The Design of Corporate Debt Structure and

Bankruptcy∗
Erik Berglöf† Gérard Roland‡
Ernst-Ludwig von Thadden§
This version: August 2009

Abstract
This paper integrates the problem of designing corporate bank-
ruptcy rules into a theory of optimal debt structure. We show that, in
an optimal contracting framework with imperfect renegotiation, hav-
ing multiple creditors increases a firm’s debt capacity while increasing
its incentives to default strategically. The optimal debt contract gives
creditors claims that are jointly inconsistent in case of default. Bank-
ruptcy rules are therefore a necessary part of the overall financing
contract, to make claims consistent and to prevent a value reducing
run for the assets of the firm. We characterize these rules, with pre-
dictions about the allocation of security rights, the right to trigger
bankruptcy and the symmetry of treatment of creditors in default.
Keywords: Bankruptcy, Debt Structure, Contracts.
JEL Classification: G3, K2.
∗
We are grateful to the Center for Advanced Studies in the Behavioral Sciences, Stan-
ford, for providing a stimulating environment for early discussions about the topic of this
paper. Thadden thanks the German National Research Foundation (DFG) for support.
We thank two anonymous referees for their detailed comments. We benefited from dis-
cussions and suggestions from Ron Anderson, Lucian Bebchuk, Bernard Black, Patrick
Bolton, Eric Maskin, Katharina Pistor, Howard Rosenthal, Yossi Spiegel, Javier Suarez,
Jean Tirole, and in particular, Oliver Hart, as well as from several seminar and conference
audiences.
†
Stockholm School of Economics, European Bank for Reconstruction and Development,
CEPR and WDI.
‡
University of California at Berkeley and CEPR
§
Universität Mannheim, ECGI, and CEPR, email: [email protected].

1
 Bankruptcy law regulates the interaction between debtors and creditors
when debtors default and the parties cannot work out their differences outside
the courts. The law addresses two main types of conflicts: conflicts between
a debtor and her creditors, and conflicts among creditors themselves. Empir-
ically, this latter type of conflict is the major source of complexity in modern
bankruptcy law, and has therefore given rise to a substantial literature, much
of which in law. This literature typically takes an ex-post perspective: how
to sort out claims once the firm is bankrupt, given the contractual and legal
arrangements in place. But the way ex-post conflicts are resolved also influ-
ences the initial financing and valuation of the firm, which suggests that the
two problems should be analyzed jointly. In fact, the problem of bankruptcy
is most interesting when posed in an ex-ante framework as it raises what
seems like a paradox: if bankruptcy with multiple creditors is so complex,
why would a firm contract with several creditors in the first place? Put dif-
ferently: if conflicts of interest must be resolved ex post anyhow and these
resolutions are costly, why create them and how structure them ex ante?
We attempt to answer these questions in an optimal contracting approach to
corporate debt and bankruptcy.
The paper analyzes a firm’s choice of debt structure and its effect on
incentives for strategic default. We start from the observation that multiple
creditors make contract renegotiations more difficult and emphasize a) the
ex-post conflicts among multiple creditors, b) the design of individual claims
and their impact on these conflicts and c) the role of bankruptcy rules in
resolving such conflicts from an ex ante perspective. In our model, having
multiple creditors gives rise to potential ex-post inefficiency, which stems
from frictions in multilateral conflict resolution. This corresponds to the
well-known inefficiencies in the workouts of financial distress, documented,
e.g., by Asquith, Gertner, and Scharfstein (1994). However, in the spirit of
the literature on strategic capital structure design,1 we note that this ex-post
inefficiency also has a positive incentive effect, as it forces the firm to honor
several claims instead of only one if it wants to avoid the ex-post inefficiency.
To provide an intuition for why two investors are better than one in our
model, consider a firm seeking outside finance from two investors. Following
Hart and Moore (1998), we describe the moral-hazard problem of the firm
in repaying its financiers by assuming that the project generates some un-
1
See, in particular, Hart and Moore (1998), Bolton and Scharfstein (1996), Dewatripont
and Tirole (1994), Berglöf and von Thadden (1994).

2
verifiable cash flows next to the verifiable assets. In this setup, repayment is
limited by how much asset value the financiers can credibly threaten to liq-
uidate. Under perfect renegotiation (or with one single creditor), the firm’s
commitment ability is in principle given by the amount of assets available
for foreclosure should the firm default. However, this constraint is relaxed if
the firm is forced to renegotiate individually with its creditors. If this is the
case, the firm can promise ex ante up to the full amount of available assets
to each one of the investors. When the firm only defaults on one investor ex
post, this investor has the right to foreclose on the firm’s assets to collect her
debt. As each creditor has this individual right, the firm must pay out twice
its asset value if it wants to protect its assets. Hence, a renegotiation with
two creditors credibly commits the firm to pay out twice in order to avoid
renegoatiation. These higher repayment obligations, however, also increase
the incentives for strategic default, in which case the inefficiency of the cred-
itor interaction ex post destroys value. This creates a counterveiling force
to the commitment effect of high individual debt claims and thus leads to a
tradeoff in the design of these claims.
If the firm defaults on both creditors, and one creditor calls the sheriff to
enforce the payment, the firm or the other creditor can file for bankruptcy. In
this case, the sum of the two claims are larger than the available amount of
verifiable assets, and individual claims must be adjusted. This is the essential
role of bankruptcy in our model. Bankruptcy provides a means to pare down
individual claims if the debtor is unable or unwilling to honor them and the
creditors must confiscate what is available. We thus emphasize the view,
shared by many scholars of law and economics, but rarely modeled in full,
that “bankruptcy is a situation in which existing claims are inconsistent ”
(Hart, 1995).
Our model is the first to bring out the distinction between debt collection
and bankruptcy that is a fundamental feature of debt finance in practice and
legal theory. Debt collection law is basically bilateral and defines the rights
of a single creditor in a bilateral conflict with the debtor. In our model,
each creditor has the right to collect his debt if the debtor defaults against
him, and as every creditor wields this threat, the debtor is under strong
pressure to pay out. Yet, if the debtor defaults against several creditors,
these threats cannot necessarily be executed simultaneously, individual debt
collection ceases, and the debtor is in bankruptcy. In this sense, we model
bankruptcy as “collective debt collection" (Jackson, 1986).
Next to the distinction between individual and collective debt collection,

3
our model yields the following empirical predictions.
The model predicts that debt claims should be “undersecured”, i.e., that
nominal debt claims should be larger than what creditors receive in bank-
ruptcy. While this is a standard consequence of single-creditor models, it
is not obvious in a multi-creditor model. In our model this result is true
because, holding total bankruptcy payout constant, too high a bankruptcy
payout for one creditor would give other creditors so little in bankruptcy
that the firm has incentives to strategically default against them (which is
not ex ante optimal, as we show). Our prediction is consistent with standard
practice and the empirically documented observation in many countries of
systematically low retrieval rates of creditors in bankruptcy (see, e.g., Weiss
(1990)).
Further, the model predicts that for a range of borrowing requirements,
the debtor should retain some of her assets in bankruptcy. In our model, this
is true because the debtor on average uses the firm’s assets more efficiently.
Hence, even if ex post the debtor has bad prospects, the ex ante allocation of
security rights should protect the debtor. As a corollary, absolute priority -
the notion that creditors must be satisfied fully in bankruptcy before owners
are to retain something - can be violated in the present model. Again, this is
consistent with the empirical literature (see, e.g., Franks and Torous (1989)
and Bharath, Panchapegesan, and Werner (2008)).
Next, the model predicts that all creditors should optimally be treated
symmetrically ex post, in the sense that either all creditors are repaid or
all are defaulted upon. In our model, this is true because unilateral foreclo-
sures by one creditor that impair the value of claims of other creditors do
not minimize expected liquidation and are therefore value-reducing ex ante.
This corresponds to, and gives an efficiency justification for, the widespread
use of “equal treatment” rules in bankruptcy legislation, such as the Trust
Indenture Act in the U.S.
Furthermore, our model allows to distinguish between liquidity-driven
defaults and value-driven defaults. In the equilibrium of the model, firms
default either because they lack liquidity but are fundamentally sound or
because they have liquidity but little long-term value. In our model, value-
driven default can occur in equilibrium because it can be optimal ex ante to
contract high face values of debt (to extract much cash on average in order to
keep overall liquidation low), which in turn makes strategic default optimal
in some ex-post contingencies. Davydenko (2005) provides evidence that
empirically both motives are important and have independent explanatory

4
power.
Our simple model of bankruptcy captures some important elements of
existing bankruptcy procedures. We show that each creditor’s right to liq-
uidate assets, which protects him against opportunism by the debtor, must
be complemented by the right to trigger bankruptcy, which in turn limits
the individual liquidation rights. Bankruptcy is triggered when a creditor
files to prevent his claims from being eroded through debt collection of other
creditors. The procedure demands an “automatic stay” ensuring that liqui-
dation claims are executed simultaneously, and distributes liquidation values
according to a pre-specified rule. Interestingly, however, giving the creditors
the right to trigger bankruptcy is not always sufficient to rule out runs for the
assets. If individual creditors stand to gain relatively little from bankruptcy,
but recover their full debt claim if they are the first to foreclose individually,
they may have an incentive not to trigger bankruptcy ex post, but rather run
for the assets. In such a constellation it is optimal to give the debtor the right
to trigger bankruptcy. In this respect, our model provides a new efficiency
argument for debtor-friendly bankruptcy legislation (see Baird, 1991).
The remainder of this paper is organized as follows. In Section 1, we
review some of the literature on debt structure and bankruptcy. Section 2
describes the basic structure of the model. Section 3 studies optimal con-
tracting under the assumption that bankruptcy is triggered automatically.
Section 4 extends the model and provides a more detailed institutional analy-
sis of the bankruptcy process, in particular the right to trigger bankruptcy
and equal treatment rules. Section 5 concludes with a discussion of the role
of contracts versus the law and further research avenues. Some proofs are
contained in the appendix.

