mss # Cai; AP art. # 04; RAND Journal of Economics vol.

34(1)

RAND Journal of Economics

Vol. 34, No. 1, Spring 2003
pp. 6377

A theory of joint asset ownership

Hongbin Cai

I offer a theory of joint ownership by extending the standard property right theory of the firm to
situations where parties can endogenously choose the degree of specificity of their investments
(i.e., both the type of investmentspecific and generaland the level of each). When specific and
general investments are complements, the standard GHM results are obtained and joint ownership
is suboptimal. When specific and general investments are substitutes, joint ownership is optimal as
long as trade takes place within the relationship. Joint ownership provides stronger incentives to
make specific investments than any other forms of ownership by discouraging general investments.

1. Introduction
Although joint ownership is quite common in the real world, where examples include partnerships, cooperatives, and joint ventures, it is not well understood in the existing literature.
A common misconception about joint ownership is to define it as a precommitment to certain
profit-sharing rules. But a straightforward profit-sharing contract would achieve the same outcome independent of ownership structure. Assuming that profit-sharing contracts are not feasible
due to nonverifiability, the property right theory of the firm, or the GHM approach (Grossman
and Hart, 1986; Hart and Moore, 1990; Hart, 1995), defines asset ownership as allocations of
residual control rights over assets. Joint ownership means that residual control rights of an asset
are shared by its co-owners, that is, each co-owner has veto power over how to use the asset and
cannot be excluded from accessing the asset. However, according to the GHM approach, joint
ownership of an asset is suboptimal, because it provides the fewest investment incentives for
every co-owner (Hart, 1995, p. 48).
To gain some concrete ideas about joint control over assets, let us consider Chinese joint
ventures. Bai, Tao, and Wu (1999) collect a sample of 200 joint venture contracts, in which partners
of the venture (foreign and local) specify board voting rules for various important corporate
issues. Three voting rules are used in these contracts: unanimous, two-thirds majority, and simple
majority. Both partners will have veto power over a certain issue if (1) a unanimous voting rule is
required, or (2) a two-thirds majority rule is required and no partner has more than two-thirds of

University of California at Los Angeles; cai@econ.ucla.edu.

An earlier version of this article was circulated under the title Endogenous Specificity and Joint Asset Ownership. I
am grateful to Joe Ostroy for stimulating conversations and insightful comments and to Jeff Zwiebel for valuable guidance
and encouragement. The Editor, Lars Stole, and two anonymous referees made extremely constructive suggestions that
greatly improved the article. I have also benefited from comments of Harold Demsetz, Bryan Ellickson, Matthew Jackson,
Naomi Lamoreaux, Phillip Leslie, David Levine, Bentley MacLeod, Gregory Mark, Michael McGill, John Riley, JeanLaurent Rosenthal, Ilya Segal, Steve Tadelis, Earl Thompson, Michael Whinston, Joel Watson, and Bill Zame. Participants
in the Southern California Theory Conference (1999), Econometric Society Summer Meetings (1999), Stanford SITE
Workshops (1999), and seminars at Caltech, UCLA, and USC have made helpful suggestions as well. Financial support
from the James Collins Faculty Fellowship at UCLA is gratefully acknowledged. All remaining errors are mine.

Copyright 2003, RAND.

63

mss # Cai; AP art. # 04; RAND Journal of Economics vol. 34(1)

64

THE RAND JOURNAL OF ECONOMICS

the total shares (board composition reflects share contributions), or (3) a simple majority rule is
required and the share contribution is 50:50. Bai, Tao, and Wu find that joint control is pervasive
in these Chinese joint ventures. In particular, 198 out of 200 firms adopt a unanimous rule for the
following issues: amendment of company constitution, merger with a third party, termination of
venture, and increase in registered capital and equity transfer to a third party. For other important
issues, most of the contracts with specific voting rules give veto power to each partner as well.
It seems quite clear that such control structures make it very difficult for a party to exclude its
partner and take away the assets of the venture for some other use.
Despite the inferiority of joint ownership predicted by the GHM approach, joint ventures
have commanded a majority share of the vast amount of foreign direct investment (FDI) in China
and have been an important force in Chinas rapid economic growth for the last two decades.
This is not unique to China: joint ventures are common in FDI in other countries, developing or
developed (see, for example, Gomes-Casseres (1989) and the references cited therein). A recent
study by Allen and Phillips (2000) finds some positive effects on financial performance when
firms form joint ventures or other alliances.1
This article provides a simple explanation for why joint ownership can be optimal. The
explanation is based on the GHM premise, but with one additional feature. In the standard GHM
model, two parties make noncontractible relation-specific investments and then bargain over the
distribution of the total surplus generated by their investments. Through its effects on the two
parties outside options, asset ownership affects the surplus distribution and hence investment
incentives. Notice that relation specificity is often exogenously imposed in the GHM approach.
That is, the parties in question can only choose the levels of investment within the relationship
(which have positive spillover effects on their outside options) but cannot make other types of
investments that may enhance their values outside the existing relationship. This assumption may
not be reasonable in many circumstances. For example, each party may often want to exert efforts
in searching for alternative business partners or to make investments tailoring to the needs of
potential alternative partners. Such searching or investment can be valuable to the investing party
even if it does not add value to his trade with his partner, because it may enhance his bargaining
position against his partner. In this article, I allow players to have choices between relation-specific
investments and general investments.2
When specific and general investments are complementary, then the GHM results are obtained and joint ownership is suboptimal. In light of this, the standard GHM model can be thought
of as the special case where general and specific investments are perfect complements.3 The idea is
that independently owning an asset increases the owners outside option, giving the owner greater
incentives to make general investments and hence greater incentives to make specific investments.
In this case, joint ownership destroys the incentive instruments of asset ownership and thus is
suboptimal. But when specific and general investments are substitutes, joint ownership becomes
an optimal ownership structure. By sharing residual control rights, the co-owners of an asset have
the fewest incentives to make general investments, and hence the greatest incentives to make
specific investments. Therefore, joint ownership serves as a mutual commitment device when the
parties have a choice between relation-specific and general investments. To a large extent, joint
ownership in my model amounts to mutual hostage, which can promote cooperation and improve
efficiency by committing the parties to their relationship (Schelling, 1960).
The analysis of the article builds on the following key assumptions of the property right
1 Chinas foreign ownership policies have certainly forced many foreign firms to accept joint ventures they
otherwise would not choose, especially in the early reform years. And although regulation restrictions eased significantly
in the late 1990s, joint ventures remain the leading ownership structure for FDI in China, accounting in 1998 for more
than 40% of FDI, with fully foreign firms accounting for less than but close to 40% of FDI and various cooperative
arrangements taking the remaining share (China Statistic Bureau, 1999).
2 We can call the latter type of investment general investment because the analysis applies equally well to cases
in which investment in outside markets also generates returns within the relationship; see the discussion at the end of
Section 5.
3 See also Segal and Whinston (2000) for this interpretation of the standard GHM model.
RAND

2003.

mss # Cai; AP art. # 04; RAND Journal of Economics vol. 34(1)

