research paper series China and the World Economy Research Paper 2009 / 05 The More the Better ? Foreign Ownership and Corporate Performance in China

We examine the relationship between the degree of foreign ownership and performance of recipient firms, using of panel of 21,582 Chinese firms over the period 2000-2005. We find that joint-ventures perform better than wholly foreign owned and purely domestic firms. Although productivity and profitability initially rise with foreign ownership, they start declining once foreign ownership reaches beyond 64%. This suggests that some domestic ownership is necessary to ensure optimal performance. We rationalize these findings with a model of a joint-venture, where strategic interactions between a foreign and a domestic owner’s inputs may lead to an inverse U-shaped ownership-performance relationship. JEL Classification: F2, G32, L25, O5


Introduction
A vast literature has investigated the effects of Foreign Direct Investment (FDI) on economic growth reaching a consensus that, especially in developing countries, FDI is an engine of growth (see De Mello, 1997, for a survey). Yet, there is still uncertainty about the exact ways in which FDI may foster growth. Factors such as knowledge spillovers from foreign owned to domestic enterprises or technological upgrading could represent possible channels. A wave of papers has tried to measure these spillovers by estimating regressions of the productivity of domestic firms on alternative indicators of the importance of foreign firms in a given sector. Much of this work, however, fails to find evidence of positive spillovers, with some studies reporting negative effects (see Blomström and Kokko, 1998;and Görg and Greenaway, 2004, for surveys).
One condition for spillovers to take place is that foreign owned firms are more productive than their domestic counterparts. A number of papers within the international economics literature, have tested whether or not this is the case, obtaining once again, mixed results 1 . The issue has also been tackled within the privatization literature, which has analyzed the effects of ownership on corporate performance in transition economies, including China (see Megginson and Netter, 2001;and Estrin et al., 2008, for surveys). Although their main focus is generally on the effects of state ownership or private ownership, most studies in this literature control for foreign ownership, and find that foreign owned firms tend to perform better than their domestic counterparts.
A shortcoming of many of these studies, especially those in the international economics literature, is that they simply divide firms into foreign owned and purely domestic using some a priori criterion, and compare various measures of firm performance across the two groups 2 , ignoring the considerable heterogeneity that typically characterizes foreign owned firms. These include wholly foreign owned firms, as well as joint-ventures between a domestic and a foreign partner, with different degrees of foreign ownership. Joint-ventures represent a popular form of 1 Papers that find little or no evidence of superior performance of foreign owned firms compared to their domestic counterparts are Globerman et al. (1994), Griffith (1999aGriffith ( , 1999b, and Benfratello and Sembenelli (2006). Papers that find positive evidence are Doms and Jensen (1998), Girma et al. (2001), Harris (2002), Harris and Robinson (2003), and Temouri et al. (2008). 2 For instance, focusing on UK data, Girma et al. (2001) define a firm as foreign owned if the country of origin of its ultimate holding company is not the UK.
Our aim is to provide, for the first time, an in-depth analysis of this relationship. For this purpose, we use a panel of 21,582 Chinese unlisted firms operating in the entire economy, over the period 2000-2005 4 . The Chinese case represents an ideal laboratory for two reasons. First, China is among the top FDI recipients in the world. Over the past decade, it has accounted for about one-third of gross FDI flows to all emerging markets and about 60% of these flows to Asian emerging markets (Prasad and Wei, 2005). Together with a considerable decline in the number of state-owned and collective firms, and a rise of privately owned enterprises, the increasing number of foreign firms in China has contributed dramatically to its changing corporate landscape 5 . Second, joint-ventures and wholly foreign owned firms coexist, which makes a study of the degree of foreign ownership on corporate performance particularly relevant 6 .
We measure the degree of foreign ownership as the share of a firm's equity owned by foreign investors. We make use of several measures of corporate performance and find that joint-ventures generally perform better than wholly foreign owned and purely domestic firms. This finding, which is robust to defining the degree of foreign ownership on the basis of registration information, can be explained considering that both the domestic and the foreign parties of a joint-venture bring in attributes essential to achieving high performance. Specifically, the former contribute knowledge of the Chinese market and legal environment, as well as important political connections with local governments; and the latter, modern technologies, capital, better corporate governance through monitoring and market discipline, and managerial and international networking skills. We then investigate the exact nature performance. Moreover, he himself admits that his results might not be representative given the small number of foreign owned firms in his sample. 4 Also see Abraham et al. (2007) and Liu (2008) who investigate whether domestic Chinese firms benefit from horizontal spillovers from foreign firms, controlling respectively for whether or not firms are joint-ventures, and for the percentage of the firms' equity owned by foreign investors. Neither of these studies allows for a non-monotonic relationship between the degree of foreign ownership and the performance of recipient firms. Another related paper is Du and Girma (2008), who use data for Chinese firms to investigate the effects of foreign acquisitions on export markets dynamics. 5 Specifically, while until 1993, state-owned enterprises still dominated the scenario, in 2005, they represented less than 5% of the total number of firms operating in China, while private firms represented 73.53%; collective firms, 6.33%; and foreign firms, 15.38% (Guariglia et al., 2008). 6 In the early days of China's opening up, foreign agents invested mainly into large and well performing state owned enterprises (SOEs) in the form of joint-ventures. Fully foreign owned firms were only allowed later. Until 2000, they were required to either provide advanced technology or to be primarily export-oriented, and were located mainly in the special economic zones, where they received favorable tax treatment, and benefited from streamlined regulations (Dougherty and McGuckin, 2002). In recent years, fully foreign owned firms have become increasingly widespread, and joint-ventures have more and more frequently become partnerships with private rather than state owned enterprises. of the relationship between foreign ownership and corporate performance of our Chinese firms, and find that the two variables are linked by an inverted U-shaped relationship: corporate performance initially rises with foreign ownership, but declines once foreign ownership reaches beyond 64%. This suggests a certain degree of domestic ownership is necessary to ensure optimal performance. Furthermore, we show that it is those firms owned by investors other than those originating from Hong Kong, Macao, and Taiwan that benefit most from their foreign ownership.
To underpin these empirical findings, we construct a simple theoretical model of a joint-venture, where a domestic and a foreign owner contribute non-contractible inputs to boost the company's productivity. The choice of the level of such inputs by each owner depends on his/her share in the firm's ownership and profits, as well as on the inputs contributed by the other owner. Under plausible conditions, the model yields predictions of a non-monotonic relationship between the firm's degree of foreign ownership and its performance similar to that found in our Chinese data.
Our findings contribute to understanding the link between FDI and economic growth: in countries where FDI inflows are large and mainly take the form of jointventures between domestic and foreign firms, the effect of FDI penetration on the performance of recipient firms could be an important channel through which FDI affects economic growth. Attracting more FDI in the form of joint-ventures could hence be beneficial to long-run growth.
The remainder of the paper is organized as follows. In section 2, we describe our data and present some descriptive statistics. Section 3 illustrates our baseline specification and our estimation methodology. Section 4 describes and evaluates our regression results. Section 5 presents our theoretical model, and section 6 concludes.