1 Related Literature
Our paper touches on two large strands of the literature that up to now have
rarely been brought together: the literature on corporate bankruptcy and on
capital structure. Here we review some contributions to both strands that
are relevant to our analysis.
An important part of the large literature on bankruptcy law focuses on
optimal procedural and substantial rules, taking as given pre-existing debt
contracts and the decision to enter bankruptcy (next to the large legal lit-
erature, see, e.g., Bebchuk (1988), Aghion, Hart and Moore (1992) or Baird

5
and Bernstein (2005)). This work rightly points out that the choice of debt
structure influences what happens and what should happen in bankruptcy.
Yet, it is silent on the determinants of debt structure, which is problematic as
the bankruptcy procedure has an impact on the firm’s capital structure de-
cision. An important exception is the work by Harris and Raviv (1995) who
study the impact of different games played ex post on the ex-ante efficiency
of the contract. Different from our work, however, Harris and Raviv (1995)
are only concerned with games between the debtor and one single investor.
Another interesting avenue of research has looked at the bankruptcy prob-
lem from an ex-ante perspective. Contributions such as those by Bebchuk
(2002), Berkovitch and Israel (1999), Cornelli and Felli (1997), Schwartz
(1998) or Acharya, John, and Sundaram (2010) analyse the impact of bank-
ruptcy on debtors’ investments, debt levels, and incentives prior to bank-
ruptcy. These ex-ante analyses are not concerned with the key question of
our paper, which is the role of multiple creditors in bankruptcy. In fact, all
of these articles consider the conflict between a debtor and a single represen-
tative creditor.
Conceptually, our work is closest to the legal literature on bankruptcy,
which emphasizes that the main role of bankruptcy law “is that of bankruptcy
as a collective debt-collection device, and it deals with the rights of credi-
tors ... inter se" (Jackson, 1986). For example, Kordana and Posner (1999)
study the complex voting features associated with Chapter 11 in the U.S.
bankruptcy code. Similar to our paper, they discuss the tradeoff between
reducing the cost of liquidation by lowering individual pre-bankruptcy enti-
tlements and discouraging strategic default. However, like most of the legal
literature, they do not formally model the full ex-ante contracting problem,
which makes it difficult to evaluate the effects they are discussing.
There are only very few papers that have modelled multiple creditor prob-
lems from an ex-ante perspective. Of interest in our context are, in particular,
the contributions by Winton (1995), Bolton and Scharfstein (1996), Bris and
Welch (2005), Hege and Mella-Barral (2005), Bisin and Rampini (2006), and
Hackbarth, Hennessy, and Leland (2007).
In Bolton and Scharfstein’s (1996) seminal theory of debt structure, bor-
rowing from multiple sources can be optimal, but need not. In their model,
multiple (two) creditors increase firm value on the one hand because of in-
creased bargaining pressure in strategic default, but decrease firm value on
the other hand because of less efficient continuation choices in liquidity de-
fault. The optimal number of creditors emerges as a trade-off between these

6
two tendencies. Their analysis does not consider the key problem of our pa-
per, the ex-post tension between individual and collective liquidation rights
of creditors. As a consequence, they are not concerned with the optimal allo-
cation of individual collection and security rights and their impact on default
and bankruptcy.
Bris and Welch (2005) take a different approach to the coordination prob-
lem between multiple creditors in ex-post bargaining. They note that such
bargaining is wasteful because of lobbying and other dead-weight losses, but
that free-riding reduces the incentives for such wasteful activities when the
number of creditors increases. Hence, although payments to creditors in fi-
nancial distress decrease with the number of creditors, this is priced in the
debt claims ex ante and only the beneficial effect of reduced influence costs
ex post matters, which creates a tendency to contract with multiple credi-
tors. Differently from our theory, Bris and Welch (2005) ignore the problem
of strategic default, by simply assuming that the debtor pays out in good
states of nature. Our theory is consistent with theirs in that we emphasize
the advantage of contracting with multiple creditors, but we differ from their
theory by pointing out the downside of multiple creditors when it comes to
strategic default. Our theory of the design of individual debt claims and the
role of bankruptcy is driven by the tension between collective and individual
creditor rights, which is absent in their model.
Winton (1995) approaches the problem of multiple creditors from the per-
spective of costly state verification, generalizing the work of Townsend (1979)
and Gale and Hellwig (1985). His results provide a theoretical rationale for
seniority and absolute priority, and predict an ordering of monitoring activ-
ities among investors. These monitoring activities are reactions to financial
distress and can therefore be interpreted as gradual bankruptcy provisions.
Different from our work, in Winton (1995) the debtor borrows from several
creditors by assumption.2
Bisin and Rampini (2006) are interested in the ex-ante incentive effects
of bankruptcy in an environment in which a debtor can borrow from several
lenders to smooth consumption. They show that a bankruptcy-like contract
allows the main lender to relax the debtor’s incentive compatibility con-
straint, because it is a means for the main lender to commit to confiscate
2
It is actually interesting to note that Winton (1995) himself plays down the link of
his theory to bankruptcy. He rather stresses examples such as asset securitization or
reinsurance, where the individual verification activities do not necessarily imply that a
joint “bankruptcy” decision is negotiated and implemented.

7
returns in low-return states (which is not optimal for consumption smoothing
reasons, but increases the borrower’s effort incentives). Different from our
model, in their model lending is for consumption smoothing and not produc-
tion (so the focus is rather on consumer bankruptcy), and exclusive lending
contracts are superior to contracts with multiple creditors, but cannot be
enforced by assumption.
Hege and Mella-Barral (2005) and Hackbarth, Hennessy and Leland (2007)
consider continuous-time models of debt renegotiation with multipe credi-
tors. Like in our model, equity can make opportunistic debt exchange offers
to force concessions on coupon payments. Different from our paper, however,
Hege and Mella-Barral (2005) assume that all debt claims are ex-ante iden-
tical and do not study the design of individual claims and its relationship
with bankruptcy. Hackbarth, Hennessy, and Leland (2007) extend the work
by Hege and Mella-Barral (2005) by assuming that there are two types of
debt, market debt and bank debt. Bank debt can be costlessly and efficiently
renegotiated, while market debt cannot be renegotiated at all. Ideally, firms
would only contract bank debt, but that claim is limited by the firm’s col-
lateral value. In order to increase debt-capacity, firms must take out market
debt. The paper shows in particular, that optimally both types of debt coex-
ist. Thus, Hackbarth, Hennessy, and Leland (2007) study how exogenously
different types of debt coexist, while our paper makes no a priori assumptions
about debt characteristics and derives the differentiation of individual claims
and their treatment in bankruptcy endogenously.

2 The Model
A firm can invest a fixed amount I at date 0 and lives for two periods after
that date. As in the “incomplete contracts” literature on debt contracts
following Hart and Moore (1998), at date 1, the firm has assets in place
worth A that generate a cash flow Y . Asset value A at date 1 is verifiable and
deterministic, known to everybody in advance. Cash flow, Y , is observable,
but not verifiable, and accrues to the firm’s management. The difference
between A and Y is that foreclosure on the firm’s property by the sheriff can
only reach A, not Y . Hence, while contractual transfers of assets in place
can be enforced by courts, transfers of cash must be incentive-compatible for
the firm.
If the firm is not liquidated at date 1, final firm value V is realized at

8
date 2, where V is a continuous random variable with cumulative distribution
function F (V ) and support [V , V ]. We assume that F is differentiable on
(V , V ) with density f , and extend F to F (V ) = 0 on [0, V ] if V > 0.3 In
this paper we assume for simplicity that all of V is non-verifiable, i.e. that
management cannot credibly promise to transfer value to creditors at date 2.
Hence, we focus on short-term debt and ignore issues such as debt maturity
structure and debt rescheduling.4
At date 1, there is a public signal v about V . To simplify the presentation
(no loss of generality here), we assume that the signal is perfect, i.e. that
V is known already at date 1. At date 1, assets in place can be liquidated.
If B ≤ A is the total amount liquidated, the firm continues on the scale
(1 − B/A). This means that the firm and its owners obtain (1 − B/A)V at
date 2. This assumption amounts to assuming that long-term firm value is
produced with constant returns to scale.5
Interest rates across periods are normalized to 0. In order to simplify
the presentation, we assume that it is always inefficient to liquidate assets in
place:
V ≥ A. (1)
This assumption describes the more interesting case of the bankruptcy
problem: although this is inefficient ex post, the parties may be forced to
liquidate assets in order to motivate the debtor to repay. All our results hold
in the more general case of arbitrary V ≥ 0. (see the appendix).
Short term cash flow Y , realized at date 1, is given by
½
0 with probability 1 − q
Y =
YH with probability q.
Here Y = 0 represents the situation in which the firm is liquidity-constrained,
and Y = YH the case of no liquidity constraints. To simplify, we assume that
cash flows in the the good state are sufficiently high so as to avoid any liq-
uidity constraints in that state. For reasons that will become clear soon (we
focus on the case of two creditors), we impose
YH ≥ 2A (2)
3
Hence, management can always claim at date 2 that it has nothing to pay out (although
this is a zero-probability event, it can occur).
4
See Gertner and Scharfstein (1991) and Berglöf and von Thadden (1994) for models
that address some of these issues.
5
As for example in Hart and Moore (1998) and Harris and Raviv (1995).

9
 At date 0, the firm is run by a risk-neutral owner/manager who has no
funds and raises them from external investors.
P Investors are risk-neutral and
competitive, they each put up Ii > 0, Ii ≥ I, i = 1, ..., n. We shall focus
here on the case of two creditors. We discuss in the conclusion how the
analysis extends to n creditors.
In order to capture more of the richness of multilateral debt contracts
in practice, we go beyond the basic incentive-efficient contracting framework
and allow for renegotiation. We model renegotiation as a simple bilateral
take-it-or-leave-it offer game between the firm and each creditor, a set up that
corresponds to a simple form of exchange offers.6 The fact that bargaining
is bilateral represents the frictions that are inherent in typical out-of-court
bargaining and means that the creditors as a group cannot get together
with the firm to negotiate their way efficiently around bankruptcy. There is
substantial empirical evidence on the difficulty of out-of-court agreements.
Among others, Gilson, John, and Lang (1990), and Asquith, Gertner, and
Scharfstein (1994) find that out-of-court restructurings have a substantial
risk of failure, and this risk is the higher, the larger the number of creditors,
in particular public debtholders. Gilson (1997) argues that the relatively
high transactions costs of private restructurings that he finds (compared to
Chapter 11 cases) are the result of unanimity requirements and coordination
problems in out-of-court settlements.
At date 0, the firm and the creditors sign a contract and investor i pays Ii
to the firm. The latter promises a repayment of Ri to investor i. The contract
contains an individual foreclosure right Di ≤ A, which can be collected if
only creditor i forecloses, and individual claims Bi under collective debt
collection, with B1 + B2 ≤ A. Apart from these constraints, we place no
a priori restriction on Bi and Di . In particular, we do not restrict Di and Bi
to be equal: investor claims under individual foreclosure and under collective
debt collection can differ.
At date 1, nature determines Y and V . Since we allow for renegotiation,
the firm then offers each creditor separately to pay pi ≤ Ri , i = 1, 2. If
pi = Ri , creditor i has no more claim against the firm. If pi < Ri for one
or both creditors, these creditors simultaneously and separately choose to
accept the payment (strategy a) or to foreclose on the firm’s assets (strategy
f). If only one creditor accepts the firm’s cash offer and the other forecloses,
6
See Detriagache and Garella (1996) for an ex-post analysis of exchange offers (that
does not address the ex-ante problems we discuss here).

