A finite number of investors compete by simultaneously offering

menus of financial contracts. If the entrepreneur’s aggregate investment is observable, contracts

can include financial covenants. The entrepreneur selects one contract in each menu, possibly

raising funds from several investors, **and** chooses an effort level. If, in a given state of nature,

the sum of payments promised to investors exceeds the firm’s cash flow, then, given her limited

liability, the entrepreneur strategically defaults. 3 All the cash flow is then seized by investors, but

the entrepreneur’s private benefit is inalienable.

Our benchmark is a fully nonexclusive setting: investors cannot write financial covenants, i.e.

they are restricted to trade bilateral contracts with the entrepreneur. In this context, we provide a

complete characterization of equilibrium allocations. Despite the fact that investors earn a positive

profit at equilibrium, all aggregate allocations are constrained efficient. The result is robust, in the

sense that it is established by allowing investors to post arbitrary menus of contracts.

We next introduce financial covenants. **Covenants** prescribe punishments to the entrepreneur

when she departs from a target financing policy. To the extent that the latter is determined by

the entire profile of investors’ offers, this introduces a degree of "contractibility of contracts" **and**

allows us to explore the investors’ ability to write exclusive contracts in a strategic setting. In

our analysis, the set of punishments available to investors depends on when the information on

outside financing is obtained, **and** it is sufficiently large to include the penalties commonly observed

following a covenant violation: acceleration of the loan repayment, adjustment of the initial loan

size, or raise in the interest rate.

We first consider that investors only observe the firm’s financing policy after the project is un-

2 The possibility that exclusive contracts play an anticompetitive role has been suggested by Aghion **and** Bolton

(1987) in a monopolistic setting with no side trading externality.

3 **Default** is denoted strategic because it is determined by the contracts chosen by the entrepreneur.

2

dertaken. **Covenants** then allow them to modify interest rates according to the entrepreneur’s level

of outside debt. The entrepreneur’s limited liability sets an upper bound on the monetary penalties

that any investor can impose when she trades with others. As a consequence, financial covenants

do not have the power to simply forbid multiple financial relationships. They do however introduce

new strategic interactions compared to the benchmark case of fully unobservable side trades. Each

deviating investor can exploit the protection of covenants to serve the market alone by punishing,

up to the realized cash flows, the entrepreneur’s attempt to raise funds from multiple financiers. At

the same time, any incumbent investor can write covenants to prevent his competitors from providing

additional investment **and** complementing existing offers. We show that the combination

of these two effects leads to an indeterminacy: a large number of Pareto ranked allocations are

sustained at equilibrium.

We next modify the information structure **and** allow investors to observe outside financing

before production takes place. This enlarges de facto the class of available financial covenants: investors

can adjust their contractual terms according to the entrepreneur’s intended financial policy.

In particular, they can include loan acceleration clauses if they observe excessive debt issuance,

or propose additional financing if they observe insufficient borrowing. This second feature can be

interpreted as a contingent line of credit. It turns out that the strategic provision of lines of credit

discourages more competitive offers backed by a loan acceleration clause. This reinforces the idea

that enlarging investors’ contractual opportunities does not enhance competition. Overall, the same

indeterminacy result holds.

On the normative side, indeterminacy of market equilibria provides a natural setting to study

how a proper design of the market fosters investors’ incentives to supply capital. We construct a

system of subsidies from investors to the entrepreneur, that grants her a transfer if her investment is

too low. This system enables one investor to exploit gains from trade with a more competitive offer

by alleviating the entrepreneur’s resulting incentive to default. Investors do not choose the size of

the subsidies but have the option to finance them or not. The decision results from comparing

the cost of the subsidy for each investor **and** his gain from avoiding default when the transfer

is provided to the entrepreneur. We show that a subsidy scheme can be designed so that all gains

from trade are exploited, **and** only the competitive allocation is sustained at equilibrium. This result

provides a possible foundation for regulatory schemes that make banks or financial intermediaries

3

financially liable whenever their strategic interactions exacerbate the risk of default. Examples

of related market institutions include the guarantee funds of central clearinghouses in derivative

markets or the recent Single Resolution Fund (SRF) of the European banking union.

Relationship to the literature

Our paper is linked to the general issue of side trading in financial markets, 4 **and** more directly

speaks to the literature on nonexclusive competition that endogenizes side trading opportunities as

part of the strategies of competing investors. Bizer **and** DeMarzo (1992) **and** Kahn **and** Mookherjee

(1998) are the first to establish that nonexclusivity creates an externality in financial markets

subject to moral hazard. Parlour **and** Rajan (2001) **and** Bisin **and** Guaitoli (2004) explicitly consider

strategic competition among lenders. They show that, compared to a hypothetical exclusive

competition benchmark, nonexclusivity typically reduces trades **and** generates an extra profit for

investors. 5 In these models, financial contracts are bilateral in essence: no supplier can condition

his offers on the customer’s entire profile of trades. Bilateral contracting does not fit well with the

practice of modern capital markets in which investors set punishments based on entrepreneurs’ financial

policies. We therefore study financial relationships when investors write general contracts

that may include exclusive clauses. We show that the use of such contracts generates a severe

market failure. **Multiple** Pareto ranked allocations are sustained at equilibrium.

Equilibrium indeterminacy is a major result of the competing mechanism literature in which

principals compete through mechanisms in face of several privately informed agents. Yamashita

(2010), Peters **and** Troncoso-Valverde (2013), Peters (2015), **and** Han (2014) provide different

versions of a Folk Theorem: if no restriction is put on the set of available mechanisms, then a

very large number of allocations can be supported at equilibrium. These results crucially hinge

on having multiple agents. A contribution of our analysis is hence to present a st**and**ard economic

setting in which indeterminacy arises with a single agent thanks to the strategic role of covenants. 6

**Strategic** default is also modeled as the entrepreneur’s ability to divert cash instead of repaying

4 See for instance Allen **and** Gale (2004) or Farhi et al. (2009).

5 A form of market breakdown also arises in adverse selection settings, where nonexclusivity often leads to Akerlof

(1970)-like results (Attar et al. (2011) **and** Attar et al. (2014)).

6 Similar insights arise in the game-theoretic analysis of Szentes (2015) who characterizes equilibria of games in

which several principals can write potentially sophisticated contractible contracts.

4

investors in multiple lending relationships: Bolton **and** Scharfstein (1996) offer the view that having

multiple lenders weakens the incentive to default by affecting ex post renegotiation between

parties. Bolton **and** Jeanne (2009) apply this analysis to sovereign lending, showing that governments

prefer to issue different types of debt. In these models, the incentive to default arises from

the impossibility to write contracts contingent on cash flows. In contrast, we allow for contingent

contracts, **and** we emphasize that the possibility to have multiple financiers is at the origin of the

entrepreneur’s willingness to default. 7

The paper is also linked to the literature on covenants in financial contracts. The conventional

approach is to study how covenants can mitigate agency conflicts between shareholders **and**

debtholders (Smith **and** Warner (1979), Myers (1977)). Much attention has been devoted to explain

how covenants affect renegotiation between parties (Gorton **and** Kahn (2000), Gârleanu **and**

Zwiebel (2009)). We take a different view **and** investigate how the design of covenants affects

competition between debtholders.

Several regulatory mechanisms have been suggested to mitigate the damaging implications of

side trading. Some rely on enhancing the observability of trades. Acharya **and** Bisin (2014) construct

a centralized clearing mechanism that sustains efficiency of market equilibria. Bennardo et

al. (2015) study the role of credit bureaus in a fully nonexclusive setting **and** show that information

sharing systems decrease investors’ rents, although they can worsen credit rationing. Both mechanisms

fail to take into account the strategic design of contracts by individual investors. We show

that this possibility creates barriers to entry that resist regulations enhancing the observability of

trades. In our analysis, financial covenants are responsible for market inefficiencies whenever investors

can observe outside debt. Others point out the role of repayment rules in case of default.

In particular, Bisin **and** Rampini (2006) put forward the idea that the institution of bankruptcy can

improve on nonexclusive contractual relationships but is not a perfect substitute for exclusivity.

In our setting, market failures arise even if a bankruptcy procedure imposes repayment priority

or captures part of the entrepreneur’s private benefit. In both cases, the threat of overborrowing

is sufficiently strong to impede perfect competition. A third class of mechanisms involves transfers

to distressed agents in order to affect their willingness to trade. Tirole (2012) **and** Philippon

**and** Skreta (2012) stress the fact that a regulator can repurchase assets, reducing adverse selection

7 We argue in Section 4 that our results extend to the case in which the entrepreneur can also divert funds.

5

**and** restoring agents’ participation on the private market. Our subsidy scheme exhibits a similar

feature in a moral hazard setting. From the viewpoint of an investor, providing a subsidy modifies

the entrepreneur’s incentive to default which in turn affects the investor’s willingness to pay

for it. An important difference with Tirole (2012) **and** Philippon **and** Skreta (2012) is that in our

case the subsidy is entirely financed by the lending sector without resorting to an external costly

source of funds. We therefore provide a rationale for the private regulation of counterparty risk

through a collective transfer system, in the spirit of the guarantor function performed by central

clearinghouses.

The remainder of the paper is organized as follows. The model is exposed in Section 2. The

benchmark case in which investors offer bilateral contracts is analyzed in Section 3, while Section

4 studies the impact of covenants on equilibrium outcomes. Section 5 is devoted to market design.

Proofs are provided either in the Appendix or in the additional Appendix A. The additional

Appendices B, C **and** D provide robustness results.

2 The model

We build on the st**and**ard capital market model of Holmstrom **and** Tirole (1997, 1998).

Agents, technology **and** preferences. We consider a production economy populated by a single

representative entrepreneur **and** a finite number N of investors. At date 0, the entrepreneur owns

a variable size project that generates at date 1 a r**and**om output over two verifiable states: an

investment of I ≥ 0 yields a cash flow GI with G > 0 if the project succeeds, **and** a cash flow of

0 if it fails. The probability distribution over states depends on an unobservable effort e = {L, H}

chosen by the entrepreneur. Let (π e , 1 − π e ) be the distribution induced by effort e, where π e is the

probability of success. Denote e = H the high level of effort, **and** assume that π H > π L . If the

entrepreneur selects e = L, she receives a private benefit B ≥ 0 per unit invested in the project.

As in Holmstrom **and** Tirole (1997, 1998), the investment project has a positive net present value

if **and** only if the entrepreneur selects e = H, that is:

π H G > 1 > π L G + B. (1)

6

The entrepreneur is risk-neutral **and** protected by limited liability. She has a verifiable endowment

of A > 0 **and** can raise additional funds by trading financial contracts issued by competing investors.

If she raises I units of funds, invests her endowment A, pays back R in case of success,

( )

**and** 0 in case of failure, her net payoff is π H G(I + A) − R − A if she chooses e = H, **and**

( )

π L G(I + A) − R + B(I + A) − A if she chooses e = L. The expected profit of investor i

when he lends I i to the entrepreneur **and** obtains R i in case of success **and** 0 in case of failure, is

π e R i − I i for a given effort e ∈ {L, H}.

Contracts, default, **and** equilibrium. Investors compete by offering arbitrary menus of financial

contracts to the entrepreneur. 8 The entrepreneur can simultaneously trade with any subset

of investors, optimally choosing the aggregate investment **and** the corresponding repayment by

combining available contracts. The nature of a financial contract depends on the variables that

we assume to be observable. In most of our analysis, a contract is an individual investment **and**

a function that associates a repayment to each amount of aggregate outside financing. Given the

entrepreneur’s limited liability, the repayment is always set equal to zero if the project fails. More

formally, a contract proposed by investor i is an array C i = (I i , R i (.)), where I i is the investment

he supplies, **and** R i (.) : R + → R is the function specifying the repayment he requires when the

project succeeds, for each I. In Section 4.2, we allow investors to also observe the aggregate financing

raised by the entrepreneur before production takes place **and** to modify their initial offer

(I i , R i (.)) accordingly. The following timing of events describes our competition game:

1. Each investor i offers a menu M i , that is a set of contracts which includes the null one (0, 0). 9

2. Given these offers, the entrepreneur chooses one contract in each menu **and** an effort level.

3. The uncertainty is realized **and** payments are made.

Being protected by limited liability, the entrepreneur can trade contracts which involve conflicting

prescriptions. In these situations, the outside financing I = ∑ I i is such that the aggregate

i

contractual repayment R = ∑ R i (I) > G(I + A), **and** we say that strategic default takes place.

i

8 Investors can therefore include contracts that are not traded at equilibrium. Menus are sufficiently general to

reproduce any bilateral communication between investors **and** the entrepreneur (Martimort **and** Stole (2002), Peters

(2001)).

9 This is done to incorporate agents’ participation decisions in a simple way.

7

The entrepreneur’s payoff can hence be written as

⎧

⎨

( ) +

π H G(I + A) − R − A

if e = H

U(I, R, e) =

⎩

( ) +

(2)

π L G(I + A) − R + B(I + A) − A if e = L,

**and** her reservation utility is U(0) ≡ (π H G − 1)A, which is strictly positive given (1).

Under strategic default, the entrepreneur chooses e = L to obtain her private benefit. Investors

cannot be repaid according to contractual terms, **and** a pro rata rule applies: each investor receives

a share of the cash flow proportional to his investment. 10 The expected profit of investor i is

V i (I i , R i (I), I, R, e) = π e

[

G

(

I + A

)I i

I 1 {R>G(I+A)} + R i (I)1 {R≤G(I+A)}

]

− Ii . (3)

A pure strategy for the entrepreneur is a mapping that associates to each profile of investors’

menus ( M 1 , M 2 , ..., M N

)

a vector of contracts

(

C1 , C 2 , ..., C N

)

**and** an effort choice e ∈ {L, H}.

A pure strategy for investor i is a menu M i . 11 Throughout the paper, we restrict attention to pure

strategy subgame perfect equilibria. Observe that the entrepreneur necessarily selects e = H at

equilibrium: given (1), investors’ aggregate profit is negative if e = L, which implies that at least

one active investor would strictly prefer not to participate.

Feasible **and** second-best allocations. We denote H = {(I, R) ∈ R 2 + : U(I, R, H) ≥

U(I, R, L)} the set of (I, R) pairs inducing e = H as an optimal choice, **and** L = {(I, R) ∈ R 2 + :

U(I, R, H) ≤ U(I, R, L)} those inducing e = L. For a given profile of menus, the entrepreneur’s

choices determine agents’ final allocations, that is, investments **and** actual repayments. If there is

no strategic default, investors are repaid according to contractual terms, in which case we refer to

(I, R) as an aggregate allocation.

We say that an allocation {(I 1 , R 1 (I)), ..., (I N , R N (I)), (I, R)} is feasible if (I, R) ∈ H, **and**

if it guarantees to any agent at least his reservation utility. The corresponding set F of aggregate

10 We discuss in the additional Appendix C the case in which one investor can issue a senior claim.

11 We make no specific assumption on the menus available to investors. We only require that they are compact sets so

that the entrepreneur’s choice problem of maximizing (2) over contracts **and** effort admits a solution for every profile

(

M1 , M 2 , ..., M N

)

. A direct way to ensure this is to assume for instance that each Ii belongs to a closed interval **and**

each R i (.) is a bounded function.

8

R

U(0)

R = G(I + A)

L

R c

R m

O

G − B ∆π

1

G πH

C m

•

I m

F

Ψ

C c

•

I c H

I

Figure 1: Set F of aggregate feasible allocations.

feasible allocations is depicted by the grey area in Figure 1. All such (I, R) allocations lie to the

right of the indifference curve relative to the entrepreneur’s reservation utility U(0), **and** to the

left of the line (O, C c ), which represents the aggregate zero profit condition for investors when

e = H. Condition (1) ensures that F is non-empty: when e = H, the entrepreneur’s marginal

rate of substitution G is greater than the investors’ one 1

π H

. Last, to ensure that e = H is optimal

for the entrepreneur, any aggregate feasible allocation must lie below the line that corresponds to

U(I, R, H) = U(I, R, L), which has a slope G − B . In line with Holmstrom **and** Tirole (1998),

∆π

we also assume that

G − B ∆π < 1

π H

, (4)

where ∆π = π H − π L , which ensures that F is closed. The frontier Ψ ≡ {(I, R) ∈ F :

U(I, R, H) = U(I, R, L)} in Figure 1 represents the set of second-best aggregate allocations:

a social planner who cannot control the entrepreneur’s effort but who can control final trades finds

optimal to select an aggregate allocation on Ψ. Indeed, (1) **and** (4) guarantee that starting from

any (I, R) ∈ Ψ, it is not possible to simultaneously increase the entrepreneur’s payoff **and** the

profit of all investors. Conversely, **and** using the same inequalities, one gets that for any allocation

9

{(I 1 , R 1 (I)), ..., (I N , R N (I)), (I, R)} such that (I, R) /∈ Ψ, there is another feasible allocation

which Pareto-dominates it. Two relevant second-best aggregate allocations are C c = (I c , R c ),

which maximizes the entrepreneur’s payoff, **and** C m = (I m , R m ), which maximizes the investors’

joint profit.

3 Market equilibria with bilateral contracts

We first assume that investors post menus of contracts contingent on the success or failure state,

but not on outside financing or on the final cash flow. This setting constitutes our benchmark

**and** corresponds to the st**and**ard representation of nonexclusive competition under moral hazard.

The credit card industry, in which credit card issuers cannot observe their customers’ portfolio of

credit cards, or the pure OTC markets, in which assets are traded outside of organized exchanges,

are good examples of this environment. A contract is then a pair (I i , R i ) ∈ R 2 +, where I i is an

investment level, **and** R i is the repayment required by investor i in the success state.

3.1 Equilibrium analysis

We derive below necessary conditions for an aggregate allocation (I ∗ , R ∗ ) ∈ F to be supported at

equilibrium.