2.
Data and summary statistics 2. Kong, Macao, and Taiwan. We will also provide specifications showing whether the effects of ownership by the two groups of foreign investors are different. We will verify the robustness of our results to the use of registration-based ownership measures. However, defining ownership categories on the basis of the fraction of capital paid in by various groups is preferable to using registration codes. The latter are in fact updated only with considerable delay (Dollar and Wei, 2007). Moreover, firms might have an incentive to falsely register as foreign simply to take advantage of the tax benefits accorded to the latter.

Summary statistics:
We divide our observations into four categories on the basis of the share of capital paid in by foreign investors 9 . Our first category encompasses those firm-years with no foreign participation, which make up 60.95% of our sample. Our second category contains those observations with a share of foreign capital, which is positive but lower than 50% (10.47% of our sample); our 7 We have excluded listed firms from our analysis as information on their ownership was not available.
Most of the studies that tried to assess the effects of foreign ownership on firm performance focused on the manufacturing sector. In the Chinese context, it is important to consider other sectors as well, as non-manufacturing sectors also attract significant levels of foreign investment (see Table A2 in Appendix 1C for details). All our results were robust to considering only firms operating in the manufacturing sector. 8 See Appendixes 1A and 1B for details about the structure of our panel, and complete definitions of all variables used. 9 Our observations refer to firm-years. We therefore allow our firms to switch across ownership categories each year. third category includes observations with a share higher than or equal to 50% but lower than 100%, (8.61% of our sample); our final category contains firms that are 100% foreign owned, (19.07% of our sample).
Our empirical analysis focuses on four measures of corporate performance: the Return on Assets (ROA, the ratio of the firm's net income to total assets); the return on sales (ROS, the ratio of the firm's net income to its total sales); labor productivity (PROD, the ratio of the firm's net income to number of employees); and Total Factor Productivity (TFP, measured using the Levinsohn and Petrin, 2003, method) 10 . All four have been frequently used in the literature assessing the effects of government ownership on corporate performance (Tian and Estrin, 2008;Jiang et al., 2008) 11 . Although they exhibit better performance than the purely domestic firms, firms that are fully (or almost fully) owned by foreign investors are unlikely to perform as well as joint-ventures, due to a limited knowledge of the Chinese market, legal, regulatory, 10 A key issue in the estimation of production functions is the correlation between unobservable productivity shocks and input levels. Profit-maximizing firms respond to positive productivity shocks by expanding output, which requires additional inputs; and to negative shocks, by decreasing output and input usage. Olley and Pakes' (1996) estimator uses investment as a proxy for these unobservable shocks. This could cause problems as any observation with zero investment would have to be dropped from the data. Levinsohn and Petrin (2003), by contrast, introduce an estimator which uses intermediate inputs as proxies, arguing that these (which are generally non-zero) are likely to respond more smoothly to productivity shocks. 11 Jiang et al. (2008) note that in the Chinese context, extraordinary income and income from non-core operations may be subject to manipulations: they therefore suggest an alternative measure of corporate performance, which excludes them, namely the ratio of operating profits to total assets. All our results were robust to using this alternative measure of firm performance. 12 See Appendix 1C for descriptive statistics of foreign presence by industry and province. and bureaucratic environment, and workers' attitudes towards incentives, as well as to a lack of political connections with local governments (Guanxi), which are often considered as a key factor in determining firm performance in China (Hsieh and Klenow, 2007). The best performers in the Chinese economy are therefore jointventures with a sufficiently large share of domestic ownership. A minority of these joint-ventures are between foreign and state firms. Specifically, those firm-years with a positive but lower than 100% share of capital paid in by foreign investors exhibit an average share of private capital of 35.74%, and an average share of state capital of 8.93%.
Focusing on the other variables reported in Table 1, we see that firm size, measured by the logarithm of its total assets, also increases with the degree of foreign ownership, and declines for 100% foreign owned firms. Leverage, defined as the ratio of the firm's total liabilities to total assets declines monotonically as the share of foreign ownership increases; and collateral, defined as the firm's ratio of tangible fixed assets to total assets, remains approximately constant across the four categories.
Finally, only 36.7% of the purely domestic firm-years export, while the corresponding percentages for joint-ventures and fully foreign owned firm years are, respectively, 70.7% and 86.0%.
In the sections that follows, we first formally analyze the extent to which, controlling for other relevant variables, the performance of minority foreign-owned, majority foreign owned, and fully foreign owned firms differs from that of their purely domestic counterparts. We then investigate the exact nature of the relationship between foreign ownership and corporate performance.

