International Business Review 24 (2015) 677689

Contents lists available at ScienceDirect

International Business Review

journal homepage: www.elsevier.com/locate/ibusrev

The termination of international joint ventures: Closure

and acquisition by domestic and foreign partners
Jose Mata a,*, Pedro Portugal a,b
a
b

Nova School of Business and Economics, Campus de Campolide, P1099-032 Lisboa, Portugal
Banco de Portugal, Lisboa, Portugal

A R T I C L E I N F O

A B S T R A C T

Article history:
Received 12 December 2013
Received in revised form 8 December 2014
Accepted 14 December 2014
Available online 23 January 2015

We study different modes of terminating international joint ventures, namely closure and acquisition,
and nd that different forces govern the two termination modes. Decisions regarding asset specicity
and the size of the venture affect the likelihood of closure, but not that of acquisition. In contrast, full
acquisition by one of the partners is related to history of the venture before the joint venture was formed,
to decisions made at the time of the creation with respect to equity split between partners, and to
subsequent changes of these initial decisions. Joint ventures that were created de novo are more likely to
be closed down than those that were previously fully owned by one of the parties. The proportion of
equity initially held by each partner and subsequent increases in this proportion increase the likelihood
of the venture being fully acquired by that partner.
2014 Elsevier Ltd. All rights reserved.

Keywords:
Joint ventures
Dissolution
Termination
Survival
Closure
Acquisition

1. Introduction
Many studies have shown that international joint ventures
(IJVs) confront high chances of termination (see Kogut, 1988 for a
seminal paper and Nemeth & Nippa, 2013 for a recent survey).
Termination can occur in different ways: by dissolving the venture
or by having one of the partners fully acquire the venture, and
these different modes of dissolution may be prompted by different
causes. For example, Kogut (1991) found that unexpected industry
growth increases the likelihood of acquisition by one of the
partners, but unexpected fall in industry shipments does not
increase likelihood of dissolution. Hennart, Kim, and Zeng (1998)
also found that the determinants of termination of JVs explain the
selling of JVs, but not their liquidation, while Chang and Singh
(1999) found that older rms shut down businesses, but younger
rms sell them. Furthermore, they found that businesses that have
entered by acquisition are more likely to exit by sell-off, a nding
that was also reached by Mata and Portugal (2000) in the context of
foreign rms. Even if the possibility of acquisition by different
partners is generally acknowledged, (e.g. Meschi & Riccio, 2008),
studies looking at dissolution of joint-ventures by acquisition

* Corresponding author. Tel.: +351 213801653.

E-mail address: jmata@novasbe.pt (J. Mata).
http://dx.doi.org/10.1016/j.ibusrev.2014.12.004
0969-5931/ 2014 Elsevier Ltd. All rights reserved.

typically focus on acquisition by the foreign partners (e.g. Puck,

Holtbrugge, & Mohr 2009), and the study by Hennart, Roehl, and
Zietlow (1999) remains as one of the few that have examined the
issue of which partners acquire and which divest the joint venture.
In this paper we analyze the process of joint venture
termination and consider the alternatives by which termination
can occur. We distinguish the factors that lead to closure from
those that lead to acquisition by the domestic and foreign partners.
Most of the studies mentioned above discuss a single set of
determinants of joint venture termination and report evidence that
while one type of termination is affected by the hypothesized
forces, the other is not. In contrast, we develop explicit hypotheses
for the factors that lead to the different types of termination. We
explicitly distinguish acquisition by foreign and domestic partners
and are able to identify forces that have opposite effects upon each
type of termination.
Specically, we are able to uncover the impact that ownership
split at the formation of the joint venture and at different points of
its evolution exerts upon the termination of joint ventures by
acquisition and upon which partners end up fully acquiring the
joint venture. First, ownership before the formation of the joint
venture matters; those joint ventures being created by a partial
acquisition of a previously fully owned business have a greater
likelihood of terminating by being bought back by the former
owner. Second, ownership arrangements at the time of the

678

J. Mata, P. Portugal / International Business Review 24 (2015) 677689

formation of the joint venture also matter for the likelihood of full
acquisition and for which partner ends up becoming the sole
owner of the venture. We nd that disproportionate equity splits
not only create instability, but also tip the acquisition process
toward the majority owner. This is further reinforced by
renegotiations over equity split that occur after the formation of
the joint venture. Joint ventures in which one of the partners
increases its equity share are more likely to be later fully acquired
by that partner.
Closure, in contrast, is related to the antecedents of the joint
venture, to joint venture size and to intangible investments made
at the inception of the joint venture. Joint ventures that are created
from scratch, those that are small, and those that do not rely much
on intangible assets are more likely to be shut down than those
that are created from a partial acquisition of a previously existing
rm, those that are large, and those that rely more on intangibles.
Our analysis carefully controls for the evolution of joint venture
termination over time. Even though some studies have controlled
for age in their analysis, to our knowledge, the way the likelihood
of termination of IJVs evolves over time has not been thoroughly
examined, except for the studies of Park and Russo (1996) and Park
and Ungston (1997). The fact that we distinguish between different
termination modes may exacerbate a problem that is common to
all the studies relying on age dependence, that is, the relationship
between the probability of an event and age: the evidence that the
probability of terminating a joint venture decreases over time can
be spurious and arise because the sample includes a proportion of
rms that do not confront the risk of termination via that particular
termination mode. We account for this possibility and, indeed, nd
that there is a non-negligible fraction of joint ventures that may
never terminate by any of the three modes considered. We nd
that termination by domestic acquisition is roughly constant over
time. In contrast, the chances of termination by acquisition by the
foreign partner increases over time, which ts well the view of
joint ventures as options to expand (Kogut, 1991). Finally, the
chances of termination by closure increase after a few rst years,
which is consistent with the view of multinationals as footloose
organizations (Gorg & Strobl, 2003).

& Tihanyi, 2007; Puck et al., 2009). Indeed, Zaheer, Hernandez, and
Banerjee (2010) report that those international acquisitions that
were preceded by a form of alliance between the acquired and
the acquiring companies show better returns than those that
were not preceded by such alliances. When uncertainty is high the
asymmetry of information is exacerbated and sequential investment and divestment become more likely (Reuer & Shen, 2004;
Folta & Miller, 2002).
However, not all acquisition joint ventures will be successes
and develop according to plan. While there is little knowledge
about what proportion of joint ventures terminate according to the
plan, one of the few studies on this matter (Makino, Chan, Isobe, &
Beamish, 2007) provides evidence that unanticipated termination
by far dominates with only 10% of the joint ventures being
terminated according to what had been planned. Steensma et al.
(2007) show that joint ventures in which there is conict between
partners are likely to become fully owned, especially if decision
power is markedly unbalanced between partners.
One way of terminating a joint venture that came about
through the partial acquisition of an ongoing rm is a full
reacquisition by its former full owner (Chi & Seth, 2009). When
partners of such a joint venture conclude that the match between
them is not good, a buyback is an easy way to end the venture, as it
amounts to a return to the previous position. It has been argued
that this would be a natural route for terminating joint ventures
partially acquired by a foreign party as this party would be
insufciently committed to the venture (Hennart et al., 1998;
Steensma et al., 2007). This argument extends naturally to any
acquiring partner and thus, rms that were previously wholly
domestic are more likely to return to their wholly domestic status,
while those that were previously fully owned by foreigners are
more likely to become wholly owned by foreigners again.
Hypothesis 1. Joint ventures that were created from an already
existing rm are more likely to be bought back by the original
owner than to be acquired by the joining party.

2.2. Contractual arrangementsInitial equity share

2. Modes and determinants of joint venture termination
International joint ventures can be terminated by shutting
down their facilities or by continuing operations under the full
ownership of either the domestic or foreign partner. These three
different outcomes are likely to be governed by different forces and
the attributes of the joint ventures are likely to exert disparate
impacts upon the probabilities of terminating in different ways.
2.1. Ownership before the formation of the joint venture
While most of the literature implicitly regards joint ventures as
new entities created by the partnering rms, the truth is that joint
ventures can be created by partial acquisition of equity in an
ongoing rm (Chen & Hennart, 2004; Chen, 2008). A common
reason for making a partial acquisition is to mitigate the
asymmetric information problem inherent to full-acquisitions
(Chen & Hennart, 2004; Reuer & Ragozzino, 2008). By keeping the
former owner with an interest in the rm following the partial
acquisition, his incentives to mis-represent the true state of the
rm and to engage in non-observable value reducing activities is
diminished.
To the extent that partial acquisitions are a way of overcoming
problems of asymmetric information associated with full-acquisitions, it is reasonable to expect that the successful development of
these acquisitions eventually terminates with the joining party
fully acquiring the ownership (Steensma, Barden, Dhanaraj, Lyles,

Conditions that are relevant for the longevity of joint ventures

include the initial contractual arrangements established between
partners. Different partners make different contributions to the
joint venture and these contributions are reected in the
agreements under which JVs are formed (Blodgett, 1991). Although
control of a joint venture cannot be taken to be identical to the
distribution of equity among partners (Zhang & Li, 2001), the initial
distribution of equity reects the distribution of bargaining power
among partners and control over the rm (Yan & Gray, 1994;
Mjoen & Tallman, 1997).
For the joint venture to be stable, the arrangements must be
such that all parties are satised with them. Uneven distributions
of equity may have costs for the stability of the joint venture,
because the smaller the share that one partner has in the joint
venture, the greater the likelihood that it will behave opportunistically (Inkpen & Currall, 2004), and free-ride on the other partner.
Joint ventures with uneven equity splits have been found to be
more prone to termination (Blodgett, 1992), which is likely to
occur due to the initiative of the dominant partner to avoid this
opportunistic exploitation. Furthermore, to the extent that large
equity shares reect a partners high contribution to the joint
venture, a large share in the venture is an indication that the rm
may more easily survive without the other party than vice-versa.
Therefore, if one partner holds a disproportionately high equity
share in the rm, the chances are that it will eventually acquire full
control of the rm.