For a given profile of menus (M 1 , M 2 , ..., M N ), the entrepreneur determines her credit dem**and**

(I, R) by optimally combining investors’ offers. Since for each e ∈ {H, L}, her preferences are

linear in I **and** R, the corresponding choice problem is simple. As an illustration, suppose that

M i = {(I i , R i ); (0, 0)}, with I i > 0. If strategic default does not take place, the entrepreneur is

willing to trade (I i , R i ) if its price R i

I i

is smaller than her corresponding marginal rate of substitution,

irrespective of the other offered menus. 12 If the entrepreneur chooses to default, she selects

the highest investment contract in each menu in order to maximize her private benefit.

The analysis of investors’ behavior is more involved. A crucial feature of nonexclusive competition

is that every investor may gain by offering contracts that the entrepreneur accepts in combination

with other offers. This makes the set of unilateral deviations very large compared to the

12 This idea is formalized in Lemma 2 in the Appendix, which is key to the proof of our results.

10

case in which exclusive relationships are assumed, **and** potentially difficult to characterize. Since

there are gains from trade when the entrepreneur chooses e = H, there is always room for deviations

that increase aggregate investment provided that they do not trigger e = L. The following

proposition builds on this intuition **and** provides a central result of this section.

Proposition 1 If (I ∗ , R ∗ ) is an equilibrium aggregate allocation, then

U(I ∗ , R ∗ , H) =

max ((I i ,R i )∈M i

)

U(I, R, L). (5)

i=1,...,N

Since the right-h**and** side of (5) cannot be greater than the equilibrium payoff U(I ∗ , R ∗ , H),

Proposition 1 is straightforward if (I ∗ , R ∗ ) ∈ Ψ. If (I ∗ , R ∗ ) ∉ Ψ, the intuition for the proposition

can easily be understood in a free entry equilibrium. 13 Suppose that an entrant deviates by offering,

together with the null contract (0, 0), the additional contract ( x, ( 1 + ε)x ) , of price 1 + ε, with

π H π H

x **and** ε strictly positive. Choose ε small enough to guarantee that the price of this contract is

strictly less than the entrepreneur’s marginal rate of substitution G when e = H. Clearly, she picks

this contract at the deviation stage. For (I ∗ , R ∗ ) to be supported at equilibrium, this choice must

induce e = L. Given that both x **and** ε can be arbitrarily small, investors’ equilibrium menus must

be such that (5) holds. In this nonexclusive context, (5) can be interpreted as an aggregate incentive

compatibility constraint. The allocations satisfying (5) are such that the threat of overborrowing

prevents any investor from profitably complementing existing offers. The welfare implications of

aggregate incentive compatibility are derived in the next corollary.

Corollary 1 If (I ∗ , R ∗ ) is an equilibrium aggregate allocation, then (I ∗ , R ∗ ) ∈ Ψ, i.e. equilibrium

allocations are second-best efficient.

Corollary 1 exploits a direct implication of Proposition 1: each contract traded at equilibrium,

combined with her preferred options in all other menus, must yield the entrepreneur the same

payoff whether she selects e = H or e = L. If this individual incentive compatibility condition

is violated, a single investor can straightforwardly deviate, even if his rivals post arbitrary menus.

The proof of Corollary 1 shows that individual incentive contraints are not compatible with the

aggregate condition (5) if (I ∗ , R ∗ ) does not belong to the frontier Ψ.

13 The proof of Proposition 1 is established for any fixed number of investors.

11

Therefore, with bilateral contracts, investors successfully exploit the opportunity to complement

their rivals’ offers, exhausting thereby all gains from trade. This efficiency result sheds light

on the welfare properties of nonexclusive markets under moral hazard. Indeed, while Kahn **and**

Mookherjee (1998) **and** Bisin **and** Guaitoli (2004) obtain similar findings, 14 they only provide partial

characterizations of market equilibria. In contrast, our result is robust in that it does not rely

on exogenous restrictions on investors’ menus or on the structure of equilibrium strategies.

3.2 Equilibrium aggregate allocations

We now provide a full characterization of equilibrium aggregate allocations in terms of the private

benefit B. To this end, we say that moral hazard is mild if B ≤ π H G − 1, **and** that it is strong

if B > π H G − 1. If moral hazard is mild, the entrepreneur’s marginal private benefit is lower

than the marginal return of the project with e = H. The severity of moral hazard affects the set of

profitable deviations **and** the resulting equilibrium outcomes, as formally stated below.

Proposition 2 The following holds:

1. If B ≤ π H G − 1, any aggregate allocation (I ∗ , R ∗ ) ∈ Ψ can be supported at equilibrium.

2. If B > π H G − 1, (I m , R m ) is the unique equilibrium aggregate allocation.

Irrespective of the severity of moral hazard, there is always an equilibrium in which investors

earn a monopolistic profit. Suppose that one investor, say investor 1, offers M 1 = {C m =

(I m , R m ), (0, 0)}, **and** all other investors i ≠ 1 propose M i = {(0, 0)}. The proof of Proposition

2 shows that no investor can profitably deviate by inducing the entrepreneur to select e = H.

This is illustrated in Figure 2. Consider any investor i ≠ 1: a deviation C i ′ = (I i, ′ R i) ′ is profitable

only if R′ i

> 1 . However, following any such deviation, it is optimal for the entrepreneur to

I i

′ π H

trade C i ′ together with C m **and** to select e = L, achieving U(C i ′ + C m , L) > U(C i, ′ H). Therefore,

the opportunity to trade the monopolistic allocation together with a deviating contract makes the

low effort’s threat fully credible.

14 Specifically, Proposition 5 in Kahn **and** Mookherjee (1998) **and** Proposition 4 in Bisin **and** Guaitoli (2004) provide

the following constrained efficiency result: even when equilibria fail to be second-best efficient, they are typically

efficient from the viewpoint of a planner who cannot control final trades (third-best efficiency). In our setting, all

equilibrium aggregate allocations belong to the second-best frontier Ψ, which therefore coincides with the third-best

one.

12

R

U(0) U(C i, ′ e(C i, ′ )) U(C i ′ + C m , e(C i ′ + C m ))

R = G(I + A)

L

C ′ i + C m •

Ψ

C m •

• C ′ i

H

O

I ′ i + I m

I

Figure 2: The monopolistic allocation equilibrium.

Proposition 2 also states that, if moral hazard is strong, the monopolistic allocation is the only

one supported at equilibrium. The result is established by contradiction. Consider any aggregate

allocation on the frontier Ψ, **and** any investor who provides funding at equilibrium. Linearity of

the entrepreneur’s preferences guarantees that this investor is indispensable. This implies that if

he deviates by offering a higher-price, smaller-investment contract, the entrepreneur’s payoff is

reduced (irrespective of her effort choice). When B is high, her private benefit sharply decreases

**and** she chooses e = H which makes the deviation profitable. Monopoly then arises as the only

equilibrium allocation.

When B is small, **and** moral hazard is mild, any aggregate allocation on Ψ can be sustained at

equilibrium. The reason is that the above deviations induce e = L as the private benefit decreases

less than the entrepreneur’s payoff under e = H. In this case, only one investor is active **and** there

is at least another investor offering a positive investment contract. This contract is latent: despite

not being traded it needs to be issued for the equilibrium to exist. The latent contract serves the role

of a threat, that is, it ensures that the entrepreneur selects e = L following any investor’s deviation.

Our benchmark incorporates the main insights of the literature on nonexclusive competition

13

under moral hazard. As in the credit economy of Parlour **and** Rajan (2001), monopoly is sustained

at equilibrium. As in the insurance settings of Kahn **and** Mookherjee (1998), Bisin **and** Guaitoli

(2004) or Attar **and** Chassagnon (2009), latent contracts are used to sustain credit rationing **and**

positive profit at equilibrium. 15 At the same time, we contribute to this literature by providing a

complete equilibrium characterization **and** a full strategic analysis of a financial market in which

no information is available on aggregate trades. 16

4 Market equilibria with covenants

Corporate financing relationships are not well-described by bilateral contracts. Most financial

contracts include covenants aimed at monitoring firms’ financial decisions by specifying target

debt ratios. In case of violations, creditors can accelerate the payment of their outst**and**ing debt.

Empirical evidence suggests that they can also increase interest rates or reduce future lending. 17

Building on these insights, we allow financial contracts to be contingent on outside financing.

To provide a thorough analysis of the strategic role of financial covenants, we distinguish two informational

structures. First, we assume that outside financing is observed only after production takes

place. This enables investors to require higher interest rates when undesired levels of debt are observed.

We next allow investors to also observe the funds raised before they are invested. Investors

can exploit this information to accelerate the repayment of their loan, or to modify their initial

investment. We therefore consider the most favorable setting to enforce exclusive contracting: investors

observe aggregate trades, **and** can use a very large set of punishments to deter departures

from their desired financing policy.

15 A traditional criticism of latent contracts is that, to be attractive for the entrepreneur, they might be issued at a

loss-making price for investors. In our context, credit rationing **and** positive profits are either supported by such latent

contracts, or by the active C m contract which is sufficient to sustain the monopolistic allocation equilibrium.

16 Parlour **and** Rajan (2001) **and** Kahn **and** Mookherjee (1998) restrict investors to post single offers, while Bisin

**and** Guaitoli (2004) also consider piecewise linear menus. Allowing investors to compete over arbitrary menus may

sustain additional equilibria, as it is the case in the complete information setting of Martimort **and** Stole (2003) or in

the adverse selection one of Attar et al. (2011).

17 Roberts **and** Sufi (2009) report that when creditors respond to covenant violations, they increase interest rates

on their loans in about half of the cases **and** that covenant violators exhibit a decline in their net debt issuing activity.

Relatedly, Chava **and** Roberts (2008) find that covenant violations induce a 13% to 20% decline in the level of

investment.

14

4.1 Covenant violation after investment

We explore here the situation in which outside funding is only observed ex post. Enlarging investors’

contracting opportunities to include such debt-based covenants introduces new strategic

interactions. On the one h**and**, they have access to an additional class of deviations. Any investor

can set punishments to circumvent the entrepreneur’s threat to overborrow. In particular, they can

stipulate a raise in interest rates to eliminate the entrepreneur’s rent when she chooses e = L **and**

repays all investors. To illustrate this procompetitive effect, consider the monopolistic equilibrium

arising with bilateral contracts. With covenants, the simple offer (I m , R m ) is not robust to entry.

Any inactive investor can undercut it by proposing a small investment at a unit price slightly above

1

π H

, writing covenants that trigger default if additional funding is raised. By construction, the

entrepreneur is better off trading this deviating contract **and** choosing e = H, which makes the

deviation profitable.

On the other h**and**, any investor can use covenants to prevent his competitors from proposing

contracts to be traded in addition to his own. Under bilateral contracting, such deviations are key to

establish Corollary 1 **and** to show that all gains from trade are exploited at equilibrium. **Covenants**

could then undermine the constrained efficiency result. To isolate this latter anticompetitive effect,

we first present a stripped-down version of our model in which π L = 0, that is, strategic default

always occurs when e = L is chosen. This case corresponds to the st**and**ard framework of Parlour

**and** Rajan (2001) which is extended to allow investors to use covenants contingent on outside

financing.

Proposition 3 If π L = 0, any aggregate allocation (I ∗ , R ∗ ) ∈ F such that I ∗ ≥ I m is sustained

at equilibrium.

Proposition 3 states that the use of covenants leads to an indeterminacy of equilibrium allocations.

To provide an intuition for this result it is useful to describe equilibrium strategies. Every

investor proposes the same menu, which includes two non degenerate contracts on top of the null

one. The entrepreneur chooses the same contract in each menu, i.e. every investor is active. Equilibrium

contracts incorporate the following financial covenants. If the entrepreneur raises a large

amount of debt from other investors, covenants impose large penalties which can be interpreted as

an increase in interest rate. Given these covenants, the entrepreneur strategically defaults whenever

15

an investor tries to gain by complementing existing offers. Symmetrically, if the entrepreneur does

not raise enough debt, covenants specify a reduction in the interest rate. This lower interest rate

is designed to guarantee that no investor is indispensable at equilibrium. This in turn blocks any

deviation of an investor who wishes to increase the price of his loan.

Equilibrium menus also include an additional high price, large investment contract. Although

investors anticipate that it will not be traded at equilibrium, this second contract is used as a threat

against the deviations of any investor trying to exp**and** his market share. The contract is therefore

strategically issued by active investors to protect their rents **and** to induce collusive outcomes at

equilibrium. 18

To clarify how inefficient allocations can be sustained at equilibrium, consider any (I ∗ , R ∗ ) ∉

Ψ. In contrast with bilateral contracts, the covenants issued by his competitors prevent a deviating

investor from gaining by complementing existing offers. He must therefore endeavor to replace

some (or all) of his competitors’ offers. Such deviations entail proposing a substantial amount

of investment. Since moral hazard is strong when π L = 0, 19 the entrepreneur’s private benefit

from investing this amount together with the large loans issued by the others is greater than the

payoff of trading with the deviator, provided that I ∗ ≥ I m . This shows that any (I ∗ , R ∗ ) such

that I ∗ ≥ I m is supported at equilibrium: equilibrium allocations are therefore Pareto-ranked.

Although all investors are active at equilibrium **and** each of them typically earns a strictly positive

profit, the corresponding outcome is entry-proof: given equilibrium covenants, no potential entrant

can do better than offering the null contract. 20

If π L > 0, a deviating investor can fully exploit the procompetitive power of covenants to

undercut competing offers. The following proposition shows that indeterminacy **and** inefficiency

of equilibrium allocations still arise for any value of π L .

Proposition 4 There exists a threshold N **and** a sequence of payoffs (U N ) N>N

with U N < U(I c , R c , H)

18 The entrepreneur is indifferent between trading equilibrium contracts **and** defaulting on these large investment

contracts. In the latter case, investors lose money in the aggregate, which leads to question their rationality to post

such contracts in the first place. The fact that all investors are active **and** make a positive profit at equilibrium ensures

that issuing them can be optimal against the entrepreneur’s trembling behaviors.

19 If π L = 0, then, given (4), we get B > π H G − 1.

20 Our equilibria can therefore exhibit an arbitrary number of inactive investors: the proof of Proposition 3 suggests

that equilibrium allocations may alternatively be supported with two active investors who equally share the market **and**

N − 2 inactive ones who only offer the null contract (0, 0).

16

such that any aggregate allocation (I ∗ , R ∗ ) ∈ F satisfying

U(I ∗ , R ∗ , H) ≥ U N (6)

is sustained at equilibrium. When B > π H G − 1, lim N→∞ U N = U(0).

From (6), the set of equilibrium allocations is non-empty whenever N > N. These allocations

are characterized in terms of a threshold payoff U N , which role can be understood as follows.

For each N > N, any inefficient equilibrium is in principle vulnerable to the undercutting of an

investor who offers a contract including covenants designed to enforce exclusivity. When moral

hazard is strong, as suggested in the discussion of Proposition 3, these offers are more likely to

induce default as the amount of funds offered by the deviator gets larger, **and** can easily be circumvented.

When moral hazard is mild, the opposite holds **and** deviations that entail a sufficiently

large investment can in principle induce e = H. The threshold U N is constructed to limit the

maximal amount of funds that a deviator can propose to exploit this opportunity, thereby deterring

such deviations.

A striking result of Proposition 4 is that, when moral hazard is strong, U N tends to U(0) as N

goes to infinity, that is, the entrepreneur can achieve any payoff above U(0). This st**and**s in sharp

contrast with st**and**ard insights that oligopolistic rents are dissipated in the limit. The reason is that

when N is large, each investor’s contract has a negligible impact on the entrepreneur’s threat of

default. In this case, the threat of default is sufficient to prevent even those deviations that provide

a payoff to the entrepreneur slightly above U(0). We then get a Folk Theorem-like result: every

feasible allocation can be sustained at equilibrium. Therefore in the limit the level of B determines

the set of equilibrium aggregate allocations, as in the bilateral contracting setting of Section 3.

In contrast with that of Proposition 3, the proof of Proposition 4 relies on asymmetric equilibrium

strategies. A subset of investors are active **and** equally share the market by offering contracts

with covenants similar to those exhibited in the case π L = 0. The remaining investors post latent

contracts in order to render the threat of default effective even when moral hazard is mild. 21 The

investment available in each latent contract is decreasing with the total number of investors. As

21 Observe that, since Proposition 4 also holds for π L = 0, Propositions 3 **and** 4 together imply that some aggregate

allocations can be sustained with different equilibrium strategies. Therefore indeterminacy also prevails regarding

investors’ equilibrium menus.

17

N goes to infinity, the investment included in each latent contract tends to zero, **and** no investor is

indispensable to provide the threat of default. The aggregate available funds can then be seen as a

barrier to entry **and** the equilibrium interpreted as entry-proof.

The minimal number of investors N is finite, but it can in principle be large. We provide

however in the additional Appendix B an alternative profile of investors’ strategies guaranteeing

that indeterminacy arises with N = 2 for the general case π L ≥ 0 if the project’s profitability is

sufficiently high.

4.2 Covenant violation before investment

So far, our modeling only allows covenants to specify contractual penalties that can be enforced after

investment is undertaken. In practice, covenant violations also entitle investors to accelerate the

repayment of their loan or to renegotiate credit facilities. We extend below our setting to introduce

the possibility for investors to observe all funds raised by the entrepreneur before investment takes

place. Investors can then dem**and** the early repayment of all or part of their loan or alternatively

provide additional liquidity. We interpret these features as loan acceleration clauses **and** contingent

lines of credit, respectively.

To fix ideas, suppose that before investment takes place, any investor i can observe the aggregate

initial financing, denoted I 0 , raised by the entrepreneur. He can therefore commit to modify

his initial arrangement according to I 0 . In particular, he can withdraw any fraction of his initial

offer in order to protect himself against the threat of overborrowing. 22 This could in principle

foster competition. We show below that these procompetitive effects of loan acceleration clauses

are undermined by the simultaneous issuance of lines of credit. As a consequence, equilibrium

allocations remain indeterminate.