Baseline specification and estimation methodology
3.1 Baseline specifications: We initially estimate the following equation: (1) PERF it = a 0 + a 1 PERF i(t-1) + a 2 Minority foreign it + a 3 Majority foreign it + + a 4 All foreign it + a 5 Size it + a 6 Leverage it + a 7 Collateral it + a 8 Expdum it + where the subscript i indicates firms, and t, time. PERF it indicates in turn our four performance indicators. We control for foreign ownership by including the dummy variables Minority foreign it , Majority foreign it , and All Foreign it . Minority foreign it is equal to 1 if the share of foreign ownership is positive but lower than 50%, and 0 otherwise; Majority foreign it is equal to 1 if the share is greater than or equal to 50% but lower than 100%, and 0 otherwise; and All Foreign it is equal to 1 if the firm is 100% foreign owned, and 0 otherwise. The omitted category is a dummy equal to 1 for purely domestic firms, and 0 otherwise. This exercise is aimed at determining the extent to which the performance of fully foreign owned firms and joint-ventures characterized by different degrees of foreign participation differs from that of purely domestic firms 13 .
The other regressors in equation (1) are motivated by the finance literature (e.g. McConnel and Servaes, 1990). Size is included to control for the fact, that as suggested by Chhibber and Majumdar (1999), larger firms may benefit from economies of scale and better access to external finance, which might enhance their profitability. Collateral it is expected to affect profitability negatively, as firms with more intangible assets are expected to have more investment opportunities and grow faster (Tian and Estrin, 2008). Finally, Leverage it is also expected to have a negative impact on firm performance, due to the debt overhang problem (Jensen and Meckling, 1976;Myers, 1977) 14 . Because more than 50% of Chinese firms in our sample export, we also include a dummy Expdum it , which is equal to 1 if the firm exports, and 0 otherwise 15 .
The error term in equation (1)  13 It has been argued that comparing foreign firms with all domestic firms may lead to a selection problem as domestic firms may include domestic multinationals, which are likely to be as productive as foreign multinationals (Criscuolo and Martin, 2005). This is not an issue in the Chinese case considering the small size of China's outward FDI (Morck et al., 2008). 14 In the presence of a high debt to assets ratio, debt holders will share future investment returns, which might lead the firm to forego some profitable investment opportunities. 15 One could argue that foreign ownership and other firm characteristics affect corporate performance with a lag. All our results were robust to using lagged values of our regressors. This also addresses the issue of possible reverse causality in the relationship between foreign ownership and corporate performance, which is extensively discussed in Section 4.2.
To better understand the nature of the relationship between foreign ownership and firm performance, we next estimate the following variant of equation (1), which includes the actual percentage of the firm's capital paid in by foreign investors, (2) PERF it = a 0 + a 1 PERF i(t-1) + a 2 Foreigncap it + a 3 Foreigncap 2 it + a 4 Size it + +a 5 Leverage it + a 6 Collateral it + a 7 Expdum it + a 8 Statecap it + + a 9 Privatecap it +v i + v t + v jt + e it Equation (2) includes both Foreigncap it and Foreigncap 2 it to account for the possible non-linearity of the relationship between Foreigncap it and PERF it . Considering the large literature on the effects of privatization on corporate performance , the percentage of the firm's total capital paid in by the state (Statecap it ) and by private investors (Privatecap it ) are also included 16 .

3.2
Estimation methodology: All equations are estimated in first-differences, to control for firm-specific, time-invariant effects. Given possible endogeneity of the regressors, we use a first-difference Generalized Method of Moments (GMM) approach 17 . Two or more lags of each of the regressors are used as instruments.
To check whether the first-difference GMM estimator is likely to suffer from finite sample bias, we compared the GMM and the Within Groups estimates of the coefficient on the lagged dependent variable in equation (1). Because the Within Groups estimate is typically downward biased in short panels (Nickell, 1981), one would expect a consistent estimate of the coefficient on the lagged dependent variable to lie above this. As our GMM coefficient was larger than its Within Groups counterpart, we concluded that the first-difference GMM estimates are unlikely to be subject to serious finite sample bias 18 .
To evaluate whether our instruments are legitimate and our model is correctly specified, we use the test for second-order serial correlation of the residuals in the differenced equation (m2). The m2 test is asymptotically distributed as a standard normal under the null of no second-order serial correlation of the differenced residuals, and provides a check on the specification of the model and legitimacy of variables dated t-2 as instruments in the differenced equation 19 .

