Determinants of privatization in China: The role of the presence of foreign firms

China Economic Review 41 (2016) 196–221

Contents lists available at ScienceDirect

China Economic Review

Determinants of privatization in China: The role of the presence of foreign firms夽 Yi Liu a,1 , Xue Li b,2 , Sajal Lahiri c, * a

The College of Economics and Trade, Hunan University, North Campus, Shi Jia Chong, Changsha, Hunan 410079, PR China Center for Economics, Finance and Management Studies, Hunan University, North Campus, Shi Jia Chong, Changsha, Hunan, 410079, PR China c Southern Illinois University Carbondale, 1000 Faner Drive, Carbondale, IL62901, United States b

A R T I C L E

I N F O

Article history: Received 20 April 2016 Received in revised form 9 October 2016 Accepted 10 October 2016 Available online 13 October 2016 JEL classification: D21 D43 D69 O53

A B S T R A C T This study investigates how the presence of foreign firms in a sector influences the privatization policy of domestic firms in that sector in China. We consider several variables to proxy for such a presence, from the perspective of the relative production scale, R&D and marketing, and labor productivity. By using the enterprise surveys carried out by the World Bank in 2005 of nearly 12,400 Chinese firms located in over 100 cities, we find that a rise in the presence of foreign firms increases the extent of the private ownership of domestic firms in a nonlinear fashion. Additionally, we apply IV-Tobit estimation with valid instruments and Tobit estimation with lagged key variables to deal with the possible endogeneity in the relationship between the presence of foreign firms and the privatization decision. Empirical evidence supports the main findings. Published by Elsevier Inc.

Keywords: Privatization Foreign presence China Endogeneity

1. Introduction Since the early 1990s, the Chinese government has reformed its economy in two directions: encouraging foreign direct investment (FDI) and (partially) privatizing government-owned firms. In this study, we examine how these two policy initiatives interact with each other in China. Specifically, we investigate whether the presence of foreign firms (termed foreign presence hereafter) in an industry affects the level of public ownership of domestic firms in that industry. Moreover, we examine the extent to which the competition provided by foreign firms in an industry is a significant determinant of privatization.

夽 This work was supported by the Fundamental Research Funds for the Central Universities of China (531107050701, 2012); the Scientific Research Starting Foundation for the Returned Overseas Chinese Scholars, Ministry of Education of China ([2014]1685, 2014); and the National Natural Science Foundation of China (NSFC, 71603078, 2016). We are also indebted to Xiaojun Wang, Mahelet Fikru, and the anonymous referee(s) for their helpful comments. Any remaining mistakes are our own. * Corresponding author. Tel.: +1(618) 453-9472; fax: +1(618) 453-2717. E-mail addresses: [email protected] (Y. Liu), [email protected] (X. Li), [email protected] (S. Lahiri). 1 2

Tel./Fax: +86(731) 8868-4825. Tel./Fax: +86(731) 8868-4822.

http://dx.doi.org/10.1016/j.chieco.2016.10.002 1043-951X/Published by Elsevier Inc.

Y. Liu et al. / China Economic Review 41 (2016) 196–221

197

While numerous studies investigate the consequences of privatization,3 the literature on the determinants of privatization is relatively small. Clarke and Cull (2002) investigated the political determinants of privatization among Argentine public banks and found, inter alia, that political incentives such as provincial deficits, federal transfers to provinces, and the political affiliation of the provincial policymaker significantly affect the probability of privatization. Mizutani and Uranishi (2010) examined the determinants of privatization by using a cross-sectional dataset of 74 Japanese public corporations in 2001 and showed that market conditions and the extent of regulations are important. Roberts and Saeed (2012) focused on the economic, political, and institutional determinants of privatization, using an unbalanced panel macro dataset with 50 countries over 1988–2006. Sprenger (2011) examined the effect of firm quality on the privatization decision, using data for 530 Russian manufacturing firms, and found that the privatization decision is positively associated with firm size, labor productivity, and average wages. Zohlnhofer, Obinger, and Wolf (2008) analyzed the effect of political control in the privatization process in EU and OECD countries between 1990 and 2000 and found significant partisan influences on privatization decisions in the OECD sample but not in the EU sample. Since China began its privatization policy in the early 1990s, several studies have shown that government-owned firms (including nationally, provincially, and locally owned) were suffering from low productivity, wasted resource, and over-staffing. From 1995 to 2001, the then prime minister Rongji Zhu implemented a policy to reduce the number of state-owned enterprises (SOEs) from 1.2 million to 0.47 million. However, as opposed to the privatization experiences of other countries, the Chinese government still controls a large share of those partially privatized firms. In other words, the country’s privatizations have tended to be partial. Indeed, the ratio of the general assets of those state-controlled firms to the total general assets of all industrial firms in China is still more than 50%, according to Garnaut, Song, and Tenev (2005). The process of privatization has also been selective in China. Among studies of the determinants of privatization in China, Ng, Yuce, and Chen (2009) examined the relationship between state ownership and the performance of privatized firms, using data for 1996–2003, and found a convex relationship between the two variables. Further, Li (2003) showed that a firm is more likely to be privatized in the presence of fiercer product market competition and more binding governmental budget constraints. In this study, we examine the degree to which foreign presence in an industry determines the extent of the public ownership of firms in that industry. In particular, we hypothesize that local firm-level competition, as measured by foreign presence, influences the privatization decision in China. Since it is expected that foreign presence fosters local privatization efforts as new capital inflows, the technology and managerial skills that accompany the setup of foreign plants in the local market make the environment more prone to competition, providing the government with an incentive to privatize inefficient SOEs. The increase in foreign presence in the local market thus benefits inefficient SOEs in need of restructuring by raising funds, providing new technologies, improving human capital, offering new managerial skills, and enhancing corporate governance. Thus, foreign presence is assumed to raise domestic privatization since economies that are more open to foreign plants usually have better institutional frameworks and the higher protection of private property rights, which facilitate foreign firms’ participation in the local market (Boubakri et al., 2013). We measure foreign presence from various dimensions such as relative production scale, expenditure on R&D and marketing, and labor productivity. Including different measures of foreign presence as an explanatory variable allows us to account for the different ways in which domestic firms in an industry face competition from foreign firms in the same industry in China. We first represent foreign presence from the perspective of production scale by using the average net fixed assets of all foreign firms in a sector relative to that of domestic plants in the same industry, plus the average net fixed assets of all other domestic firms in the same sector relative to that of the same domestic plants (termed FP via NFA). The idea here is that a large foreign cohort indicates a tremendous production scale and enormous foreign competition for local firms. Additionally, the large production scale provided by the domestic cohort would promote foreign presence.4 We also follow Fagerberg (1988) and consider a composite measure including relative average foreign R&D expenditure and advertising expenditure in the sector relative to that of the local firm in the same sector (FP via RDA). Larger foreign presence is expected to have a competitive edge because of foreign firms’ ability to provide internal financing for a number of activities including R&D and advertising. Following studies that consider the efficiency levels of firms as a determinant (see, for example, Li, 2003; Liu & Woo, 2001; Yang, 1998), we consider the average unit labor cost of all foreign firms in a sector relative to that of domestic firms in the same industry as a measure foreign presence (FP via ULC). As stated, these three variables, FP via NFA, FP via RDA, and FP via ULC, represent foreign presence relative to the domestic plant in the same industry. In addition, we examine the nonlinearity of FP via NFA and FP via RDA on the extent of the public ownership of firms in that industry, while controlling for lagged FP via ULC and assuming firm-level foreign presence measured by the productive efficiency constant. Since the prevailing hypothesis asserts that partial privatization is the optimal solution under foreign competition, if foreign presence indeed represents competition, increasing state ownership would cause productive inefficiency while

3 The consequences of privatization can be categorized into the effects on firm productivity (Bai, Lu, & Tao, 2009; Bridgman, Gomes, & Teixeira, 2011), the effects on employment (Bhaskar & Khan, 1995; Chang & Brada, 2011; Naito, 2013), and the effects on macro-level FDI flows (Boubakri, Cosset, Debab, & Valery, 2013; Reece & Abdoul, 2012). A large volume of the literature on the consequences of privatization has been based on demand. 4 The average production scale of the local cohort affects foreign entry in two contrasting ways. First, a large average scale of the local cohort indicates the prosperous development of the industry, which may ensure the profitability of foreign entry and thus spur foreign presence. Second, the large average scale of the local cohort may also indicate the presence of monopolists or oligopolists in the industrial local cohort, preventing foreign entry and reducing foreign presence. To verify the effect that dominates in our case, we examine the correlation between the average NFA of the foreign cohort and that of the local cohort. We find that these are positively correlated, indicating that the spurring effect dominates.

198

Y. Liu et al. / China Economic Review 41 (2016) 196–221

guaranteeing resources and funding (Matsumura, 1998). We also test whether FP via NFA and FP via RDA interact with FP via ULC in affecting the firm-level privatization decision in China. This test aims to check if FP via ULC complements or substitutes for FP via NFA/RDA in the privatization decision: for firms with higher FP via ULC, the effect of FP via NFA/RDA on privatization could be larger or smaller. Endogeneity is a vital issue to be tackled in the study. While foreign presence is expected to enhance the pursuit of privatization, privatization might also reinforce foreign presence. For example, privatization, through share issues, has an optimistic impact on stock market development and is thus likely to draw FDI inflows into the local market (Boubakri et al., 2013). Thus, this study may be subject to two types of endogeneity. The first type is reverse causality running from privatization to foreign presence. The second type is omitted variable bias (Chang, Fu, Low, & Zhang, 2015). In terms of the econometric methodology, by using the Enterprise Surveys of the World Bank (ESWB; Word Bank, 2005) of 12,400 industrial plants in more than 100 Chinese cities, we first test the existence of endogeneity by using the Smith–Blundell (SB) test. Second, based on this test, we estimate both an instrumental variable (IV)-Tobit model with instruments for foreign presence and a Tobit model with lagged variables, by controlling for the variables that affect both the privatization decision and foreign presence to avoid omitted variable bias. Finally, we use the weak instrument robust test to check the validity and shortcomings of the instruments used. The main contribution of this study is that rather than simply defining foreign presence at the macro level as FDI volume or the number of foreign firms in an industry, we aim to measure foreign presence at the micro level by using diversified aspects such as NFA, RDA, and ULC. As a result, we document that foreign presence fosters the local firm-level privatization decision. Second, the study uses both IV-Tobit estimation with exogenous instruments and Tobit estimation with lagged foreign presence proxies to tackle the endogeneity problem and avoid bias. The main conclusion is that foreign presence in the local market, as measured by the reciprocal of FP via NFA, FP via RDA, or FP via ULC, decreases the extent of the state ownership of domestic firms in that sector. Moreover, FP via NFA has a convex impact on the extent of plant-level state ownership in the local market in China, suggesting that partial privatization is optimal under foreign competition. Further, FP via NFA interacts positively with FP via ULC in affecting the firm-level privatization decision in China, indicating that local firms challenged by foreign presence from production scale also face difficulties competing with foreign firms in terms of productive efficiency. Hence, the domestic government must reduce the state ownership of local plants further. The remainder of this paper is arranged as follows. The nature of privatization in China is discussed in Section 2; the empirical model is presented in Section 3; Section 4 describes the data; the empirical results are presented in Section 5, and concluding remarks are made in Section 6. 2. Privatization in China The privatization of Chinese SOEs has been carried out in four phases: implicit privatization (1997–2000), pension-driven privatization (2001–2002), share conversion (2005–2006), and market-oriented privatization (2013 to present).5 During the implicit privatization phase, the average ratio of state ownership decreased from 45% to 31% by eliminating the weaknesses of most large and medium-sized SOEs through “strategic restructuring” (Wang, Xu, & Zhu, 2004). Second, before pension-driven privatization, the pension system in China ran on a pay-as-you-go basis, and it was unable to sustain its rapidly aging population. Thus, in 2000, raising money for a newly established national pension fund emerged as the major driving force for selling state assets. Third, the share conversion reform did not necessarily mean diminishing the state share because of its tradability, since it may still be retained by the state. In 2005, the relative size of state ownership in Chinese listed companies fell sharply, from 62% to 55%. The ideology of market-oriented privatization was at the top of the government’s agenda in November 2013 for two reasons. First, market-oriented privatization aims to revitalize the state-owned sector of the economy through development in the nonpublic sector, which stimulates the creativity of the whole economy. Second, market-oriented privatization is supposed to speed up the improvement of the modern market system, particularly the market price mechanism. According to Naughton (2007), the widespread abuse of managerial power in SOEs prevails because of the lack of market-oriented mechanisms, leading to resource waste/misallocation and inefficiency. Market-oriented privatization is supposed to improve matters in this respect. The top three Chinese privatization cases from 2000 to 2008 in terms of value were the Industrial and Commercial Bank of China (ICBC), Bank of China (BOC), and Petro China Co. Ltd. (PCCL). The ICBC privatization was the world’s largest initial public offering (IPO) in 2006, with its value, $22 billion, surpassing the second largest IPO by Japan’s NTT DoCoMo in 1998 ($18.4 billion). It is also the first establishment listed simultaneously on both the Hong Kong Stock Exchange and the Shanghai Stock Exchange. The large shareholders of ICBC are the Chinese Ministry of Finance, Goldman Sachs (an American multinational investment banking firm), Dresdner Bank (a wholly owned subsidiary of Commerce bank), and American Express. The BOC privatization started from its listing on the Hong Kong Stock Exchange on June 1, 2006, by raising $9.7 billion in the H-share Global Offering. The over-allotment option was then exercised on June 7, 2006, raising the total value of its IPO to $13.7 billion. The largest shareholders of BOC are the State Administration of Foreign Exchange (an investment arm of the Chinese central government), the Royal Bank of Scotland China, Asia Finance Holdings (a wholly owned subsidiary of Temasek), and the National Council for Social Security Fund (PRC state pension fund).

5

See Ma (2008) for a discussion of the first three phases.