More formally, define a contract for investor i as ( I i , R i (.), I + i (.), R+ i (.)) with I i ≥ 0, I + i (I 0) ≥

−I i **and** R i (.) **and** R + i (.) functions of the aggregate initial debt I 0 **and** of the final outside financing

I. In this setting, a loan acceleration clause following an undesired level of debt I 0 is written:

I + i (I 0) = −I i **and** R + i (I 0, I) = −R i (I 0 , I) ∀I. Similarly, a line of credit can be written: I + i (I 0) >

0, **and** R + i (I 0, I) = pI + i (I 0), with I + i (I 0) being the credit limit available at a unit price p. The

22 This is indeed the most common rationale for the use of loan acceleration clauses in debt contracts.

18

following proposition summarizes our results.

Proposition 5 Propositions 3 **and** 4 extend to the case in which investors can write covenants

contingent on the initial debt I 0 .

When the initial debt is observable, an investor can try to enforce exclusivity by asking to

be reimbursed if he observes additional debt issuance. Such a deviation is however blocked by

investors’ ability to offer lines of credit contingent on the initial debt. At equilibrium, investors do

not need to ask for an acceleration of their loan, **and** the entrepreneur does not draw on any line of

credit. Yet the presence of this additional source of liquidity undermines the threat of asking for

an early loan repayment. In general terms, Proposition 5 shows that the joint issuance of ex post

**and** ex ante debt-based covenants constitutes a barrier against any attempt of investors to enter the

market using sophisticated contracts. 23

It is interesting to discuss to what extent these financial covenants are consistent with some

empirical observations on firms’ credit relationships. To start with, the equilibrium strategies of

Proposition 5 are such that investors do not ask for any acceleration of their loans. They do however

specify changes in interest rates following the violation of financial covenants resulting from firms’

multiple borrowing. This is in line with the observation that creditors do not ask for the early

repayment of their loan following a covenant violation but rather use this right to modify other

features of the firm’s financial policy, like the interest rate (Roberts **and** Sufi (2009)).

Next, at equilibrium, lines of credit are issued, **and** are not drawn upon, unless the borrower

modifies her investment policy. At the same time, investment is typically inefficiently low, **and**

firms are credit-rationed. Relatedly, Sufi (2009) points out that lines of credit are not fully committed

**and** that their access is contingent on firms’ financial performance. This illustrates the idea

that lines of credit are not a perfect substitute for liquidity, **and** are strategically offered by banks,

as also emphasized by Acharya et al. (2014).

23 The distinction between raised **and** invested funds creates a possibility for the entrepreneur to divert funds, which

can constitute an additional source of moral hazard. To illustrate its impact, suppose that the entrepreneur ( can divert )

a fraction

(

γ ∈ (0, 1) of the

)

investment

( )

I + A. The relevant incentive constraint is then π H G(I + A) − R ≥

π L G(1 − γ)(I + A) − R + γ + B (I + A), since diverting is optimal when e = L, but not when e = H. The

constraint is binding whenever R = ( G − B + γ(1 − π LG) )

(I + A). If γ is not too large, conditions (1) **and** (4) are

∆π

satisfied, **and** the possibility of fund diversion is hence equivalent to an increase in B, which reinforces our results.

19

More fundamentally, in our analysis, the strategic role of covenants is incorporated in investors’

out-of-equilibrium offers. The entire equilibrium capital supply therefore matters to determine

firms’ investment policy. The recent paper of Degryse et al. (2015) is the first to investigate the

effect of multiple lending on the supply of credit of a given bank. Their main result is that the

bank’s willingness to lend decreases when its customers contract loans with other lenders, unless

the bank loan is fully secured. Our work, which explicitly derives investors’ equilibrium best

responses, provides further insights. Both the willingness to lend **and** the contracts issued by a

bank depend on those covenants included in its competitors’ loans, which should therefore be

taken into account to identify banks’ credit supply. When competitors do not use covenants, a

bank can issue covenants that reduce the threat of dilution, leaving its capital supply unaffected.

Facing such covenants, a bank might be unable to secure its loan, which induces a reduction in

investment.

5 Market design under the threat of default

The protection offered by covenants turns out to be a double-edged sword. It encourages investors

to offer more competitive contracts, but it also prevents entry of potential competitors. The latter

effect dominates **and** a large number of low investment, inefficient allocations are sustained at

equilibrium. On the normative side, this calls for a proper design of the market architecture.

At the root of market inefficiency is the investors’ individual inability to prevent overborrowing

**and** strategic default. It is therefore natural to ask whether this threat can be mitigated by appropriately

designed bankruptcy rules. We show however in the additional Appendices C **and** D that

our results survive the introduction of st**and**ard procedures. Precisely, Proposition C.1 analyzes

the case in which one investor has repayment priority over the others, **and** Proposition D.1 that in

which part of the entrepreneur’s private benefit is seized under bankruptcy. In both cases, a large

number of inefficient allocations are sustained at equilibrium.

We therefore consider an alternative way to circumvent overborrowing based on a market

mechanism in which investors can contribute to the provision of a subsidy to the entrepreneur.

This mechanism induces the entrepreneur not to default **and** sustains the competitive allocation as

the unique equilibrium one. We introduce in the next subsection a subsidy mechanism **and** discuss

20

its empirical relevance in Subsection 5.2.

5.1 The subsidy game

Consider a subsidy scheme, conceived, say, by a market designer, such that for every observed

outside financing I, investor i is asked to pay a transfer T i (I), with T i (0) = 0. After the cash

flow realization, the entrepreneur receives a (possibly null) subsidy equal to the sum of investors’

transfers. This mechanism induces the following game:

1. Given (T 1 (.), T 2 (.), ..., T N (.)), each investor i offers a menu M i of contracts that can include

any class of financial covenants. 24

2. After observing all offers, each investor decides whether to pay his transfer or not.

3. The entrepreneur chooses a contract in each menu **and** makes her effort choice.

4. The uncertainty is realized, **and** payoffs are distributed according to financial contracts **and**

to the subsidy scheme.

We focus on the case π L = 0 which exacerbates the anticompetitive effect of covenants, **and** we

consider pure strategy subgame perfect symmetric equilibria. The following proposition highlights

the role of the subsidy in disciplining investors **and** resolving indeterminacy.

Proposition 6 If π L = 0, there is a profile of transfers (T 1 (.), T 2 (.), ..., T N (.)) such that (I c , R c ) is

the unique aggregate allocation supported in a symmetric equilibrium of the subsidy game.

The intuition for Proposition 6 is the following. Suppose that an allocation different from

the competitive one is supported at equilibrium. In the absence of a subsidy scheme, we know

from the previous section that the entrepreneur’s threat to default hinders deviations by an investor

who tries to exploit gains from trade. We show that there exist individual transfers that guarantee

a profitable deviation for a well-chosen investor. The design of the subsidy scheme involves the

following trade-off. On the one h**and**, the subsidy provided to the entrepreneur has to be sufficiently

large to alleviate the threat of default **and** induce e = H at the deviation stage. On the other

24 That is, the analysis applies to both the informational structures of Subsections 4.1 **and** 4.2.

21

h**and**, investors must be willing to pay their individual transfers. To ensure their participation,

the mechanism is such that the entrepreneur’s subsidy, that is, the sum of individual transfers, is

strictly positive only if all investors accept to pay. This implies that the entrepreneur defaults in

the absence of the subsidy. Also, given that investors observe all offers when deciding whether

to pay or not, they can anticipate the entrepreneur’s behavior. Each investor prefers to contribute

whenever the cost to induce e = H is smaller than his loss following default.

We construct

a subsidy scheme satisfying both requirements. Given this mechanism, no aggregate allocation

different from (I c , R c ) can be supported in a symmetric equilibrium. The corresponding transfers

are simple: investors’ payments are linear **and** increasing in the entrepreneur’s investment unless

I = I c , in which case they are null. 25

allocation.

This leaves (I c , R c ) as the unique equilibrium aggregate

To implement the above mechanism, the market designer only needs to observe investors’ participation

decisions on top of the realized investment I. The subsidy scheme therefore relies on

similar instruments compared to those available to investors. 26 Also, in contrast with st**and**ard taxation

schemes, the mechanism cannot exclude non-participating agents from entering the market:

it relies on a voluntary contribution of investors. The question therefore arises of whether the same

subsidy scheme can be otherwise enforced by competing investors who individually design their

transfer schedules. Intuition suggests that this is not the case. Allowing investor i to pay a transfer

according to I is equivalent to letting him free to set a possibly negative repayment R i (.). Since

we already allow repayments to take negative values in the general model of Section 2, the equilibrium

characterization of Proposition 3 would not be affected. We therefore stress the need for a

centralized entity with the power to commit to collect **and** redistribute resources according to the

volume of trades.

25 To ensure equilibrium uniqueness in the investors’ participation subgame, we design transfers to guarantee that

participating is a dominant strategy. Precisely, each contributor is granted a reward if he participates **and** at least one

of the others does not.

26 Indeed, financial covenants contingent on realized investment allow de facto to control the entrepreneur’s participation

decision with every investor.

22

5.2 Guarantee funds as subsidy mechanisms

The history of financial regulation offers many examples of privately financed mechanisms that

perform functions similar to those of our subsidy scheme. In modern futures markets, central

clearinghouses set up **and** manage guarantee funds to provide insurance against counterparty default

(Kroszner (1999)). 27

Credit markets also offer different types of rescue mechanisms supported

by banks. During the Comptoir d’Escompte crisis in 1889 in Paris, the Banque de France

granted a loan to the insolvent Comptoir d’Escompte, asking other banks to provide a guarantee

to absorb losses on this loan (Hautcoeur et al. (2014)). A similar role was performed by the New

York Clearing House Association (NYCHA) before the creation of the Federal Reserve System in

1913. During a crisis, the NYCHA could allow its members to issue loan certificates, the payment

of which was guaranteed by the clearinghouse itself, that is, by the whole banking industry (Gorton

(1985)). More recently, the emerging European banking union has decided the creation of a

Single Resolution Fund that will be financed by European banks **and** that will help to restructure

distressed banks. 28 Our analysis offers a rationale for market-based regulatory schemes that make

banks or financial intermediaries liable for the negative externalities stemming from their strategic

interactions. In the subsidy game, the threat to provide a subsidy has a disciplining effect on credit

markets, enhancing competition **and** decreasing rents. While it is hard to assess to what extent

actual guarantee mechanisms cope with counterparty externalities, some authors argue that they

do have a disciplinary role. In his analysis of the 1987 financial markets crash, Bernanke (1990)

points at the effectiveness of the clearinghouse institution as an insurance company. Riva **and**

White (2011) find evidence that during the nineteenth century, the number of brokers’ defaults at

the Paris Bourse decreases as the size of the guarantee fund relative to the volume of transactions

increases. Whether these insurance systems also affect the volume of trades **and** the size of the

market remains an open question.

Our mechanism shares important features with the above institutions. In particular, the subsidy

27 OTC derivative markets can also organize some form of mutualization of losses across traders upon default of one

party. For instance, the stock brokers on the forward market at the Paris Bourse in the nineteenth century organized a

Common Fund to rescue defaulting brokers (Riva **and** White (2011)).

28 Vice-President Michel Barnier, responsible for Internal Market **and** Services recently said: "To respond to the

financial crisis, we have worked hard to improve the financial system so that banks pay for themselves if they have

problems, **and** not the taxpayers. The detailed rules on resolution funds financed by the banking sector, adopted today,

are an important step to making that a reality." European Commission Press Release, October 21, 2014.

23

scheme we suggest does not have the power to shut markets down: investors can refuse to subscribe

to the system of transfers, **and** still offer financial contracts. Relatedly, the European banking

union lets non-Euro-area banks free to participate to the Single Resolution Mechanism. Also, the

financial contributions of banks increase in the volume of the available investment. In practice,

the guarantee funds of clearinghouses are financed by members’ contributions that depend on the

volume of trades they generate, **and** therefore on their risk exposure. 29 An important difference

with actual mechanisms is that our subsidy scheme does not involve a direct transfer from some

banks to others. It is therefore immune from the criticism that such bailouts create moral hazard

**and** induce the insured agents (banks or traders) to take excessive risk. Indeed, subsidizing the

entrepreneur when she invests too little does not provide incentives for banks to lend excessively.

6 Conclusion

This paper shows that investors’ competition over financial contracts incorporating exclusivity

clauses does not necessarily lead to efficient outcomes in the presence of side trading. That is,

letting investors free to design financial covenants may exacerbate the negative effects of the counterparty

externalities. Our results suggest new ways to empirically identify investors’ behaviors

**and** the associated market outcomes. In particular, to measure the impact of side trading externalities

in capital markets, one should look into the entire profile of investors’ offers. Observing

financial contracts that, although not traded, discourage the entry of rivals, along with covenants

contingent on outside financing would provide (indirect) evidence of such externalities.

On the theoretical side, our analysis calls for further research in two directions. The first one

relates to the degree of observability of outside financing. We focus on the two extreme cases of

either fully unobservable or fully observable outside debt. However, it might be easier for investors

to observe cash flows rather than total debt. For instance, some liabilities may be hidden in offbalance

sheet items. This raises the question of whether cash flow-based covenants introduce new

strategic effects. Intuitively, it could be difficult for an investor to detect multiple trades if the

29 These are recurrent features in the history of common funds: at the Paris Bourse in the nineteenth century, most

of the Common Fund’s revenue came from a stamp tax on the paper used by brokers for their operations (White

(2007)). In the US, the Board of Trade Clearing Corporation created in 1925 built its reserve fund from the clearing

fees charged to its members (Kroszner (1999)).

24

observed cash flow only provides an imperfect signal of outside debt, which might undermine the

anticompetitive role of covenants. 30 More generally, since accounting principles determine the

informativeness of financial variables, this analysis could provide novel insights on the economic

implications of accounting st**and**ards.

The other interesting extension would be to explicitly consider repeated lending relationships

in the presence of strategic default. In such contexts, covenants can threaten to reduce borrowers’

future financing capacity. Reinforcing this coercitive role over a long horizon could thereby enhance

their procompetitive effect. At the same time, the opportunities to side-trade extend to future

periods, which could generate a novel source of market failure. Exploring this tradeoff is relevant

to our underst**and**ing of the relationship between debt constraints **and** the performance of dynamic

economies.

30 This effect cannot be captured in our model, in which both variables are informationally equivalent.

25

Appendix

We first state Lemma 1 which is used in most of our proofs.

Lemma 1 The following holds:

1. Let (M 1 , M 2 , ..., M N ) be a profile of arbitrary menus. Any entrepreneur’s best response

leading to strategic default induces a strictly negative profit to each investor i = 1, 2, ..., N.

2. Let (M 1 , M 2 , ..., M N ) be a profile of menus of bilateral contracts. Any entrepreneur’s best

PROOF

response leading to e = L induces a strictly negative profit to each investor i = 1, 2, ..., N.

1. Suppose that (M 1 , M 2 , ..., M N ) is a profile of arbitrary menus, **and** let (I i , R i (.)) be the

contract chosen by the entrepreneur in M i in a given best response. If the entrepreneur strategically

defaults, the profit to investor i is V i = π L G( ∑ I i + A) I i

(

∑ − I i = I i πL G − 1 + π LGA) ∑ . It is

i I i I i

immediate to check that V i has the same sign as V = ∑ V i = I ( π L G − 1 ) + π L GA. Observe also

i

that since the entrepreneur finds optimal to strategically default, then B(I + A) ≥ π H GA > A,

otherwise she would have not invested. We therefore have: V < I ( π L G − 1 ) + π L GA + B(I +

A) − A = (B + π L G − 1)(I + A) < 0. It follows that V i < 0 for each i = 1, 2, ..., N.

2. Suppose next that (M 1 , M 2 , ..., M N ) are menus of bilateral contracts. Let (I i , R i ) i=1,2,...,N be

the array of entrepreneur’s optimal choices in each menu M i in a given best response. By assumption,

(I, R) = (∑ I i , ∑ )

R i ∈ L, which implies that U(I, R, L) ≥ U(I − Ii , R − R i , L) for each

i i

i = 1, 2, ..., N. To complete the proof, consider the case in which the entrepreneur’s best response

entails no strategic default. In this case, the former inequality is equivalent to R i

I i

i

i

≤ B π L

+ G. The

profit to each investor i = 1, 2, ..., N is V i = π L R i −I i ≤ I i (π L ( B π L

+G)−1) = I i (π L G+B−1) <

0, where the last inequality follows from I i ≥ 0, **and** from (1).

The following lemma characterizes the entrepreneur’s optimal choices in the bilateral contracting

setting of Section 3.

Lemma 2 Consider the bilateral contracting game of Section 3 **and** assume that the entrepreneur

chooses e = H. Then, for each given menu M i one has:

26

1. If choosing (0, 0) in M i is optimal for the entrepreneur, then each (I i , R i ) ∈ M i \(0, 0) is

such that p i ≥ G.

2. If each (I i , R i ) ∈ M i \(0, 0) is such that p i > G, then (0, 0) is the unique optimal choice in

M i for the entrepreneur.

PROOF: Consider a given menu M i , **and** let p i = R i

I i

be the price of every contract (I i , R i ) ∈

M i \(0, 0). The entrepreneur’s marginal rate of substitution is equal to G when e = H.