4.1
Are joint-ventures the best performers? Estimates of equation (1) are reported in columns 1 to 4 of Table 2, which are respectively based on ROA, ROS, PROD, and TFP as measures of corporate performance 20 . We can see that the lagged dependent variable always has a positive and statistically significant coefficient, suggesting persistence. Minority foreign always has a positive and precisely determined coefficient, while the coefficient on Majority foreign is also positive but is only significant for ROS and TFP. Finally, the coefficient on All Foreign is never precisely determined, suggesting that being fully foreign owned is not associated with a statistically significant performance advantage. These findings confirm that jointventures generally perform better than purely domestic and purely foreign owned firms. Yet, the fact that in two out of four cases, the coefficient associated with Majority Foreign is not precisely determined indicates that, as suggested by the descriptive statistics in section 2.2, there may be a level of foreign ownership beyond which corporate performance starts to decline. Hence, foreign ownership is associated with improved firm performance only as long as it is accompanied by a sufficient degree of local investors' participation, without which foreign agents may be unable to perform optimally.
As for the other regressors included in equation (1), Size is positively and significantly associated with ROS and PROD, but negatively associated with TFP.
Leverage is negatively linked with ROA and ROS, suggesting evidence of debt overhang. Collateral has a negative and precisely determined coefficient for ROA and TFP, indicating that a relative prevalence of tangible assets in the firm's capital structure may hamper its performance. Finally, our export dummy (Expdum) is positively associated with TFP. This is in line with the vast literature, which has shown that exporters are typically more productive than non-exporters (see Greenaway and Kneller, 2007, for a survey). Yet, surprisingly, Expdum is negatively associated with PROD. In all cases, the m2 test does not indicate problems with the specification or choice of instruments.
Next, because our ownership measures based on the share of capital may suffer from miscoding problems, we verify the robustness of our results to the use of registration-based firm ownership characteristics. In particular, we replace the Minority foreign, Majority foreign and All foreign dummies in equation (1)  On the other hand, the CJV parties' profit, control, and risks are divided according to negotiated contract terms (Folta, 2005). Focusing on these new definitions of foreign ownership, 15.95% of our firm-years are made up of equity joint-ventures; 2.78%, by contractual joint-ventures; and 17.28%, by wholly foreign owned firms. This leaves 63.98% of domestically owned firm-years, which compares favorably with the corresponding share based on the capital paid in by non foreign agents (60.95%).
Estimates of equation (1), using registration information are reported in columns 5 to 8 of Table 2. In column 5, ROA is used as our measure of corporate performance. Column 6 refers to ROS; column 7, to PROD; and column 8, to TFP.
We can see that the JV dummy always has a positive and strongly significant coefficient for all our measures of corporate performance, while the coefficient on the WFO dummy is generally smaller and only marginally significant in three out of four 21 Unfortunately, using registration-based ownership information does not allow us to differentiate between different degrees of foreign ownership in joint-ventures. cases 22 . In line with Abraham et al. (2007), Yusuf et al. (2006), andPan et al. (1999), these findings indicate that while both joint-ventures and fully foreign owned firms perform better than purely domestic firms, the former have a larger advantage. Our main result that joint-ventures are top performers in China is therefore robust to using registration-based ownership measures. It should be noted, however, that registrationbased measures may be inaccurate, since they are typically updated with significant delay, and firms may have incentives to falsely register as foreign to take advantage of tax benefits (Dollar and Wei, 2007).

