Does Foreign Direct Investment Improve the Productivity of Domestic Firms? Technology Spillovers, Industry Linkages, and Firm Capabilities

Transcription

1 Does Foreign Direct Investment Improve the Productivity of Domestic Firms? Technology Spillovers, Industry Linkages, and Firm Capabilities Feng Helen Liang Haas School of Business University of California, Berkeley This version: February 3, 2008 Abstract Many developing countries attract foreign investment in an attempt to improve the productivity of domestic industries. However, it is unclear whether local firms learn from foreign direct investment, and if so, which local firms benefit, what forms of foreign direct investment are most beneficial, and why these effects occur. This paper explores how industrial linkages, firm capabilities, and the geographic location of domestic firms affect the diffusion of technology brought by foreign direct investment. I hypothesize that local firms are more likely to improve efficiency when they receive better product inputs from foreign suppliers and technology support by foreign customers, and such transfer of knowledge is more effective when the recipient has high absorptive capacity and is located near the source of knowledge. I analyze plant-level data in China for over 20,000 plants between 1998 and I find positive productivity spillovers between foreign suppliers and their domestic customers. However, there are not positive spillovers from foreign-owned customers or competitors. Domestic firms inhouse R&D capital facilitates learn from foreign firms. Local firms learn from both joint ventures and wholly-owned foreign subsidiaries and the effects are larger from wholly-owned subsidiaries. Keywords: Foreign Direct Investment, Technology Spillovers, Productivity, China fenliang@haas.berkeley.edu. I am grateful to Ping He, Jinchang Qian, and Weining Yu at the National Bureau of Statistics of China for their assistance in obtaining and understanding the data. I received helpful advice from Henry Chesbrough, Paul Gertler, Bronwyn Hall, Canfei He, David Levine, David Mowery, Yingyi Qian, Catherine Wolfram, and participants in Innovation Seminar in UC Berkeley, ITEC Beijing Forum, and Academy of Management. I gratefully thank David Levine, Henry Chesbrough, and Alfred P. Sloan Foundation for their generous financial support of the project. 1

2 I. Introduction Policymakers in many developing countries spend considerably in an effort to attract foreign direct investment, by giving costly tax holidays, building infrastructure, and issuing regulatory exemptions. These policymakers expect foreign-invested facilities to bring new technology, capital, and management expertise and therefore improve the productivity of domestic industries. However, evidence remains limited whether and how domestic firms actually benefit from technology spillovers from the inflow of foreign investment. For example, Aitken and Harrison s (1999) study on Venezuela finds foreign-invested joint ventures actually have negative effects on the productivity of domestic firms in the same industrial sector. Several recent studies point out that positive productivity spillovers are more likely to happen between vertically linked industries, rather than within the same industry sector. This is because multinational firms have an incentive to prevent knowledge leakage to competitors, but may transfer technology to local suppliers to get higher quality inputs at lower prices. Local firms could also improve efficiency for the same reasons when dealing with multinational suppliers. Javorcik (2004) finds positive spillovers from foreign-invested joint ventures to domestic firms in Lithuania in upstream industries, but not in horizontal or downstream industries. Blalock and Gertler (2005a) find evidence of productivity gains and lower market prices among Indonesian firms supplying industrial sectors with a large foreign presence. Other factors beyond industrial linkages may affect the effectiveness of technology spillover. These factors include firm capabilities, the geographic locations of the source and the recipient of knowledge, and the ownership structure of foreign-owned firms, among others. Blalock and 2

3 Gertler (2005b) find domestic firms with previous R&D investment and highly educated work forces are more likely to adopt the technology brought by foreign firms. Jaffe et al (1993) use patent citation data as a measure of spillovers and show the spillover effect is more evident at the local level. Keller (2002) finds the productivity benefit from R&D expenditures decreases with geographic distance between technology sender and recipient. Javorcik (2004) suggests that joint ventures are more likely than wholly-owned foreign subsidiaries to source locally and thus transfer technology to local suppliers. This study builds on the above research and examines whether the productivity of domestic firms is associated with the presence of foreign firms in the downstream, upstream, and horizontal industrial sectors, and whether this relationship is affected by firm capabilities, geographic distance and types of foreign ownership. Understanding the channels and moderating factors of firms adoption of foreign technology has important policy implications, as the current empirical evidence on the welfare effects of foreign investment is mixed. A clearer understanding of the exact mechanisms of learning via foreign investment will allow policymakers to better target appropriate forms of FDI. I hypothesize that local firms are more likely to improve efficiency when they purchase higher quality inputs from foreign suppliers or when they receive technology support from foreign customers. I also hold that this transfer of know-how is more effective in improving productivity when the local firm has higher absorptive capacity, when the recipient is close to the source of knowledge, and when the multinational firm is a joint venture rather than a wholly-owned subsidiary. 3

4 This study contributes to the existing literature by exploring how the heterogeneity of firms and other microeconomic factors influence the spillover effects between local firms and foreign firms via industrial linkages. I examine these questions using detailed firm-level data on production, finances, and R&D activities. The data is from national enterprise surveys and science and technology surveys conducted by the National Bureau of Statistics of China which covers all large- and medium-sized firms in China. These are the most comprehensive datasets covering Chinese firms and have been widely used in other research. For example, Hu and Jefferson (2002) study the impact of foreign investment on domestic firms productivity in the textile and electronic industries. Hu, Jefferson, and Qian (2005) examine technology transfer and the R&D activities of Chinese firms and how foreign investment influences the technology activities and domestic firms productivity. Chang, Chung, and Xu (forthcoming) study the spillover effects from foreign investment from related industries in China and how the ownership types of recipients influence the adoption of foreign technology. The panel structure of the data set and the rich information on firm production and R&D activities provide several advantages for this study. First, panel data allows within firm estimation to control for unobserved firm-level variations and better identification of efficiency improvement than cross-sectional data. Second, detailed information on firms R&D activities allows analysis on the absorptive capacity of firms. Third, China itself provides a salient context for this research. The country has received the largest amount of foreign direct investment among developing countries, and the investment is distributed unevenly across regions due to the country s semi-landlocked geographic feature and vast size. In addition, the 4

5 privatization process in the country has introduced different incentives to domestic firms and led to fierce competition and a shake-out in many domestic sectors. The study finds strong evidence of positive spillovers from foreign suppliers in the upstream sectors to local firms. A one-standard-deviation increase in foreign presence in the upstream sector, or an increase of 6.4 percentage points in the supplying FDI output share, is associated with a rise of 1.2 percent in the output of domestic firms in the supplied sector. But there is no evidence of positive spillovers from horizontal or downstream sectors. In other words, the productivity of Chinese firms is positively correlated with the presence of multinational suppliers, but not with foreign customers and competitors. This result is contrary to the findings of previous studies on technology spillovers set in other developing countries that find positive spillovers from foreign customers (Blalock and Gertler 2005a; Javorcik 2004), but is consistent with the findings on China s manufacturing firms in recent years (Chang, Chung, and Xu 2007). Firm capabilities, measured as R&D capital stock, are found to be positively associated with spillovers from foreign suppliers, but have little effect on spillovers from horizontal or downstream sectors. Both negative and positive spillover effects on productivity are significantly larger when the source of the spillover is wholly owned foreign subsidiaries rather than joint ventures, contrary to the findings in previous research (Javorcik 2004). Geographic distance does not have significant effect on learning from multinationals. These findings are stable in a number of robustness checks and after controlling for simultaneity issues (Olley and Pakes 1996; Blalock and Gertler 2004; Chang, Chung, and Xu 2007). 5

6 The paper is structured as follows: Section II gives an overview of the literature on the channels and moderating factors of technology spillovers. Section III discusses FDI inflows into China, the data, and the estimation strategy. Section IV presents results and Section V concludes. II. Literature Review on the Channels of Spillover and Moderating Factors Multinational firms investing in manufacturing facilities in foreign countries are believed to possess advantages that enable them to compete with better informed domestic firms. These advantages include intangible productive assets, such as technology know-how, management skills, reputation, etc. Since these assets are gained through experience in operation or related to tacit knowledge, they cannot be easily replicated by local firms, but can be transferred via several channels and increase the productivity of a local firm. These channels include: 1) foreign customers may directly transfer knowledge to local suppliers; 2) foreign customers may have higher quality requirement and better supply-chain management skills that prompt domestic firms to improve production technology and management; 3) local firms may observe and imitate the technology and management practice of the multinationals; 4) local firms may recruit employees trained by multinationals or benefit from interaction with the personnel in multinationals; 5) local firms may benefit from the externalities brought by multinationals, such as higher quality input from the upstream foreign suppliers, larger demand from foreign customers and the economy of scale, better infrastructure subsidized by the government to attract foreign investment, etc. 6