1. Assume that (0, 0) is an entrepreneur’s optimal choice in M i . If there is a contract (I i , R i ) ∈

M i such that p i = R i

I i

U( ∑ j≠i

I j + I i , ∑ j≠i

< G, then

R j + R i , H) = π H

(

G(

∑

R j + R i ) ) +

− A

I j + I i + A) − ( ∑

j≠i j≠i

= U ( ∑

R j , H ) + π H (GI i − R i ) > U ( ∑

j≠i

I j , ∑ j≠i

j≠i

I j , ∑ j≠i

R j , H ) ,

for each ( ∑ I j , ∑ j ). This contradicts the assumption that (0, 0) is optimal in M i .

j≠i j≠iR

2. Assume that each contract in M i \(0, 0) is such that p i > G, **and** let (I i , R i ) ≠ (0, 0) be an

optimal choice in M i . Then

U( ∑ j≠i

I j + I i , ∑ j≠i

R j + R i , H) = π H

(

G(

∑

R j + R i ) ) +

− A

I j + I i + A) − ( ∑

j≠i j≠i

= U ( ∑

R j , H ) + π H (GI i − R i ) < U ( ∑

j≠i

I j , ∑ j≠i

j≠i

I j , ∑ j≠i

R j , H ) ,

for each ( ∑ I j , ∑ j ). This contradicts the assumption that (I i , R i ) is optimal in M i .

j≠i j≠iR

PROOF OF PROPOSITION 1 : We introduce the indirect utility functions

{

Ū −i (I i , R i , e) ≡ max U ( I i + ∑ I j , R i + ∑ R j , e ) }

: (I j , R j ) ∈ M j for all j ≠ i ,

j≠i

j≠i

for i = 1, 2, ..., N. For a fixed profile of menus (M 1 , M 2 , ..., M N ) **and** for a given effort e, the

function Ū−i(I i , R i , e) denotes the maximum payoff of the entrepreneur when she trades (I i , R i )

with investor i **and** optimally chooses one contract in each of the menus offered by his rivals. Thus,

27

if (I ∗ , R ∗ ) is an equilibrium allocation, we have U(I ∗ , R ∗ , H) = Ū−i(I ∗ i , R ∗ i , H) for i = 1, 2, ..., N.

Observe that each Ū−i function is continuous in (I i , R i ) by Berge’s maximum theorem.

We now develop the proof, letting (M ∗ 1 , M ∗ 2 , ..., M ∗ N ) be the equilibrium menus. If (I∗ , R ∗ ) ∈

Ψ, Proposition 1 is straightforward. Consider the case (I ∗ , R ∗ ) ∉ Ψ. If (I ∗ , R ∗ ) ≠ (0, 0) there

exists at least one investor i such that the entrepreneur chooses a contract (I ∗ i , R ∗ i ) ≠ (0, 0) in

M ∗ i . It follows from Lemma 2 that R∗ i

I ∗ i

≤ G. Assume by contradiction that U(I ∗ , R ∗ , H) >

max U(I, R, L). Then, since max U(I, R, L) ≥ Ū−i(Ii ∗ , Ri ∗ , L), we get

((I i ,R i )∈M i ) i=1,...,N ((I i ,R i )∈M i ) i=1,...,N

U(I ∗ , R ∗ , H) = Ū−i(I ∗ i , R ∗ i , H) > Ū−i(I ∗ i , R ∗ i , L). (7)

Suppose now that investor i deviates to the menu Mi ε = {(0, 0), (Ii ∗ + ε, Ri ∗ + ε + ε 2 )} with

π H

Ri ∗ + ε + ε 2

ε

π

ε > 0. Observe that

H

Ii ∗ + ε < max { Ri

∗ + ε 2

π

, H

}

≤ G for ε small. It follows from

Ii

∗ ε

Lemma 2 that the entrepreneur chooses (Ii ∗ + ε, Ri ∗ + ε + ε 2 ) in Mi ε . Moreover, given (7) **and** the

π H

continuity of Ū−i, we have Ū−i(Ii ∗ + ε, Ri ∗ + ε + ε 2 , H) >

π Ū−i(Ii ∗ + ε, Ri ∗ + ε + ε 2 , L) which

H π H

guarantees that the entrepreneur chooses e = H at the deviation stage. Since π H (Ri ∗ + ε + ε 2 ) −

π H

Ii ∗ − ε > π H Ri ∗ − Ii ∗ the deviation is profitable, which contradicts the assumption that (I ∗ , R ∗ ) is

an equilibrium aggregate allocation. If (I ∗ , R ∗ ) = (0, 0), the same reasoning applies: letting any

investor i deviate to Mi

ε ε

= {(0, 0), (ε, + ε 2 )}, it is immediate to check that e = H is chosen

π H

**and** that the deviation is profitable. In particular, the proof establishes the stronger result

∀i = 1, 2, . . . , N : Ū −i (I ∗ i , R ∗ i , H) = Ū−i(I ∗ i , R ∗ i , L) = max

((I i ,R i )∈M i ) i=1,...,N

U(I, R, L). (8)

PROOF OF COROLLARY 1: Consider the equilibrium menus (M1 ∗ , M2 ∗ , ..., MN ∗ ). Let z ji ∈ Mj

∗

be an optimal choice in Mj ∗ , given that the entrepreneur has chosen (Ii ∗ , Ri ∗ ) in Mi ∗ , **and** selects

e = L. In particular, we denote z ii = (Ii ∗ , Ri ∗ ), **and** we let Z i = ∑ z ji ∈ L be the corresponding

j

aggregate allocation. Consider now the vector d ji = ( z ji − (Ij ∗ , Rj) ) ∗ ≡ (α ji , β ji ). Given the

definition of Z i , we have Z i = (I ∗ , R ∗ ) + ∑ d ji . The system of equations in (8) can hence be

j

rewritten as

28

∀i = 1, 2, . . . , N : U(I ∗ , R ∗ , H) = U(Z i , L) = U ( (I ∗ , R ∗ ) + ∑ j

d ji , L ) . (9)

Suppose, by contradiction, that (I ∗ , R ∗ ) ∉ Ψ. Then, for (9) to be satisfied, it should be that

∑

d ji ∈ R 2 ++ for each i. Take any (i, j) ∈ {1, 2, .., N} 2 . See that each allocation Z j + d ji =

j

Z j + (z ji − (I ∗ j , R ∗ j)) is available: the entrepreneur can achieve it by trading z ij in each menu M ∗ i ,

with i ≠ j, **and** z ji instead of (Ij ∗ , Rj) ∗ in Mj ∗ . In a similar way, each allocation Z i − d ji can be

achieved by letting the entrepreneur trade the same contracts needed to obtain Z i , but choosing

(Ij ∗ , Rj) ∗ instead of z ji in Mj ∗ . Condition (9) therefore implies that

∀(i, j) ∈ {1, 2, .., N} 2 : U(Z j + d ji , L) ≤ U(Z j , L), (10)

∀(i, j) ∈ {1, 2, .., N} 2 : U(Z i − d ji , L) ≤ U(Z i , L). (11)

We now show that, given (10) **and** (11), each d ji vector is such that

∀(i, j) ∈ {1, 2, .., N} 2 : β ji ≥ (G + B )α ji . (12)

π L

To establish the result, it is useful to denote D = {(I, R) ∈ R 2 + : R > G(I + A)}. Consider

any (i, j) pair. Suppose first that α ji > 0. In this case, (10) implies that Z j ∉ D **and** that β ji ≥

(G + B )α ji , otherwise the payoff associated to Z j + d ji would be greater than that corresponding

π L

to Z j , **and** Z j would fail to be optimal. Suppose next that α ji < 0. The same reasoning (starting

by considering Z j − d ji ) guarantees that (11) implies that Z i ∉ D **and** that β ji ≥ (G + B )α ji .

π L

Suppose finally that α ji = 0. Then, by definition, z ji = (Ij ∗ , Rj ∗ + β ji ) ∈ Mj ∗ . Since (Ij ∗ , Rj) ∗ is

optimal in Mj ∗ , (10) implies that β ji ≥ 0 at equilibrium. This in turn establishes (12).

Taking the sum of all the inequalities (12), we get ∑ j

β ji ≥ (G+ B π L

) ∑ j

α ji . That is, the vector ∑ j

d ji

has a slope greater than or equal to G + B π L

. It hence follows from (9) that if (I ∗ , R ∗ ) ∉ Ψ, then

for each i ∈ {1, 2, ..., N}, Z i ∈ D. Take now any i. Since ∑ j

d ji ∈ R 2 ++ whenever (I ∗ , R ∗ ) ∉ Ψ,

there exists an associated j such that α ji > 0. Condition (10) then implies that Z j

/∈ D, which

contradicts the assumption that for each i ∈ {1, 2, ..., N} Z i ∈ D, **and** completes the proof.

PROOF OF PROPOSITION 2:

29

Mild moral hazard : B ≤ π H G − 1. We show that any aggregate allocation (I ∗ , R ∗ ) ∈ Ψ

is supported at equilibrium by the following strategies. Investor 1 offers M ∗ 1 = {(I ∗ , R ∗ ), (0, 0)},

investor 2 offers M ∗ 2 = {(Ī, ¯R), (0, 0)}. The investment level Ī = π L

∆π (I∗ + A) is such that

U(I ∗ , R ∗ , H) = U(I ∗ , R ∗ , L) = B(I ∗ + Ī + A) − A, (13)

**and** ¯R = G(I ∗ + Ī + A). Each other investor i = 3, 4, ..., N offers M ∗ i = {(0, 0)}.

Investing only Ī is not an optimal choice for the entrepreneur. This is because, given (1), we

have ¯R

Ī = G + G∆π > G + B implying that, unless she defaults, the entrepreneur gets a payoff

π L π L

smaller than the reservation one. Observe also that, given ¯R, if I ∗ + Ī is invested the entrepreneur

defaults **and** gets a payoff B(I ∗ + Ī + A) − A. From (13), choosing (I∗ , R ∗ ) in M ∗ 1 , (0, 0) in M ∗ 2

**and** selecting e = H is an optimal choice for the entrepreneur. We now prove that no investor has

a unilateral profitable deviation.

Deviations of investor 1. Without loss of generality, any deviation by investor 1 can be represented

by a menu M ′ 1 = {(I ′ 1, R ′ 1), (0, 0)}. Given Lemma 1, any profitable deviation should induce

the entrepreneur to choose e = H. In this case, the deviation is profitable only if

π H R ′ 1 − I ′ 1 > π H R ∗ − I ∗ . (14)

Since (I ∗ , R ∗ ) ∈ Ψ, (4) implies that (14) is satisfied only if I 1 ′ < I ∗ **and** R 1 ′ < R ∗ . We therefore

write (I 1, ′ R 1) ′ = (I ∗ − ε, R ∗ − ηε), with η ∈ (G − B ∆π , 1

) **and** ε ∈ (0, GI∗ − R ∗

). The lower

π H G − η

bound on η ensures that, if she only trades the deviating contract, the entrepreneur selects e = H,

while the upper bound ensures that (14) is satisfied. The upper bound on ε ensures that, if she only

trades the deviating contract, the entrepreneur’s payoff is above her reservation one. It follows

from Lemma 2 that when she chooses e = H, the entrepreneur selects (I 1, ′ R 1) ′ in M 1, ′ **and** (0, 0)

in M2 ∗ , since ¯R

Ī > G. Given η **and** ε, choosing only (I′ 1, R 1) ′ optimally induces e = H. Once

again, when choosing e = L, she prefers not to invest rather than investing only Ī. She therefore

strategically defaults, since π L

(

G(I ′ + Ī + A) − (R′ + ¯R) ) < 0. Comparing her respective payoff

under e = H **and** e = L, one gets

π H (G(I ∗ − ε + A) − (R ∗ − η ε)) = B(I ∗ + Ī + A) − π H(G − η)ε < B(I ∗ − ε + Ī + A),

where the last inequality obtains since the assumption of mild moral hazard together with η < 1

π H

30

**and** ε > 0 imply that B ≤ π H (G − 1/π H ) < π H (G − η). Thus, the entrepreneur finds optimal to

default, contradicting the assumption that e = H is an optimal choice.

Deviations of investor 2. Let M ′ 2 = {(I ′ 2, R ′ 2), (0, 0)} be a deviating menu of investor 2. Once

again, we restrict attention to deviations inducing the entrepreneur to choose e = H. In that case,

it follows from Lemma 2 that for (I ′ 2, R ′ 2) to be traded by the entrepreneur, we must have R′ 2

I ′ 2

Since (I ∗ , R ∗ ) ∈ F, we also have R∗

I ∗

≤ G. It follows that

U(I ∗ + I ′ 2, R ∗ + R ′ 2, H) ≥ max{U(I ′ 2, R ′ 2, H), U(I ∗ , R ∗ , H)}.

For (I ′ 2, R ′ 2) to be a profitable deviation, it must be R′ 2

R 2

′

I 2

′

I ′ 2

≤ G.

> 1

π H

. It then follows from (4) that

> G − B ∆π **and** U(I∗ + I ′ 2, R ∗ + R ′ 2, L) > U(I ∗ + I ′ 2, R ∗ + R ′ 2, H). This implies that the

entrepreneur chooses e = L following the deviation to M ′ 2, which constitutes a contradiction. The

same reasoning yields that none of the inactive investors i = 3, ..., N has a unilateral deviation.

Strong moral hazard: B > π H G − 1. First, we show that no aggregate allocation different

from (I m , R m ) can be supported at equilibrium. Suppose that (I ∗ , R ∗ ) ∈ Ψ \ {(I m , R m )} is an

aggregate equilibrium allocation. There must hence be at least one traded contract, say (I ∗ i , R ∗ i ),

of price p i < G. We then show that investor i can profitably deviate to M ′ i = {(0, 0), (I ∗ i − ε, R ∗ i −

ηε)}, with η ∈ (G − B ∆π , 1

π H

) **and** ε small enough. If the entrepreneur selects e = H, then, by

Lemma 1, the contract (I ∗ i −ε, R ∗ i −ηε) is chosen in the deviating menu. In that case, the deviation

is profitable by construction. The maximum payoff available at the deviation stage when e = H is

selected is

Ū −i (I ∗ i − ε, R ∗ i − ηε, H) = ε π H (η − G) + Ū−i(I ∗ i , R ∗ i , H),

**and** the maximum payoff available at the deviation stage if e = L is chosen is

Ū −i (I ∗ i − ε, R ∗ i − ηε, L) =

⎧

⎨

⎩

ε π L (η − G) − Bε + Ū−i(I ∗ i , R ∗ i , L)

−Bε + Ū−i(I ∗ i , R ∗ i , L)

if no default,

if default.

In case of no default, following (8), we have that Ū−i(Ii ∗ − ε, Ri ∗ − ηε, H) > Ū−i(Ii ∗ − ε, Ri ∗ −

ηε, L) ⇔ η − G − B > 0, which is satisfied by definition of strong moral hazard. Similarly, in

∆π

case of default, Ū−i(Ii ∗ − ε, Ri ∗ − ηε, H) > Ū−i(Ii ∗ − ε, Ri ∗ − ηε, L) ⇔ π H (η − G) + B > 0, which

is satisfied by construction whenever B > π H G − 1. Thus, (I ∗ , R ∗ ) ∈ Ψ \ {(I m , R m )} cannot be

31

an equilibrium allocation.

Second, we show that (I m , R m ) is supported at equilibrium. Consider the following strategies.

Investor 1 offers M 1 = {(I m , R m ), (0, 0)} **and** each other investor i = 2, ..., N offers

M i

= {(0, 0)}. Then, it is a best reply for the entrepreneur to select the contract (I m , R m ) in

M 1 **and** to choose e = H. Since investor 1 earns the monopolistic profit, only deviations of the

inactive investors must be considered. The same reasoning developed for the deviations of investor

2 in the mild moral hazard case guarantees that no investor has a profitable deviation.

PROOF OF PROPOSITION 3: Take any (I ∗ , R ∗ ) ∈ F such that I ∗ ≥ I m **and** consider the following

profile of strategies. Each investor i = 1, 2, ..., N offers M ∗ = {(0, 0); ( I ∗

N , R∗ (.) ) ; ( Î

N , ˆR(.) ) },

with R ∗ (I) = G(I + A) for I /∈ {I ∗ , I∗

}, **and**

⎧

N

⎪⎨

R ∗

if I = I ∗ ,

R ∗ (I) = N

⎪⎩ R ∗ − N − 1

N GI∗ if I = I∗

N .

The investment level Î is such that

U(I ∗ , R ∗ , H) = B(Î + A) − A, (15)

which guarantees that Î ≥ I∗ ≥ 0, 31 **and** ˆR(I) = G(I + A) ∀I ∈ R + . It is immediate to verify

that choosing the investment I∗

in each menu **and** selecting e = H is an optimal choice for the

N

entrepreneur. Observe also that the repayment R ∗ ( I∗

) guarantees that the entrepreneur achieves

N

her equilibrium payoff by trading with only one investor. This is turn implies that U(I ∗ , R ∗ , H) is

available to the entrepreneur if any of the investors withdraws his offer. Consider now investors’

deviations. Without loss of generality, any unilateral deviation can be represented by a menu

M ′ = { ( I ∗

N + I′ , R∗

N + R′ (.) ) ]

, (0, 0)} with I ′ ∈

[− I∗

N , Ic − I∗

. Since π L = 0, it follows from

N

Lemma 1 that any profitable deviation necessarily induces e = H. We then have that

R ′ (I) > 1

π H

I ′ , (16)

with I being the aggregate investment traded at the deviation stage. If, following the deviation, the

31 As U(I ∗ , R ∗ , H) ≥ B(I ∗ + A) − A.

32

entrepreneur chooses e = L, then, as π L = 0, **and** given (15), she gets

U sd = U(I ∗ , R ∗ , H) + B ( I ′ + I∗

N − Î )

.