CAI

65

theory of the firm: (i) disagreement payoffs affect the final division of trade surplus, and (ii)
disagreement payoffs of the two parties are determined by the combinations of their general
investments and the assets they control. In the framework of the property right theory of the firm,
the fundamental property of joint asset ownership is that neither of the two parties can get the
marginal benefits of their general investments combined with the jointly owned assets. In terms of
providing investment incentives, the co-ownership of an asset by the two parties is equivalent to no
one owning the asset, which is bad for general investments but good for specific investments. In the
next section, I shall argue that the legal framework of partnerships in the United States governing
default dissolution rules is consistent with this definition of joint asset ownership. I shall also show
that my result is unchanged when I allow renegotiation of asset ownership following a breakdown
of trade negotiations, under fairly reasonable bargaining assumptions of asset renegotiations that
are consistent with those of trade negotiations.
The result of this article seems to square well with casual observations about joint ownership.
In professional service industries such as law, accounting, consulting, and architecture, specific
human capital investments are crucial to firm success, and it is difficult to prevent people from
overinvesting in general human capital (or searching for outside opportunities) through contractual
means. Because specific and general human capital investments are likely to be substitutes (if total
time available is fixed), it makes sense in light of my model that partnerships dominate in such
industries. Similarly, joint ventures are often found in R&D cooperation in high-tech industries and
between multinational companies and local firms in foreign direct investments, where contractual
solutions are hard to find and marginal incentives to invest outside the relationship (e.g., searching
for other partners) are strong.4
A closely related article is Segal and Whinston (2000), who consider the effects of exclusive
trade contracts on investment incentives in a similar environment with complementary or substitute
internal and external investments. My analysis focuses on ownership structures. Ownership differs
from exclusive trade contracts in that asset ownership gives a party the right to exclude others
from accessing an asset, whereas an exclusive trade contract prevents a party from trading with
outsiders. In many cases, e.g., joint ventures involving joint production, it may be difficult to
define trade with outsiders ex post but easy to specify control over assets ex ante.5 Several other
articles have also studied various organizational design issues with complementary or substitute
multidimensional investments, e.g, Holmstrom and Tirole (1991), Rajan and Zingales (1998),
Holmstrom (1999), Tirole (1999), and Baker, Gibbons, and Murphy (2002). In addition, Bai, Tao,
and Wu (1999) argue that joint control can sometimes help commit to revenue-sharing contracts.
Maskin and Tirole (1999) show that joint ownership can lead to the first-best investment levels if
combined with an option-to-sell contract. Halonen (1997) shows that joint ownership may help
maintain reputation in a dynamic setting through renegotiation of ownership contracts.6
The rest of the article is organized as follows. Section 2 presents the basic model, and Section
3 presents the assumptions and some preliminary analysis. In Section 4 I characterize the set of
equilibrium for the game. Then I derive the optimality of joint ownership in Section 5, followed
by concluding remarks in Section 6.

2. The model
The model is an extension of Hart (1995). To highlight the real differences, I use Harts
symbols and notation whenever possible. Two parties, M1 and M2, are engaged in a business

4 In this regard, marriage might be viewed as a special form of joint venture when community property is the rule for
household assets. One difference is that family and divorce laws usually specify the asset ownership arrangement between
the husband and wife, so individuals may not have complete freedom to choose their desired ownership arrangements.
5 Besides having different focuses, the two models also differ in several other aspects. First, Segal and Whinston
study a more general three-party game (a buyer and two sellers), whereas outsiders in my model are passive. Second, they
mainly consider exclusivity on one side of the market (the buyer), whereas the two parties in my model are symmetric.
6 Hart and Moore (1998) compare cooperatives and outside ownership from a perspective different from investment
incentives.
RAND

2003.

mss # Cai; AP art. # 04; RAND Journal of Economics vol. 34(1)

66

THE RAND JOURNAL OF ECONOMICS

relationship. There are two assets, a1 and a2 , and M1 (M2) has to work on a1 (a2 ) to be
productive. The timing of the model is as follows.
At date 0, ownership structure is chosen. Let A = {a1 , a2 } and be the empty set. Let A1 A
and A2 A be the assets owned by M1 and M2 respectively. Denote an asset ownership structure
by A = {A1 , A2 }. In this simple setting, there are five ownership structures: (i) no integration
A N I = {A1 , A2 }, where A1 = {a1 } and A2 = {a2 }; (ii) type-1 integration AT 1 = {A1 , A2 },
where A1 = {a1 , a2 } and A2 = ; (iii) type-2 integration AT 2 = {A1 , A2 }, where A1 = and
A2 = {a1 , a2 }; (iv) joint ownership A J = {A1 , A2 }, where A1 = {a1 , a2 } and A2 = {a1 , a2 };
and (v) outside ownership A1 = A2 = and A0 = {a1 , a2 }, where player M0 is an outsider who
does not make any kind of investment. Here, the overbar on an asset means it is jointly owned by
both parties. The exact definition of joint ownership will be made clear shortly. Besides these five
basic ownership structures, there can be various hybrid forms of ownerships, for example, type-1
partially joint ownership A J 1 = {A1 , A2 }, where A1 = {a1} and A2 = {a1 , a2 }. It will become
transparent that the main result of the article (superiority of joint ownership) holds equally well
when we count in those hybird forms. I assume that ownership structure is chosen at date 0 to
maximize the total net surplus.
At date 1, M1 and M2 make investment decisions noncooperatively. I assume that each party
makes two kinds of investments, specific investment and general investment. Let (i 1 , i 2 ) be
M1s investment choice, where i 1 0 is specific and i 2 0 is general; let (e1 , e2 ) be M2s
investment choice, where e1 0 is specific and e2 0 is general. The investment cost to M1
is given by C1 (i 1 , i 2 ), to M2 by C2 (e1 , e2 ). The joint surplus is S = R(i 1 ) + Q(e1 ) if M1 and M2
choose to trade with each other at date 2. The assumption that S is independent of ownership
structure captures the notion that information or technology is unaffected by ownership structure.
The general investment by M1 (M2) enhances his (her) value in the outside market if M1 and M2
do not trade with each other at date 2. The idea is that there is a demand for M1s (M2s) service
or product outside of the relationship between M1 and M2, and M1s (M2s) general investment
increases the value he (she) can realize in the outside market. Let r (i 2 ; A1 ) and q(e2 ; A2 ) denote
the outside option of M1 and M2, respectively. Let s = r + q be the total outside surplus M1 and
M2 can generate.
The outside options depend on asset ownership because the owner of an asset can take the
asset with him/her to the outside market and thus enhance his/her value. This is clear when an asset
is owned solely by one party. Let us suppose that there is an active asset market, and the market
value for a set of assets is B(A).7 Let us also denote B1 , B2 , and B12 as the market value for a1 , a2 ,
and {a1 , a2 }, respectively. When M1 and M2 break up with each other, each of them actually has
two outside choices: (i) they can sell the assets they own to the outside asset market and then trade
with some other partner, or (ii) they can use their assets in trading with some other partner. M1
and M2 will choose to sell their assets before trading with some other partner whenever selling
assets yields greater payoff. So the outside option (value) of each player is the larger of his payoffs
from these two choices.8 The first choice gives M1 a payoff of r (i 2 ; ) + B(A1 ) and M2 a payoff
of q(e2 ; ) + B(A2 ). Since it is always possible to sell the assets before trading with some other
partner, we should have r (i 2 ; A1 ) r (i 2 ; ) + B(A1 ) and q(e2 ; A2 ) q(e2 ; ) + B(A2 ) when
every asset is owned solely by one party.
When an asset is jointly owned by M1 and M2, what can be done with the asset when they
break up? By definition, neither party can take the asset to trade in the outside market (either sell
or use) without the consent of the other party. Naturally they will try to come to an agreement on
how to dispense the jointly owned assets through negotiations. Some default rule regarding asset
disposal is needed in the event that such negotiations fail. In this article, my working definition of
7

When there is no outside asset market, we can simply let B = 0, so this is without loss of generality.
One may expect that the second choice is more likely to generate higher payoff than the first, since general
investments may be asset-specific. But this does not have to be true all the time; for example, the asset market might
experience some favorable shocks from time to time. In any case, my model works well no matter which outside choice
is better.
8

RAND

2003.

mss # Cai; AP art. # 04; RAND Journal of Economics vol. 34(1)

