WHEN DO WHOLLY OWNED SUBSIDIARIES PERFORM BETTER THAN JOINT VENTURES? · subsidiaries that continued as joint ventures, we examine whether their performance diverges from the moment

Strategic Management JournalStrat. Mgmt. J., 34: 317–337 (2013)

Published online EarlyView in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/smj.2016

Received 11 November 2010; Final revision received 21 September 2011

WHEN DO WHOLLY OWNED SUBSIDIARIESPERFORM BETTER THAN JOINT VENTURES?

SEA-JIN CHANG,1* JAIHO CHUNG,2 and JON JUNGBIEN MOON2

1 NUS Business School, National University of Singapore, Singapore2 Korea University Business School, Korea University, Seoul, Korea

This study explores when wholly owned subsidiaries outperform joint ventures with localpartners. In order to avoid the endogeneity problem inherent in foreign subsidiaries’ operatingmode decisions that might confound performance measurement, we employ the propensityscore matching method, along with the difference-in-differences approach, and compare theperformances of joint ventures turned wholly owned subsidiaries vis-a-vis continuing jointventures. Based on foreign subsidiaries’ financial data in China for 1998–2006, we find strongevidence that converted wholly owned subsidiaries outperform continuing joint ventures inindustries characterized by high levels of intangible assets such as technology or brand, aftercontrolling for factors that may affect the conversion decision. This finding is consistent with theprediction of transaction cost theory. Copyright 2012 John Wiley & Sons, Ltd.

INTRODUCTION

Conventional wisdom in strategy research suggeststhat when multinational firms possess intangibleresources such as technology or brand, they shouldenter a foreign country via wholly owned sub-sidiaries rather than joint ventures. According totransaction cost theory, the use of a partially ownedventure or a contractual mode to transfer tacitor proprietary intangible assets is subject to theconsiderable risks of free riding and other oppor-tunistic behavior (Hennart, 1982). Despite abun-dant research on ex ante mode choice at the timeof entry, little is known about the ex post rela-tive performance of operating mode. That is, dowholly owned subsidiaries indeed perform better

Keywords: joint venture termination; entry mode choice;wholly owned subsidiaries; subsidiary performance; trans-action cost theory∗ Correspondence to: Sea-Jin Chang, NUS Business School,National University of Singapore Singapore 117592.E-mail: [email protected]

than joint ventures after entry, as their ex antemode choice would predict (Brouthers, Brouthers,and Werner, 2003; Shaver, 1998), or vice versa?The absence of work in this area takes root in bothendogeneity issues inherent to examining post-entry performance and the lack of financial data forforeign subsidiaries. We overcome both limitationsby employing two recently developed empiricaltechniques and using financial data for foreign sub-sidiaries in China.

The choice of operating mode is central to afirm’s foreign investment strategy.1 Depending onthe need to maintain control, firms can use con-tracts (e.g., licensing), joint ventures with localpartners, or wholly owned subsidiaries. Transac-tion cost theory provides several reasons whywholly owned subsidiaries might perform betterthan joint ventures under certain circumstances.First, foreign parents may transfer more intangible

1 Brouthers and Hennart (2007) review research on internationalentry modes.

Copyright 2012 John Wiley & Sons, Ltd.

318 S.-J. Chang, J. Chung, and J. J. Moon

resources, such as more sophisticated technol-ogy or brands, to wholly owned subsidiaries thanthey do to joint ventures because they are moreconcerned about local joint venture partners appro-priating their intangible resources. Second, the per-formance of a joint venture depends not only onthe congruence of the partners’ goals but also onwhether opportunism by partners can be contained.When a firm uses a wholly owned subsidiary, itdoes not have such concerns. In contrast, transac-tion cost theory suggests that a joint venture maybe preferable when a local partner offers comple-mentary knowledge, such as a deep understandingof local markets or access to distribution chan-nels and natural resources, which cannot be easilypurchased on the market (Hennart, 2009). Thus,transaction cost theory indicates that the optimalchoice of operating mode is contingent upon vari-ous factors. Our study focuses on one of those con-tingencies by examining whether wholly ownedsubsidiaries perform better than joint ventures inindustries characterized by high levels of intangi-ble and tacit resources.

Previous attempts to evaluate post-entry sub-sidiary performance of different mode choiceshave been limited in two important ways. First,because a firm’s choice about whether to use ajoint venture or a wholly owned subsidiary is, inpart, a function of the firm’s own characteristics,the endogeneity of this choice must be consid-ered when the subsidiary’s post-entry performanceis evaluated. For instance, Shaver (1998) demon-strates that the apparently higher survival rate ofgreenfield investment vis-a-vis acquisition disap-pears when the self-selection aspect of entry modeis incorporated into the model. Second, financialdata for subsidiary-level performances are seldomavailable, as most foreign subsidiaries are privatelyheld, and thus not required to disclose performancedata. As a consequence, previous studies have beenlimited to the use of subsidiary exit, sales growth,or subjective evaluation as a proxy for subsidiaryperformance (Delios and Beamish, 2001; Shaver,1998).

This study overcomes these limitations. First,we employ two recently developed techniques:propensity score matching and the difference-in-differences approach. It would be ideal to comparethe performance of a joint venture and that ofa wholly owned subsidiary from the same firm,but it is impossible to observe performance for anoperating mode that the subsidiary did not choose

(i.e., counterfactual performance). The propensityscore matching technique provides a way to cre-ate counterfactual performances for the sake ofcomparison. Second, we use foreign subsidiaries’financial data in China from 1998 to 2006. TheAnnual Industrial Survey Database of the ChineseNational Bureau of Statistics provides financialinformation for all industrial firms, including for-eign subsidiaries above a certain size. These dataallow us to track financial performance.

In order to choose an empirical setting thatcan handle the endogeneity issue, we focus onsubsidiaries that converted from joint ventures towholly owned subsidiaries. This focus allows us tomatch a joint venture that did not change its oper-ating mode even though its ex ante likelihood ofdoing so (i.e., its propensity score) is roughly equalto that of the converted wholly owned subsidiary.In order to do so, we control for factors that mayaffect the conversion decision based on the jointventure termination/instability literature, includingownership structure (Blodgett, 1992; Dhanaraj andBeamish, 2004), changes in environmental andfirm-specific conditions (Kogut, 1991), and thenontransferable complementary resources of localpartners (Hennart, 2009). We are then equippedto harness the performance of a matched jointventure as an appropriate counterpart, even if wecannot observe the counterfactual performance hadthe converted subsidiary remained a joint venture.Once we match all the converted subsidiaries withsubsidiaries that continued as joint ventures, weexamine whether their performance diverges fromthe moment of conversion by using the difference-in-differences approach.

Our empirical context of joint ventures in Chinafrom 1998 to 2006 provides an ideal setting forour research questions. When China adopted a‘reform and open-door policy’ in 1978, it requiredforeign multinationals to take interests in jointventures in order to enter the market. The jointventure requirement was relaxed in most indus-tries, excluding strategic industries like automo-tive and steel, as the country prepared itself tojoin the World Trade Organization (WTO). Asa consequence, many joint ventures were con-verted to wholly owned subsidiaries. Yet a sub-stantial number of joint ventures continued evenafter this requirement was lifted, thus providing anopportunity to apply the propensity score matchingand the difference-in-differences methods, whichrequire both converted wholly owned subsidiaries

Copyright 2012 John Wiley & Sons, Ltd. Strat. Mgmt. J., 34: 317–337 (2013)DOI: 10.1002/smj

Wholly Owned Subsidiaries vs. Joint Ventures 319

(the treatment group) and continuing joint ventures(the control group).

Our findings confirm transaction cost theory’skey prediction for post-entry performance: con-verted wholly owned subsidiaries demonstratesuperior financial performance in industries char-acterized by high levels of intangible assets liketechnology or brand. We also explore which mech-anisms may enhance performance when ownershipchanges. To do so, we examine temporal changesof key strategic variables including sales, intangi-ble assets, and fixed assets. We find that convertedwholly owned subsidiaries invest more in intangi-ble assets and increase sales revenue faster thancontinuing joint ventures.

THEORY AND HYPOTHESES

Literature review

Internalization theory, an early attempt to explainforeign operation mode choice, holds that a parentfirm’s need to internalize intangible resources dic-tates the choice between exporting, licensing, andforeign direct investment (Dunning, 1988). Sev-eral other researchers subsequently proposed thatas firms gain experience in foreign operations, theyperceive less uncertainty and therefore tend to usewholly owned subsidiaries rather than contractualor partial ownership modes (Johanson and Vahlne,1977).

Transaction cost theory pushes the internaliza-tion logic further by specifying the conditionsfor market failures, including asset specificity andinformation asymmetry, coupled with uncertaintyand opportunism (Hennart, 1982). According tothis theory, multinational firms can efficientlytransfer resources by using wholly owned sub-sidiaries rather than joint ventures when poten-tial partners could free ride off their proprietaryintangible assets and act in other opportunis-tic ways. Furthermore, because proprietary assetsare typically tacit, the information asymmetrybetween transacting parties would make multina-tional firms more likely to exploit these assetsvia wholly owned subsidiaries. Several empiri-cal investigations demonstrate that entry via out-right ownership is preferred by multinational firmsthat possess higher amounts of intangible assets,which are often operationalized by research and

development (R&D) and advertising intensities(e.g., Gatignon and Anderson, 1988).2

Considering its strategic importance, it is sur-prising that few studies evaluate the relationshipbetween operating mode choice and subsidiary per-formance. Nitsch, Beamish, and Makino (1996)identify two reasons for the scant attention paidto this issue. First, it is very difficult to obtainsubsidiary performance data because of differentnational financial reporting conventions, the reluc-tance of parent firms to divulge nonconsolidateddata, and the problems of reconciling internal datafrom different firms, even when such data areavailable. Firms’ use of internal transfer pricingand tax havens further complicates evaluationsof subsidiary-level financial performance. Second,performance likely depends on mode choice; for-eign parents evaluate the available mode choiceson a risk-return basis and select the one with thehighest expected performance. If this choice isoptimal for a given firm ex ante, an ex post per-formance comparison among firms suffers fromendogeneity issues unless we explicitly considerthe choice of operating mode.

