# Impact of outward FDI on firms’ productivity over the food industry: evidence from China

Ting-gui Chen (School of Economics and Management, Shanghai Ocean University, Shanghai, China)
Gan Lin (School of Economics and Management, Shanghai Ocean University, Shanghai, China)
Mitsuyasu Yabe (Laboratory of Environmental Economics, Department of Agricultural and Resource Economics, Faculty of Agriculture, Fukuoka, Japan)

ISSN: 1756-137X

Publication date: 25 October 2019

## Abstract

### Purpose

The purpose of this paper is to study the impact of outward foreign direct investment (OFDI) on the productivity of parent firms over the food industry.

### Design/methodology/approach

The main data in this paper are derived from the China Industrial Enterprise Database 2005–2013 and a data set of Chinese firms’ OFDI information. Then this paper uses propensity score matching to match the treatment and control groups with firm characteristics and combines that with the differences-in-differences method to estimate the real effect of OFDI on total factor productivity.

### Findings

The food firm’s OFDI significantly improves the parent firm’s productivity (known as the OFDI own-firm effect), but this promotion only exists in the short term. The OFDI own-firm effect of food firms differs remarkably as the sub-sectors, regions and ownership of firms vary. The food firm’s OFDI in “non-tax havens” and high-income destinations has a significantly stronger effect on the parent firm’s productivity. FDI, R&D and exporting can effectively strengthen the OFDI own-firm effect of food firms.

### Originality/value

The effect of OFDI on food industry productivity has not been researched yet. This paper aims to fill this gap. This paper further divides the characteristics of food firms into different sub-sectors, regions and ownership types for a comparative analysis, with the aim of conducting a more comprehensive study at the micro-level of firms. In addition, an investigation into which factors influence the degree of the OFDI own-firm effect at the micro-level has not been found in the literature. This paper will draw its own conclusions.

## Keywords

#### Citation

Chen, T., Lin, G. and Yabe, M. (2019), "Impact of outward FDI on firms’ productivity over the food industry: evidence from China", China Agricultural Economic Review, Vol. 11 No. 4, pp. 655-671. https://doi.org/10.1108/CAER-12-2017-0246

### Publisher

:

Emerald Publishing Limited

## 1. Introduction

With the deepening of economic globalization and regional economic integration, China’s outward foreign direct investment (OFDI) has entered the fast lane in the context of China’s “Going out” and “the Belt and Road” Initiatives. The USA, Japan and China, respectively, are the largest investors worldwide and China is responsible for 8.65 percent of the world’s OFDI flows (UNCTAD, 2016). China has become a major investor in some developed destinations, especially in cross-border mergers and acquisitions (CM&A). Outward foreign direct investment is abbreviated as OFDI, which is one of the main forms of outward channel of international investment compared with foreign direct investment (FDI). OFDI is defined separately by the International Monetary Fund and the Ministry of Commerce of the People’s Republic of China. In this paper, we define OFDI as Chinese firms that invest in foreign destinations for operation and management rights of foreign firms with emphasis on capacity-building and adoption of improved production standards. This study only investigates Chinese food firms that have undertaken OFDI. OFDI helps a parent firm stimulate exports (Mucchielli and Soubaya, 2002; Jiang and Jiang, 2014a, b), enhance the quality of export products (Du and Li, 2015; Jing and Li, 2016), promote industrial upgrading (Blomstrom et al., 2000; Li, 2012), ease overcapacity (Wen, 2017) and improve profit margins (Yang and Cao, 2017). OFDI also increases the parent firm’s total factor productivity (TFP) because intellectual capital or other nontechnical information through external channels promotes a firm’s productivity (Jiang and Jiang, 2014a, b), which is known as the OFDI own-firm effect. This effect may be different among various industries because firms have their own motives, abilities and methods in OFDI (Blonigen, 2005; Chawla and Rohra, 2015).

Chinese food firms have begun to actively explore the overseas markets. For example, the Shuanghui Group spent 7.10bn to acquire the largest US pork producer “Smithfield” in 2013. In 2014, the Yili Group established an oceanic production base in New Zealand. Bright Diary acquired a 76.70 percent stake in an Israeli dairy firm, Troyes, in 2015.

China’s main motivations for these FDI are as follows. First, the productivity of the food industry is relatively low in the manufacturing sector (Jin et al., 2017), which influences the efficiency of production processes. Second, the demand for high-quality food is increasing as consumers’ income increases. Third, food security incidents have become more frequent, the total degree of trust in the domestic food industry is decreasing (Li and Shi, 2014) and the domestic firms can produce safer food to meet domestic demand by acquiring overseas firms or introducing higher-safety production lines. Technological breakthroughs are required to improve this situation. However, the effect of OFDI on food industry productivity is not in the literature. This paper aims to study this issue. The OFDI in the food industry gradually increased in the background of China’s “Going out” and “the Belt and Road” initiatives; this paper put forward suggestions regarding the performance of China’s OFDI that may be considered as good practice in innovation promotion within the food industry. In addition, it can help create more specific policies for food firms wanting to engage beneficially in OFDIs. The remainder of the paper is as follows. Section 2 is the literature review and hypotheses. Section 3 describes the data and the calculation of the food firm’s TFP. Section 4 presents the model and method of empirical research. Section 5 examines the results of the empirical analysis. Section 6 concludes.

## 2. Literature review and hypotheses

### 2.1 Literature review

Two types of research are related to this paper. The first type is the study of the relationship between OFDI and TFP for different destinations. Helpman et al. (2004) posited that the most productive firms choose to serve the overseas market through OFDI, the more productive firms choose to export and the lowest productive firms only serve the domestic market. This enterprise heterogeneous trade theory has been verified by a series of empirical studies in different destinations, such as Keller and Yeaple (2004) for the USA, Damijan et al. (2008) for Slovenia and Ryuhei and Takashi (2012) for Japan. Many scholars have noticed there may be a mutual causal relationship between OFDI and productivity in developed destinations, that is, OFDI may also increase the parent firm’s TFP. The possible theoretical explanation for this phenomenon is that incomplete markets make multinationals gain monopoly advantages and utilize these advantages through OFDI to enhance their technological superiority (Hymer, 1969).