CAI

67

joint asset ownership assumes the following default rule for joint assets: in the event of an M1 and
M2 breakup, the assets will be sold in the outside asset market and revenue from selling them will
be divided between the two. The sharing rule of the asset sale revenue can be specified ex ante
in the joint venture agreement. So under joint ownership, M1s outside option is r (i 2 ; {a1 , a2 }) =
r (i 2 ; ) + 1 B12 , and M2s outside option is q(e2 ; {a1 , a2 }) = q(e2 ; ) + (1 1 )B12 , where 1 is
M1s share of asset sale revenue.
A few points are worth noting about this definition of joint asset ownership. First, my definition is consistent with the common definition of joint asset ownership in the property right theory
of the firm. According to Hart (1995), joint asset ownership means that when trade negotiation
breaks down, neither M1 nor M2 has independent access to the jointly owned assets. He then
elaborates this by saying that it is equivalent to M1 and M2 independently owning some assets
that are strictly complementary, which means that having access to those assets alone has no
effect on M1 or M2s marginal returns of (general) investments. This is the same as my definition,
by which we also have r  (i 2 ; {a1 , a2 }) = r  (i 2 ; ) and q  (e2 ; {a1 , a2 }) = q  (e2 ; ). Second, and
perhaps more important, my definition is consistent with the common practice of joint ownership
firms in the real world. One of the important forms of joint ownership is partnership. The Uniform Partnership Act drafted by the National Conference of Commissioners on Uniform State
Laws provides a standard legal framework of partnership firms for all states. In a recent version
of the act (1997), its Section 807 that provides the default rules for the settlement of accounts
and contributions among partners in winding up the business requires that (i) Each partner is
entitled to a settlement of all partnership accounts upon winding up the partnership business
and (ii) Any surplus must be applied to pay in cash the net amount distributable to partners in
accordance with their right to distributions under subsection (b). In the Comment on the section,
it is further explained that After the payment of all partnership liabilities, any surplus must be
applied to pay in cash the net amount due the partners under subsection (b) by way of a liquidating
distribution, and the in-cash rule . . . provides that a partner has no right to receive, and may not
be required to accept, a distribution in kind, unless otherwise agreed.9 Clearly, the default rule in
my definition of joint ownership is consistent with that specified in the Uniform Partnership Act,
in particular the entitlement of each partner and the in-cash rule. A third point to note about my
definition of joint asset ownership is that the default rule of liquidation is just that, the rule that
specifies a default for asset disposal in the event M1 and M2 break up with each other. It may well
happen that one partner values the jointly owned assets more than the outside asset market and
thus attempts to buy them from the other partner, in which case they need to reach a deal about
asset reallocation (the unless otherwise agreed part in the Uniform Partnership Act). As will
become clear below, my model works fine allowing such asset ownership renegotiations under
fairly reasonable bargaining environments of asset renegotiations. Thus, what is important in the
definition of joint ownership is that it defines an ultimate default rule in the event of disagreement,
which does not mean that M1 and M2 need to commit ex ante to such an arrangement (probably
quite inefficient ex post) if trade breaks down. In fact, when one party has the highest value for
an initially jointly owned asset at date 2, he/she will always get the asset through efficient asset
renegotiations.
At date 2, after investment has been made and observed by both parties, M1 and M2 bargain
over whether and how to trade with each other. Since there is no information problem, bilateral
bargaining will lead to efficient outcomes. As in GHM, we suppose that the bargaining outcomes
are given by the split-the-net-surplus solution, i.e., the Nash bargaining solution. This solution is
appropriate in situations where the bargainers receive their outside options during the bargaining
process or can fall back on their outside options if bargaining breaks down.10 For simplicity,
assume that M1 and M2 have equal bargaining power so they split the net surplus 50:50.
9 From the Uniform Partnership Act (1997), available at www.law.upenn.edu/ bll/ulc/ulc final.htm. The default
rule in this version was little changed from earlier versions. See also Lamoreaux (2002) for an excellent historical account
and analysis of partnerships and corporations in the United States.
10 Theoretically, the point can be made with an alternating-offer bargaining model in which bargaining breaks
down with a small probability whenever an offer is rejected, and the two players get their outside options in a bargaining
RAND

2003.

mss # Cai; AP art. # 04; RAND Journal of Economics vol. 34(1)

68

THE RAND JOURNAL OF ECONOMICS

Specifically, for a fixed asset ownership A and investment levels i = (i 1 , i 2 ) and e = (e1 , e2 ),
we have:
Trade equilibria. When S s = R(i 1 ) + Q(e1 ) [r (i 2 ; A1 ) + q(e2 ; A2 )] 0, then M1
and M2 will trade with each other, and the division of the total surplus S depends on the
two parties outside options: M1 will get r + (S s)/2 and M2 will get q + (S s)/2.
No-trade equilibria. When S s = R(i 1 ) + Q(e1 ) [r (i 2 ; A1 ) + q(e2 ; A2 )] < 0, then
M1 and M2 will not trade with each other and M1 will get a surplus of r (i 2 ; A1 ) and
M2 will get a surplus of q(e2 ; A2 ) from their corresponding outside markets.
Note that in the above description of date-2 outcomes, I assume asset ownership is not renegotiable in the event M1 and M2 are to break up their relationship. Although this nonrenegotiable
asset ownership assumption is maintained universally in the property right theory of the firm, it
is rather unreasonable because it requires too much commitment ability on the two parties and
is inconsistent with asset reallocations often observed in the disintegration and dissolution of
firms. However, as Cai (2001) shows, this assumption is actually not as limited as it seems. In
fact, under quite natural bargaining assumptions of asset renegotiation, allowing asset ownership
renegotiation does not affect the model.
Specifically, let us modify the model at date 2 to include an intermediate stage of asset
ownership renegotiation. That is, if M1 and M2 are to break up with each other, instead of
simply getting r (i 2 ; A1 ) and q(e2 ; A2 ) as the default payoffs determined by their initial ownership
agreement, they can renegotiate the asset ownership so as to achieve the maximal dissolution
surplus. Given general investment levels i 2 and e2 chosen at date 1, let A (i 2 , e2 ) be an ownership
structure that maximizes r (i 2 ; A1 ) + q(e2 ; A2 ), and let s (i 2 , e2 ) = s(i 2 , e2 ; A ) = r (i 2 ; A1 ) +
q(e2 ; A2 ) denote the highest joint surplus M1 and M2 can possibly achieve through asset ownership
renegotiations. The net gain from asset ownership renegotiation is then s (i 2 , e2 ) s(i 2 , e2 ; A),
where s(i 2 , e2 ; A) = r (i 2 ; A1 ) + q(e2 ; A2 ) is the joint surplus from the initial ownership structure.
Consistent with the bargaining assumptions made in the model, let us suppose that M1 and
M2 will divide the net gain from asset ownership renegotiation through the Nash bargaining
solution with equal bargaining power. Basically, we assume that the two parties have the same
relative bargaining power in asset renegotiation as in trade negotiations.11 This seems to be quite
natural, given that the relative bargaining power is exogenously given in both the Nash bargaining
solution and in noncooperative Rubinstein-type bargaining (determined by relative patience or
relative risk of bargaining breakdown). Under our assumption about asset ownership renegotiation,
after renegotiation of asset ownserhip, M1s payoff is u 1 (i 2 , e2 ; A) = r (i 2 ; A1 ) + .5[s (i 2 , e2 )
s(i 2 , e2 ; A)] = .5s (i 2 , e2 )+.5r (i 2 ; A1 ).5q(e2 ; A2 ), and M2s payoff is u 2 (i 2 , e2 ; A) = q(e2 ; A2 )+
.5[s (i 2 , e2 ) s(i 2 , e2 ; A)] = .5s (i 2 , e2 ) + .5q(e2 ; A2 ) .5r (i 2 ; A1 ).
From the above discussion, we see that because of asset ownership renegotiation, the date-2
outcome of the model needs to be modified as follows. When S s < 0, then M1 and M2 will
not trade with each other and M1 will get a surplus of u 1 (i 2 , e2 ; A) and M2 will get a surplus
of u 2 (i 2 , e2 ; A). When S s 0, then M1 and M2 will trade with each other. Now M1 will
get u 1 (i 2 , e2 ; A) + (S s )/2 and M2 will get u 2 (i 2 , e2 ; A) + (S s )/2, since the disagreement
payoffs should take into account asset ownership renegotiation. The following result from Cai
(2001) shows that allowing renegotiation of asset ownership has no effect in my model.
Lemma 1. For any initial asset ownership A and investment choices, as long as M1 and M2 trade
with each other in both cases of renegotiable or nonrenegotiable asset ownership, M1 and M2
have identical payoffs in these two cases.
breakdown. The unique subgame-perfect equilibrium is exactly the split-the-net-surplus solution in the limit when the
probability of bargaining breakdown goes to zero (e.g., see Binmore, Shaked, and Sutton, 1989; Osborne and Rubinstein,
1990).
11 The main result of the model will still hold even if the relative bargaining power of the two parties changes
modestly from trade negotiation to asset ownership renegotiations.
RAND

2003.

mss # Cai; AP art. # 04; RAND Journal of Economics vol. 34(1)

CAI

69

Proof. See the Appendix.