Y. Liu et al. / China Economic Review 41 (2016) 196–221

199

PCCL, the largest oil and gas producer in China, first traded on the Hong Kong and New York Stock Exchanges and then returned to the mainland capital market by listing on the Shanghai Stock Exchange on November 5, 2007 for A share trading. PCCL raised 73.6 billion yuan ($9.2 billion) in Shanghai by selling four billion A shares, or 2.2% of its expanded share capital, in the world’s biggest IPO up to November 2007. The top three shareholders of PCCL are the China National Petroleum Corporation (state-owned shareholder), Hong Kong Securities Clearing Co. Ltd. (H share), and China Life Insurance Co. Ltd. (A share) (Table 1). 3. Empirical model When selecting a suitable estimation strategy for the present study, two issues demand attention. First, privatization may influence the inflow of FDI (Roberts & Saeed, 2012) and therefore some of our key explanatory variables. Thus, possible twoway causality needs to be tackled. The second issue is that our dependent variable (i.e., public ownership as a proportion of total ownership) is restricted to being between zero and 100, and thus ordinary least squares (OLS) estimates are inconsistent because the censored sample cannot represent the population. Thus, we need to deploy a censored regression model when the dependent variable is observed only over some interval of its support. To address these issues, we deploy both the IV-Tobit estimation technique (Newey, 1987) and the Tobit model with lagged values of regressors (Bortolotti, D’Souza, Fantini, & Megginson, 2002; Boubakri, Cosset, Guedhami, & Saffar, 2011; Roberts & Saeed, 2012). In the IV-Tobit model, we emphasize the two-way causality between micro-level foreign presence and its privatization policy, by including the exogenous instrumental variables, which only explain privatization policy through firmlevel foreign presence. In the following phase of the empirical analysis, we adopt the Tobit model with lagged key regressors (i.e., lagged proxies of foreign presence measured in diversified dimensions). Lagged regressors are beneficial to solve simultaneity between foreign presence and privatization policy. Additionally, it is crucial to investigate the interactions among the diversified dimensions of foreign presence on micro-level privatization policy in the Tobit model. In the robustness check, we use an alternative categorical dependent variable, ownership structure, rather than state ownership to describe privatization

Table 1 Privatization cases in China, 2000–2008. Company name

Year

Sector

Deal type

Value ($millions)

PetroChina China United Communications China Telecom China Life Insurance Co. Ltd. Yangtze Electric Power Co. Bank of Communications China Netcom Bank of China Bank of China China Construction Bank Industrial & Commercial Bank of China China Communications Construction Company Bank of China China Merchant Bank Industrial and Commercial Bank of China Daqin Railway Co. Ltd. Guangshen Railway Co. Ltd. China Coal Energy Co, LTD. China Molybdenum Co., Ltd. China Railway Group China Railway Group China Shipper Container Lines Sinotrans Shipping Sinotruk China Shenhua Energy Co PetroChina Co. Ltd. Bank of Beijing Bank of Communications China CITIC Bank Corporation China Construction Bank Corporations China Life Insurance Company China Pacific Insurance Ping An Insurance (Group) Company of China Aluminum Corporation of China (Chalco) China Coal Energy Co. Ltd. China Railway Construction Corporation Ltd. China Railway Construction Corporation Ltd.

2000 2000 2002 2003 2003 2004 2004 2005 2005 2005 2005 2006 2006 2006 2006 2006 2006 2006 2007 2007 2007 2007 2007 2007 2007 2007 2007 2007 2007 2007 2007 2007 2007 2007 2008 2008 2008

Energy Infrastructure Infrastructure Financial Infrastructure Financial Infrastructure Financial Financial Financial Financial Competitive Financial Financial Financial Infrastructure Infrastructure Primary Competitive Competitive Competitive Competitive Competitive Competitive Energy Energy Financial Financial Financial Financial Financial Financial Financial Primary Primary Competitive Competitive

Partial Partial

2890.00 5653.00 1430.00 3021.60 1200.00 2100.00 1140.00 3100.00 3100.00 3000.00 3000.00 2380.00 13,714.00 2660.00 22,041.00 1922.00 1324.00 1944.00 1037.00 3075.00 2835.00 2100.00 1470.00 1200.00 9124.00 9154.00 2000.00 3454.00 6051.00 7955.00 3881.00 7700.00 5327.00 1010.00 3761.00 3259.00 2606.00

Partial Partial 20% stake 10% stake 10% stake 9% stake 10% stake Partial (24.5%) Partial Partial Partial (15%) Partial Partial Partial (29%) H-shares A-shares A-shares A-shares H-shares H-shares A-shares A-shares A-shares

A-shares A- and H-share A shares A shares H shares

Note: Data source: Privatization Database, Financial and Private Sector Development, World Bank. The privatization cases reported in this table are only those above $1billion.

200

Y. Liu et al. / China Economic Review 41 (2016) 196–221

policy at the firm level. In this case, we employ multinomial Logit (MNL) estimation with lagged values of the key explanatory variables to verify the influence of foreign presence on privatization, given the existence of simultaneity.6 3.1. IV-Tobit & Tobit method According to Newey (1987) and Smith and Blundell (1986), IV-Tobit uses maximum likelihood estimation, taking into account that the observations of the dependent variable (state ownership y1i ) are censored. The basic equation is ⎧ ∗ ∗ ⎪ ⎪ ⎨y1i , if 0 < y1i < 1 ∗ y1i = 1, if y1i ≥ 1, ⎪ ⎪ ⎩0, if y∗1i ≤ 0

(1)

In ordinary Tobit estimation, we estimate ∗   = y2i c1 + x1i c2 + e i , y1i

(2)

by taking Eq. (1) into account. The variable y∗1i is the unobserved latent variable and ei ∼ N(0, s 2 ). Further, the vector x1i comprises the firm-, sector-, or region-specific control variables. All the variables on the right-hand side of Eq. (2) are taken to be exogenous. In IV-Tobit estimation, the variables y2i are endogenous and we consider a third equation:   y2i = x1i v1j + x2i v2j + 1i ,

(3)

where the variables x2i are instruments for y2i , while 1 i is presumed to be i.i.d. with E(e|x2 ) = 0. In both Tobit and IV-Tobit estimation, y2i measures foreign presence, or rather the reciprocal of it in the form of either FP via NFA or FP via RDA, but lagged by one period in the Tobit model and current values in the IV-Tobit model. x1i is the exogenous control variable vector, including total employment, location, and the share of industrial output to control for the possible fixed effects of the plant, sector, and regional levels. The vector x2i comprises IVs for the current values of both FP via NFA and FP via RDA in the IVTobit estimation to tackle the endogeneity of foreign presence in the current period on the privatization decision. In addition, the empirical analysis is carried out at the cross-sectional plant level. Thus, we have to specify that the standard errors allow for intra-group correlation, relaxing the usual requirement that individual plants are independent. That is, individual plants are independent across groups (e.g., across diversified sectors and regions), but not necessarily within a region or sector. We specify to which group, including regions and sectors, each individual firm belongs to further control for the unobservable effect among different individuals. 4. Data, variables, and summary statistics 4.1. Data and sample Our main data are taken from ESWB (Word Bank, 2005), which provides comprehensive firm-level data. This is a crosssectional survey dataset of more than 12,000 firms in China in 2004. The firms are located in over 100 Chinese cities (not all metropolitan cities) and represent various sectors. Table 13 in Appendix A lists the sectors included in the survey along with the unique sector ID. As shown in Table 13, all the sectors included in the following empirical test are manufacturing industries; the financial sector is excluded. The relationship between the privatization decision and foreign presence might vary in accordance with the distinctive sectoral features. Therefore, accurately defining the sectors by using an empirical test helps investigate the causal relationship between the privatization decision and foreign presence. We also use regional- and sector-level data on each specified individual plant such as geographical location and sector industry output, as taken from the Chinese Statistical Yearbook 2005 (National Bureau of Statistics of China, 2005). We first discuss how we combine the firm-level data (Word Bank, 2005) with the regional and sector-level data (National Bureau of Statistics of China, 2005). The city in which the firm is located and the relevant sector are available from ESWB (Word Bank, 2005). Then, we match the regional (provincial) data and sector-level data in the main dataset in terms of the firm’s physical location provided by ESWB (Word Bank, 2005).

6 Roberts and Saeed (2012) and Bortolotti et al. (2002) took lags of the vector of country-specific explanatory variables in their Tobit estimations to avoid endogeneity between economic/political/institutional factors and privatization. Similarly, Sprenger (2011) also took the lag of firm quality to deal with the endogeneity issue. Clarke and Cull (2002) exploited a discrete time hazard estimation in the Probit model to control for endogeneity. Ng et al. (2009) developed a Granger causality test between state ownership and performance to identify simultaneity. Li (2003) adopted two-stage least squares (2SLS) methods to deal with the potential two-way causality between the hardness of the government’s budget constraint and the privatization decision.

Y. Liu et al. / China Economic Review 41 (2016) 196–221

201

According to the Chinese Ministry of Commerce: “For a foreign funded enterprise established after takeover by a foreign investor, if the foreign investor’s proportion of investments exceeds 25% of the registered capital of this enterprise, this enterprise shall be entitled to enjoy the treatments to foreign funded enterprises...” Accordingly, we define a foreign firm in the dataset as having more than a 25% share ownership by a foreign investor. In the following empirical tests, we exclude 2519 foreign firms from the sample, as we are only interested in the privatization of domestic Chinese enterprises (see also the footnote to Table 13 in Appendix A). However, for the construction of the independent variables of foreign presence, we use information on foreign firms. Another important contribution of the foreign firm subset is that we adopt it to develop IVs. These include the average GM’s schooling years across all foreign plants located in the same city as the local firm and belonging to the firm’s sector (foreign GM schooling mean) and the average ratio of staff regularly using computers across all foreign firms located in the same region as the local firm and belonging to the firm’s industry (percentage using computers mean). 4.2. Variable construction 4.2.1. Measuring privatization To fully examine the effect of firm-level foreign presence on the privatization decision, we follow Boubakri et al. (2011), Boubakri, Cosset, and Saffar (2013) and John, Litov, and Yeung (2008) to choose the percentage of the state ownership of a firm (State Ownership) as the proxy for the firm-level privatization decision, the main dependent variable that lies between 0% and 100%. However, because of the limitation of using a cross-sectional dataset, we cannot depict the change of state ownership across sequential years. Thus, we adopt a micro-level registration type (Registration Status) to control for the initial corporate status and to address the impact of foreign presence on the privatization decision. It is worth noting that decreasing state ownership (or increasing private ownership) is not equivalent to privatization because many privately owned firms start as private firms rather than as privatized SOEs. However, for political, economic, and ideological reasons, “privatization” has been taboo in China. Nonetheless, de facto privatization has been underway since the mid-1980s, under the name of “Shareholding System Reform.” When this reform was first proposed in the mid-1980s, an era of the planned economy in China, more than 90% of GDP was created by SOEs. Therefore, this is the Chinese route of privatization, in the sense that it provides a framework through which assets are transferred from the state into private hands (Ma, 2008). Thus, the large proportion of private ownership in China could come from privatization. Hence, we posit that decreasing state ownership (or increasing private ownership) is somewhat equivalent to privatization. 4.2.2. Measuring foreign presence Unlike previous studies that have measured foreign presence either as a ratio of employment/output in foreign-owned firms as a share of total employment/output within the same sector and region (Haller, 2014; Wang & Chen, 2014) or as a proportion of the number of foreign firms to total firms within an industry (Anwar & Sun, 2013), our three key explanatory variables, defined below, represent the extent of foreign presence in a sector and aim to examine plant-level foreign presence objectively. 4.2.2.1. Foreign presence via NFA . Previous studies have examined the significance of firm size on the employment consequences of the privatization decision.7 In this study, we test whether average foreign firm size in terms of NFA scale, relative to the size of the individual local firm and spurred by the relative average NFA scale of the local cohort in the same sector, could be an important determinant of the privatization decision. Therefore, we develop FP via NFA, comprising two natural logarithm ratios of average NFA from both the foreign cohort and the local cohort, relative to the specific observed local plant, to proxy for the relative foreign production scale with respect to that of local firm(s). Plant-level FP via NFA depends not only on the average scale of NFA across all foreign plants in that sector, but also on the relative degree compared with the specific observed local plant. In addition, plant-level FP via NFA may rely on the average production scale of the local cohort in the sector, which would spur or prevent the entry of foreign plants and thus affect foreign presence. The average production scale of the local cohort affects foreign entry in two contrasting ways. First, a large average scale of the local cohort indicates the prosperous development of the industry, which may ensure the profitability of foreign entry and thus spur foreign presence. Second, the large average scale of the local cohort may also indicate the presence of monopolists or oligopolists in the industrial local cohort, preventing foreign entry and reducing foreign presence. To verify the effect that dominates in our case, we examine the correlation between the average NFA of the foreign cohort and that of the local cohort. We find that these are positively correlated, indicating that the spurring effect dominates. A large NFA of the specified observed local plant, on the contrary, may signal inefficiency in production. Therefore, the study also investigates and controls for the role of foreign presence in the form of relative labor productivity (ULC) as well as the interactions between the two. In addition, we control for total employment in our empirical estimations. FP via NFA is a firm-specific variable, as the denominators, in the two natural logarithm ratios, are the firm’s net fixed assets. Here, we measure this by comparing the average net fixed assets owned by the foreign cohort relative to the local firm with those owned by the local cohort. Thus, an increasing FP via NFA for local firm i indicates that the plant-level foreign presence

7 See, for example, Bhaskar and Khan (1995), Bortolotti et al. (2002), Dewenter and Malatesta (2001), Frydman, Gray, Hessel, and Rapaczynski (1999), Porta and Lopez-de-Silane (1999), McKenzie, Mookherjee, Castañeda, and Saavedra (2003), Okten and Arin (2002) and Omran (2004).