Many scholars have conducted empirical research on the OFDI own-firm effect. Using Swedish manufacturing data, Braconier et al. (2001) demonstrated that OFDI can significantly facilitate technology imports. Kimura and Kiyota (2006) used Japanese firm-level vertical panel data to demonstrate that enterprises with OFDI have higher productivity growth. Imbriani et al. (2011) used the 2003–2006 Italian firm-level data to analyze the effect of OFDI and indicated that OFDI would increase the productivity of manufacturing enterprises. Gazaniol and Peltrault (2013) used propensity score matching (PSM) to study the impact of OFDI on the microeconomic performance of French enterprises and demonstrated that part of this business group is more inclined to invest abroad and could significantly improve its business performance through investment.

Does the OFDI own-firm effect still exist in emerging economies and developing destinations? The firms in these destinations cannot improve their productivity by using monopolistic advantages. However, the parent firm may improve their productivity by obtaining advanced technologies and management skills abroad, such as building factories and through CM&A, then utilizing them to improve the product quality and production technology (Desai et al., 2005; Syverson, 2010). Jiang and Jiang (2014a, b) used the 2004–2006 Chinese Industrial Enterprises Database to study the relationship between OFDI and productivity and discovered that OFDI could significantly improve the productivity of enterprises, but the promotion gradually reduces over time. Mao and Xu (2014) used China’s 2004–2009 firm-level data to conclude that a significant causal effect exists between OFDI and corporate innovation. Huang and Zhang (2017) used China’s firm-level data from 2002–2007 to examine the effect of OFDI based on the heterogeneity of a firm’s productivity. They divided firms in terms of absorptive capacity and whether they receive national support and demonstrated the following: enterprises effectively improve productivity if they invest abroad for the first time and the degree of impact varies greatly with the characteristics of the enterprise. Yang et al. (2013) used the 1987–2000 data from Taiwan’s manufacturing industry to study the effect of OFDI on the technological efficiency of enterprises and demonstrated a positive correlation between the enterprises’ OFDI activities and technological progress.

By contrast, some scholars believe that OFDI will not improve the productivity of firms (Hijzen et al., 2007; Bai, 2009) and has a negative effect on them (Dhyne and Guerin, 2014). This may be due to the differences in sample selection, such as Bai (2009) found that the reverse technology spillover effect of China’s OFDI was not statistically significant based on the macro data of national statistical yearbooks of 14 destinations. It may also be due to research methods that have resulted in different outcomes despite use of the same Japanese firm-level data as shown by Hijzen et al. (2007) and Kimura and Kiyota (2006). Hijzen et al. (2007) used a differences-in-differences (DID) model and found that “Going out” had no significant effect on the improvement of firms’ productivity. Some scholars even find that OFDI has a negative impact on productivity (Dhyne and Guerin, 2014). So, whether the differences in these conclusions are due to the differences in sample selection or research methods, no consensus exists as to whether OFDI can significantly improve a firm’s productivity. Because the literature on the OFDI own-firm effect has been mostly at the macro-level the question then becomes, “Will the OFDI own-firm effect of firms with different micro characteristics be different?” Therefore, this paper further divides the characteristics of food firms into different sub-sectors, regions and ownership types for a comparative analysis, with the aim of conducting a more comprehensive study at the micro-level of firms.

The second type of literature has examined the OFDI own-firm effect in specific industries. Pradhan and Singh (2008) examined the Indian auto industry and demonstrated that enterprises can enhance their productivity when they invest in developed or developing destinations. Shen and Ju (2016) demonstrated that the reverse technology spillover effect of OFDI in China’s electronic information industry is significant and increases in the technical level strengthen this effect. Driffield and Love (2003) examined the reverse spillover effect of OFDI in manufacturing in the UK and demonstrated that OFDI does increase the technological level of its manufacturing industry; however, this increase is limited to research and development (R&D) intensive industries. The literature on OFDI in the food industry is insufficient; thus, the impact of OFDI on a firm’s productivity is a valuable topic for the food industry.

In addition, an investigation into which factors influence the degree of the OFDI own-firm effect at the micro-level has not been found in the literature. In this paper, we will draw our own conclusions in the final part of the empirical analysis results.

### 2.2 Hypotheses

As mentioned, China’s food firms may conduct OFDI due to efficiency seeking motivation, food quality motivation and food safety motivation, but what is the specific mechanism of the OFDI own-firm effect? This paper generalizes the conduction process into three phases (Figure 1).

The first phase is the acquisition of advanced technology and management experience. Food firms can acquire foreign advanced technology and management experience through four channels: technology transfer, learning and imitation, the flow of talent and sharing platform. The technology transfer refers to the transfer of technological achievements within firms. At present, more and more Chinese firms are engaged in mergers and acquisitions of firms with more advanced technology levels in developed destinations, internalizing the external market, and acquiring patent technologies, supply chain management, R&D teams, etc. Learning and imitation means that foreign subsidiaries of multinational firms could track, learn and imitate the technology research methods and directions of local leading firms. Compared with domestic firms, multinational subsidiaries are more likely to use their convenience to obtain the latest research and marketing methods, management models and cooperation with scientific research institution due to they are closer to local advanced firms and research centers. The flow of talent means that multinational subsidiaries could improve the technical level by introducing local capable technical talents and management talents, in addition, they can also share and exchange technologies and enhance the abilities of technological innovation through cooperation with local firms. Sharing platform means that multinational subsidiary could absorb their advanced technologies by using local resource platforms, R&D facilities, scientific research culture and scientific research achievements.

The second phase is the absorption and transformation of food firms. After acquiring the advanced technology and management experience of foreign subsidiaries, the multinational parent food firm needs a process of absorbing and transforming to internalize it into its own technological advantages, which can be achieved through the personnel and product flows of multinational subsidiaries and parent firms.