The basic idea behind Lemma 1 is the following. Suppose M1 and M2 will trade (which is
the case we are primarily interested in). Allowing renegotiation of asset ownership means that the
disagreement point determined by the initial ownership structure can be moved upward. But as
long as the relative bargaining power of the two parties remains the same in the asset renegotiation
game, this new disagreement point will be on the line connecting the initial disagreement point and
the agreement point. Therefore, bargaining from this new disagreement point will lead to the same
agreement pointthat is, the same payoff functions for both parties as in the case without asset
ownership renegotiation. See Cai (2001) for more details. Given Lemma 1, purely for simplicity,
I shall maintain the assumption of nonrenegotiable asset ownership.
In this article I use the Nash bargaining solution to arrive at the division of the surplus from
trade. It should be noted that the general idea here can be extended to other bargaining situations
whose outcomes are not characterized by the Nash bargaining solution. For example, it can be
shown that joint ownership is still optimal under conditions similar to those in this article when
outside options serve as constraints on bargaining agreements rather than disagreement payoffs in
the Nash bargaining solutionthe so-called outside-options principle solution (see Cai (1999) for
details).12 What is critical for my result is that when the parties share residual control rights over
an asset, they are forced to share at least some of the marginal returns from general investments
combined with the asset in the ex post renegotiations. As long as this is the case, there is a role
for joint ownership in encouraging specific investments in the relationship.

3. Assumptions and preliminary analysis

If M1 and M2 can write contracts on investment or on how to trade, the efficient outcome is
easily reached so asset ownership is irrelevant. For asset ownership to be relevant, I assume that
no contract (except ownership arrangements) is possible between the two players.
I maintain the following standard assumptions throughout: (i) R(i 1 ) and Q(e1 ) are positive,
increasing, and concave; (ii) r (i 2 ; A1 ) and q(e2 ; A2 ) are nonnegative, nondecreasing, and concave
in the first element (i.e., investment); and (iii) both C1 and C2 are nonnegative, increasing, and
convex.
I also make a number of other assumptions. The first two are standard in the GHM approach.

Assumption 1. Asset ownership increases the marginal product of general investment in the
outside options. Formally, for any i 2 , r  (i 2 ; {a1 , a2 }) r  (i 2 ; {a1 }) r  (i 2 ; {a1 , a2 }) = r  (i 2 ; );
for any e2 , q  (e2 ; {a1 , a2 }) q  (e2 ; {a2 }) q  (e2 ; {a1 , a2 }) = q  (e2 ; ).
Note that since no party has exclusive control over the asset under joint control, for each
co-owner, the marginal product of general investment under joint ownership is the same as that
if he owns no asset. This follows from our definition of joint asset ownership: r (i 2 ; {a1 , a2 }) =
r (i 2 ; ) + 1 B12 and q(e2 ; {a1 , a2 }) = q(e2 ; ) + (1 1 )B12 .
Assumption 2. Investments are relation-specific: investments yield higher values within the relationship than in the outside markets. Formally, for any ownership structure A = {A1 , A2 },
R(i) > r (i; A1 ) and Q(e) > q(e; A2 ), where i and e are any scalar investment levels.
Assumption 2 requires that M1 (M2) is essential to realizing Q (R). Specificity on marginal
terms (i.e., R  > r  , Q  > q  ) is not necessary for my model, but it is sufficient for Assumption 4
in the next section.
Assumption 3. Specific investment and general investment are perfect substitutes for both parties:
C1 (i 1 , i 2 ) = C1 (i 1 + i 2 ) and C2 (e1 , e2 ) = C2 (e1 + e2 ).
12 Recently, de Meza and Lockwood (1998) and Chiu (1998) have shown that the results of GHM are sensitive to
the bargaining solution used in modelling the distribution of surplus in the ex post bargaining stage. Specifically, using the
outside-options principle solution, these authors show that it is possible to derive implications about ownership structures
that are directly opposite to those in GHM.
RAND

2003.

mss # Cai; AP art. # 04; RAND Journal of Economics vol. 34(1)

70

THE RAND JOURNAL OF ECONOMICS

Assumption 3 departs from the standard GHM approach and is the driving force behind
the optimality of joint ownership. In the general model where choices between specific and
general investments are endogenous, the GHM approach can be thought of as making the implicit
assumption that specific and general investment are perfect complements.13 It is easy to see that
the qualitative results of the current article hold as long as the two kinds of investments are
substitutes (i.e., C1 and C2 both have positive cross-derivatives), hence perfect substitution is a
simplifying assumption. Likewise, if the investments are complementary, the GHM results are
obtained.
If M1 and M2 trade with each other at date 2, the net total surplus is  = R(i 1 ) + Q(e1 )
C1 (i 1 + i 2 ) C2 (e1 + e2 ). To maximize , obviously both i 2 and e2 should be zero and the optimal
i 1 and e1 satisfy the following first-order conditions:
R  (i 1 ) = C1 (i 1 )
Q  (e1 ) = C2 (e1 ).

(1)
(2)

I assume that the solution (i 1 , e1 ) is interior, and that the maximum net total surplus  is strictly
positive.
If M1 and M2 do not trade with each other at date 2, the net total surplus for a given asset
ownership structure A = {A1 , A2 } is (A) = r (i 2 ; A1 ) + q(e2 ; A2 ) C1 (i 1 + i 2 ) C2 (e1 + e2 ).
To maximize , obviously both i 1 and e1 should be zero and the optimal i 2 and e2 satisfy the
following first-order conditions:
r  (i 2 ; A1 ) = C1 (i 2 )


q (e2 ; A2 ) =

C2 (e2 ).

(3)
(4)

I assume that the solution (i 2 (A1 ), e2 (A2 )) is interior for every A, and that M1s payoff r (i 2 ; A1 )
C1 (i 2 ) and M2s payoff q(e2 ; A2 ) C2 (e2 ) are both positive.
Under Assumption 2, the first-best solution is for M1 and M2 to trade: the optimal investment
levels are (i 1 , 0) for M1 and (e1 , 0) for M2, and the maximum net total surplus is  . Because
of contract incompleteness, the first-best solution is in general not attainable. For a benchmark
second-best solution, consider the case in which there are no outside options for either party,
i.e., r (i 2 ; A1 ) = q(e2 ; A2 ) = 0 for any i 2 , e2 , A1 , and A2 . This can happen if the two parties sign
an enforceable, mutually exclusive trade contract. In this case, clearly i 2 and e2 should be zero.
At date 2, M1 and M2 are going to split the ex post surplus of S = R(i 1 ) + Q(e1 ). The firstorder conditions for the equilibrium investment levels are (again, assuming interior solutions) as
follows:
1 
R (i 1 ) = C1 (i 1 )
2
1 
Q (e1 ) = C2 (e1 ).
2

(5)
(6)

Let the solutions to (5) and (6) be 1 and e 1 . Comparing (5) and (6) with (1) and (2) shows that
the equilibrium investments will be lower than in the first-best solution. That is, there will be
underinvestment as a result of ex post surplus sharing.