10
the latter receives Di whereas if both foreclose, they each receive Bi .
This model of individual versus collective liquidation rights adds a new,
more realistic dimension to the contracting problem. In practice, debt con-
tracts typically specify individual, non-interactive collection rights, which
are governed by debt collection law, and are less specific concerning collec-
tive debt collection. This problem is addressed by covenants, priority rules
or collateral assignments, but many interactive collection rights of creditors
are left to bankruptcy law and judges, as a way to implement multilateral
debt collection. Our approach to modelling these decisions in this and the
next section is to ask what a contract would optimally stipulate if it included
complete provisions for multilateral debt collection and if these provisions
were executed automatically as soon as both creditors decide to foreclose
on the firm’s assets. Such multilateral debt collection therefore is akin to
bankruptcy. In Section 4, we will investigate the bankruptcy problem more
deeply by asking how such multilateral debt collection can be implemented
if it does not happen automatically.
A contract in this renegotiation framework thus is given by eight non-
negative numbers (I1, R1 , D1 , B1 , I2, R2 , D2 , B2 ) with I1 + I2 ≥ I, Di ≤ A,
B1 + B2 ≤ A, R1 + R2 ≤ YH . Let D = D1 + D2 and B = B1 + B2 denote
aggregate claims. Furthermore, we normalize notation by assuming that
creditor 1 is the one with a (weakly) higher individual foreclosure right:

D1 ≥ D/2 ≥ D2 . (3)

Note that, once (3) is fixed, a further such normalization is not possible
with respect to the collective liquidation claims Bi .
Date-1 outcomes are payments by the firm of (0, 0), resp. (R1 , R2 ) in
the two cash-flow states if there is no renegotiation, and are given by the
following payment matrix if there is renegotiation:

Creditor 2
a f
(4)
a p1 , p2 p1 , D2
Creditor 1
f D1 , p2 B1 , B2

Contracts are chosen to maximize the firm’s expected payoff subject to

the investors’ participation at date 0. Once a contract is in place, the firm
and the investors behave according to subgame-perfect Nash equilibrium,
taking the contract as given.

11
3 Optimal Contracts
We first analyze the interaction at date 1 for any given contract and then
study the choice of contract at date 0, assuming that the contracting parties
anticipate how the contract influences their behavior later on. We maintain
the assumption that the liquidation outcome (B1 , B2 ) obtains automatically
if both creditors want to foreclose simultaneously in game (4). In Section 4,
we investigate what game structure can achieve this ex-post outcome.

3.1 Date 1 interaction:

Consider first the low cash-flow state, Y = 0. The firm has nothing to pay
out, and whether or not the firm attempts to renegotiate, the payments to
the creditors will be (0, 0) and liquidation (B1 , B2 ) will ensue.
Now consider the case Y = YH . If the firm renegotiates with both credi-
tors, the outcome of the foreclosure game between the two creditors depends
on the offered payments. We make the (standard) assumption that creditors
accept the firm’s payment whenever it is weakly greater than their liquida-
tion return. A sure way for the firm to induce the outcome (f, f ) is to offer
p1 = p2 = 0. To prevent foreclosure by both creditors it is necessary that
p1 ≥ D1 and p2 ≥ D2 . If D1 ≥ B1 or D2 ≥ B2 then the firm does not
have to pay more than Di to each creditor to obtain acceptance from both
creditors. In this case, setting pi = Di , i = 1, 2, achieves (a, a) as the unique
equilibrium of the foreclosure game, and this is the cheapest way to obtain
this outcome.
However, if D1 < B1 and D2 < B2 then setting pi = Di , i = 1, 2, leads
to two equilibria in game (4), (a, a) and (f, f). This multiplicity makes it
difficult to evaluate contracts ex ante, since their value depends on what
equilibrium will be played ex post.7 To simplify the analysis, we assume here
that creditors play the (a, a) equilibrium when being offered the payments
pi = Di , i = 1, 2, regardless of the contract in place.8
7
For a similar problem in the context of debt contracts under asymmetric information,
see Gale and Hellwig (1989).
8
If the creditors will play the (f, f )-equilibrium when being offered pi = Di , the firm
must pay pi = Bi to at least one creditor in order to induce (a, a). The cheapest way to
achieve (a, a) as the unique equilibrium outcome then is to set pi = Bi and pj = Dj , j 6= i,
where Bi + Dj ≤ Bj + Di . This yields a payoff of YH − Bi − Dj + V . This payoff is lower
than the payoff from offering pi = Di to both creditors, but guarantees that there is no

12
 By a similar reasoning, the firm can induce the asymmetric outcome (a, f)
as a unique equilibrium by offering p1 = B1 and p2 = 0, and analogously
induce (f, a) by offering p1 = 0 and p2 = B2 . Here, the firm defaults and
treats its creditors asymmetrically: it does not pay one creditor and has
him collect his debt, and it pays the other creditor just enough to prevent
him from sending the firm into bankruptcy. Note that such asymmetry of
treatment between the two creditors is typically illegal in most jurisdictions.
However, as we construct the interaction from first principles, we cannot rule
out such behaviour. Rather, we shall show that optimality considerations
will rule out such asymmetries.
Going back one stage in the bargaining game, which of the four bargaining
outcomes in matrix (4) does the debtor want to induce when renegotiating at
date 1?9 Clearly, this will depend on the parameters of the original contract
and on the realization of V . As an example that we will continue below, the
asymmetric default (a, f) is optimal if the firm’s payoff from it, YH − B1 +
(1 − D2 /A)V , exceeds the payoffs from the other three alternatives, i.e. if

YH − B1 + (1 − D2 /A)V > YH − D + V
YH − B1 + (1 − D2 /A)V > YH + (1 − B/A)V
YH − B1 + (1 − D2 /A)V > YH − B2 + (1 − D1 /A)V

These conditions are equivalent to

V D − B1
< (5)
A D − D1
V B1
> (6)
A B − D + D1
V 2B1 − B
> (7)
A 2D1 − D
plus the condition that D1 > D − B.10 Hence, if these conditions are com-
patible, there is an intermediate range for V such that asymmetric default
with liquidation by creditor 2 is optimal.
liquidation.
9
Of course, at date 0 management prefers the (a, a) outcome because liquidation is
inefficient. But at date 1, its preferences are guided by the Di , Bi , and no longer by
overall efficiency considerations.
10
If D1 ≤ D − B it is easy to see that the payoff from (a, f ) must be smaller than the
payoff from (f, f ).

13
 Since the precise conditions under which the different choices are ex-post
optimal are of little interest, we relegate their description to the appendix.
The above discussion shows that if the firm pays out to both creditors
then it will not pay more than (D1 , D2 ). Hence, without loss of generality we
can assume that the contract will not be renegotiated in this case and that
(R1 , R2 ) = (D1 , D2 ). This is consistent with our earlier definition of D1 and
D2 as individual foreclosure rights, because in all jurisdictions we know of,
an individual creditor has the right to sue the debtor for the full amount of
her due debt, unless the firm seeks bankruptcy protection.

3.2 Date 0 contracting:

3.2.1 Individual debt claims:
Contract design at date 0 consists of choosing the optimal individual claims
Di (the face values) and bankruptcy claims Bi . A first important observation
is that it is never optimal to induce default and liquidation by only creditor.

Proposition 1 (Individual Debt Design): Optimal contracts satisfy

D D2 D
D1 + D − ≤ B1 ≤ D1 − D + B (8)
B B B
This implies that in equilibrium there is no asymmetric default.

The proof of Proposition 1 is given in the appendix. It states that

B1 , B2, D1 , D2 must not be chosen such that the firm will want to default
asymmetrically ex post, i.e. such that the firm induces the outcomes (a, f) or
(f, a) in the renegotiation game. To see this, note that the first inequality in
(8) implies that the two conditions (5) and (6) above are incompatible, hence
that there is no value V for which the firm wants to default asymmetrically
with liquidation by creditor 2. The second inequality in (8) analogously rules
out the other asymmetric default.
The intuition for Proposition 1 is that while asymmetric default can be
optimal for the debtor ex post, it is not optimal ex ante. From an ex-ante per-
spective, liquidation is a necessary evil in order to give the debtor incentives
to repay ex post. But of course, liquidation destroys value. It is unavoidable
if the debtor is liquidity constrained (Y = 0), and in this case assets worth
B are liquidated. Therefore, the classical tradeoff between lowering B to

14
reduce inefficiency in the bad state and increasing B to improve incentives
to repay in the good state is unavoidable also in this context. But partial
default can only occur in the good state, and in this state default, which is
strategic, should be minimized. Indeed, making the debtor repay Di in cash
is ex ante more efficient than having her deliver Di through liquidation. The
key insight is that this latter outcome can be avoided through the proper
design of individual claims, namely by choosing individual bankrupty claims
in the range defined in (8). The first inequality of (8) shows, in particular,
that if creditor 1 has a high share of nominal debt (high D1 , holding total
debt D fixed), then his bankruptcy claim B1 cannot be too low, either. The
reason is that the contract must mute the debtor’s ex-post incentive to have
creditor 2 foreclose (on the relatively small D2 ) and avoid bankruptcy by
paying B1 to creditor 1 in cash. Proposition 1 shows that this is possible for
all ex-post realizations of long-term value V .
Note that Proposition 1 rules out the case D1 < B1 and D2 < B2 . Hence,
the optimal contract induces a unique equilibrium in the foreclosure game at
date 1.11
Proposition 1 is of interest for several reasons. First, it shows that opti-
mal contracts treat creditors equally ex post in the sense that they do not
create a situation where the debtor will want to induce unilateral foreclo-
sure by one creditor as a means to renegotiate downward the payment to the
other creditor. Under the optimal contract, either both creditors are repaid,
or both are defaulted upon. Note, however, that condition (8) is compati-
ble with heterogeneity in individual bankruptcy recovery ratios, Bi /Di . The
symmetry induced by the optimal contract is in line with the many provi-
sions in different bankruptcy codes that prohibit the unequal treatment of
11
Remember that we have assumed that creditors play the (a, a) equilibrium when being
offered the payments pi = Di , i = 1, 2. As noted in footnote 8, if creditors will play the
(f, f )-equilibrium when the contract stipulates D1 < B1 and D2 < B2 and the creditors
are being offered pi = Di , then the firm must pay pi = Bi to at least one creditor to induce
(a, a). This would change the payoffs obtained under outcome (a, a) to YH − Bi − Dj + V
and therefore invalidate the argument in the proof of Proposition 1 (which uses the payoff
YH − D + V for outcome (a, a)). If this is the behavior for contracts with D1 < B1 and
D2 < B2 then it can be shown along similar lines to those in the proof of Proposition 1
that it is not optimal to set D1 < B1 and D2 < B2 in the first place.
Hence, no matter how the problem of multiple equilibria is solved - through the exoge-
nous selection of one equilibrium as done here, or by assuming the worst-case scenario for
the debtor, which makes him overpay - the debt structure identified in Proposition 1 is
the optimal one.