Despite such difficulties, several researchershave attempted to evaluate post-entry performancewith exit rates, sales growth, or subjective man-agerial assessments as proxies for subsidiary per-formance. One such attempt is developed intothe joint venture termination literature. Severalresearchers consider the termination (or morebroadly instability) of joint ventures as an indi-cation of failure. According to this view, jointventures fail because multinational parents mayreorient their organizational structure towardgreater centralization (Franko, 1971), have diffi-culty in coordinating parent-joint venture productportfolio due to shared control (Killing, 1983),and be unable to manage cultural differences(Barkema, Bell, and Pennings, 1996). Similarly,several researchers argue that ‘initial conditions,’including task definitions, partners’ routines, andexpectations, may create divergent learning and

2 In contrast to our focus on the choice of ownership andcontrol, others have studied how joint ventures and whollyowned subsidiaries are set up (i.e., via greenfield investmentor acquisition; see Hennart and Park, 1993). When firms seek toexploit their superior organizational and technical expertise, theyfrequently prefer greenfield entry, which is the most efficientway to transfer these advantages to foreign countries. By way ofcontrast, a multinational firm will choose to acquire a firm if itis going to acquire complementary inputs that can be purchasedmore inexpensively as a bundled form in a going concern.



frustrated expectations (Doz, 1996; Reuer, Zollo,and Singh, 2002). Blodgett (1992) further exam-ines whether uneven ownership structure promptsjoint venture instability. Finally, challenging thenotion that joint ventures are inherently fragile,Hennart, Kim, and Zeng (1998) find that interna-tional joint venture instability might be comparableto that of wholly owned subsidiaries after control-ling for age and size.

More recent studies, however, emphasize thatjoint venture termination should not be inter-preted as a failure but as an optimal adjustmentin response to changing environmental or firm-specific conditions. According to this view, for-eign firms are likely to enter new markets viajoint ventures because they confer an option toexpand/divest under conditions of uncertainty. Asuncertainty resolves, foreign firms can either divestjoint ventures by exercising a put option or acquirethem by exercising a call option (Kogut, 1991).Reuer (2001) further supports this option view,finding that joint venture buyouts positively impactmultinational parents’ abnormal stock returns,especially for firms with high R&D intensity.

Another line of work employs simple compar-ative analyses across different modes of foreignoperation. As these works do not consider endo-geneity, however, they tend to generate inconsis-tent results. Mixing the ownership decision, thatis, joint ventures vs. wholly owned subsidiaries,with the mode of setting up their subsidiaries, thatis, greenfield investment vs. acquisition, furthercomplicates analysis. For instance, using a sam-ple of Japanese firms entering the North Americanmarket, Woodcock, Beamish, and Makino (1994)discover that greenfield ventures offer the best per-formance, followed by joint ventures, and thenacquisitions. In a similar empirical study based ondata of Japanese foreign investments in the whole-sale and retail sector, Anand and Delios (1997)suggest a different order: joint ventures, green-field ventures, and acquisitions. Also using data onJapanese subsidiaries, Delios and Beamish (2001)find that the host country’s experience directlyand positively influences a subsidiary’s survival,regardless of chosen entry mode. Yet the authorsalso suggest that host country experience influ-ences subsidiary profitability for joint venturesless than it does for wholly owned subsidiaries,where profitability was subjectively assessed bymanagers.

Taking on this line of research, Shaver (1998)argues that studies examining the performanceimplications of mode choice fail to account forendogeneity. To demonstrate the importance ofendogeneity, he uses a sample of foreignsubsidiaries operating in the United States to deter-mine whether entry mode (acquisition vs. green-field investment) affects subsidiary performance,as set by the exit rate. Although he demonstratesthat greenfield entries survive more often thanacquisitions, he also shows that the significanceof this effect disappears when he includes self-selection in his model. Brouthers et al. (2003) like-wise address the endogeneity issue, as they employthe Heckman method. Based on data from largeDutch, German, and British firms entering Centraland Eastern European markets via wholly ownedsubsidiaries or joint ventures, they find that firmsusing mode choices predicted by transaction costtheory perform significantly better than firms thatchoose entry modes inconsistent with transactioncost theory.

Economists, on the other hand, address the endo-geneity issue more carefully. However, they typi-cally do not distinguish joint ventures from whollyowned subsidiaries. Earlier empirical studies, mostof which use cross-sectional and aggregate data,find that foreign subsidiaries are more productiveand pay higher wages than domestic firms (Domsand Jensen, 1998; Ries, Globerman, and Vertinsky,1994). Conyon et al. (2002), based on firm-levelpanel data, find a positive effect of foreign own-ership on performance, after controlling for unob-served heterogeneity. More recent works counterthese findings, arguing that panel regressions fail tocontrol for all of the unobservable firm character-istics that likely influence subsidiary performance.For example, Girma and Gorg (2007) employthe difference-in-differences estimation based onpropensity score matching to identify substan-tial heterogeneity in post-acquisition wages. Simi-larly, employing the firm-level data from Indonesiaand the difference-in-differences estimation withpropensity score matching, Arnold and Javorcik(2009) find that foreign-owned plants demonstratehigher total factor productivity and labor produc-tivity than their local counterparts.

To summarize, the management literature onsubsidiary-level performance suffers from both alack of data and measurement problems, whileeconomics research on this topic has not paiddue attention to the question of which foreign



operations mode (e.g., joint ventures vs. whollyowned subsidiaries) generate superior perfor-mance. Building on the strengths of the manage-ment and economic literatures, this paper attemptsto enhance our understanding of the subsidiary-level performance consequences of different modesby employing propensity score matching and thedifference-in-differences approach.

Hypotheses

In this study, we examine whether wholly ownedsubsidiaries perform better than joint ventures inindustries characterized by high levels of intan-gible assets such as technology and brand, asspecified in the transaction cost theory. First, asintangible assets such as technology or trademarksare inherently subject to market failure (Arrow,1974), multinational firms that transfer intangi-ble assets will find it difficult to prevent jointventure partners from leaking knowledge or freeriding on their reputation, or to price them dueto information asymmetry. Second, joint venturescan also be more costly when there is a highlevel of ‘asset specificity,’ which refers to durableinvestments that cannot be easily redeployed toalternative uses or alternative users without a sacri-fice of productive value (Williamson, 1991). Highasset specificity may result in self-interested small-number bargaining and invite opportunism (Linand Png, 2003). Third, most proprietary knowl-edge is uncodified, making it difficult for foreignparents to transfer it except through close, long-term relationships, which are difficult to set up andmaintain in joint ventures.

Furthermore, wholly owned subsidiaries canmake decisions more quickly, as they do notneed to obtain consent from local partners. Inaddition, wholly owned subsidiaries can increasecontrol by installing their own control mecha-nisms, most notably corporate culture and manage-ment systems. As industries with a high level ofintangible assets frequently face rapid changes intechnology and competition, which require quickdecision making and implementation, the whollyowned option can be attractive. Consistent with thelogic of transaction cost theory, previous studiesfind that wholly owned subsidiaries are preferredwhen parent firms’ R&D and/or advertising inten-sities, which proxy for tacit and poorly protectedintangible assets, are high (Gatignon and Ander-son, 1988). Reuer’s (2001) finding that positive

abnormal stock returns from joint venture acquisi-tions for parent firms with high R&D intensity isalso in line with the transaction cost logic.

Yet the benefits of joint ventures exist when themarket for complementary assets owned by localpartners fail. Joint ventures may generate superiorperformance in conditions where local partners’deep understanding of local markets and access todistribution channels and natural resources cannotbe purchased on the market (Madhok, 1997). Wethus need to control for a local partner’s contribu-tion that can help the performance of a continuingjoint venture be superior to that of a convertedwholly owned subsidiary.

To summarize, wholly owned subsidiaries per-form better than joint ventures due to higher invest-ment in intangible and tangible resources, fastdecision making, and better control. Even whenlocal knowledge or resources from local partnersare critical, our empirical context provides a con-servative test, as we expect wholly owned sub-sidiaries to outperform joint ventures. This leadsus to the following two hypotheses:

Hypothesis 1: Wholly owned subsidiaries per-form better than joint ventures in industries witha higher level of R&D intensity.

Hypothesis 2: Wholly owned subsidiaries per-form better than joint ventures in industries witha higher level of advertising intensity.

METHODS

Data and sample

This study utilizes the Annual Industrial SurveyDatabase (1998–2006) from the Chinese NationalBureau of Statistics (NBS). The NBS collectsfinancial information on industrial firms and pub-lishes aggregate information in the official ChinaStatistics Yearbooks. All firms in China are re-quired to cooperate with the survey and sub-mit the appropriate financial information. Before1998, this database did not include informationon private firms. In 1998, the NBS expanded itsdatabase coverage to all firms with annual salesof at least 5 million RMB (approximately USD626,000 using the average exchange rate in 2006)in the year prior to the survey, including state-owned enterprises (SOEs), all non-SOE firms, and



foreign firms. In China, foreign firms are legallydefined by a variety of relevant laws. Article 4of the Joint Ventures Law states that the foreignpartner must hold at least 25 percent of the regis-tered capital for equity investment. The databaseis confirmed accurate and consistent for statisticalanalysis (Chow, 1993), with several prior studiestaking advantage of its data (e.g., Chang and Xu,2008; Park, Li, and Tse, 2006).

Beginning in 1978, China adopted a ‘reform andopen-door policy’ to move to a market-based econ-omy. With this policy, the Chinese governmentsought to increase the efficiency of local firms byprivatizing and restructuring. It also allowed for-eign firms to enter the market, beginning with theestablishment of four special economic zones in1980. In the early stages of this ‘open-door’ policy,foreign firms were required to form joint ventureswith domestic firms in order to enter China. TheChinese government often designated local part-ners and dictated the exact terms for the initialcontract (Puck, Holtbrugge, and Mohr, 2009). WithChina’s accession to the WTO in 2001, however,this requirement was relaxed, as foreign firms wereallowed to maintain full ownership and acquirelocal firms in most ‘non-strategic industries.’ As aconsequence, foreign direct investment inflow viawholly owned subsidiaries increased dramatically,while many joint ventures converted to whollyowned subsidiaries.