J. Mata, P. Portugal / International Business Review 24 (2015) 677689

679

Hypothesis 2. The greater the share of initial equity held by one

partner, the more likely it is that this partner takes full control of
the rm.

Hypothesis 4. Joint ventures that were created from an already

existing rm are less likely to be shut down than greeneld joint
ventures.

2.3. Changes in contractual arrangementsIncreases in equity share

2.5. Efciency advantages and asset specicity

Many joint ventures adjust equity ownership splits over their

life and continue operating after these adjustments (Hennart et al.,
1999; Reuer, Zollo, & Singh, 2002; Chung & Beamish, 2010;
Iriyama, Shi, & Prescott, 2014). Changes in ownership structure
have been related to conditions that already existed at the time of
the formation of the joint venture, such as pre-formation
experience in several domains (Reuer et al., 2002) and balance
in the partners equity distribution (Chung & Beamish, 2010;
Iriyama et al., 2014). These changes also typically follow periods
of poor performance (Zhang & Li, 2001; Chung & Beamish, 2012)
and may be seen as an attempt to repair sentiments of
inequity and improve partners involvement in the joint
venture and subsequently improve performance (Brouthers &
Bamossy, 2006).
There is little evidence on what the real effect of changes in
ownership structure is upon subsequent performance of the joint
venture, but the available evidence offers little support to the idea
that adjustments lead to improved performance (Zhang & Li, 2001).
On the contrary, the ndings by Chung and Beamish (2010),
showing that the changes in the ownership structure are followed
by increases in the chances of dissolution, seem to indicate that in
most cases these changes in ownership were not able to overcome
the problems that were at the origin of the changes. Therefore, we
predict that changes of the initial ownership structure are followed
by increases in the chances of subsequent termination of the joint
venture.
Moreover, increasing ones own equity signals an increased
degree of commitment to the venture, while decreasing ones
equity indicates a lower involvement in the venture. Therefore, we
predict that those partners who have increased their equity share
in the joint venture will be more likely to fully acquire the venture
at a later point in time.

The source of efciency advantages is typically associated with

the ability of rms to develop rm-specic assets that cannot be
imitated by competitors and that provide the basis for their
competitive advantage (Wernerfelt, 1984; Barney, 1991). Firms
with such assets are often associated with those that undertake
large R&D activities and Delios and Beamish (2001) found that
R&D expenditures affect the survival of international joint
ventures.
Although activities such as R&D entail substantial spending on
physical facilities and equipment it is not the physical technology
that consists the basis of sustainable competitive advantage. As , (p.
110) puts it, Physical technology, whether it takes the form of
machine tools or robotics or complex information management
systems, is by itself imitable. One of the few classes of assets that
are not tradeable today are knowledge assets (Teece, 1998) and
these assets are thus the real basis of sustainable competitive
advantage. We therefore expect that rms that use knowledge
assets more intensively are less likely to be shut down.

Hypothesis 3. Partners who have increased/decreased its equity

share in a rm are subsequently more/less likely to take full control
of the rm.

2.4. Antecedents of the joint venture and closure

The fact that some joint ventures start as greeneld investments while others are created from an already existing rm has
implications for the longevity of the joint ventures and for the
mode of how these ventures are eventually terminated. To the
extent that joint ventures are also subject to the normal risk of
doing business, those that have been created from scratch are more
likely to be shut down than those that have already been operating
for some time. The fact that a rm has been partially acquired is an
indication that the acquiring rm saw value in the acquired unit.
Those productive units that existed prior to the formation of the
joint venture and that were thought to be sufciently valuable to
be kept in operation within the joint venture are likely to already
have a minimum level of efciency that greeneld units do not. As
such, these units are less likely to be unviable and are less likely to
be terminated by closure. If they are to be terminated, this value
makes it likely that the termination will be by passing ownership
to a single party, in very much the same way that rms that have
been acquired are more likely to be sold off than shut down (Chang
& Singh, 1999).

Hypothesis 5. Joint ventures that make a more intensive use of

knowledge assets experience a lower probability of being terminated by closure.
2.6. Firm size
Joint ventures face the same normal business risks as any other
rm. Virtually all studies on rm survival have found that large
rms tend to survive more than small ones (Dunne, Roberts, &
Samuelson, 1989; Audretsch & Mahmood, 1994; Mata & Portugal,
1994; Mitchell, 1994; Haveman, 1995; Sharma & Kesner, 1996). A
number of reasons have been suggested to account for this. The
rst two are related to superior efciency of large rms. On the one
hand, large rms may benet from economies of scale, thereby
boosting efciency (Audretsch & Mahmood, 1994). On the other
hand, size may also be a consequence of superior efciency. Even if
economies of scale are not present, rms with superior managerial
capabilities will have lower costs regardless of size and these lower
costs encourage rms to operate at a large scale (Lucas, 1978). A
third reason is related to constraints in accessing funds (Fazzari,
Hubbard, & Petersen, 1988). If there are cash constraints, rms for
which these are more severe will be forced to operate at a smaller
scale. Even if these constraints do not push smaller rms into a cost
disadvantage via economies of scale, they make it harder to survive
unexpected contingencies in comparison to competitors with
easier access to funds (Zingales, 1998). Large rms are also likely to
be more diversied than their smaller counterparts and this may
reduce risk and improve survival prospects. All of these reasons
suggest that size is likely to be related to the closure of rms, not
necessarily to the other modes of termination. Therefore, we
predict that
Hypothesis 6. Larger joint ventures experience a lower probability
of being terminated by closure than smaller joint ventures.

3. Empirical model
For analyzing the time pattern of the longevity of joint ventures
we rely on a class of statistical models known as duration analysis

680

J. Mata, P. Portugal / International Business Review 24 (2015) 677689

(Lancaster, 1990) or event history analysis (Allison, 1984). Before

going into the details of the specication of our empirical model, it
is convenient to examine two issues that have implications for our
modeling strategy: the rst is that joint ventures may be
temporary by design and that their durations may be determined
by partners when the joint ventures are formed. The second is that
some of the joint venture may never terminate, at least in one given
mode.
3.1. The stochastic nature of the longevity of joint ventures
An objection that is sometimes raised to the modeling of the
longevity of joint ventures as a stochastic process is the belief that
many joint ventures are temporary in nature and terminate when
the goal for which they were created in the rst place is achieved.
For example, Geringer and Killing (1988, p. 120) assert that quite a
few joint ventures are formed to solve what are, for at least one of
the partners, temporary problems. In their suggestions to
managers on how to deal with this issue the authors recommend:
Set up the venture on a project basis, with a denite termination
date. Gulati (1998, p. 307) expresses the opinion that many
successful alliances terminate because they are predestined to do
so by the parent rms at the very outset. To the extent that this is a
pervasive phenomenon in reality, the time patterns of terminations might well be determined by the duration clauses as
expressed in the joint-venture contracts. In this case, research
attempting to analyze the instability of joint ventures should be
directed to the reasons that lead partners to stipulate different
durations in their contractual arrangements, and ultimately to the
specicity of the projects developed by the joint venture.
However, although many joint venture contracts may indeed
have a duration clause, very little is known about the time-pattern
of these clauses in different contracts. In one of the few studies that
analyze data on duration clauses, Luo (2002) reported an average
duration of 24.6 years with a standard deviation of 3.5, which is in
sharp contrast with the much shorter durations generally observed
in empirical studies on joint venture terminations. These two
ndings can only be reconciled if no more than a minority of the
joint ventures do actually terminate according to the plan. Very
little is known too about the extent to which terminations come
about as a consequence of the exhaustion of the project that gave
rise to the joint venture or as a result of something that was
previously unexpected. The few studies on this matter indicate
that the latter is by far dominant, only 10% of the joint ventures
terminating according to what had been planned (Makino et al.,
2007) and only 6% terminating as a consequence of the
achievement of the joint-ventures objectives (Meschi & Riccio,
2008). This suggests that the timing of actual terminations does
indeed have an important non-deterministic element, and supports the statistical modeling of the termination of joint ventures.
3.2. Some JVs may never terminate in one given mode
The second issue that we examine is the possibility that some
termination modes may never be considered by some joint
ventures. This does not mean that some joint ventures will never
terminate at all; only that some will never choose some modes of
terminating.
Closure may never be considered by some joint ventures.
Utilities are a clear example, but others are possible: producers of
goods with low value content per unit of weight or produced in
highly specic facilities, such as mineral water, cement, or
providers of services for which a highly specic distribution
network is important are all unlikely to be shut down. If these rms
encounter problems, they may be traded, but actual closure is
highly unlikely. This, of course, applies within a foreseeable future.