7 2.1 Industrial Linkage The spillover effects on productivity from these channels depend crucially on industrial linkage. As previous studies suggest, there is little evidence of horizontal spillovers from foreign invested firms to domestic firms in developing countries (Aitken and Harrison 1997; Blalock and Gertler 2005a; Javorcik 2004), although such horizontal spillover is found in developed countries such as the United States (Keller and Yeaple 2003; Chung, Mitchell, and Yeung 2003) and United Kingdom (Haskel 2002; Liu et al. 2000). In fact it is not surprising that horizontal spillovers are limited in developing countries. Domestic firms and foreign firms might produce for different markets: domestic firms supply local market with low demand on quality and product specifications, while multinationals produce for the international market with higher demand on product quality, as Chesbrough and Liang (2007) find in China s semiconductor industry. Multinationals might also implement measures to prevent knowledge leakage to local competitors, such as paying higher wages to employees to prevent domestic firms from bidding them away. In addition, domestic firms may have limited absorptive capacity to recognize and adopt the new technology or management skills from the multinationals (Blalock and Gertler 2005b; Cohen and Levinthal 1990). These factors prevent domestic firms from reaping the benefit of technology spillovers through the channels of personnel turnover and imitation. In addition, the competition from multinational might generate negative externalities on domestic firms, particularly in the short run, when the multinationals bid away high quality labor, reduce the market share of domestic firms if the multinationals supply the host country market and force domestic firms to cut production and incur higher unit costs (Aitken and Harrison 1999). 7

8 On the other hand, the positive spillovers are more likely to happen between vertically related industries, either through backward linkage when foreign customers transfer technology to local suppliers or through forward linkage when domestic firms get higher-quality inputs or equipment from foreign suppliers. Because multinationals in China are more export oriented than domestic firms (as shown in Figure A1) the percentage of export of total output is 42% among foreign affiliates, and 10% among domestic affiliates during , spillovers may happen through backward linkage more than forward linkage. Why would multinationals transfer technology through supplier chain? While multinationals try to prevent knowledge leakage to local competitors, they have incentives to support their suppliers through training in technology and management, so as to improve the quality of supply and reduce cost. To avoid being held up by a single supplier, the multinational may establish such relationship with multiple suppliers, and encourage the spread of the knowledge in the supplying sector (Blalock and Gertler 2005a). Although more productive suppliers and lower supply prices could also benefit the multinational s rivals and encourage entrance into the downstream sector, previous research has shown that the benefits of a competitive supply base to the multinational buyer outweigh the lost to free-riding entrants when the new competition is not too great (Pack and Saggi 2001). Local firms may also benefit from foreign suppliers. Foreign-owned firms in the upstream industry supplying local market may produce components and equipment of higher quality, and may provide domestic customers with technology support when they purchase the equipment. 8

9 Thus, domestic firms in the downstream sectors may improve productivity when there is foreign participation in the supplying sector. Empirical and anecdotal evidence confirms positive spillovers through backward and forward linkages in developing economies. Javorcik (2004) suggests multinational customers help improve local supplier s quality control system in Czech Republic after signing the contract, and the Czech supplier applied these improvements to its other production lines not related to this particular foreign customer. In the empirical analysis employing Lithuanian firm data, Javorcik finds positive and significant spillover effects through backward industrial linkage: a one-standard-deviation increase in the foreign presence in the sourcing sectors is associated with a 15-percent rise in output of each domestic firm in the supplying sector. Similarly, Blalock and Gertler (2005a) find positive spillovers through backward linkage in Indonesia firms supplying foreign customers. Both studies find little evidence of spillover within the same industry sector and through forward linkage (from suppliers to buyers). Chang, Chung, and Xu (2007) find positive spillovers from foreign suppliers to Chinese manufacturing firms. Based on the previous literature, I expect the spillover effects from upstream and downstream foreign presence to be positive, and the effects from horizontal foreign presence to be ambiguous. Hypothesis 1.1. Firms receive either positive or negative spillovers from foreign investment in the same industry sector; 9

10 Hypothesis 1.2. Firms receive positive productivity spillovers from foreign investment in upstream and downstream industry sectors. 2.2 Firm Capability Firms absorptive capacity, the capability to recognize and adopt the new technology or management skills from others, might also impact whether they benefit from spillovers (Cohen & Levinthal 1990). Cohen and Levinthal argue that a firm can build absorptive capacity through activities requiring related knowledge. Thus absorptive capacity can be measured with firms previous experience and investment in research and development and human capital. Previous studies on Indonesia manufacturing firms find that a firm benefits more from foreign technology when it has invested in-house R&D or has more highly educated workers (Blalock & Gertler 2005b). In a dynamic industry environment characterized with rapid technology changes, higher absorptive capacity is believed to lead to better adaptation to changing technology environment and exploitation of the opportunities (Zahra & George 2002, Todorova & Durisin, 2007). Thus, firms with higher existing knowledge capital are expected to benefit more from the external knowledge. Hypothesis 2. Firms with larger in-house R&D capital are more likely to receive positive spillovers from foreign presence. 2.3 Foreign Ownership: Joint Ventures and Wholly-owned Subsidiaries The form of foreign ownership may influence the tendency and effects of local sourcing. Multinationals may establish affiliates in a host country through merger and acquisition or 10

11 green field establishment, joint ventures with local firms or wholly-owned subsidiaries. It is argued that multinationals may choose joint venture as the entry mode in order to utilize domestic partner s knowledge of local market, distribution channel, and supply chain. Empirical studies find evidence to support this view for Japanese multinationals (Belderbos et al. 2001) and for Swedish affiliates in Eastern Europe (UNCTC 2000). Therefore, joint ventures are expected to source more locally, while wholly-owned foreign affiliates rely more on imported inputs. Javorcik (2004) shows that among the foreign affiliates operating in Latvia, over 50% of joint ventures purchase their intermediate inputs locally, and only 9% of whollyowned foreign subsidiaries do so. The same rationale applies to foreign suppliers. When multinational firms set up factories in the host country, they are more likely to choose joint venture mode if the goal is to supply local market instead of exporting. As a result joint ventures may have more interaction with domestic customers. If supply chain is a major channel of technology spillover, then it is expected that positive spillover effects are related with joint ownership more than with wholly-owned foreign subsidiaries. Hypothesis 3. The spillover effects from foreign investment in vertically related industries are larger from joint ventures than from wholly-owned foreign subsidiaries. 2.4 Geographic Location and Spillover Effects The extent to which domestic firms benefit from technology spillovers also depends on the geographic distance from the source of knowledge. When the source of knowledge is located nearby, it is less costly for the recipient of knowledge to imitate or communicate with the 11

12 source, to visit factory floors, and to get training for its personnel. When the spillovers are through employee turnover and social interaction, geographic proximity is also crucial, since recruiting and communication are more likely to be local. Using patent citation data as a measure of spillover, Jaffe, Trajtenberg, and Henderson (1993) show that knowledge diffusions are geographically localized. Specifically, patent citations are more likely to come from the same state and standard metropolitan statistical area (SMSA) as the cited patents, and the effects are particularly significant at the local (SMSA) level. Separately, in a cross-country study, Keller (2002) finds that the productivity benefit from R&D expenditure decreases with geographic distance between technology sender and recipient country. Mowery and Ziedonis (2001) find that knowledge flows from university to industry firms through licensing is more geographically localized due to concerns with incomplete contract and high transaction cost. To examine the moderating effects of geographic distance on spillovers is especially salient in the context of China s economy. The distribution of industrial sectors and foreign direct investment is very uneven across the country. As the country is half land-locked, most of the foreign entrance has occurred in the eastern/coastal region, while the western/inland region remains lagged behind in attracting foreign investment. This variation across regions enables a test on the effect of geographic distance on the spillovers 1. 1 While geographic location can be considered an exogenous determinant of the distribution of foreign investment, other endogenous factors related to local government behavior could confound the effect. For example, the local protectionism at provincial and city level may result in high cost to buy inputs from a distant supplier. In a study using province level data on 32 industrial sectors in China from 1985 to 1997, Bai and coauthors (2004) find that local governments tend to protect industries that yielded high profit and tax, thus reducing the geographic concentration in those industries, and that local protectionism is significant for industries with large shares of state ownership. Using industrial sector data at province level in China, Amiti and Javorcik (2003) find foreign investors are less likely to invest in provinces where protectionism is high, measured by the relative size of state-owned sectors. They also find local-based supplier network and good infrastructure attract foreign investment. These effects lead to the concern that foreign presence is determined by the local supplies productivity, rather 12