13 For example, if C (i , i ) = i /2 + i /2 + c(i i )2 and the coefficient c is sufficiently large relative to other
1 1 2
1
2
1
2
parameters in the model, then M1 will always choose i 1 = i 2 = i. Then M1s benefit functions and cost function are
reduced to R(i), r (i; A1 ), and C(i) = i, exactly as in Hart (1995).
RAND

2003.

mss # Cai; AP art. # 04; RAND Journal of Economics vol. 34(1)

CAI

71

4. Equilibrium characterization
In this section I characterize the set of equilibria for the game for a fixed asset ownership
structure A. In Section 5 I find the optimal ownership structure that gives rise to the largest total
surplus. Throughout, I shall focus on pure-strategy equilibria only.
Given investment levels i = (i 1 , i 2 ) and e = (e1 , e2 ), if S s = R(i 1 ) + Q(e1 ) [r (i 2 ; A1 ) +
q(e2 ; A2 )] 0, then M1 and M2 will trade with each other and the split-the-net-surplus solution
says that M1 will get a surplus of r (i 2 ; A1 ) + (S s)/2 and M2 will get a surplus of q(e2 ; A2 ) +
(S s)/2. The payoff functions can be rewritten as

U1 = [R(i 1 ) + r (i 2 ; A1 )]/2 + [Q(e1 ) q(e2 ; A2 )]/2 C1 (i 1 + i 2 )

U2 = [R(i 1 ) r (i 2 ; A1 )]/2 + [Q(e1 ) + q(e2 ; A2 )]/2 C2 (e1 + e2 ).

(7)
(8)

Assuming interior solutions, the optimal investment levels satisfy the following first-order
conditions:
1 
R (i 1 ) =
2
1 
Q (e1 ) =
2

1 
r (i 2 ; A1 ) = C1 (i 1 + i 2 )
2
1 
q (e2 ; A2 ) = C2 (e1 + e2 ).
2

(9)
(10)

Denote the solutions to (9) and (10) by ( 1 , 2 )(A1 ) and (e1 , e 2 )(A2 ), respectively. Let U 1 (U 2 )
denote M1s (M2s) payoff with these investment choices in the trade equilibrium (if it exists).
For these investment choices to constitute a trade equilibrium, a necessary condition is that the
total surplus within the relationship exceed the sum of the two parties outside options. If this
condition fails for investment choices ( 1 , 2 )(A1 ) by M1 and (e1 , e 2 )(A2 ) by M2, one may expect
that there can exist a trade equilibrium in which both parties make more specific investments. A
bit surprisingly, this is not the case.
Lemma 2. A necessary condition for any trade equilibrium to exist is S s = R( 1 ) + Q(e1 )
[r ( 2 ; A1 ) + q(e2 ; A2 )] 0.
Proof. See the Appendix.
Lemma 2 says that to sustain a trade equilibrium, the surplus must be higher within the
relationship than outside it when each party invests optimally in specific and general investments.
Otherwise the two parties have to overinvest in specific investments to generate enough surplus
for trade to be mutually beneficial. But this is not sustainable in equilibrium because both would
shirk on specific investments and increase general investments, eventually leading to no trade.
Corollary 1. There can be at most one trade equilibrium, in which the investment choices are
{( 1 , 2 (A1 )), (e1 , e 2 (A2 ))}.
Proof. From Lemma 2, if a trade equilibrium exists, then S s 0. It follows that {( 1 , 2 (A1 )),
Q.E.D.
(e1 , e 2 (A2 ))} are dominant strategies.
To concentrate on interesting cases, I make the following assumption, which is slightly
stronger than saying that S s > 0.
Assumption 4. For any ownership structure, assume R( 1 ) > r ( 2 ; A1 ) and Q(e1 ) > q(e2 ; A2 ).
Assumption 4 is satisfied when R  (i) r  (i; A1 ) and Q  (e) q  (e; A2 ). This is because
(9) and (10) will then imply 1 2 and e 1 e2 , Assumptions 1 and 2 then imply that R( 1 )
R( 2 ) > r ( 2 ; A1 ) and Q(e1 ) Q(e2 ) > q(e2 ; A2 ). In other words, if investments are relationspecific in a marginal sense (as assumed in Hart (1995)), then it is beneficial for the two parties
to trade with each other once they make the investment choices given by (9) and (10).
Now let us consider a no-trade equilibrium. If M1 and M2 do not trade, then M1s payoff
is r (i 2 ; A1 ) C1 (i 1 + i 2 ) and M2s payoff is q(e2 ; A2 ) C2 (e1 + e2 ). Clearly, both parties will
RAND

2003.

mss # Cai; AP art. # 04; RAND Journal of Economics vol. 34(1)

72

THE RAND JOURNAL OF ECONOMICS

make zero specific investment, and the optimal general investments i 2 (A1 ) and e2 (A2 ) are given
by (3) and (4). A necessary condition for these investment choices to be a no-trade equilibrium
is that the total surplus generated within the relationship is less than the sum of the two parties
outside options. Similar to Lemma 2, this condition is necessary for the existence of any no-trade
equilibrium.
Lemma 3. A necessary condition for any no-trade equilibrium to exist is R(0)+ Q(0)[r (i 2 ; A1 )+
q(e2 ; A2 )] 0.
Proof. See the Appendix.
Lemma 3 says that to sustain a no-trade equilibrium, the surplus must be higher outside the
relationship than within it when each party invests optimally in general investment. Otherwise the
two parties have to overinvest in general investments to generate enough surplus for no trade
to be mutually beneficial. But this is not sustainable in equilibrium because both would shirk on
general investments to the point that trade with each other is preferred.
Corollary 2. There can be at most one no-trade equilibrium, in which the investment choices are
{(0, i 2 (A1 )), (0, e2 (A2 ))}.
Assumption 5. For any ownership structure, assume R(0) < r (i 2 ; A1 ) and Q(0) <
q(e2 ; A2 ).
To further simplify matters, I make the following assumption.
Assumption 6. For any ownership structure, assume that R(0)+ Q(e1 )[r (i 2 ; A1 )+q(e2 ; A2 )] 0
and R( 1 ) + Q(0) [r ( 2 ; A1 ) + q(e2 ; A2 )] 0.
This says that when one party invests only in general investment, the surplus within the
relationship is sufficiently large to justify trade as long as the other party invests as in the trade
equilibrium. Assumption 6 will be satisfied if R( 1 ) r ( 2 ; A1 ) and Q(e1 ) q(e2 ; A2 ) are sufficiently large (see Assumption 4) or if specific investments are sufficiently more productive than
general investments for both parties. Roughly speaking, Assumption 6 requires that both parties
be critical to the trade (i.e., sufficiently productive within the relationship). If Assumption 6 does
not hold, then the no-trade equilibrium can be the unique equilibrium for some parameter values,
in which case joint ownership will be inferior (see the discussion at the end of Section 5).14
Lemma 4. Under Assumption 6, both parties prefer a trade equilibrium to a no-trade equilibrium
(if the equilibria exist), that is, U 1 r (i 2 ; A1 ) C1 (i 2 ) and U 2 q(e2 ; A2 ) C2 (e2 ).
Proof. See the Appendix.
Define U 1 = [R( 1 ) + r ( 2 ; A1 )]/2 + [Q(0) q(e2 ; A2 )]/2 C1 ( 1 + 2 ) and U 2 = [R(0)
A1 )]/2 + [Q(e1 ) + q(e2 ; A2 )]/2 C2 (e1 + e 2 ). In words, U 1 is the payoff M1 would get if
M1s strategy and M2s strategy are mismatched: M1 invests as in the trade equilibrium while M2
invests as in the no-trade equilibrium. Clearly, we have U 1 > U 1 and U 2 > U 2 . The following
proposition characterizes the set of equilibria for the game.

r (i 2 ;

Proposition 1. Under Assumptions 4, 5, and 6,

(i) when either U 1 > r (i 2 ; A1 ) C1 (i 2 ) or U 2 > q(e2 ; A2 ) C2 (e2 ) holds, the trade equilibrium
is the unique equilibrium;
(ii) when r (i 2 ; A1 ) C1 (i 2 ) U 1 and q(e2 ; A2 ) C2 (e2 ) U 2 , both the trade equilibrium and
the no-trade equilibrium are equilibria for this game.
Proof. See the Appendix.
14 When Assumption 6 does not hold, there are also parameter values under which both trade and no-trade equilibria
exist, and there are some other parameter values under which there is no pure-strategy equilibrium at all. As Proposition
1 shows, co-existence of trade and no-trade equilibria can also happen under Assumption 6. For our purposes, I mainly
focus on the trade equilibrium.
RAND

2003.

mss # Cai; AP art. # 04; RAND Journal of Economics vol. 34(1)

CAI

73

5. Optimality of joint ownership

Based on the equilibrium characterization in Proposition 1, let us now turn to the comparison
of ownership structures. The conventional approach is to examine the first-order conditions of
(9) and (10): under conventional differentiability and convexity conditions, higher A1 leads to
greater i 2 and hence lower i 1 . A more powerful approach is to use the monotone comparative
statics developed by Milgrom and Shannon (1994), in which differentiability and convexity are
not necessary for the purpose of comparative statics. To use this approach, let us examine the
two parties payoff functions given by (7) and (8). Under Assumptions 1 and 3, U1 is supermodular in (i 1 , i 2 ) (i.e., having positive cross-partial derivatives) and has increasing differences in
(i 1 , i 2 ; A1 ) (i.e., marginal returns of i 1 and i 2 are both increasing in A1 ). Therefore, by Milgrom and Shannon (1994), the optimal (i 1 , i 2 ) is nondecreasing in A1 . Similarly, the optimal
(e1 , e2 ) is nondecreasing in A2 . Thus, we have the following proposition.