It is highly unlikely that people will stop drinking water, but if a

close substitute for cement is discovered that can be produced at
much lower cost, it is not impossible that cement plants will
eventually disappear. Also, although there may be some industry
characteristics that make this more likely, it need not be the case
that all rms in a given industry do not shut down. It may be that
some rms in that industry consider shutdown as an option while
others do not.
Some other rms may never be fully acquired by foreign
partners. A rst reason is that there may be government
restrictions in some countries. Restriction may apply across the
board (e.g., limit of 74% foreign ownership in India) or to a
particular industry (e.g. airlines in most OECD countries, Conway,
Janod, & Nicoletti, 2005) or to companies considered to be of
national interest (e.g. France) or when acquisitions are considered
to be threats to national security (Moran, 2009). Even when there
are no such restrictions in the law, countries often seek to impede
acquisitions of some rms by foreigners. One such attempt
occurred with success in April 2007 when, despite the nonexistence of any foreign ownership limits on telecommunications
in Italy, the Italian Prime Minister Romano Prodi made a call on
Italian banks to help stave off the takeover of Telecom Italia by the
Mexican America Movil. The Spanish operator Telefonica acquired
a minority equity share in the Italian telecom operator, but the
government succeeded in achieving its goal of preventing Telecom
Italia from becoming foreign controlled. In other countries, even if
it is not legally required, it may be difcult to do business without a
local partner. A second reason for some joint ventures never
becoming fully foreign owned is a heavy reliance on geographically
distributed resources. Running a highly decentralized distribution
network, for example, requires constant monitoring and will be
best done by someone based in the country. While this need not be
done through a joint venture (other alternative arrangements may
be available, e.g., franchising), if an own distribution network is
preferred, the foreign partner may never consider operating it
itself. Typically, the advantage of foreign partners lies elsewhere,
and they will not wish to invest resources locally in areas that are
not related to their core advantage. The foreign partner may
consider nding another domestic partner or, if this is not feasible,
divest from the country.
Finally, domestic partners may never consider acquiring some
international joint ventures when the contribution of foreign
partners is typically highly specic. Foreign partners must possess
some specic assets (normally associated with knowledge of a
particular technology, or possession of rmspecic goodwill, in the
form of brands, trademarks, etc.) that enable them to compensate
for the liability of foreignness. While learning from foreign
partners is certainly possible (Anh, Baughn, Hang, & Neupert
2006), it may be very difcult for a domestic rm to replace the
contribution of the foreign partner, as the knowledge required to
eliminate foreign dependency is usually more difcult to acquire
than that required to eliminate dependency upon local partners
(Inkpen & Beamish, 1997).
3.2.1. The consequence of defective risks for the estimation of age
dependence
The presence in the sample of joint ventures that never
terminate in one given mode may bias the estimates of age
dependence for that given mode. Age dependence refers to the way
the hazard rates evolve over time. Negative (positive) age
dependence indicates that the phenomenon is increasingly less
(more) likely to be observed as joint ventures get older. Negative
age dependence may, however, be observed for spurious reasons if
the population under analysis contains unobserved (or unknown)
heterogeneous groups of rms, each exhibiting different levels of
risk. In this case, even if there is no genuine age dependence, that is,

J. Mata, P. Portugal / International Business Review 24 (2015) 677689

even if the risk confronting each rm is constant over time, the

observer may conclude that the risk is diminishing over time. This
occurs because rms in the group with higher risk will leave the
sample more rapidly than do those in the other group. The
remaining sample will, therefore, be made up of an increasing
proportion of rms with a low risk of exit.
To make these ideas clearer, consider a simple extreme case in
which we have 160 rms, in two groups of 80 rms each. One of the
groups confronts a constant hazard rate of 50% while the other
confronts no risk of exit (0% hazard rate). In the rst period,
40 rms from the rst group will exit, and an observer will
calculate the overall hazard rate to be 25% = 40/160. In the second
period, 20 rms (one half of the remaining rms in the rst group)
will exit again, and the observer will now calculate the hazard rate
to be 16.7% = 20/120. Therefore, if the analyst cannot identify the
group to which each JV belongs, he runs the risk of concluding for
negative age dependence, while this is not warranted by the data.
Note that the opposite is not true: one may never be led into the
conclusion of positive duration dependence due to the presence of
these defective risks. In our modeling, we take into account the
possibility of such defective risks.
3.3. A Statistical model for analyzing the termination of JVs over time
As our durations are observed only at year intervals, we will use
a simple discrete time duration model: the complementary log-log
(clog-log) model (Prentice & Gloeckler, 1978; Jenkins, 1995) (see
the appendix for a formal exposition of our model). Applying the
same approach as in the conventional proportional hazards model
(Cox, 1972), we specify an effect for age (the baseline hazard
function) and a multiplicative effect of the explanatory variables
upon survival, regression coefcients being interpreted as in
standard proportional hazards models.
The model can be estimated straightforwardly, by transforming
the duration data into binary outcomes, a procedure known as
episode splitting, and using maximum likelihood methods to t a
generalized linear model with binomial error and complementary
log-log link (Jenkins, 1995). In the likelihood function, a distinction
has to be made between joint ventures that were terminated and
those that continued to be run as joint ventures until the end of the
survey. To the former we can assign discrete durations. To the
latter all we know is that their duration exceeds a given limit, and
thus the observations are right censored. This same statistical
methodology applies to the three modes of terminating a joint
venture and three equations are estimated. In order to separate the
determinants of these three different exit modes, a clear
distinction has to be made between joint ventures that terminate
because the rm is shut down and those that are fully acquired by
domestic or foreign owners. We adopt the conventional assumption that the risks of exit from a joint venture are independent, and
when any one of these events occurs, we treat the observation as
censored in the other two exit mode equations.
3.3.1. Modeling age dependence
A common approach to the modeling of age dependence is to
assume that the baseline hazard function follows one given
distribution. Among continuous time models, popular choices are
the Weibull and the lognormal distributions, used by Park and Russo
(1996) in their seminal analysis of the age dependence of joint
venture termination. There are, however, serious potential drawbacks with an a priori use of this approach. Those distributions are
not general enough to accommodate sufciently varied patterns of
age dependence, and the choice of an inappropriate distribution to
model the baseline hazard function may seriously endanger our
conclusions about the nature of the evolution of the hazard rates
over time. In addition, as the most common distributions are not

681

nested with each other, it is not easy to choose between them. These
problems are compounded when duration data are grouped into
time intervals. As before, if the discrete nature of the duration
variable is not taken into account, the estimation procedure will lead
to inconsistent regression coefcient estimates and to a misleading
picture of age dependence.
In our model we avoid the imposition of severe distributional
assumptions and use a rather exible specication that models the
hazard rate as a polynomial function of age (see, e.g., Kennan,
1985; Ham & Rea, 1987). Estimation proceeds from a rst-order
polynomial by adding as many higher order terms as necessary.
The process stops when higher-order terms are found not to be
signicant. This allows the hazard function to have as many
inection points as is most appropriate to t the data well, without
the parametric constraint that predetermined distribution functions would impose.
3.3.2. Handling terminations that will never occur
To incorporate the possibility of defective risks, that is, the
possibility that some units may survive forever, we redene the
survival function so that the survival probability includes the
(unknown) proportion of long-term survivors, which do not
terminate in one given mode with certainty, plus the proportion of
susceptible rms, multiplied by their corresponding probability
of remaining a joint venture until each moment. Models of this
type have been used with a single risk in the duration analysis of
the acquisition of new products (Anscombe, 1961), deaths by AIDS
(Struthers & Farewell, 1989), criminal recidivism (Schmidt & Witte,
1989), timing of births (Heckman & Walker, 1990), and job stability
(Yamaguchi, 1992). Generalization to multiple independent risks is
straightforward (Addison & Portugal, 2003), with one additional
parameter to be estimated for each mode of termination (the
probability that joint ventures never terminate via that mode).

4. Data
The data used in this paper were obtained from an annual
survey (Quadros de Pessoal, hereinafter QP) which is conducted by
the Portuguese Ministry of Employment. Unlike most databases
employed in the analysis of alliance and foreign direct investment,
our data are not restricted to the largest companies, and include
rms of all sizes, as the survey covers all rms employing paid
labor in Portugal. We worked with the original raw data les from
1982 to 2009, which include over 100,000 rms in each year.
The survey has two characteristics that make this data set an
advantageous source for analyzing the survival of joint ventures.
First, the survey has a longitudinal dimension, i.e., rms are
identied by a unique number allowing them to be followed over
time. Second, the survey records the share of equity held by nonresidents, which we use for identifying joint ventures.
We are concerned here with foreign joint ventures, that is, rms
that have considerable (but not total) foreign equity participation.
Because of this we restricted our analysis to those rms having a
foreign participation between 10% and 90%. The 10% threshold is
usually employed to distinguish foreign direct investment from
portfolio investment, as this is the threshold that normally grants
the right to designate one board member. Using this criterion, we
were able to identify 3697 newly formed joint ventures, which
comprise our sample. An important limitation in the database is
that we do not know the identity of the rms owners. This is
unfortunate because we are not able to identify the number of
partners in the joint venture nor are we able to identify joint
ventures in which all partners are foreign companies. Moreover,
we are not able to trace the acquisition of a share held by one
foreign rm that is sold to another foreign rm and, similarly, we