4.2
What is the exact nature of the relationship between the degree foreign ownership and corporate performance? Table 3 provides estimates of equation (2) Turning to the other control variables, Statecap generally displays a negative coefficient, statistically significant for ROA and ROS, which suggests the higher the state's participation in a firm's capital, the lower its performance. This is consistent with the findings in many of the studies on privatization surveyed by Estrin et al. (2008). Privatecap, on the other hand, is positively related only with PROD. As in the previous specifications, Leverage and Collateral display either negative or poorly determined coefficients. Size is positively associated with ROS and PROD, but 22 These results were robust to including two separate dummies for equity joint-ventures (EJV) and contractual joint-ventures (CJV).
negatively associated with TFP, and Expdum displays a positive and significant coefficient in the TFP regression. In none of the specifications does the m2 test highlight problems with the specification of the model or the choice of instruments.
As foreign presence typically varies considerably across provinces (see Appendix 1C), we next verify whether our estimates of equation (2)  investors. It has been argued that although investors from HMT may enjoy an advantage based on cultural and geographical proximity to China, they are likely to be fundamentally different from investors from other parts of the world. Specifically, it may be that investment by HMT firms simply represents "round-tripping" by domestic Chinese investors hoping to take advantage of the favorable tax and regulatory treatment received by foreign investors (Huang, 2001). In such cases, one would not necessarily expect firms owned by investors originating from Hong Kong, Macao, and Taiwan to perform better than domestic firms.
Tables A3 and A4 in Appendix 1D report descriptive statistics similar to those presented in Table 1, where observations are partitioned on the basis of the share of total capital paid in by investors originating from countries other than HMT, and by HMT entrepreneurs, respectively. Like Table 1, Table A3 indicates that corporate performance increases with the share of non HMT capital participation, but declines for 100% non HMT owned firms. The pattern in Table A4 is less clear, as very little 23 These dummies also control for factors such as the prevalence of foreign owned firms in a given province and year. difference in corporate performance is observed among firms with HMT participation between 1% and 49.99%, and between 50% and 99.99%. Moreover 100% HMT owned firms exhibit the worst performance (column 4). In summary, ownership by HMT investors does not seem to lead to a clear-cut productivity advantage: this can be seen as indirect evidence in favor of the "round-tripping" hypothesis.
To shed further light on this issue, we estimate a new version of equation (2) where PERF represents in turn one of our four indicators of corporate performance.
The structure of the error term in equation (3) is similar to that in equations (1) and (2). In this framework, PERF is said not to Granger cause Foreigncap if the coefficients on PERF i(t-1) and PERF i(t-2) in equation (3) are not significantly different from 0, i.e. if b 1 = b 2 = 0. The results of the estimates of equation (3), together with the p-values associated with the F-test aimed at testing our null hypothesis are presented in Table 6. We can see that for each of our four measures of corporate performance, the null hypothesis cannot be rejected. We therefore conclude that there is no obvious evidence that corporate performance Granger causes foreign ownership 24 .
Overall, our findings indicate that foreign ownership is only beneficial to the performance of Chinese firms, as long as the degree of foreign ownership does not pass a certain threshold which ranges between 47% and 64%, depending on the measure of corporate performance used. This suggests that a certain degree of domestic ownership is necessary to ensure optimal performance. We next construct a simple theoretical model aimed at underpinning these empirical findings.

Theoretical model
We now construct a simple theoretical model which can generate predictions of a non-monotonic relationship between a firm's degree of foreign ownership and its performance similar to that found in the previous section 25 .

The model:
We assume that the market is made up of I firms labeled with the subscript i (i = 1...I), each of which produces a different product, also labeled with i. 24 These results were robust to adding other control variables to the regressions. 25 Some features of this model resemble the "property rights" approach to the analysis of firm behavior pioneered by Hart and Moore (1990) and Grossman and Hart (1986), where business partners choose their relation-specific investments on the basis of the allocation of the project's joint surplus. Also see Chapter 5.2 in Barba Navaretti and Venables (2006). We thank Arijit Mukherjee for offering inspiring ideas for this section.
Let and denote the quantity of firm i's product, and that of a competitively supplied numeraire good, respectively. Consumers maximize the following utility function: Utility maximization yields the following demand curve: where i p represents the price of good i, and is a demand shifter, which is exogenous to individual firms. Labor is the only factor of production, and production technology exhibits constant returns to scale, with total labor costs given by where w and v i represent respectively the wage (common to all firms) and productivity (firm-specific).
Each firm is potentially a joint-venture between two owners: a foreign (F) and a domestic owner (D). Each may contribute a non-contractible input (such as effort) to affect the productivity of the joint-venture (v i ). Productivity can hence be expressed as follows (for simplicity, we hereafter suppress the firm subscript i): x and y represent the non-contractible inputs supplied respectively by F and D. x may be interpreted as the foreign owner's effort aimed at improving the quality of the design of the product, or any other form of "knowledge capital". y could be seen as the domestic owner's effort aimed at promoting the sale and marketing of the product in the local market (e.g. through the organization and monitoring of a sales team), or at facilitating the political connection (Guanxi) with local governments, which is often considered as a key factor in determining firms' performance in China (Hsieh and Klenow, 2007). Both inputs are assumed to be greater than or equal to 0, (i.e. x≥0 and y≥0) 26 . Furthermore, equation (6) implies decreasing returns to scale of x and y (i.e. v xx <0, v yy <0) 27 . 26 This non-negativity assumption means that neither party can contribute negative inputs to incur "damage" to the productivity of the joint-venture, by making v fall below its benchmark level given by = . ) The parameter ) (β α captures the "relative importance" of the contribution of to productivity. The relative marginal return to x is in fact given by: , which is increasing in α . The greater α , the greater the contribution of x to firm productivity. In the extreme case in which 0 = α , x is completely "unimportant" to v, which is purely determined by y.

The parameter
inversely captures the "absolute indispensability" of x (y): the lower a (b), the greater the marginal return to x (y). In the extreme case in which (b=0), v=0 if x=0 (y=0), indicating that x (y) is completely "indispensable", since x(y) must be strictly positive in order to achieve a positive productivity.
We further assume that there are implicit costs borne by each party to increase their inputs. These can be expressed as: Note that these implicit costs are assumed not to be incurred by the joint-venture, but only by the individual owners. For example, the domestic party may need to devote some of his/her own time to monitor the local sales team, incurring a disutility.
Finally, the ownership shares of the firm allocated to parties F and D are respectively S and 1-S. S is assumed to be exogenous: it could partly depend on the bargaining power of the foreign party and on other factors beyond the control of individual investors (such as government interventions).