Proposition 2. In the trade equilibrium, asset ownership by one party decreases his/her incentives
to make specific investments and increases his/her incentives to make general investments.
Since general investments are a complete waste when M1 and M2 eventually trade with each
other, Proposition 2 reaches the opposite conclusion with the GHM approach: the more assets
one party owns exclusively, the fewer incentives to make specific investments.15 The intuition
for this result is simple. Because asset ownership increases the marginal productivity of general
investments, more assets lead to greater general investments. And because general and specific
investments are substitutes in my model, more assets lead to lower specific investments. In contrast,
when general and specific investments are complementary (as in GHM), owning more assets leads
to greater specific investments as well as general investments.
Although owning more assets discourages a party from making specific investments, it
benefits the owner individually. From the expression of U 1 , we can see that U 1 is increasing in
A1 and decreasing in A2 , because r (i 2 ; A1 ) is increasing in A1 and q(e1 ; A2 ) is increasing in A2 .
The idea is that owning more assets allows one party to hold up his opponent more strongly and
hence enables him to get a larger share of the total surplus.
Note that Proposition 2 holds in more general settings. For example, suppose one partys
investments affect the other partys values, and that investments by the two parties are complementary. Specifically, suppose R(i 1 , e1 ) and Q(e1 , i 1 ) are increasing in both elements and have
positive cross-partial derivatives, and r (i 2 , e2 ; A1 ) and q(e2 , i 2 ; A2 ) are increasing and have positive cross-partial derivatives in the first two elements. Assuming that M1 and M2 will always
trade at date 2, then the game has strategic complementarities. From Theorems 12 and 13 of
Milgrom and Shannon (1994), equilibrium investment levels can be ranked, i.e., (i 1 , i 2 ) and
(e1 , e2 ) move in the same direction. Moreover, (i 1 , i 2 ) ((e1 , e2 )) is nondecreasing in A1 (A2 )
in the two extreme equilibria (would be the same when the equilibrium is unique), which is what
Proposition 2 says.
Since the total net surplus in the trade equilibrium, R( 1 ) + Q(e1 ) C1 ( 1 + 2 ) C2 (e1 + e 2 ),
is increasing in the two parties specific investments and decreasing in their general investments,
the following result is a direct consequence of Proposition 2.
Proposition 3. In the trade equilibrium, joint ownership achieves the greatest total surplus among
all ownership structures.
Proof. Since under joint ownership r (i 2 ; {a1 , a2 }) = r (i 2 ; ) + 1 B12 and q(e2 ; {a1 , a2 }) =
q(e2 ; ) + (1 1 )B12 , the first-order conditions for equilibrium investments, (9) and (10),
are (1/2)R  (i 1 ) = (1/2)r  (i 2 ; ) = C1 (i 1 + i 2 ) and (1/2)Q  (e1 ) = (1/2)q  (e2 ; ) = C2 (e1 +
e2 ). If M1 owns both assets a1 and a2 , equilibrium investments are given by (1/2)R  (i 1 ) =
(1/2)r  (i 2 ; {a1 , a2 }) = C1 (i 1 + i 2 ) and (1/2)Q  (e1 ) = (1/2)q  (e2 ; ) = C2 (e1 + e2 ). Clearly, M2s
15 This same conclusion is reached by Rajan and Zingales (1998) in a restricted setting (e.g., only one party makes
investments).
RAND

2003.

mss # Cai; AP art. # 04; RAND Journal of Economics vol. 34(1)

74

THE RAND JOURNAL OF ECONOMICS

investment choice is identical under both ownership structures. By Proposition 2, M1 will make
more specific investment i 1 and less general investment i 2 under joint ownership than under M1
integration. Thus, joint ownership dominates M1 integration, and it dominates M2 integration
by symmetry. It is also easy to see that under nonintegration (M1 owns a1 and M2 owns a2 ),
both managers make more general investments and fewer specific investments than under joint
ownership, so nonintegration is dominated by joint ownership.
Joint ownership also dominates outside ownership. To see this, note that with an outside
owner (the conclusion holds more strongly for a larger number of outside owners), each player
gets his/her disagreement payoff plus one-third of the net surplus from trade. Hence, we have
U1 = r (i 2 ; ) + [R(i 1 ) + Q(e1 ) r (i 2 ; ) q(e2 ; ) B12 ]/3 C1 (i 1 + i 2 )
U2 = q(e2 ; ) + [R(i 1 ) + Q(e1 ) r (i 2 ; ) q(e2 ; ) B12 ]/3 C2 (e1 + e2 )
U0 = R(i 1 ) + Q(e1 ) r (i 2 ; ) q(e2 ; ) B12 ]/3.
The first-order conditions are (1/3)R  (i 1 ) = (2/3)r  (i 2 ; ) = C1 (i 1 + i 2 ) and (1/3)Q  (e1 ) =
(2/3)q  (e2 ; ) = C2 (e1 + e2 ). Compared with joint ownership, we can see that both M1 and
M2 will make fewer specific investments and more general investments, because private marginal
returns of specific investments are more diluted and private marginal returns of general investments
are better protected. Therefore, joint ownership dominates outside ownership.
Q.E.D.
The intuition for Proposition 3 is simple. Because owning more assets increases ones incentives to make general investment but decreases ones incentives to make specific investment, the
optimal ownership structure should take away assets from the exclusive control of either party as
much as possible. Joint ownership does exactly that. Outside ownership also takes away control
of assets from M1 and M2, but an outside owner will share some of the surplus created by M1 and
M2, which reduces M1s and M2s incentives to make specific investments. Moreover, diluted
ownership makes the payoffs of M1 and M2 depend more on their disagreement payoffs, which
further strengthens their incentives to make general investments (without the help of assets).
Next I want to determine the efficiency property of the trade equilibrium under asset ownership structures relative to some contracting solutions. Comparing (9) and (10) with (5) and (6),
we can see that for any asset ownership structure, 1 1 and e 1 e 1 because of the convexity
of C1 and C2 (i.e., substitution of specific and general investments). The equalities hold only
when 2 = e 2 = 0. Therefore, relative to the case without outside options, disincentives to make
productive specific investment are more severe under any ownership structure.
Proposition 4. In the trade equilibrium, specific investments will be underprovided under any
ownership structure even relative to the case without outside options.
Since only the second-best can be achieved when outside options are not available (e.g.,
through exclusive trade contracts), Proposition 4 implies that the outcomes in cases where the
two parties have outside options are third-best solutions. When outside options cannot be ruled
out, spending on general investments is equivalent to building an arsenal against the other party
in order to hold up her/him more strongly. Relative to the first-best solution, there are two sources
of underprovision of productive investment in my model: ex post surplus sharing (i.e., hold-up)
and ex ante spending on an arsenal. Joint ownership gives rise to the optimal third-best solution
because it reduces the incentives to engage in general investments.
So far we have assumed that general investments do not generate surplus within the relationship. This can be relaxed under reasonable conditions (and thus justify the term general).
Suppose general investment i 2 by M1 generates a surplus of H (i 2 ) within the relationship as well
as r (i 2 ; A1 ) outside the relationship. Assume that R  (i) > r  (i) H  (i). The first-best investment
choice for M1 satisfies the following first-order conditions (assuming interior solutions):
R  (i 1 ) = H  (i 2 ) = C1 (i 1 + i 2 ),
RAND