J. Mata, P. Portugal / International Business Review 24 (2015) 677689

682

cannot identify the transfer of ownership if both the buyer and the
seller are domestic rms.
Our denition of entry involves the creation of a new equity
alliance between foreign and domestic partners. These new JVs
may be created in three different ways. The rst involves the
creation of a new legal entity. The second is by having a foreign
party acquire a stake in an already existing rm that was until that
moment entirely held by domestic owners. The third is by having a
domestic partner acquire a stake in an ongoing rm that was
previously entirely owned by foreign owners. Symmetrically, we
identify three ways in which an equity joint venture may
terminate: by shutting down the rm, by being totally acquired
by domestic partners, or totally acquired by foreign partners.
We were able to identify the longevity of joint ventures because
rms are identied in the survey by numbers, which are assigned
sequentially when they rst report to the survey. The moment at
which joint ventures are formed was identied by comparing
rms identiers over the years. Greeneld joint ventures, i.e., joint
ventures that did not exist as independent legal entities prior to
their formation were located by comparing the rms number with
the highest identication number in the le in the previous year.
The creation of joint ventures when such creation came about by
acquisition was identied by locating the rst year in which a
previously existing rm exhibited a percentage of foreign equity
between 10% and 90%. Our analysis includes joint ventures that
were formed during the period 19832008. This period was chosen
largely based on the availability of data. It is, however, convenient
to stop in 2008, so that our estimates are not contaminated by the
effect of the 2009 international crisis. We use data from 2009 only
for checking that rms classied as closures are indeed not active.
To compute termination by closure we located the moment
when rms cease to permanently report to the survey. To be on the
safe side in computing life spans with such a large database, we
required that a rm be absent from the le for at least two years in
order to be classied as a closure. In our analysis we use data only
until 2008, with information from year 2009 being used only for
the purpose of classifying observations as complete or censored.
We do not know the longevity of the JVs that are still active in 2008,
and therefore our duration measure is right-censored at 2008. We
use information from 2009 to classify rms that are not in the le
in 2008 as censored or complete. Firms are classied as censored at
2008 if they appear again in 2009, and as complete if they are also
absent from the 2009 le. For identifying termination of joint
ventures by acquisition we located the rst year in which the rms
foreign equity is outside the 1090% interval. It is, of course,
possible that foreign equity goes outside the 1090% interval in
one year and goes inside that interval in a subsequent year. When
such multiple episodes of joint ventures occur, to avoid selection
issues, we use only the rst of such episodes.
Table 1 shows the numbers of joint ventures created in each of
the three ways above. For each of these groups Table 1 also shows
the number of those that terminated via the three modes, along
Table 1
Formation and termination types of joint ventures.
Termination type

Continuation Total

Total
Closure Domestic Foreign
acquisition acquisition termination
Formation
type
Greeneld
642
Formerly
510
Domestic
Formerly
42
Foreign
1194
Total

578
624

176
189

1396
1323

356
459

1752
1782

43

35

120

43

163

1245

400

2839

858

3697

with those that were still active as joint ventures at the end of our
observation period. There are a total of 3697 joint ventures being
created, of which only 858 are still operating as joint ventures by
the end of the period. This may look like a very high exit rate, but
keep in mind that, while some of our joint ventures are observed
for only one year, others are observed for a period of 26 years.
Second, while the literature has focused on acquisitions by
foreign partners (e.g. Puck et al., 2009), this is the least common
way of terminating a joint venture in our sample (only 15% of
the total number of terminations). The most common way of
terminating a joint venture is by acquisition by the domestic
partner, closely followed by closure. Third, greeneld joint
ventures and those that were previously owned by domestic
partners are roughly of equal importance, while joint ventures that
are formed from a previously fully foreign owned rm represent a
minor fraction (4%) of the total.
Fourth, the number of closures (1194) is smaller than the
number of greeneld joint ventures (1752), and this holds even if
we subtract the number of greeneld joint ventures that are still
operating as joint ventures by the end of the period (net of 1396).
Similarly, the number of joint ventures that are created from a
previously domestic owned company (1782) is also greater than
the number of joint ventures that eventually become wholly
domestic owned (1245) and this also holds if we subtract the
number of still ongoing businesses (1323). In contrast, the number
of joint ventures that are formed from an existing wholly foreign
owned rm (163) is much smaller than the number of those that
eventually become wholly owned by foreigners (400).
4.1. Variables
We use the information in our data set to develop measures for
the variables outlined in Section 3 that account for the survival of
rms.
We measure the antecedents of the joint venture with two
dummies indicating the rm status prior to becoming a joint
venture. One dummy indicates whether the rm was previously
wholly foreign owned, while the other indicates whether it was
previously wholly owned by domestic owners. The omitted
category includes rms that were created simultaneously with
the creation of the JV.
We also wish to include a measure of the share of equity held by
foreigners. While the foreign share can vary on a continuous scale
between 0 and 100, earlier studies (e.g. Franko, 1989) typically
used categories such as minority, equal stake, majority owned joint
ventures to account for partners control over joint ventures. More
recently, Dhanaraj and Beamish (2004) suggested that equity share
should be used to explain survival of international joint ventures
rather that these broad categories. The variable used in this work is
the logarithm of equity share.
We account for changes in the ownership split with two
dummy variables that indicate whether the last change in equity
was in favor of reinforcing the domestic or the foreign position in
the joint venture. The rst of our dummy variables (Foreign Equity
Increase) takes the value 1 from the year when foreign equity
increases until the moment foreign equity is decreased if this ever
happens. Similarly, the dummy variable Domestic Equity Increase
takes the value 1 from the year when domestic equity increases
until the moment domestic equity is decreased, if ever. These
variables can both take the value 0 if equity has not changed, but
they cannot both take the value 1. As a robustness check we
dened another pair of dummy variables that take the value 1 only
in the year when foreign or domestic equity increases, as Chung
and Beamish (2010) report that the effect of changes in equity is
particularly prominent immediately after the change in equity.
Results remain qualitatively the same.

J. Mata, P. Portugal / International Business Review 24 (2015) 677689

683

Table 2
Descriptive statistics.

Formerly domestic
Formerly foreign
Initial equity share
Foreign equity increased
Domestic equity increased
Knowledge intensive
Size
Foreign presence

(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)

Mean

Standard
deviation

0.489
0.042
0.776
0.104
0.081
0.357
2.432
0.149

0.533

1.641
0.170

Correlations
(1)

(2)

(3)

(4)

(5)

(6)

(7)

0.206
0.116
0.022
0.004
0.031
0.276
0.018

0.075
0.025
0.006
0.015
0.118
0.059

0.112
0.015
0.032
0.155
0.104

0.102
0.004
0.102
0.086

0.022
0.089
0.035

0.013
0.089

0.251

We measure the propensity to develop knowledge assets by

including a dummy that indicates whether the sector in which the
rm operates is knowledge intensive. This is dened aggregating
industries classied as high-technology and medium-high-technology in manufacturing and as Knowledge Intensive Services. This
classication is provided by Eurostat (available at http://epp.
eurostat.ec.europa.eu/cache/ITY_SDDS/en/htec_esms.htm)
and
has been used in several academic studies (e.g. Godin, 2004;
Garca-Manjon & Romero-Merino, 2012).
Size is measured here by the logarithm of the number of
persons in the rm. Earlier reported evidence on the effect of rm
size on the survival of rms suggests a very robust negative effect
(Mitchell, 1994; Haveman, 1995; Sharma & Kesner, 1996). The
relationship between size and the likelihood of divestiture is less
obvious and the empirical studies that have analyzed exit by
divestment have not found any signicant relationship between
divestment and the size of rms (Schary 1991; Mitchell, 1994).
In addition we control for the extent of foreign presence in the
industry in which entry is attempted. The impact of previous
presence of foreign rms upon the survival of the new foreign
owned rms has been scrutinized in several studies (e.g.,
Mascarenhas, 1992; Mitchell, Shaver, & Yeung, 1994; Shaver,
Mitchell, & Yeung, 1997). Most of the arguments developed in this
line of research are of a time-series nature, comparing the positions
of rst-movers with those of late-movers. In our case variation is
largely cross-sectional. We thus expect previous foreign presence
to signal the presence of those characteristics, such as advertising
and technological intensity, which make foreign survival more
likely. These are characteristics that we are not able to observe
directly, but that are also related to the previous presence of
foreign rms in the market (Dunning, 1993; Caves, 1996). We
include previous foreign presence in the industry as a means of
controling for these unobserved industry characteristics, which
may be related to the survival of foreign rms. Foreign presence is
measured by the proportion of employment in the industry that is
accounted for by foreign rms.
Summary statistics for the independent variables above and
correlations between them are shown in Table 2.
4.2. Patterns of termination of joint ventures
Fig. 1 shows our estimates of the survival rates for each of the
three termination modes discussed above. Although our data cover
a span of 26 years (19832008), and all the available data are used
in the regressions, the table displays the survival rates for the rst
18 years only. As age increases, the number of observations falls for
two reasons. One reason is that there are joint ventures that
terminate. Only 78% of the total number of joint ventures in our
sample are able to make it through the following year, only 50%
remain as joint ventures after the third year, and fewer than 10%
remain after the thirteenth year. The second reason is that not all
joint ventures are observed over the same number of years. While