N

Consider the entrepreneur’s payoff when she chooses e = H. Assume first that several contracts

are traded at the deviation stage. Given equilibrium covenants, e = H is an optimal choice only

if the corresponding investment is I = I ∗ . The deviating contract should hence be such that

I ′ = k I∗

, with k ∈ {1, ..., N − 2} **and** N − k − 1 being the number of non deviating lenders who

N

actively trade at the deviation stage. 32 The corresponding entrepreneur’s payoff is

π H

(

G(I ∗ + A) − N − k

N R∗ − R ′ (I ∗ ) ) +

− A < U(I ∗ , R ∗ , H) + k N (π HR ∗ − I ∗ ),

where the inequality follows from (16). Observe that, since π H R ∗ − I ∗ = (π H G − 1)I ∗ −

(U(I ∗ , R ∗ k (

, H) − U(0)), the entrepreneur strategically defaults if U(I ∗ , R ∗ , H) − U(0) ) +

N

B

N (I∗ − Î) ≥ k N (π HG − 1 − B)I ∗ or, equivalently, if

U(I ∗ , R ∗ , H) − U(0) − B k (Î − I∗ ) ≥ (π H G − 1 − B)I ∗ , (17)

which right-h**and** side is negative because I ∗ ≥ 0, **and** B > π H G − 1. 33 Since, for any (I ∗ , R ∗ ) ∈

F, the left-h**and** side of (17) is increasing in k, **and** the right-h**and** side is decreasing in I ∗ , a

sufficient condition for the inequality to hold for all (k, I ∗ ) is that it is satisfied for k = 1 when the

right-h**and** side equal to zero. That is, we have to check that

U(I ∗ , R ∗ , H) − U(0) − B(Î − I∗ ) ≥ 0, (18)

which, given (15), is equivalent to B(I ∗ + A) − A ≥ U(0) ⇔ I ∗ ≥ I m since π L = 0.

Suppose now that the entrepreneur only trades with the deviating investor, that is I = I∗

N + I′ .

In this case, choosing e = H yields the payoff

π H

(

G(

I ∗

N + I′ + A) − R∗

N − R′ ( I∗

N + I′ ) ) + U(0)(N − 1) + U(I ∗ , R ∗ , H)

− A <

N

+ (π H G − 1)I ′ .

Since the entrepreneur’s equilibrium utility remains available at the deviation stage, the right-h**and**

32 Observe that one cannot have N − 1 non deviating lenders who actively trade. In that case, I ′ is equal to zero **and**

the deviating lender who offers I∗

N cannot increase his profit since U(I∗ , R ∗ , H) is available to the entrepreneur at the

deviation stage.

33 The last inequality follows from (4) when π L = 0.

33

side of the last inequality must a fortiori be strictly greater than U(I ∗ , R ∗ , H), which implies that

I ′ > 0. The entrepreneur therefore finds optimal to strategically default if

N − 1( U(I ∗ , R ∗ , H) − U(0) ) − B N

N (Î − I∗ ) ≥ (π H G − 1 − B)I ′ . (19)

Since N ≥ 2, **and** because π H G − 1 − B < 0 when π L = 0, any (I ∗ , R ∗ ) which satisfies (18) also

satisfies (19). This implies that each (I ∗ , R ∗ ) ∈ F such that I ∗ ≥ I m is supported at equilibrium.

PROOF OF PROPOSITIONS 4 **and** 5: See additional Appendix A.

PROOF OF PROPOSITION 6: We design the schedules T 1 (.), ..., T k (.), ..., T N (.), which identify

the transfers to be paid by investors if the project succeeds contingent on the array of investors’

participation decisions. Fix an arbitrary k ∈ {1, . . . , N}. For each investor i ≠ k, we let

⎧

⎪⎨

1 1 − δ

T i (I) = π H N I if I ∈ [0, Ic ),

⎪⎩ 0 if I ≥ I c ,

(20)

with δ ∈ (0, 1 − B), if all investors agree to participate. In addition, we let

⎧

⎨ −ν if I ∈ [0, I c ),

T i (I) =

⎩ 0 if I ≥ I c ,

(21)

for some ν > 0, if investor i agrees to participate **and** at least one of the others does not. In this

case, investor i receives a positive reward. Finally, T i (I) = 0 for all I ≥ 0 if investor i does not

participate. Consider then investor k. We have T k (I) = 0 ∀I ≥ 0 if all remaining N − 1 investors

pay for the transfer, **and** T k (I) = Kν ∀I ≥ 0 if only K of them decide to do so. In this last case,

investor k effectively rewards his competitors, which guarantees feasibility of the mechanism. The

subsidy to the entrepreneur is therefore

⎧

⎪⎨

1 N − 1

(1 − δ)I if all investors participate,

T (I) = π H N

⎪⎩ 0 if at least one investor does not participate.

The proof is by contradiction. Consider any equilibrium allocation (I ∗ , R ∗ ) ≠ (I c , R c ) supported

by investors’ menus M1

∗ = ... = MN ∗ . In a symmetric equilibrium, (I∗ i , Ri ∗ ) = ( I∗

N , R∗

N ) for

each i **and** each investor earns the same profit. Thus, given that the transfer schedules are not

34

(22)

symmetric, no transfer is provided at equilibrium **and** we have U(I ∗ , R ∗ , H) ≥ B(Î + A) − A,

with Î = ∑ Î i , **and** Îi = Î

i

N being the highest investment level in M i ∗ . We show that, given the

schedules (T 1 (.), T 2 (.), ..., T N (.)), investor k has a profitable deviation. Suppose that he deviates

to the menu M k ′ = ((0, 0), (I′ k + ε, R′ k (.))) with I′ k = I∗ − N − 1 π H R ∗ − I ∗

**and** ε > 0. The

N π H G − 1

function R k ′ (.) is such that

⎧

⎪⎨

R k(I) ′ =

⎪⎩

R ∗

N + 1 (

I

′

π k − I∗

H N + ε) + ε 2 if I = I ′ k + ε,

G(I + A) if I ≠ I ′ k + ε.

For ε = 0, the pair (I ′ k + ε, R′ k (I′ k + ε)) is such that U(I′ k , R′ k (I′ k ), H) = U(I∗ , R ∗ , H) **and**

R ∗

π H R k ′ (I′ k ) − I′ k = π H

N − I∗

N . That is, (I′ k + ε, R′ k (I′ k + ε)) lies at the intersection of the entrepreneur’s

equilibrium indifference curve with the investor k’s equilibrium isoprofit line. In

addition, since (I ∗ , R ∗ ) ≠ (I c , R c ), (I k ′ + ε, R′ k (I′ k + ε)) ∈ int(F) for ε small enough. The deviation

is designed to induce the entrepreneur to trade (I k ′ + ε, R′ k (I′ k + ε)) **and** to choose e = H.

It is immediate to see that, given M k ′ , if the entrepreneur chooses e = H, she can get a payoff

above the equilibrium one by choosing (I ′ k + ε, R′ k (I′ k + ε)). In this case, however, given R′ k (.),

she finds optimal to trade with investor k alone. It follows that, as long as ε > 0, M k ′ is a profitable

deviation for investor k whenever e = H is chosen. We now show that if investors agree to finance

the subsidy to the entrepreneur, she strictly prefers to choose e = H following this deviation. This

is indeed the case if

U(I ∗ , R ∗ , H) + (π H G − 1 − ε)ε + N − 1

N

(1 − δ)(I′ k + ε) > B(Î + A) − A + B( I k ′ + ε − Î )

,

N

where the left-h**and** side is equal to U(I k ′ + ε, R′ k (I′ k + ε), H). Choose ε small enough to ensure

(π H G − 1 − ε) > 0; since I′ k

N ≤ I∗

N ≤ Î , a sufficient condition to verify the above inequality is

N

U(I ∗ , R ∗ , H) − (B(Î

+ A) − A) +

N − 1

N (1 − δ)I′ k > B N − 1

N I′ k,

which is satisfied because U(I ∗ , R ∗ , H) ≥ B(Î + A) − A, **and** δ < 1 − B.

Consider then the investors’ participation subgame. Given (20) **and** (21), it is a dominant

strategy for each nondeviating investor i ≠ k to pay for the transfer. In particular, paying is an

optimal choice when investor i anticipates that all other investors pay if 1 − δ

N (I′ k + ε) < Î

N

35

which, given that Î > 0, δ < 1, **and** I′ k ≤ I∗ ≤ Î, is satisfied for ε sufficiently small. It follows that

providing the subsidy to the entrepreneur is the unique equilibrium outcome of the continuation

game. This in turn guarantees that M k ′ induces the entrepreneur to choose e = H, **and** constitutes

a profitable deviation for investor k.

To complete the proof, see that given the schedules (T 1 (.), T 2 (.), ..., T N (.)), there exists a profile

of symmetric menus (M ∗ 1 , ..., M ∗ N ) which supports (Ic , R c ) as an equilibrium allocation. This

follows straightforwardly from the fact that T (I) = 0 for I ≥ I c **and** that (I c , R c ) can be supported

at equilibrium with the symmetric strategies exhibited in Proposition 3.

The proof extends to the case in which investors can write covenants contingent on the initial

debt I 0 (see additional Appendix A).

36

References

Acharya, Viral **and** Alberto Bisin, “Counterparty Risk Externality: Centralized versus over-the-

Counter Markets,” Journal of Economic Theory, 2014, 149(C), 153–182.

, Heitor Almeida, Filippo Ippolito, **and** Ander Perez, “Credit Lines as Monitored Liquidity

Insurance: Theory **and** Evidence,” Journal of Financial Economics, 2014, 112(3), 287 – 319.

Aghion, Philippe **and** Patrick Bolton, “Contracts as a Barrier to Entry,” American Economic

Review, 1987, 77(3), 388–401.

Allen, Franklin **and** Douglas Gale, “Financial Intermediaries **and** Markets,” Econometrica, 2004,

72(4), 1023–1061.

Attar, Andrea **and** Arnold Chassagnon, “On Moral Hazard **and** Nonexclusive Contracts,” Journal

of Mathematical Economics, 2009, 45(9-10), 511–525.

, Thomas Mariotti, **and** François Salanié, “Nonexclusive Competition in the Market for

Lemons,” Econometrica, 2011, 79(6), 1869–1918.

, , **and** , “Nonexclusive Competition under Adverse Selection,” Theoretical Economics,

2014, 9(1), 1–40.

Bennardo, Alberto, Marco Pagano, **and** Salvatore Piccolo, “**Multiple**-Bank Lending, Creditor

Rights **and** Information Sharing,” Review of Finance, 2015, 19(2), 519–570.

Bernanke, Ben, “Clearing **and** Settlement during the Crash,” The Review of Financial Studies,

1990, 3(1), 133–151.

Bisin, Alberto **and** Adriano Rampini, “Exclusive Contracts **and** the Institution of Bankruptcy,”

Economic Theory, 2006, 27(2), 277–304.

**and** Danilo Guaitoli, “Moral Hazard **and** Nonexclusive Contracts,” RAND Journal of Economics,

2004, 35(2), 306–328.

Bizer, David **and** Peter DeMarzo, “Sequential Banking,” Journal of Political Economy, 1992,

100(1), 41–61.

37

Bolton, Patrick **and** David S Scharfstein, “Optimal Debt Structure **and** the Number of Creditors,”

Journal of Political Economy, 1996, 104(1), 1–25.

**and** Olivier Jeanne, “Structuring **and** Restructuring Sovereign Debt: The Role of Seniority,”

The Review of Economic Studies, 2009, 76(3), 879–902.

Chava, Sudheer **and** Michael Roberts, “How Does Financing Impact Investment? The Role of

Debt **Covenants**,” Journal of Finance, 2008, 63(5), 2085–2121.

Degryse, Hans, Vasso Ioannidou, **and** Erik von Schedvin, “On the Non-Exclusivity of Loan

Contracts: An Empirical Investigation,” Management Science, 2015. forthcoming.

Demiroglu, Cern **and** Christopher James, “The Information Content of Bank Loan **Covenants**,”

Review of Financial Studies, 2010, 23(10), 3700–3737.

Farhi, Emmanuel, Mikhail Golosov, **and** Aleh Tsyvinski, “A Theory of Liquidity **and** Regulation

of Financial Intermediation,” Review of Economic Studies, 2009, 76, 973–992.

Gorton, Gary, “Clearinghouses **and** the Origin of Central Banking in the United States,” The

Journal of Economic History, 1985, 45(2), 277–283.

**and** James Kahn, “The Design of Bank Loan Contracts,” Review of Financial Studies, 2000,

13(2), 331–64.

Gârleanu, Nicolae **and** Jeffrey Zwiebel, “Design **and** Renegotiation of Debt **Covenants**,” Review

of Financial Studies, 2009, 22(2), 749–781.

Han, Seungjin, “Implicit Collusion in Non-exclusive Contracting under Adverse Selection,” Journal

of Economic Behavior **and** Organization, 2014, 99(C), 85–95.

Hautcoeur, Pierre-Cyrille, Angelo Riva, **and** Eugene White, “Floating a "Lifeboat": the Banque

de France **and** the Crisis of 1889,” Journal of Monetary Economics, 2014, 65, 104–119.

Holmstrom, Bengt **and** Jean Tirole, “Financial Intermediation, Loanable Funds **and** the Real

Sector,” Quarterly Journal of Economics, 1997, 112(3), 663–691.

**and** , “Private **and** Public Supply of Liquidity,” Journal of Political Economy, 1998, 106(1),

1–40.

38

Kahn, Charles **and** Dilip Mookherjee, “Competition **and** Incentives with Nonexclusive Contracts,”

RAND Journal of Economics, 1998, 29(3), 443–465.

Kroszner, R**and**all, “Can the Financial Markets Privately Regulate Risk?: The Development of

Derivatives Clearinghouses **and** Recent over-the-Counter Innovations,” Journal of Money, Credit

**and** Banking, 1999, 31(3), 596–618.

Martimort, David **and** Lars Stole, “The Revelation **and** Delegation Principles in Common

Agency Games,” Econometrica, 2002, 70(4), 1659–1673.

**and** , “Contractual Externalities **and** Common Agency Equilibria,” Advances in Theoretical

Economics, 2003, 3(1).

Myers, Stewart, “Determinants of Corporate Borrowing,” Journal of Financial Economics, 1977,

5(2), 147 – 175.

Parlour, Christine **and** Uday Rajan, “Competition in Loan Contracts,” American Economic Review,

2001, 91(5), 1311–1328.

Peters, Michael, “Common Agency **and** the Revelation Principle,” Econometrica, 2001, 69(5),

1349–1372.

, “Reciprocal Contracting,” Journal of Economic Theory, 2015, 158, 102–126.

**and** Cristian Troncoso-Valverde, “A Folk Theorem for Competing Mechanisms,” Journal of

Economic Theory, 2013, 148(3), 953–973.

Philippon, Thomas **and** Vasiliki Skreta, “Optimal Interventions in Markets with Adverse Selection,”

American Economic Review, 2012, 102(1), 1–28.

Rauh, Joshua **and** Amir Sufi, “Capital Structure **and** Debt Structure,” Review of Financial Studies,

2010, 23(12), 4242–4280.

Riva, Angelo **and** Eugene White, “Danger on the Exchange: How Counterparty Risk was Managed

on the Paris Exchange in the Nineteenth Century,” Explorations in Economic History, 2011,

48(4), 478–493.

Roberts, Michael **and** Amir Sufi, “Control Rights **and** Capital Structure: An Empirical Investigation,”

Journal of Finance, 2009, 64(4), 1657–1695.

39

Smith, Clifford **and** Jerold Warner, “On Financial Contracting:

**Covenants**,” Journal of Financial Economics, 1979, (7(2)), 117–161.

An Analysis of Bond

Sufi, Amir, “Bank Lines of Credit in Corporate Finance: An Empirical Analysis,” Review of

Financial Studies, 2009, 22(3), 1057–1088.

Szentes, Balász, “Contractible Contracts in Common Agency Problems,” Review of Economic

Studies, 2015. forthcoming.

Tirole, Jean, “Overcoming Adverse Selection: How Public Intervention Can Restore Market

Functioning,” American Economic Review, 2012, 102(1), 29–59.

White, Eugene, “The Crash of 1882 **and** the Bailout of the Paris Bourse,” Cliometrica, 2007, 1,

115–144.

Yamashita, Takuro, “Mechanism Games with **Multiple** Principals **and** Three or More Agents,”

Econometrica, 2010, 78(2), 791–801.

40

PROOF OF PROPOSITION 4: Define

FOR ONLINE PUBLICATION

ADDITIONAL APPENDIX A

**and**

γ ≡ max {

( ) 2

G

G

πH G − B 1 }

B(G − B ,

∆π ) (π H G − 1)(G − B ,

> 0,

∆π ) B

π H (G − B ∆π )

U N = U B +

2γ √

N − 1

(U(I c , R c , H) − U B ) with U B = max{U(0), U(I c , R c , H) − BI c }.

Let us consider a number of investors N satisfying N > N = (2γ + 1) 2 . Observe that U N <

U(I c , R c , H) for any N > N. Note also that, when B > π H G − 1, we have U B = U(0) so that

lim U N = U(0) as stated in Proposition 4.

N−→∞

We exhibit a profile of investors’ strategies that supports any allocation (I ∗ , R ∗ ) satisfying (6) at

equilibrium. Each investor i = {1, 2, ..., K}, with K ≡ ⌊ √ N⌋, offers Mi

∗ = {(0, 0); ( I ∗

K , R∗ i (.) ) },

with Ri ∗ (I) = G(I + A) for I /∈ {I ∗ , I∗

}, **and**

⎧ K

⎪⎨ Ri ∗ (I ∗ ) = R∗

K ,

⎪⎩ Ri ∗ ( I∗

K ) = (R∗ − K − 1

K GI∗ ).

Each investor j = {K + 1, ..., N} offers Mj ∗ = {(0, 0); ( Ī

N − K , ¯R j (.) ) } with ¯R j (I) = G(I +

A) ∀I ∈ R + . The investment level Ī is such that

U(I ∗ , R ∗ , H) = B(I ∗ + Ī + A) − A.