We constrict our attention to joint ventures atthe time they first appear in our dataset. We definea firm as a joint venture when the foreign part-ner holds between 25 percent and 95 percent ofthe venture’s equity. We define a firm in whichthe foreign partner owns more than 95 percentof the venture’s equity as a wholly owned sub-sidiary. Joint ventures can be further subdividedinto majority joint ventures (where the foreignpartner owns more than 50% of the equity) andminority joint ventures (where the foreign partnerowns between 25% and 50% of the equity).3 In oursample period, the number of foreign firms (i.e.,the sum of minority and majority joint venturesand wholly owned subsidiaries) grew from 24,168in 1998 to 55,819 in 2006 (see Figure 1a). Among

3 Gatignon and Anderson (1988) use these criteria to distinguishamong minority and majority joint ventures and wholly ownedsubsidiaries. We include 50 : 50 joint ventures in the category ofminority joint venture.

(a) The number of foreign firms by ownership type

(b) The number of minority joint ventures converted towholly owned subsidiaries by year

Figure 1. Ownership structure of foreign firms in Chinaduring 1998–2006

foreign firms, the fraction of wholly owned sub-sidiaries increased from 40.3 percent in 1998 to64.9 percent in 2006.

We focus on minority joint ventures in order tomaximize the contrast between before and after theconversion from joint ventures to wholly ownedsubsidiaries. We identify 31,435 minority jointventures that appear during our sample period.Of these, 2,991 converted to wholly owned sub-sidiaries and 28,444 remained joint ventures, eitheruntil they disappeared from the database or until2006, the end of our period of analysis. The num-ber of conversions from minority joint ventures towholly owned subsidiaries also grew, increasingfrom 177 cases in 1999 to 649 cases in 2006 (seeFigure 1b).

Because we measure changes in performancefrom year t through year t+3 and because weinclude lagged explanatory variables when we cal-culate the propensity score, we restrict our sampleto firms for which we have at least five consecutiveyears of observations, from year t−1 through yeart+3, where the firm converted to a wholly ownedsubsidiary in year t. Among the 19,557 minorityjoint ventures lasting for at least five consecu-tive years, 840 firms converted to wholly owned



subsidiaries (see the distribution by the year ofconversion in Figure 1b), which is 4.30 percentof all observations in our sample. As we requiredata for up to three years after conversion giventhe parameters outlined above, we consider onlythe cases in which conversion occurred by 2003.We match each of these firms with the most similarminority joint venture that did not convert through-out the sample period within the three-digit Stan-dard Industrial Classification (SIC) code industryand same year.

Variables

We use return on assets (ROA), the most commonmeasure of profitability, as our measure of finan-cial performance. ROA is defined as net incomedivided by total assets. In order to address the con-cern that ROA may be sensitive to financial lever-age or nonoperating income, for example, assetsales and tax payments, we also use operatingreturn on assets (Operating ROA), which is definedas operating income divided by total assets. It isimportant to note that financial performance indi-cators based on subsidiary-level accounting dataare susceptible to potential biases from transferpricing, scope economies, and cross-selling. Asfar as transfer pricing is concerned, our researchdesign poses a conservative test against any down-ward bias, as we expect higher profitability im-provement in joint venture turned wholly ownedsubsidiaries than in continuing joint ventures. Itis usually harder to repatriate profits via trans-fer pricing schemes in joint ventures compared towholly owned subsidiaries, as joint venture part-ners can monitor multinational partners’ transferpricing schemes.4 Because our analysis is limitedto three years after conversion, we assume thatscope economies and cross-selling practices do notchange in this time frame. In fact, there is littlechange in export ratio after the conversion.

We include several firm indicators to predictownership conversion. To control for any size-related factors leading to conversion, we measurefirm size using the logarithm of assets (Hennartet al., 1998). We operationalize firm age as the

4 While it is also possible to have the upward bias in perfor-mances of wholly owned subsidiaries via transfer pricing (i.e.,multinationals use transfer pricing to shift profits from multina-tional headquarter to their Chinese subsidiaries), it is unlikelygiven the high corporate tax rates and regulations on profit repa-triation in China.

calendar year since a firm’s establishment. Weinclude firm age in order to control for the localknowledge or resources gained from a firm’s jointventure experiences, as well as to match convertedwholly owned subsidiaries and continuing jointventures with comparable joint venture experi-ences (Hennart, 1991). Leverage is defined as totaldebt divided by total assets. We measure exportratio as export sales divided by total sales, whichreflects a firm’s export orientation. In calculatingthe propensity score, we include ROA, leverage,and export ratio of a joint venture at one year priorto its conversion to control for any real option val-ues (Kogut, 1991; Kogut and Chang, 1996). Forinstance, a multinational parent will be more likelyto acquire a joint venture with higher profitability,a sound balance sheet, and export platform.

Intangible assets ratio measures the relativeimportance of a firm’s intangible assets, definedas intangible assets divided by total assets. Webelieve that the intangible assets item in a for-eign subsidiary’s balance sheet is a key indicatorof the transfer of intangible assets from multi-national parents to foreign subsidiaries. R&D oradvertising-related activities carried out in indi-vidual foreign subsidiaries would be regarded ascurrent expenditures in subsidiaries’ income state-ments, but technology, brands, and trademarks thatare developed by multinational parents elsewhereand then transferred to their foreign subsidiariesare usually treated as investments in intangibleassets that can be amortized over several years(Kieso, Weygandt, and Warfield, 2010; Chap. 12).Fixed assets ratio, defined as fixed assets dividedby total assets, indicates the firm’s capital inten-sity. We attend to foreign subsidiaries’ fixed assets,as technology transfer often occurs in the form ofmore sophisticated machinery or equipment, whichwill show up as an increase in fixed assets. Theseintangible and fixed assets are important considera-tions for joint venture conversion decisions. Theyalso serve as key strategic indicators to observechanges after the conversion.

The Annual Industrial Survey Database does notcontain any information about individual firms’ultimate parents. In order to capture thecharacteristics of foreign parents and local jointventure partners, we look to the ownership struc-tures of the joint ventures. Building from priorworks that highlight the importance of initial con-ditions underlying the stability of joint ventures(Barkema et al., 1996; Blodgett, 1992; Dhanaraj



and Beamish, 2004; Doz, 1996), we categorizeboth multinational and local parents in our sampleinto two types of firms that likely have differentinitial conditions such as routines, interfaces, andexpectations toward joint ventures. Conventionallocal parent refers to joint ventures in which localparent firms are conventional local firms such asstate-owned enterprises or collectives, as opposedto modernized local firms like private enterprisesor incorporated firms. This variable is coded as 1 ifthe ownership shares held by central or provincialgovernments and collectives are greater than thoseheld by private individuals and incorporated firms,including shareholding firms or limited liabilityfirms, and 0 otherwise. HMT foreign parent cap-tures the differences in two types of foreign par-ents, that is, ethnic Chinese investors from HongKong, Macao, and Taiwan (HMT) and nonethnicChinese foreign multinational parents. This vari-able is coded as 1 if the ownership shares heldby ethnic Chinese investors are greater than thoseheld by nonethnic Chinese foreign investors, and 0otherwise. Finally, we control for the level of for-eign ownership by including foreign share, whichis the percentage share held by foreign parents,and its squared term. We expect that the likeli-hood of conversion to a wholly owned subsidiarywill have an inverted U-shaped relationship withthe level of foreign ownership, as a higher level offoreign ownership would obviate the need for con-version, while a lower level of foreign ownershipwould make conversion more challenging.

We also use the U.S. industry-level averageR&D and advertising intensities for each three-digit SIC industry code (from Compustat datafrom 1998 to 2006) in order to create subsamplesof relatively high/low levels of intangible assetslike technology or brand. Our primary reason forusing U.S. firms’ R&D and advertising intensitiesis to classify industries with exogenous informa-tion unaffected by the sample firms’ strategic R&Dand advertising investment decisions, analogous toRajan and Zingales (1998). We use the medianpoints within our sample, that is, 1.19 percentfor R&D intensity and 2.15 percent for adver-tising intensity. Thus, by design, we create twosubsamples with roughly equal numbers of firmsthat are relatively higher or lower in the techno-logical assets (or marketing know-how) to observedifferential impact on performance associated withconversion.

Finally, we lag all explanatory variables by oneyear in order to help clarify causal relationships. Indealing with outliers, we delete observations withextreme ROA values, namely, those firms with anROA greater than 100 percent or less than −50percent. Table 1 provides the descriptive statisticsand pairwise correlations for the variables used inour analysis.

Propensity score matching method anddifference-in-differences approach

A key challenge in comparing the performance ofjoint ventures and wholly owned subsidiaries isthat the choice of operating mode is endogenous(Shaver, 1998). Simple comparisons of perfor-mance between joint ventures and wholly ownedsubsidiaries are problematic because inherentlybetter firms may have chosen a particular operatingmode, thereby making it difficult to conclude thatthe performance difference is, in fact, caused bythe choice of operating mode. The central ques-tion thus becomes how to construct more reli-able comparison groups between joint ventures andwholly owned subsidiaries. We address this issueby using the propensity score matching method-ology (Arnold and Javorcik, 2009; Imbens andWooldridge, 2009; Rosenbaum and Rubin, 1983).

In propensity score matching, the treatment andcontrol groups are constructed based on a scalar‘similarity’ measure calculated from many differ-ent observable firm characteristics. According toRosenbaum and Rubin (1983: 41), ‘[t]he propen-sity score is the conditional probability of assign-ment to a particular treatment given a vector ofobserved covariates. Both large and small sampletheory show that adjustment for the scalar propen-sity score is sufficient to remove bias due to allobserved covariates.’ The propensity score is cal-culated as the predicted probability of treatmentusing the following probit estimation:

y∗it = Xi,t−1 β + εit , yit = I (y∗

it > 0)

Pr(yit = 1|Xi,t−1 ) = Pr(y∗it > 0|Xi,t−1 )

= Pr(εit > −Xi,t−1 β|Xi,t−1 )

= 1 − �(−Xi,t−1 β) = �(Xi,t−1 β)

where y∗it denotes the unobserved likelihood for

treatment conditional on observable characteris-tics, Xi,t−1 ; yit is a latent dummy variable that



Tabl

e1.

Sum

mar

yst

atis

tics

and

corr

elat

ions

Var

iabl

esM

ean

S.D

.1

23

45

67

89

1011

1213

1C

onve

rsio

n(t

)(%

)4.

3020

.28

1.00

2Fi

rmsi

ze(t−1

)10

.26

1.34

−0.0

4∗1.

003

Firm

age

(t−1

)8.