those that are formed in 1983 are observed for 26 years, those that
are formed in 2003 are observed for only 6 years. These two effects
compound to produce smaller samples for older ages, and thus less
precise estimates. Consequently, for older ages the precision of the
estimates is lower than for younger ones. The plots in Fig. 1 show
that the different functions seem to converge to values that are
clearly above zero, unlike what happens to the overall termination
rate. This suggests that there may be a fraction of joint ventures
that, indeed, never terminate in one given mode. Second, the
concavity of the different functions seems to be quite different. In
particular, termination by acquisition by the foreign partner is
visibly less concave, which suggests that the probability of
termination increases with age. These plots also show that these
survival rates for the different termination modes converge to
some value above zero, while overall survival converges toward
zero. This suggests that it is possible that some of the joint ventures
do not confront termination by one specic mode and that our
modeling that takes into account such a possibility is indeed
appropriate.
Fig. 2 goes into greater detail in the analysis of the chances of
termination over time and shows the observed empirical hazard
rates of a joint venture being terminated by closure, by acquisition
of the domestic partner, and by acquisition of the foreign partner,
respectively. The observed patterns of the hazard function are
clearly different for the three types of exit. The hazard function
decreases for termination by closure, at least during the rst years,
to increase at a later point in time. The hazard rates of being
acquired by a domestic partner are decreasing over time, although
they are somewhat irregular. Finally, the hazard rates of being
acquired by the foreign partner are increasing over time, again
with greater irregularities for the later periods of life.
The regressions presented in the next section also allow us to
take into account the effects of the determinants of termination
upon these hazard rates and the possibility that some joint
ventures never terminate in a given mode, as discussed above.
5. Results
Results of our regression analysis are reported in Tables 3 and
4. Table 3 shows the results of the conventional complementary
log-log model, while the results in Table 4 take into account
the possibility that there exists a fraction of joint ventures
that does not confront the risk of being terminated via the
mode under analysis. For each model we report a specication
with only a linear term on age, and one with a quadratic term
as well.
The results give partial support to our rst hypothesis, which
posited that the acquiring partner is the one that leaves the rm
more easily if the joint venture is to be dissolved. The coefcient
associated with the Formerly Foreign variable in the foreign
acquisitions equation is positive, indicating that rms that were
once fully owned by foreign owners are more likely to become fully

J. Mata, P. Portugal / International Business Review 24 (2015) 677689

684

Closure

0.75

0.75

Survival rate

Survival rate

Overall

0.5
0.25

0.5
0.25
0

0
0

10

12

14

16

18

12

14

16

18

14

16

18

Age

Domesc Acquision

Foreign Acquision

0.75

0.75

Survival rate

Survival rate

10

Age

0.5
0.25

0.5
0.25

0
0

10

12

14

16

18

Age

10

12

Age
Fig. 1. Survival rates by different types of termination.

owned by foreigners again. Similarly, rms that were fully owned

by domestic owners before the joint venture are more likely to
return to full domestic ownership, but in this case the effect is not
statistically signicant.

Our next set of ndings has to do with the ownership

arrangements, as reected in the share equity held by foreign
partners. We nd that the initial split of equity between domestic
and foreign partners is clearly relevant to the probability of

Overall

Closure

0.25

0.12
0.1
Hazard rate

Hazard rate

0.2
0.15
0.1
0.05

0.08
0.06
0.04
0.02

0
3

-2

13

18

-2

Age

Domesc Acquision

13

18

0.08
0.07

0.1

0.06

0.08

Hazard rate

Hazard rate

18

Foreign Acquision

0.12

0.06
0.04

0.05
0.04
0.03
0.02

0.02

0.01

0
-2

13

Age

0
3

13

18

-2

Age

8
Age

Fig. 2. Hazard rates for different types of termination.

J. Mata, P. Portugal / International Business Review 24 (2015) 677689

685

Table 3
Regression resultscloglog model.
Variables

(1)

(2)

Closure
Formerly domestic
Formerly foreign
Initial equity share
Foreign equity increased
Domestic equity increased
Knowledge intensive
Size
Foreign presence
Age
Age squared
Constant
Log L

0.111* (0.062)
0.058 (0.163)
0.011 (0.057)
0.206* (0.120)
0.148 (0.117)
0.121* (0.062)
0.197*** (0.022)
0.211 (0.198)
0.061*** (0.010)
1.657*** (0.076)
4016.3

(3)

(4)

Domestic acquisition
0.109* (0.062)
0.059 (0.163)
0.009 (0.057)
0.187 (0.121)
0.170 (0.118)
0.121* (0.062)
0.196*** (0.022)
0.208 (0.198)
0.098*** (0.024)
0.003* (0.002)
1.591*** (0.085)
4014.9

0.035 (0.061)
0.094 (0.161)
0.312*** (0.051)
0.234** (0.110)
0.062 (0.112)
0.092 (0.059)
0.030 (0.020)
0.745*** (0.198)
0.034*** (0.009)
2.374*** (0.076)
4190.1

(5)

(6)

Foreign Acquisition
0.034 (0.061)
0.094 (0.161)
0.313*** (0.051)
0.243** (0.111)
0.053 (0.113)
0.092 (0.059)
0.030 (0.020)
0.746*** (0.198)
0.017 (0.024)
0.001 (0.002)
2.404*** (0.087)
4189.8

0.028 (0.111)
0.367* (0.194)
1.306*** (0.133)
0.566*** (0.144)
0.400* (0.211)
0.081 (0.105)
0.152*** (0.032)
0.630** (0.273)
0.039*** (0.013)
3.482*** (0.137)
1734.4

0.029 (0.111)
0.360* (0.194)
1.306*** (0.133)
0.532*** (0.145)
0.441** (0.212)
0.082 (0.105)
0.152*** (0.032)
0.616** (0.274)
0.092** (0.038)
0.003 (0.002)
3.594*** (0.158)
1733.2

Standard errors in parentheses. Signicance is indicated as follows *** p < 0.01, ** p < 0.05, * p < 0.1. Number of observations is 14,340. Number of zero outcomes is 1194 for
closure, 1245 for domestic acquisition, and 400 for foreign acquisition.

subsequent acquisition by these partners. The greater the share of

equity that one partner holds in a joint venture, the more likely it is
that this partner eventually becomes the full owner of the rm.
Partners with very small stakes in the rm may tend to free-ride on
the dominant partner, and this partner may nd it necessary to
dissolve the joint venture. In addition, an even split of ownership at
the beginning of the joint venture also may reect disproportionate contributions from both parties to the success of the venture,
and the party with the greater involvement may nd it easier to put
in the extra effort required to fully operate the venture. We also
nd that foreign equity share is not important to the probability of
closure, which seems to indicate that a greater or smaller foreign
share does not really increase overall efciency of the joint
venture.
Changes in equity split seem to have a very clear effect upon the
probability of dissolving the joint venture by acquisition. Increasing the share of foreign equity has a positive and signicant effect
upon the probability of being acquired by the foreign partner and a
negative and equally signicant effect upon the probability of
being acquired by the domestic partner. Decreasing the share of
foreign equity has a negative and signicant effect upon the
probability of being acquired by the foreign partner, but does not
increase the chances of being acquired by the domestic partner. If
anything, decreasing the share of foreign equity leads to an
increased probability of closure.
The results clearly corroborate our fourth hypothesis. Joint
ventures formed from an already existing rm are less likely to be
shut down. This holds true regardless of whether the rm was

previously domestic or foreign owned, although the coefcient

associated with being previously domestic owned is not signicantly different from zero. However, the two estimated coefcients
are remarkably similar in magnitude and the hypothesis that they
are identical cannot be rejected. Firms that were already operating
are likely to have value independently of the joint venture and are,
therefore, less likely to see their operations discontinued.
Being in knowledge intensive industries is found to affect the
closure of joint venture as expected. Joint ventures in a knowledge
intensive industry are less likely to be shut down. In contrast, joint
ventures in such industries are more likely to be acquired by the
domestic or foreign partners. We did not nd any signicant effect
of knowledge intensity upon acquisition by one of the partners. We
hypothesized that to the extent that high knowledge intensity
signals the presence of highly specic assets, joint ventures would
be more difcult to be dissolved by acquisition by one of the
partners. The evidence, however, does not support this hypothesis.
It may be that high knowledge intensity also makes the rm more
valuable for acquisition and that this effect mayoffset the other.
The available data do not allow us to go deeper into this issue.
We obtained the expected pattern of results for the relationship
between size and closure. Larger rms are less likely to be closed
down. The above relationship between size and closure conrms
our predictions. Larger rms, being likely to be more efcient and
have fewer constraints, are more apt to survive. We also found that
larger rms are more likely to be acquired by foreign partners and
less likely to be acquired by domestic partners (although the effect
is not signicant in the latter case).