202

Y. Liu et al. / China Economic Review 41 (2016) 196–221

of firm i is increasing, which signals that firmi is a small firm compared with both the foreign cohort and the local cohort in the sector. ⎞ ⎛ n 1 NFA of the foreign firmf in sectorx ⎟ ⎜ n f =1 ⎟ ⎜ Foreign Presence via NFAi = ln ⎜ ⎟ ⎝ NFA of the local firmi in the same sector ⎠ ⎛

k

1 k

⎞ NFA of the local firmj in sectorx

⎟ ⎜ ⎟ ⎜ j=1 ⎟ + ln ⎜ ⎜ NFA of the local firm in the same sector ⎟ , where i = j i ⎠ ⎝

(4)

4.2.2.2. Foreign presence via RDA. To investigate the influence of foreign presence on privatization in relation to R&D exploration and product marketing, we take the calculation of international competitiveness proposed by Fagerberg (1988) as a reference point and develop the variable FP via RDA:

Foreign Presence via RDAi =

 std(RG) std(AG) ∗ AGi + ∗ RGi ∗ 100 std(RG) + std(AG) std(RG) + std(AG)

where  RGi =

RDf TSf



RDi − TSi

 AGi =

Af TSf

 −

Ai , TSi

Updated from Fagerberg (1988), RGi (AGi ) is the average value of the foreign firm’s R&D expenditure (advertisement fee) over the corresponding total sales revenue minus the domestic firm’s R&D expenditure (advertisement fee) over its total sales revenue in the same sector. std(RG) (std(AG)) is the standard deviation of RGi (AGi ), where i is the observation of all domestic firms and f is the observation of all foreign firms. FP via RDA measures relative average expenditure on R&D and product marketing by foreign firms compared with that of local firms in the same sector. Thus, an increase in the FP via RDA of a local plant signals an increase in foreign presence for that local plant, from the perspective of R&D expenditure and product advertisement fee. 4.2.2.3. Lagged foreign presence via ULC . Ehrlich, Gallais-Hamonno, Liu, and Lutter (1994) argued that “causality goes from ownership to productivity, and not vice versa.” However, large unproductive firms may become nationalized because of deteriorating unemployment in the Chinese labor market. Additionally, a vast number of articles control for labor productivity when studying privatization.8 For example, Frydman et al. (1999) used revenue per employee as a proxy for firm performance. This evidence supports the fact that productivity (measured by unit labor cost) is one perspective of foreign presence in the privatization issue. In addition, Liu and Wang (2003) and Waldkirch and Ofosu (2010) considered micro-level FDI inflows when computing plant-level total factor productivity, suggesting the existence of an interaction between productive efficiency and other foreign presence measures such as FP via NFA and FP via RDA. Thus, we implement the lagged average unit labor cost of the foreign cohort in a sector relative to that of local plants in the same sector as a proxy for plant-level FP via ULC. Note that we develop both current (for the IV-Tobit model) and lagged (for the Tobit model) FP via NFA and FP via RDA, but only lagged FP via ULC, used in both the IV-Tobit and the Tobit model, to eliminate the endogeneity caused by productive efficiency. FP via ULC is measured as the ratio of the lagged average unit labor cost from the foreign cohort in a sector to the lagged local firm’s unit labor cost in the same sector. Therefore, a rise in FP via ULCi indicates an increase in foreign presence for local firms, as measured by productive efficiency: ⎛

n

2003

⎞ labor cost of the foreign firmf in sectorx

⎜ f =1 t=2002 ⎟ ⎜ ⎟ 2003 n ⎜ ⎟ annual total profit of firm in sector x f ⎜ ⎟ ⎜ f =1 t=2002 ⎟ 9 Foreign Presence via ULCi = log ⎜ 2003 ⎟, ⎜ ⎟ labor cost of the local firm in the same sector ⎜ ⎟ i ⎜ t=2002 ⎟ ⎝ 2003 ⎠ t=2002

8

annual total profit of firmi in the same sector

See Chirwa (2004), Estache (2003), Gupta (2005), Megginson and Netter (2001), and Okten and Arin (2002). Most of the questions surveyed in ESWB (Word Bank, 2005) were answered for fiscal year 2004; thus, the current FP via NFA and FP via RDA values are measured in 2004. However, those questions in the Investment Climate Survey of ESWB (Word Bank, 2005) are available from 2002 to 2004, which makes possible the computation of lagged FP via NFA, FP via RDA, and FP via ULC. 9

Y. Liu et al. / China Economic Review 41 (2016) 196–221

203

4.2.3. IV approach To address our endogeneity concerns, especially those not specifically identified previously in the IV-Tobit model, we employ an IV approach similar to that of Hochberg and Lindsey (2010) and Chang et al. (2015). In particular, we use IVs correlated with FP via NFA and FP via RDA, but unrelated to the plant-level privatization decision. We develop two IVs to proxy for FP via NFA: average GM’s schooling years across all foreign firms located in the same city and belonging to the same sector (foreign GM schooling mean) and that of local firms except the local planti located in the same city and belonging to the same industry as firmi (local GM schooling mean). The IV defined for FP via RDA is the average ratio of staff regularly using computers across all foreign firms located in the same region as the local firm and belonging to the firm’s industry (percentage using computers mean). Not only does the selection of instruments satisfy the relevance and exclusion criteria, but also it is not correlated with the common factors causing both the privatization decision and foreign presence. See also Section 5.2 of the main results for the details of the construction and qualification of these IVs.10 4.2.4. Control variables 4.2.4.1. Registration status. Registration status is a categorized variable, with nine groups depicting the ownership change across all local individual plants in terms of foreign presence. We expect that the firm-level privatization decision is highly correlated with the registration status of that firm. A firm registered as a state firm is more likely to be state-owned or jointly owned by state and private investors than a foreign invested enterprise (FIE) after a few years of cooperation. In addition, registration status also corresponds to plant-level foreign presence. For example, the plant-level foreign presence of a state-owned monopolist is much lower than that of a private local corporation. ⎧ ⎪ 1, if the firm is state-owned; ⎪ ⎪ ⎪ ⎪ ⎪ 2, if the firm is collectively owned; ⎪ ⎪ ⎪ ⎪ ⎪ 3, if the firm is a jointly owned unit; ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨4, if the firm is limited liability corporation; Registration Status = 5, if the firm is a shareholding corporation; ⎪ ⎪ ⎪ ⎪ 6, if the firm is privately owned; ⎪ ⎪ ⎪ ⎪ ⎪ 7, if the firm is an enterprise invested by HK, Macau, and Taiwan; ⎪ ⎪ ⎪ ⎪ ⎪ 8, if the firm is an FIE; ⎪ ⎪ ⎪ ⎩ 9, if the firm is another type.

4.2.4.2. Total employment. Total employment measures the number of employees in each local plant. Note that, in line with assets, total firm employment is commonly used as an alternative measure of plant size, corresponding to firm-level net fixed assets. Further, SOEs often undertake social responsibilities to release local unemployment pressure; therefore, we would expect a lower level of privatization in firms with high employment (see also footnote 4). 4.2.4.3. Share of exports. Share of exports is a firm-specific ratio describing the percentage of firm-level export volume compared with total output. We control for this variable in empirical estimations since being internationally competitive might indicate a large export volume, which may also signal high plant-level foreign presence and influence the privatization decision.11 4.2.4.4. Location. Location is a dummy variable that,  1, if the firm is in the Open Coastal City (OCC) or Special Economic Zone (SEZ) Location = 0, otherwise Fig. 1 shows that firms in SEZ or OCC (with icons) are close to the coastal line of the East China Sea and South China Sea (except Kashgar in northwest China), which are expected to face lower transportation costs for international trade. In fact, the state adopts preferential policies in SEZ and OCC, such as the reduction and elimination of customs duties and income tax, plus less restricted censorship on the entry of new businesses in order to develop the foreign-oriented economy and generate foreign exchanges by exporting products and importing advanced technologies. Primarily geared to exports and imports, the SEZ and OCC are foreign-oriented areas that integrate advanced scientific and industrial technologies with trade, thereby benefiting from the tilt of state policies and up-to-date managerial systems. Thus, compared with their counterparts in inland Chinese cities, the plants in SEZ and OCC could receive customs duty and income tax concessions and are less restricted in the entry of an industry, leading to a competitive and attractive market environment for private local and foreign entrepreneurship.

10 Note that we adopt lagged FP via ULC throughout the empirical tests to alleviate the endogeneity concern that arises between plant-level productivity efficiency and the privatization decision. There is no specific instrument for FP via ULC, only its lagged term. 11 See also Megginson and Netter (2001) and Zinnes, Eilat, and Sachs (2001).

204

Y. Liu et al. / China Economic Review 41 (2016) 196–221

Fig. 1. Allocation of Chinese coastal cities and SEZs. Note: Source: Open news released by the central government of P.R. China, collated by the authors. The OCCs are Dalian, Qinhuangdao, Tianjin, Yantai, Qingdao, Lianyungang, Nantong, Shanghai, Ningbo, Wenzhou, Fuzhou, Guangzhou, Zhanjiang, and Behai. The SEZs are Shenzhen, Zhuhai, Shantou, Xiamen, Kashgar, and Hainan (province). The cities/regions with icons are either coastal cities or SEZs. There is only one SEZ, located in the northwest part of China, Kashgar. The purpose of establishing this SEZ in northwestern China is to promote western expansion in the country.

4.2.4.5. Share of industrial output.

Share of Industrial Output =

sectoral output ∗ 100% sum of all sectoral outputs

Share of industrial output is a sector-level variable controlling for the sector effect. As pointed by Dewenter and Malatesta (2001) and Frydman et al. (1999), since the industry specification is closely correlated with corporate-level performance, labor, and technology intensity, it can be used to examine the effect of privatization. In addition, with everything else equal, firms in large sectors such as petroleum processing and coking and transportation equipment are less likely to be privatized, since these sectors are strategic for the national economy in China. We do not argue that all plants in these sectors with a high percentage of industrial output are SOEs. Nonetheless, we control for the potential factors that may affect both foreign presence and the privatization decision. 4.2.4.6. Sector City ID. At first glance, sector city ID seems to partially repeat the function of the following variables, namely the location and share of industrial output, which controls for the two-dimensional effect by combining both the regional effect (location) and the sectoral effect (share of industrial output). However, to mitigate the threat of the statistical association between privatization and foreign presence because of the nature of sector or of region and to avoid the failure to control for the share of industrial output if two or more manufacturing industries contribute a similar GDP in a specific period through industrial proximity, the inter-regional and inter-sectoral effects need to be emphasized simultaneously. Foreign presence may reflect the openness of the sector/region, and such a sector/region may have higher private ownership. In particular, a sector/region with specific features may attract foreign plants as well as encourage privatization. For instance, the nature of a city suitable for agricultural cultivation and that of a sector specialized in agriculture and its byproduct might be the cause of the lower level of statistical association between foreign presence and privatization because of the feature of the sector and region, but not a causal relationship.

Y. Liu et al. / China Economic Review 41 (2016) 196–221

205

4.3. Descriptive statistics Except for the share of industrial output and sector/city information, which are taken from the Chinese Statistical Yearbook (CSY, National Bureau of Statistics of China, 2005), all the other variables including the instruments are obtained from ESWB (Word Bank, 2005). Note that we implement the same methodology to develop the indices for different periods. For example, FP via NFA (in 2004) and lagged FP via NFA (in 2002 and 2003) are calculated by using the same methodology; hence, we only describe the construction of one of these below. The same applies to current FP via RDA and lagged FP via RDA. Table 2 provides the summary statistics for the variables. The largest firm in our sample has 730.88 thousand employees, whereas the smallest plant has just six employees. For a specific region and sector, no foreign plant existed or was surveyed. Thus, we can see the missing data in the IVs of foreign GM schooling mean and percentage using computers mean. Since ownership structure and lagged FP via TE are the variables involved in the robustness test, they are discussed in the section describing the robustness test. Meanwhile, ownership structure is a categorical variable with four ownership types. The construction of lagged FP via TE is similar to that of FP via NFA, but using data on 2002 and 2003 to avoid the endogeneity issue. The surveyed local plants, belonging to 30 industrial sectors, are randomly located in almost 123 Chinese cities. As a result, we generate a unique ID (i.e., sector city ID) for those plants located in the same city and belonging to the same sector to capture both the regional effect and the sector effect. Therefore, the minimum and maximum values of sector city ID are not magnitudes, but simply IDs representing the diversified combination of 123 cities and 30 sectors. See also Table 14 in Appendix B for an additional description of the variables. Note that the variable, ownership structure, and lagged FP via TE are alternative measures for the micro-level privatization decision and foreign presence in the robustness test. 5. Main results 5.1. Causality inspection The main contribution of this study is to investigate the causal relationship between foreign firm presence and privatization in the local market in China by using the IV approach to avoid the endogeneity issue. However, we admit that the correlation between foreign presence and privatization, rather than being via causality, may also be caused by the set of common factors, which we identify in this section. The set of potential common factors affecting both the presence of foreign firms and the plant-level privatization decision in local firms are as follows: • • • •

WTO trade agreement; positive spillover effect from the presence of foreign firms to the presence of domestic private firms; local firm’s location; local firm’s registration status;

Table 2 Summary statistics. Variables

Obs.

Mean

Std. dev.

Min.

Max.

State Ownership Ownership Structure FP via NFA Lagged FP via NFA FP via RDA Lagged FP via RDA Lagged FP via ULC Lagged FP via TE Registration Status Sector City ID Total Employment Location Share of Industrial Output Share of Exports IV1: Local GM Schooling Mean IV2: Foreign GM Schooling Mean IV3: Percentage using Computers Mean

9882 9880 9867 9867 9879 9873 7364 9879 9882 9882 9882 9840 9882 9881 9688 6280 6280

15.53 3.03 5.01 4.95 0.15 1.12 −1.56 2.18 4.00 16,414.36 0.92 0.11 0.05 9.73 15.22 15.92 23.08

34.17 1.09 4.52 4.51 2.99 10.01 1.86 2.30 1.62 8316.05 7.89 0.31 0.02 23.9 1.06 1.27 18.39

0 1 −14.97 −15.18 −61.59 −779.43 −11.87 −8.73 1.00 1002 0.006 0 0.0003 0 8.00 9.00 0.00

100 4 26.97 26.22 6.62 11.43 5.31 10.57 9.00 31,080 730.88 1 0.12 100 18.00 18.00 100.00

Note: Data source: ESWB (Word Bank, 2005) and Chinese Statistical Yearbook (National Bureau of Statistics of China, 2005), constructed by the authors. Refer to Appendix B for variable definition and construction. State ownership is the measure of the privatization decision, which is the dependent variable in the empirical model. However, ownership structure is also a dependent variable to proxy for the privatization decision, but in the robustness test, FP via NFA and FP via RDA are the key explanatory variables. Lagged FP via NFA, lagged FP via RDA, lagged FP via ULC, and lagged FP via TE are included to solve the endogeneity issue. The firm-level control variables are registration status, total employment, and the share of exports. The region-level control variable is the location dummy. The sector-level control variable is the share of industrial output. The two-dimensional control variable at the sector and regional level is sector city ID. Local GM schooling mean, foreign GM schooling mean, and percentage using computers mean are the IVs, which are used to address endogeneity instead of the lagged explanatory variables.