078.

06−0

.04∗

0.24

∗1.

004

RO

A(t−1

)(%

)5.

3110

.09

−0.0

3∗−0

.03∗

−0.0

4∗1.

005

Lev

erag

e(t−1

)(%

)56

.95

25.1

60.

000.

000.

04∗

−0.3

3∗1.

00

6E

xpor

tra

tio(t−1

)(%

)36

.47

42.3

50.

05∗

−0.2

2∗−0

.04∗

0.00

0.08

∗1.

00

7In

tang

ible

asse

tsra

tio(t−1

)(%

)1.

984.

240.

010.

11∗

−0.0

2∗−0

.06∗

−0.1

0∗−0

.07∗

1.00

8Fi

xed

asse

tsra

tio(t−1

)(%

)32

.15

18.4

80.

000.

17∗

−0.0

2∗−0

.13∗

−0.1

9∗−0

.10∗

0.08

∗1.

00

9C

onve

ntio

nal

loca

lpa

rent

(t−1

)0.

420.

49−0

.05∗

0.08

∗0.

15∗

−0.0

6∗0.

01−0

.07∗

−0.0

3∗0.

04∗

1.00

10H

MT

fore

ign

pare

nt(t−1

)0.

510.

500.

04∗

−0.0

2∗0.

01−0

.07∗

0.06

∗−0

.03∗

−0.0

5∗0.

000.

011.

00

11Fo

reig

nsh

are

(t−1

)(%

)37

.43

15.5

90.

01−0

.02∗

−0.0

5∗0.

00−0

.07∗

0.07

∗−0

.01

0.00

−0.0

9∗−0

.01

1.00

12In

dust

ryR

&D

inte

nsity

(%)

3.58

8.50

−0.0

2∗0.

06∗

0.03

∗0.

06∗

−0.0

5∗−0

.10∗

−0.0

1∗0.

00−0

.01

0.03

∗−0

.01

1.00

13In

dust

ryA

dv.

Inte

nsity

(%)

3.29

6.11

0.00

−0.0

10.

02∗

0.03

∗−0

.04∗

0.00

−0.0

2∗0.

03∗

−0.0

3∗0.

04∗

0.00

0.80

∗1.

00

14O

pera

ting

RO

A(t−1

)(%

)5.

3310

.15

−0.0

1∗−0

.05∗

−0.0

5∗0.

64∗

−0.2

2∗0.

01−0

.05∗

−0.0

8∗−0

.07∗

−0.0

9∗0.

010.

04∗

0.02

∗

[Not

e]N

=19,

557;

∗in

dica

tes

that

the

corr

elat

ion

coef

ficie

ntis

stat

istic

ally

sign

ifica

ntat

the

5%le

vel.



equals 1 if treatment occurs at time t , and 0 other-wise; εit is the error term, which is assumed to benormally distributed; and �(•) is the cumulativenormal distribution function, which ranges from0 to 1 and is symmetric around the midpoint ofthe distribution. In our study, the treatment groupconsists of joint ventures converted into whollyowned subsidiaries and the control group refers tothe continuing joint ventures.

In order to calculate the propensity score, wedraw on the joint venture termination literatureto control for various ‘known’ factors for con-version. First, parent firm type and foreign firmshare reflect the ‘initial conditions’ and culturaldifference that may affect the stability of jointventures (Barkema et al., 1996; Blodgett, 1992;Dhanaraj and Beamish, 2004; Doz, 1996). Sec-ond, we include profitability, as well as export andleverage ratios to explain the real-option relatedmotivations for conversion (Kogut, 1991; Reuerand Tong, 2005), as multinational firms are morelikely to acquire joint ventures with high profitabil-ity, sound balance sheets, and export platforms.Third, we include firm age to control for the num-ber of years partners learn from each other. Fourth,intangible and tangible asset ratios capture assetspecificity that may favor internalization (Hennart,1982). Lastly, we include industry, region, andyear fixed effects to control for their respectiveshocks.

Yet the explanatory variables included in ourprobit model may not capture every factor affect-ing the conversion decision. For instance, prof-itability and leverage and export ratios may notfully indicate the option value of a joint ven-ture. Similarly, firm age may be an imperfectmeasure for the differential learning among part-ners. Some multinational parents may simply favorwholly owned subsidiaries due to their organi-zational change toward more centralization orchanges in their business portfolios (Franko, 1971;Killing, 1983). There are also some unknown ratio-nal reasons. For example, even though multina-tional parents with high R&D/advertising intensityseek to convert joint ventures to wholly ownedsubsidiaries, the negotiation may stalemate if theyhave difficulty finding local managers to overseeoperations or if local partners demand an exorbi-tant price in return for a breakup. Sometimes, theconversion decision can be driven by irrational fac-tors. For instance, some joint ventures, particularlythose in industries with low intensity in R&D and

advertising, may favor conversion out of imitativeor herd behavior, despite no clear benefits (DiMag-gio and Powell, 1983; Yiu and Makino, 2002).Other joint ventures, on the other hand, may notconvert despite being natural cases for conversion,for example, those in industries with high R&Dand advertising. The propensity score matchingmethod assumes that the treatment of ownershipconversion is exogenous given the propensity scorethat is conditional upon observable firm character-istics. In other words, unknown rational or irra-tional factors are commonly held by all firms andare treated as the error term in the probit model.

With the propensity score calculated from theprobit model, we match converted, wholly ownedsubsidiaries with continuing joint ventures withinthe three-digit SIC industry code and within thesame year using the STATA command psmatch2(Leuven and Sianesi, 2003). It is also critical toassess how well the propensity score matchingprocedure created the comparable samples betweenthe treatment and control groups (Dehejia andWahba, 2002; Smith and Todd, 2005). We performseveral widely used ‘balancing tests’ to ensurethat firms in the treatment and control groups arenot statistically different from each other prior tothe treatment (i.e., conversion to wholly ownedsubsidiaries).5

After constructing the treatment and the controlgroups using the propensity score matching tech-nique and confirming that these two groups areindeed comparable using various balancing tests,we employ the difference-in-differences methodto compare the performance of joint venturesturned wholly owned subsidiaries (i.e., the treat-ment group) to the remaining joint ventures (i.e.,the control group). Our adoption of the difference-in-differences approach exploits the panel natureof our dataset and eliminates the effect of unob-servable nonrandom elements of the conversion

5 First, we compare the sample means of all variables includedin the matching procedures by performing individual t-tests.Second, we regress each variable k included in the propensityscore regression on the quartic function of the propensity scoreof firm i in year t , and its interactions with the treatmentvariable. We then use the F-test with the null hypothesis thatthese interaction terms are jointly insignificant, as the treatmentdummy should not provide any further information about theexplained variable if the propensity score satisfies the balancingcondition. Third, we perform a Hotelling T2 test on a joint nullhypothesis that the means of all the variables included in thepropensity score calculation are equal. These balancing testssuggest that the treatment and control groups are comparable oneyear prior to the conversion. Results are available upon request.



decision on performance, which is common to thetreatment group and the control group. In doingso, we compare the differences in within-firm per-formance differences prior to and after treatment(i.e., conversion to a wholly owned subsidiary)between the treatment group and the control group.This design has an advantage over the simplecross-sectional regression approach because theeffects of within-firm unobservable characteristicsare neutralized by our estimations.

Why do we focus on the conversion of joint ven-tures, even though we are fundamentally interestedin understanding the performance implications forjoint ventures vis-a-vis wholly owned subsidiaries?Let us consider the performance function of a sub-sidiary i at time t, yit , which is conditional onmit , the mode choice of subsidiary i at time t

where mit is J (joint venture) or W (wholly ownedsubsidiary), and conditional on Xi,t−1 , a vector ofobservable firm characteristics that also determinesmit . Previous studies using a simple comparativeanalysis without accounting for the endogeneityproblem measured:

� = E[yit |mit(Xi,t−1 ) = W, Xi,t−1 ]

− E[yjt |mjt(Xj,t−1 ) = J, Xj,t−1 ]. (1)

This approach does not account for the fact thatmode choice mit is conditional upon the observedfirm characteristics, Xi,t−1 . This oversight is prob-lematic; even though the model identifies a positivedifference in performance using Equation (1), thisdifference might be due to the fact that more com-petent firms chose to operate as wholly ownedsubsidiaries, not because the wholly owned sub-sidiary mode is the cause of superior performance.Such endogeneity would prevent us from conclud-ing that the wholly owned subsidiary mode is fun-damentally better than the joint venture mode, eventhough we obtained a positive and significant dif-ference in performance, �.

The ideal way to identify comparative perfor-mance would be to measure the performance dif-ference of the same firm with different modechoices, as in Equation (2):

�∗ = E[yit |mit(Xi,t−1 ) = W, Xi,t−1 ]

− E[yit |mit(Xi,t−1 ) = J, Xi,t−1 ]. (2)

Yet we cannot observe the counterfactual perfor-mance for the operating mode that a firm did notchoose.

The propensity score matching technique pro-vides a way to create counterfactual performancefor the purpose of comparison. In order to con-trol for endogeneity, this technique focuses onsubsidiaries that converted from joint ventures towholly owned subsidiaries at time t . We focus onthe converted wholly owned subsidiary and com-pare its performance to that of a remaining jointventure that did not change its operating mode,even though its ex ante likelihood of conversion(i.e., propensity score) is almost identical to thatof the converted subsidiary. Specifically, we canfind a set of counterexample firms j such that|Pc(Xi,t−1 ) − Pc(Xj,t−1 )| < ε, where Pc(Xi,t−1 ) isthe probability of conversion for firm i, and ε isa small positive number. As such, we measure thefollowing difference, which we refer to the averagetreatment effect on the treated (ATT):

�AT T = E[yit |mit(Xi,t−1 ) = W, Xi,t−1 ]

− E[yjt |mjt(Xj,t−1 ) = J, Xj,t−1 ]

wherePc(Xi,t−1 ) ≈ Pc(Xj,t−1 ). (3)

This comparison between Equations (1) and (3)highlights that our new empirical strategy allowsus to focus on a restricted group of whollyowned subsidiaries (which converted at time t)and a restricted comparison group of joint ventures(which ex ante had a similar predicted probabilityof conversion) in order to control for the endo-geneity problem inherent to the choice of foreignmode of operation. To be more specific, ATT iscalculated using the formula below, where sub-script k represents one, two, or three years sinceconversion, similar to Arnold and Javorcik (2009):

AT Tk = 1

n

∑(ROAtreated

t+k − ROAcontrol

t+k )

− 1

n

∑(ROAtreated

t − ROAcontrol

t ).