Table 4
Regression resultscloglog model with defective risks.
Variables

(1)

(2)

0.154** (0.073)
0.149 (0.183)
0.007 (0.066)
0.123 (0.142)
0.240* (0.137)
0.129* (0.068)
0.239*** (0.026)
0.063 (0.251)
0.004 (0.015)
1.260*** (0.097)
0.291* (0.035)
4008.06

(4)

Domestic acquisition

Closure
Formerly domestic
Formerly foreign
Initial equity share
Foreign equity increased
Domestic equity increased
Knowledge intensive
Size
Foreign presence
Age
Age squared
Constant
Prob. never fail
Log L

(3)

0.150** (0.073)
0.168 (0.183)
0.008 (0.066)
0.083 (0.144)
0.265* (0.138)
0.127* (0.068)
0.237*** (0.026)
0.064 (0.251)
0.051* (0.029)
0.004* (0.002)
1.184*** (0.101)
0.295* (0.032)
4006.3

0.050 (0.070)
0.091 (0.178)
0.354*** (0.058)
0.276** (0.122)
0.010 (0.125)
0.105* (0.064)
0.042* (0.023)
0.735*** (0.225)
0.011 (0.019)
2.204*** (0.098)
0.211* (0.050)
4187.2

(5)

(6)

Foreign acquisition
0.050 (0.070)
0.092 (0.179)
0.355*** (0.058)
0.273** (0.123)
0.012 (0.125)
0.106* (0.064)
0.042* (0.023)
0.730*** (0.227)
0.006 (0.027)
0.001 (0.002)
2.185*** (0.118)
0.218* (0.052)
4187.1

0.053 (0.122)
0.540** (0.248)
1.372*** (0.142)
0.551*** (0.159)
0.456** (0.224)
0.107 (0.111)
0.165*** (0.035)
0.658** (0.316)
0.071*** (0.020)
3.209*** (0.178)
0.294** (0.093)
1732.6

0.051 (0.121)
0.522** (0.251)
1.361*** (0.142)
0.537*** (0.159)
0.466** (0.223)
0.106 (0.111)
0.164*** (0.035)
0.656** (0.312)
0.097** (0.041)
0.002 (0.003)
3.303*** (0.224)
0.267*** (0.117)
1732.3

Standard errors in parentheses. Signicance is indicated as follows *** p < 0.01, ** p < 0.05, * p < 0.1. Number of observations is 14,340. Number of zero outcomes is 1194 for
closure, 1245 for domestic acquisition, and 400 for foreign acquisition.

686

J. Mata, P. Portugal / International Business Review 24 (2015) 677689

Finally, previous foreign presence in the industry is important

to termination by acquisition, but not for closure. This variable
has opposite signs in the two acquisition equations. Foreign
acquisition is more likely in industries with greater foreign
presence, while domestic acquisitions are less likely in these
industries. We expected that closures of joint ventures would
also be less likely in industries with a strong foreign presence, on
the grounds that the industries having a strong foreign presence
would be those that are more conducive to the survivability of
international joint ventures. We did not nd such and effect,
however.
Our exible specication of the effect of age allowed us to
uncover a rather disparate effect of age upon termination modes.
We found a U-shape pattern for closure, an upward sloping effect
for acquisition by foreign partners, and no effect of age for the
acquisition by domestic partners. In no case did age attract a
negative and signicant coefcient. We report only a linear and a
quadratic specication for each of the exit modes, as cubic terms
were never signicant. The quadratic term was signicant only in
the closure equation and the estimates imply that the minimum of
the hazard rate is at the age of 6.4 years.
Table 4 reports that results of estimating a model in which we
account for the possibility that some joint ventures never
terminate in one given mode. The probability that this happens
is estimated to be sizable, with a magnitude of around 30% for
closure and foreign acquisition and 21% for domestic acquisition.
Note that this result also means that very few of the joint ventures
are sheltered from the overall risk of terminating. The overall
probability of never terminating is given by the product of the
probabilities that joint ventures never terminate in each mode.
Given our estimates of these probabilities, the overall probability
that joint ventures never terminate is a mere 2%.
Taking into account the possibility that some joint ventures
never terminate in one given mode produces somewhat larger
coefcient estimates, except those concerning the effect of age, but
does not change the qualitative results. As expected, the impact of
controlling for the possibility that some joint ventures do not
confront the risk of terminating in one particular mode attenuates
the negative effect of age. In the complementary log-log model, the
coefcient of Age is negative and signicant for the closure and
domestic acquisition equations (Table 3, columns 1 and 3).
Accounting for the possibility of no termination produces age
coefcients that are not signicant in both of the equations (see
Table 4, columns 1 and 3). The effect of age is positive in the foreign
acquisition equation. For this equation, accounting for the
possibility of no termination produces age coefcients that are
larger than when such possibility is not accounted for (see column
5 in Tables 3 and 4). As explained above, not accounting for such a
possibility biases the results toward negative age dependence and
can generate a spurious negative relationship, but not a positive
one. We nd that not accounting for it leads to a sizeable
underestimation of the true effect. The quadratic specication is
signicant only in the closure equation. Accounting for the
possibility of no terminations in the quadratic specication in
this equation produces changes in the coefcients such that the
minimum of the hazard function is found at the seventh year rather
than at the eighteenth, as was the case when no such possibility
was accounted for. The cubic specication never produces any
signicant coefcients for the cubic term.
6. Discussion
Our results suggest a complex pattern for the effect of age upon
termination of joint ventures. During the rst years the likelihood
of closure decreases, but after a number of years (not many,
according to our estimates) the likelihood of exit starts to increase.

Acquisitions by the domestic partners the other mode by which

foreign equity can leave a country remains at roughly the same
level over time, which means that overall, foreign owned rms are
less likely to remain active in the host country as time goes by. This
pattern is at odds with the suggestion of Zaheer and Mosakowski
(1997). They present evidence suggesting that the chances of exit
by foreign rms decreases over time and the difference relative to
the corresponding pattern of domestic rms attenuates over time.
According to their explanation, foreign rms would be more likely
to exit during the rst years of operation, due to a liability of
foreignness, but as experience brings knowledge about local
conditions, this initial disadvantage vanishes and exit becomes as
likely as that for a comparable domestic rm. Our results are also at
odds with the suggestion that age should improve the chances of
joint venture survival, as a consequence of age enabling the
development of trust among partners (Inkpen & Currall, 2004). On
the contrary, our results for international joint ventures are more
consistent with a view of footloose multinationals (Gorg & Strobl,
2003). Foreign rms may enter a country to exploit an opportunity
that is limited in timeor will stay in the country as long as an
alternative does not emerge that is more interesting. Gorg and
Strobl (2003) have found that foreign rms are more likely to exit
the country than comparable domestically owned rms. The same
ndings are reported by Mata and Freitas (2012), who further
indicate that foreign rms become more and more likely to exit
over time. Our observation that foreign rms exit the country with
probabilities that increase over time is closer to the view of the
footloose multinationals than to that of a diminishing liability of
foreignness.
On the contrary, foreign partners will fully acquire the joint
ventures with increasing probability over time. This observed
pattern for foreign acquisition ts well the view of joint ventures
as options to expand in those cases where the foreign owner
would wish to do so (Kogut, 1991). However, Buckley and Casson
(1998) argued persuasively that the domestic partner may also be
a ready buyer in those cases in which the foreign rm decides to
divest. The option value of joint ventures would reside in the
possibility of acquiring information about market prospects for
some time before making the subsequent decision on whether to
acquire or divest (Kumar, 2005). As formulated by Buckley and
Casson (1998), the theory of joint ventures as bilateral options
does not necessarily assign asymmetrical positions to partners in
the process of joint venture termination. Our evidence, however,
suggests an important asymmetry between the acquisitions and
divestments by foreign partners, which involves a rather more
passive role for domestic than for foreign partners. Foreign
partners will never consider acquiring some joint ventures. In
those cases in which they do, however, they do so with increasing
probabilities over time. Still, the probability of fully acquiring the
joint venture is low as compared to the probability of dissolving
it. The two combined ndings indicate that foreign partners will
exit the country with an increasing probability over time. On the
contrary, domestic partners will not become more active in
seeking to do it (or in successfully doing it) over time. Taken
together with the observation that multinationals may be
footloose, it is tempting to speculate that domestic partners
take full control of the joint ventures when, and if, foreign
partners are no longer interested in taking part in them. These
different patterns for the evolution of the probability of joint
ventures being terminated through the acquisition by one or the
other partner is consistent with the idea that the assets of the
domestic partner are easier to learn than the assets of the foreign
partner (Inkpen & Beamish, 1997). Foreign partners may learn
from domestic ones what they need for operating a whollyowned business, but there is no evidence that domestic partners
easily do the same.