13 Hypothesis 4. Firms receive larger spillover effects from nearby foreign investment than from foreign investment in the distance. III. Data, Measurement, and Empirical Strategy 3.1 Foreign Direct Investment in China China had been virtually closed to foreign investment until 1978, when the party leader Deng Xiaoping implemented economic reforms (Qian 2000). These reforming measures were phased in gradually, and the central government imposed different trade policies in different regions. In the first stage, China established four special economic zones (SEZs) in 1980 and designated 14 coastal cities open to overseas investment in Foreign firms investing in these places receive tax breaks and other regulatory privileges. The policy led to large amount of FDI inflows and boosted economic development in these cities and the surrounding areas. In the following years, the government granted favorable policies to other regions and more autonomy in policy making to local government. Local governments across the country established hundreds of industrial parks in an effort to attract investment and increase local employment and tax income. By 2004, China has become the largest recipient of foreign direct investment among developing countries, receiving US$61 billion in foreign direct investment, among which 71% was invested in manufacturing sectors 2 (Figure 1). Among manufacturing than predict it, or, foreign presence may be determined by other unobserved variables that determine local firms productivity and the attractiveness to foreign capital simultaneously, such as the efficiency of municipal government. Assuming these factors are persistent over time, this selection of the location can be controlled with sector and region fixed effects. More details will be discussed in the empirical section that follows. 2 Source: China Statistics Yearbooks 2005, 13

14 sectors, foreign investments are concentrated in electronics, transportation equipment, rubber products, and paper products, etc, as measured by the share of capital and output (Table A2). In addition to trade policy, China s geographic features also lead to uneven distribution of foreign investment 3. China is landlocked on the northwest side and surrounded by mountains, deserts, and foreign countries that barely participate in international trade, most of the trade activities and foreign investment come from the coastal areas on the east and southeast side. A city s distance to the coast and several major trade hubs on the coast significantly influences its access to foreign investment, introducing a source of exogenous variation in the access to foreign direct investment. 3.2 Data Description The data used in this study is based on the establishment level data from Large and Medium Enterprise Survey and Science and Technology Activity Survey conducted by the China National Bureau of Statistics (NBS). Both surveys are of census type and cover all the stateowned enterprises and non-state-owned enterprises with sales above 5 million Yuan (about US$ 600,000 according to the exchange rate in 2006). These enterprises account for 25% of all the registered enterprises and 90% of sales 4. The original data set includes about 20,000 establishments each year from 1998 to NBS removed industrial sectors related to national defense and some precious metal industries. I further excluded agricultural and services sectors. 3 See Figure A2 for Geographic distribution of foreign direct investment in Jefferson et. al provides a comprehensive review of the enterprise survey data collected by National Bureau of Statistics. 14

15 For each establishment, the variables include production/financial indicators such as total output/sales, capital, employment, wage bill, intermediate input, investment, ownership structure, export, profit, etc, and research and development activity indicators such as R&D expenditure, funding sources, personnel, patenting activity, etc. These variables are deflated to starting year (1998) price in the analysis. I use 2002 Input/Output table at national level to construct the industrial linkage variables, and assume that the technology is constant throughout the period 5. The I/O table includes 35 manufacturing and mining sectors, which I concord with the industrial survey data. Thus, the industrial linkage variables are defined at 35 2-digit industrial sector levels. Table 1 provides summary statistics of all the variables. 3.3 Empirical Strategy and Measurement Productivity Estimation Function To examine the correlation between productivity and FDI in the same industry or other industries, I use an approach similar to previous literature and estimate the following equation: (1) ln Yijrt = ß0 + ß1 ln Cijrt + ß2 ln Lijrt+ ß3 ln Mijrt + ß4 ln Kijrt + ß5 FDIjrt+ ß6 Capability ijrt*fdi jrt + γ Xijrt + αi + αt + ξijrt where Xijrt is a vector of firm level control variables, and FDIjrt is a vector of horizontal, downstream, and upstream foreign presence, FDIjrt = ( Horizontal_FDIjrt, Downstream_FDI jrt, Upstream_FDI jrt ) 5 Ideally I would use yearly I/O table but such information is not available. An alternative is to interpolate the yearly I/O ratios from 1997 and 2002 I/O table. 15

16 Yijrt stands for output in 1998 price of establishment i in industrial sector j and region r at time t, reported as industrial output, or revenue, in the original data set 6. Cijrt is capital defined as the net fixed asset average balance, deflated by fixed asset investment price index at province level. Lijrt is labor input measured by total employment 7. Mijrt is intermediate input deflated by intermediate input price index at national level. Kijrt is the firm s knowledge stock, defined as is the firm s own R&D capital constructed using a perpetual inventory method (Hall & Mairesse 1995), assuming a depreciation rate of 15% and an annual growth rate of 3% of R&D expenditure 8. Xijrt is a set of control variables to be specified later in this section. In all the variables index j stands for the 2-digit industrial sector in which foreign presence is measured. A translog production function is used in the estimation and includes second order terms of the inputs Industrial Linkage and Foreign Presence I use three proxies to measure foreign presence in the horizontal, upstream, and downstream sectors, following the definition used by Blalock and Gerlter (2005a) and Javorcik (2004). 6 Using revenue as the measurement of output raises the concern that an increase in prices may show up as an increase of productivity. Such measurement errors could be controlled when information on price is available at industry-region level. Unfortunately I do not have such information. In the final section, I will discuss how previous studies use industry level prices to test the price effects related to foreign entry. 7 Both labor and capital are adjusted for double counting of R&D activities by subtracting R&D employment from employment and subtracting an R&D capital stock constructed from the equipment investment component of R&D expenditure from the capital stock, following Hall and Mairesse (1995) and Schankerman (1981). The estimation result is qualitative similar before and after the adjustment. The coefficients of return to knowledge stock are slightly larger after the adjustment. 8 The annual growth rate of 3% is based on the sample mean of the data set I use for this study. The growth rate is often assumed to be 5% in previous research, including Hall and Mairesse 1995, Jefferson The results from estimating a Cobb-Douglas production function are similar to those from translog production function. 16

17 Horizontal_FDI jrt measures FDI in sector j, region r, at time t and is defined as the share of the sector s output produced by foreign-owned firms: (2) Horizontal _ FDI jrt = Yijrt _ Foreign Yijrt _ All i jrt i jrt The measure of horizontal FDI varies by 2-digit industrial sector, region, and time. The regions are defined at city-level and province-level alternately. Downstream_FDI is a proxy for foreign presence in the industries that are supplied by sector j in region r at time t. It captures the extent of linkage between domestic suppliers and multinational customers. Downstream_FDI is defined as: (3) Downstream _ FDI jrt = α Horizontal _ FDI jk k j krt where αjk is the proportion of sector j s output supplied to sector k taken from the 1997 input/output table at the 2-digit industry level. The proportion is calculated excluding output supplied for final consumption but including imports of intermediate products, following Javorcik (2004). Note that αjk doesn t have a region subscript because the input/output table is at national level. 17