The third phase is the stage in which the firm’s technology spreads to the food industry. Domestic firms could bring technological upgrading through demonstration effect and competitive effect. The demonstration effect means that multinational corporations will play a demonstration role for other firms within the food industry and encourage domestic non-multinational firms to strengthen the construction of R&D institutions after acquiring advanced technologies of foreign leading firms. The competitive effect refers to that the acquisition of advanced technology by multinational corporations will increase the pressure of firms’ competition in the food industry, and then force food firms to enhance their innovation capabilities in order to survive. It is worth noting that such a technological upgrading process can be achieved not only between non-multinational corporations and multinational corporations, but also between different multinational corporations.

Based on the above analysis, we propose the first hypothesis:

H1.

OFDI could increase the productivity of China’s food industry, and the OFDI own-firm effect has hysteresis due to the time required for the absorption, transformation and diffusion of technologies.

China’s food industry consists of three sub-sectors: agricultural food processing industry (AFPI), food manufacturing (FM) and beverage manufacturing (BM), they have different levels of development. There may be differences when the OFDI own-firm effect spreads from firms to the industry. Therefore, food firms affiliated with different sub-sectors may have different effects on productivity through OFDI. The role of FDI in productivity improvement may also be different. Second, the economic level and corporate culture of different regions are also different, so the OFDI own-firm effects of food firms belonging to different regions may also be different, which needs to be verified by subsequent empirical studies. Third, we could divide Chinese firms into state-owned firms and non-state-owned firms according to the type of ownership. OFDI of these two types of food firms may have different effects on productivity due to the operation of non-state-owned firms are freer and more efficient than state-owned firms (Huang and Zhang, 2017), thus, the non-state-owned firms could make better use of reverse technology spillovers from OFDI. Finally, the different destinations of OFDI represent different investment objectives. So, the difference in the destinations of OFDI may have an impact on the OFDI own-firm effect. The manner in which a firm’s OFDI can obtain reverse technology spillovers is through investment, CM&A and other activities to obtain advanced technologies and management skills, the parent firms may become the users and creators of technologies and skills, the developed degree of an investment destinations may affect the efficiency of technology and experience absorption, and OFDI in high-income destinations may be more conducive to the promotion of the firm’s productivity. Moreover, some firms have “system to escape or speculate” motives to invest abroad, meaning the firm invests in “tax havens,” such as Hong Kong, the British Virgin Islands, and the Cayman Islands, to obtain domestic investment preferential policies – their main purpose is not to obtain advanced technologies and management skills. This phenomenon also exists in the food industry, thus, this part of OFDI may have no significant effect on enhancing TFP.

Based on the above analysis, we propose the second hypothesis:

H2.

There is a firm heterogeneity in the OFDI own-firm effect. Specifically, the OFDI own-firm effect of food firms in different sub-sectors and regions may be different. Non-state-owned firms could gain greater productivity enhancement through OFDI than state-owned firms. The OFDI in high-income destinations may be more effective than it in low-income destinations, and the OFDI in “tax havens” may not achieve effective productivity gains.

In addition to objective factors such as sub-sectors and region, the OFDI own-firm effect may also be affected by the firm’s own characteristics such as the level of FDI, export status and innovation ability. FDI allows firms to obtain financial advantages in learning that allow them to absorb and apply advanced technologies and management skills better than the other types of firms. Second, firms focused on R&D absorb the advanced technologies, gain a dominant initiative advantage and apply more efficiently them to the production than other types of firms. Finally, the received advanced technologies can produce the parent firm’s own products through OFDI, and the export of these products is one of the communications between the parent and overseas firm. Additional exchanges may promote the parent firm to continue to use the experiences and technologies to enhance TFP.

Based on the above analysis, we propose the third hypothesis:

H3.

FDI, R&D and exporting could enhance the OFDI own-firm effect.

This paper then uses the PSM and DID methods to examine the following four questions based on the information from the Chinese Industrial Enterprises Database for the period 2005–2013: does the OFDI own-firm effect exist in the food industry? Could this effect persist? Considering the heterogeneity of firms, do the different types of food firms have different effects? Does the food firm’s OFDI in different types of destinations affect the effect? Finally, may some characteristics of food firms affect the OFDI own-firm effect?

## 3. Data and TFP

### 3.1 Data sources and processing

The main data in this paper are derived from the China Industrial Enterprise Database (CIED), which is maintained by the China National Bureau of Statistics and includes all state-owned and above-scale (enterprise annual sales above RMB 5 and 20m, respectively, since 2011) non-state-owned enterprises. The subject of this paper is food firms, which corresponds to the following categories in the CIED: AFPI (industry code: 13), FM (industry code: 14) and BM (industry code: 15). We process the data according to Xie et al. (2008) and Yang (2015): excluding if industrial output value, total assets, capital stock, product sales or other key variables are missing, zero or negative; excluding if the number of employees in a firm is less than 8; excluding if the firm was established before 1950; and keeping if the paid-in capital of a firm is greater than 0. The CIED does not contain OFDI observations. Thus, this paper will process the information from the CIED and a data set of Chinese firms’ OFDI information (CFOFDI) acquired from the Chinese Ministry of Commerce to make matches according to the name of firms and obtain the combined data that comprise firms’ OFDI activities. We demonstrate that the observations of OFDI in the combined data before 2004 are very few and begin to increase significantly in 2005. Therefore, the time span of this paper is 2005–2013. The combined data contain 258,182 observations (337 identified as OFDI observations) and 72,981 firms (307 identified as OFDI firms).