The corresponding standard errors are calculatedusing the following method with STATA’spsmatch2 command, where n is the number ofmatches (Leuven and Sianesi, 2003):

SEk

=√√√√ 1

nV ar(ROAt+k − ROAt |treatment = 1)

+1nV ar(ROAt+k − ROAt |control = 1)

.



Table 2. Probit models of conversion decision from joint ventures to wholly owned subsidiaries

Dep. variable:JV to WOS conversion

(1)All firms

(2)High R&D

(3)Low R&D

(4)High adv.

(5)Low adv.

Firm size (t−1) −0.07∗∗∗ −0.09∗∗∗ −0.06∗∗ −0.07∗∗∗ −0.07∗∗∗

(0.02) (0.02) (0.02) (0.03) (0.02)Firm age (t−1) −0.02∗∗∗ −0.01∗∗∗ −0.02∗∗∗ −0.02∗∗∗ −0.01∗∗

(4.0 × 10−3) (5.0 × 10−3) (6.0 × 10−3) (6.0 × 10−3) (5.0 × 10−3)ROA (t−1) −4.5 × 10−3∗∗ −5.1 × 10−3∗ −4.3 × 10−3 −9.8 × 10−3∗∗∗ 0.3 × 10−3

(2.0 × 10−3) (3.0 × 10−3) (3.0 × 10−3) (3.0 × 10−3) (3.0 × 10−3)Leverage (t−1) 0.5 × 10−3 1.1 × 10−3 −0.4 × 10−3 −1.9 × 10−3 2.6 × 10−3∗∗

(1.0 × 10−3) (1.0 × 10−3) (1.0 × 10−3) (1.0 × 10−3) (1.0 × 10−3)Export ratio (t−1) 0.08∗ 0.13∗ 0.06 0.01 0.17∗∗

(0.05) (0.07) (0.07) (0.07) (0.07)Intangible assets ratio (t−1) 6.0 × 10−3 1.2 × 10−2∗∗ −1.3 × 10−3 3.0 × 10−3 7.2 × 10−3

(4.0 × 10−3) (6.0 × 10−3) (6.0 × 10−3) (6.0 × 10−3) (5.0 × 10−3)Fixed assets ratio (t−1) 2.7 × 10−3∗∗ 3.2 × 10−3∗∗ 2.5 × 10−3∗ 2.8 × 10−3∗ 2.5 × 10−3

(1.0 × 10−3) (1.6 × 10−3) (1.4 × 10−3) (1.8 × 10−3) (2.0 × 10−3)Conventional local −0.20∗∗∗ −0.23∗∗∗ −0.18∗∗∗ −0.21∗∗∗ −0.19∗∗∗

parent (t−1) (0.04) (0.06) (0.06) (0.06) (0.06)HMT foreign 0.04 0.02 0.04 0.09∗ −0.02parent (t−1) (0.04) (0.06) (0.05) (0.06) (0.05)Foreign share (t−1) 13.28∗∗∗ 12.94∗∗∗ 14.26∗∗∗ 16.76∗∗∗ 10.70∗∗∗

(1.78) (2.62) (2.50) (2.80) (2.23)Foreign share −15.80∗∗∗ −15.79∗∗∗ −16.68∗∗∗ −20.27∗∗∗ −12.42∗∗∗

squared (t−1) (2.33) (3.47) (3.25) (3.69) (2.90)Industry fixed effects Included Included Included Included IncludedYear fixed effects Included Included Included Included IncludedRegion fixed effects Included Included Included Included IncludedPseudo-R2 0.14 0.14 0.15 0.18 0.11Chi-squared 953.37∗∗∗ 446.40∗∗∗ 538.85∗∗∗ 619.31∗∗∗ 376.14∗∗∗

(p-value) (0.00) (0.00) (0.00) (0.00) (0.00)Observations 19,557 9,195 9,603 8,757 10,262

[Note] Standard errors in parentheses; ∗,∗∗,∗∗∗ indicate statistical significance at the 10%, 5%, and 1% level, respectively.

A simple t-test is then employed to confirmwhether the accumulated difference between thesetwo groups is significant. As we expect that thegains from conversion will be greater for firms inindustries characterized by a high level of intan-gible assets as predicted by the transaction costtheory, we contrast the difference in two subgroupsthat are relatively high or low in intangible assets.

RESULTS

Operating mode choice and performance

Table 2 displays our probit regression results forthe conversion decision from joint venture towholly owned subsidiary. All regressions controlfor industry, year, and region fixed effects. The firstcolumn reports the coefficients from the regressionusing the full sample; the remaining four columnsreport coefficients from subsamples of high/low

R&D and advertising-intensive industries. We con-duct matching separately for each subsample ofindustries with high/low R&D or advertising inten-sity.6 Despite being estimated with different sub-samples, these regressions demonstrate generallyconsistent results.

Joint ventures with more assets are less likely tobe converted, as the foreign partner requires morecapital to buy out the local partner. Older joint ven-tures are less likely to be converted, contrary to theexpectation of the learning perspective (Hennart,2009). Joint ventures that are more profitable areless likely to be converted in the whole sample andin the subsample of high advertising intensity, aslocal joint venture partners are less willing to partwith profitable joint ventures, as predicted by the

6 As we conduct matching for each subsample, the number ofobservations in high and low R&D/advertising-intensive indus-tries does not necessarily add up to that in the whole sample.



real option perspective (Kogut, 1991). Leverageinfluences the conversion decision negatively onlyin the case of low advertising industries. Exportratio is positively associated with the likelihood ofconversion in the whole sample, as well as thehigh R&D and low advertising-intensive indus-try samples, reflecting the real option value ofexport platform (Kogut and Chang, 1996). Ratiosof intangible assets to total assets are positivelyassociated with the likelihood of conversion in thehigh R&D-intensive industry, in line with the pre-diction of the transaction cost theory (Hennart,1982). This thereby suggests that the higher theportion of intangible assets, the more likely it isthat a venture in a high R&D-intensive industrywill convert. The ratios of fixed assets to totalassets have a significantly positive association withthe likelihood of conversion in all samples exceptfor the low advertising-intensive industry subsam-ple, thereby suggesting that firms that require ahigh fixed asset investment are more likely to beconverted to wholly owned subsidiaries.

Among the parent type variables, which reflectvarious ‘initial conditions’ and the cultural dis-tances that affect the joint venture stability/termination, joint ventures with conventional localfirms are less likely to convert, as the foreignpartner finds it harder to sever its relationshipwith state-owned enterprises or collectives thanwhen local partners are private firms or incorpo-rated firms. On the other hand, the type of for-eign parents—HMT firms or nonethnic-Chinesefirms—does not seem to have much influence onthe conversion decision. The foreign share variableis significant and positive, while its squared termis significant and negative, suggesting that, as pre-dicted, the relationship between the level of foreignownership and the likelihood of conversion is aninverted U shape. This result may be due to the factthat the likelihood of converting increases whenforeign investors have higher shares but decreasesafter a certain level since the additional benefitof converting may be smaller for joint venturesin which foreign investors already have relativelylarge shares.

Table 3a shows the value of ROA for boththe converted and the remaining joint venturesover time, with the matching performed within thethree-digit SIC industry code and within the sameyear. This procedure generates matching for 799out of 840 converted wholly owned subsidiarieswith the same number of continuing joint ventures

from the full sample of 19,557 minority joint ven-tures that remained in our database for at least fiveyears. The median matching distance is 0.0025.Table 3a shows the results from the difference-in-differences estimation, which denotes the differ-ences in the performance measure estimates forthe 799 converted wholly owned subsidiaries rela-tive to the performance for their counterparts in thecontrol group. The ATT measures the difference incumulative changes in ROA since the year of con-version (at time t) between the two groups. Theresults demonstrate that, for each of the first threeyears after conversion, converted wholly ownedsubsidiaries, on average, have 1.12, 1.21, and 1.55percentage points greater increase in ROA thanfirms that remained joint ventures with similarobserved characteristics. These estimates are sig-nificant at the five percent level. Such superior per-formance of converted wholly owned subsidiariesover continuing joint ventures seems jointly con-ditioned by improved performance of the formerand declining performance of the latter. Whileconverted wholly owned subsidiaries might haveimproved or maintained their profitability facedwith intensifying competition in China during ourstudy time period, continuing joint ventures mightnot have been able to do so.

Table 3a also displays the performance differ-ences between the converted wholly owned sub-sidiaries and continuing joint ventures in subsam-ples of high/low R&D- and advertising-intensiveindustries in an analogous way to De Loecker(2007), Girma and Gorg (2007), and MacGarvie(2006), who presented separate results betweentwo subgroups. The difference in ROA increaseof the converted wholly owned subsidiaries com-pared to continuing joint ventures in high R&D-intensive industries is 1.49 percentage points inthe first year after conversion, 1.60 percentagepoints in the second year, and 1.76 percentagepoints in the third. These differences are statis-tically significant at the five percent level. In con-trast, there are no significant differences in ROAbetween these two groups for firms in low R&D-intensive industries. ROA for converted whollyowned subsidiaries relative to continuing joint ven-tures in high advertising-intensive industries is2.36 percentage points greater in the first yearafter conversion, which is significant at the onepercent level. In the second year, this differenceis 1.84 percentage points, and it is 2.08 per-centage points in the third year. These estimates



Tabl

e3.

Fina

ncia

lpe

rfor

man

ceof

join

tve

ntur

estu

rned

who

llyow

ned

subs

idia

ries

vs.

cont

inui

ngjo

int

vent

ures

a)R

OA

(%)

over

time

Who

lesa

mpl

eH

igh

R&

Din

tens

ityin

dust

ries

Low

R&

Din

tens

ityin

dust

ries

Yea

rt−

1t

t+1

t+2

t+3

Yea

rt−

1t

t+1

t+2

t+3

Yea

rt−

1t

t+1

t+2

t+3

JV-t

o-W

OS

4.20

4.86

5.73

5.49

5.31

JV-t

o-W

OS

3.96

4.43

5.65

5.63

4.92

JV-t

o-W

OS

4.43

5.21

5.68

5.28

5.55

JV4.