15
creditors.12 What is interesting is that the traditional justifications against
unequal treatment are ethical: it is deemed unjust to treat similar creditors
differently. Our model justifies such behavior on efficiency grounds: unequal
treatment ex-post provides a way to pay off one creditor cheaply, which re-
laxes the ex-ante discipline of the liquidation threat.13
Second, the proposition shows that optimally, individual creditor claims
in bankruptcy are bounded below and above. For small values of individual
nominal debt Di , the lower bound in the proposition may not bind, hence
in this case optimality alone does not require debt claims to be secured in
bankruptcy. Yet, if D1 is sufficiently large (creditor 1 being the larger one
by normalization), creditor 1’s bankruptcy claim must be positive, and if
D1 approaches D it must be almost fully secured (the lower bound for B1
approaches D1 ).
The third point of interest of Proposition 1 is that it shows that optimal
contracts only depend on the aggregate claims B and D, as long as individual
claims respect the constraint (8). In fact, if partial liquidation is eliminated
by ex-ante design, then the firm will ex post either repay fully or go bankrupt,
and this decision, to the extent that it is strategic, depends only on B and
D.
It can be easily shown that for any level of B and D, there are many
combinations of Bi ’s and Di ’s that satisfy D/2 ≤ D1 ≤ D, 0 ≤ B1 ≤ B and
(8). Hence, the optimal contract is not unique in terms of individual claims
(D1 , D2 , B1 , B2 ).14

3.2.2 Aggregate debt claims and debt capacity

In order to characterize the optimal aggregate claims, note that in outcome
(a, a) the debtor’s payoff is
YH − D + V, (9)
12
The par conditio omnium creditorum (equal treatment of all creditors) in Roman law
is the mother of these clauses.
13
The liquidation transfer Di made to the liquidating creditor i can be viewed as a
fraudulent conveyance, ruled out in most bankruptcy codes. A payment to one creditor
is a fraudulent conveyance if it reduces the assets available to other creditors of equal or
higher seniority in bankruptcy.
14
Hence, our model leaves room for investor heterogeneity in the design of individual
debt claims. This is in interesting extension that we do not model explicitly. We discuss
some possible exogenous restrictions on individual claims, such as equal treatment in
bankruptcy, in Section 4.

16
while the payoff under (f, f) is
B
YH + (1 − )V. (10)
A
Comparing (9) and (10) shows that the debtor prefers (a, a) over (f, f)
iff
D V
≤ . (11)
B A
In words, if the firm is not liquidity constrained it prefers full repayment
if the continuation value V is higher than the threshold DA/B and prefers
complete strategic default otherwise. The firm gets (1 − B/A)V in the latter
(low cash flow) state and either YH −D+V (if V ≥ DA/B) or YH +(1−B/A)V
(if V < DA/B) in the good cash flow state. Letting
D
θ = Prob (V ≥ A), (12)
B
the firm’s expected profit at date 0 is, therefore,
¯
B ¯ D
S0 = (1 − q)(1 − )EV + qθ(YH − D + E[V ¯¯V ≥ A]) (13)
A B
¯
B ¯ D
+q(1 − θ)(YH + (1 − )E[V ¯¯V < A])
A B
¯
EV B ¯ D
= qYH + EV − qθD − (1 − q) B − q(1 − θ) E[V ¯¯V < A](14)
.
A A B
The last two terms of (14) show the two sources of inefficiency in the
contracting problem. First, there is liquidation in the bad cash flow state
(occuring with probability 1 − q), and second there is liquidation in the good
cash flow state following strategic default (occuring with probability q(1−θ)).
The first inefficiency is minimized by making bankruptcy liquidation B as
small as possible. Yet, this tends to aggravate the second inefficiency, because
it increases the threshold for strategic default in the last term of (14), thus
making (ex ante) wasteful strategic default more likely. The optimal contract
must strike a balance between these two payouts and the average cash payout
qθD, which a priori has no negative efficiency effects but must be supported
by liquidation threats to be effective.
The investors’ participation constraints are
(1 − q)Bi + qθDi + q(1 − θ)Bi = Ii , i = 1, 2, (15)

17
where the Ii are, in fact, defined by (15), and I1 + I2 ≥ I. Furthermore, the
feasibility constraints

0 ≤ D1 , D2 , B1 , B2 , B ≤ A (16)

and the constraints (8) and D1 ≥ D/2 must hold.

Summing (15), the investors’ aggregate participation constraint is

(1 − qθ)B + qθD ≥ I. (17)

Clearly, the participation constraint binds, and total borrowing is equal

to investment. Hence, the contract optimization problem at date 0 is

max S0 (18)
D,B
s.t. (1 − qθ)B + qθD = I (19)
0≤B≤A (20)
0 ≤ D ≤ 2A. (21)

Here, (20) is the constraint on bankruptcy liquidation and (21) the con-
straint on cash payout.
The left-hand side of (19) is the equilibrium repayment to investors. Its
maximum is the firm’s debt capacity, i.e. the maximum money the firm can
raise such that the above problem has a solution. Clearly, the firm can raise
any amount I ≤ A by setting B = D = I. In this case the firm has no
incentive for strategic default (θ = 1) and its repayment is I for sure: with
probability q in cash, with probability 1 − q through bankruptcy liquidation.
The interesting question is whether the firm’s debt capacity can be strictly
greater than its asset base.
The following estimate provides an upper bound for the debt capacity:

max(1 − qθ)B + qθD ≤ (1 − qθ)A + 2qθA (22)

B,D
≤ (1 + q)A (23)

This is intuitive because (1 + q)A = (1 − q)A + q2A: the creditors can

never get more than all assets in the bad state and cash of double that value
in the good state. But (12) shows that this reasoning actually gives the
exact debt capacity if V ≥ 2A, i.e. if long-term firm value is always very
high. In this case, increasing B and D all the way up to their maximum

18
values maximises the payments to investors without creating incentives for
strategic default for the debtor. On the other hand, if V < 2A, the debt
capacity is lower than (1 + q)A because the incentive for strategic default at
low values of V provides a countervailing effect.
Note also that if financing needs are large, individual claims must be
similar: in the limit, if I = (1 + q)A, then necessarily D1 = D2 = D/2.
In order to analyze the contracting problem in greater depth, we introduce
the variable t = D/B, the inverse of the aggregate bankruptcy recovery rate.
By making appropriate substitutions and rearrangements, remembering that
θ = 1−F (At), we can rewrite the contract optimization problem at date 0 as
the following constrained payout minimization problem in one real variable:
min H(ϕ(t), t) (24)
t
s.t. t≥0 (25)
ϕ(t) ≤ A (26)
tϕ(t) ≤ 2A (27)
where
∙ Z At ¸
H(B, t) = (1 − q)EV + q(1 − F (At))At + q V dF (V ) B (28)
0

corresponds to the expected payout by the firm (which it wants to minimize)

and
I
ϕ(t) = B = (29)
1 + q(1 − F (At))(t − 1)
uses the investors’ participation constraint (19) to express bankruptcy liqui-
dation B by the new variable t. In the new problem, (26) corresponds to (20),
and (27) to (21). The auxiliary function ϕ is defined for t ≥ 0, continuous,
and is differentiable for t > 0 except at t = V /A where it has a kink.
To simplify the presentation we assume from now on that the distribution
of V satisfies the Monotone Hazard Rate Condition (MHRC):
f (V )
Assumption (MHRC): The hazard rate of V , 1−F (V )
, is non-decreasing.

(MHRC) is a frequent assumption in contracting problems that is met by

most standard distributions. It simplifies the analysis of the present problem
because it implies that the auxiliary function ϕ has a unique minimum. The
following lemma is proved in the appendix.

19
Lemma 1: Under Assumption (MHRC),

(a) ϕ has exactly one minimum t0 .

(b) If f (V )(V − A) ≥ 1, then t0 = V /A. Otherwise t0 > V /A and ϕ0 (t0 ) =

0.

(c) t0 > 1.

(d) The function tϕ(t) is strictly increasing for all t > 0.

Comparing the two constraints of the optimization problem (24) - (27), we

have tϕ(t)/2 < ϕ(t) ⇔ t < 2. Hence, constraint (27) on D is redundant if t <
2, and constraint (26) on B is redundant if t > 2. Constraint (25) is always
redundant. Since the constraint set defined by (25) - (27) is compact and
the objective function is continuous, the contracting problem has a solution
if and only if constraints (26) and (27) are compatible. These constraints
have the investment level I on their left-hand side, hence they characterize
the maximum investment levels that can be financed.

Proposition 2 (Debt Capacity): The contracting problem has a solution

if and only if
ϕ(min(2, t0 )) ≤ A (30)
where t0 is the unique minimum of ϕ.