### 3.2 Calculation of TFP

TFP is a key variable in the subsequent analysis of this paper. Studies have used different methods for its estimation, such as the ordinary least squares (OLS) method, the Olley–Pakes (OP) method, the Levinsohn-Petrin method, the fixed effects (FE) method and so on. The OP method could solve the selectivity and simultaneity biases (if we use OLS to estimate TFP) by constructing a survival probability function to estimate the entry and exit of firms and an investment function as the proxy variable of the firm’s observable efficiency effect (Olley and Pakes, 1996). Thus, this paper chooses the OP method as its primary method of estimating the TFP of firms for subsequent analysis. We also use the FE model to calculate TFP, to increase the robustness of the results. In this paper, the output elasticities of capital and labor are estimated by the following OP regression model. Next, TFP is calculated according to the Cobb–Douglas production function:

(1) ln Y i t = β 0 + β 1 ln K i t + β 2 ln L i t + β 3 age i t + m δ m year m + n θ n reg n + k φ k ind k + ε i t ,
where lnYit is the log of output (total value) from firm i at time t, lnKit is its capital input measured by total fixed assets and lnLit is its labor input measured by practitioners. The output value and capital input are based on the industrial producer and fixed asset investment price indexes (base year is 2005) to reduce. The regression uses the OP semiparametric three-step regression method. The state variables are lnKit. The firm’s age is ageit. The free variables are lnLit, a regional dummy variable (regn) and a three-digit code industry dummy variable (indk). The control variable is time trend variable (yearm). The proxy variable is the investment variable (lnIit). The exit variable is exit according to whether the firm is not included in the combined data, and if the firm is out of the data, exit is 1, otherwise 0. Table I presents the description of the relevant variables.

The results in Table I demonstrate that the TFP estimated by the OP method is less than that of the FE method and the regular pattern that an OFDI firm’s TFP is higher than that of non-OFDI firm in the three sub-sectors of the food industry. This phenomenon may be because of the OFDI own-firm effect, but it may also be because the firm had a higher TFP before its OFDI (Helpman et al., 2004) – known as the self-selection effect of the firm. Therefore, determining if OFDI effectively contributes to the improvement of food firms’ TFP requires verification by following empirical analysis.

## 4. Empirical methodology

To verify the aforementioned problems and explore the effect of OFDI on food firms’ TFP, this paper constructs a treatment group (OFDI firms) and control group (non-OFDI firms). Considering that the differences between the TFP of OFDI firms and non-OFDI firms are also likely to be caused by other unobservable and non-time-varying factors, this paper first uses the PSM method to match the treatment and control groups with firm characteristics and then combines that with the DID method to estimate the real effect of OFDI on TFP.

### 4.1 PSM for the sample

According to Heckman et al. (1997), we first divided the sample into treatment and control groups. The treatment group consists of OFDI firms, the control group consists of non-OFDI firms. After merging the CIED and CFOFDI, it can be clearly seen which firms have OFDI records, if the firm has OFDI records in the combined data, we set its OFDI value to 1, otherwise 0. And the logit model calculates the probability score of a firm’s OFDI. We select the control group for the OFDI firms (the treatment group) based on the proximity of the probability score. This paper selects labor productivity, capital intensity, firm size, export, age, FDI, ownership type and R&D as matching variables, see Jiang and Jiang (2014a, b) and Ye and Zhao (2016). Table II presents the matching variables calculation method.

Table III shows the summary of matching variables. It indicates that the value of each firm characteristic of OFDI firms is significantly different from that of non-OFDI firms before matching. Therefore, we next use PSM to screen out a certain number of non-OFDI firms so that the values of their firm characteristics are close to OFDI firms.

The matching methods of the PSM include radius matching, caliper matching, K-nearest neighbor matching and kernel matching. K-nearest neighbor matching is the most commonly used. We use K-nearest neighbor matching to pair the treatment and control groups. Because the number of OFDI firms in the sample is relatively small, we make the “k” value equal to 4, meaning we match four non-OFDI firms for each OFDI firm with similar firm characteristics. To test the matching effect, Table IV lists the standardized deviation of the matching variables. We can observe that the standardized deviations of all the variables are less than 5 percent, and the t-test results of the variables demonstrate no significant difference between the two groups after matching. According to Rosenbaum and Rubin (1983), the results demonstrate that K-nearest neighbor matching balances the combined data well.

Is the TFP of OFDI firms still higher than non-OFDI firms after controlling the characteristics of firms? First, we use the PSM method for the entire sample. The results (Table V) demonstrate that the TFP of the treatment and control groups is 6.185 and 5.668, respectively, before matching: the difference between the two is 0.517. After K-nearest neighbor matching, the TFP of the control group is 6.101: the difference between the two is reduced to 0.085. However, the t-test results demonstrate that the difference in the TFP between the two groups is still significant after matching, that is, the TFP mean of OFDI firms is still higher than that of non-OFDI firms after controlling the firm characteristics. We also use radius and kernel matching to do a robust test. The results demonstrate that the other two matching methods support that the TFP of OFDI firms is significantly higher than non-OFDI firms.

Notably, matching the entire sample does not allow us to locate the non-OFDI firms, the firm characteristics or the TFP closest to the OFDI firms when not yet have the OFDI records because the full sample also contains the observations of the OFDI firms after investing abroad. To meet the common trend assumption acquired by the DID method, this paper matches the treatment and control groups year by year based on the firm characteristics when the OFDI firms have not yet invested abroad (refers to Jiang and Jiang, 2014a, b). Additionally, if the firm has an OFDI record in the first year of its entry into the combined data, we base the firm’s characteristics on this year to select the non-OFDI firms that match the OFDI firms. Table VI shows the results of yearly matching. The TFP of the treatment and control groups is very close after yearly matching; the difference is significantly reduced before matching; and the results basically meet the condition that there is no difference between the TFP of OFDI firms before investing abroad and non-OFDI firms.

### 4.2 DID for the sample after matching

We obtain a new control group in which the firm characteristics are similar to the treatment group after the yearly matching. Next, we add the observations of the treatment group to form new combined data. The new combined data contain 9,120 observations (315 identified as OFDI observations) and 1,444 firms (296 identified as OFDI firms). The ATT results of the whole sample in Table V demonstrate that the TFP of OFDI firms is still higher than non-OFDI firms after matching. Therefore, this paper next uses the DID method to examine whether OFDI can improve the TFP of food firms. The DID model is usually used to examine whether the effect of the policy has significant statistical significance, it has the advantage of avoiding endogeneity compared with the traditional method, that is controlling the possible interaction between the dependent variable and the independent variables. Meanwhile, the DID model as a classical method in empirical research can make causal inference of independent variables influencing dependent variables. But using the DID model requires certain conditions, one of the most important is what is called the “natural experiment,” that is the policy impact or the firm’s OFDI decision in this paper must be exogenous, this is one of the reasons why PSM is used in this paper to control the firm’s characteristic and productivity.