18 To illustrate the meaning of the variable, suppose that the wheat milling industry sells half of its output to whiskey sector and half to pasta sector. If no multinationals are producing pasta but 75% of all whiskey production comes from foreign affiliates, the Downstream_FDI variable will be = 0.5* *0. To separate the effects from horizontal foreign presence, the formula excludes inputs supplied within the sector, since this effect is already captured by the Horizontal_FDI variable. Downstream_FDI increases with foreign presence in the sectors supplied by industry j and with the share of intermediates supplied to industries with foreign presence. The Upstream_FDI variable is defined as the weighted share of output in upstream sector produced by foreign affiliates. This variable is defined similarly as Downstream_FDI, except that goods produced by foreign affiliates for exports are excluded, following Javorcik (2004), since domestic customers do not capture spillover from these intermediates. (4) Upstream_ FDI jrt = σ jk k j i krt ( Y ikrt _ Foreign Export ikrt _ Foreign) i krt ( Y ikrt _ All Export ikrt _ All) where σjk is the share of inputs purchased by industry j from industry k in total intermediate inputs sourced by sector j. As before, inputs supplied within the sector are excluded. 18

19 The three proxies for industrial linkage vary by year, region, and 2-digit industrial sectors. Table 2 shows the values of the three variables from 1998 to 2005 at city level and province level. There is a trend of steady increase in all the three variables, indicating foreign presence increases over time through all the three industrial linkages. Table A2 shows the distribution of foreign presence by industry in In addition to output, I use knowledge stock of foreign firms as an alternative way to measure foreign presence in related industry sectors. The share of knowledge stock is calculated using the same formulas as output Firm Capability Firm capability is measured with knowledge stock defined in the previous section. The moderating effect of firm capability on spillover is captured by the coefficients of the interactive terms of firm s own knowledge stock and foreign presence variables. A positive sign on these coefficients indicates that firm capability facilitates the absorption of foreign technology Types of Ownership: Wholly-owned Subsidiaries and Joint Ventures To examine the spillover effects of different types of foreign ownership, namely wholly-owned subsidiaries and joint ventures, foreign presence in the related industries are separately defined for the two ownership types. For example, the share of wholly-owned (WOS) foreign investment in the downstream sectors is defines as: 19

20 (5) Downstream_ FDI( WOS) jrt = α jk WOSikrt Yikrt _ Foreign k j i krt Y ikrt i krt _ All where WOS is a dummy for wholly-owned subsidiaries. It is equal to one if the establishment is reported as a wholly-owned subsidiary. The share of joint ventures Downstream_FDI (JV) is defined in a similar manner Geographic Distance China s administrative units are currently based on a five-tier system, dividing the nation into provinces, prefecture cities, counties, townships, and villages, each consisting of the units at one level below itself. By the end of 2005, the country has 33 province-level divisions 10, subdivided into 333 prefecture cities, and in turn into 2,862 counties. The industry census data collected by the National Bureau of Statistics is generally compiled and reported at these three levels. To examine the effects of geographic distance between the source and recipient of knowledge on productivity spillovers, the three variables for foreign presence, Horizontal, Downstream, and Upstream_FDI are calculated at both city level and province level. It is expected that citylevel variables have larger effects on productivities. 10 The 33 province-level units include four types: 22 provinces, 5 autonomous regions, 4 centrally administered municipalities, and 2 special administrative regions (Hong Kong and Macau). The study excludes Hong Kong and Macau, and treats them as the sources of foreign investment, in accordance to the way statistics data is reported in China. The 4 centrally administered municipalities include Beijing, Shanghai, Tianjin, and Chongqing. These are actually large cities consisting of urban districts and rural counties. In the analysis these 4 municipalities are treated as provinces and as cities alternately. The results are similar when excluding them from either level. 20

21 3.3.6 Control for Unobserved Factors and Simultaneity Bias A set of control variables are included in the regression. The analysis needs to address the omission of unobserved variables, such as firm-, time- and region-specific factors unknown to the researcher but known to the firm that may affect the relation between firm productivity and foreign presence. Examples of these unobserved variables include a pre-existing efficient supplier base, better infrastructure, more efficient local government, macroeconomic shocks such as exchange rate fluctuation or trade restriction on certain industries, etc. firm fixed effects αi and year fixed effects αt are included in the regression to remove these unobserved effects 11. The control variables also include the level of export activities in each establishment-year and a dummy variable for any export in that year. Previous studies suggest a positive correlation between firm productivity and export activities (Blalock and Gertler 2004; Hallak and Sivadasan 2006) in the developing country context. The rationale is similar to the backward linkage with industries with foreign presence overseas customers may have higher demand on product quality and on-time delivery, and prompt exporting firms to improve productivity. Such effects are establishment-time specific and cannot be removed by fixed effects, therefore export ratio defined as export divided by total sales is included in the regression. Because a large number of establishments report zero export in the sample, a dummy variable indicating 11 As a robustness check, city-year fixed effects, αrt, are also added to the estimation to control for city-year specific effects, such as infrastructure improvement in a specific city in a certain year. The results are similar to those in the baseline model specification. 21

22 that an establishment is involved in export activities in a certain year is also included 12. It is defined as one if an establishment reports positive value of export in a certain year. The positive correlation between foreign presence and productivity could result from multinationals choosing to locate in the areas where domestic supplier s productivity is improving, or domestic customer base is expanding. Although firm and year fixed effects control for time trend and firm heterogeneities, they do not address time-variant productivity shocks that could simultaneously determine foreign presence and productivity. Examples include technology breakthrough in a supplying sector not recorded in the data but observed by domestic supplier and foreign customer, who react to the shock by adjusting inputs and output. Although it is hard to believe that multinationals making direct investment decisions respond frequently to these contemporaneous shocks given the high cost of entry and relocation, nonetheless a semi-parametric method proposed by Olley and Pakes (1996) is employed to capture the idiosyncratic shocks in the estimation, following Blalock and Gertler (2005a) 13. The method is based on assumption that investment is monotonically increasing with the productivity shock, conditional on capital. Because capital responds to the shock with a lag through contemporaneous investment, the return to the other inputs can be obtained by inverting the function of investment and capital to proxy for the unobserved shock. In 12 Similar concerns about zero foreign presence in horizontal, downstream, and upstream sectors are addressed in the same way. The estimation results on the coefficients are qualitatively similar before and after including these dummy variables. The estimation restricted to the sample with non-zero foreign shares yields qualitatively similar results too. The results reported include all these dummy variables. 13 Controlling for the simultaneity of FDI presence and productivity shocks is not the only reason to implement Olley-Pakes proxy, as discussed in Javorcik (2004). This proxy is originally constructed to controls for firmspecific productivity differences that change idiosyncratically over time (Olley and Pakes 1996). 22

23 implementation of the method, a third order polynomial of investment and capital are included in the regression 14. IV. Estimation Results 4.1 Baseline Model: Industrial Linkage The results from a model estimated with establishment fixed effects and Olley-Pakes proxies are consistent with the hypothesis that domestic sectors benefit from the foreign presence in the sectors they source from. But foreign presence in the same industrial sector and downstream sectors has no effect on or a negative effect on domestic sector s productivity. The first two columns of Table 3, Panel A, report the results for the whole sample, and the third and fourth columns are results for the sample of domestic establishments, with foreign presence measured as the share of output by foreign-owned factories in the same city in current year 15. Panel B of Table 3 reports the results when foreign presence is measured as the share at province level. To save space, the coefficients on inputs (labor, capital, R&D capital, and intermediate input) estimated with a translog productivity function are not reported. Foreign presence in upstream sectors has a positive and significant effect on productivity for the whole sample and the domestic sample. The coefficient is smaller for the domestic sample. 14 A complete procedure of implementing Olley-Pakes proxies involves two stages to yield unbiased estimation of the coefficients of inputs. Since the coefficients of the variables of interest in this study, i.e. the coefficients of FDI presence, are estimated unbiased in the first stage, I will skip the second stage to simplify the analysis. 15 The regressions on foreign presence in the previous one or two periods yield qualitatively similar results as those on current period. The coefficients on lagged foreign presence variables are about one half the scales of current period coefficients. 23