Proof: If t0 < 2 the two constraints are compatible iff ϕ(t0 ) ≤ A. If t0 > 2

they are compatible iff ϕ(2) ≤ A. ¥

If t0 > 2 condition (30) is equivalent to [1+q(1−(F (2A))]A ≥ I. If t0 < 2

(30) is equivalent to [1+q(1−(F (t0 A))(t0 −1)]A ≥ I. Hence, because t0 > 1,
the firm’s debt capacity is strictly greater than A iff min(2, t0 )A < V . This
reasoning also shows that the debt capacity can be as small as A, despite
the double liquidation threat with two creditors. This is the case if the
firm’s long-term returns are concentrated at values close to A. In this case,
any attempt to provide high-powered repayment incentives leads to strategic
default.

20
3.2.3 Optimal contracts
We now turn to the determination of the optimal contract, i.e. the determi-
nation of the value t∗ that solves problem (24) - (27). Remember that higher
t means lower bankruptcy liquidation relative to the face value of debt D.

Proposition 3 (Bankruptcy recovery): The solution to the minimization

problem (24) - (27) satisfies t∗ > 1. Hence, under the optimal contract,
creditors are not fully repaid in bankruptcy: B ∗ < D∗ .

The proposition states that creditors should get less than the face value of
their claims in bankruptcy. This result is not trivial because in the model it is
quite possible to have full or even excessive recovery of claims in bankruptcy
if D < A. Yet, our model highlights that bankruptcy liquidation has the cost
of destroying value ex post, which is an effect the parties want to minimize
ex ante by relying more strongly on individual liquidation threats than on
aggregate ones. Proposition 3 only concerns aggregate values of debt and
repayment. Individual claims will be analyzed in the next section.
Proposition 3 has an immediate consequence for the optimal number of
creditors in our model. If the firm contracted with only one creditor, this
creditor would hold a liquidation claim of D regardless of the type of default.
This implies that the creditor would be repaid D in both states: in the good
state in cash (because the firm prefers paying out to liquidating the same
amount), and in the bad state through liquidation (because the firm is cash
constrained). Such a contract can be replicated in the two creditor case by
setting D = B (in which case the firm never defaults strategically, just as
with one creditor). Proposition 3 implies that such a contract is not optimal.

Proposition 4 (Optimal Number of Creditors): The firm strictly prefers

to have two creditors rather than one.

Hence, the ability of the debtor to pledge his assets individually to more
than one creditor strictly improves the terms of contracting. Because the
debtor can credibly promise a higher cash payout than his bankruptcy liqui-
dation value, the creditors can lower the amount of liquidation in bankruptcy.
Although this strengthens the debtor’s incentives for strategic default, these
incentives are outweighed by the improved repayment incentives. The next
proposition provides a sufficient condition for strategic default to occur in
equilibrium.

21
Proposition 5 (Strategic Default): Strategic default occurs with positive
probability if

f(V )(V − A)((1 − q)EV + qV ) < (1 − q)(EV − A) (31)

Proof: Condition (31) implies that the right-hand derivative of H(ϕ(t), t)

at t = V /A, (44), is strictly negative. Hence, t∗ > V /A. This implies the
claim, because by (11), the firm defaults strategically iff V < At.¥

Proposition 5 illustrates the basic tradeoff faced by the debtor. In the

original two-dimensional contracting problem (18)-(21), on the one hand,
minimising bankruptcy liquidation B is desirable. However, the participation
constraint (19) shows that this comes at the cost of increasing t, which means
that the face value of debt (D = tB = tϕ(t)) rises, which in turn favors
strategic default. Yet, as long as it is possible to decrease B such that t
does not exceed V /A, strategic default is not an issue and decreasing B
is unambiguously advantageous. Otherwise, we have t∗ > V /A, and the
contract induces some strategic default. This will happen, as condition (31)
shows, if V is sufficiently small. In this case, there are some realizations
of long-term firm value such that the firm prefers to keep the current cash
rather than the right to future returns.
We finally investigate under what conditions the firm is not fully liqui-
dated in bankruptcy.

Proposition 6 (Deviation from Absolute Priority): If I < qV + (1 −

q)A the bankruptcy liquidation constraint (20) is slack: B ∗ < A.
∗
Proof: If V /A > 2 we have t∗ ≥ V /A > 2. Therefore, ϕ(t∗ ) < t2 ϕ(t∗ ) ≤ A
by (27).
If V /A ≤ 2 (26) is slack if ϕ(V /A) < A, because ϕ is decreasing on
[V /A, t0 ] and because of (43). This last condition is equivalent to I < qV +
(1 − q)A.¥

Proposition 6 only provides a sufficient condition, but this suffices to

make the point. The point is that the constraint on B, (20) or (26), does not
need to bind at the optimum. If this is the case, the debtor retains some of
the assets after bankruptcy. This is due to the fact that continuation in the
present model is always efficient. Therefore, the parties have a strong ex ante

22
incentive not to punish the debtor too hard in case of bankruptcy. However,
the point continues to be true if continuing the firm ex post is efficient with a
sufficiently high probabability (V < A but F (A) sufficiently small), because
the level of bankruptcy liquidation must be independent of V . Hence, only
the expected continuation efficiency matters. On the other hand, if total
borrowing I is sufficiently high, it is optimal to make maximum use of the
liquidation threat and set B = A.
The basic message of Proposition 6 is fairly general: if the debtor has a
comparative advantage using the assets, it is ex ante costly to separate her
from them ex post, and, therefore, an optimal contract will aim at reduc-
ing this incidence as much as possible. This insight, simple as it is, is in
sharp contrast with traditional legal reasoning that demands to satisfy cred-
itors first in case of bankruptcy (absolute priority). On the other hand, our
model also spells out the costs of deviations from absolute priority (reduced
debt capacity and increased incentives for strategic default). It is therefore
consistent with the observed wide variety of practices of handling absolute
priority. In particular, it may be useful in explaining the long-term decline in
absolute priority violations in the U.S. observed by Bharath, Panchapegesan,
and Werner (2008).

4 Bankruptcy versus Debt Collection

In the base model of the last section, we have assumed that a simultaneous
attempt by creditors to collect their debt automatically triggers bankruptcy.
In reality, of course, bankruptcy must be triggered by someone, and the
base model is silent on this issue. In this section, we generalize the base
model to a model in which bankruptcy is not an automatic consequence of
simultaneous debt collection, but the result of deliberate individual decisions.
For these decisions the individual payoffs are important, which were left
largely unspecified in the optimal payment/liquidation schedule of the last
section.
Proposition 1 already implies some interesting structure for individual
claims.

Proposition 7 (No Over-Collateralization): The creditors’ individual

claims satisfy Bi∗ ≤ Di∗ for i = 1, 2.

Proposition 7 makes a familiar but not obvious statement. From Propo-

23
sition 3 we know that total bankruptcy liquidation must be strictly smaller
than the total face value of debt. Proposition 7 states that this must even
hold at the individual level: no single creditor must have the right to liqui-
date more in bankruptcy than his nominal debt claim. This is what is true
in practice, but upon reflection not a priori clear. In fact, our model provides
one possible reason why it might be optimal to have one creditor who is over-
collateralized, i.e. who holds a bankruptcy claim Bi > Di : such a creditor
would always trigger bankruptcy when the firm defaults strategically. Since
bankruptcy by Proposition 3 protects the firm, this is ex ante desirable and
would also have a dissuasive effect on strategic default. However, Proposition
7 also shows why such an arrangement is impossible: Giving one creditor a
high bankruptcy claim would imply that the other creditor gets such a small
bankruptcy claim that there are ex-post firm values V for which the firm will
default partially, which is not optimal ex ante.
We now generalize the contracting model of the last two sections by asking
what type of interaction at date 1 can bring about the (second-best) optimal
aggregate repayment/liquidation schedule (B ∗ , D∗ ) if bankruptcy does not
obtain automatically. For this, we must reconsider the assumption about si-
multaneous foreclosure in the debt collection game (4). While bankruptcy is
a coordinated attempt to recover funds from the debtor in default, simultane-
ous foreclosure is a priori uncoordinated. Hence, instead of the collective debt
collection procedure assumed in game (4) in Section 2, we will now assume
that simultaneous foreclosure results in an uncoordinated run for the assets,
in which the first to collect his debt liquidates Di , and the second receives
min(Dj , A−Di ) where j 6= i and i is the creditor who collects first. Assuming
that each creditor has the same probability of being first, the payoff matrix
becomes

a f
a p1 , p2 p1 , D2 (32)
1
f D1 , p2 D + 12 min(D1 , A − D2 ), 21 D2 + 12 min(D2 , A − D1 )
2 1

where p1 = p2 = 0 in the case of Y = 0.

In this framework, which represents a “pre-bankruptcy”, primitive state,
ex-post interactions and ex-ante contracting will be as in Section 3, with the
exception that B is fixed exogenously at B = min(A, D). This simplifies the
problem of Section 3 considerably, but, as Propositions 3 and 6 have shown,

24
the resulting contract will typically not be optimal. The reason is that the
deadweight loss from high liquidation in the bad state more than outweighs
the improved incentives for payout in the good state. Ex ante it is therefore
optimal to reduce the threat of liquidation from min(A, D) to B ∗ .
This can be done by introducing bankruptcy into the model; more pre-
cisely, by giving each creditor or the firm the right to trigger bankruptcy
when the firm defaults on (some of) its payments. Bankruptcy then means
that individual debt collection is suspended, and creditors receive Bi instead
of Di .
The possibility of triggering bankruptcy changes nothing in the analysis
of Section 3 if the firm has defaulted partially. In fact, the creditor who is
defaulted upon recuperates his full claim through individual debt collection
(which is individually optimal by Proposition 7), whereas the other creditor
weakly prefers to accept the renegotiated payment from the firm. The firm
prefers partial default to bankruptcy by revealed preferences.
If the firm has defaulted on both creditors, be it for liquidity or strategic
reasons, the analysis changes. In this case, creditor i when observing the
attempt to foreclose by creditor j, will call bankruptcy if this makes him
better off than waiting, i.e. if
Bi > min(Di , A − Dj ). (33)
Proposition 7 implies that for optimal values of Bi , Di condition (33) is
equivalent to
Bi∗ > A − Dj∗ . (34)
Since either of the two creditors may be last in line for debt collection,
condition (34) must hold for i = 1, 2, if an (inefficient) run for the assets shall
be avoided for sure. Adding up (34) for i = 1, 2 yields the joint condition
B ∗ + D∗ > 2A. (35)

Whether or not the optimal (B ∗ , D∗ ) satisfy this condition depends on

the parameters of the model. Using the definition of ϕ and t of the previous
section, (35) is equivalent to
(1 + t∗ )ϕ(t∗ ) > 2A (36)
If the optimal aggregate repayment/liquidation schedule (D∗ , B ∗ ) satisfies
the joint condition (35), the next question is whether one can find values

25
B1 ≤ D1 with D1 ≥ D∗ /2 such that (34) and the conditions (8) jointly
hold. If this is the case, the solution (D∗ , B ∗ ) can be implemented by giving
each creditor the right to trigger bankruptcy. If this is not the case or if
B ∗ + D∗ < 2A, this will not be possible, and at least one creditor will have
an incentive to run for the assets even if he has the right to trigger bankruptcy.
In this case, however, another simple contractual remedy is available:
allowing the firm to trigger bankruptcy whenever at least one creditor tries
to collect his debt.15 In fact, if B ∗ < A, the firm strictly prefers bankruptcy
to uncoordinated liquidation. If on the other hand B ∗ = A, the firm is
indifferent, but then there is no efficiency case for bankruptcy in the first
place.
The following proposition shows that the condition B ∗ + D∗ > 2A is
indeed sufficient for creditor-initiated bankruptcy to implement the optimal
repayment/liquidation schedule.