The classical DID regression model is as follows:

(2) ln ( TFP i t ) = β 0 + β 1 OFDI i t × TIME t + β 2 Z i t + β 3 OFDI i t + β 4 TIME t + ε i t ,
where the interactive term OFDIit × TIMEt is the product of the OFDI dummy variable (takes 1 if the firm belongs to the treatment group, and 0 otherwise) and time dummy variable (takes 1 for the periods in and after the firm undertakes first OFDI, and 0 otherwise). Zit includes the control variables. εit is the error term. Theoretically, the coefficient of the interaction term β1 represents the effect of OFDI on the firm’s TFP. However, the model is more suitable for the two-stage model. In the data used in this paper, food firms undertake their OFDI in different years, and the period of the same firm’s OFDI is not unique; therefore, this paper refers to Beck et al. (2010) and estimates the following model:
(3) ln ( TFP i t ) = β 0 + β 1 OFDI i t × TIME t + β 2 Z i t + β 3 IND i t + β 4 YEAR t + ε i t ,
where INDit represents the industry FE, and YEARt represents the year FE. In this paper, we select KLR, SIZE, EXP, AGE, FDI, OWN and RD as the control variable Zit, according to the literature. The coefficient of OFDIit × TIMEt represents the impact of OFDI on the firm’s TFP, meaning that OFDI has a positive effect on TFP if β1>0, and OFDI has a negative effect on TFP if β1<0.

## 5. Empirical results

First, to intuitively perceive the causal relationship between OFDI and the TFP, this paper starts to estimate Equation (3) based on the new combined data. Second, the dynamic trend of the OFDI own-firm effect is investigated by the hysteresis effect test. Third, we examine the effect of OFDI on TFP from the perspective of sub-sectors, sub-regions, the different types of ownership and the different investment destinations. Finally, we introduce microcosmic characteristic variables to investigate the influence of firm characteristics on the OFDI own-firm effect.

### 5.1 Baseline results

Table VII reports the baseline results. From Column 1, we can see that the coefficient of OFDIit × TIMEt is significantly positive without controlling any variables, the year or industry FE, indicating that the food firms’ OFDI can significantly increase their productivity. Additionally, although the year and industry FE are controlled, the results are still robust, according to Columns 2. Columns 3 and 4 are the regression results after controlling the firm characteristics variables, where Column 4 is the result of the regression of Equation (3), and the coefficient of OFDIit × TIMEt is still significantly positive with controlling of the firm characteristics variables, year FE and industry FE. These results prove, again, that food firms’ OFDI can promote the promotion of TFP. Column 5 reports the result of the TFP calculated by the FE method. The coefficient of OFDIit × TIMEt is 0.0675, which is significant at a 5 percent significance level. These results indicate that the result reported by Column 4 is robust.

Table VII also reports the values of other variables. The coefficient of the capital labor ratio (KLR) is significantly negative. This result indicates that capital intensity will inhibit the increase in TFP, which may be because of the inefficiency of the re-allocation of food industry resources, and the capital the firms will not have what they require to increase output and TFP, resulting in a waste of capital. This situation inhibits TFP growth. The coefficient of firm size (SIZE) is significant, which indicates that firms with a larger scale have a higher TFP. The coefficient of export (EXP) is also positive, indicating that exports will promote the growth of TFP, which is consistent with most studies. Thus, we can conclude that a food firm’s OFDI can significantly enhance its TFP.

### 5.2 Hysteresis effect

After a firm undertakes OFDI, it may affect more than the TFP in the current period. Generally, firms require some time to learn, absorb, make technological advancements and improve their management skills. Therefore, the impact of OFDI on TFP may have a hysteresis effect. Jiang and Jiang (2014a, b) also confirmed that the OFDI own-firm effect does have a hysteresis effect, by using the Chinese firm-level data. Does the same law apply to food firms? Table VIII shows the results.

Columns 6–8 in Table VIII are the results without controlling for the year and industry FE. We observe that the coefficients of the core interaction term (OFDIit × TIMEt) from the lag one period to three periods are significantly positive, and the OFDI own-firm effect is the strongest in the first period of lag, weakens in the second period and rebounds in the third period. After we control for the year and industry FE (see the results in Columns 9–11), the core interaction term’s coefficient of the lag one period is still significantly positive and the highest; however, the coefficients of the second and third periods of lag are no longer significant. The results in Table VII indicate that the OFDI own-firm effect has a hysteresis effect in the food industry and the food firm can also obtain the enhancement after one year of OFDI, but the enhancement will be weakened after two years of OFDI.

This conclusion is significantly different from the findings of Jiang and Jiang (2014a, b). Their research on China’s manufacturing shows that the OFDI own-firm effect increases first and then declines and is the strongest in the lag two years. The difference may be because food firms have a relatively short time to learn advanced technologies and gain experience and can continue to effectively increase TFP in the short term; however, its impact on TFP will diminish after full absorption.

In summary, H1 has been verified.

### 5.3 Firm heterogeneity effect

To study the impact of OFDI on TFP, this paper classifies food firms according to sub-sectors (two-digital), regions and ownership types. We divide the food industry into three sub-sectors: the AFPI, FM and BM. We divide the regions into the eastern region (ER), northeastern region (NER), central region (CR) and western region (WR), according to the China National Bureau of Statistics. We divide ownership types into state-owned (SO) and non-state-owned (NSO). Table IX presents the results of heterogeneity effect test.