(A.1)

Choosing I∗

in each menu **and** selecting e = H is an optimal choice for the entrepreneur. Observe

K

also that the repayment Ri ∗ ( I∗

K ) guarantees that U(I∗ , R ∗ , H) is available to the entrepreneur if any

of the investors withdraws his offer.

Now consider the investors’ deviations. We first remark that every profitable deviation must

induce the entrepreneur to choose e = H. Given Lemma 1, a single investor may achieve a positive

profit by inducing e = L only if the entrepreneur trades several contracts out of equilibrium **and**

does not default. Given the equilibrium covenants, this is possible only if she takes up an aggregate

loan of I ∗ **and** chooses e = L, which yields the entrepreneur a payoff smaller than the available

equilibrium one, **and** cannot be an optimal choice.

41

We next show that there are not unilaterally profitable deviations for investors, which sustains

the allocation (I ∗ , R ∗ ) at equilibrium. In particular, a deviation of investor i ∈ {1, · · · , K} can be

characterized, without loss of generality, by the menu M i ′ = { ( I ∗

K + I′ i, R∗

K + R′ i(.) ) , (0, 0)}, with

]

I i ′ ∈

[− I∗

K , Ic − I∗

. Similarly, a deviation of any investor j ∈ {K + 1, · · · , N} can be characterized

by the menu M j ′ = { ( Ī

K

N − K + I′ j, R j(.) ) [

′

, (0, 0)}, with I j ′ ∈ − Ī

]

N − K , Ī

Ic − .

N − K

Observe that for any such deviation to be profitable when the entrepreneur chooses the aggregate

investment I **and** the effort e = H, one needs that

R i(I) ′ > 1 I

π

i, ′ R j(I) ′ > 1 (

)

I ′ Ī

j + , (A.2)

H π H N − K

for i ∈ {1, · · · , K}, j ∈ {K + 1, · · · , N}. If, following any such deviation, the entrepreneur

defaults, her payoff is

U sd = U(I ∗ , R ∗ , H) + BI ′ h,

with h = i, j. We now evaluate the entrepreneur’s payoff when e = H is chosen.

First, consider a deviation of an active investor i. If the entrepreneur only trades with investor

i at the deviation stage, her payoff is

( I ∗

π H G(

K +I′ i+A)− R∗

K −R′ i(I i+ ′ I∗

K )) + K − 1

−A <

K U(0)+ U(I∗ , R ∗ , H)

+I ′

K

i(π H G−1), (A.3)

where the inequality follows from (A.2). Since the equilibrium utility is available at the deviation

stage, the right-h**and** side of (A.3) must be strictly larger than U(I ∗ , R ∗ , H), which implies that

I ′ i > 0. Thus, if U sd is greater than the right-h**and** side of (A.3), the entrepreneur finds it optimal to

strategically default. This is the case whenever

K − 1( U(I ∗ , R ∗ , H) − U(0) ) ≥ (π H G − 1 − B)I ′

K

i. (A.4)

We show that each (I ∗ , R ∗ ) ∈ F satisfying (6) also satisfies (A.4). We consider two cases:

1. If B > π H G − 1, then since the right-h**and** side of (A.4) is negative for each I ′ i > 0, (A.4) is

satisfied by any (I ∗ , R ∗ ) ∈ F.

2. If B ≤ π H G − 1, a sufficient condition for (A.4) to hold is

U(I ∗ , R ∗ , H) − U B ≥ 1

K − 1 (U B − U(0)),

in which I ′ i is replaced by I c in (A.4). Note that, as K − 1 ≥ √ N − 1, we have that

1

(A.5)

2

√

N − 1

≥

K − 1 . Moreover, as 42

γ ≥

G

B(G − B > π HG

∆π ) B

> π HG − 1 − B

B

=

U B − U(0)

U(I c , R c , H) − U B

,

condition (6) implies (A.5) thus the result.

If several contracts are traded at the deviation stage, then, given the equilibrium covenants,

e = H is an optimal choice only if the aggregate investment is I ∗ . In this case, we necessarily have

I i ′ = k I∗

, with k ∈ {1, ..., K − 2}. The entrepreneur’s payoff when e = H is chosen is then

K

(

π H G(I ∗ [K − (k + 1)] + 1

+ A) − R ∗ − R ′

K

i(I ∗ ) ) − A < U(I ∗ , R ∗ , H) + k K (π HR ∗ − I ∗ ), (A.6)

where the inequality follows from (A.2). The entrepreneur strategically defaults if U sd is greater

than the value in the right-h**and** side of (A.6) which, since π H R ∗ −I ∗ = (π H G−1)I ∗ −(U(I ∗ , R ∗ , H)−

U(0)), yields

k (

U(I ∗ , R ∗ , H) − U(0) ) ≥ k K

K (π HG − 1 − B)I ∗ . (A.7)

Since each (I ∗ , R ∗ ) ∈ F satisfying (6) is such that U(I ∗ , R ∗ , H) ≥ U B , any of these allocations

also satisfies (A.7).

Suppose now that an inactive investor j deviates. If the entrepreneur only trades with investor

j at the deviation stage, given (A.2), her payoff when e = H is chosen is bounded above by

U(0) + (π H G − 1) ( Ī

N − K + ) I′ j . The entrepreneur finds therefore optimal to strategically default

if

U(I ∗ , R ∗ , H) − U(0) ≥ I j(π ′ Ī

H G − 1 − B) +

N − K (π HG − 1). (A.8)

From (A.1), we get

BĪ ≤ (π HG − 1 − B)I ∗ + U(0) + (1 − B)A.

Given (A.9), a sufficient condition for (A.8) is then

(A.9)

U(I ∗ , R ∗ , H) − U(0) ≥(π H G − 1 − B)I j

′

1 π H G − 1( +

U(0) + (1 − B)A + (πH G − 1 − B)I ∗) .

N − K B

(A.10)

We show that each (I ∗ , R ∗ ) ∈ F satisfying (6) also satisfies (A.10). We consider two cases:

1. If B > π H G−1, we remark that the inequality U(I ∗ , R ∗ , H)−U(0) ≥ (− Ī

N − K )(π HG−1−

43

Ī

B) +

N − K (π Ī

HG − 1) = B is weaker than (A.8). Indeed, we obtain the right-h**and** side

N − K

by replacing in (A.8) I j ′ by its lower bound − Ī

N − K . Given (A.9), **and** since π HG − 1 − B < 0,

(A.8) is a fortiori weaker than U(I ∗ , R ∗ 1

, H) − U(0) ≥ (U(0) + (1 − B)A), which in

N − K

turn implies (A.10). To show that this inequality holds for each (I ∗ , R ∗ ) ∈ F, observe that,

2

1

since N ≥ 3 by construction, we get √ ≥

N − 1 N − √ N − 1 ≥ 1

. In addition, using

N − K

π H G − 1 < B < 1 we have

γ ≥

G

(π H G − 1)(G − B ∆π ) >

( 1

(π H G − B) − (G − B )

π H ∆π )

(π H G − 1)(G − B = (π HG − B)A U(0) + (1 − B)A

=

∆π ) (π H G − 1)Ic U(I c , R c , H) − U(0) .

Condition (6) implies then that

U(I ∗ , R ∗ , H)−U(0) ≥

thus the result.

1

N − K

U(0) + (1 − B)A

U(I c , R c , H) − U(0) (U(Ic , R c , H)−U(0)) =

2. If B ≤ π H G − 1, replacing I j ′ **and** I ∗ by I c , a sufficient condition for (A.10) is

U(I ∗ , R ∗ 1 π H G − 1( , H) − U(0) ≥ (U B − U(0)) +

UB +(1 − B)A ) .

N − K B

2

√

N − 1

≥

1

N − K . Moreover,

1

N − K (U(0)+(1−B)A),

As in the former case,

⎛

( ) 2 ( ) ( )

1

πH G − B 1 πH G − 1 πH G − B

− (G − B ⎞

γ ≥

B

π H (G − B ≥

⎜

∆π ) B B ⎝ 1 + π H ∆π )

(G − B ⎟

∆π ) ⎠

( ) (

πH G − 1 πH G − B − 1

≥

+ π )

HG − B A

B

B

B I c

≥ π HG − 1 (π H G − 1 − B)I c + A(π H G − B)

= π HG − 1 U B + (1 − B)A

. (A.11)

B

BI c B U(I c , R c , H) − U B

Then, (6) implies that

thus the result.

U(I ∗ , R ∗ , H) − U B ≥

1

N − K

π H G − 1

B

U B + (1 − B)A

(U(I c , R c , H) − U

U(I c , R c B ),

, H) − U B

If several contracts are traded at the deviation stage, then, given the equilibrium covenants,

44

e = H is an optimal choice only if the aggregate investment is I ∗ . In this case, we necessarily have

I j ′ = k I∗

K − Ī , with k ∈ {1, ..., K − 1}. The entrepreneur’s payoff when e = H is chosen is

N − K

then

(

π H G(I ∗ + A) − K − k

K R∗ − R j(I ′ ∗ ) ) − A ≤ U(I ∗ , R ∗ , H) + k K (π HR ∗ − I ∗ ), (A.12)

where the inequality follows from (A.2). The entrepreneur finds optimal to strategically default if

U sd is greater than the value in the right-h**and** side of (A.12), which is the case if

B ( k Ī ) k

K I∗ − ≥

N − K K (π HR ∗ − I ∗ ).

Using (A.9) **and** π H R ∗ − I ∗ = (π H G − 1)I ∗ − (U(I ∗ , R ∗ , H) − U(0)), a sufficient condition for

(A.12) is

k (

(1+B−πH G)I ∗ +U(I ∗ , R ∗ , H)−U(0) ) ≥

K

1

N − K ((π HG − 1 − B)I ∗ + U(0) + (1 − B)A) .

(A.13)

The left-h**and** side of (A.13) is increasing in k. This is straightforward if B > π H G−1. In the case

B ≤ π H G − 1, the result follows from U(I ∗ , R ∗ , H) − U(0) ≥ U B − U(0) = (π H G − 1 − B)I c

**and** (I c − I ∗ ) ≥ 0. It is hence enough to verify (A.13) for k = 1, that is,

U(I ∗ , R ∗ , H) − U(0) ≥

K (U(0)+(1 − B)A) +

N

N − K N − K (π HG − 1 − B)I ∗ .

(A.14)

We show that each (I ∗ , R ∗ ) ∈ F satisfying (6) also satisfies (A.14). We consider two cases:

1. If B > π H G − 1, we show that U(I ∗ , R ∗ , H) − U(0) ≥ K (U(0) + A), which is stronger

N − K

than (A.14), is satisfied. Note that ( √ N − 1) 2 ≥ 0, or equivalently, 2N − 2 √ √

N ≥ N − 1. This

2 N + 1

implies that √ ≥

N − 1 N − √ N ≥ K . In addition,

N − K

G

γ ≥

(π H G − 1)(G − B = π HG 1/π H

∆π ) π H G − 1

G − B ≥

π HG A

π H G − 1 I = U(0) + A

c U(I c , R c , H) − U(0) .

∆π

To conclude, observe that (6) implies

U(I ∗ , R ∗ , H) − U(0) ≥

K U(0) + A

N − K U(I c , R c , H) − U(0) (U(Ic , R c , H) − U(0)).

2. If B ≤ π H G − 1, we remark that

45

≤

K

N − K (U(0)+(1 − B)A) + N(π HG − 1 − B)I ∗

K

N − K (U(0) + (1 − B)A) + N(U B − U(0)) = U B − U(0) +

K

N − K (U B + (1 − B)A) .

We show that U(I ∗ , R ∗ , H) − U B ≥

K

N − K (U B + (1 − B)A), which is stronger than (A.14). As

2

in the former case, we have √ ≥

K . Then, using (A.11), condition (6) implies that

N − 1 N − K

U(I ∗ , R ∗ , H)−U B ≥

K

N − K

U B + (1 − B)A

U(I c , R c , H) − U B

(U(I c , R c , H)−U B ) =

K

N − K (U B+(1−B)A).

Thus, any allocation (I ∗ , R ∗ ) ∈ F satisfying (6) also satisfies (A.4), (A.7), (A.10), **and** (A.14). This

proves that any such allocation is supported at equilibrium by the investors’ strategies M ∗ 1 , ..., M ∗ N .

PROOF OF PROPOSITION 5: We extend the results of Proposition 3 **and** Proposition 4 to the case

in which covenants can be contingent on the initial debt I 0 . Assume π L = 0, take any (I ∗ , R ∗ ) ∈ F

such that I ∗ ≥ I m **and** consider the following profile of strategies. Each investor i = 1, 2, ..., N

offers the same menu M ∗ = {(0, 0, 0, 0), ( I ∗

N , R∗ (.), I + (.), R + (.) ) , ( 0, 0, Î+ (.), ˆR + (.) ) }. We shall

refer to (0, 0, 0, 0) as the null contract, to ( I∗

N , R∗ (.), I + (.), R + (.)) as the equilibrium contract, **and**

to (0, 0, Î+ (.), ˆR + (.)) as the latent contract. In each equilibrium contract, R ∗ (.) is such that

⎧

⎪⎨ R ∗ − N − 1

R ∗ (I 0 , I F (I 0 )) =

⎪⎩

R ∗

N

N GI∗ if I 0 = I F (I 0 ) = I∗

N ,

otherwise,

where I F (I 0 ) is the amount ultimately invested for a given initial I 0 . The additional offer (I + (.), R + (.))

is such that

⎧

⎨ 0 if I

I + 0 = k I∗

, for k = 1, 2, ..., N,

(I 0 ) =

N

⎩

I CL otherwise,

46

⎧

⎪⎨

R + (I 0 , I F (I 0 )) =

⎪⎩

0 if I 0 = I F (I 0 ) = I ∗ , or I 0 = I F (I 0 ) = I∗

N ,

G(I F (I 0 ) + A) if I 0 = I ∗ , I F (I 0 ) ≠ I ∗ , or I 0 = I∗

N , IF (I 0 ) ≠ I∗

N ,

p ∗ I CL otherwise,

with p ∗ = R∗

I ∗ . The investment ICL is the additional credit line which any investor st**and**s ready to

provide against any competitors’ threat of asking for an accelerated repayment. It is such that

U( I∗

N + ICL , p ∗ ( I∗

N + ICL ), L) = U(I c , R c , H).

In each latent contract, the additional offer (Î+ (.), ˆR + (.)) is such that

⎧

⎪⎨

Î + (I 0 ) =

⎪⎩

Î

N

if I 0 = 0,

1 I∗

(Î −

N − 1 N ) if I 0 = I∗

N ,

0 otherwise,

**and** ˆR + (I 0 , I F (I 0 )) = G(I F (I 0 ) + A) for all I 0 **and** I F (I 0 ). As in the proof of Proposition 3, the

investment Î is characterized by U(I∗ , R ∗ , H) = B(Î + A) − A.

Given these offers, the entrepreneur cannot obtain a payoff higher than U(I ∗ , R ∗ , H). At equilibrium,

she achieves U(I ∗ , R ∗ , H) by trading the same equilibrium contract with each of the

investors, receiving thereby no additional funds at the second stage, **and** selecting e = H. As in

the proof of Proposition 3, none of the investors is indispensable to provide the equilibrium payoff:

the entrepreneur can get U(I ∗ , R ∗ , H) by trading the equilibrium contract with only one investor. 34

Consider then investors’ deviations. Without loss of generality, any unilateral deviation can

be represented by a menu M ′ = {(0, 0, 0, 0), ( I ′ , R ′ (.), I ′+ (.), R ′+ (.) ) }. By Lemma 1, **and** given

that π L = 0, each profitable deviation must induce the entrepreneur to choose e = H. Hence, it

must be that I ′ ∈ {0, I∗

N } in any profitable deviation. Indeed, if I′ ∉ {0, I∗

}, the entrepreneur

N

can combine the deviating contract with (at least) one equilibrium contract, **and** obtain the line of

credit (I CL , p ∗ I CL ). This guarantees her (at least) the payoff U( I∗

N + ICL , p ∗ ( I∗

N + ICL ), L) =

U(I c , R c , H). The entrepreneur’s strategy can therefore be constructed so that she strategically

defaults when trading I ′ with the deviating investor. In addition, if the entrepreneur chooses e = H,

then the aggregate initial financing induced by the deviation to M ′ must be I 0 ∈ {0, I∗

N , I∗ }.

34 Alternatively, she can also get U(I ∗ , R ∗ , H) by trading all latent contracts **and** defaulting.

47

Indeed, the repayment R + (.) is such that, if she trades any equilibrium contract together with the

deviating one, the entrepreneur strategically defaults unless her initial outside financing is I∗

N or

I ∗ . 35 Assuming that e = H is chosen, we distinguish two cases, depending on whether the borrower

trades at least one of the equilibrium contracts, or she only trades with the deviating investor. In

the first case, given R + (.), it must be that I 0 = I F (I 0 ) ≤ I ∗ **and** I ′+ (I 0 ) = I + (I 0 ) = 0, otherwise

she would default. It follows that, since I ′ ∈ {0, I∗

N }, we must have I′ = I∗

for the deviation to

N

be profitable. Thus, the corresponding entrepreneur’s payoff is

π H (G( I∗

N + k I∗

N + A) − R′ (I 0 , I F (I 0 )) − R ′+ (I 0 , I F (I 0 )) − k R∗

N ) − A

< π H G(I ∗ + A) − I ′ + I∗

N − π HR ∗ − A = U(I ∗ , R ∗ , H) + ( I∗

N − I′ ),

(A.15)

where k ∈ {1, ..., N − 1} is any number of equilibrium contracts optimally traded by the en-

(

trepreneur when e = H. The latter inequality obtains since π H R ′ (I 0 , I F ) + R ′+ (I 0 , I F ) ) − I ′ >

R ∗

(π H

N − I∗

), which guarantees that the deviation is profitable, **and** by observing that GI∗

N N − R∗

N >

0 by construction. Thus, (A.15) implies that, following the deviation, the payoff achieved by the

entrepreneur when choosing e = H is strictly below U(I ∗ , R ∗ , H), which contradicts the fact that

no investor is indispensable to provide the equilibrium payoff under e = H.