5.2
Nash equilibrium: Each firm's profit is given by ( ) v pq c π = − . Profit maximization leads to the following optimal price: ( 1) , which yields the following optimal profit: Parties F and D receive respectively shares S and 1-S of the total profits. Each simultaneously chooses the inputs that it contributes to maximize his/her individual net payoff, taking the other party's inputs as given. The two parties' maximization problems can therefore be expressed as follows: (9) F: (9) and (10) imply a standard Nash equilibrium problem, whereby each party strategically chooses its input, depending on the choice made by the other party.
For the purpose of the presentation, we will hereafter focus on the simple case in which the cost functions of x and y are symmetric ( r t = ) and linear (k=1) 28 . In this case, the equilibrium is determined by: where . To ensure that the second order conditions for each party's optimization problem hold, we assume that (1 )( 1) 1 0 z 28 Although our qualitative results were largely unaffected in more general cases, the example allows us to derive simple closed form solutions to the problem. 29 The second order conditions associated with (11) and (12)  his/her share in total profit (S) increases, party F tends to provide more input x, as this raises the reward to his/her input. We call this the "share effect". On the other hand, an increasing share S means a decreasing share of party D in total profits (1-S), leading to a decreasing input from D, which in turn tends to reduce productivity and hence total profit. This decreases the marginal return to F's inputs, and hence induces F to provide less input. We call this the "strategic effect". The overall effect turns out to depend on the value of S: when S is low (high), the "share effect" ("strategic effect") dominates, leading to a positive (negative) relationship between S and x.
Equations (13) and (14) imply that x and y could possibly be negative. Yet, we assumed that neither x nor y can take a value below zero (non-negativity assumption).
It follows that: Note that x ( ) is monotonically increasing in S(1-S). Using the non-negativity assumption and combining equations 2 2 y (13) to (16), the equilibrium levels of the inputs and contributed respectively by F and D are given by :

Relationship between ownership and productivity:
We now investigate the central issue of the model, i.e. its predictions regarding the relationship between the share of foreign ownership (S) and firm productivity (v*) 31 . Substituting (17) and (18) into (6), we obtain the equilibrium value of productivity as a function of S, i.e.: What conclusions can we draw from the above analysis about the effects of the degree of foreign ownership on owners' inputs and corporate productivity? As can be seen from equations (17)- (19), depending on the model parameters, x*, y*, and v* could either be linked to S by an inverted U-shaped relationship, or be monotonically increasing or decreasing in S, or be independent of S. This implies that, in general, the impact of S on x*, y*, and v* is ambiguous.
However, under reasonable parameter ranges, the model yields clear-cut predictions on the relationship between foreign ownership (S) and firm productivity (v*) that broadly fit the empirical results obtained in section 4. Figures 2a and 2b, which plot the values of x*, y*, and v* against S, illustrate a situation of this type 32 .
The following important features emerge. First, the best performing jointventure is more productive than the wholly foreign owned company, which in turn outperforms the purely domestic company. This can be expressed as follows: Note that according to (19) and (13)-(14), As shown in Appendix 2A, it follows that necessary and sufficient conditions for (20) are that both a and b are sufficiently small, and that a is sufficiently small relative to b (given , and Q α β ). In word, the above pattern of productivity ordering requires that both the foreign and domestic party's inputs are crucial to the joint-venture's productivity performance, with the foreign party's inputs being relatively more important. This is consistent with the fact that, in developing countries, the foreign parties of joint-ventures usually provide core technology and design of the products, which are often regarded as "key inputs" to production and, hence, to productivity.
Second, for a joint-venture with an intermediate ownership share S (between 5% and 95% in our example), there exists an inverse U-shaped relationship between the share of foreign ownership and productivity, with maximum productivity level at α = S . As shown in Appendix 2B, this feature requires that v*(S) is dominated by the inverse U-shaped function , which in turn requires that both parameters a and b are sufficiently small. To understand the intuition behind this inverse U-shaped relationship, it is important to note that the firm's productivity depends on the joint inputs from both parties. As can be seen from figures 2a and 2b, when S is low, the input contributed by party F is low, leading to relatively low productivity. Similarly, when S is high, party D's contribution is low, which once again leads to relatively low productivity. This is because, as shown in section 5.1, the contributions of inputs x and y to overall firm productivity are characterized by decreasing return to scales.
Hence, a high S (1-S) leads to "undersupply" of input y(x), as the marginal return to this input is relatively high, and to "oversupply" of input x (y), whose marginal return is low. It is only when S is at an intermediate level, which depends on the relative marginal returns of the two inputs, and equals the foreign owner's share in the productivity function ( 1 ( ) v S α ) that the joint inputs contributed by both parties lead to the highest productivity of the joint-venture. This result mirrors a well-known conclusion from the "property rights" theory of the firm, according to which the optimal allocation of property rights should assign more assets to the party whose investment has greater impact on production (Grossman andHart 1986, Hart andMoore 1990).
Third, when foreign ownership (S) is very low (high), firm productivity (v*) is monotonically decreasing (increasing) in S. This is because, as can be seen from figure 2a, when S is very low (<5% in this example), party F will always provide zero inputs, meaning that a marginal increase in S will decrease party D's share in total profits (1-S). This will reduce party D's incentive to contribute to production (y*), leading to a lower level of total productivity. Reasoning analogously, when S is very high, party D will always provide zero inputs, whilst party F's incentive to contribute to production will increases in S. Consequently, productivity will increase in S.
To conclude, although the above theoretical predictions on the relationship between ownership and performance depend on model parameters, and in particular on the degree of "indispensability" of both parties' inputs (a, b), the key merit of this theoretical model is to provide a rationale for why there could exist a productivity ordering between joint-ventures, wholly foreign owned firms, and purely domestic firms, and an inverse U-shaped relationship between the degree of foreign ownership and corporate performance, similar to those empirically revealed in sections 2 and 4.
This rationale hinges on the strategic interactions between the owners' noncontractible inputs, and the optimal allocation of ownership according to the relative importance of each owner's inputs.