J. Mata, P. Portugal / International Business Review 24 (2015) 677689

7. Conclusion
This paper reports the results of a detailed investigation into the
patterns of joint venture termination, distinguishing three modes
of terminating a joint venture: closure, acquisition by the foreign
partner, and acquisition by the domestic partner.
First, we were able to uncover the role played by ownership
structure at very different moments upon the termination of joint
ventures: previous ownership of the business unit before the joint
venture is created, equity arrangements at the time the joint
venture is formed, and subsequent changes in such arrangements
all affect the likelihood of the joint venture becoming fully owned
by domestic or foreign partners. Specically, we nd that joint
ventures that were fully owned by foreigners before the formation
of the joint venture are signicantly more likely to become wholly
owned by foreigners again. The more equity one of the partners
owns when the joint venture is created, the more likely it is that
this partner will become the sole owner of the venture. Finally, any
changes in the equity split that occur after the formation of the
joint venture at the time of the formation tip the acquisition in
favor of the partner that increased its equity share. Which partner
acquires sole ownership of the joint venture is also found to be
related to the previous foreign presence in the industry. Foreign
presence favors acquisitions by foreign partners and decreases the
odds of acquisition by the domestic partner. In contrast,
termination by closure is not related to ownership structure.
Instead, it is associated with factors that have been identied as
determinants of rm exits in general, such as size, intangible
assets, and existence previous to the formation of the joint venture.
Second, we nd that there is a non-negligible proportion of joint
ventures that never terminate in any of the three modes
considered. Not controlling for this effect would lead us to
conclude for a much greater negative effect of age upon
termination than is warranted when such an effect is accounted
for. Furthermore, the temporal patterns of exit are complex and
differ depending on the termination mode. The odds of a joint
venture being acquired by a domestic partner are fairly constant
over time. The chances of a joint venture being shut down increase,
at least after a period which is not too long. In contrast, we nd that
the odds of termination by foreign acquisition increase over time.
Our results for the effect of age upon acquisition by foreign
partners are consistent with a view of joint ventures as options to
expand (Kogut, 1991) and that partners use the rst years of the
joint venture to learn about its prospects. When joint ventures
terminate by acquisition by one of the partners, the evidence
indicates that this is much more likely to occur via the acquisition
by foreign partners than by domestic ones. The absolute number of
acquisitions by the foreign partner is much smaller than that of
acquisitions by domestic partners. Even so, the pattern of such
acquisitions over time is much different, which suggests that
domestic and foreign partners are not symmetrical.
Our results have implications for both policy makers and
practitioners. Some countries especially developing countries
require or provide incentives for the formation of joint ventures in
the hope that local partners learn from their foreign counterparts
(see the discussion in Desai, Foley, & Hines 2004 and also UNIDO,
2006). To the extent that this learning would be effective and the
foreign partner would become expendable, we should observe an
increasing probability of joint ventures being converted to wholly
owned domestic rms over time. We do not observe such a
tendency in our data. This result is especially notable as it is
obtained from a sample in which the local partners are from a
developed country, which may have a greater capacity to learn
from foreign partners than that existing in managerial teams in
developing countries. Our results also speak to practitioners who
may be confronted with problems in their joint ventures. If

687

confronted with problems, partners may be tempted to change the

equity split as part of the corrective actions. Practitioners should be
aware that this typically does not translate into greater stability of
the joint ventureand be prepared to deal with an increased
likelihood of termination.
No study is without limitations and ours is no exception. First,
we do not claim to cover all possible explanations for the
termination of joint ventures. In particular, as mentioned above,
we do not know the identity of the partners and, as such, we are not
able to control for the many facets of distance, such as cultural,
linguistic, geographic or socio-economic, that have been suggested
in the literature as relevant for the operations of joint ventures
(Barkema & Vermeulen, 1997; Makino & Beamish, 1998; Zhang &
Li, 2001; Glaister, Husan, & Buckley 2003; Dhanaraj & Beamish,
2004; Anh et al., 2006; Meschi & Riccio, 2008; Yamin & Golesorkhi,
2010), and these may also be an important part of the story. In
addition, our study is performed with data from a single country.
While we do not have any reason to expect that the specicities of
Portugal may greatly inuence the results, replication of the
analysis for other host countries would, of course, be welcome.
Acknowledgments
We are grateful to the audiences in several seminars and
conferences and three referees for comments and to Lucena Vieira
for research assistance. Support from Nova Forum and Fundacao
para a Ciencia e Tecnologia is also acknowledged

Appendix A. Our empirical model

Consider time to be divided into k intervals t 0 ; t 1 , t 1 ; t 2
t k1 ; 1. We observe joint ventures at discrete points in time
T 2 {1, . . ., k}, where T = t denotes the termination of a joint venture
within the interval t t1 ; t t . The hazard function, which gives the
probability of terminating the joint venture during interval t, given
that it was still active at the beginning of this interval, is given by

ht P T tjT  t ;

t 1; 2; . . . ; k  1

and the survivor function, which gives the probability of staying

active up until time T is dened as
St PT  t

t
Y
1  h j:
j1

To incorporate the effect of explanatory variables upon survival,

we apply the same approach as in the conventional proportional
hazards model (Cox, 1972), and dene
0

Stjxi S0 t expxi b
where S(t|xi) is the probability that the individual joint venture i
with covariates xi (which measure those of its characteristics that
are relevant to survival), will remain active up to time t, and S0(t)
denotes the baseline survivor function (that is, when the covariates
equal zero). Given the relationship between the hazard and the
survivor functions above, one can write

1  htjxi 1  h0 t

expx0i b

which leads to the clog-log hazard function

htjxi 1  1  h0 t 

expx0i b

688

J. Mata, P. Portugal / International Business Review 24 (2015) 677689

To incorporate the possibility of defective risks, that is, the

possibility that some units may survive forever, we redene the
survival function, which represents the proportion of joint
1  p pSt,
ventures that did not terminate until t as St
where p is the proportion of joint ventures that face a risk of
dissolution, that is, which are indeed susceptible to the risk of
failure. The survival probability is, therefore, given by the
proportion of long-term survivors, (1  p), which do not exit into
a given destination with probability 1, plus the proportion of
susceptible rms, p, multiplied by their corresponding probability of remaining a joint venture until t, S(t).
Models of this type have been used in a broad number of
applications of duration models. Generalization to multiple
independent risks is straightforward, the maximization of the
likelihood function producing estimates for one additional
unknown parameter p for each mode of termination. In order to
guarantee that each p lies between zero and one, the logit
reparameterization for p = exp(m)/(1 + exp(m)) was employed.
This has no other consequence in terms of nding evidence of
long-term survivors, since it does not preclude p from being as
close to one (or zero) as needed.

References
Addison, J., & Portugal, P. (2003). Unemployment duration: Competing and defective
risks. The Journal of Human Resources, 38(1), 156191.
Allison, P. (1984). Event history analysis. California: Sage Publications Beverly Hills.
Anh, P. T. T., Baughn, C. C., Minh Hang, N. T., & Neupert, K. E. (2006). Knowledge
acquisition from foreign parents in international joint ventures: An empirical
study in vietnam. International Business Review, 15(5), 463487.
Anscombe, F. (1961). Estimating a mixed exponential response law. Journal of the
American Statistical Association, 56(295), 493502.
Audretsch, D., & Mahmood, T. (1994). The rate of hazard confronting new rms and
plants in U.S. manufacturing. Review of Industrial Organization, 9(1), 4156.
Barkema, H. G., & Vermeulen, F. (1997). What differences in the cultural backgrounds of
partners are detrimental for international joint ventures? Journal of International
Business Studies, 28(4), 845864.
Barney, J. (1991). Firm resources and sustained competitive advantage. Journal of
Management, 17(1), 3456.
Blodgett, L. (1991). Partner contributions as predictors of equity share in international
joint ventures. Journal of International Business Studies, 22(1), 6378.
Blodgett, L. (1992). Factors in the instability of international joint ventures: An event
history analysis. Strategic Management Journal, 13(6), 475481.
Brouthers, K., & Bamossy, Gary J.. (2006). Post-formation processes in eastern and
western European joint ventures. Journal of Management Studies, 43(2), 203229.
Buckley, P., & Casson, M. (1998). Models of the multinational enterprise. Journal of
International Business Studies, 29(1), 2144.
Caves, R. (1996). Multinational enterprise and economic analysis. Cambridge: Cambridge
University Press.
Chang, S., & Singh, H. (1999). The impact of modes of entry and resource t on modes of
exit by multi-business rms. Strategic Management Journal, 20(11), 10191035.
Chen, S. F. S. (2008). The motives for international acquisitions: Capability procurements. Strategic considerations, and the role of ownership structures. Journal of
International Business Studies, 39(3), 454471.
Chen, S. F. S., & Hennart, J. F. (2004). A hostage theory of joint ventures: Why do
Japanese investors choose partial over full acquisitions to enter the United States?
Journal of Business Research, 57(10), 11261134.
Chi, T., & Seth, A. (2009). A dynamic model of the Choice of mode for exploiting
complementary capabilities. Journal of International Business Studies, 40(3), 365387.
Chung, C. C., & Beamish, P. W. (2010). The trap of continual Change in international
equity joint ventures. Organization Science, 21(5), 9951015.
Chung, C. C., & Beamish, P. W. (2012). Multi-party international joint ventures: Multiple
post-formation Change processes. Journal of World Business, 47(4), 648663.
Conway, P., Janod, V., & Nicoletti, G. (2005). Product market regulation in OECD
countries: 1998 to 2003. In OECD economics department working papers, no. 419.
OECD Publishing, http://dx.doi.org/10.1787/783417550348
Cox, D. (1972). Regression models and life-tables. Journal of the Royal Statistical Society.
Series B (Methodological), 34(2), 187220.
Delios, A., & Beamish, P. (2001). Survival and protability: The roles of experience and
intangible assets in foreign subsidiary performance. Academy of Management
Journal, 44(5), 10281038.
Desai, M. A., Foley, C. F., & Hines, J. R., Jr. (2004). The costs of shared ownership:
Evidence from international joint ventures. Journal of Financial Economics, 73(2),
323374.
Dhanaraj, C., & Beamish, P. (2004). Effect of equity ownership on the survival of
international joint ventures. Strategic Management Journal, 25(3), 295305.
Dunne, T., Roberts, M., & Samuelson, L. (1989). The growth and failure of U.S.
manufacturing plants. Quarterly Journal of Economics, 104(4), 678691.