We next consider the case in which, when choosing e = H, the entrepreneur only trades with

the deviating investor, which implies that I 0 = I ′ ∈ {0, I∗

}. Her corresponding payoff is

N

π H

(

G(I ′ + I ′+ (I ′ ) + A) − R ′ (I ′ , I F (I ′ )) − R ′+ (I ′ , I F (I ′ )) ) − A

< π H (G(I ′ + A) − R∗

N ) − A + (π HG − 1)I ′+ (I ′ ) + ( I∗

N − I′ )

(A.16)

< U(I ∗ , R ∗ , H) + (π H G − 1)I ′+ (I ′ ). (A.17)

Inequality (A.16) follows from

π H (R ′ (I ′ , I F (I ′ ) + R ′+ (I ′ , I F (I ′ )) > I ′ + I ′+ (I ′ ) + (π H

R ∗

N − I∗

N ),

which guarantees that the deviation is profitable. Inequality (A.17) obtains because I ′ ≤ I∗

N **and**

π H G − 1 > 0. Since U(I ∗ , R ∗ , H) is available to the entrepreneur at the deviation stage, (A.17)

implies that I ′+ (I ′ ) > 0. We now prove that, by strategically defaulting, the entrepreneur gets

35 ˆR+ (.) is such that trading any of the latent contracts straightforwardly leads to strategic default.

48

a payoff which is above the upper bound in (A.16). Suppose that, together with the deviating

contract, she takes N − 1 latent contracts at the deviation stage. Given ˆR + (.), she then finds

optimal to default. Her corresponding payoff is

U sd = B(I ′ + I ′+ (I ′ ) + (N − 1)Ī+ (I ′ ) + A) − A.

(A.18)

If I 0 = I ′ = I∗

N , (A.18) yields U sd = U(I ∗ , R ∗ , H)+BI ′+ (I ′ ) > U(I ∗ , R ∗ , H)+(π H G−1)I ′+ (I ′ ),

as I ′+ (I ′ ) > 0, **and** B > π H G − 1 since π L = 0. If I 0 = I ′ = 0, (A.16) together with the fact that

U(I ∗ , R ∗ , H) remains available at the deviation stage imply π H (G(I ′+ (0)+A)−I ′+ (0)−(π H

R ∗

N −

I ∗

N ) > U ∗ (I ∗ , R ∗ , H). Since U ∗ (I ∗ , R ∗ , H)−U(0) = (π H G−1)I ∗ −(π H R ∗ −I ∗ ), we get I ′+ (0) >

I ∗ . Thus, without loss of generality, we can write I ′+ (0) = I∗

N + I ′′ with I ′′ > N − 1

N I∗ > 0.

Then, (A.16) implies that the entrepreneur’s payoff is bounded by U(0)(N − 1) + U(I∗ , R ∗ , H)

+

N

(π H G − 1)I ′′ , **and** (A.18) can be rewritten as

U sd = B ( N − 1

N Î + I∗

N + I ′′ + A ) − A = U(I ∗ , R ∗ , H) + B(I ′′ + I∗

N − Î

N ).

As shown in the proof of Proposition 3, we have U sd ≥ U(0)(N − 1) + U(I∗ , R ∗ , H)

+ (π H G −

N

1)I ′′ for each I ′′ > 0. This guarantees that the entrepreneur strategically defaults **and** reestablishes

that any aggregate allocation (I ∗ , R ∗ ) ∈ F satisfying I ∗ ≥ I m is sustained at equilibrium.

We now extend the result of Proposition 4. We exhibit a profile of investors’ strategies that

supports at equilibrium any allocation (I ∗ , R ∗ ) satisfying (6). Each investor i = {1, 2, ..., K}, with

K ≡ ⌊ √ N⌋, offers

M ∗ i

= {(0, 0, 0, 0); ( I ∗

K , R∗ i (.), I + i (.), R+ i (.)) , ( 0, 0, Ī+ i (.), ¯R + i (.))},

**and** each investor j = {K + 1, ..., N} offers

Mj ∗ = {(0, 0, 0, 0); ( Ī

N − K , ¯R j (.), 0, 0 ) ; ( 0, 0, Ĩ+ j (.), ˜R + j (.)) }.

We shall refer to (0, 0, 0, 0) as the null contract, to ( I∗

K , R∗ i (.), I + i (.), R+ i (.)) as the equilibrium

Ī

contract, to (

N − K , ¯R j (.), 0, 0) as the type-1 latent contract, to (0, 0, Ī+ i (.), ¯R + i (.)) as the type-2

latent contract, **and**, to (0, 0, Ĩ+ j (.), ˜R + j (.)) as the type-3 latent contract.

In each equilibrium contract, Ri ∗ (.) is such that

49

⎧

⎪⎨

⎪⎩

Ri ∗ (I 0 ) = (R ∗ − K − 1

K GI∗ ) if I 0 = I F (I 0 ) = I∗

K ,

Ri ∗ (I 0 ) = R∗

K ,

otherwise.

The additional offer (I + i (.), R+ i (.)) is such that:

⎧

⎪⎨

I + i (I 0 if I 0 = k I∗

0) =

K + l Ī

, for k = 1, 2, ..., K **and** l = 0, ..., N − K,

N − K

⎪⎩ I CL otherwise,

**and**

⎧

⎪⎨

R + i (I 0, I F (I 0 )) =

⎪⎩

0 if I 0 = I ∗ , I F (I 0 ) = I ∗ , or I 0 = I∗

K , IF (I 0 ) = I∗

K ,

G(I F (I 0 ) + A) if I 0 = I ∗ , I F (I 0 ) ≠ I ∗ , or I 0 = I∗

K , IF (I 0 ) ≠ I∗

K ,

or I 0 = k I∗

K + l Ī

N − K , ∀IF (I 0 )

for k = 1, 2, ..., K **and** l = 0, ..., N − K, with (k, l) /∈ {(K, 0), (1, 0)},

G(I CL + A) otherwise,

where I CL is such that B( I∗

K + ICL + A) − A = U(I c , R c , H) **and** Ī is such that U(I∗ , R ∗ , H) =

B(I ∗ + Ī + A) − A. The additional offer (Ī+ i (.), ¯R + i (.)) is such that

⎧

⎪⎨ I ∗

Ī + i (I 0) = K

if I 0 = 0 or, I 0 = I∗

K or, I 0 =

⎪⎩ 0 otherwise,

Ī

N − K ,

**and** ¯R + i (I 0, I F (I 0 )) = G(I F (I 0 ) + A) for all I 0 **and** I F (I 0 ).

For any j ∈ {K + 1, 2, ..., N}, the repayment ¯R j (.) satisfies ¯R + j (I 0, I F (I 0 )) = G(I F (I 0 ) + A)

for all I 0 **and** I F (I 0 ) **and** the offers (Ĩ+ j (.), ˜R + j (.)) are such that

⎧

⎪⎨ Ī

Ĩ + j (I 0) = N − K

if I 0 = I∗

K or, I 0 = 0,

⎪⎩ 0 otherwise,

**and** ˜R + j (I 0, I F (I 0 )) = G(I F (I 0 ) + A) for all I 0 **and** I F (I 0 ).

Given these offers, the entrepreneur cannot obtain a payoff higher than U(I ∗ , R ∗ , H).

At

50

equilibrium, she achieves U(I ∗ , R ∗ , H) by trading the equilibrium contract with each lender i ∈

{1, 2, ..., K}, the null contract with each lender j ∈ {K + 1, 2, ..., N}, obtains no additional funds

in the second stage **and** selects e = H. As before, none of the investors is indispensable: the

entrepreneur can get U(I ∗ , R ∗ , H) by trading the equilibrium contract with only one investor. 36

Consider then investors’ deviations. Without loss of generality, a deviation by any investor can

be represented by a menu M ′ = {(0, 0, 0, 0), ( I ′ , R ′ (.), I ′+ (.), R ′+ (.) ) }. Every profitable deviation

must induce the entrepreneur to choose e = H. Indeed, given Lemma 1, a single investor may

achieve a positive profit by inducing e = L only if the entrepreneur trades several contracts out

of equilibrium **and** does not default. Given the equilibrium contracts, this is possible only if she

takes up an aggregate initial financing I 0 = I ∗ , invests ultimately I F (I ∗ ) = I ∗ **and** chooses e = L,

which yields the entrepreneur a payoff smaller than the available equilibrium one, **and** thus cannot

be an optimal choice.

Suppose an investor i ∈ {1, ..., K} deviates. Any profitable deviation must be such that I ′ ∈

{0, I∗

K }. Indeed, if I′ ∉ {0, I∗

}, then the entrepreneur can combine the deviating contract with

K

(at least) one equilibrium contract, **and** get access to the line of credit (I CL , G(I CL + A)). which

ensures her a payoff at least equal to B( I∗

K + ICL + A) − A = U(I c , R c , H). This shows that

in any profitable deviation the borrower defaults, which constitutes a contradiction. In addition,

at the deviation stage, the initial investment I 0 belongs to the set {0, I∗

K , I∗ }. Indeed, R + i (.) is

such that, if she trades any equilibrium contract together with the deviating one, the entrepreneur

necessarily defaults if I 0 /∈ { I∗

K , I∗ }. Assuming that e = H is chosen, we distinguish below two

cases, depending on whether the entrepreneur trades at least one of the equilibrium contracts, or

she only trades with the deviating investor.

Consider that the entrepreneur trades equilibrium contracts at the deviation stage. Given R + i (.),

we have I 0 = I F (I 0 ) which implies I ′+ (I 0 ) = I + (I 0 ) = 0, otherwise the entrepreneur would

default. It follows that, since I ′ ∈ {0, I∗

K }, it must be that I′ = I∗

for the deviation to be profitable.

K

Thus, the entrepreneur’s payoff is

π H (G( I∗

K + k I∗

K + A) − R′ (I 0 , I F (I 0 )) − R ′+ (I 0 , I F (I 0 )) − k R∗

K ) − A < U(I∗ , R ∗ , H), (A.19)

where k ∈ {1, ..., N − 1} is any number of equilibrium contracts optimally traded by the entrepreneur

when e = H. The latter inequality obtains since π H

(

R ′ (I 0 , I F ) + R ′+ (I 0 , I F ) ) − I∗

K >

36 She may also get U(I ∗ , R ∗ , H) by trading the type-1 latent contract with each investor j ∈ {K + 1, 2, ..., N}.

51

R ∗

(π H

K − I∗

), which guarantees that the deviation is profitable, **and** by observing that GI∗

K K − R∗

K >

0 by construction. Thus, (A.19) implies that, following the deviation, the payoff achieved by the

entrepreneur when choosing e = H is strictly below U(I ∗ , R ∗ , H), which contradicts the fact that

no investor is indispensable to provide the equilibrium payoff under e = H.

Consider the case where the entrepreneur trades the null contract with all non deviating lenders.

This implies that I 0 = I ′ ∈ {0, I∗

} **and** her payoff is

K

π H

(

G(I ′ + I ′+ (I ′ ) + A) − R ′ (I ′ , I F (I ′ )) − R ′+ (I ′ , I F (I ′ )) ) − A

< π H (G(I ′ + A) − R∗

K ) − A + (π HG − 1)I ′+ (I ′ ) + ( I∗

K − I′ )

(A.20)

< U(I ∗ , R ∗ , H) + (π H G − 1)I ′+ (I ′ ), (A.21)

where (A.20) obtains since

π H (R ′ (I ′ , I F (I ′ )) + R ′+ (I ′ , I F (I ′ )) > I ′ + I ′+ (I ′ ) + (π H

R ∗

K − I∗

K )

(A.22)

for the deviation to be profitable. The second one obtains because I ′ ≤ I∗

K **and** π HG−1 > 0. Since

the payoff U(I ∗ , R ∗ , H) is available to the entrepreneur at the deviation stage, (A.21) implies that

I ′+ ( I∗

K ) > 0.

We prove that the upper-bound (A.20) of the entrepreneur’s payoff is less than what she gets

if she strategically defaults. Indeed, if, following the deviation, the entrepreneur strategically defaults,

she can select the contract ( 0, 0, Ī+ i (.), ¯R + i (.)) in the menu of each non-deviating investor

i ∈ {1, ..., K} **and** the contract ( 0, 0, Ĩ+ j (.), ˜R + j (.)) in the menu of each lender j ∈ {K + 1, ..., N}.

By doing that, she obtains

U sd = B ( I ′ + I ′+ (I ′ ) + (K − 1)Ī+ i (I′ ) + (N − K)Ĩ+ j (I′ ) + A) ) − A.

(A.23)

If I ′ = I∗

K , then U sd = U(I ∗ , R ∗ , H)+BI ′+ ( I∗

). Thus, using (A.21), a sufficient condition for the

K

entrepreneur to strategically default is K − 1

K

(U(I∗ , R ∗ , H) − U(0)) ≥ (π H G − 1 − B)I ′+ ( I∗

K ).

This corresponds to (A.4), which holds from the proof of Proposition 4. If I ′ = 0, we deduce from

(A.22) that

π H (R ′ (0, I F (0)) + R ′+ (0, I F (0))) > I ′+ (0) + (π H

R ∗

K − I∗

K ),

(A.24)

from which it follows that I ′+ (0) > I ∗ . Thus, without loss of generality, we write I ′+ (0) = I∗ ′′

+I

K

52

with I ′′ > K − 1

K

I∗ > 0. Then, using again (A.20) **and** (A.24), we get an upper bound for the

entrepreneur’s payoff:

(

π H G(I ′+ (0)+A)−R ′ (0, I F (0))−R ′+ (0, I F (0)) ) −A < U(0)(K − 1) + U(I∗ , R ∗ , H)

+(π H G−1)I ′′ .

K

But (A.23) becomes U sd = B ( K − 1

K

I∗ + I∗

K +I ′′ +Ī+A) −A = U(I ∗ , R ∗ , H)+BI ′′ . Thus, a sufficient

condition for the entrepreneur to default is K − 1

K (U(I∗ , R ∗ , H)−U(0)) ≥ (π H G−1−B)I ′′

with I ′′ > 0. Again, we have established this relation in the proof of Proposition 4.

Suppose an investor j ∈ {K + 1, ..., N} deviates. Any profitable deviation must be such that

I ′ Ī

∈ {0,

N − K }. Indeed, if Ī

I′ ∉ {0, }, then the entrepreneur can combine equilibrium

N − K

contracts **and**/or latent contracts of type-1, get access to the line of credit, **and** earn at least the

payoff B( I∗

K + ICL + A) − A = U(I c , R c , H). This shows that, in any profitable deviation,

the entrepreneur defaults, which constitutes a contradiction. Below, we consider the two cases

I ′ Ī

=

N − K **and** I′ = 0 **and** we show that in each case the entrepreneur strategically defaults

following the deviation.

First, consider the case I ′ Ī

= . Because e = H is chosen at the deviation stage, the

N − K

entrepreneur does not trade latent contracts. Furthermore, given the additional offers (I + i (.), R+ i (.))

she trades null contracts with each lender i ∈ {1, ..., K}.

Thus, following the deviation, the

entrepreneur trades the null contract with each non-deviating investor, which implies that I 0 =

Ī

N − K . When the entrepreneur chooses e = H, her payoff is bounded above by U(0) + (π HG −

Ī

1)(

N − K + Ī

I+′ (

N − K )), where Ī

I+′ ( ) ≥ 0. Under default, the entrepreneur’s payoff is

N − K

bounded below by

Ī

U sd = B(

N − K + Ī

I+′ (

N − K ) + N − K − 1

N − K Ī + KĪ+ i ( Ī

N − K ) − A) − A

= U ∗ (I ∗ , R ∗ Ī

, H) + BI +′ (

N − K ).

Thus, a sufficient condition for the entrepreneur to default is

U ∗ (I ∗ , R ∗ Ī

, H) − U(0) ≥ (π H G − 1 − B)I +′ (

N − K ) + (π Ī

HG − 1)

N − K ,

which corresponds to (A.8) established in the proof of Proposition 4.

53

Second, consider the case I ′ = 0. Again, because e = H is chosen at the deviation stage, the

entrepreneur does not trade latent contracts together with the deviating contract. Remark that we

must have I ′+ (I 0 ) > 0 for the deviation to be profitable. Furthermore, given the additional offers

(I + i (.), R+ i (.)), the entrepreneur raises the amount I 0 = I∗

K , or the amount I 0 = I ∗ .

If the entrepreneur chooses I 0 = I∗

together with e = H, she must trade one equilibrium

K

contract at the deviating stage. The inequality I ′+ ( I∗

K ) > 0 implies that IF ( I∗

K ) > I∗

. Thus, given

K

(.), the entrepreneur cannot get more than her reservation utility. Thus, when choosing e = H,

R + i

she prefers not to trade the deviating contract.

The same logic applies if the entrepreneur chooses I 0

= I ∗ together with e = H. In that

case she must trade an equilibrium contract with each investor i ∈ {1, ..., K}. The inequality

I ′+ (I ∗ ) > 0 implies that I F (I ∗ ) > I ∗ , **and**, given R + i (.), the entrepreneur cannot get more than her

reservation utility.

PROOF OF PROPOSITION 6 (continued): The proof extends to the case in which investors can

write covenants contingent on the initial debt ( I 0 . Take any aggregate allocation ) (I ∗ , R ∗ ) ≠ (I c , R c )

I

supported in a symmetric equilibrium. Let

N , R(I∗ )

N

, I+ (I)

N , R+ (I, I ∗ )

be the equilibrium

N

trades of each lender, with I + I + (I) = I ∗ .