206

Y. Liu et al. / China Economic Review 41 (2016) 196–221

• local firm size; and • features of the local firm sector. In the previous section, the local firm’s location, registration status, size, and sector features were discussed as the control variables that influence both foreign presence and privatization. The WTO trade agreement and positive spillover effect from the presence of foreign firms to the presence of domestic private firms is investigated in this section. 5.1.1. WTO trade agreement When adopting cross-sectional data, the time variation has already been eliminated. If the trade agreement is a driver of both foreign presence and privatization, we must control for it as a policy dummy in the empirical test. Indeed, the time trend dummy (or the policy dummy) would be ineffective in the model given the adoption of the cross-sectional data of 2004 (surveyed in 2005) in the empirical model. However, we still detect a significantly positive relationship between foreign presence and the privatization decision by using the cross-sectional data, suggesting that plant-level foreign presence accelerates the privatization process after eliminating the policy effect. Additionally, when referring to the Right to Trade, the fifth provision in the Protocol on the Accession of the People’s Republic of China runs as follows: “...Without prejudice to China’s right to regulated trade in a manner consistent with the WTO Agreement, China shall progressively liberalize the availability and scope of the right to trade, so that, within three years after accession, all enterprises in China shall have the right to trade in all goods throughout the customs territory of China, except for those goods listed in Annex 2A which continue to be subject to state trading in accordance with this protocol...” The Right to Trade came into effect on December 11, 2004. We posit that most sectors including manufacturing industries and the Chinese government are selfpreserving. Thus, the impact of the Right to Trade began from 2005. Since the cross-sectional dataset adopted in the present study is based on data on 2004 (i.e., surveyed in 2005), the trade agreement was not a significant driver of both foreign presence and privatization during 2004. 5.1.2. Positive spillover effect from the presence of foreign firms to the presence of domestic private firms This positive spillover effect, which may increase the presence of local private firms, aligns with the higher micro level of foreign presence, signaling an easy entry or a policy tilt in the industry or sector. If a positive spillover effect mainly exists, the causal relationship between foreign presence and the privatization decision described herein might be disturbed. However, as Aitken and Harrison (1999) stated, “productivity in domestically owned plants declines when foreign investment increases.” This so-called market stealing effect indicates that a rise in foreign presence weakens the performance of local private plants, which decreases the number of private local firms and/or weakens the degree of privatization if the plant goes bankrupt owing to inefficient corporate performance. On the contrary, a strand of the literature examines and supports the existence of a positive technological spillover effect between FIEs and domestic owned plants, suggesting that with an increasing volume of FDI inflow (or FIE), domestic enterprises would become technologically advanced and thus efficient in productivity (Blomstrom & Sjoholm, 1999; Ito, Yashiro, Xu, Chen, & Wakasugi, 2012; Keller & Yeaple, 2009; Kokko, 1994). By investigating the empirical evidence from our data, we use the correlation between the plant-level unit labor cost in 2004 and foreign presence via net fixed assets, R&D and advertisement expenditure, and total employment in 2002 and 2003 to examine which effect dominates here, the market stealing effect or technological spillover effect (see Table 3). As shown in Table 3, micro-level foreign presence via NFA, RDA, and TE, regardless of whether it is privately owned or stateowned, rises (i.e., productivity decreases) as its unit labor cost increases. Thus, we can infer that the market stealing effect dominates the technological spillover effect in our study. Thus, the causal relationship investigated herein does not seem to be interrupted by the positive spillover effect from foreign firm presence to domestic private firms. Table 3 Correlation between productivity and foreign presence. Variables

Log of unit labor cost

Lagged FP via NFA

Lagged FP via RDA

Lagged FP via TE

Log of unit labor cost Lagged FP via NFA Lagged FP via RDA Lagged FP via TE

Subsample: private local firms 1.00 0.61 0.05 0.59

1.00 0.03 0.78

1.00 0.01

1.00

1.00 0.06 0.84

1.00 0.06

1.00

Subsample: SOEs Log of unit labor cost Lagged FP via NFA Lagged FP via RDA Lagged FP via TE

1.00 0.64 0.15 0.62

Note: Data source: ESWB (Word Bank, 2005), constructed by the authors. Data period: 2004 for log of unit labor cost. The rest are constructed based on data in 2002 and 2003. Log of unit labor cost is the logarithm of total wage of permanent and temporary employees in a plant divided by the total profit of that plant. Refer to Appendix B for the remaining variable definition and construction. Only private local firms (state ownership=0) and SOEs (state ownership≥51%) were used to examine the correlation between the plant-level performance (i.e., unit labor cost) and foreign presence proxies.

Y. Liu et al. / China Economic Review 41 (2016) 196–221

207

Additionally, the features of China’s “One Country, Two System” have conferred several advantages on FIEs mainly to the detriment of domestic private firms. For example, there is internal coherence between actively courting FDI and restricting domestic private firms (Huang, 2003). Therefore, Chinese law inhibits the effectiveness of the positive spillover effect from foreign presence to domestic private presence in the local market in China. 5.2. Endogeneity The correlations between the dependent variable, suspected endogenous regressors, and instruments are reported in Table 4. Although Table 4 shows a strongly negative (slightly weak) association between current FP via NFA (RDA) and plant-level state ownership, a proxy for the privatization decision, the empirical results may suffer from two types of endogeneity. To address omitted variable bias, we first attempt to control for a standard set of variables that have been found by previous studies to affect both the privatization decision and plant-level foreign presence, such as firm-level registration status and total employment. The other possible endogeneity issue is reverse causality running from the privatization decision to plant-level foreign presence. In both cases, the coefficient estimates from the OLS regressions are biased and inconsistent. To emphasize these endogeneity issues, in our second strategy, we use the IV approach (in IV-Tobit and Tobit with the lagged key regressors) to alleviate any remaining endogeneity effect (Chang et al., 2015). 5.3. IV-Tobit estimation Here, we use instruments correlated with corporate-level FP via NFA and FP via RDA, but uncorrelated with both the plantlevel privatization decision and those common factors aforementioned. In detail, these are as follows: • Local GM Schooling Mean (IV1): average GM’s schooling years across all local firms except the local firmi located in the same city and belonging to the same sector as firmi ; • Foreign GM Schooling Mean (IV2): average GM’s schooling years across all foreign firms located in the same city and belonging to the same sector as local firmi ; and • Percentage Using Computers Mean (IV3): the average ratio of staff regularly using computers across all foreign firms located in the same city and belonging to the same sector as local firmi . IV1 (IV2) represents the local (foreign) GM schooling mean and IV3 the percentage using computers mean. IV1 and IV2 are instruments for FP via NFA in terms of its set-up, which comprises the absolute foreign average production scale and indirect foreign presence stimulated by the local average production scale. IV2 and IV3 are instruments for FP via RDA. When constructing these instruments, we use firms in geographical and industrial proximity because Kedia and Rajgopal (2009) and Chang et al. (2015) described the effect of local similarity. We posit that this similarity also exists within industries. Therefore, our three instruments satisfy the relevance criteria. In other words, IV1 and IV2 seem to be associated with the average local (foreign) production scale, which further affects FP via NFA. Apparently, a well-educated GM makes good strategic decisions about the production scale, such as enlarging the plant factory/production line in order to increase corporate efficiency and profitability. IV3 relates to R&D activities and product marketing. A high IV3 indicates active R&D and marketing by the foreign cohort, which thus affects FP via RDA with respect to that of the local firm. Generally, the vitality of R&D and product marketing in a plant is affected by the GM’s educational level of that firm as well. Therefore, our instruments are also likely to satisfy the exclusion criteria. That is, the average educational level of the GM or percentage of staff regularly using computers in the local or foreign cohort in the same region and industry as the local firm is unlikely to be associated with the privatization decision of the local firm. However, the absolute correlation of local GM schooling mean (IV1) and state ownership (Y) is unexpectedly higher than that of local GM schooling mean and current FP via NFA (Key X), reminding us to examine the weak instrument of IV1 later on. Additionally, we verify whether the set of instruments used in the model is uncorrelated with those common factors causing both foreign presence and privatization to vary. In this way, we can test the causal relationship between foreign presence and plant-level privatization by using the 2SLS method. Once this uncorrelated relationship is identified, we can conclude that the

Table 4 Correlation between suspected endogenous regressors and IVs. Variables

State Ownership

FP via NFA

FP via RDA

IV1

IV2

IV3

State Ownership FP via NFA FP via RDA IV1:Local GM Schooling Mean IV2:Foreign GM Schooling Mean IV3:Percentage using Computers Mean

1.00 −0.20 −0.04 0.13 0.07 0.08

1.00 0.06 −0.08 0.0002 −0.01

1.00 −0.10 −0.05 −0.10

1.00 0.21 0.31

1.00 0.27

1.00

Note: Data source: ESWB (Word Bank, 2005), constructed by the authors. Refer to Appendix B for variable definition and construction.

208

Y. Liu et al. / China Economic Review 41 (2016) 196–221

IVs affect the privatization decision only through their effect on foreign presence (i.e., not the second channel), which indicates a causal relationship between the privatization decision and foreign presence. IV1–IV3 are developed based on the regional and industrial proximity of a specific local plant, firmi . Thus, the adoption of a macro-level WTO trade policy, a potential positive spillover effect partially restricted by the Chinese legal system, firm location, firm registration status, and firm size, are uncorrelated with those IVs. Nevertheless, those instruments seem to be correlated with sector features. For instance, in high-tech industries, the average GM’s education, including both domestic and foreign cohorts, and average ratio of staff using computers in the same sector are higher than those of traditional industries. However, the present study controlled for specific sector features by using the share of industrial output as well as the sector and city dummies to keep the industrial-level characteristics constant when studying the causal relationship between foreign presence and the privatization decision. We now use the IV-Tobit approach to address the endogeneity issue. To detect two-way causality, we use the wellknown SB test. Table 5 suggests rejecting the null hypothesis and therefore confirms the presence of endogeneity. The J-test confidence interval with the entire grid suggests the validity of the instruments and absence of the weak instrument

Table 5 IV-Tobit-correcting endogeneity with IVs. Dept. var.: State Ownership Independent variable

FP via NFA IV1: Local GM Schooling Mean IV2: Foreign GM Schooling Mean

(1) Panel A: FP via NFA −5.83 (1.50)*** −0.45 (0.09)*** 0.005 (0.05)

Lagged FP via ULC Registration Status Sector City ID Other Controls SB Test p-Value J-Test Confidence Set Amemiya Lee Newey Test p-value Number of Obs.

−9.54 (0.62)*** 0.001 (0.00)*** No 0.00 Null set 0.01 6222

(2)

(3)

−5.37 (1.10)*** −0.54 (0.09)*** −0.02 (0.05) −4.17 (0.74)*** −7.55 (0.61)*** 0.0004 (0.00)*** No 0.00 Entire grid 0.13 4717

−4.64 (1.13)*** −0.51 (0.09)*** −0.01 (0.05) −3.68 (0.72)*** −7.68 (0.57)*** 0.0003 (0.00)*** Yes 0.00 Entire grid 0.16 4701

−9.32 (3.01)*** −0.04 (0.03) −0.01 (0.004)*** −1.91 (0.45)*** −8.93 (0.55)*** −0.00001 (0.00) No 0.00 Entire grid 0.48 4756

−8.71 (3.10)*** −0.03 (0.03) −0.01 (0.003)*** −1.88 (0.40)*** −8.68 (0.54)*** −0.0001 (0.00) Yes 0.00 Entire grid 0.61 4740

Panel B: FP via RDA FP via RDA IV2: Foreign GM Schooling Mean IV3: Percentage using Computers Mean

−7.52 (2.13)*** −0.05 (0.03)* −0.02 (0.003)***

Lagged FP via ULC Registration Status Sector City ID Other Controls SB Test p-Value J-Test Confidence Set Amemiya Lee Newey Test p-Value Number of Obs.

−10.85 (0.50)*** 0.00001 (0.00) No 0.00 Entire grid 0.56 6280

Note: Data source: ESWB (Word Bank, 2005) and Chinese Statistical Yearbook (National Bureau of Statistics of China, 2005), constructed by the authors. Refer to Appendix B for variable definition and construction. The dependent variable is the percentage of state ownership. Standard errors are in parentheses and *, **, and *** indicate significance at the 10%, 5%, and 1% levels. Panels A and B provide the empirical results by including FP via NFA and FP via RDA to measure foreign presence, respectively. The p-value of the SB test suggests endogeneity between foreign presence in the current period and the privatization decision. The J-test confidence interval with the entire grid suggests the validity of the instruments and absence of the weak instrument problem. The Amemiya Lee Newey test is adopted for the over-identification restriction. The other control variables include total employment, share of industrial output, share of exports, and location. Columns 1 and 2 in both Panels A and B do not include the other control variables, but just plant-level registration status and sector city ID, whereas column 3 in both Panels A and B does. The other control variables include the share of industrial output, the share of exports, and location. The standard errors are robust by allowing for intra-group (group of enterprises in the same sector and city) correlation, relaxing the usual requirement that observations must be independent.