244.

924.

674.

343.

82JV

4.44

4.68

4.40

4.27

3.41

JV4.

915.

055.

084.

324.

00A

TT

1.12

∗∗1.

21∗∗

1.55

∗∗A

TT

1.49

∗∗1.

60∗∗

1.76

∗∗A

TT

0.44

0.80

1.39

S.E

.0.

510.

610.

68S.

E.

0.75

0.77

0.89

S.E

.0.

750.

960.

97#

Mat

ches

799

799

799

799

799

#M

atch

es38

838

838

838

838

8#

Mat

ches

410

410

410

410

410

Hig

had

vert

isin

gin

tens

ityin

dust

ries

Low

adve

rtis

ing

inte

nsity

indu

stri

es

Yea

rt−

1t

t+1

t+2

t+3

Yea

rt−

1t

t+1

t+2

t+3

JV-t

o-W

OS

3.44

4.08

5.63

5.37

5.11

JV-t

o-W

OS

5.16

5.54

5.78

5.45

5.54

JV3.

414.

603.

794.

053.

54JV

5.39

5.77

5.15

4.56

4.94

AT

T2.

36∗∗

∗1.

84∗∗

2.08

∗∗A

TT

0.86

1.12

0.83

S.E

.0.

710.

850.

93S.

E.

0.66

0.78

0.88

#M

atch

es41

441

441

441

441

4#

Mat

ches

386

386

386

386

386

b)O

pera

ting

RO

A(%

)ov

ertim

e

Who

lesa

mpl

eH

igh

R&

Din

tens

ityin

dust

ries

Low

R&

Din

tens

ityin

dust

ries

Yea

rt−

1t

t+1

t+2

t+3

Yea

rt−

1t

t+1

t+2

t+3

Yea

rt−

1t

t+1

t+2

t+3

JV-t

o-W

OS

4.38

4.82

5.57

5.57

5.56

JV-t

o-W

OS

4.05

4.49

5.54

5.55

5.16

JV-t

o-W

OS

4.71

5.08

5.47

5.47

5.82

JV3.

994.

744.

494.

074.

05JV

4.17

4.46

4.10

3.79

3.82

JV4.

615.

355.

154.

484.

43A

TT

1.00

∗1.

41∗∗

1.43

∗A

TT

1.40

∗1.

73∗∗

1.31

AT

T0.

591.

261.

65S.

E.

0.54

0.63

0.73

S.E

.0.

780.

800.

94S.

E.

0.86

1.06

1.15

#M

atch

es79

979

979

979

979

9#

Mat

ches

388

388

388

388

388

#M

atch

es41

041

041

041

041

0

Hig

had

vert

isin

gin

tens

ityin

dust

ries

Low

adve

rtis

ing

inte

nsity

indu

stri

es

Yea

rt−

1t

t+1

t+2

t+3

Yea

rt−

1t

t+1

t+2

t+3

JV-t

o-W

OS

3.69

4.16

5.57

5.31

5.59

JV-t

o-W

OS

5.30

5.37

5.50

5.76

5.68

JV3.

364.

193.

793.

683.

48JV

5.12

5.47

5.07

4.39

5.44

AT

T1.

80∗∗

1.66

∗2.

13∗∗

AT

T0.

521.

46∗

0.33

S.E

.0.

840.

870.

97S.

E.

0.66

0.84

0.96

#M

atch

es41

441

441

441

441

4#

Mat

ches

386

386

386

386

386

[Not

e]∗,∗

∗,∗∗∗

indi

cate

stat

istic

alsi

gnifi

canc

eat

the

10%

,5%

,an

d1%

leve

l,re

spec

tivel

y.



are significant at the five percent level. In lowadvertising-intensive industries, however, the dif-ferences in ROA are insignificant. These resultsconfirm Hypotheses 1 and 2, which predict that theperformance-enhancing effects of wholly ownedsubsidiaries are stronger in industries character-ized by higher R&D and advertising intensity; theyare also consistent with transaction cost theory. Asillustrated by Table 3a, we do not find any sig-nificant results in industries with low R&D andadvertising intensity.

In Table 3b, we use Operating ROA as an alter-native performance measure. For each of the firstthree years after conversion, converted whollyowned subsidiaries, have, on average, 1.00, 1.41,and 1.43 percentage points greater increase inOperating ROA than firms that remained joint ven-tures with similar observed characteristics. Theseestimates are significant at the 10 percent, five per-cent, and 10 percent level, respectively. Similarto difference in ROA, the difference in OperatingROA increase between converted wholly ownedsubsidiaries relative to continuing joint venturesis more pronounced in high R&D- or advertising-intensive industries. These results confirm that ourfindings are robust over different choices of finan-cial performance variables.

Potential sources of performance improvement

In order to identify potential sources for per-formance improvement, Tables 4a–4c show thechanges of key strategic indicators for continu-ing joint ventures and converted wholly ownedsubsidiaries. We trace sales, intangible assets, andfixed assets that are salient in explaining the per-formance differences between converted whollyowned subsidiaries and continuing joint venturesover time. We use the natural logarithm of sales,intangible assets, and fixed assets since the dif-ferences in the natural logarithm of these vari-ables can be interpreted as percentage changesover time.7 The results suggest that the con-verted wholly owned subsidiaries show substantialincreases in sales and intangible assets, when com-pared to continuing joint ventures.

Table 4a displays the results for the natural log-arithm of sales over time. In the first and second

7 Let rt be the rate of change in xt over time, then: ln xt −ln xt−1 = ln

(xt

xt−1

)= ln(1 + rt ) ≈ rt , when rt is small (typi-

cally, less than 0.2) as in most of our cases.

year after conversion, there is a five percentagepoint and a six percentage point difference in salesgrowth rate, which is significant at the 10 percentlevel. These differences are greater in high R&D-intensive industries than in low R&D-intensiveindustries. These differences are also greater inhigh advertising-intensive industries than in lowadvertising-intensive industries, but the estimatesin high advertising-intensive industries are not sta-tistically significant.

We expected that foreign parents would bringmore sophisticated technology or brands to theirconverted wholly owned subsidiaries. If a technol-ogy transfer or new brand introduction is associ-ated with conversion, there should be an increasein subsidiary-level intangible assets in the bal-ance sheet. Table 4b examines the natural loga-rithm of intangible assets over time, which cap-tures new transfers of technology or brand fromforeign parents to local subsidiaries. By the sec-ond year after conversion, there is a 38 percent-age point difference in intangible asset growthrate, which is significant at the five percent level.This difference grows to 52 percentage points inthe third year after conversion, which is signifi-cant at the one percent level. The different growthrates for these two groups is more pronounced inhigh advertising-intensive industries than in lowadvertising-intensive industries, and also in highR&D-intensive industries compared to low R&D-intensive industries. The fact that the increase ofintangible assets is more pronounced in high R&D-and advertising-intensive industries is consistentwith our expectation that conversions improveperformance by allowing multinational parents tobring more intangible assets to their now whollyowned subsidiaries. In low R&D- or advertising-intensive industries, we find no significant differ-ences between converted subsidiaries and contin-uing joint ventures.

Table 4c examines the natural logarithm of fixedassets, which captures new investment in physicalcapital, for example, the transfer of technology thatis embedded in sophisticated machinery or equip-ment, especially in technology-intensive industries.By the third year after conversion, there is a 10 per-centage point difference, which is significant at thefive percent level. The net increase in fixed assetsis more pronounced in high R&D-intensive indus-tries than in low R&D-intensive industries, in linewith our expectation. However, we also find thatthe asset growth is slightly more pronounced in



Tabl

e4.

Cha

nges

inke

yst

rate

gic

indi

cato

rsa)

Log

sale

sov

ertim

e Who

lesa

mpl

eH

igh

R&

Din

tens

ityin

dust

ries

Low

R&

Din

tens

ityin

dust

ries

Yea

rt−

1t

t+1

t+2

t+3

Yea

rt−

1t

t+1

t+2

t+3

Yea

rt−

1t

t+1

t+2

t+3

JV-t

o-W

OS

10.1

310

.24

10.3

710

.48

10.5

2JV

-to-

WO

S10

.12

10.2

710

.38

10.5

210

.59

JV-t

o-W

OS

10.1

110

.22

10.3

410

.41

10.4

4JV

10.1

210

.21

10.2

910

.38

10.4

2JV

10.1

410

.25

10.3

310

.41

10.4

7JV

9.98

10.0

610

.14

10.2

010

.23

AT

T0.

05∗

0.06

∗0.

06A

TT

0.03

0.10

∗∗0.

10∗

AT

T0.

040.

060.

05S.

E.

0.03

0.03

0.05

S.E

.0.

030.

050.

06S.

E.

0.04

0.06

0.08

#M

atch

es79

979

979

979

979

9#

Mat

ches

388

388

388

388

388

#M

atch

es41

041

041

041

041

0

Hig

had

vert

isin

gin

tens

ityin

dust

ries

Low

adve

rtis

ing

inte

nsity

indu

stri

es

Yea

rt−

1t

t+1

t+2

t+3

Yea

rt−

1t

t+1

t+2

t+3

JV-t

o-W

OS

10.0

610

.18

10.2

910

.40

10.4

5JV

-to-

WO

S10

.22

10.3

310

.44

10.5

510

.59

JV10

.05

10.1

410

.23

10.3

010

.33

JV10

.14

10.2

610

.36

10.4

410

.52

AT

T0.

030.

060.

08A

TT

0.02

0.05

0.01

S.E

.0.

040.

050.

07S.

E.

0.03

0.05

0.06

#M

atch

es41

441

441

441

441

4#

Mat

ches

386

386

386

386

386

b)L

ogin

tang

ible

asse

tsov

ertim

e

Who

lesa

mpl

eH

igh

R&

Din

tens

ityin

dust

ries

Low

R&

Din

tens

ityin

dust

ries

Yea

rt−

1t

t+1

t+2

t+3

Yea

rt−

1t

t+1

t+2

t+3

Yea

rt−

1t

t+1

t+2

t+3

JV-t

o-W

OS

2.57

2.55

2.90

3.09

3.10

JV-t

o-W

OS

2.78

2.75

3.12

3.19

3.31

JV-t

o-W

OS

1.99

2.08

2.09

2.10

2.10

JV2.

783.

013.

153.

163.