24 The results suggest that a one-standard-deviation increase in the foreign presence in the upstream sectors in the same city, i.e. an increase of 6.4 percentage points in the supplying FDI output share, is associated with a rise of 1.2 percent point in the output of domestic firms in the supplied sector. While a one-standard-deviation increase in the upstream foreign presence in the same province (5.5 percentage points) leads to a rise of 2.3 percent point in the output of domestic firms in the downstream sector 16. These results are similar when foreign presence is measured with foreign firms knowledge stock (Table 5). Foreign presence in the horizontal sector has no effect at city level, and has a relatively small positive effect at province level. One-standard-deviation increase in the foreign presence in the same sectors in the same province, i.e. 20 percentage points, is associated with 1.1 percent increase in domestic firms output (Table 3, Panel B, Column 3, Row 1). Foreign presence in the downstream sectors produces negative effects on the productivity for the whole sample and for domestic sectors, contrary to previous research that have found positive spillovers from foreign customers. The negative effects are relatively larger on domestic-owned firms and of similar scale at different regional level. One-standard-deviation increase in the foreign presence in the downstream sectors in the same city, i.e. 7.3 percentage points, is associated with 0.8 percent decrease in domestic firms output (Table 3, Panel A, Column 3, Row 2). 16 The calculation is based on the coefficient from the fixed effect regression with the Olley-Pakes proxies on the domestic sample with foreign presence measure at city-level (Table 3, Panel A, Column 3, Row 3) and provincelevel (Table 3, Panel B, Column 3, Row 3) respectively. 24

25 4.2 Firm Capability I use interactive terms of a firm s own knowledge stock and foreign presence variables to capture the moderating effect of firm capability on the adopting foreign technology. Column 2 and 4 in Table 3 report the results when these interactive terms are added to the estimation. Firms in-house knowledge stock is found to facilitate the adoption of foreign technology in the upstream sectors (Table 3, Panel A/B, Row 6), but not in other industrial sectors. The effects are similar at different regional levels, and when foreign presence is measured with either output (Table 3) or knowledge stock (Table 5). 4.3 Foreign Ownership: Joint Ventures and Wholly-owned Subsidiaries The estimation results on different types of foreign ownership are presented in Table 4 and Table 6. Because affiliates jointly owned by foreigners and domestic partners are expected to have a higher tendency to buy intermediate inputs locally, or to sell more output to local customers, the presence of joint ventures (JV) is expected to have a larger effects on productivity than the presence of wholly-owned foreign subsidiaries (WOS). To examine this question, foreign share in related industry sectors is divided into the shares of joint ventures and of wholly-owned subsidiaries 17. The results in Table 4 do not support the hypothesis. To the contrary, the coefficients of the shares of wholly-owned subsidiaries are larger than those of joint ventures. The differences between the coefficients of WOS share and JV share are statistically significant. The sign of all the coefficients show similar patterns as the previous results before dividing the share of two 17 The estimation results on WOS share only or JV share only are similar as those in Table 4 and Table 6. 25

26 types of foreign investment: horizontal and downstream foreign presence has small or negative effects on the productivity of domestic sectors, and upstream foreign presence has positive effects. A possible explanation of this counterintuitive result is wholly-owned foreign affiliates may utilize more advance technology and choose the ownership over joint ventures in order to protect their technology secret. When the extent of local sourcing or local supplying is controlled for both foreign ownership types, local firms are more likely to receive larger spillover effects from the more advanced knowledge source. 4.4 Geographic Distance: Cities and Provinces To examine the effects of geographic location on spillovers, I compare the coefficients of foreign presence measured at city-level and at province-level. The results on city- and province-level variables are presented in Panel A and B of Table 3-6, respectively. If spillover effects decrease with geographic distance, as previous literature predicts, province-level variables are expected to have a smaller effect on productivity. In fact, the coefficients on most variables are similar between city-level variable and provincelevel variables. The coefficients on upstream variables are slightly larger at province level than at city-level 18 (Table 3 and 5, Panel A/B, Row 3). 18 An estimation including FDI shares at both city and province level yields similar results as those in Table 3 and Table 5. An F-test does not reject the hypothesis that the coefficients at the two regional levels are same. 26

27 There are several explanations for the lack of distance effect. While foreign affiliates located nearby have better communication with local firms, thus enabling a more effective technology transfer, they might also bring about negative externalities by bidding up the price of labor, land, and energy supply. The spillover effects and crowding effects could cancel each other at city-level. 4.5 Robustness Checks This section describes several robustness checks to test whether the results from the previous analysis varies by geographic location and industry characteristics. First, it is concerned that firms in the regions and industries with high foreign presence may exhibit a different pattern of technology advance from those firms in regions and industries with low foreign presence. For example, a firm in an inland province mainly supplying domestic customers may have less incentive to improve technology and product quality but focus more on price competition. Thus the effects of foreign presence in the related industries may have a smaller effect on the productivity of the firm. I test the regional effects with three additional estimations: 1) the sample is divided into two parts each including the establishments in the coastal provinces and those in the inland provinces 19. The results are similar to the previous analysis. 2) The sample is divided into 18 coastal cities and the other cities. The 18 cities include 14 coastal open cities and 4 special economic zones that received favorable trade and investment policies in the 1980s and have attracted larger amounts of foreign investment in the past. The results are qualitatively similar 19 Of the 31 provinces in China, excluding Hong Kong and Macau, there are 11 provinces on the coast, 20 inland. 27

28 to the previous analysis, although the coefficients in the 18 coastal cities are of smaller scales. 3) A time trend interactive with the 18 coastal open cities is added to the regression to test whether the first-mover advantage of these cities persists over time, or, whether the firms in these cities improve productivity faster than in other cities. The results on foreign presence indicators are similar, and the coefficient on time trend is negative and significant 20. All the three tests indicate that being located near the coastal areas (or the hubs of foreign investment and international trade) is not associated with positive productivity effects, or that the advantage of receiving foreign investment early declines over time. Another explanation is that domestic factories in the coastal area are more advanced in technology and thus have less room to improve than their inland counterparts. To examine industry effects, the previous estimation is repeated on the sample of industries with high horizontal and upstream foreign presence. Eleven industries with foreign share in output or capital higher than 20% (according to Table A2, Column 1 and 2) are included in sample of high horizontal foreign presence 21. Twelve industries with upstream foreign share higher than 8% (according to Table A2, Column 4) are included in the sample of high upstream foreign presence 22. The results are stable and similar to previous analysis in both samples. In addition, there is a concern that some industries may experience a surge of productivity when a domestic factory starts to produce large orders for multinational customers. To control for the 20 The coefficient of the time trend interactive with coastal cities is -0.12, and the coefficients for year dummies are 0.5 for the period of analysis. These numbers might indicate that productivity improves overtime for the whole country, while the advantage of being located in the coastal open cities is declining over time. 21 The eleven high-horizontal-fdi industries include: garments, leather, furniture, paper, culture and sports good, rubber, plastic, metal, transportation equipment, electronics, and instruments. 22 The twelve high-upstream-fdi industries include: petroleum, paper, printing, culture and sports good, chemical, medical and pharmaceutical, chemical fiber, plastic, ordinary machinery equipment, special purposes equipment, electric equipment, and machinery, and instruments. 28

29 effects from these outliers and entrance, the previous estimations are repeated on a sample excluding large output and input changes 23, and also on a balanced sample excluding cases of empty output in any of the years from 1998 to The results are qualitatively similar to previous ones, although the coefficients become smaller and less statistically significant after excluding outliers. To test whether the effect is stable over time, I estimate all the models using foreign share in the previous one and two period. The coefficients show similar pattern as those using current period foreign shares, while coefficients on lagged variables are of smaller scales. The effect of foreign share in the previous period is about half the scale of foreign share in current period. V. Conclusion and Discussion This study examines the channels and moderating factors of productivity spillovers taking place between foreign invested firms and local firms in a developing country context. I hypothesize that technology spillovers takes place through industry linkage, either horizontally or vertically, and firms previous investment in research and development, distance to the source of knowledge, and the types of knowledge source will influence the adoption of foreign technology. Using firm level panel data from China, the analysis finds positive spillovers from multinational suppliers to local customers, and no positive spillovers within the same industrial 23 This step excludes the establishments that experienced the top and bottom 25 percentile of changes in output, capital, labor, and material, or about more than 30% increase or decrease from the previous year. 29