Y. Liu et al. / China Economic Review 41 (2016) 196–221

209

problem. The Amemiya Lee Newey test is adopted for the over-identification restriction.12 Thus, the test results reported in Table 5, except those in column 1 of Panel A, signal that the instruments explain the dependent variable mainly through the suspected endogenous regressors and that the key structural parameters are effectively estimated at the 5% significance level. The results obtained by using the IV approach with the IV-Tobit method in the framework of the 2SLS regressions are reported in Table 5. The coefficients of the instruments in Table 5 are estimated from the reduced form of the IV-Tobit approach (Eq. (3)) and the rest are estimated from the structural equation (Eq. (2)). In Table 5, the local GM’s schooling mean is significantly and negatively related to FP via NFA, indicating that the GM’s educational level of the local plant is positively affected by the local cohort. Moreover, a high educational level for the GM in the local firm decreases plant-level foreign presence from the aspect of production scale. By the same token, the percentage of staff frequently using computers affects FP via RDA significantly and negatively. We only display the estimates of registration status, sector city ID, and omit the remaining control variables to address the ownership change effect caused by foreign presence. Note that lagged FP via ULC is included in the empirical estimation in both the IV-Tobit and the Tobit models with its lagged term, since productive efficiency is a crucial omitted variable associated with the privatization decision and foreign presence. Second, FP via ULC, a proxy of foreign presence measured from productive efficiency, also has the reverse causality effect running from privatization to corporate performance, suggesting that the government might privatize those inefficient firms to reduce the fiscal loss. Additionally, there already exist two suspected endogenous regressors, current FP via NFA and current FP via RDA. The affiliation of current FP via ULC would worsen and complicate the endogeneity issue. Therefore, we choose to adopt lagged FP via ULC in all the empirical estimations to simplify and thus tackle the endogeneity concern. The coefficients for FP via NFA (RDA) are statistically negative, indicating that increasing corporate-level foreign presence raises the chance of decreasing the state ownership of that domestic plant (or the so-called privatization of that domestic firm). Recall that decreasing state ownership (or increasing private ownership) is not equivalent to privatization because many privately owned firms start as private firms rather than as privatized SOEs. However, the Chinese route of privatization provides a framework through which assets are transferred from the state into private hands (Ma, 2008). Thus, the large proportion of private ownership in China could come from privatization. Hence, decreasing state ownership (or increasing private ownership) is somewhat equivalent to privatization. Nonetheless, this study carries out a robustness test by removing those plants registered as private firms and re-verifying the causality between foreign presence and the privatization decision among state-owned, collectively owned, and jointly owned domestic firms in China. From column 1 in Panel A, we find that the marginal effect of FP via NFA on state ownership is about 5.83% and the same figure for current FP via RDA is approximately 7.52% (see column 1 in Panel B). This marginal effect of FP via RDA on state ownership turns out to be the highest when lagged FP via ULC (i.e., the proxy for productive efficiency) is controlled for (see also column 2 in Panel B). In addition, lagged FP via ULC negatively affects firm-level state ownership. The statistically negative estimates of registration status on the privatization decision expectedly suggest that a firm registered as a SOE is least likely to be privatized compared with the other types of plants. Although the coefficients of sector city ID cannot be explained as a marginal effect, the control of regional and sectoral effect could extensively avoid the endogeneity issue from the omitted variables. The other control variables, not presented in Table 5, also have mostly significant coefficients and the expected signs. 5.4. Tobit model with lagged explanatory variables In this subsection, we interpret the 2SLS result as strong evidence for a causal effect of micro-level foreign presence on the privatization decision because the instruments are far from perfect. For instance, if the local plant is large in production (NFA), active in innovation and marketing (RDA), and efficient in corporate performance (ULC), the other plants including both foreign and local firms in the same region or same industry might mimic the local firm’s recruitment of GMs or its utilization of advanced technology. Hence, those instruments developed from the average level of the local (foreign) cohort may be associated with the privatization decision of the local firm (Chang et al., 2015). However, the IV approach could give us a baseline result, thereby shedding light on the influence of plant-level foreign presence on the privatization decision. To mitigate the above concerns, we adopt the lagged terms of FP via NFA/RDA, aligned with lagged FP via ULC, to replace the current terms in the framework of the Tobit estimation and expect the same effect of plant-level foreign presence on the privatization decision. Thus, in this subsection, we attempt to emphasize the effect of foreign presence on privatization, coupled with the interaction between the three foreign presence measures (FP via NFA, RDA, and ULC) and nonlinearity

12 The null hypothesis for the SB test is H0 : the suspected variable is exogenous. If the p-value of the SB test is ¡0.05, we have to reject the null hypothesis at the 5% significance level and conclude that the suspected variable is endogenous (Smith & Blundell, 1986). The null hypothesis for the J-test is H0 : E(Zu) = 0. The entire grid of the confidence interval suggests that the exogeneity conditions of the instruments are generally satisfied, while the null set argues the opposite fact (Finlay & Magnusson, 2009). The Amemiya Lee Newey test computes the tests of over-identifying restrictions for a regression estimated by using IVs in which the number of instruments exceeds the number of regressors. That is, for an over-identified equation, these are tests of the joint null hypothesis that the excluded instruments are valid, namely uncorrelated with the error term and correctly excluded from the estimated equation (Baum, Wiggins, Stillman, & Schaffer, 2010).

210

Y. Liu et al. / China Economic Review 41 (2016) 196–221

of these measures (FP via NFA and RDA) on the plant-level privatization decision. Worth noting, too, is that Waldkirch and Ofosu (2010) and Liu and Wang (2003) included FDI inflows in the computation of total factor productivity, suggesting that foreign presence measured by FDI inflows correlates with productivity. Aitken and Harrison (1999) also stated that “...foreign presence would raise the productivity of domestically owned firms. But it can also reduce productivity of domestically owned firms, particularly in the short run...” Similarly, Sun (2014) investigated the determinants of advertising intensity at the firm level by focusing on the role of foreign entry and found that foreign entry significantly affects advertising intensity. Thus, it would be interesting to investigate the interaction of foreign presence measured by production scale (R&D and product advertising) and foreign presence measured by productive efficiency on the plant-level privatization decision. In addition, given the dual impact of foreign presence on state ownership (i.e., the market stealing effect indicates that the presence of a foreign cohort might crowd out the profit of domestic enterprises), government intervention is needed to protect local industry. On the contrary, the positive technological spillover effect on local plants caused by foreign presence could improve corporate-level performance and the productivity of local firms. Mohsni and Otchere (2014) suggested a nonlinear relationship between the government/private ownership of banks and risk-taking. Borisova and Megginson (2011) explored whether government ownership affects the cost of debt and found a non-monotonic association between state ownership and the cost of debt. All this intuitive and literature-based evidence inspires us to tentatively examine the nonlinearity between foreign presence and the privatization decision. Following Bortolotti et al. (2002) and Roberts and Saeed (2012), who considered lagged country-specific explanatory variables in a Tobit model, Table 6 presents the empirical results from the Tobit model of correcting for endogeneity by considering the lagged values of the explanatory variable FP via NFA (Panel A) and FP via RDA (Panel B). We also correct for heteroskedasticity and specify the robust standard errors allowing for an intra-group correlation, relaxing the usual requirement that individual plants are independent. We find that the lagged values of both FP via NFA and FP via RDA—coupled with FP via ULC accordingly—have significantly negative effects on an establishment’s state ownership at the 1% significance level, implying that a firm is more likely to be privatized if there is a larger foreign presence in that industry. The Chinese central government tends to privatize more firms in those industries where competition from foreign-owned firms is higher. Note that a higher value of FP via NFA and FP via RDA signifies a higher foreign presence. In both Panel A and Panel B, the coefficient of FP via ULC is statistically negative at the 1% significance level, asserting that a high foreign presence measured from productive efficiency would lead the local firm to be privatized. In other words, if the average productivity cultivated by the foreign cohort is fairly efficient in a sector, the domestic government would privatize the local plant in the same sector, as long as foreign presence measured through production scale (NFA) and R&D and product marketing expenditure (RDA) stay constant. Hence. not only does privatization lead to efficiency, but efficiency also causes privatization.13 As expected, the coefficients of plant-level registration status in Table 6 are again significantly negative, in accord with those in Table 5. It is expected that foreign presence is nonlinearly associated with the privatization decision, since the negative market stealing effect and positive technological spillover effect of foreign presence work contrary to privatization. Thus, partial privatization is found to be an optimal solution compared with complete privatization and nationalization (Matsumura, 1998). In Table 6, the nonlinear correlation between foreign presence and privatization only seems to exist when foreign presence is measured by production scale (FP via NFA, in Panel A), but not by R&D and advertising expenditure (FP via RDA, in Panel B). For instance, the coefficient of squared lagged FP via NFA in column 5 of Panel A is 0.16, suggesting that foreign presence measured in production scale is significantly and negatively correlated with privatization. This finding means that before approaching the turning point, as FP via NFA increases, the state ownership of a local plant reduces. Afterwards, sequentially increasing FP via NFA leads to micro-level nationalization. This negative correlation demonstrates the attitude of the Chinese government towards foreign presence: it allows the existence and encourages the entry of foreign presence, however only within a specific domain of plantlevel FP via NFA. As far as it grows beyond the limitation, increasing FP via NFA results in government intervention and raises state ownership. Intuitively, as expected, the production scale, NFA (R&D and marketing expenditure, RDA) is normally associated with productive efficiency, ULC. For instance, a large firm with a higher number of net fixed assets is generally productive, since it may provide its employees with sufficient production facilities. Furthermore, a plant active in innovation and product advertising, by offering advanced technology and strategic marketing policy, might be efficient in production as well. Thus, the reciprocal action of FP via NFA (RDA) and FP via ULC on the plant-level privatization decision must be further analyzed. According to the empirical estimation above, if FP via NFA increases by 1%, keeping everything else constant, the state ownership of that firm drops by 3.02% (see also column 5 in Panel A of Table 6). Nevertheless, this is not the ultimate partial effect of FP via NFA on the privatization decision. In accord with our hypothesis that FP via NFA is related to FP via ULC, increasing FP via NFA may also affect privatization through FP via ULC. Therefore, a 1% increase in lagged FP via NFA causes the state ownership of that firm to drop by 2.58%, meaning that a local firm challenged by FP via NFA also faces difficulty in competing with FP via ULC. Hence, the domestic government must reduce the state ownership of that plant to slow the rate of reduction. All these empirical facts are statistically insignificant for lagged FP via RDA (see Panel B of Table 6).

13

See footnote 1 for studies discussing how privatization affects a firm’s productive efficiency.

Y. Liu et al. / China Economic Review 41 (2016) 196–221

211

Table 6 Tobit-correcting endogeneity with lagged explanatory variables. Dept. var.: State Ownership Independent variable

Lagged FP via NFA Squared Lagged FP via NFA

(1)

(2)

Panel A: Lagged FP via NFA −1.80 −3.00 (0.09)*** (0.20)*** 0.11 (0.01)***

Lagged FP via ULC Lagged FP via NFA* Lagged FP via ULC Registration Status Sector City ID Other Controls Number of Obs. Prob>F Pseudo R2 Predicted State Ownership

−10.95 (0.37)*** 0.0002 (0.00)*** No 9867 0.00 0.04 17.13%

−10.91 (0.37)*** 0.0002 (0.00)*** No 9867 0.00 0.04 17.13%

(3)

(4)

(5)

−3.27 (0.20)*** 0.13 (0.01)*** −2.01 (0.22)***

−3.01 (0.19)*** 0.17 (0.02)*** −4.22 (0.39)*** 0.43 (0.05)*** −8.40 (0.38)*** 0.0001 (0.00)*** No 7356 0.00 0.04 14%

−3.02 (0.19)*** 0.16 (0.02)*** −4.20 (0.40)*** 0.44 (0.05)*** −8.31 (0.37)*** 0.0001 (0.00)* Yes 7326 0.00 0.04 14%

−0.29 (0.13)** 0.01 (0.01) −1.01 (0.22)*** 0.01 (0.06) −9.43 (0.40)*** 0.00004 (0.00) No 7364 0.00 0.02 14.03%

−0.33 (0.13)*** 0.001 (0.01) −0.99 (0.23)*** 0.03 (0.06) −9.36 (0.40)*** −0.00002 (0.00) Yes 7334 0.00 0.02 14.03%

−8.59 (0.38)*** 0.0001 (0.00)*** No 7356 0.00 0.03 14%

Panel B: Lagged FP via RDA Lagged FP via RDA

−0.003 (0.02)

−0.14 (0.09) −0.0002 (0.0001)*

−0.31 (0.11)*** 0.007 (0.01) −0.99 (0.22)***

−11.56 (0.38)*** 0.0001 (0.00)** No 9873 0.00 0.03 17.19%

−11.57 (0.38)*** 0.0001 (0.00)* No 9873 0.00 0.03 17.19%

−9.43 (0.40)*** 0.00004 (0.00) No 7364 0.00 0.02 14.03%

Squared Lagged FP via RDA Lagged FP via ULC Lagged FP via RDA* Lagged FP via ULC Registration Status Sector City ID Other Controls Number of Obs. Prob>F Pseudo R2 Predicted State Ownership

Note: Data source: ESWB (Word Bank, 2005) and Chinese Statistical Yearbook (National Bureau of Statistics of China, 2005), constructed by the authors. Refer to Appendix B for variable definition and construction. The dependent variable is the percentage of state ownership. Standard errors are in parentheses and *, **, and *** indicate significance at the 10%, 5%, and 1% levels. Panels A and B provide the empirical results by including lagged FP via NFA and lagged FP via RDA to measure foreign presence, respectively. The p-value of the F-test indicates that all the variables jointly significantly affect firm-level state ownership. The other control variables include total employment, the share of industrial output, the share of exports, and location. The standard errors are robust by allowing for intra-group correlation, relaxing the usual requirement that observations must be independent.

5.5. Robustness Test This study has thus far adopted cross-sectional data on the state ownership of local plants, as a proxy of the privatization decision, to explain the inter-firm and inter-sector/region variation of foreign presence on privatization, which does not change over time. Using state ownership as a proxy of the privatization decision has some shortcomings, such as confusing the definition of privatization and private ownership, since many enterprises are private firms from the start rather than privatized SOEs. However, as aforementioned, in terms of the data limitation and “Chinese route of privatization,” a large proportion of private ownership in China is transformed from privatization. Hence, to tackle the discrepancy between privatization and private ownership, the following three subsections treat this issue in terms of the adoption of a new subsample, use of a new proxy for the privatization decision, and application of new proxies for both foreign presence and the privatization decision.

5.5.1. Adoption of a new subsample First, we remove those private enterprises from the original sample to form a new state-owned subsample. Private enterprises are firms registered as limited liability corporations, shareholding corporations, privately owned firms, enterprises

212

Y. Liu et al. / China Economic Review 41 (2016) 196–221

invested in by Hong Kong, Macau, and Taiwan, FIEs, or other types. The individuals left in the subsample are state-owned, collectively owned, and jointly owned firms. Second, we re-conduct the IV-Tobit and Tobit estimation. The aim is to separate firms that privatized from state ownership to private ownership from privately owned firms that started as private firms to identify the relationship between the privatization decision (for the former firms) and foreign presence in accordance with the Chinese route of privatization. Tables 7 and 8 show the IV-Tobit and Tobit estimation results based on the subsample composed of state-owned, collectively owned, and jointly owned enterprises, at least holding state shareholdings. We compare the empirical results in Table 5 with those in Table 7 and the results in Table 6 with those in Table 8 to analyze how they change. We find that the results in Table 7 and Table 8 change little, signaling that the key causal relationship between foreign presence and the privatization decision is robust. That is, as plant-level foreign presence increases, the state ownership of enterprises declines, meaning that privatization proceeds. Nevertheless, by excluding private enterprises, the number of sample observations decreases sharply from approximately 6000 firms to only 3000. As a result, the instruments in Panel A of Table 7 turn out to be weak and invalid, and the over-identification issue appears.