04JV

2.57

2.64

2.64

2.90

2.84

JV1.

662.

252.

262.

272.

28A

TT

0.20

0.38

∗∗0.

52∗∗

∗A

TT

0.38

∗∗0.

180.

37A

TT

−0.0

10.

00−0

.01

S.E

.0.

140.

160.

19S.

E.

0.19

0.23

0.25

S.E

.0.

010.

010.

01#

Mat

ches

799

799

799

799

799

#M

atch

es38

838

838

838

838

8#

Mat

ches

410

410

410

410

410

Hig

had

vert

isin

gin

tens

ityin

dust

ries

Low

adve

rtis

ing

inte

nsity

indu

stri

es

Yea

rt−

1t

t+1

t+2

t+3

Yea

rt−

1t

t+1

t+2

t+3

JV-t

o-W

OS

2.38

2.34

2.77

2.96

2.97

JV-t

o-W

OS

2.86

2.87

3.04

3.28

3.28

JV2.

442.

722.

782.

932.

86JV

3.11

3.21

3.34

3.38

3.34

AT

T0.

36∗∗

0.41

∗0.

49∗∗

AT

T0.

050.

240.

28S.

E.

0.18

0.22

0.24

S.E

.0.

190.

210.

25#

Mat

ches

414

414

414

414

414

#M

atch

es38

638

638

638

638

6



Tabl

e4.

(Con

tinu

ed)

c)L

ogfix

edas

sets

over

time

Who

lesa

mpl

eH

igh

R&

Din

tens

ityin

dust

ries

Low

R&

Din

tens

ityin

dust

ries

Yea

rt−

1t

t+1

t+2

t+3

Yea

rt−

1t

t+1

t+2

t+3

Yea

rt−

1t

t+1

t+2

t+3

JV-t

o-W

OS

8.60

8.70

8.67

8.73

8.76

JV-t

o-W

OS

8.71

8.85

8.92

9.01

9.00

JV-t

o-W

OS

8.49

8.57

8.61

8.66

8.73

JV8.

668.

718.

708.

728.

72JV

8.76

8.84

8.85

8.86

8.83

JV8.

408.

468.

498.

508.

50A

TT

0.02

0.07

∗0.

10∗∗

AT

T0.

070.

14∗∗

∗0.

17∗∗

AT

T0.

020.

060.

12∗

S.E

.0.

030.

040.

05S.

E.

0.05

0.05

0.07

S.E

.0.

050.

060.

07#

Mat

ches

781

781

781

781

781

#M

atch

es37

837

837

837

837

8#

Mat

ches

402

402

402

402

402

Hig

had

vert

isin

gin

tens

ityin

dust

ries

Low

adve

rtis

ing

inte

nsity

indu

stri

es

Yea

rt−

1t

t+1

t+2

t+3

Yea

rt−

1t

t+1

t+2

t+3

JV-t

o-W

OS

8.59

8.68

8.73

8.78

8.82

JV-t

o-W

OS

8.64

8.75

8.83

8.90

8.92

JV8.

588.

688.

768.

768.

72JV

8.67

8.71

8.77

8.78

8.76

AT

T−0

.04

0.01

0.10

AT

T0.

030.

080.

12∗

S.E

.0.

040.

050.

07S.

E.

0.05

0.05

0.07

#M

atch

es40

340

340

340

340

3#

Mat

ches

379

379

379

379

379

[Not

e]∗,∗

∗,∗∗∗

indi

cate

stat

istic

alsi

gnifi

canc

eat

the

10%

,5%

,an

d1%

leve

l,re

spec

tivel

y.



low advertising-intensive industries than in highadverting-intensive industries. This suggests thatconverted wholly owned subsidiaries may bringin more brands in the form of intangible assetinvestments but not necessarily by increasing fixedasset investments like building factories or pur-chasing machinery. This contrast between intangi-ble and fixed assets reconfirms that technology isoften embodied by more sophisticated machineryand equipment in high R&D- intensive industries,but brand is not necessarily embodied in physicalassets.

In addition, we also trace export ratio and finan-cial leverage for each time period in order to seewhether the improved performance of convertedwholly owned subsidiaries might be attributable toan increase in multinational parents’ export salesafter the conversion or to financial restructuring.We find no difference between these two groups.These results are available upon request.

While these additional analyses do not neces-sarily confirm the causal relationships betweenthese key strategic indicators and performance,they provide circumstantial evidence that suchincreases in intangible assets may contribute tohigher sales growth and profitability for convertedwholly owned subsidiaries over continuing jointventures. This notion finds support in the resource-based theory (Wernerfelt, 1984).

Robustness check

We perform additional robustness tests to ensurethe validity of our results. First, we check whetherour results are sensitive to how we define a jointventure, which is currently set as a 25–50 percentforeign share. We relax the lower bound of ourdefinition of minority joint venture by includingfirms with 20–25 percent shares. We also lower theupper bound to 45 percent without losing too manyobservations, and then raise the upper bound to 55percent without adding too many more majorityjoint ventures. Results are consistent with thosereported in this study. Second, we estimate resultswithout dropping extreme outliers whose ROA val-ues are greater than 100 percent or below −50percent, which also generates consistent results.8

8 If we do not drop firms whose ROAs are greater than 100 per-cent or below −50 percent, we are left with 801 instead of 799matched samples of converted wholly owned subsidiaries andcontinuing joint ventures. While results are consistent with those

Third, we employ different cutoff points for highand low R&D or advertising intensity, for example,rounding the median values for R&D and adver-tising intensity to one percent and two percent,respectively. Our findings are also robust to differ-ent cutoff points for high/low R&D and advertisingintensities. All additional robustness test results areavailable upon request.

DISCUSSION AND CONCLUSION

This paper evaluates the differences in perfor-mance between joint ventures and wholly ownedsubsidiaries. To circumvent the endogeneity prob-lem inherent to research in this vein, we adopta propensity score matching technique combinedwith the difference-in-differences approach, whichtogether allow us to focus on the comparisonbetween joint ventures that were converted towholly owned subsidiaries and continuing jointventures. We compare these two groups’ per-formance for three years following this conver-sion, after controlling for endogeneity in operatingmode choice. We detect a steady improvementin the performance of converted wholly ownedsubsidiaries, as measured by ROA and Operat-ing ROA, which far exceeds that of continuingjoint ventures. We also find this improvement tobe more pronounced in industries characterized byhigh R&D and advertising intensity. Lastly, ourresults suggest that enhanced performance fromownership conversion is associated with increasedsales and increased intangible and fixed assets.

This study offers several contributions to theliterature on international operating mode choice.First, it supports transaction cost theory’s predic-tion that wholly owned subsidiaries perform betterthan joint ventures in industries characterized byhigh levels of intangible resources, after control-ling for the local knowledge or resources that jointventure partners can contribute. In contrast to mostprior work, which focuses on the ex ante modechoice, we find that the operation mode dictatedby transaction cost theory significantly and posi-tively affects a firm’s ex post performance. Whilesome previous studies, for example, Reuer (2001),

presented in the paper, these two additional matches of extremeoutliers tend to depress the performance of converted joint ven-tures in low R&D-intensive industries. As a consequence, con-verted wholly owned subsidiaries outperform continuing jointventures, even in low R&D-intensive industries.



show that joint venture buyouts positively impactabnormal stock returns of parent firms with highR&D intensity, this study directly tests the per-formance change at the foreign subsidiary levelafter acquisition, which strongly supports transac-tion cost theory.

Our study also makes several important empir-ical contributions. First, it sheds light on post-entry performance, a previously understudied phe-nomenon due to methodological obstacles and dataconstraints. Relying on the Annual Industrial Sur-vey Database, we could evaluate subsidiary-levelfinancial performance of foreign firms in China.Second, we explore potential mechanisms wherebywholly owned subsidiaries might perform moreeffectively than joint ventures. Our results show,albeit indirectly, that converted wholly owned sub-sidiaries enable multinational firms to bring inmore sophisticated technology and brands thanjoint ventures. The demonstration of the potentialmechanism of ownership transition further sup-ports our hypotheses. Finally, we exploit a uniqueempirical setting to observe the performance dif-ference between joint ventures and wholly ownedsubsidiaries. This empirical strategy allows us toaddress the endogeneity issue by finding coun-terfactual samples of joint ventures for matching,using the propensity score matching technique, andthen compare their performances with those of theconverted wholly owned subsidiaries.

Furthermore, our empirical context of joint ven-tures in China from 1998 to 2006 provides an idealsetting for our research questions. As China pre-pared to join the WTO, the joint venture require-ment was relaxed in most industries and manyjoint ventures converted to wholly owned sub-sidiaries. Even after this requirement was lifted, asubstantial number of joint ventures continued asthey were. This empirical setting provides enoughcases for both groups to compare their respec-tive performances. It is, however, important tostress that our findings are not driven by this jointventure requirement policy per se, because ourpropensity matching technique creates counterfac-tual foreign joint ventures in the same three-digitindustry that are just as likely to convert to whollyowned subsidiaries as the ventures that actuallyconverted. While joint venture requirements couldhave affected ‘all’ firms in the same three-digitindustries, our study shows that only some jointventures converted while others in the same indus-tries did not, despite the same hazard rate. This

suggests that the joint venture requirements inChina were not the main driver for our results.Rather, our findings suggest that the existence oftacit and poorly protected proprietary intangibleassets to be the main reason for the differences; theconversion to wholly owned subsidiaries improvedperformance only in industries characterized byhigh levels of intangible assets. If the joint venturerequirement led to suboptimal investment and theconversion to wholly owned subsidiaries was theanswer to this problem, firm performance shouldimprove after conversion regardless of the degreeof intangible assets. We find no evidence of suchimprovement in industries characterized by lowlevels of intangible assets. In other words, ourresults are not context-specific; they are applicableto other countries not characterized by constraintson foreign ownership.

It is worthwhile to note that not all joint ven-tures in industries with high levels of intangibleassets converted. At the same time, we find thatmany joint ventures with low levels of intangibleassets did convert, despite having no good rea-son to do so from a transaction cost perspective.We acknowledge that the conversion decision fromjoint ventures to wholly owned subsidiaries canbe driven by ‘unknown’ rational or irrational rea-sons that our probit model cannot fully capture,such as bandwagon effects or stalemate in bar-gaining with local partners. The propensity scorematching technique assumes such unknown ratio-nal or irrational reasons for conversion is commonin both treatment and control groups, and treatsthem as the error term in the probit model. Whilewe try our best to control for all ‘known’ factorsof conversions based on the joint venture termi-nation/instability literature and while we identifyconverted wholly owned subsidiaries and continu-ing joint ventures in the same three-digit industries,future research should consider additional factorsfor conversion.