2003.

mss # Cai; AP art. # 04; RAND Journal of Economics vol. 34(1)

CAI

75

while the equilibrium conditions from (9) become

1 
1
1
R (i 1 ) = H  (i 2 ) + r  (i 2 ; A1 ) = C1 (i 1 + i 2 ).
2
2
2
By the monotone comparative statics of Milgrom and Shannon (1994), the equilibrium specific
investment i1 (A1 ) must be less than the first-best level i 1 and the equilibrium general investment
i2 (A1 ) must be greater than the first-best level i 2 . Similar to Proposition 2, i1 (A1 ) is decreasing
in A1 and i2 (A1 ) is increasing in A2 . Therefore joint ownership is still optimal, since it reduces
overinvesting in general investments and underinvesting in specific investment.
From Proposition 1, no trade can be an equilibrium outcome of the game. In fact, it is
possible that a no-trade equilibrium is the unique equilibrium of the game when Assumption 6
does not hold. With nonrenegotiable asset ownership, if M1 and M2 do not trade with each other,
joint ownership is the worst possible ownership structure, because incentives to make general
investments are minimal (no specific investments will be made under any ownership structure
if no trade is expected). This is obvious: if they are not going to trade with each other, why tie
themselves up with joint ownership? Thus, if M1 and M2 are to play the no-trade equilibrium,
then they will choose to avoid joint ownership at date 0. Since Assumption 6 requires that both
parties be critical (i.e., very productive within the relationship), trade is not likely to take place in
a business relationship that has a high degree of asset specificity but in which one party has high
productivity outside it. Of course, in such cases joint asset ownership will not be observed.

6. Conclusion
I have shown that joint ownership can serve as a mutual commitment mechanism to promote
relation-specific investments when players endogenously choose the level of specificity in their
relationships. The argument can be extended to a dynamic setting with multiple investment stages.
It can be shown that joint ownership provides the best dynamic investment incentives because
gains from deviations (i.e., to make general investments) are the smallest and punishments for
deviations are the largest (i.e., the no-trade equilibrium). Even with only two investment stages,
it is possible to achieve the first-best outcome.16
Given that the model predicts that joint ownership is optimal, one must wonder why it is not
a dominant ownership form in the real world. Several reasons can be given. First, some of the
key assumptions in the model may not hold in many contexts. Perhaps the most important is that
specific and general investments are substitutes. When investments are in human capital, and the
cost of investments is mainly time, this assumption seems plausible. Otherwise, it is less clear
whether the assumption makes sense. Of course, whether specific and general investments are
substitutes or complements depends on contexts and is ultimately an empirical issue. Assumption
6, which requires both parties to be sufficiently productive within the relationship, may be hard to
satisfy, too. Secondly, joint ownership may result in considerable governance costs. When multiple
parties jointly own a firm, they have to reach agreements on how to use the assets effectively.
The costs in making collective decisions can be quite substantial, especially when these owners
have diverse preferences (Hansmann, 1996; Cai and Milgrom, 1999). The difficulties in making
effective decisions may increase as the number of owners increases (Cai, 1997). These ex post
governance costs may outweigh the ex ante investment-incentives benefits under joint ownership,
making joint ownership less attractive. Finally, ownership structure may respond to important
issues other than providing investment incentives (Holmstrom and Roberts, 1998), and joint
ownership may lose its appeal on those issues. Asset ownership can be a useful instrument in
resolving agency problems (Holmstrom and Milgrom, 1994), and joint ownership may score
badly in this field. For example, if asset maintenance is important but noncontractible, then asset

16 The analysis of the dynamic setting requires that asset ownership is fixed and nonrenegotiable. See Cai (1999) for
details. Halonen (1997) and Woodruff (1997) analyze the dynamic optimality of joint ownership in different settingsin
particular, neither considers two-dimensional investments, and both models need infinite periods.
RAND

2003.

mss # Cai; AP art. # 04; RAND Journal of Economics vol. 34(1)

76

THE RAND JOURNAL OF ECONOMICS

value will be better protected under single ownership than under joint ownership. Moreover, asset
ownership provides monitoring incentives to the owners (Alchian and Demsetz, 1972), and joint
ownership can create additional free-riding probles in monitoring.
Appendix

Proofs of Lemmas 14 and Proposition 1 follow.

Proof of Lemma 1. Fix intial ownership structure A = (A1 , A2 ) and investment choices (i 1 , i 2 ) and (e1 , e2 ). Without asset
renegotition, M1s payoff is W1 = .5S + .5r (i 2 ; A1 ) .5q(e2 ; A2 ) C1 (i 1 , i 2 ). With asset renegotiation, his payoff is
U1 = .5S + .5u 1 (i 2 , e2 ; A) .5u 2 (i 2 , e2 ; A) C1 (i 1 , i 2 )
= .5S + .5{r (i 2 ; A1 ) + .5[s (i 2 , e2 ) s(i 2 , e2 ; A)]}
.5{q(e2 ; A2 ) + .5[s (i 2 , e2 ) s(i 2 , e2 ; A)]} C1 (i 1 , i 2 )
= .5S + .5r (i 2 ; A1 ) .5q(e2 ; A2 ) C1 (i 1 , i 2 ).
Therefore, W1 = U1 . Similarly W2 = U2 .

Q.E.D.

Proof of Lemma 2. If the investment levels derived from (9) and (10) indeed satisfy the condition in the lemma, then
the two parties will trade at date 2. Suppose otherwise, that is, S s < 0 and that there is a trade equilibrium with
{(i 1 , i 2 (A1 )), (e1 , e2 (A2 ))} such that S  s  = R(i 1 ) + Q(e1 ) [r (i 2 ; A1 ) + q(e2 ; A2 )] 0. I first show that it must
be the case in which S  s  = R(i 1 ) + Q(e1 ) [r (i 2 ; A1 ) + q(e2 ; A2 )] = 0. If not, then S  s  > 0. Define (i 1 , i 2 ) =
(1)(i 1 , i 2 )+( 1 , 2 ), where is a positive number sufficiently small so that R(i 1 )+ Q(e1 )[r (i 2 ; A1 )+q(e2 ; A2 )] > 0.
Since M1s payoff function is continuous in (i 1 , i 2 ) and ( 1 , 2 ) achieves the maximal payoff, (i 1 , i 2 ) yields a greater
payoff for M1 than (i 1 , i 2 ), which contradicts that (i 1 , i 2 ) is M1s equilibrium strategy. However, it is not possible for
{(i 1 , i 2 (A1 )), (e1 , e2 (A2 ))} to be a trade equilibrium when S  s  = R(i 1 ) + Q(e1 ) [r (i 2 ; A1 ) + q(e2 ; A2 )] = 0. If
S  = s  , then M1s payoff is r (i 2 ; A1 ) C1 (i 1 + i 2 ). But in this case, M1 should deviate to (0, i 2 ) and obtain a payoff of
r (i 2 ; A1 ) C1 (i 2 ), because there would be no trade between M1 and M2 after such a deviation and hence M1 would get
his outside option. Therefore, if the condition specified in the lemma is not satisfied, there will be no equilibrium with
trade.
Q.E.D.
Proof of Lemma 3. If the levels of general investments (i 2 (A1 ), e2 (A2 )) from (3) and (4) indeed satisfy the condition
in the lemma, then the two parties will not trade at date 2 with these investment choices. Suppose otherwise, that is,
R(0) + Q(0) [r (i 2 ; A1 ) + q(e2 ; A2 )] > 0 and that there is a no-trade equilibrium with {(0, i 2 (A1 )), (0, e2 (A2 ))}
such that R(0) + Q(0) [r (i 2 ; A1 ) + q(e2 ; A2 )] 0. Then at least one of the following must hold: i 2 (A1 ) > i 2 (A1 ),
e2 (A2 ) > e2 (A2 ). Suppose the first strict inequality holds. We show that M1 will want to deviate to i 2 (A1 ). If R(0) +
Q(0) [r (i 2 ; A1 ) + q(e2 ; A2 )] < 0, then deviating to i 2 (A1 ) is clearly profitable for M1 by definition of i 2 (A1 ). If
R(0) + Q(0) [r (i 2 ; A1 ) + q(e2 ; A2 )] 0, then M1s payoff from deviation is r (i 2 ; A1 ) C1 (i 2 ) + (S s)/2
Q.E.D.
r (i 2 ; A1 ) C1 (i 2 ) > r (i 2 ; A1 ) C1 (i 2 ).
Proof of Lemma 4. From Assumption 6, Q(e1 ) q(e2 ; A2 ) r (i 2 ; A1 ) R(0). Then,
U 1 [r (i 2 ; A1 ) C1 (i 2 )]