30 sector or from multinational customers to local suppliers. This result is consistent when foreign presence is measured with either the share of output or the share of R&D capital. Firms previous R&D investment facilitates the adoption of technology from upstream sectors. The analysis also finds that the spillovers along supply chain are associated with both joint ventures and wholly-owned foreign subsidiaries, and the effects from wholly-owned subsidiaries are larger. The results show a similar pattern of spillover effects from foreign presence at city level and at province level. The study can be improved in several ways. First, the foreign presence indicators are calculated at 2-digit industry level. Calculating these variables at 3-digit industry level using a more detailed input/output table would reduce the noises. Second, the measurement of geographic distance from knowledge sources could be refined by taking the average of foreign share in neighboring cities weighted by the distance from where the industry is located. Another concern with this study, as well as with previous studies on firm productivity, are the price changes when domestic sectors start producing for export and multinational customers and charge higher prices for these outputs, or when domestic sectors buy input from multinational suppliers, as anecdotes suggest in developing economies and confirmed by Blalock and Gertler (2005a) using industry price indices of Indonesia. Lacking more detailed information on prices, this study cannot separate the price effects from productivity effects. One way to identify such effects at national level is to run regression of price indices of output by industry on foreign presence indicators and export or import ratio. A positive correlation between foreign presence and industrial output prices indicate that supplying multinational customers or sourcing from 30

31 multinational supplier increase prices. It is hoped that these questions be answered in the future when more detailed information is available. References Aitken, Brian J., and Ann E. Harrison, Do Domestic Firms Benefit from Direct Foreign Investment? Evidence from Venezuela, American Economic Review, 89(3), Amiti, Mary, and Jozef Konings, Trade Liberalization, Intermediate Inputs, and Productivity: Evidence from Indonesia, American Economic Review 97(5): Amiti, Mary, and Javorcik, Beata Smarzynska, Trade Costs and Location of Foreign Firms in China, World Bank Working Paper, Aug Audretsch, D. B. and Feldman, M. P R&D spillovers and the geography of innovation and production, American Economic Review 86(3): Bai, C., Du, Y., Tao, Z., and Tong, S.Y Local Protectionism and Regional Specialization: Evidence from China s Industries, Journal of International Economics, 63(2), Belderbos, Rene; Capannelli, Giovanni and Kyoji Fukao, Backward Vertical Linkages of Foreign Manufacturing Affiliates: Evidence from Japanese Multinationals. World Development, 29(1), Blalock, G. and Paul Gertler, Learning from Exporting Revisited in a Less Developed Setting, Journal of Development Economics, 75(2), Blalock, G. and Paul Gertler, 2005a. Welfare Gains from Foreign Direct Investment through Technology Transfer to Local Suppliers, forthcoming, Journal of International Economics. Blalock, G. and Paul Gertler, 2005b. How Firm Capabilities Affect Who Benefits from Foreign Technology, working paper. Blömstrom, M Foreign Investment and Productive Efficiency: The Case of Mexico, Journal of Industrial Economics, 35: Blömstrom, M. and Kokko, A Multinational Corporations and Spillovers, Journal of Economic Surveys, 12: Blomström, M. and Sjöholm, F Technology Transfer and Spillovers: Does Local Participation with Multinationals Matter? European Economic Review, 43: Chang, Sea-Jin, Chung, Jaiho, and Dean Xu, forthcoming. Spillovers and Competition among Foreign and Local Firms in China, Strategic Management Journal. 31

32 Chesbrough, Henry W., The Globalization of R&D in the Chinese Semiconductor Industry, Sloan Foundation Research Report, Chesbrough, Henry W. and Feng Liang Return to R&D Investment and Spillovers in the Chinese Semiconductor Industry: A Tale of Two Segments, University of California, Berkeley, Institute of Management, Innovation, and Organization, working paper. Chung, W., Mitchell, W., and Yeung, B Foreign Direct Investment and Host Country Productivity: the American Automotive Component Industry in the 1980s, Journal of International Business Studies, 34: Coelli, Tim, Rao, D.S. Prasada, and George E. Battese, An Introduction to Efficiency and Productivity Analysis (2 nd edition), New York: Springer. Cohen, Wesley M., and Daniel A. Levinthal Absorptive Capacity: A New Perspective on Learning and Innovation, Administrative Science Quarterly, 35(1), Griliches, Zvi, and Jacques Mairesse, R&D and Productivity Growth: Comparing Japanese and US Manufacturing Firms, in Hulten, C. edit Productivity Growth in Japan and the United States, Chicago: University of Chicago Press. Hale, Galina and Cheryl Long, What Determines Technological Spillovers of Foreign Direct Investment: Evidence from China, Yale University, Economic Growth Center, working paper. Hale, Galina and Cheryl Long, Are there Productivity Spillovers from Foreign Direct Investment in China? Federal Reserve Bank of San Francisco, working paper. Hall, Bronwyn H., and Jacques Mairesse, Exploring the Relationship between R&D and Productivity in French Manufacturing Firms, Journal of Econometrics, Vol. 65, Hallak, Juan Carlos, and Jagadeesh Sivadasan, Productivity, Quality and Exporting Behavior under Minimum Quality Requirements, University of Michigan, working paper. Haskel, Jonathan E., Pereira, Sonia C. and Matthew J Slaughter, Does Inward Foreign Direct Investment Boost the Productivity of Domestic Firms? National Bureau of Economic Research (Cambridge, MA) Working Paper No. 8724, January Hippel, Eric von, The Source of Innovation, New York: Oxford University Press. Hu, Albert G.Z.; Jefferson, Gary H. and Jinchang Qian, R&D and Technology Transfer: Firm- Level Evidence from Chinese Industry, Review of Economics and Statistics, 87(4), Hu, Albert G.Z. and Gary H. Jefferson, FDI Impact and Spillover: Evidence from China's Electronic and Textile Industries, World Economy, 25(8), Jaffe, Adam B., Trajtenberg, Manuel, and Rebecca Henderson, Geographic Localization of Knowledge Spillovers as Evidenced by Patent Citations, Quarterly Journal of Economics, 108(3),

33 Javorcik, Beata Smarzynska, Does Foreign Direct Investment Increase the Productivity of Domestic Firms? In Search of Spillovers through Backward Linkages, American Economic Review, 94(3), Javorcik, Beata Smarzynska, Technological leadership and the choice of entry mode by foreign investors, in: B. Hoekman and B. Javorcik eds. Global Integration and Technology Transfer, Palgrave Macmillan and the World Bank, Jefferson, Gary H Returns to Research and Development in Chinese Industry: Evidence from State-owned Enterprises in Beijing, China Economic Review, 15(4). Jefferson, Gary H.; Hu, Albert G.Z.; Guan X. and Yu X, Ownership, performance, and innovation in China's large- and medium-size industrial enterprise sector, China Economic Review, 14(1), Keller, Wolfgang, Geographic Localization of International Technology Diffusion, American Economic Review, 92(1), Keller, Wolfgang and Stephen R.Yeaple, Multinational Enterprises, International Trade and Productivity Growth: Firm Level Evidence from the United States, National Bureau of Economic Research (Cambridge, MA) Working Paper No. 9504, February Krugman, Paul, 1991a. Increasing Returns and Economic Geography. Journal Political Economy, 99(3), Krugman, Paul, 1991b. Geography and Trade. Cambridge, MA: MIT Press. Liu, Xiaming, Siler, Pamela, Wang, Chengqi and Yingqi Wei, Productivity spillovers from Foreign Direct Investment: Evidence from UK Industry Level Panel Data, Journal of International Business Studies, 31(3), Mairesse, Jacques and Mohamed Sassenou, R&D and productivity: A survey of econometric studies at the firm level, Science-Technology-Industry Review 8 (OECD, Paris) Mansfield, Edwin, Price Indexes for R and D Inputs, , Management Science, 33(1), Marshall, Alfred, Principles of Economics, 8 th edition. London: Macmillan, Markusen, James R. and Anthony J.Venables, Foreign Direct Investment as a Catalyst for Industrial Development, European Economic Review, 43(2), Martin, X. and Robert M. Salomon, Knowledge Transfer Capacity: Implications for the Theory of the Multinational Corporation, Journal of International Business Studies, 34: Melitz, Marc J The Impact of Trade on Intra-industry Reallocations and Aggregate Industry Productivity, Econometrica, 71(6),