Table 7 IV-Tobit-correcting endogeneity with IVs. Dept. var.: State Ownership Independent variable

FP via NFA IV1: Local GM Schooling Mean IV2: Foreign GM Schooling Mean

(1) Panel A: FP via NFA −5.05 (1.40)*** −0.65 (0.12)*** 0.02 (0.08)

Lagged FP via ULC Registration Status Sector City ID Other Controls SB Test p-Value J-Test Confidence Set Amemiya Lee Newey Test p-Value Number of Obs.

−10.53 (0.59)*** 0.0002 (0.00)* No 0.00 null set 0.003 3368

(2)

(3)

−4.74 (1.09)*** −0.77 (0.11)*** 0.04 (0.08) −4.67 (0.90)*** −8.55 (0.61)*** 0.00003 (0.00) No 0.00 null set 0.03 2539

−4.4 (1.21)*** −0.72 (0.12)*** 0.05 (0.07) −4.31 (0.90)*** −8.57 (0.58)*** −0.00003 (0.00) Yes 0.00 0.03 (p-value) 0.05 2529

−9.62 (3.45)*** −0.07 (0.04) −0.01 (0.005)*** −2.17 (0.56)*** −9.82 (0.58)*** −0.0003 (0.00)** No 0.00 entire grid 0.59 2562

−9.73 (3.78)*** −0.06 (0.04) −0.01 (0.005)*** −2.22 (0.55)*** −9.55 (0.58)*** −0.0003 (0.00)*** Yes 0.00 entire grid 0.68 2552

Panel B: FP via RDA FP via RDA IV2: Foreign GM Schooling Mean IV3: Percentage using Computers Mean

−8.4 (2.54)*** −0.09 (0.04)** −0.02 (0.01)***

Lagged FP via ULC Registration Status Sector City ID Other Controls SB Test p-Value J-Test Confidence Set Amemiya Lee Newey Test p-Value Number of Obs.

−11.65 (0.51)*** −0.0002 (0.00)* No 0.00 entire grid 0.51 3402

Note: Data source: ESWB (Word Bank, 2005) and Chinese Statistical Yearbook (National Bureau of Statistics of China, 2005), constructed by the authors. Refer to Appendix B for variable definition and construction.The dependent variable is the percentage of state ownership. Standard errors are in parentheses and *, **, and *** indicate significance at the 10%, 5%, and 1% levels. Panels A and B provide the empirical results by including FP via NFA and FP via RDA to measure foreign presence, respectively. The p-value of the SB test suggests endogeneity between foreign presence in the current period and the privatization decision. The J-test confidence interval with the entire grid suggests the validity of the instruments and absence of the weak instrument problem. The Amemiya Lee Newey test is adopted for the over-identification restriction. The other control variables include total employment, the share of industrial output, the share of exports, and location. Columns 1 and 2 in Panels A and B do not include the other control variables, just plant-level registration status and sector city ID, whereas column 3 in both Panels A and B does. The other control variables include the share of industrial output, the share of exports, and location. The standard errors are robust by allowing for intra-group correlation, relaxing the usual requirement that observations must be independent.

Y. Liu et al. / China Economic Review 41 (2016) 196–221

213

Table 8 Tobit-correcting endogeneity with lagged explanatory variables. Dept. var.: State Ownership Independent variable

(1)

Lagged FP via NFA Squared Lagged FP via NFA

(2)

Panel A: Lagged FP via NFA −2.13 −3.00 (0.12)*** (0.25)*** 0.08 (0.02)***

Lagged FP via ULC Lagged FP via NFA* Lagged FP via ULC Registration Status Sector City ID Other Controls Number of Obs. Prob>F Pseudo R2 Predicted State Ownership

−11.56 (0.38)*** 0.00001 (0.00) No 5372 0.00 0.05 25.47%

−11.53 (0.38)*** 0.000003 (0.00) No 5372 0.00 0.05 25.47%

(3)

(4)

(5)

−3.42 (0.26)*** 0.1 (0.02)*** −2.7 (0.34)***

−3.27 (0.25)*** 0.15 (0.02)*** −5.23 (0.60)*** 0.47 (0.07)*** −8.91 (0.40)*** −0.0001 (0.00) No 3967 0.00 0.05 19.99%

−3.32 (0.28)*** 0.15 (0.02)*** −5.24 (0.60)*** 0.48 (0.07)*** −8.81 (0.40)*** −0.0001 (0.00) Yes 3953 0.00 0.05 19.97%

−0.66 (0.22)*** 0.02 (0.01) −1.16 (0.35)*** −0.07 (0.10) −10.22 (0.42)*** −0.0002 (0.00)*** No 3973 0.00 0.04 20.07%

−0.52 (0.22)** 0.02 (0.01) −1.35 (0.36)*** −0.05 (0.10) −9.85 (0.42)*** −0.0002 (0.00)*** Yes 3959 0.00 0.04 20.13%

−9.16 (0.39)*** −0.0001 (0.00) No 3967 0.00 0.05 19.99%

Panel B: Lagged FP via RDA Lagged FP via RDA

−0.36 (0.16)**

−0.33 (0.17)* 0.01 (0.01)

−0.56 (0.18)*** 0.01 (0.01) −1.24 (0.35)***

−12.29 (0.39)*** −0.0001 (0.00) No 5379 0.00 0.04 25.60%

−12.29 (0.39)*** −0.0001 (0.00) No 5379 0.00 0.04 25.60%

−10.22 (0.42)*** −0.0002 (0.00)*** No 3973 0.00 0.04 20.07%

Squared Lagged FP via RDA Lagged FP via ULC Lagged FP via RDA* Lagged FP via ULC Registration Status Sector City ID Other Controls Number of Obs. Prob>F Pseudo R2 Predicted State Ownership

Note: Data source: ESWB (Word Bank, 2005) and Chinese Statistical Yearbook (National Bureau of Statistics of China, 2005), constructed by the authors. Refer to Appendix B for variable definition and construction. The dependent variable is the percentage of state ownership. Standard errors are in parentheses and *, **, and *** indicate significance at the 10%, 5%, and 1% levels. Panels A and B provide the empirical results by including lagged FP via NFA and lagged FP via RDA to measure foreign presence, respectively. The p-value of the F-test indicates that all the variables jointly significantly affect firm-level state ownership. The other control variables include total employment, the share of industrial output, the share of exports, and location. The standard errors are robust by allowing for intra-group correlation, relaxing the usual requirement that observations must be independent.

5.5.2. An alternative measure for privatization In this section, we use a new dependent variable, termed ownership structure, as a proxy of privatization, while continuing to measure foreign presence by using FP via NFA and FP via RDA, to analyze the relationship between firm-level foreign presence and the privatization decision. The particulars of the alternative measure for privatization are as follows:

Ownership

⎧ ⎪ 1, ⎪ ⎪ ⎪ ⎨2, Structure = ⎪ 3, ⎪ ⎪ ⎪ ⎩ 4,

if the firm’s ownership type is state; if the firm’s ownership type is collective; if the firm’s ownership type is corporation; if the firm’s ownership type is private.

Since ownership structure is a categorical variable with four ownership types (i.e., it takes a higher value when the degree of state ownership is lower), we employ multinomial Logit (MNL) estimation with lagged FP via NFA (Table 9) and lagged FP via RDA (Table 10) as the key explanatory variable to verify the negative impact of plant-level foreign presence on the privatization

214

Y. Liu et al. / China Economic Review 41 (2016) 196–221

Table 9 Multinomial logit — impact of FP via NFA on ownership structure. Dept. var.: Ownership Structure Independent variable

(1)

Lagged FP via NFA

State-owned −0.21 (0.01)***

(2)

(3)

(4)

(5)

−0.29 (0.02)*** 0.01 (0.002)***

−0.33 (0.02)*** 0.01 (0.002)*** −0.15 (0.03)***

−1.24 (0.04)*** 0.00003 (0.00)*** No

−1.15 (0.05)*** 0.00003 (0.00)*** No

−0.32 (0.02)*** 0.01 (0.002)*** −0.21 (0.03)*** 0.01 (0.01)*** −1.15 (0.05)*** 0.00003 (0.00)*** No

−0.31 (0.02)*** 0.01 (0.003)*** −0.2 (0.04)*** 0.01 (0.02)*** −1.14 (0.05)*** 0.00002 (0.00)*** Yes

0.02 (0.01)*

0.03 (0.03) −0.0004 (0.002)

0.002 (0.03) 0.002 (0.002) 0.05 (0.03)*

−1.07 (0.04)*** 0.00002 (0.00)*** No

−1.07 (0.04)*** 0.00002 (0.00)*** No

−1.11 (0.04)*** 0.00002 (0.00)** No

−0.002 (0.03) 0.001 (0.002) 0.16 (0.06)*** −0.02 (0.01)*** −1.12 (0.04)*** 0.00002 (0.00)** No

0.01 (0.03) 0.0001 (0.002) 0.16 (0.07)*** −0.02 (0.02)*** −1.12 (0.05)*** 0.00002 (0.00)* Yes

−0.09 (0.01)***

−0.15 (0.01)*** 0.01 (0.001)***

−0.14 (0.01)*** 0.01 (0.001)*** 0.02 (0.02)

−0.29 (0.02)*** 0.00002 (0.00)*** No

−0.3 (0.02)*** 0.00002 (0.00)*** No

−0.29 (0.02)*** 0.00002 (0.00)*** No

−0.13 (0.01)*** 0.01 (0.001)*** −0.04 (0.02) 0.01 (0.004)*** −0.29 (0.02)*** 0.00002 (0.00)*** No

−0.12 (0.02)*** 0.01 (0.001)*** −0.04 (0.03) 0.01 (0.004)*** −0.29 (0.02)*** 0.00002 (0.00)*** Yes

7354 0.00 0.18

7354 0.00 0.18

7324 0.00 0.18

Squared Lagged FP via NFA Lagged FP via ULC Lagged FP via NFA* Lagged FP via ULC Registration Status Sector City ID Other Controls

−1.23 (0.04)*** 0.00003 (0.00)*** No Collectively owned

Lagged FP via NFA Squared Lagged FP via NFA Lagged FP via ULC Lagged FP via NFA* Lagged FP via ULC Registration Status Sector City ID Other Controls

Corporation Lagged FP via NFA Squared Lagged FP via NFA Lagged FP via ULC Lagged FP via NFA* Lagged FP via ULC Registration Status Sector City ID Other Controls

Privately owned: base outcome Number of Obs. Wald test Prob>Chi2 Pseudo R2

9865 0.00 0.18

9865 0.00 0.19

Note: Data source: ESWB (Word Bank, 2005) and Chinese Statistical Yearbook (National Bureau of Statistics of China, 2005), constructed by the authors. Refer to Appendix B for variable definition and construction. The dependent variable has a categorical ownership structure. Standard errors are in parentheses and *, **, and *** indicate significance at the 10%, 5%, and 1% levels. The proxy for foreign presence in this table is lagged FP via NFA. There are four groups of empirical panels, indicating that the dependent variable has been categorized into state-owned, collectively owned, corporation, and privately owned. The latter is set as the base outcome. The p-values of the Wald test suggest that all the variables are jointly significant in affecting the ownership structure. The other control variables include the share of industrial output, the share of exports, and location. The standard errors are robust by allowing for intra-group correlation, relaxing the usual requirement that observations must be independent.

decision, while tackling the endogeneity issue by using lagged explanatory variables. The marginal effect with regard to Tables 9 and 10 is presented in Panels A and B in Table 12, respectively. This MNL estimation is based on the original sample rather than on the state-owned one adopted in Section 5.5.1. To ensure the MNL model identification, the coefficients of the category privately owned are set to zero and those of the other categories (e.g., state-owned, collectively owned, and corporations) are then interpreted with respect to the base category of

Y. Liu et al. / China Economic Review 41 (2016) 196–221

215

Table 10 Multinomial logit — impact of FP via RDA on ownership structure. Dept. var.: Ownership Structure Independent variable

(1)

Lagged FP via RDA

State-owned −0.01 (0.01)*

(2)

(3)

(4)

(5)

−0.01 (0.01) 0.001 (0.001)

−0.03 (0.01)** 0.001 (0.001) −0.04 (0.02)*

−1.23 (0.03)*** 0.00002 (0.00)*** No

−1.16 (0.04)*** 0.00002 (0.00)*** No

−0.03 (0.02)* 0.001 (0.001) −0.04 (0.02)* 0.0005 (0.01) −1.16 (0.04)*** 0.00002 (0.00)*** No

−0.02 (0.02) 0.001 (0.001) −0.05 (0.02)** 0.001 (0.01) −1.16 (0.04)*** 0.00001 (0.00) Yes

0.05 (0.02)***

0.06 (0.02)*** −0.002 (0.002)

0.10 (0.02)*** −0.01 (0.004)* 0.06 (0.03)*

−1.03 (0.03)*** 0.00003 (0.00)*** No

−1.03 (0.03)*** 0.00003 (0.00)*** No

−1.06 (0.04)*** 0.00003 (0.00)*** No

0.13 (0.03)*** −0.01 (0.003)** 0.02 (0.03) 0.02 (0.01)** −1.06 (0.04)*** 0.00003 (0.00)*** No

0.12 (0.03)*** −0.01 (0.003)** 0.02 (0.03) 0.02 (0.01)** −1.08 (0.04)*** 0.00003 (0.00)*** Yes

−0.02 (0.01)***

−0.01 (0.01)* 0.001 (0.001)**

−0.02 (0.01)* 0.001 (0.001)** 0.06 (0.02)***

−0.29 (0.02)*** 0.00001 (0.00)*** No

−0.29 (0.02)*** 0.00002 (0.00)*** No

−0.29 (0.02)*** 0.00002 (0.00)*** No

−0.02 (0.01) 0.001 (0.001)** 0.06 (0.02)*** 0.001 (0.004) −0.29 (0.02)*** 0.00002 (0.00)*** No

−0.01 (0.01) 0.002 (0.001)*** 0.05 (0.02)*** 0.002 (0.004) −0.29 (0.02)*** 0.00002 (0.00)*** Yes

7362 0.00 0.14

7362 0.00 0.14

7332 0.00 0.16

Squared Lagged FP via RDA Lagged FP via ULC Lagged FP via RDA* Lagged FP via ULC Registration Status Sector City ID Other Controls

−1.23 (0.03)*** 0.00002 (0.00)**** No Collectively owned

Lagged FP via RDA Squared Lagged FP via RDA Lagged FP via ULC Lagged FP via RDA* Lagged FP via ULC Registration Status Sector City ID Other Controls

Corporation Lagged FP via RDA Squared Lagged FP via RDA Lagged FP via ULC Lagged FP via RDA* Lagged FP via ULC Registration Status Sector City ID Other Controls