Columns 12–14 in Table IX report the results of the inspection of the sub-sectors. The coefficients of the core interaction term are both significantly positive in the AFPI (0.068) and BM (0.229). This result indicates that OFDI effectively promotes the TFP of the AFPI and BM firms, and the role of OFDI in promoting BM is greater than that of the AFPI. The coefficient of the core interaction term is not significant in FM, demonstrating that OFDI has no obvious effect on the improvement of TFP in FM. Because of the large differences in productivity levels across regions, the OFDI own-firm effects may have different effects in different regions.

Columns 15–18 report the results of the inspection of the regions. The coefficients of the core interaction term are both significantly positive in Columns 15 (0.077) and 17 (0.340). These results demonstrate that the OFDI own-firm effect exists, obviously, in the ER and CR, and the role in the CR is greater, whereas OFDI cannot significantly improve the TFP of firms in the NER and WR, according to Columns 16 and 18. By calculating the food firm’s TFP, we demonstrate that the firm productivity in the CR is the lowest. Thus, the biggest role of OFDI in the CR is probably because of the law of diminishing marginal utility on TFP, that is, the impact of external shocks on productivity is faster and more pronounced in areas with lower productivity.

Columns 19 and 20 report the results of the ownership types inspection. We can observe that the coefficient of the core interaction term is significantly positive in Column 20 and not noticeable in Column 19. This result shows that OFDI cannot significantly improve the TFP of SO firms but can significantly improve the TFP of NSOs.

### 5.4 Investment destinations

What types of destinations to invest in may be one of the factors that affect the OFDI own-firm effect? Table X shows the inspection results according to the classification of investment destinations.

Table X demonstrates that the core interaction term coefficient of non-tax havens is significantly positive, whereas the core interaction term of tax havens is nonsignificant. These results demonstrate that the food firm’s OFDI in non-tax havens can promote the improvement of TFP, and OFDI in tax havens cannot noticeably enhance the firm’s TFP. Second, the core interaction term coefficient of high-income destinations is positive, and the core interaction term coefficient is nonsignificant in middle- and low-income destinations. These results demonstrate that food firms can achieve effective productivity promotion through investing in high-income destinations. The results in Table X validate the previous hypothesis. So, H2 has been verified.

### 5.5 Influence of firm characteristics on the OFDI own-firm effect

Through the aforementioned analysis, we conclude that the food firm’s OFDI can significantly improve the TFP. If that is true, then what factors will affect the effect? We will further investigate if the firm characteristics influence the OFDI own-firm effect by constructing the interaction term of OFDI and FDI, RD, capital labor ratio (KLR), export (EXP), firm size (SIZE) and firm age (AGE). The following equation is shown in detail:

(4) ln ( TFP i t ) = β 0 + β 1 OFDI i t × TIME t + β 2 OFDI i t × TIME t × M i t + β 3 Z i t + β 4 IND i t + β 5 YEAR t + ε i t ,
where M represents FDI, RD, KLR, EXP, SIZE and AGE, respectively, Table XI reports the results of the joint effect test.

Table XI shows that the coefficients of the interaction term of OFDI and M are significantly positive except the results of Columns 30 and 31, it can be seen that when FDI, RD and EXP values are 1, the coefficient values of OFDIit × TIMEt are 0.222, 0.229 and 0.139, respectively, from the results of Columns 26, 27 and 29, they are all significantly higher than 0.111 in Column 25. When FDI, RD and EXP values are 0, the coefficient values of OFDIit × TIMEt are 0.083, 0.105 and 0.077, respectively, and they are all significantly lower than 0.111 in Column 25. The results demonstrate that FDI, R&D and exporting can significantly enhance the OFDI own-firm effect. In Column 28, if the value of KLR is 1, the coefficient value of the interaction term of OFDI and KLR is −0.005, and in fact the value of KLR is between 0 and 1, so in any case the coefficient value of the interaction term of OFDI and KLR is less than that of Column 25, indicating that capital intensity has an inhibitory effect on the OFDI own-firm effect. In Columns 30 and 31, the coefficients of interaction term of OFDI and SIZE, AGE are not significant, it shows that firm size and firm age may not have an outstanding influence on the OFDI own-firm effect. In summary, H3 has been verified.

## 6. Conclusions

Related studies confirm that OFDI has a significant effect on TFP. Does this phenomenon exist in the food industry? What are the characteristics of the OFDI own-firm effect in the food industry? This paper empirically studied the effect of OFDI on TFP by using information from the CIED and the data set of CFOFDI from 2005 to 2013 and draws the following conclusions: first, the OFDI of food firms significantly enhances their own TFP. Therefore, the government should encourage food firms to “Going out” for technological improvement and learn advanced technology and management experience from firms in developed destinations under the conditions permitting. Second, the OFDI own-firm effect has a hysteresis effect on the food industry, and the food firm can obtain strong enhancement in the short term; however, this enhancement will weaken in the long run. Third, the OFDI own-firm effect shows obvious firm heterogeneity. OFDI can significantly improve the TFP of the AFPI and BM at the industry level, and cannot significantly improve the TFP of FM. At the regional level, OFDI can significantly improve the TFP of food firms in the ER and CR but not in the WR and NER. At the ownership level, OFDI can significantly improve the TFP of non-state-owned firms but not state-owned firms. The government should formulate different policies for different sectors and regions to encourage food firms’ OFDI and needs to promote the exchange of experiences and lessons among firms, to jointly explore how to strengthen innovation and promote industrial development. The government needs to continue to promote the reform of state-owned firms, encourage fair competition between state-owned and non-state-owned firms, promote the rational use of resources by state-owned firms and inspire their innovation potential. Fourth, the food firm’s OFDI in high-income destinations and non-tax havens can have a significant own-firm effect, whereas productivity cannot improve if the firm invests in medium- and low-income destinations and tax havens. The government should improve the legal system, limit domestic food firms’ investments for speculative purposes and establish a strict regulatory system to pay close attention to the OFDI of food firms. Fifth, FDI, R&D and exporting can strengthen the OFDI own-firm effect of food firms, whereas the capital intensity will inhibit the effect.