34 Mowery, David C., Oxley, Joanne E., and Brian S. Silverman, Strategic Alliances and Interfirm Knowledge Transfer, Strategic Management Journal, 17 (Special Issue: Knowledge and the Firm.), Mowery, David C. and Arvids A. Ziedonis, The Geographic Reach of Market and Non-market Channels of Technology Transfer: Comparing Citations and Licenses of University Patents, NBER working paper Olley, Steven G., and Ariel Pakes, The Dynamics of Productivity in the Telecommunications Equipment Industry, Econometrica, 64(6), Pack, Howard, and Kamal Saggi, Vertical Technology Transfer via International Outsourcing, Journal of Development Economics, 65(2), Schankerman, M The Effects of Double-Counting and Expensing on the Measured Returns to R&D, Review of Economics and Statistics, Salomon, Robert M Spillovers to Foreign Market Participants: Assessing the Impact of Export Strategies on Innovative Productivity, Strategic Organization, Vol.4 (2): Spencer, J.W (forthcoming). The Impact of Multinational Enterprise Strategy on Indigenous Enterprises: Horizontal Spillovers and Crowding out in Developing Countries, Academy of Management Review, 2008: 33(2). Todorova, Gergana, and Boris Durisin, Absorptive Capacity: Valuing a Reconceptualization, Academy of Management Review, 32(3): Qian, Yingyi, The Process of China s Market Transition : the Evolutionary, Historical, and Comparative Perspectives, Journal of Institutional and Theoretical Economics, 156(1), Teece, David J Profiting from Technological Innovation: Implications for Integration, Collaboration, Licensing, and Public Policy, Research Policy, 15: Tybout, James R Manufacturing Firms in Developing Countries: How Well Do They Do, and Why? Journal of Economic Literature, 38(1), United Nations Centre on Transnational Corporations (UNCTC), The competitiveness challenge: Transnational corporations and industrial restructuring in developing countries. New York: United Nations, Zahra, Shaker A., and Gerard George, Absorptive capacity: A review, reconceptualization, and extension, Academy of Management Review, 27: Zhang, Yan, Li, Haiyang, Hitt, Michael A., and Geng Cui R&D Intensity and International Joint Venture Performance in an Emerging Market: Moderating Effects of Market Focus and Ownership Structure, Journal of International Business, 38(6): Zhao, Minyuan, Conducting R&D in Countries with Weak Intellectual Property Rights Protection, Management Science, 52:

35 Tables and Graphs Table 1 Summary Statistics on Output, Input, and Spillover Variables (in 1998 price) Number of Standard Variables Mean Min Max Observations Deviation Log Output (1,000 Yuan*) 178, Log Capital (1,000 Yuan) 175, Log Labor (persons) 177, Log Material (1,000 Yuan) 178, Log Knowledge Stock (1,000 Yuan) 178, Log Investment (1,000 Yuan) 178, Horizontal Foreign Share city level: Output 178, Downstream Foreign Share city level: Output 169, Upstream Foreign Share city level : Output 169, Horizontal Foreign Share province level: Output 178, Downstream Foreign Share province level: Output 169, Upstream Foreign Share province level: Output 169, Horizontal Foreign Share city level: Knowledge Stock 177, Downstream Foreign Share city level: Knowledge Stock 169, Upstream Foreign Share city level: Knowledge Stock 169, Horizontal Foreign Share province level: Knowledge Stock 177, Downstream Foreign Share province level: Knowledge Stock 169, Upstream Foreign Share province level: Knowledge Stock 169, Export ratio 177, * The official exchange rate between Yuan and US$ is 8.27 Yuan/US$ during Table 2 Percentage of foreign output share at City and Province Level Number of Horizontal Downstream Upstream Year Industries Mean Mean Mean City-Level Province-Level

36 Table 3 Total Factor Productivity: Spillover Effects from Foreign Share of Output Dependent Variable: Log industrial output at establishment level (in 1,000 Yuan) Panel A Fixed Effect Regressions on Foreign Share in the Cities All Domestic Horizontal_FDI (0.0098) (0.0406) (0.0131) (0.0505) Downstream_FDI (0.0345)*** (0.1632) (0.0394)*** (0.1844) Upstream_FDI (0.0405)*** (0.1947) (0.0467)*** (0.2191)* Horizontal_FDI*Knowledge stock (0.0045) (0.0057) Downstream _FDI*Knowledge stock (0.0186) (0.0211) Upstream _FDI*Knowledge stock (0.0214)** (0.0240)*** Number of Observations Number of Establishments R-squared Panel B Fixed Effect Regressions on Foreign Share in the Provinces All Domestic Horizontal_FDI (0.0115)*** (0.0470)** (0.0146)*** (0.0593) Downstream_FDI (0.0491)* (0.2232) (0.0577)*** (0.2653) Upstream_FDI (0.0545)*** (0.2650) (0.0666)*** (0.3215) Horizontal_FDI*Knowledge stock (0.0053) (0.0067) Downstream _FDI*Knowledge stock (0.0253) (0.0300) Upstream _FDI*Knowledge stock (0.0292) (0.0351)** Number of Observations Number of Establishments R-squared Notes: Robust clustered standard errors in parentheses. The error terms are corrected for clustering for each establishment. All the regressions include establishment fixed effects, year dummies, and Olley Pakes proxy. The dependent variable is establishment output. The right-hand side variables include ln capital stock, ln labor, ln materials, ln R&D capital stock, and their second order terms in a translog production function. All the regressions include dummy variables for foreign presence in horizontal, downstream, and upstream industries, export ratio, export sector dummy, and exclude upper 1% outliers of input variables. *, **, *** significant at 10%, 5%, and 1% level. 36

37 Table 4 Total Factor Productivity: Spillover Effects from Foreign Share of Output divided between Wholly-owned Subsidiaries and Joint Ventures Dependent Variable: Log industrial output at establishment level (in 1,000 Yuan) Panel A Fixed Effect Regressions on JV and Wholly-owned Subsidiaries Share in the Cities All Domestic Horizontal_FDI (Wholly-owned) (0.0136)** (0.0816) (0.0174) (0.0798) Downstream_FDI (Wholly-owned) (0.0547)*** (0.2918)** (0.0705)*** (0.3516) Upstream_FDI (Wholly-owned) (0.0688)*** (0.3978)** (0.0877)*** (0.4648) Horizontal_FDI (Joint Ventures) (0.0114) (0.0430) (0.0162) (0.0631) Downstream_FDI (Joint Ventures) (0.0411) (0.2092)* (0.0453) (0.2279) Upstream_FDI (Joint Ventures) (0.0494)** (0.2325) (0.0527) (0.2559)** Horizontal_FDI (Wholly-owned)*Knowledge stock (0.0088) (0.0085) Downstream_FDI (Wholly-owned) *Knowledge stock (0.0311) (0.0375) Upstream_FDI (Wholly-owned) *Knowledge stock (0.0424) (0.0488) Horizontal_FDI (Joint Ventures) *Knowledge stock (0.0048) (0.0072) Downstream_FDI (Joint Ventures) *Knowledge stock (0.0241)* (0.0262) Upstream_FDI (Joint Ventures) *Knowledge stock (0.0260)** (0.0282)** Number of Observations Number of Establishments R-squared Panel B Fixed Effect Regressions on JV and Wholly-owned Subsidiaries Share in the Provinces All Domestic Horizontal_FDI (Wholly-owned) (0.0142) (0.0700)*** (0.0182)* (0.0900)** Downstream_FDI (Wholly-owned) (0.0696)*** (0.3926)*** (0.0891)*** (0.4725)* Upstream_FDI (Wholly-owned) (0.0962)*** (0.5885)** (0.1212)*** (0.6777)* Horizontal_FDI (Joint Ventures) (0.0148)*** (0.0630) (0.0192)** (0.0772) Downstream_FDI (Joint Ventures) (0.0768)*** (0.3620) (0.0898) (0.4249) Upstream_FDI (Joint Ventures) (0.0719)** (0.3678) (0.0870) (0.4258) Horizontal_FDI (Wholly-owned)*Knowledge stock (0.0076)*** (0.0096)** Downstream_FDI (Wholly-owned) *Knowledge stock (0.0418) (0.0494) Upstream_FDI (Wholly-owned) *Knowledge stock (0.0637) (0.0694) Horizontal_FDI (Joint Ventures) *Knowledge stock (0.0071) (0.0088) Downstream_FDI (Joint Ventures) *Knowledge stock (0.0419) (0.0491) 37