Proposition 8 (Implementation): If B ∗ + D∗ > 2A, the optimal repay-

ment/ liquidation scheme (B ∗ , D∗ ) can be implemented by giving each credi-
tor the individual right to trigger bankruptcy following default. If B ∗ + D∗ <
2A, the firm must have the right to trigger bankruptcy.

As noted above, whether or not the optimal contract satisfies the condi-
tion B ∗ + D∗ > 2A depends on the parameters of the model. It is easy to
see that both alternatives are possible. More detailed comparative statics of
(36) require a more detailed analysis of the properties of t∗ derived in Section
3.2 and are not attempted here.
The no-run conditions (34) impose further restrictions on B1 and D1 .
For example, they imply that no creditor can be fully secured, i.e. that any
optimal contract must satisfy Bi < Di , i = 1, 2.16 The reason is that full
bankruptcy protection of one creditor would induce the other creditor not
to trigger bankruptcy if the firm defaults, because he would get more from
an uncoordinated liquidation than from bankruptcy. Hence, if the firm is
not allowed to trigger bankruptcy, no creditor must hold safe claims. Yet,
even with the restriction of the no-run conditions, optimal individual claims
15
Whatever the other creditor does in this case - run for the assets trying to collect
Dj instead of A − Di , or standing by the sidelines to collect Dj for sure - the resulting
liquidation is inefficient, and the firm has an incentive to trigger bankruptcy.
16
Formally, the condition Bj > A − Di is equivalent to Bi < B − A + Di which implies
Bi < Di .

26
are not pinned down uniquely, and there is some freedom to design them
according to specific investor preferences (which we do not model here).
Note in particular that the restrictions on optimal individual claims im-
posed by Propositions 1, 7, and 8 do not rule out the possibility of treating
creditors equally in bankruptcy. Since D∗ = t∗ B ∗ , identical recovery rates in
bankruptcy for both creditors can (only) be achieved by setting Di = t∗ Bi
for i = 1, 2. For such individual claims, condition (8) of Proposition 1 can
easily be shown to hold if and only if B1 is chosen such that
t∗
B ∗ /2 ≤ B1 ≤ B∗ (37)
t∗ + 1
which is possible because t∗ > 1 by Proposition 3. If t∗ is relatively small
(which means, following the discussion in Section 3.2, that long-term firm
value does not exceed asset value A by much), then condition (37) forces the
claims of both creditors to be similar in size (B1 close to B ∗ /2).
Hence, if the no-run conditions (34) are compatible with the inequalities
(37), then individual claims can be designed such that creditors have the
same recovery rates in bankruptcy. Otherwise (both alternatives are possible,
depending on I and A), the no-run conditions cannot be satisfied and the
firm must have the right to trigger bankruptcy, if creditors are to be treated
equally in bankruptcy. If equal treatment is desirable for reasons outside the
present model, this provides a further justification for giving borrowers the
right to trigger bankruptcy.
Proposition 8 is interesting because it provides a new efficiency justifi-
cation for debtor-friendly bankruptcy rules. There is a broad consensus in
academia and practice that creditors should have the right to trigger bank-
ruptcy. The consensus is less developed with respect to debtor rights. While
most of the criticism of debtor-friendly rules concerns debtor-in-possession
rules such as those of Chapter 11 in the U.S., it could be argued that the
individual right of a debtor to evade the discipline of debt collection is harm-
ful. Proposition 8 shows that there is an efficiency reason for such a right: if
the creditors do not have the ex-post incentives to select the efficient contin-
uation decision, the debtor may have these incentives. Hence, any reform of
possibly excessive managerial discretion rights under Chapter 11 should be
careful not to throw out the child with the bathwater and scrap the debtor’s
right to trigger bankruptcy.17
17
On a related note, see Berkovitch, Israel and Zender (1998) and Povel (1999). An
excellent discussion of the initiation problem in bankruptcy is given by Baird (1991).

27
5 Conclusion
We have analyzed the design of bankruptcy rules and debt structure in an
optimal-contracting perspective. If cash flow is not verifiable and only the
asset value of the firm is verifiable, then when a firm borrows from a single
creditor and has all bargaining power, its debt capacity is limited to the
value of its asset base. The reason is that the creditor can never expect
to receive more than the asset value in liquidation and in renegotiation.
However, when a firm borrows from more than one creditor, it can increase
its debt capacity by pledging its asset base to more than one creditor by
giving each the right to foreclose individually. If the debt structure of the
firm is designed appropriately, the resulting non-cooperative game between
creditors ex post creates a commitment for the firm to pay out more in good
states to prevent the exercise of individual foreclosure rights and thus raises
the firm’s debt capacity. Having multiple creditors thus helps to reduce the
negative effects of the lack of commitment in contracting by distinguishing
between individual foreclosure rights and joint liquidation rights achieved
under bankruptcy.
Our theory provides a bridge between corporate finance and the legal
theory of debtor-creditor law. The key distinction in debtor-creditor law in
most jurisdictions is that between debt collection law and bankruptcy law.
The former governs the interaction between the debtor and a single cred-
itor, the latter the interaction between the debtor and several creditors.18
Our analysis shows how this same distinction can be made in a contract-
theoretic approach to debt. Individual foreclosure rights (corresponding to
debt-collection law) are crucial to generate repayment incentives, but need
to be complemented by collective liquidiation rights (corresponding to bank-
ruptcy law) in order to maximize ex-post efficiency.
Our results on debt structure and overleverage under multiple creditors
depend on the fact that creditors have unilateral foreclosure rights that they
can exercise in case of default, independently of what other creditors decide.
These rights should be seen as an important element of investor protection.
The renegotiation procedure modeled in this paper emphasises the effect of
these rights since renegotiation is assumed to happen on an individual basis.
The ensuing non-cooperative game between creditors forces the debtor to
18
The U.S. is a particularly telling example of this distinction. Here, the two bodies of
law are even governed by different authorities: debt-collection law is state law, bankruptcy
law is federal law.

28
respect contractual claims as given by individual foreclosure rights whenever
he wants to avoid default. The key assumption in this approach is that it
is difficult and costly to bring the creditors together to renegotiate the debt
contract collectively. Only bankruptcy brings all the contracting parties to-
gether at one table, but in bankruptcy the debtor has given up his residual
ownership rights, and the procedure is mostly concerned with the reconcilia-
tion of individual liquidation claims. This is the classical “vis attractiva” of
bankruptcy.19
Yet, it is theoretically conceivable that the debtor can unite the group of
creditors and extract from them joint concessions under the threat of bank-
ruptcy. If such workouts are frictionless, debt structure becomes irrelevant,
because individual collection rights have no bite in enforcing claims. As docu-
mented, e.g., by Asquith, Gertner, and Scharfstein (1994) and Gilson (1997),
however, frictions in such negotiations are usually substantial and increase
with the number of creditors. One important reason for these frictions is
the hold-up problem of individual creditors, which is precisely the reason for
institutionalised bankruptcy rules as discussed in Section 4. Further reasons
(which we ignore in this paper) include the aggregation of asymmetric in-
formation among multiple creditors and the legal uncertainty accompanying
out-of-court debt renegotiations, if individual creditors have the possibility
of contesting the new arrangement in court.
Our model has been kept deliberately simple and parsimonious. The
assumption that long-term asset value V is not verifiable has allowed us to
focus solely on short-term debt contracts. In a more general model, V would
have a verifiable and an unverifiable component. The verifiable component
would give rise to long-term debt and to the possibility to reschedule short-
term debt in debt renegotiation. The assumption that V ≥ A has focused
the analysis on the case of efficient ex-post continuation. In the more general
case V < A (analyzed in the appendix) there are situations in which ex
post there is no conflict of interest and the only question is how to divide
the liquidation value of the firm. None of the insights of the analysis are
changed, but the analysis becomes more complicated.
The analysis has been confined to the case of two creditors. To what ex-
19
The notion of “vis attractiva” (“attracting force”) is one of the classical principles of
bankruptcy theory going back to Roman times. It formulates the historical experience
that creditors typically fail to reach agreement when left to their own devices, and that
only bankruptcy - when the debtor gives up the right to his estate - has the force to bring
them together.

29
tent would our insights generalize to more than two creditors? The analysis
would become more complicated because now contracts have to specify what
each creditor gets in liquidation for every subset of creditors who choose to
foreclose in the renegotiation game. Yet, in most simple extensions of our
foreclosure game creditors would still have the incentive to liquidate individ-
ually if they are not paid their face value and all others do not liquidate.
Hence, the basic logic of our model applies, and our analysis suggests that
more creditors are better able to extract higher repayments ex post. Of
course, with many creditors, the simplifying assumption (2) of no liquid-
ity constraints becomes unreasonable. Now the optimal number of creditors
should emerge from trading off higher ex-post extraction in good states, up
to the point where liquidity is exhausted, against more inefficient liquidation
in bad states. Avoiding ex post asymmetric default would still be optimal
for fundamentally the same reason as with two creditors even though the
analysis would be more complicated. The extension of Propositions 7 and 8
to more than two creditors would also follow the same logic as in the case
of two creditors. But a full analysis of the many-creditor case is beyond the
scope of this paper.
In our model, all results are derived as parts of an optimal ex-ante con-
tract between debtor and creditors. Strictly speaking this suggests that there
is no need for a law. In practice, however, there may well be, if individuals
are unable to join the initial grand contract and write contracts specifying
procedures of collective behavior. In fact, this is the classical Rawlsian justi-
fication of legislation as a substitute for contracting in the “original position”
(Rawls, 1971), an approach to law, and bankruptcy law in particular, that
is wide-spread in legal thinking.20 The classic text of Jackson (1986), for ex-
ample, when exploring the foundations of bankruptcy law, only argues that
a “collective system of debt collection law” is needed, relegating the issue of
private contracting to a footnote.21 Building on this approach to the founda-
tions of bankruptcy law, our contribution is to make the hypothetical private
contract explicit.
The role for a law in our model therefore is that of a general rule that
completes individual incomplete contracts (which are mostly concerned with
individual financial claims such as the Ii and Ri ), when these contracts cannot
20
See, e.g., Schwartz (1998) and more generally Schwartz and Scott (2003).
21
“As such, it reflects the kind of contract that creditors would agree to if they were
able to negotiate with each other before extending credit” (Jackson, 1986, p. 17).