Privately owned: base outcome Number of Obs. Wald test Prob>Chi2 Pseudo R2

9871 0.00 0.15

9871 0.00 0.15

Note: Data source: ESWB (Word Bank, 2005) and Chinese Statistical Yearbook (National Bureau of Statistics of China, 2005), constructed by the authors. Refer to Appendix B for variable definition and construction. The dependent variable has a categorical ownership structure. Standard errors are in parentheses and *, **, and *** indicate significance at the 10%, 5%, and 1% levels. The proxy for foreign presence in this table is lagged FP via RDA. There are four groups of empirical panels, indicating that the dependent variable has been categorized into state-owned, collectively owned, corporation, and privately owned. The latter is set as the base outcome. The p-values of the Wald test suggest that all the variables are jointly significant in affecting the ownership structure. The other control variables include the share of industrial output, the share of exports, and location. The standard errors are robust by allowing for intra-group correlation, relaxing the usual requirement that observations must be independent.

privately owned. The model fit is ordinary with a pseudo R2 of about 18% in Table 9 and about 15% in Table 10, based on the crosssectional dataset. Additionally, the regressors are jointly statistically significant at the 5% level, because the p-values of the Wald test in both Tables 9 and 10 are 0.00. The empirical results presented in Table 9 (Table 10) signal that a one-unit increase in lagged FP via NFA (lagged FP via RDA) leads to relative odds of being an SOE rather than being a privately owned firm that are 0.72 (0.97)

216

Y. Liu et al. / China Economic Review 41 (2016) 196–221

Table 11 Multinomial logit — impact of FP via TE on ownership structure. Dept. var.: Ownership Structure Independent variable

(1)

Lagged FP via TE

State-owned −0.30 (0.02)***

(2)

(3)

(4)

(5)

−0.34 (0.02)*** 0.01 (0.004)***

−0.36 (0.02)*** 0.01 (0.004)*** −0.05 (0.03)*

−1.22 (0.04)*** 0.00002 (0.00)*** No

−1.16 (0.05)*** 0.00003 (0.00)*** No

−0.36 (0.03)*** 0.01 (0.005)*** −0.06 (0.03)** 0.01 (0.01) −1.16 (0.05)*** 0.00003 (0.00)*** No

−0.37 (0.03)*** 0.01 (0.005)*** −0.05 (0.03)* 0.01 (0.01) −1.15 (0.05)*** 0.00002 (0.00)*** Yes

0.01 (0.02)

−0.003 (0.03) 0.003 (0.005)

−0.01 (0.04) 0.01 (0.01) 0.05 (0.03)

−1.09 (0.03)*** 0.00002 (0.00)*** No

−1.07 (0.03)*** 0.00002 (0.00)*** No

−1.1 (0.04)*** 0.00002 (0.00)*** No

−0.02 (0.04) 0.01 (0.005) 0.07 (0.05) −0.01 (0.01) −1.11 (0.04)*** 0.00002 (0.00)*** No

−0.03 (0.04) 0.01 (0.005) 0.07 (0.05) −0.01 (0.01) −1.1 (0.04)*** 0.00002 (0.00)** Yes

−0.09 (0.01)***

−0.15 (0.01)*** 0.01 (0.002)***

−0.13 (0.02)*** 0.01 (0.003)*** 0.06 (0.02)***

−0.28 (0.02)*** 0.00002 (0.00)*** No

−0.29 (0.02)*** 0.00002 (0.00)*** No

−0.29 (0.02)*** 0.00002 (0.00)*** No

−0.11 (0.02)*** 0.01 (0.003)*** 0.02 (0.02) 0.01 (0.01)*** −0.28 (0.02)*** 0.00002 (0.00)*** No

−0.11 (0.02)*** 0.01 (0.003)*** 0.02 (0.02) 0.01 (0.01)*** −0.28 (0.02)*** 0.00002 (0.00)*** Yes

7362 0.00 0.17

7362 0.00 0.17

7332 0.00 0.17

Squared Lagged FP via TE Lagged FP via ULC Lagged FP via TE* Lagged FP via ULC Registration Status Sector City ID Other Controls

−1.22 (0.04)*** 0.00002 (0.00)*** No Collectively owned

Lagged FP via TE Squared Lagged FP via TE Lagged FP via ULC Lagged FP via TE* Lagged FP via ULC Registration Status Sector City ID Other Controls

Corporation Lagged FP via TE Squared Lagged FP via TE Lagged FP via ULC Lagged FP via TE* Lagged FP via ULC Registration Status Sector City ID Other Controls

Privately owned: base outcome Number of Obs. Wald test Prob>Chi2 Pseudo R2

9877 0.00 0.18

9877 0.00 0.18

Note: Data source: ESWB (Word Bank, 2005) and Chinese Statistical Yearbook (National Bureau of Statistics of China, 2005), constructed by the authors. Refer to Appendix B for variable definition and construction. The dependent variable has a categorical ownership structure. Standard errors are in parentheses and *, **, and *** indicate significance at the 10%, 5%, and 1% levels. The proxy for foreign presence in this table is lagged FP via TE. There are four groups of empirical panels, indicating that the dependent variable has been categorized into state-owned, collectively owned, corporation, and privately owned. The latter is set as the base outcome. The p-values of the Wald test suggest that all the variables are jointly significant in affecting the ownership structure. The other control variables include the share of industrial output, the share of exports, and location. The standard errors are robust by allowing for intra-group correlation, relaxing the usual requirement that observations must be independent.

times those before the change (i.e., the relative odds have declined).14 In addition, the marginal effect listed in Panel A (Panel B) of Table 12 confirms this result: a one-unit increase in foreign presence measured by lagged FP via NFA, decreases by 0.02 (0.002) the probability of being an SOE than a privately owned firm, indicating that an increase in foreign presence leads to privatization.

14

By examining column 4 of the state-owned category in Table 9 (Table 10), we calculate a relative odds ratio equal to 1/e0.32 ≈ 0.72 (1/e0.03 ≈ 0.97).

Y. Liu et al. / China Economic Review 41 (2016) 196–221

217

Table 12 Marginal effect of multinomial logit model. Dependent variable: Ownership Structure Privately owned: base outcome Independent variable

(1)

(2)

(3)

(4)

(5)

dy/dx

dy/dx

dy/dx

dy/dx

dy/dx

−0.02 (0.001)*** 0.001 (0.0001)*** −0.01 (0.002)***

−0.02 (0.001)*** 0.001 (0.0001)*** −0.01 (0.002)*** 0.001 (0.0003)** −0.06 (0.002)*** 0.000001 (0.00)*** Yes

Panel A: Lagged FP via NFA −0.01 −0.02 (0.001)*** (0.001)*** 0.001 (0.0001)***

Other Controls

−0.08 (0.003)*** 0.000002 (0.00)*** No

−0.08 (0.003)*** 0.000002 (0.00)*** No

−0.06 (0.003)*** 0.000001 (0.00)*** No

−0.02 (0.001)*** 0.001 (0.0001)*** −0.01 (0.002)*** 0.001 (0.0004)*** −0.06 (0.003)*** 0.000001 (0.00)*** No

Y=Pr(Ownership Structure=1)

0.09

0.09

0.07

0.07

0.07

−0.002 (0.001)** 0.0001 (0.0001) −0.01 (0.002)*** −0.0001 (0.0004) −0.08 (0.003)*** 0.000001 (0.00)*** No 0.09

−0.002 (0.001) 0.0001 (0.0001) −0.01 (0.002)*** −0.0001 (0.0004) −0.08 (0.003)*** 0.0000003 (0.00) Yes 0.09

−0.02 (0.002)*** 0.001 (0.0003)** −0.005 (0.002)*** 0.00004 (0.001) −0.06 (0.003)*** 0.000001 (0.00)*** No 0.07

−0.02 (0.002)*** 0.001 (0.0003)** −0.004 (0.002)*** 0.0001 (0.001) −0.06 (0.003)*** 0.000001 (0.00)*** Yes 0.07

Lagged FP via NFA Squared Lagged FP via NFA Lagged FP via ULC Lagged FP via NFA* Lagged FP via ULC Registration Status Sector City ID

Panel B: Lagged FP via RDA Lagged FP via RDA

−0.001 (0.001)*

−0.001 (0.001) 0.00003 (0.0001)

−0.002 (0.001)** 0.0001 (0.0001) −0.01 (0.002)***

−0.09 (0.003)*** 0.000001 (0.00)*** No 0.1

−0.1 (0.01)*** 0.000001 (0.00)*** No 0.1

−0.1 (0.003)*** 0.000001 (0.00)*** No 0.1

Squared Lagged FP via RDA Lagged FP via ULC Lagged FP via RDA* Lagged FP via ULC Registration Status Sector City ID Other Controls Y=Pr(Ownership Structure=1)

Panel C: Lagged FP via TE Lagged FP via TE

−0.02 (0.001)***

−0.02 (0.002)*** 0.001 (0.0003)**

−0.02 (0.001)*** 0.001 (0.0003)** −0.005 (0.002)***

−0.08 (0.003)*** 0.000001 (0.00)*** No 0.09

−0.08 (0.003)*** 0.000001 (0.00)*** No 0.09

−0.06 (0.003)*** 0.000001 (0.00)*** No 0.07

Squared Lagged FP via TE Lagged FP via ULC Lagged FP via TE* Lagged FP via ULC Registration Status Sector City ID Other Controls Y=Pr(Ownership Structure=1)

Note: Data source: ESWB (Word Bank, 2005) and Chinese Statistical Yearbook (National Bureau of Statistics of China, 2005), constructed by the authors. Refer to Appendix B for variable definition and construction. The dependent variable is the categorization of ownership structure. Standard errors are in parentheses and *, **, and *** indicate significance at the 10%, 5%, and 1% levels. (∗) dy/dx is for the discrete change in the dummy variable from 0 to 1. The other control variables include the share of industrial output, the share of exports, and location.

5.5.3. Alternative measures for foreign presence and privatization Now, we develop FP via TE according to FP via NFA, since total employment is another common measure of production scale (Haller, 2014; Wang & Chen, 2014). Plant-level foreign presence measured by production scale not only depends on the average relative scale of the foreign cohort, but is also associated with the average physical size cultivated by the local cohort in that

218

Y. Liu et al. / China Economic Review 41 (2016) 196–221

sector. In addition, we use lagged FP via TE (2002 and 2003) to examine its negative impact on ownership structure (2004) to avoid the endogeneity issue. ⎛ Foreign Presence via

1 n

n

⎞ TE of foreign firmf in sectorx

⎟ ⎜ f =1 ⎟ ⎜ TEi = ln ⎜ ⎟ ⎝ TE of local firmi within the same sector ⎠ ⎞ k 1 TE of local firm in sector x j ⎟ ⎜ k ⎟ ⎜ j=1 ⎟ + ln ⎜ ⎜ TE of local firm within the same sector ⎟ , where i = j i ⎠ ⎝ ⎛

(5)

The MNL estimation results are presented in Table 11 and the marginal effects are in Panel C of Table 12. Overall, from the p-values of the Wald test in each regression in Table 11, the regressors are jointly statistically significant at the 5% level. The category “privately owned” is the base outcome group, and the results for the other categories of state-owned, collectively owned, and corporation are presented in Table 11. Panel C of Table 12 provides the marginal probability estimation evaluated at the sample mean of the regressors by including lagged FP via TE. We once again find that the results are generally robust. The marginal probabilities evaluated at the sample mean of the regressors in the table show that a one-unit increase in FP via TE drops the probability of being an SOE rather than being a privately owned enterprise by 0.02%. A one-unit increase in FP via ULC decreases the probability of being state-owned than privately owned by approximately 0.01%. The marginal effect of FP via ULC on the ownership structure decreases when the control variables are added. Nevertheless, the interaction of FP via TE and FP via ULC on the ownership structure is no longer statistically significant.

6. Conclusion In November 2013, the central government in China pushed the idea of market-oriented privatization to the top of its agenda, prompting a preliminary round of privatization from 2001 to 2003. Since then, privatization and the development of a mixed economy have become core strategies to enhance the efficiency of state-owned capital and market vitality. Therefore, an empirical investigation using data from ESWB (Word Bank, 2005), which captures the first wave of privatization (2001–2003) in China, would highlight the real-world development of market-oriented privatization in the new millennium. Although a large number of researchers have investigated the consequences of privatization in China, few studies have examined the role of foreign presence as a determinant of the privatization decision, despite the rising global interaction between China and the international community. Our study bridges this gap. In this study, we examine whether foreign presence in a sector influences the determination of the public ownership of domestic firms in the same sector in China and discuss whether the extent of the competition activated by foreign firms in an industry is a significant determinant of privatization. With the support of the empirical evidence presented herein, the answers to the two research questions proposed in the Introduction are positive. In a market economy, foreign presence may stimulate local firms to be competitive in light of intensive market competition. Therefore, the policies of encouraging inward FDI and (partially) privatizing state-owned domestic firms seem to have interacted with each other in China. By using a large sample of firms covered by ESWB (Word Bank, 2005) and the Chinese Statistical Yearbook (2002–2004), we find that a firm facing a higher foreign presence is more likely to be privatized. We deal with the possible existence of a two-way causal relationship between state ownership and foreign presence in two ways: IV-Tobit estimation with IVs and Tobit estimation with lagged key explanatory variables. In addition, we find that a rise in foreign presence increases the extent of private ownership of domestic enterprise in that sector in a nonlinear fashion, suggesting that partial privatization is optimal under foreign competition, if foreign presence indeed represents competition. Third, we show that FP via NFA positively interacts with FP via ULC, thereby affecting the plant-level privatization decision. This fact asserts that increasing both FP via NFA and via ULC slows the original privatization progress. We also carry out a number of robustness checks including MNL estimation and its marginal effects by using a new subsample removing those private firms, a new categorical dependent variable, and a new proxy for foreign presence measured by total employment, showing that the results remain robust and in line with our basic findings. However, our results are subject to a number of important limitations. First, although we attempted to proxy for foreign presence from as many aspects as possible, some unexplained variance in foreign presence could not be depicted by production scale (NFA), R&D and advertising expenditure (RDA), and productivity efficiency (ULC). Second, we took care to interpret 2SLS in the IV-Tobit estimation result as strong evidence for a causal effect of micro-level foreign firm presence on the privatization decision, because the instruments are far from perfect. For instance, in inland regions, which lack FDI inflows, the foreign GM schooling mean and percentage using computers mean of foreign firms might be biased. To summarize, when endogeneity is a perennial issue that no empirical test can entirely rule out, we conduct the Tobit estimation with lagged explanatory variables to reconfirm and alleviate the endogeneity concerns, finding that our main conclusions hold. Thus, we anticipate that we have measured foreign firm presence from diversified points of view, while updating the main dataset for future study.