38 Upstream_FDI (Joint Ventures) *Knowledge stock (0.0417) (0.0473)* Number of Observations Number of Establishments R-squared Notes: Robust clustered standard errors in parentheses. The error terms are corrected for clustering for each establishment. All the regressions include establishment fixed effects, year dummies, and Olley Pakes proxy. The dependent variable is establishment output. The right-hand side variables include ln capital stock, ln labor, ln materials, ln R&D capital stock, and their second order terms in a translog production function. All the regressions include dummy variables for foreign presence in horizontal, downstream, and upstream industries, export ratio, export sector dummy, and exclude upper 1% outliers of input variables. *, **, *** significant at 10%, 5%, and 1% level. 38

39 Table 5 Total Factor Productivity: Spillover Effects from Foreign Share of Knowledge Stock Dependent Variable: Log industrial output at establishment level (in 1,000 Yuan) Panel A Fixed Effect Regressions on Foreign Share in the Cities All Domestic Horizontal_FDI (0.0075) (0.0337) (0.0110) (0.0510) Downstream_FDI (0.0433)*** (0.2066) (0.0541)*** (0.2431) Upstream_FDI (0.0446)*** (0.2098)* (0.0524)*** (0.2428) Horizontal_FDI*Knowledge stock (0.0038) (0.0060) Downstream _FDI*Knowledge stock (0.0236) (0.0281) Upstream _FDI*Knowledge stock (0.0230) (0.0264) Number of Observations Number of Establishments R-squared Panel B Fixed Effect Regressions on Foreign Share in the Provinces All Domestic Horizontal_FDI (0.0085) (0.0388)*** (0.0111)** (0.0520) Downstream_FDI (0.0502)*** (0.2540) (0.0654)*** (0.3049) Upstream_FDI (0.0550)*** (0.2819) (0.0677)*** (0.3571) Horizontal_FDI*Knowledge stock (0.0045)*** (0.0062) Downstream _FDI*Knowledge stock (0.0290) (0.0351) Upstream _FDI*Knowledge stock (0.0306) (0.0387)** Number of Observations Number of Establishments R-squared Notes: Robust clustered standard errors in parentheses. The error terms are corrected for clustering for each establishment. All the regressions include establishment fixed effects, year dummies, and Olley Pakes proxy. The dependent variable is establishment output. The right-hand side variables include ln capital stock, ln labor, ln materials, ln R&D capital stock, and their second order terms in a translog production function. All the regressions include dummy variables for foreign presence in horizontal, downstream, and upstream industries, export ratio, export sector dummy, and exclude upper 1% outliers of input variables. *, **, *** significant at 10%, 5%, and 1% level. 39

40 Table 6 Total Factor Productivity: Spillover Effects from Foreign Share of Knowledge Stock divided between Wholly-owned Subsidiaries and Joint Ventures Dependent Variable: Log industrial output at establishment level (in 1,000 Yuan) Panel A Fixed Effect Regressions on JV and Wholly-owned Subsidiaries Share in the Cities All Domestic Horizontal_FDI (Wholly-owned) (0.0128)* (0.0672) (0.0182) (0.0899) Downstream_FDI (Wholly-owned) (0.0793)*** (0.4082)* (0.1038)*** (0.6105) Upstream_FDI (Wholly-owned) (0.0825)*** (0.4606) (0.1046)*** (0.6606) Horizontal_FDI (Joint Ventures) (0.0085) (0.0374)** (0.0131)* (0.0592) Downstream_FDI (Joint Ventures) (0.0494)** (0.2478) (0.0606)*** (0.2777) Upstream_FDI (Joint Ventures) (0.0503)*** (0.2632) (0.0571)*** (0.2778) Horizontal_FDI (Wholly-owned)*Knowledge stock (0.0072) (0.0094) Downstream_FDI (Wholly-owned) *Knowledge stock (0.0437) (0.0627) Upstream_FDI (Wholly-owned) *Knowledge stock (0.0497) (0.0664) Horizontal_FDI (Joint Ventures) *Knowledge stock (0.0042)** (0.0071) Downstream_FDI (Joint Ventures) *Knowledge stock (0.0290) (0.0329) Upstream_FDI (Joint Ventures) *Knowledge stock (0.0297) (0.0307) Number of Observations Number of Establishments R-squared Panel B Fixed Effect Regressions on JV and Wholly-owned Subsidiaries Share in the Provinces All Domestic Horizontal_FDI (Wholly-owned) (0.0132)** (0.0686)** (0.0178) (0.0856) Downstream_FDI (Wholly-owned) (0.0714)*** (0.4411) (0.0887)*** (0.5068) Upstream_FDI (Wholly-owned) (0.0954)*** (0.5153) (0.1031)*** (0.5795) Horizontal_FDI (Joint Ventures) (0.0099) (0.0441)** (0.0134) (0.0600) Downstream_FDI (Joint Ventures) (0.0673) (0.3397) (0.0895)* (0.4170) Upstream_FDI (Joint Ventures) (0.0682)*** (0.3907) (0.0860)** (0.4752) Horizontal_FDI (Wholly-owned)*Knowledge stock (0.0077)*** (0.0097) Downstream_FDI (Wholly-owned) *Knowledge stock (0.0481) (0.0550) Upstream_FDI (Wholly-owned) *Knowledge stock (0.0574)* (0.0607)* Horizontal_FDI (Joint Ventures) *Knowledge stock (0.0052)* (0.0071) Downstream_FDI (Joint Ventures) *Knowledge stock (0.0392) (0.0477) 40

41 Upstream_FDI (Joint Ventures) *Knowledge stock (0.0437) (0.0521) Number of Observations Number of Establishments R-squared Notes: Robust clustered standard errors in parentheses. The error terms are corrected for clustering for each establishment. All the regressions include establishment fixed effects, year dummies, and Olley Pakes proxy. The dependent variable is establishment output. The right-hand side variables include ln capital stock, ln labor, ln materials, ln R&D capital stock, and their second order terms in a translog production function. All the regressions include dummy variables for foreign presence in horizontal, downstream, and upstream industries, export ratio, export sector dummy, and exclude upper 1% outliers of input variables. *, **, *** significant at 10%, 5%, and 1% level. Table 7 Pair-wise Correlation Matrix Log Output 1 2. Log Capital Log Labor Log Material Log Knowledge Stock Log Investment Horizontal Foreign Share city level: Output Downstream Foreign Share city level: Output Upstream Foreign Share city level : Output Horizontal Foreign Share province level: Output Downstream Foreign Share province level: Output Upstream Foreign Share province level: Output Note: All the correlations are significant at 1% level. 41

42 Figure 1 Foreign Direct Investment Inflows into China: Source: China Statistic Yearbook FDI in China Million US$ Number of Firms Actually Utilized FDI (Million US$) Number of Foreign Invested Firms FDI in Manufacturing (Million US$) 42

43 Appendix: Additional Tables and Graphs Table A1 Mean of Output, Input, and Spillover Variables by Domestic and Foreign Sectors (in 1998 price) Variables Domestic Foreign Log Output (1,000 Yuan*) Log Capital (1,000 Yuan) Log Labor (persons) Log Material (1,000 Yuan) Log Knowledge Stock (1,000 Yuan) Log Investment (1,000 Yuan) Horizontal Foreign Share city level: Output Downstream Foreign Share city level: Output Upstream Foreign Share city level : Output Horizontal Foreign Share province level: Output Downstream Foreign Share province level: Output Upstream Foreign Share province level: Output Horizontal Foreign Share city level: Knowledge Stock Downstream Foreign Share city level: Knowledge Stock Upstream Foreign Share city level: Knowledge Stock Horizontal Foreign Share province level: Knowledge Stock Downstream Foreign Share province level: Knowledge Stock Upstream Foreign Share province level: Knowledge Stock Export ratio (export/sales ) Number of Observations 134,671 23,034 * The official exchange rate between Yuan and US$ is 8.27 Yuan/US$ during Figure A1 Export Ratio (export/sales) of domestic industry sectors and foreign-owned sectors 43

44 Figure A2 Geographic Distribution of Foreign Direct Investment in China, 2000 Source: The Atlas of Population, Environment and Sustainable Development of China 2003 Figure A3 Per Capita GDP by regions in China, 2000 Source: The Atlas of Population, Environment and Sustainable Development of China