## Figures

#### Figure 1

The mechanism of OFDI’s impact on the productivity of the food industry

## Table I

Descriptive statistics of combined data

Agricultural food processing industry Food manufacturing Beverage manufacturing
Variable OFDI non-OFDI OFDI non-OFDI OFDI non-OFDI
lnYit 12.179 10.847 12.182 10.675 12.255 10.747
lnKit 10.136 8.668 10.388 8.914 10.571 9.206
lnIit 8.964 7.550 9.106 7.744 9.514 8.042
lnLit 5.569 4.657 5.787 4.914 5.878 4.910
TFPit (OP) 6.565 6.093 5.493 4.978 5.596 4.987
TFPit (FE) 9.537 8.697 9.364 8.260 9.052 8.051

Note: The values in the Table are the mathematical mean

## Table II

Matching variables calculation method

Variable Variable name Calculation method
LP Labor productivity The log of output labor ratio
KLR Capital labor ratio The log of capital labor ratio
SIZE Firm size The log of output
EXP Export 1 if export delivery value is greater than 0, 0 otherwise
AGE Firm age The number of years since the creation of the firm
FDI Foreign direct investment 1 if the firm has FDI, 0 otherwise
OWN Ownership type 1 if state-owned capital accounts for more than 0.5 paid-up capital, 0 otherwise
RD Research and development 1 if the firm conducts R&D, 0 otherwise

Note: Output and capital are the real value after the reduction

## Table III

The summary of matching variables

OFDI firms Non-OFDI firms
Variable Mean SD Mean SD
LP 6.460 1.200 6.019 1.101
KLR 4.544 1.338 4.076 1.278
SIZE 11.590 1.785 9.877 1.453
EXP 0.543 0.475 0.319 0.466
AGE 11.407 8.171 9.147 7.825
FDI 0.216 0.500 0.415 0.493
OWN 0.021 0.477 0.336 0.473
RD 0.099 0.300 0.039 0.193

## Table IV

Standardized deviation of matching variables

Mean
Variable Treated Control % bias t-stat.
LP 6.460 6.446 1.2 0.37
KLR 4.544 4.502 3.0 0.91
SIZE 11.590 11.539 3.1 0.92
EXP 0.543 0.558 −3.4 −0.92
AGE 11.407 11.275 1.6 0.50
FDI 0.216 0.210 1.5 0.43
OWN 0.021 0.018 1.9 0.69
RD 0.099 0.097 0.9 0.23

## Table V

Matching results of the entire sample

Variable Sample Treated Controls Difference SE t-stat.
K-nearest neighbor matching TFP Unmatched 6.185 5.668 0.517 0.026 20.13***
ATT 6.185 6.101 0.085 0.029 2.90***
Radius matching TFP Unmatched 6.185 5.668 0.517 0.026 20.13***
ATT 6.173 6.084 0.089 0.026 3.42**
Kernel matching TFP Unmatched 6.185 5.668 0.517 0.026 20.13***
ATT 6.185 5.802 0.383 0.026 14.94***

Notes: All the TFP in the Table are estimated by OP method. *,**,***Significant at the 10, 5 and 1 percent levels, respectively

## Table VI

The results of yearly matching

Sample of year Variable Sample Treated Controls Difference SE t-stat.
2005 TFP Unmatched 5.744 5.248 0.496 0.082 6.02
ATT 5.744 5.583 0.161 0.084 1.93
2006 TFP Unmatched 5.072 5.406 −0.334 0.216 −1.55
ATT 5.072 4.828 0.244 0.289 0.85
2007 TFP Unmatched 5.458 5.503 −0.045 0.265 −0.17
ATT 5.458 5.468 −0.009 0.391 −0.03
2008 TFP Unmatched 5.573 5.525 0.048 0.302 0.16
ATT 5.573 5.578 −0.005 0.387 −0.01
2009 TFP Unmatched 5.871 5.795 0.076 0.370 0.20
ATT 5.871 5.811 0.060 0.466 0.13
2010 TFP Unmatched 6.342 5.666 0.676 0.596 1.13
ATT 6.342 6.209 0.133 0.554 0.24
2011 TFP Unmatched 5.698 5.931 −0.233 0.199 −1.17
ATT 5.698 5.888 −0.190 0.276 −0.69
2012 TFP Unmatched 5.951 6.024 −0.090 0.352 −0.26
ATT 5.951 5.744 0.207 0.599 0.35
2013 TFP Unmatched 6.315 5.938 0.377 0.343 1.10
ATT 6.315 6.318 −0.003 0.344 −0.01

## Table VII

Baseline results

(1) (2) (3) (4) (5)
VARIABLES TFP (OP) TFP (OP) TFP (OP) TFP (OP) TFP (FE)
OFDIit × TIMEt 0.409*** 0.111*** 0.213*** 0.0871** 0.0675**
KLR −0.108*** −0.0945*** −0.0486***
SIZE 0.182*** 0.159*** 0.309***
EXP 0.0279 0.127*** 0.128***
AGE 0.0222*** −0.00195 −0.000151
FDI −0.0350 0.0327 0.0259
OWN −0.154** −0.0464 −0.0403
RD −0.0995*** −0.0116 0.0481*
Constant 5.797*** 5.567*** 4.039*** 4.180*** 5.452***
INDit No Yes No Yes Yes
YEARt No Yes No Yes Yes
Observations 9,120 9,120 9,120 9,120 9,120
R2 0.01 0.103 0.08 0.129 0.261

Notes: The dependent variable is lnTFPit. *,**,***Significant at the 10, 5 and 1 percent levels, respectively