30
be executed in isolation. Going one step further, there is also an intersting
role for bankruptcy courts in our model. One obvious such role of bankruptcy
courts with great practical importance is that of a certification agency that
verifies the firm’s asset value A and the outstanding claims against the firm.
We have gone beyond that and shown that courts are important in imposing
an automatic stay to prevent a run for the assets and supervise the contrac-
tually agreed collective liquidation procedure. In fact, our model shows that
there is an important conceptual distinction between out-of-bankruptcy ne-
gotiations and bankruptcy. These two events are both inefficient ex-post, but
for different reasons: negotiations out of bankruptcy are inefficient because
the parties have difficulties to coordinate and avoid free-riding. Bankruptcy
has no coordination costs because the law forces the parties around one table,
but since the court enforces the liquidation promise of the original contract,
asset value is destroyed inefficiently. Going one step further, in our model
courts can also have a role in determining the distribution of the asset value
to the creditors. As we have argued, it is difficult in practice to fix bank-
ruptcy payouts Bi contractually because not everybody has joined the initial
grand contract. Hence, these payouts must be set ex-post in a situation
of widely diverging interests. Our model provides guidelines of the type
Bi = gi (D1 , D2 , I1 , I2 ) that determine bankruptcy payments as a function
of nominal claims. Note that aggregate bankruptcy payout in our model is
linked to total debt by B = ϕ(ψ−1 (D)) where ψ(t) = tϕ(t) (which is strictly
monotone by Lemma 1). Bankruptcy courts can use this type of relation to
implement rules for individual payments.
Another line of research to develop our model further is to analyze the
effect of different renegotiation procedures. We have considered a fixed bar-
gaining game between debtor and creditors, and analyzed its implications for
debt structure and bankruptcy. This bargaining game, bilateral exchange of-
fers, captures some important features and frictions of debt renegotiations,
but, in the spirit of Harris and Raviv (1995), it would be useful to compare
this structure to others. For these and other extensions, our model may be
a useful building block that allows to develop a more comprehensive theory
of bankruptcy and ultimately contribute to a broader comparative analysis
of the workings and effects of different bankruptcy laws.

31
6 Appendix
6.1 Proofs
Proof of Proposition 1:
As discussed in the main text, at the renegotiation stage the firm has
four alternatives that lead to the following payoffs. Under outcome (a, a),
the payoff is YH −D +V . Under outcome (f, f ) it is YH +(1−B/A)V . Under
outcome (f, a) it is YH −B2 +(1− DA1 )V , and under (a, f) YH −B1 +(1− DA2 )V .
By some straightforward but lengthy comparisons it can be shown that the
firm’s choice is, in fact, quite well structured. We summarize this choice in
the following Lemma.

Lemma A1: The renegotiation game at date 1 has the following outcomes:

1. If
D
B1 > D1 − D + B (38)
B
the firm optimally induces
V D−B+B1
• outcome (a, a) if and only if A
≥ D1
B−B1 V D−B+B1
• outcome (f, a) if and only if B−D1
≤ A
< D1
V B−B1
• outcome (f, f ) if and only if A
< B−D1

2. If
D D2
B1 < D1 + D − (39)
B B
the firm optimally induces
V D−B1
• outcome (a, a) if and only if A
≥ D−D1
B1 V D−B1
• outcome (a, f ) if and only if B−D+D1
≤ A
< D−D1
V B1
• outcome (f, f ) if and only if A
< B−D+D1

D D2 D
3. If D
B 1
+D− B
≤ B1 ≤ D
B 1
− D + B the firm optimally induces
V D
• outcome (a, a) if and only if A
≥ B
V D
• outcome (f, f ) if and only if A
< B
.

32
 The lemma states that the firm will default against both creditors if future
firm value is low, it will repay both creditors (D1 , D2 ) in cash if future firm
value is high, and it will default partially in an intermediate region. This
intermediate region is empty in the third case. Note that the condition in
this third case is condition (8) of Proposition 1.
We now show that it is not optimal to choose individual claims according
to condition (39). Define the two bounds derived in Lemma A1 for the
B1 D−B1
case (39) as VL = B−D+D 1
A and VH = D−D 1
A. We now take an arbitrary
combination (B, D, B1 , D1 ) that satisfies (39) and vary B1 , D1 for 0 < B1 <
B
B and D B1 + D − B < D1 < D such that the investors stay on their
participation constraint.
The investors’ (aggregate) participation constraint is

(1 − q + qF (VL ))B + q(1 − F (VH ))D + q(F (VH ) − F (VL ))(B1 + D − D1 ) = I.

(40)
Hence, small changes dB1 and dD1 keep the investors indifferent iff

(F (VH )−F (VL ))(dB1 −dD1 ) = (D1 −B1 )f (VH )dVH +(D−B−D1 +B1 )f (VL )dVL
(41)
where dVi = ∂B1 Vi dB1 + ∂D1 Vi dD1 .
The total ex-ante deadweight loss under any choice (D1 , B1 ) is
∙Z VL µ ¶ Z VH µ ¶ ¸
B D − D1
L = q V − B dF (V ) + V − (D − D1 ) dF (V )
V A VL A
B
+(1 − q)( EV − B)
A
(remember that cash transfers constitute no loss). Hence, under the small
change (dD1 , dB1 ) we have, after some manipulations,

1 VH VL
dL = (D − D1 )( − 1)f(VH )dVH + (B + D1 − D)( − 1)f (VL )dVL
q A A
µ Z VH ¶
V
+ F (VH ) − F (VL ) − dF (V ) dD1
VL A
VH VL
= (D − D1 )( − 1)f(VH )dVH + (B + D1 − D)( − 1)f (VL )dVL
µ A A ¶
Z
VH VL 1 VH
− ( − 1)F (VH ) − ( − 1)F (VL ) − F (V )dV dD1
A A A VL

33
where we have used partial integration for the last equality. Combining the
last formula with (41) shows, after some straightforward manipulations, that
increasing D1 strictly increases L if (39) holds. Therefore and since (39) is
a strict inequality, decreasing D1 increases ex-ante value, and the contract
satisfying (39) is not optimal.
A similar argument shows that (38) is not possible under an optimal
contract.
¥

Proof of Lemma 1: For t < V /A, ϕ is strictly decreasing. For t > V /A,
the derivative of ϕ is
1 − F (At) − Af(At)(t − 1)
ϕ0 (t) = − qI
(1 + q(1 − F (At))(t − 1))2
This derivative is negative iff
f (At) 1
<
1 − F (At) At − A
Hence, (MHRC) implies statements (a) and (b).
If f (V )(V − A) < 1, (b) implies that t0 > V /A ≥ 1. If f(V )(V − A) ≥ 1,
then V > A, hence t0 = V /A > 1, which proves (c).
Statement (d) follows by differentiation for t 6= V /A.¥

Proof of Proposition 3: It is easy to check that both partial derivatives

of the objective function H are positive:
Z At
HB (B, t) = (1 − q)EV + q(1 − F (At))At + q V dF (V ) > 0
0
Ht (B, t) = qAB(1 − F (At)) ≥ 0

This is intuitive: increasing collective liquidation (B) and total payout

obligations (D = tϕ(t), which is increasing in t by Lemma 1-(d), reduces the
firm’s profit. Given that t0 is the unique minimum of ϕ, this implies that
d
dt
H(ϕ(t), t) > 0 if t > t0 . Hence, the solution to the minimization problem
(24) - (27) satisfies t∗ ≤ t0 . Furthermore, by explicit calculation one finds
that for t < V /A (which implies F (At) = f (At) = 0),
d q(1 − q)I
H(ϕ(t), t) = − [EV − A] < 0 (42)
dt (1 + q(t − 1))2

34
 This implies that the solution satisfies
V /A ≤ t∗ ≤ t0 (43)
If V > A, (43) implies the desired result. If V = A we have t∗ > 1 iff
the right-hand derivative of H(ϕ(t), t) at V /A is strictly negative. Direct
calculation shows that this derivative is
qA2 I
[A + q(V − A) − ((1 − q)EV + qV )(1 − f (V )(V − A))]
(A + q(V − A))2
(44)
For V = A, this derivative is −q(1 − q)I[EV − A] < 0, which proves the
proposition.¥
Proof of Proposition 7: By Proposition 1, B1∗ , D1∗ , B ∗ , D∗ must satisfy
∗
constraint (8). Suppose first that B1∗ > D1∗ . Then (8) implies D1∗ < D D∗ −
B∗ 1
D∗ + B ∗ which is equivalent to D1∗ > B ∗ . Hence, D1∗ > B1∗ , a contradiction.
Now suppose that B2∗ > D2∗ . This is equivalent to B1∗ < D1∗ − D∗ + B ∗ . Now
(8) implies D1∗ < D∗ − B ∗ , which implies B1∗ < 0, a contradiction.¥
Proof of Proposition 8: Assume that B ∗ + D∗ > 2A. Because B ∗ ≤ A
(feasibility), this condition immediately implies D∗ > A. As argued in the
main text, if none of the two creditors shall prefer a run to bankruptcy, we
must have
B1 > A − D∗ + D1 (45)
B1 < B ∗ − A + D1 . (46)
A straightforward geometrical argument in the D1 −B1 plane (for B ∗ and
D∗ fixed) shows that conditions (45) and (46) are consistent with condition
(8), which we know from Proposition 1 an optimal contract must satisfy. For
this argument it is helpful to distinguish between the cases D∗ > 2B ∗ and
B ∗ < D∗ ≤ 2B ∗ . Figure 1 shows the case B ∗ < D∗ ≤ 2B ∗ : the shaded
area represents condition (8) together with the restrictions D1 ≥ D∗ /2 and
B1 ≤ B ∗ .
¥

6.2 Inefficient continuation: the case V < A

We now briefly sketch how the analysis changes if the continuation of the
firm can be inefficient ex post. Hence, we replace assumption (1) by the
assumption V < A < V .