Dunning, J. (1993). Multinational enterprises and the global economy. Massachusetts:

Addison-Wesley Reading.
Fazzari, S., Hubbard, R. G., & Petersen, B. C. (1988). Financing constraints and corporate
investment. Brookings Papers on Economic Activity, 1988(1), 141195.
Folta, T. B., & Miller, K. D. (2002). Real options in equity partnerships. Strategic
Management Journal, 23(1), 7788.
Franko, L. (1989). Use of minority and 5050 joint ventures by United States multinationals during the 1970s: The interaction of host country policies and corporate
strategies. Journal of International Business Studies, 20(1), 1940.
Garca-Manjon, J. V., & Romero-Merino, M. E. (2012). Research, development, and rm
growth. Empirical evidence from European top R&D spending rms. Research
Policy, 41(6), 10841092.
Geringer, J. M., & Killing, J. P. (1988). Joint venture partner selection: Strategies for
developed countries. New York, NY: Quorum Books.
Glaister, K. W., Husan, R., & Buckley, P. J. (2003). Learning to manage international joint
ventures. International Business Review, 12(1), 83108.
Godin, B. (2004). The obsession for competitiveness and its impact on statistics: The
construction of high-technology indicators. Research Policy, 33(8), 12171229.
Gorg, H., & Strobl, E. (2003). Footloose multinationals. The Manchester School, 71(1), 119.
Gulati, R. (1998). Alliances and networks. Strategic Management Journal, 19(4), 293
317.
Ham, J. C., & Rea, S. (1987). Unemployment insurance and male unemployment
duration in Canada. Journal of Labor Economics, 5(3), 325353.
Haveman, H. A. (1995). The demographic metabolism of organizations: Industry
dynamics, turnover and tenure distributions. Administrative Science Quarterly,
40(4), 586618.
Heckman, J., & Walker, J. R. (1990). Estimating fecundability from data on waiting times
to rst conception. Journal of the American Statistical Association, 85(410), 283294.
Hennart, J., Kim, D., & Zeng, M. (1998). The impact of joint venture status on the
longevity of Japanese stakes in US manufacturing afliates. Organization Science,
9(3), 382395.
Hennart, J., Roehl, T., & Zietlow, D. (1999). Trojan horse or Workhorse: The evolution
of US Japanese joint ventures in the United States. Strategic Management Journal,
20(1), 1529.
Inkpen, A., & Beamish, P. (1997). Knowledge, bargaining power, and the instability of
international joint ventures. Academy of Management Review, 22(1), 177202.
Inkpen, A., & Currall, S. (2004). The co-evolution of trust. Control, and learning in joint
ventures. Organization Science, 15(5), 586599.
Iriyama, A., Shi, W. S., & Prescott, J. E. (2014). Frequency and directional reversal of
equity ownership change in international joint ventures. Asia Pacic Journal of
Management, 31(1), 215243.
Jenkins, S. P. (1995). Easy estimation methods for discrete-time duration models.
Oxford Bulletin of Economics and Statistics, 57(1), 129136.
Kennan, J. (1985). The duration of contract strikes in US manufacturing. Journal of
Econometrics, 28(1), 528.
Kogut, B. (1988). Joint ventures: Theoretical and empirical perspectives. Strategic
Management Journal, 9(4), 319332.
Kogut, B. (1991). Joint ventures and the option to expand and acquire. Management
Science, 37(1), 1933.
Kumar, M. (2005). The value from acquiring and divesting a joint venture: A real
options approach. Strategic Management Journal, 26(4), 321331.
Lancaster, T. (1990). The econometric analysis of transition data. Cambridge: Cambridge
University Press.
Lucas, R. E. (1978). On the size distribution of business rms. Bell Journal of Economics,
9(2), 508523.
Luo, Y. (2002). Contract, cooperation, and performance in international joint ventures.
Strategic Management Journal, 21(10), 903919.
Makino, S., Chan, C. M., Isobe, T., & Beamish, P. W. (2007). Intended and unintended
termination of international joint ventures. Strategic Management Journal, 28(11),
11131132.
Makino, S., & Beamish, P. W. (1998). Performance and survival of joint ventures with
non-conventional ownership structures. Journal of International Business Studies,
29(4), 797818.
Mascarenhas, B. (1992). Order of entry and performance in international markets.
Strategic Management Journal, 13(7), 499510.
Mata, J., & Freitas, E. (2012). Foreignness and exit over the life cycle of rms. Journal of
International Business Studies, 43(7), 116.
Mata, J., & Portugal, P. (1994). Life duration of new rms. Journal of Industrial Economics,
42(3), 227246.
Mata, J., & Portugal, P. (2000). Closure and divestiture by foreign entrants: The impact
of entry and post-entry strategies. Strategic Management Journal, 21(5), 549562.
Meschi, P. X., & Riccio, E. L. (2008). Country risk, national cultural differences between
partners and survival of international joint ventures in Brazil. International Business
Review, 17(3), 250266.
Mitchell, W. (1994). The dynamics of evolving markets: The effects of business sales
and age on dissolutions and divestitures. Administrative Science Quarterly, 39(4),
575602.
Mitchell, W., Shaver, J., & Yeung, B. (1994). Foreign entrant survival and foreign market
share: Canadian companies experience in United States medical sector markets.
Strategic Management Journal, 15(7), 555567.
Moran, T. (2009). Foreign acquisitions and national security: What are genuine threats?
What are implausible worries? Paris: OECD.
Mjoen, H., & Tallman, S. (1997). Control and performance in international joint
ventures. Organization Science, 8(3), 257274.
Nemeth, A., & Nippa, M. (2013). Rigor and relevance of IJV exit research. Management
International Review, 53(3), 449475.

J. Mata, P. Portugal / International Business Review 24 (2015) 677689

Park, S. H., & Russo, M. V. (1996). When cooperation eclipses cooperation: An event
history analysis of joint venture failure. Management Science, 42(6), 875890.
Park, S. H., & Ungston, G. R. (1997). The effects of national culture. Organizational
complementarity, and economics motivations on joint venture dissolution. Academy of Management Journal, 40(2), 297307.
Prentice, R. L., & Gloeckler, L. A. (1978). Regression analysis of grouped survival data
with application to breast cancer data. Biometrics, 34(1), 5767.
Puck, J. F., Holtbrugge, D., & Mohr, A. T. (2009). Beyond entry mode choice: Explaining
the conversion of joint ventures into wholly owned subsidiaries in the Peoples
Republic of China. Journal of International Business Studies, 40(3), 388404.
Reuer, J. J., & Shen, J. C. (2004). Sequential divestiture through initial public offerings.
Journal of Economic Behavior & Organization, 54(2), 249266.
Reuer, J. J., & Ragozzino, R. (2008). Adverse selection and M&A design: The roles of
alliances and IPOs. Journal of Economic Behavior & Organization, 66(2), 195212.
Reuer, J. J., Zollo, M., & Singh, H. (2002). Post-formation dynamics in strategic alliances.
Strategic Management Journal, 23(2), 135151.
Schary, M. A. (1991). The probability of exit. The RAND Journal of Economics, 22(3), 339
353.
Schmidt, P., & Witte, A. (1989). Predicting criminal recidivism using Split population
survival time models. Journal of Econometrics, 40(1), 141159.
Sharma, A., & Kesner, I. F. (1996). Diversifying entry: Some ex ante explanations for
postentry survival and growth. Academy of Management Journal, 39(3), 635677.
Shaver, J., Mitchell, W., & Yeung, B. (1997). The effect of own-rm and other-rm
experience on foreign direct investment survival in the United States, 198792.
Strategic Management Journal, 18(10), 811824.
Steensma, H. K., Barden, J. Q., Dhanaraj, C., Lyles, M., & Tihanyi, L. (2007). The evolution
and internalization of international joint ventures in a transitioning economy.
Journal of International Business Studies, 39(3), 491507.

689

Struthers, C., & Farewell, V. (1989). A mixture model for time to AIDS data with left
truncation and an uncertain origin. Biometrika, 76(4), 814.
Teece, D. (1998). Capturing value from knowledge assets: The new economy. Markets
for know-how, and intangible assets. California Management Review, 40(3), 5579.
UNIDO (2006). Alliances and joint ventures: Patterns of internationalization for developing
country enterprise. Vienna: United Nations.
Wernerfelt, B. (1984). A resource-based view of the rm. Strategic Management Journal,
5(2), 171180.
Yamaguchi, K. (1992). Accelerated failure-time regression models with a regression
model of surviving fraction: An application to the analysis of permanent employment in Japan. Journal of the American Statistical Association, 87(418), 284292.
Yamin, M., & Golesorkhi, S. (2010). Cultural distance and the pattern of equity
ownership structure in international joint ventures. International Business Review,
(5), 457467.
Yan, A., & Gray, B. (1994). Bargaining power, management control, and performance in
United StatesChina joint ventures: A comparative case study. Academy of Management Journal, 37(6), 14781517.
Zaheer, A., Hernandez, E., & Banerjee, S. (2010). Prior alliances with targets and
acquisition performance in knowledge-intensive industries. Organization Science,
(5), 10721091.
Zaheer, S., & Mosakowski, E. (1997). The dynamics of the liability of foreignness: A
global study of survival in nancial services. Strategic Management Journal, 18(6),
439463.
Zhang, Y., & Li, H. (2001). The control design and performance in international joint
ventures: A dynamic evolution perspective. International Business Review, 10(3),
341362.
Zingales, L. (1998). Survival of the ttest or the fattest?. Exit and nancing in the
trucking industry. Journal of Finance, 53(3), 905938.