Y. Liu et al. / China Economic Review 41 (2016) 196–221

219

Appendix A Table 13 Description of the sample sectors. Sector name

Sector %

Agricultural and side line food processing Food production Beverages production Tobacco production Textiles manufacturing Garment, shoes, and caps manufacturing Leather, furs, down, and related products Timber processing, bamboo, cane, palm fiber and straw products Furniture manufacturing Paper-making and paper products Printing and record medium reproduction Cultural, educational and sports goods Petroleum processing and coking Raw chemical materials and chemical products Medical and pharmaceutical products Chemical fiber products Rubber products Plastic products Non-metal mineral products Smelting and pressing of ferrous metals Smelting and pressing of non-ferrous metals Metal products General machinery Equipment for special purposes Transportation equipment Electrical equipment and machinery Electronic and telecommunications equipment Instruments, meters, cultural and office machinery Handicraft products and other machinery Renewable materials processing Total

8.1% 2.1% 1.4% 0.4% 7.7% 1.2% 0.8% 1.2% 0.4% 2% 0.5% 0.3% 1.7% 12.4% 3.5% 0.4% 0.2% 2.2% 12% 4.4% 3.1% 2.6% 9.4% 4.3% 8.2% 6.4% 2.2% 0.3% 0.6% 0.04% 100%

Note: Data source: ESWB (Word Bank, 2005), collated and computed by the authors. Data period: 2004. The total number of firms in the sample is 9881, and these are domestic firms located in China. Financial sector is not included in the survey, all sectors listed above are manufacturing industries.

Appendix B Table 14 Variable description. Variable name

Type

Description

Source

State Ownership Ownership Structure

Dependent Dependent in robustness test Key explanatory in IV-Tobit

Percentage of governmental shareholding in a firm Firm-level ownership type

ESWB (Word Bank, 2005) ESWB (Word Bank, 2005)

Constructed based on current period average of all foreign firms’ net fixed asset in a sector relative to that of a local plant in the same sector, and current period average of all domestic firms’ net fixed asset of the same sector relative to that of a local plant in the sector Constructed based on lagged period average of all foreign firms’ net fixed asset in a sector relative to that of a local plant in the same sector, and lagged period average of all domestic firms’ net fixed asset of the same sector relative to that of a local plant in the sector Constructed based on current period average of all foreign firms’ R&D and advertising expenditure in a sector relative to that of a local plant in the same sector Constructed based on lagged period average of all foreign firms’ R&D and advertising expenditure in a sector relative to that of a local plant in the same sector Constructed based on lagged period average of all foreign firms’ unit labor cost in a sector relative to that of a local plant in the same sector

ESWB (Word Bank, 2005)

FP via NFA

Lagged FP via NFA

Key explanatory in Tobit

FP via RDA

Key explanatory in IV-Tobit

Lagged FP via RDA

Key explanatory in Tobit

Lagged FP via ULC

Key explanatory in both IV-Tobit and Tobit

ESWB (Word Bank, 2005)

ESWB (Word Bank, 2005)

ESWB (Word Bank, 2005)

ESWB (Word Bank, 2005)

(continued on next page)

220

Y. Liu et al. / China Economic Review 41 (2016) 196–221

Table 14 (continued) Variable name

Type

Description

Source

Lagged FP via TE

Key explanatory in robustness test

ESWB (Word Bank, 2005)

Registration Status

Control

Sector City ID

Control

Constructed based on lagged period average of all foreign firms’ number of employment in a sector relative to that of a local plant in the same sector A categorized variable indicating the type of a firm when registered A categorized variable controlling for both sectoral and regional effect

Total Employment

Control

Location

Control

Share of Industrial output

Control

Share of Exports

Control

Local GM Schooling Mean

IV1

Foreign GM Schooling Mean

IV2

Percentage using Computers Mean

IV3

Total number of employee, unit measurement is thousand person A dummy variable describing whether the firm is located in a coastal city or SEZs A sectoral effect control variable to differentiate industrial effect Percentage of firm-level export volume over total production Average GM’s schooling years across all foreign firms that are located in the same city and belong to the same sector Average GM’s schooling years across all local firms except local planti that are located in the same city and belong to the same sector Average ratio of staff regularly using computers across all foreign firms that are located in the same region as the local firm and belong to the firm’s industry

ESWB (Word Bank, 2005) ESWB (Word Bank, 2005) &CSY (National Bureau of Statistics of China, 2005) ESWB (Word Bank, 2005) ESWB (Word Bank, 2005) &CSY (National Bureau of Statistics of China, 2005) CSY (National Bureau of Statistics of China, 2005) ESWB (Word Bank, 2005) ESWB (Word Bank, 2005)

ESWB (Word Bank, 2005)

ESWB (Word Bank, 2005)

References Aitken, B. J., & Harrison, A. E. (1999). Do domestic firms benefit from direct foreign investment? Evidence from Venezuela. American Economic Review, 89, 605–618. Anwar, S., & Sun, S. (2013). Foreign entry and firm R&D: Evidence from Chinese manufacturing industries. R&D Management, 43, 303–317. Bai, C. E., Lu, J., & Tao, Z. (2009). How does privatization work in China? Journal of Comparative Economics, 37, 453–470. Baum, C. F., Wiggins, V., Stillman, S., & Schaffer, M. E. (2010). OVERID: Stata module to calculate tests of overidentifying restrictions after ivreg, ivreg2, ivreg29, ivprobit, ivtobit, reg3. Statistical Software Components. Bhaskar, V., & Khan, M. (1995). Privatization and employment: A study of the jute industry in Bangladesh. The American Economic Review, 85, 267–273. Blomstrom, M., & Sjoholm, F. (1999). Technology transfer and spillovers: Does local participation with multinationals matter? European Economic Review, 43, 915–923. Borisova, G., & Megginson, W. L. (2011). Does government ownership affect the cost of debt? Evidence from privatization. Review of Financial Studies, 24, 2693–2737. Bortolotti, B., D’Souza, J., Fantini, M., & Megginson, W. L. (2002). Privatization and the sources of performance improvement in the global telecommunications industry. Telecommunications Policy, 26, 243–268. Boubakri, N., Cosset, J. C., Debab, N., & Valery, P. (2013). Privatization and globalization: An empirical analysis. Journal of Banking and Finance, 37, 1898–1914. Boubakri, N., Cosset, J. C., Guedhami, O., & Saffar, W. (2011). The political economy of residual state ownership in privatized firms: Evidence from emerging markets. Journal of Corporate Finance, 17, 244–258. Boubakri, N., Cosset, J. C., & Saffar, W. (2013). The role of state and foreign owners in corporate risk-taking: Evidence from privatization. Journal of Financial Economics, 108, 641–658. Bridgman, B., Gomes, V., & Teixeira, A. (2011). Threatening to increase productivity: Evidence from Brazil’s oil industry. World Development, 39, 1372–1385. Chang, G. H., & Brada, J. C. (2011). A model of the macroeconomic effects of privatization on employment in transition and developing countries with an application to China. Transition Studies Review, 18, 310–327. Chang, X., Fu, K., Low, A., & Zhang, W. (2015). Non-executive employee stock options and corporate innovation. Journal of Financial Economics, 115, 168–188. Chirwa, E. W. (2004). Industry and firm effects of privatization in Malawian oligopolistic manufacturing. The Journal of Industrial Economics, 52, 277–289. Clarke, G. R., & Cull, R. (2002). Political and economic determinants of the likelihood of privatizing Argentine public banks. Journal of Law and Economics, 45, 165–197. Dewenter, K. L., & Malatesta, P. H. (2001). State-owned and privately owned firms: An empirical analysis of profitability, leverage, and labor intensity. American Economic Review, 91, 320–334. Ehrlich, I., Gallais-Hamonno, G., Liu, Z., & Lutter, R. (1994). Productivity growth and firm ownership: An analytical and empirical investigation. Journal of Political Economy, 102, 1006–1038. Estache, A. (2003). Argentina 1990s’ utilities privatization: A cure or a disease? World bank working paper series. Fagerberg, J. (1988). International competitiveness. The Economic Journal, 98, 355–374. Finlay, K., & Magnusson, L. M. (2009). Implementing weak-instrument robust tests for a general class of instrumental-variables models. Stata Journal, 9, 398–421. Frydman, R., Gray, C., Hessel, M., & Rapaczynski, A. (1999). When does privatization work? The impact of private ownership on corporate performance in the transition economies. Quarterly Journal of Economics, 114, 1153–1191. Garnaut, R., Song, L., & Tenev, S. (2005). China’s ownership transformation: Process, outcomes, World Bank Washington D.C. xvii, 235 Gupta, N. (2005). Partial privatization and firm performance. The Journal of Finance, 60, 987–1015. Haller, S. A. (2014). Do domestic firms benefit from foreign presence and import competition in Irish services sectors? The World Economy, 37, 219–243. Hochberg, Y. V., & Lindsey, L. (2010). Incentives, targeting, and firm performance: An analysis of non-executive stock options. Review of Financial Studies, 23, 4148–4186. Huang, Y. (2003). One country, two systems: Foreign-invested enterprises and domestic firms in China. China Economic Review, 14, 404–416. Ito, B., Yashiro, N., Xu, Z., Chen, X., & Wakasugi, R. (2012). How do Chinese industries benefit from FDI spillovers? China Economic Review, 23, 342–356. John, K., Litov, L., & Yeung, B. (2008). Corporate governance and risk-taking. The Journal of Finance, 63, 1679–1728.

Y. Liu et al. / China Economic Review 41 (2016) 196–221

221

Kedia, S., & Rajgopal, S. (2009). Neighborhood matters: The impact of location on broad based stock option plans. Journal of Financial Economics, 92, 109–127. Keller, W., & Yeaple, S. R. (2009). Multinational enterprises, international trade, and productivity growth: Firm-level evidence from the United States. The Review of Economics and Statistics, 91, 821–831. Kokko, A. (1994). Technology, market characteristics, and spillovers. Journal of Development Economics, 43, 279–293. Li, H. (2003). Government’s budget constraint, competition, and privatization: Evidence from China’s rural industry. Journal of Comparative Economics, 31, 486–502. Liu, G. S., & Woo, W. T. (2001). How will ownership in China’s industrial sector evolve with WTO accession? China Economic Review, 12, 137–161. Liu, X., & Wang, C. (2003). Does foreign direct investment facilitate technological progress?: Evidence from Chinese industries. Research Policy, 32, 945–953. Ma, S. Y. (2008). China’s privatization: From gradualism to shock therapy? Asian Survey, 48, 199–214. Matsumura, T. (1998). Partial privatization in mixed duopoly. Journal of Public Economics, 70, 473–483. McKenzie, D., Mookherjee, D., Castañeda, G., & Saavedra, J. (2003). The distributive impact of privatization in Latin America: Evidence from four countries. Economia, 3, 161–233. Megginson, W., & Netter, J. (2001). From state to market: A survey of empirical studies on privatization. Journal of Economic Literature, 39, 321–389. Mizutani, F., & Uranishi, S. (2010). Determinants of privatization of public corporations: Evidence from the Japanese experience. Empirical Economics, 39, 515–535. Mohsni, S., & Otchere, I. (2014). Risk taking behavior of privatized banks. Journal of Corporate Finance, 29, 122–142. Naito, T. (2013). Privatization of public firms and urban unemployment in an integrated economy. Review of Urban and Regional Development Studies, 25, 93–106. National Bureau of Statistics of China(2005). Chinese statistical yearbook 2005. http://www.stats.gov.cn/tjsj/. Naughton, B. (2007). The Chinese economy: Transitions and growth, MIT Press. Newey, W. K. (1987). Efficient estimation of limited dependent variable models with endogenous explanatory variables. Journal of Econometrics, 36, 231–250. Ng, A., Yuce, A., & Chen, E. (2009). Determinants of state equity ownership, and its effect on value/performance: China’s privatized firms. Pacific-Basin Finance Journal, 17, 413–443. Okten, C., & Arin, K. P. (2002). How does privatization affect the firm’s efficiency and technology choice?: Evidence from Turkey. Working paper.. Baton Rouge, LA: Louisiana State University. Omran, M. (2004). The performance of state-owned enterprises and newly privatized firms: Does privatization really matter? World Development, 32, 1019–1041. Porta, R.L., & Lopez-de-Silane, F. (1999). The benefits of privatization: Evidence from Mexico. Quarterly Journal of Economics, 114, 1193–1242. Reece, C., & Abdoul, G. S. (2012). Impact of pension privatization on foreign direct investment. World Development, 40, 291–302. Roberts, B. M., & Saeed, M. A. (2012). Privatizations around the world: Economic or political determinants? Economics and Politics, 24, 47–71. Smith, R. J., & Blundell, R. W. (1986). An exogeneity test for a simultaneous equation Tobit model with an application to labor supply. Econometrica: Journal of the Econometric Society, 679–685. Sprenger, C. (2011). The choice of ownership structure: Evidence from Russian mass privatization. Journal of Comparative Economics, 39, 260–277. Sun, S. (2014). Foreign entry and firm advertising intensity: Evidence from China. Review of Industrial Organization, 45, 79–97. Waldkirch, A., & Ofosu, A. (2010). Foreign presence, spillovers, and productivity: Evidence from Ghana. World Development, 38, 1114–1126. Wang, D. T., & Chen, W. Y. (2014). Foreign direct investment, institutional development, and environmental externalities: Evidence from China. Journal of Environmental Management, 135, 81–90. Wang, X., Xu, L., & Zhu, T. (2004). State-owned enterprise going public: The case of China. Economics of Transition, 12, 467–487. World Bank(2005). . Enterprise surveys 2005, http://www.enterprisesurveys.org/. Yang, Z. (1998). Get out of the mis-perception, speed up reforms of township and village enterprises. Township and Village Enterprises Research, 1, 29–32. Zinnes, C., Eilat, Y., & Sachs, J. (2001). The gains from privatization in transition economies: Is change of ownership enough? IMF Staff papers, 146–170. Zohlnhofer, R., Obinger, H., & Wolf, F. (2008). Partisan politics, globalization, and the determinants of privatization proceeds in advanced democracies. Governance, 21, 95–121.