35
 Figure 1: Consistency of conditions (45) and (46)

The main results do not change in this setting. However, some of the
analysis becomes more complicated. Already Lemma A1 changes signifi-
cantly, which reflects the different continuation incentives when V < A. In
fact, the main insight for the case V < A is that the debtor will always liq-
uidate all his assets voluntarily ex post. This means that the debtor values
an asset loss exactly like a cash loss. In particular, if V < A the debtor
prefers repayment (the cell (a, a)) over strategic bankruptcy (the cell (f, f ))
iff D < B. This condition, and the analogous ones for asymmtric default,
are different from the conditions in Lemma A1 and must therefore be added
to the (already complicated) set of conditions in Lemma A1. However, a
closer inspection shows that for the case of regime 3 (no partial liquidation),
defined by (8), the additional restrictions mean very little. More precisely, if
D > B, the default condition in regime 3 remains unchanged (i.e., the new
conditions are slack); and for D < B, it becomes even simpler, becaus there
never is strategic default.
Proposition 1 does not change at all when V < A. The reason for why
regime 3 is optimal ex ante is essentially the same as before: As seen in
Proposition 1, it is strictly suboptimal to induce partial liquidation when
V > A for sure, because it induces too much liquidation compared to what

36
it achieves on the incentives front. However, if V < A ex post, liquidation is
efficient and the parties are therefore indifferent ex ante whether to induce
partial liquidation or not. Hence, as long as there is some probability that
V > A, the parties will strictly prefer to induce regime 3.
For the analysis of Propositions 2 - 6, one additional argument is necessary
when V < A. As seen above, if D < B the debtor will always repay D
when solvent (regardless of V ). Hence, there is no concern with strategic
bankruptcy ex ante. This implies that it is optimal to increase D at the
margin (which is costless in terms of incentives and liquidation loss) and
lower B in exchange (which brings an efficiency gain if V > A ex post). It
follows that the optimal contract has D ≥ B.
This now implies that the analysis remains essentially unchanged in the
more general model. In fact, the debtor will default strategically if V < A
(because this costs him cash of B instead of cash of D - remember that
he transforms his assets into cash, anyhow), and will default strategically if
A ≤ V < tA (for the reasons discussed in Section 3.1). Taken together this
means that the debtor will default strategically iff V < tA, exactly as before.
The debtor’s ex ante objective therefore is to maximize
Z V
B
(1 − q)F (A)(A − B) + (1 − q) (1 − )V dF (V ) + qF (A)(Y + A − B)
A A
Z At Z V
B
+q Y + (1 − )V dF (V ) + q Y − D + V dF (V )
A A At

which is a direct generalization of S0 in (13). Hence, the deadweight loss to

be minimized is
∙ Z At ¸
H(B, t) = K + (1 − F (At))At + V dF (V ) B (47)
A

where Z
F (A) 1−q V
K= A+ V dF (V )
q q A
The function H as defined in (47) is exactly the same as in (28), taking
into account that F (A) = 0 in the case of Section 3.2 and that liquidation
for V < A is no deadweight loss. Similarly, the participation constraint does
not change at all, and therefore, the ϕ - function linking B and t remains
unchanged in the more general framework. Hence, the analysis can be con-
ducted as before. Intuitively, what happens is that all the realizations V < A

37
lead to voluntary liquidation by the debtor, which gives him verifiable funds
of A. This is the same as assuming that the distribution of V has V = A
with a mass point at A.
Put differently, liquidation is efficient for V < A and inefficient for V > A.
In the former case, liquidation by the creditors simply is a transfer, with no
ex-post efficiency consequences. Qualitatively, in the design of the initial
contract, the efficiency consideration of the case V ≥ A is therefore the only
one that matters. Quantitatively, shifting weight in the distribution of V to
the left of A will, of course, change things. For example, an inspection of the
ϕ - function in (29) shows that this shift will shift ϕ uniformly upward. But
this is only to be expected: if the firm’s prospects become worse, its debt
capacity goes down, and its bankruptcy loss will increase.

38
7 References
Acharya, V. K. John, and R. Sundaram (2010), "Cross-Country Variations
in Capital Structure: The Role of Bankruptcy Codes", Journal of Financial
Intermediation, forthcoming.
Asquith, P., R. Gertner and D. Scharfstein (1994), "Anatomy of Financial
Distress: An Examination of Junk Bond Issuers”, Quarterly Journal of Eco-
nomics 109, 625-658.
Aghion, Ph., Hart, O. and J. Moore (1992), ”The Economics of Bankruptcy
Reform”, Journal of Law, Economics and Organization 8, 523-46.
Baird, D. (1991), "The Initiation Problem in Bankruptcy", International
Review of Law and Economics 11, 223-232.
Baird, D. and D. Bernstein (2005), "Absolute Priority, Valuation Uncer-
tainty, and the Reorganization Bargain", mimeo, University of Chicago Law
School.
Bebchuk, L. (1988), ”A New Approach to Corporate Reorganizations” Har-
vard Law Review 101, 775-804.
Bebchuk, L. (2002), "Ex-ante Costs of Violating Absolute Priority in Bank-
ruptcy", Journal of Finance 57, 445-460.
Berglöf, E. and E.L. von Thadden, (1994), ”Short-Term Versus Long-Term
Interests : Capital Structure with Multiple Investors”, Quarterly Journal of
Economics 109, 1055-84.
Berkovitch, E. and R. Israel (1999), ”Optimal Bankruptcy Laws across Dif-
ferent Economic Systems”, Review of Financial Studies 12, 347-77.
Berkovitch, E., R. Israel and J. Zender (1998), ”The Design of Bankruptcy
Law: A Case for Management Bias in Bankruptcy Procedures,” Journal of
Financial and Quantitative Analysis 33, 441-64.
Bharath, S., V. Panchapegesan, and I. Werner (2008), "The Changing Nature
of Chapter 11", manuscript, University of Michigan.
Bisin, A. and A. Rampini (2006), ”Exclusive Contracts and the Institution
of Bankruptcy”, Economic Theory 27, 277-304.
Bolton, P. and D. Scharfstein (1996), ”Optimal Debt Structure with Multiple
Creditors”, Journal of Political Economy 104, 1-25.
Bris, A. and I. Welch (2005), "The Optimal Concentration of Creditors",
Journal of Finance 60, 2193-2212.

39
Cornelli, F. and L. Felli (1997), ”Ex-ante Efficiency of Bankruptcy Proce-
dures”, European Economic Review 41, 475-85.
Davydenko, S. (2005), "When Do Firms Default? A Study of the De-
fault Boundary" manuscript, Rotman School of Management, University of
Toronto.
Detriagache, E. and P. Garella (1996), ”Debt Restructuring with Multiple
Creditors and the Role of Exchange Offers”, Journal of Financial Interme-
diation 5, 305-336.
Dewatripont, M. and J. Tirole (1994), ”A Theory of Debt and Equity: Diver-
sity of Securities and Manager-Shareholder Congruence”, Quarterly Journal
of Economics 109, 1027-1054.
Franks, J. and W. Torous (1989), ”An Empirical Investigation of U.S. Firms
in Renegotiation”, Journal of Finance 44, 747-769.
Gale, D. and M. Hellwig (1985), ”Incentive-Compatible Debt Contracts: The
One-Period Problem”, Review of Economic Studies 52, 647-663.
Gale, D. and M. Hellwig (1989), "Repudiation and Renegotiation: The Case
of Sovereign Debt", International Economic Review 30, 3-31.
Gertner, R. and D. Scharfstein (1991), ”A Theory of Workouts and the Effects
of Reorganization Law”, Journal of Finance 46, 1189-1222.
Gilson, S. (1997), "Transactions Costs and Capital Structure Choice: Evi-
dence from Financially Distressed Firms”, Journal of Finance 52, 161-196.
Gilson, S., K. John and L. Lang (1990), "Troubled Debt Restructuring: An
Empirical Study of Private Reorganization of Firms in Default”, Journal of
Financial Economics 27, 315-353.
Hackbart, D., C. Hennessy and H. Leland (2007), "Can the Tradeoff Theory
Explain Debt Structure?", Review of Financial Studies 20, 1389-1428.
Harris, M and A. Raviv (1995), ”The Role of Games in Security Design”,
Review of Financial Studies 8, 327-367.
Hart, O. (1995), Firms, Contracts and Financial Structure, Clarendon Lec-
tures in Economics, Oxford University Press, Oxford.
Hart, O. and J. Moore (1998), ”Default and Renegotiation: A Dynamic
Model of Debt”, Quarterly Journal of Economics 113, 1-41.
Hege, U. and P. Mella-Barral (2005), ”Repeated Dilution of Diffusely Held
Debt”, Journal of Business 78, 737-786.

40
Jackson, T. (1986), The Logic and Limits of Bankruptcy Law, Harvard Uni-
versity Press, Cambridge, Massachussetts.
Kordana, K. and E. Posner (1999), ”A Positive Theory of Chapter 11,” New
York University Law Review 74, 161-234.
Povel, P. (1999) “Optimal ‘Soft’ or ‘Tough’ Bankruptcy Procedures ”, Jour-
nal of Law Economics & Organization 15, 659-84.
Rawls, J. (1971), A Theory of Justice, Harvard University Press.
Schwartz, A. (1998), ”A Contract Theory Approach to Business Bank-
ruptcy”, Yale Law Journal 107, pp. 1806-1851.
Schwartz, A. and R. Scott (2003), "Contract Theory and the Limits of Con-
tract Law", Yale Law Journal 113,
Townsend, R. (1979), ”Optimal Contracts and Competitive Markets with
Costly State Verification”, Journal of Economic Theory 21, 265-293.
Weiss, L. (1990), ”Bankruptcy Resolution: Direct Costs and Violation of
Priority of Claims”, Journal of Financial Economics 27, 285-314.
Winton, A. (1995), ”Costly State Verification and Multiple Investors: The
Role of Seniority”, Review of Financial Studies 8, 91-123.

41