This study suggests several other areas for futurestudies. First, researchers may conduct a simi-lar experiment in countries without any prece-dent of joint venture requirements. This wouldfurther validate that our findings are not context-specific. Second, researchers may also considerwhy some joint ventures were dissolved withoutbeing acquired by a multinational or local parent.As this study focuses on the performance implica-tion, we are unable to consider the cases of jointventure dissolution. Third, future studies might



also consider some parent-firm specific factors,for example, product relatedness or cross-selling,that may have a longer-run implication on theconversion decision and the post-conversion per-formance. Finally, better performance indicatorsthat take more explicit account of transfer pric-ing schemes may also build on the results pre-sented here. This study suggests that the logic oftransaction costs can help firms choose the bestoperating mode; multinational firms should con-sider this choice jointly with the technology theychoose to transfer to their subsidiaries. Taken asa whole, our results confirm that wholly ownedsubsidiaries have superior performance vis-a-visequivalent joint ventures in industries character-ized by high intangible assets because whollyowned subsidiaries can tap into better technologiesand brands owned by the parent firm. Thus, jointventures should be used selectively when resourcesor skills that local partners can bring in are nototherwise available.

ACKNOWLEDGEMENTS

We thank two anonymous reviewers and Asso-ciate Editor Jeff Reuer for helpful comments andsuggestions. We also thank Jaideep Anand, KatieBrown, Chris C. Chung, Jean-Francois Hennart,Bruce Kogut, Ivan Png, and seminar participantsat National University of Singapore, Shanghai JiaoTong University, Shanghai University of Financeand Economics, and the 2010 Academy of Interna-tional Business Annual Meetings for helpful com-ments and suggestions. Sea-Jin Chang appreciatesfinancial support from the National University ofSingapore, Research Grant # R313-000-086-133.Jaiho Chung and Jon J. Moon appreciate finan-cial support from the SK Research Fund, KoreaUniversity Business School.

REFERENCES

Anand J, Delios A. 1997. Location specificity andthe transferability of downstream assets to foreignsubsidiaries. Journal of International Business Studies28(3): 579–603.

Arnold JM, Javorcik BS. 2009. Gifted kids or pushy par-ents? Foreign direct investment and plant productivityin Indonesia. Journal of International Economics 79:42–53.

Arrow KJ. 1974. The Limits of Organization . W. W.Norton: New York.

Barkema HG, Bell JHJ, Pennings JM. 1996. Foreignentry, cultural barriers, and learning. StrategicManagement Journal 17(2): 151–166.

Blodgett LL. 1992. Factors in the instability ofinternational joint ventures: an event history analysis.Strategic Management Journal 13(6): 475–482.

Brouthers KD, Brouthers LE, Werner S. 2003. Trans-action cost-enhanced entry mode choices and firmperformance. Strategic Management Journal 24(12):1239–1248.

Brouthers KD, Hennart J-F. 2007. Boundaries of thefirm: insights from international entry mode research.Journal of Management 33(3): 395–425.

Chang SJ, Xu D. 2008. Spillovers and competitionamong foreign and local firms in China. StrategicManagement Journal 29(5): 495–518.

Chow GC. 1993. Capital formation and economic growthin China. Quarterly Journal of Economics 108(3):809–842.

Conyon MJ, Girma S, Thompson S, Wright PW. 2002.The productivity and wage effects of foreignacquisition in the United Kingdom. Journal ofIndustrial Economics 50(1): 85–102.

Dehejia RH, Wahba S. 2002. Propensity score-matchingmethods for nonexperimental causal studies. Reviewof Economics and Statistics 84(1): 151–161.

Delios A, Beamish PW. 2001. Survival and profitability:the roles of experience and intangible assets in foreignsubsidiary performance. Academy of ManagementJournal 44(5): 1028–1038.

De Loecker J. 2007. Do exports generate higherproductivity? Evidence from Slovenia. Journal ofInternational Economics 73: 69–98.

Dhanaraj C, Beamish PW. 2004. Effect of equityownership on the survival of international jointventures. Strategic Management Journal 25(3):295–305.

DiMaggio PJ, Powell WW. 1983. The iron cage revisited:institutional isomorphism and collective rationality inorganizational fields. American Sociological Review48(2): 147–160.

Doms ME, Jensen JB. 1998. Comparing wages, skills,and productivity between domestically and foreign-owned manufacturing establishments in the UnitedStates. In Geography and Ownership as Basesfor Economic Accounting, Baldwin R, Lipsey R,Richardson JD (eds.). Chicago University Press:Chicago, IL; 235–258.

Doz YL. 1996. The evolution of cooperation in strategicalliances: initial conditions or learning processes?Strategic Management Journal , Summer Special Issue17: 55–83.

Dunning JH. 1988. The eclectic paradigm of internationalproduction: a restatement and some possible exten-sions. Journal of International Business Studies 19(1):1–31.

Franko LG. 1971. Joint Venture Survival in MultinationalCorporations . Praeger: New York.

Gatignon H, Anderson E. 1988. The multinational corpo-ration’s degree of control over foreign subsidiaries: anempirical test of transaction cost explanation. Journalof Law, Economics, and Organization 4: 305–336.



Girma S, Gorg H. 2007. Evaluating the foreign owner-ship wage premium using a difference-in-differencematching approach. Journal of International Eco-nomics 72: 97–112.

Hennart J-F. 1982. A Theory of Multinational Enterprise.University of Michigan Press: Ann Arbor, MI.

Hennart J-F. 1991. The transaction costs theory of jointventures: an empirical study of Japanese subsidiariesin the United States. Management Science 37:483–497.

Hennart J-F. 2009. Down with MNE-centric theories!Market entry and expansion as the bundling of MNEand local assets. Journal of International BusinessStudies 40: 1432–1454.

Hennart J-F, Kim D-J, Zeng M. 1998. The impact of jointventure status on the longevity of Japanese stakes inU.S. manufacturing affiliates. Organization Science 9:382–395.

Hennart J-F, Park Y-R. 1993. Greenfield vs acquisition:the strategy of Japanese investors in the United States.Management Science 39(9): 1054–1070.

Imbens G, Wooldridge JM. 2009. Recent developmentsin the econometrics of program evaluation. Journal ofEconomic Literature 47(1): 5–86.

Johanson J, Vahlne J-E. 1977. The internationaliza-tion process of the firm: a model of knowledgedevelopment and increasing foreign market commit-ments. Journal of International Business Studies 8:23–32.

Kieso D, Weygandt J, Warfield T. 2010. IntermediateAccounting (13th edn). Wiley: Hoboken, NJ.

Killing JP. 1983. Strategies for Joint Venture Success .Praeger: New York.

Kogut B. 1991. Joint ventures and the option to expandand acquire. Management Science 37(1): 19–33.

Kogut B, Chang SJ. 1996. Platform investments andvolatile exchange rates: direct investment in theU.S. by Japanese electronic companies. Review ofEconomics and Statistics 78(2): 221–231.

Leuven E, Sianesi B. 2003. PSMATCH2: Stata module toperform full Mahalanobis and propensity score match-ing, common support graphing, and covariate imbal-ance testing, http://ideas.repec.org/c/boc/bocode/s432001.html version 3.1.5 (21 November 2009).

Lin C, Png I. 2003. Monitoring costs and the modeof international investment. Journal of EconomicGeography 3(3): 261–274.

MacGarvie M. 2006. Do firms learn from internationaltrade? Review of Economics and Statistics 88(1):46–60.

Madhok A. 1997. Cost, value and foreign marketentry mode: the transaction and the firm. StrategicManagement Journal 18(1): 39–61.

Nitsch D, Beamish PW, Makino S. 1996. Entry modeand performance of Japanese FDI in Western Europe.Management International Review 36: 27–38.

Park S, Li S, Tse D. 2006. Market liberalizationand firm performance during China’s economictransition. Journal of International Business Studies37: 127–147.

Puck JF, Holtbrugge D, Mohr AT. 2009. Beyond entrymode choice: explaining the conversion of jointventures into wholly owned subsidiaries in thePeople’s Republic of China. Journal of InternationalBusiness Studies 40: 388–404.

Rajan RG, Zingales L. 1998. Financial dependence andgrowth. American Economic Review 88(3): 559–586.

Reuer JJ. 2001. From hybrids to hierarchies: shareholderwealth effects of joint venture partner buyouts.Strategic Management Journal 22(1): 27–44.

Reuer JJ, Tong TW. 2005. Real options in internationaljoint ventures. Journal of Management 31(3):403–423.

Reuer JJ, Zollo M, Singh H. 2002. Post-formationdynamics in strategic alliances. Strategic ManagementJounal 23(2): 135–151.

Ries J, Globerman S, Vertinsky I. 1994. The economicperformance of foreign affiliates in Canada. CanadianJournal of Economics 27(1): 143–156.

Rosenbaum PR, Rubin DB. 1983. The central role of thepropensity score in observational studies for causaleffects. Biometrika 70(1): 41–55.

Shaver JM. 1998. Accounting for endogeneity whenassessing strategy performance: does entry modechoice affect FDI survival? Management Science44(4): 571–585.

Smith J, Todd P. 2005. Rejoinder. Journal of Economet-rics 125: 365–375.

Wernerfelt B. 1984. A resource-based view of the firm.Strategic Management Journal 5(2): 171–180.

Williamson OE. 1991. Comparative economic organiza-tion: the analysis of discrete structural alternatives.Administrative Science Quarterly 36: 269–296.

Woodcock CP, Beamish PW, Makino S. 1994. Owner-ship-based entry mode strategies and internationalperformance. Journal of International Business Studies25(2): 253–273.

Yiu D, Makino S. 2002. The choice between jointventure and wholly owned subsidiary: an institutionalperspective. Organization Science 13: 667–683.


WHEN DO WHOLLY OWNED SUBSIDIARIES PERFORM BETTER THAN JOINT VENTURES? · subsidiaries that continued as joint ventures, we examine whether their performance diverges from the moment

Documents