R( 1 ) + r ( 2 ; A1 ) r (i 2 ; A1 ) R(0)
+
C1 ( 1 + 2 ) [r (i 2 ; A1 ) C1 (i 2 )]
2
2
R(0) + r (i 2 ; A1 )
R( 1 ) + r ( 2 ; A1 )
={
C1 ( 1 + 2 )} {
C1 (i 2 )}
2
2
0.

The last inequality follows from the fact that ( 1 , 2 ) maximize [R(i 1 ) + r (i 2 )]/2 C1 (i 1 + i 2 ) and hence achieve higher
Q.E.D.
value for the maximand than (0, i 2 ). U 2 q(e2 ; A2 ) C2 (e2 ) can be proven in exactly the same way.
Proof of Proposition 1. It is easy to see that under Assumption 6, M1 investing ( 1 , 2 (A1 )) and M2 investing (e1 , e 2 (A2 ))
is indeed an equilibrium with trade. Since r (i 2 ; A1 ) C1 (i 2 ) is the largest possible payoff M1 can obtain in case he does
not trade with M2, if U 1 r (i 2 ; A1 ) C1 (i 2 ), then clearly M1 does not want to deviate to no trade. Similarly, M2 does
not want to deviate to no trade either. Since U 1 and U 2 represent the best possible outcome for M1 and M2 respectively
in case they trade, this is an equilibrium.
As long as r (i 2 ; A1 ) C1 (i 2 ) U 1 and q(e2 ; A2 ) C2 (e2 ) U 2 , then M1 investing (0, i 2 ) and M2 investing (0, e2 )
is an equilibrium with no trade. This is also straightforward because given M2s strategy, U 1 is the largest possible payoff
M1 can obtain in case he trades with M2. So under the two conditions, no trade is an equilibrium. r (i 2 ; A1 ) C1 (i 2 ) U 1
and q(e2 ; A2 ) C2 (e2 ) U 2 are also necessary for the no-trade equilibrium. Suppose r (i 2 ; A1 ) C1 (i 2 ) < U 1 . Since
RAND

2003.

mss # Cai; AP art. # 04; RAND Journal of Economics vol. 34(1)

CAI

77

R( 1 ) + Q(0) [r ( 2 ; A1 ) + q(e2 ; A2 ] 0, M1 clearly wants to deviate to ( 1 , 2 ) because M1 will get a payoff of U 1

from trading with M2, hence no trade cannot be an equilibrium. The conclusion of Proposition 1 follows immediately.
Q.E.D.

References
ALCHIAN, A.A. AND DEMSETZ, H. Production, Information Costs, and Economic Organization. American Economic
Review, Vol. 62 (1972), pp. 777795.
ALLEN, J.W. AND PHILLIPS, G.M. Corporate Equity Ownership, Strategic Alliances, and Product Market Relationships.
Journal of Finance, Vol. 55 (2000), pp. 27912815.
BAI, C.E., TAO, Z., AND WU, C. Revenue Sharing Contract and Control Right Arrangement: An Application to the Joint
Venture. Working Paper, Department of Economics, Hong Kong University of Science and Technology, 1999.
BAKER, G., GIBBONS, R., AND MURPHY, K.J. Relational Contracts and the Theory of the Firm. Quarterly Journal of
Economics, Vol. 117 (2002), pp. 3984.
BINMORE, K., SHAKED, A., AND SUTTON, J. An Outside Option Experiment. Quarterly Journal of Economics, Vol. 104
(1989), pp. 753770.
CAI, H. Bargaining Inefficiency with Multiple Players. Ph.D dissertation, Stanford University, 1997.
. A Theory of Joint Asset Ownership. Working Paper, UCLA, 1999.
. Renegotiation of Asset Ownership. Working Paper, UCLA, 2001.
AND MILGROM, P. Diversity in Committee Decisions. Working Paper, UCLA, 1999.
CHINA STATISTIC BUREAU. Statistical Yearbook of China. Beijing: China Statistics Publisher, 1999.
CHIU, S. Noncooperative Bargaining, Hostages, and Optimal Asset Ownership. American Economic Review, Vol. 88
(1998), pp. 882901.
DE MEZA, D. AND LOCKWOOD, B. Does Asset Ownership Always Motivate Managers? Outside Options and the Property
Rights Theory of the Firm. Quarterly Journal of Economics, Vol. 113 (1998), pp. 361386.
GOMES-CASSERES, B. Ownership Structures of Foreign Subsidiaries: Theory and Evidence. Journal of Economic
Behavior and Organization, Vol. 11 (1989), pp. 125.
GROSSMAN, S.J. AND HART, O.D. The Costs and Benefits of Ownership: A Theory of Vertical and Lateral Integration.
Journal of Political Economy, Vol. 94 (1986), pp. 691719.
HALONEN, M. A Theory of Joint Ownership. Department of Economics Discussion Paper no. 97/437, University of
Bristol, 1997.
HANSMANN, H. The Ownership of Enterprise. Cambridge, Mass.: Harvard University Press, 1996.
HART, O. Firms, Contracts, and Financial Structure. Oxford: Oxford University Press, 1995.
AND MOORE, J. Property Rights and the Nature of the Firm. Journal of Political Economy, Vol. 98 (1990), pp.
11191158.
AND . Cooperatives vs. Outside Ownership. NBER Working Paper no. W6421, 1998.
, B. The Firm as a Subeconomy. Journal of Law, Economics and Organization, Vol. 15 (1999), pp. 74102.
HOLMSTROM
AND MILGROM, P. The Firm as an Incentive System. American Economic Review, Vol. 84 (1994), pp. 972-991.
AND ROBERTS, J. The Boundaries of the Firm Revisited. Journal of Economic Perspectives, Vol. 12 (1998), pp.
7394.
AND TIROLE, J. Transfer Pricing and Organizational Form. Journal of Law, Economics and Organization, Vol.
7 (1991), pp. 201228.
LAMOREAUX, N.R. Partnerships, Corporations, and the Limits on Contractual Freedom in U.S. History: An Essay in
Economics, Law and Culture. Working Paper, UCLA, 2002.
MASKIN, E. AND TIROLE, J. Two Remarks on the Property-Rights Literature. Review of Economic Studies, Vol. 66
(1999), pp. 139149.
MILGROM, P. AND SHANNON, C. Monotone Comparative Statics. Econometrica, Vol. 62 (1994), pp. 157180.
OSBORNE, M.J. AND RUBINSTEIN, A. Bargaining and Markets. San Diego: Academic Press, 1990.
RAJAN, R.G. AND ZINGALES, L. Power in a Theory of Firm. Quarterly Journal of Economics, Vol. 113 (1998), pp.
387432.
SCHELLING, T. The Strategy of Conflict. Cambridge, Mass.: Harvard University Press, 1960.
SEGAL, I.R. AND WHINSTON, M.D. Exclusive Contracts and Protection of Investments. RAND Journal of Economics,
Vol. 31 (2000), pp. 603633.
TIROLE, J. Incomplete Contracts: Where Do We Stand? Econometrica, Vol. 67 (1999), pp. 741781.
WOODRUFF, C. Efficient Ownership in Bilateral Relationships with Repeated Specific Investments. Mimeo, Graduate
School of International Relations and Pacific Studies, UCSD, 1997.

RAND

2003.