Conclusions
We have used a panel of 21,582 unlisted Chinese firms over the period 2000-2005 to analyze the relationship between the degree of foreign ownership and corporate performance. Focusing on the return on assets, the return on sales, labor productivity, and TFP, we have found that joint-ventures generally perform better than purely foreign owned and purely domestic firms. This finding is robust to defining jointventures on the basis of the capital paid in by various foreign agents, and on the basis of registration information. It can be explained considering that both the domestic and foreign parties of a joint-venture bring in attributes essential to achieving high performance. Specifically, the former contribute knowledge of the Chinese market and legal environment, as well as important political connections with local governments; and the latter, modern technologies, capital, better corporate governance through monitoring and market discipline, and managerial and international networking skills.
We have then shown that foreign ownership and the performance of our

Ownership variables
Minority foreign: dummy variable (DV) equal to 1 if the share of the firm's total capital owned by foreign investors is positive but lower than 50%, and 0 otherwise.
Majority foreign: DV equal to 1 if the share of the firm's total capital owned by foreign investors is greater than or equal to 50% but lower than 100%, and 0 otherwise.
All foreign: DV equal to 1 if the share of the firm's total capital owned by foreign investors is equal to 100%, and 0 otherwise.

Corporate performance variables
Return on Assets (ROA): ratio of the firm's net income to total assets.

Return on Sales (ROS): ratio of the firm's net income to total sales.
PROD: labor productivity, calculated as the ratio of the firm's net income to total number of employees.
TFP: total factor productivity calculated using the Levinsohn and Petrin (2003) method, applied separately to different industrial groups.

Other variables
Total assets: sum of the firm's fixed and current assets, where the former include tangible, intangible, and other fixed assets; and the latter include inventories, accounts receivable, and other current assets.
Size: logarithm of the firm's total assets.
Leverage: ratio of current liabilities plus non-current liabilities to total assets, where current liabilities include loans, accounts payable, and other current liabilities; and non-current liabilities include long-term debt and other non-current liabilities.
Collateral: ratio of tangible assets to total assets.
Employees: total number of people employed by the firm.
Expdum: dummy equal to 1 if the firm reports a positive value of firm's overseas sales, and 0 otherwise.
Deflators: all variables are deflated using provincial GDP deflators, taken from various issues of the China Statistical Yearbook. geographical concentration of foreign owned firms may have its root in the FDI promotion policies adopted in the past, and in the inadequate infrastructures characterizing the Central and Western regions (Dougherty and McGuckin, 2002).

C. FDI patterns by region and industry
Focusing on specific provinces, Guandong contains the lowest percentage of domestically owned firm-years (25.74%), while Qinghai has the highest (98.61%).
Guandong also contains the highest share of wholly foreign owned firms (52.73%), while the highest share of partially foreign owned firms is in Shanghai (34.79%). In sum, Table A1 indicates that although all Chinese provinces exhibit some degree of foreign ownership, there is a substantial heterogeneity in the degree of foreign presence across provinces. Table A2 presents a similar analysis for 15 industrial groups 34 . 87.90% of our sample is made up of manufacturing firm-years. Yet, because foreign ownership also characterizes other sectors in the economy, we include some of these in our analysis.
The Table shows that except for "Mining" and "Transportation; Communication, Electric, Gas, and Sanitary Services", all sectors are characterized by a percentage of foreign owned firm-years in excess of 20% 35 . There is also considerable heterogeneity in the degree of foreign ownership across sectors. In particular, "Other Manufacturing" contains the highest share of wholly foreign owned firm years (45.95%) 36 . The highest percentage of partially foreign owned firm-years is in the wholesale and retail sector (30.84%). The Table also shows that foreign firms have entered both labor-intensive industries such as "Textiles, Clothing, and Leather", and capital-intensive industries such as "Chemicals, Petroleum, and Man Made Fibres", and "Electrical, Machinery and Computer Equipment".
34 These groups are based on two-digit SIC groups for the manufacturing sector, and on the broader SIC divisions for the non-manufacturing sector for which fewer observations are available. 35 Note that 96.74% of the observations in the broad sector "Transportation, Communication, Electric, Gas, and Sanitary Services" comes from "Electric, Gas, and Sanitary Services". This broad sector is largely controlled by the State. 36 "Other Manufacturing" contains, among others: measuring, analyzing, and controlling instruments; photographic, medical and optical goods; watches and clocks; jewellery, silverware, and plated ware; musical instruments; dolls, toys, games, sporting and athletic goods; pens, pencils and other artists materials.

D. Descriptive statistics differentiating foreign owners into those origination
from Hong Kong, Macao, and Taiwan, and those originating from other parts of the world Table A3 presents descriptive statistics similar to those presented in Table 1 for investors originating from Hong Kong, Macao, and Taiwan. Table A4 presents similar statistics from investors originating from other parts of the world.