Let investor k deviate to the menu M k ′ = ((0, 0, 0, 0), ( I N , R(I∗ )

N

, I+′ k

(.) + ε, R+′

k

(.))) with

ε > 0. The additional offer (I +′

k

(.), R+′ (.)) is such that

⎧

⎪⎨

I +′

k (I 0) =

⎪⎩

k

I ∗ − I N − N − 1 π H R ∗ − I ∗

N π H G − 1

if I 0 = I N ,

0 if I 0 ≠ I N ,

⎧

⎪⎨

R +′

k (I 0, I F (I 0 )) =

⎪⎩

R + (I, I ∗ )

N

G(I F (I 0 ) + A)

+ 1

π H

(I ′+ ( I N ) − I N

+ ε) + ε2 if

I 0 = I N **and** IF (I 0 ) = I N + I+′ k ( I N ) + ε,

otherwise.

The logic developed in the first part of the proof applies to this more general case. If ε = 0, the

pair ( I N + I+′ k ( I N ) + ε, R(I∗ )

N

+ R+′ k ( I N , I N + I+′ k

( I ) + ε)) is such that

N

U( I N + I+′ k ( I N ), R(I∗ )

N

+ R+′ k ( I N , I N + I+′ k ( I N )), H) = U(I∗ , R ∗ , H)

54

**and** π H R +′

k ( I N , I N +I+′ k

( I N ))−I+′ k

( I N +I+′ k ( I N )+ε, R(I∗ )

+R+′

( I N ) = π 1

H

N R+ (I, I ∗ )− 1 N I+ (I). Since (I ∗ , R ∗ ) ≠ (I c , R c ),

( I )+ε)) ∈ int(F) for ε small enough. The deviation

N

k

N

( I N , I N +I+′ k

is designed to induce the entrepreneur to trade( I N + I+′ k ( I N ) + ε, R(I∗ )

N

+ R+′ k ( I N , I N + I+′ k ( I N ) +

ε)), **and** to choose e = H. Given M k ′ , if the entrepreneur chooses e = H, she can get a payoff

above the equilibrium one by choosing ( I N +I+′ k ( I N )+ε, R(I∗ )

N

+R+′ k ( I N , I N +I+′ k

( I )+ε)) in the

N

deviating menu. In this case, however, given R k ′ (.), she finds optimal to trade with investor k alone.

Thus, as long as ε > 0, M k ′ is a profitable deviation for investor k whenever e = H is chosen. When

the entrepreneur chooses e = L, the only possibility to get a payoff above the equilibrium one is to

choose I 0 = I N . In this case, she trades the aggregate amount I N + I+′ k ( I N ) + ε + N − 1

N Î+ ( I N ),

where 1 N Î+ ( I ) denotes the largest additional investment offered in any equilibrium menu for

N

I 0 = I . Given the equilibrium schedules in (22), the deviation is hence profitable if

N

U(I ∗ , R ∗ , H) + (π H G − 1 − ε)ε + N − 1

N

(1 − δ)( I N + I+′ k ( I N ) + ε) >

B( I N + I+′ k ( I N ) + ε + N − 1

N Î+ ( I ) + A) − A. (A.25)

N

As in the first part of the proof, (A.25) is satisfied, since U(I ∗ , R ∗ , H) ≥ B( I N + N − 1

N Î+ ( I N ) +

A) − A by construction, π H G − 1 − ε > 0 for ε small enough, **and** δ < 1 − B. The rest of the

proof is a straightforward adaptation of the reasoning developed when I is only observed ex post.

55

FOR ONLINE PUBLICATION

ADDITIONAL APPENDIX B: Indeterminacy with N ≥ 2

√

πL

Proposition B.1 Take any N ≥ 2. If π H G − 1 > 2 , there exists an investment level I < I c

π H

such that any aggregate allocation (I ∗ , R ∗ ) ∈ F with I ∗ ≥ I can be supported at equilibrium.

PROOF: Assume

√

πL

π H G − 1 > 2 .

π H

(B.1)

We first establish a set of relationships that will be used throughout the proof. First, from (1) we

get

G − B ∆π > G + π LG − 1

∆π

where the last inequality follows from (B.1).

= π HG − 1

∆π

Second, given (B.1), (1) **and** (4) together imply that 37

Adding the two conditions, one gets

√

πL π H

1 > 2 ⇔

∆π

which yields

√

πH

π L

> 1 + √ 2,

> 2 √

πL

,

∆π π H

1 − π H

∆π B > 2 √

πL

π H

π L

∆π **and** π H

∆π B > 2 √

πL

π H

.

√

πH

π L

−

√

πL

π H

> 2 ⇔

(√ ) 2 πH

− 1 > 2,

π L

(B.2)

√

πL

π H

< √ 2 − 1 **and** π H > (3 + 2 √ 2)π L . (B.3)

We now turn to the proof of Proposition B.1. It is useful to characterize equilibrium allocations

in terms of two parameters, which we denote ε **and** η. Precisely, let (ε, η) ∈ [0, ε] ×

[ ( ]

πH G −

B

∆π)

, 1 , with

(

ε = 1 N min ∆π

B (G − 1 ), π HG − 1

−

π H B

π L

π H ∆π

1

G − B ∆π

,

(

1 − π L

∆π

1/π H

G − B ∆π

)

)

B

.

2B + (1 − π H G)

Observe that (1) **and** (4) imply that ∆π(G

− 1

B π H

) ∈ (0, 1), **and**, given (B.2), both the second **and**

third terms are strictly positive. Thus, we get 0 < Nε < 1. Consider now the aggregate allocation

(

)

(I ε , Rη) ε = I c 1

(1 − ε),

π H

I c (1 − ηε) . It is immediate to check that, if e = H, the aggregate profit

37 It is useful to rewrite (1) as π H G − 1 < ∆π

π L

(1 − π H

∆π B), **and** (4) as π HG − 1< π H

∆π B.

56

π H R ε η − I ε is strictly decreasing in η. See also that ∀ε ∈ [0, ε], I ε > I m . 38 Letting I = I ε > I m ,

**and** considering all (ε, η) ∈ [0, ε] × [ π H

(

G −

B

∆π)

, 1

]

, one can hence generate a closed subset of

F including all aggregate allocations (I ε , R ε η) ∈ F such that I ε ≥ I. The subset is represented in

the dashed area in Figure 3.

R

L

R c

R m

O

G − B ∆π

1

πH

C m

•

I m

F

Ψ

I

C c

•

I c H

I

Figure 3: Set of aggregate allocations (I ε , R ε η)

Consider now any allocation (I ε , R ε η) such that I ε ≥ I, **and** denote it (I ∗ , R ∗ ). We show that it is

supported at equilibrium by the following strategies, which are similar to those used in the proof of

Proposition 3. Each investor i = 1, 2, ..., N offers the same menu M ∗ = {(0, 0), ( I ∗

N , R∗ (.) ) , ( Î

N , ˆR(.) ) },

with R ∗ (I) = G(I + A) for I /∈ {I ∗ , I∗

}, **and**

⎧N ⎪⎨

R ∗

if I = I ∗ ,

R ∗ (I) = N

⎪⎩ R ∗ − N − 1

(B.4)

N GI∗ if I = I∗

N .

The investment level Î is such that

B(Î + A) − A = U(I∗ , R ∗ , H) = U c − εI c (π H G − η),

(B.5)

which guarantees, given that U(I ∗ , R ∗ , H) ≥ B(I ∗ + A) − A whenever (I ∗ , R ∗ ) ∈ F, that Î >

I ∗ ≥ 0.

As in the proof of Proposition 3, one can check that choosing the investment I∗

in each menu

N

**and** selecting e = H is an optimal choice for the entrepreneur. Consider now investors’ deviations.

38 Indeed, I

(1 c − ∆π (

G − 1 ))

= ∆π (

G − B )

A = I m **and** ε < ∆π (

G − 1 )

.

B π H B ∆π

B π H

57

Without loss of generality, any unilateral deviation can be represented by a menu M ′ = { ( I ∗

N +

I ′ , R∗

N + R′ (.) ) , (0, 0)}. A profitable deviation must necessarily induces e = H, 39 **and** one must

have that

R ′ (I) > 1

π H

I ′ ,

with I being the aggregate investment traded at the deviation stage, **and** that I ′ ≤ I c − I∗

N . If,

following the deviation, the entrepreneur chooses e = L **and** strategically defaults, then, given

(B.5), she gets

U sd = U(I ∗ , R ∗ , H) + B ( I ′ + I∗

N − Î )

.

N

Assume first that the entrepreneur only trades with the deviating investor. In this case, choosing

e = H yields her the payoff

π H

(

G(

I ∗

N + I′ + A) − R∗

N − R′ ( I∗

N + I′ ) ) + U(0)(N − 1) + U(I ∗ , R ∗ , H)

− A <

N

+ (π H G − 1)I ′ .

Since the entrepreneur’s equilibrium utility remains available at the deviation stage, the right-h**and**

side of the last inequality must be strictly greater than U(I ∗ , R ∗ , H), which implies that I ′ > 0.

The entrepreneur therefore finds optimal to strategically default if

N − 1( U(I ∗ , R ∗ , H) − U(0) ) − B N

N (Î − I∗ ) ≥ (π H G − 1 − B)I ′ . (B.6)

The left-h**and** side being increasing with N, a sufficient condition for (B.6) obtains with N = 2.

That is, after rearranging: B(I ∗ + A) − π H GA ≥ 2(π H G − 1 − B)I ′ or B(I c + A) − π H GA ≥

BεI c + 2(π H G − 1 − B)I ′ . See that B(I c + A) − π H GA = − π L

∆π B(Ic + A) + (π H G − 1)I c which

leads to the condition:

π H G − 1 ≥ f(I ′ , ε),

(B.7)

39 Indeed, given Lemma 1, a deviating investor may achieve a positive profit by inducing e = L only if the entrepreneur

trades several contracts out of equilibrium **and** does not default. Given equilibrium menus, this is only

possible if the entrepreneur invests I = I ∗ **and** selects e = L; this in turn provides her a payoff smaller than the

equilibrium one, which guarantees that this is not an optimal choice.

58

( )

with f(I ′ πL I c + A

, ε) = B

+ ε + 2(π

∆π I c

H G − 1 − B) I′

. Given the linearity of f, we have that

Ic (

( ))

f(I ′ I ′

, ε) = f

I c (Ic , 0) + (1 − I′

I ) ε

0,

c 1 − I′

I c

= I′

I c f(Ic , 0) + (1 − I′

I c )f (

0,

To prove (B.7), using 0 < I ′ ≤ N−1

N

Ic < I c , we simply need to show that

ε

1 − I′

I c )

.

π H G − 1 ≥ f(I c , 0), (B.8)

(

ε

π H G − 1 ≥ f 0, . (B.9)

1 − I′

(

)

(

)

Condition (B.8) is equivalent to B 2 − π L 1/π H

≥ π

∆π G − B H G−1, which holds since B 2 − π L 1/π H

>

∆π G − B

(

∆π

∆π

B 2 − 1 √ )

πL

> B(2 − 1 2 π H 2 (√ B

2 − 1)) >

1 − (3 − 2 √ 2) > π H

∆π B ≥ π HG − 1. The first inequality

comes from (B.2), the second **and** the fourth from (B.3) **and** the last one is (4). To prove

(B.9), first remark that (B.7) holds for any couple (0, ε) with ε ≤ Nε. Indeed, by definition of ε,

we have ε ≤ π HG − 1

− π L 1/π H

for any ε ≤ Nε. To complete the proof of (B.9), observe

B ∆π G − B ∆π

that, since I ′ ≤ N−1

N

Ic **and** ε ≤ ε, we have

I c )

ε

1 − I′

I c

≤ 1 N

Nε

1 − I′

I c

≤ 1 N

Nε

1 − N−1

N

= Nε.

Assume next that several contracts are traded at the deviation stage. Getting back to the proof

of Proposition 3, following any unilateral deviation, the entrepreneur strategically defaults if (17)

holds. Again, the left-h**and** side of (17) is increasing in k. A sufficient condition for (17) is

therefore, using (B.5) **and** I ∗ = I c (1 − ε)

BI c (

1 − π L

∆π

1/π H

G − B ∆π

)

≥ εI c (2B + 1 − π H G),

(

)

B

which is equivalent to ε ≤

1 − π L 1/π H

. The inequality holds by definition

of ε, which concludes the proof that (I ∗ , R ∗ ) is supported at equilibrium.

2B + (1 − π H G) ∆π G − B ∆π

59

FOR ONLINE PUBLICATION

ADDITIONAL APPENDIX C: Seniority under bankruptcy

Proposition C.1 For sufficiently small values of π L , if one investor is repaid first in case of default,

then every allocation characterized in Proposition 4 can be supported at equilibrium.

PROOF: Fix N such that the set of allocations satisfying (6) is non-empty. Consider any strategy

profile introduced in the proof of Proposition 4 with the equilibrium allocation (I ∗ , R ∗ ) **and** let j

be the investor whose claim is senior in case of default. We show that if π L is sufficiently small,

this investor cannot profitably deviate inducing strategic default. Let M ′ j

= {(0, 0), (I ′ j, R ′ j(.))}

be a deviating menu with R ′ j(I) = G(I + A) ∀I the repayment that maximizes investor j’s profit

in case of default. Suppose first that j ∈ {1, · · · , K}. Given equilibrium covenants, (I ′ j, R ′ j(.))

induces the entrepreneur to strategically default only if

U(I ∗ , R ∗ , H) = B(Ī + I∗ + A) − A < B(I j ′ + K − 1

K I∗ + Ī + A) − A,

which corresponds to

I ′ j > I∗

K .

(C.1)

A sufficient condition for the deviation not to be profitable is that it yields a strictly negative profit,

that is

(π L G − 1)I ′ j + π L G(Ī + K − 1

K I∗ + A) < 0. (C.2)

For a given N, it follows from Proposition 4 that I∗

> 0 at equilibrium. See that the strict

K

inequalities (C.1) **and** (C.2) are satisfied for π L = 0. By continuity, one can then construct an

open neighborhood of π ∗ L = 0 such that, for each π L in this interval, any I ′ j that satisfies (C.1) also

satisfies (C.2).

The same reasoning applies to the case j ∈ {K + 1, · · · , N}.

60

FOR ONLINE PUBLICATION

ADDITIONAL APPENDIX D: Reducing B under bankruptcy

We consider the following bankruptcy mechanism. Before the game is played, an authority

commits to reduce the entrepreneur’s private benefit to some B ′

observing R > G(I + A). We then have the following proposition.

= λB with λ ∈ (0, 1) after

Proposition D.1 Assume π L = 0 **and** consider an economy with a bankruptcy mechanism of parameter

λ. If

(

πH G − 1

λ > max , 1 − (π H G − B) π )

HG − 1

, (D.1)

B

B

there exists an investment level I λ such that any aggregate allocation (I ∗ , R ∗ ) ∈ F with I ≥ I λ

can be supported at equilibrium.

PROOF: Consider first any array of parameters (π H , π L , G, B) such that π L = 0 **and** (1) **and**

(4) are satisfied. Observe that, in this case, (4) can be written π HG − 1

< 1, which implies

B

1 − (π H G − B) π HG − 1

> 0. Construct then a modified array of parameters, by replacing B

B

with any B ′ = λB, with λ satisfying (D.1). We refer to these parameters as the economy E λ .

We denote F λ its feasibility set **and** Iλ

m (Ic λ ) the corresponding monopoly (competitive) investment

level. Observe that since G − λB

∆π > G − B , we have

∆π

F ⊂ F λ , I c ≤ I c λ. (D.2)

We know from Proposition 3 that any aggregate allocation (I ∗ , R ∗ ) ∈ F satisfying I ∗ ≥ I m can

be sustained at equilibrium. This implies that any (I ∗ , R ∗ ) ∈ F λ with I ∗ ≥ Iλ

m can be sustained at

equilibrium in E λ with the symmetric menus M λ = {(0, 0), ( I ∗

N , R∗ (.) ) , ( Î

N , ˆR(.) ) }, where R ∗ (.)

**and** ˆR(.) are defined as in the proof of Proposition 3, **and**

λB(Î + A) − A = U(I∗ , R ∗ , H).

(D.3)

Simple manipulations lead to I m λ

A. The inequality

(D.1) therefore implies that

+ A = π HG

λB A **and** Ic + A =

1/π H

1

− (G − B )

π H π H

I m λ < I c . (D.4)

61

Define I λ = I m λ

until the end of the proof.

We now turn to the original economy, identified by the parameters array (π H , π L = 0, G, B)

**and** suppose that a bankruptcy mechanism is introduced. Consider any (I ∗ , R ∗ ) ∈ F such that

I ∗ ≥ I λ . We show that the menu M λ defined above supports (I ∗ , R ∗ ) at equilibrium even in the

presence of a bankruptcy mechanism.

Observe first that when the entrepreneur chooses an investment different from I ∗ or I∗

N at

equilibrium, covenants ensure that she strategically defaults obtaining B ′ (I +A). 40 As in the proof

of Proposition 3, choosing I∗

in each menu **and** selecting e = H is an optimal choice for the

N

entrepreneur by (D.3). 41

Consider now investors’ deviations. As in the proof of Proposition 3, any unilateral deviation

can be represented by a menu M ′ = { ( I ∗

N + I′ , R∗

N + R′ (.) ) , (0, 0)} **and** must induce

the entrepreneur to choose e = H to be profitable. See that by (D.4) **and** (D.2), any allocation

(I ∗ , R ∗ ) ∈ F with I ∗ ≥ I λ is also supported at equilibrium with the menus M λ in the economy

E λ . This establishes that the same allocation is supported at equilibrium in the original economy

when a bankruptcy mechanism is in place.

40 That is, she cannot select e = L without triggering the bankruptcy mechanism.

41 If the entrepreneur accepts I∗

N in each menu **and** chooses e = L, we necessary have B(I∗ + A) − A ≤

U(I ∗ , R ∗ , H) since (I ∗ , R ∗ ) ∈ F.

62