Appendix 2: Proofs
A. Necessary and sufficient conditions for the productivity ordering , which is increasing in a; and that [ ] requires the following necessary and sufficient conditions to be satisfied: (A.1) implies that for given values of α, β, and Q, both a and b must be sufficiently small. Moreover, a must be sufficiently small relative to b, for the inequality chain to hold.

B. Necessary and sufficient conditions for the quasi inverse-U shaped ownership-productivity relationship
Recall that is an inverse U-shaped function of S , while ν   (14), these imply: . This implies that 1 y Δ increases as b declines.
Hence, Δ 1 = 1 x Δ ∩ will be larger, the smaller a and b . In other words, the more "indispensable" the inputs 1 y Δ x and y are to the production process, the more the relationship between S and v* will be dominated by the inverse U-shaped function v1.
In the extreme case in which 0 This represents a purely inverse U-shaped relationship between v* and S. Eq. 12  Notes: Foreigncap represents the fraction of the firm's capital paid in by foreign investors. ROA represents the firm's returns to assets and is given by its net income over its total assets. ROS represents the firm's returns to sales and is given by its net income over its total sales. PROD represents labor productivity, i.e. the ratio of the firm's net income to its number of employees. TFP is total factor productivity calculated using the Levinsohn and Petrin (2003) method. Size is the logarithm of the firm's total assets. Leverage is given by the sum of the firm's current and non-current liabilities to its total assets. Collateral is given by the ratio of the firm's fixed tangible assets to its total assets. Expdum is a dummy equal to 1 if the firm exports, and 0 otherwise. Statecap represents the fraction of the firm's capital paid in by the state. Privatecap represents the fraction of the firm's capital paid in by individual investors and legal entities. See Appendix 1B for complete definitions of all variables. Notes: Minority foreign is a dummy equal to 1 if the share of the firm's total capital owned by foreign investors is positive but lower than 50%, and 0 otherwise. Majority foreign is a dummy equal to 1 if the same share is greater than or equal to 50% but lower than 100%, and 0 otherwise. All foreign is a dummy variable equal to 1 if the share of the firm's total capital owned by foreign investors is equal to 100%, and 0 otherwise. JV is a dummy variable equal to 1 if the firm is registered as an equity jointventure or as a contractual joint-venture, and 0 otherwise. WFO is a dummy variable equal to 1 if the firm is registered as wholly foreign owned, and 0 otherwise. All specifications were estimated using a GMM first-difference estimator. The figures reported in parentheses are asymptotic t-statistics. Time dummies and time dummies interacted with industry dummies were included in all specifications. Standard errors and test statistics are asymptotically robust to heteroskedasticity. The instrument set includes two or more lags of all explanatory variables, time dummies, and time dummies interacted with industry dummies. m2 is a test for secondorder serial correlation in the first-differenced residuals, asymptotically distributed as N(0,1) under the null of no serial correlation. Also see Notes to Table 1. * indicates significance at the 10% level. ** indicates significance at the 5% level. *** indicates significance at the 1% level. Notes: All specifications were estimated using a GMM first-difference estimator. The figures reported in parentheses are asymptotic t-statistics. Time dummies and time dummies interacted with industry dummies were included in all specifications. Standard errors and test statistics are asymptotically robust to heteroskedasticity. The instrument set includes two or more lags of all explanatory variables, time dummies, and time dummies interacted with industry dummies. m2 is a test for second-order serial correlation in the first-differenced residuals, asymptotically distributed as N(0,1) under the null of no serial correlation. Also see Notes to Table 1. * indicates significance at the 10% level. ** indicates significance at the 5% level. *** indicates significance at the 1% level. Notes: All specifications were estimated using a GMM first-difference estimator. The figures reported in parentheses are asymptotic t-statistics. Time dummies and time dummies interacted with provincial dummies were included in all specifications. Standard errors and test statistics are asymptotically robust to heteroskedasticity. The instrument set includes two or more lags of all explanatory variables, time dummies, and time dummies interacted with provincial dummies. m2 is a test for second-order serial correlation in the first-differenced residuals, asymptotically distributed as N(0,1) under the null of no serial correlation. Also see Notes to Table 1. * indicates significance at the 10% level. ** indicates significance at the 5% level. *** indicates significance at the 1% level.  Notes: PERF indicates in turn our 4 performance indicators. All specifications were estimated using a GMM first-difference estimator. The figures reported in parentheses are asymptotic t-statistics. Time dummies and time dummies interacted with industry dummies were included in all specifications. Standard errors and test statistics are asymptotically robust to heteroskedasticity. The instrument set includes two or more lags of all explanatory variables, time dummies, and time dummies interacted with industry dummies. The Granger causality test is an F-test aimed at testing the null hypothesis that the coefficients on the two lags of our performance indicators are jointly equal to 0. The m2 is a test for second-order serial correlation in the first-differenced residuals, asymptotically distributed as N(0,1) under the null of no serial correlation. This test is not reported in column 4 due to the lack of a sufficient number of time series observations. Also see Notes to Table 1. * indicates significance at the 10% level. ** indicates significance at the 5% level. *** indicates significance at the 1% level.