## Table VIII

Results of the hysteresis effect

(6) (7) (8) (9) (10) (11)
Lag one period Lag two periods Lag three periods Lag one period Lag two periods Lag three periods
OFDIit × TIMEt 0.188*** 0.107** 0.127** 0.0870** 0.0047 0.0376
Firm characteristics Yes Yes Yes Yes Yes Yes
INDit No No No Yes Yes Yes
YEARt No No No Yes Yes Yes
Constant 4.362*** 4.474*** 4.464*** 4.589*** 4.837*** 5.313***
Observations 7,219 5,986 4,842 7,219 5,986 4,842
R2 0.080 0.039 0.035 0.082 0.058 0.050
Number of firms 1,348 1,289 1,149 1,348 1,289 1,149

Notes: The dependent variable is lnTFPit estimated by the OP method. *,**,***Significant at the 10, 5 and 1 percent levels, respectively

## Table IX

The results of heterogeneity effect test

(12) (13) (14) (15) (16) (17) (18) (19) (20)
AFPI FM BM ER NER CR WR SO NSO
OFDIit × TIMEt 0.068* 0.066 0.229** 0.077** −0.055 0.340*** 0.100 −0.163 0.097***
Firm characteristics Yes Yes Yes Yes Yes Yes Yes Yes Yes
YEARt Yes Yes Yes Yes Yes Yes Yes Yes Yes
Constant 4.12*** 3.03*** 3.14*** 3.83*** 4.36*** 4.26*** 3.26*** 3.75*** 3.93***
Observations 5,385 2,279 1,456 5,471 966 1,580 1,103 191 8,929
Number of firms 870 414 222 853 159 255 177 67 1,441

Notes: The dependent variable is lnTFPit, estimated by OP method. *,**,***Significant at the 10, 5 and 1 percent levels, respectively

## Table X

Results of the investment destinations

(21) (22) (23) (24)
Non-tax havens Tax havens High-income destinations Middle- and low-income destinations
OFDIit × TIMEt 0.0795* 0.0977 0.0921** 0.0774
Firm characteristics Yes Yes Yes Yes
INDit Yes Yes Yes Yes
YEARt Yes Yes Yes Yes
Constant 4.071*** 4.400*** 4.241*** 4.078***
Observations 6,485 2,635 5,866 3,254
R2 0.137 0.121 0.123 0.143

Notes: The dependent variable is lnTFPit, estimated by the OP method. This paper divides the destinations according to the World Bank GNI ranking (2010): high-income destinations have a per capita income that exceeds $12,276, middle income destinations have a per capita income between$1,006 and 12,275 and low-income destinations have a per capita income less than $1,005. Tax havens: several destinations are contained. Based on this new, combined data and tax havens are defined as: Hong Kong (China), the British Virgin Islands and Macao (China). *,**,***Significant at the 10, 5 and 1 percent levels, respectively ## Table XI Results of the joint effect (25) (26) (27) (28) (29) (30) (31) Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 OFDIit × TIMEt 0.111*** 0.0829** 0.105*** −0.0370 0.0766 −0.225 0.157** OFDIit × TIMEt × FDI 0.139*** OFDIit × TIMEt × RD 0.124* OFDIit × TIMEt × KLR 0.0320* OFDIit × TIMEt × EXP 0.0622* OFDIit × TIMEt × SIZE 0.028 OFDIit × TIMEt × AGE −0.003 INDit Yes Yes Yes Yes Yes Yes Yes YEARt Yes Yes Yes Yes Yes Yes Yes Constant 5.567*** 5.567*** 5.566*** 5.567*** 5.567*** 5.568*** 5.567*** Observations 9,120 9,120 9,120 9,120 9,120 9,120 9,120 R2 0.103 0.104 0.104 0.104 0.103 0.104 0.103 Number of firms 1,444 1,444 1,444 1,444 1,444 1,444 1,444 Notes: The dependent variable is lnTFPit, estimated by the OP method. *,**,***Significant at the 10, 5 and 1 percent levels, respectively ## References Bai, J. (2009), “The effect of the reverse technology spillover of Chinese outward direct investment on TFP: an empirical analysis”, World Economy Study, Vol. 8 No. 10, pp. 65-69+89 (in Chinese). Beck, T., Levine, R. and Levkov, A. (2010), “Big bad banks? The winners and losers from bank deregulation in the united states”, Journal of Finance, Vol. 65 No. 5, pp. 1637-1667. Blomstrom, M., Konan, D.E. and Lipsey, R.E. (2000), “FDI in the restructuring of the Japanese economy”, NBER Working Paper No. 7693, 1050 Massachusetts Avenue, Cambridge. Blonigen, B.A. (2005), “A review of the empirical literature on FDI determinants”, Atlantic Economic Journal, Vol. 33 No. 4, pp. 383-403. Braconier, H., Ekholm, K. and Knarvik, K.H.M. (2001), “In search of FDI-transmitted R&D spillovers: a study based on Swedish data”, Review of World Economics, Vol. 137 No. 4, pp. 644-665. Chawla, K. and Rohra, N. (2015), “Determinants of FDI: a literature review”, The International Journal of Business & Management, Vol. 3 No. 3, pp. 227-250. Damijan, J.P., Kostevc, Č. and Polanec, S. (2008), “From innovation to exporting or vice versa? Causal link between innovation activity and exporting in Slovenian microdata”, LICOS Discussion Paper No. 204, Catholic University of Leuven. Desai, M.A., Foley, C.F. and Hines, J.R. (2005), “Foreign direct investment and domestic economic activity”, NBER Working Paper No. 11717, 1050 Massachusetts Avenue, Cambridge. Dhyne, E. and Guerin, S.S. (2014), “Outward foreign direct investment and domestic performance: in search of a causal link”, NBB Working Paper No. 272, National Bank of Belgium. Driffield, N. and Love, J.H. (2003), “Foreign direct investment, technology sourcing and reverse spillovers”, The Manchester School, Vol. 71 No. 6, pp. 659-672. Du, W.J. and Li, M.J. (2015), “Relationship between OFDI and quality of export products: a variable weight estimation based on propensity score matching”, Journal of International Trade, Vol. 8 No. 12, pp. 112-122 (in Chinese). Gazaniol, A. and Peltrault, F. (2013), “Outward FDI, performance and group affiliation: evidence from French matched firms”, Economics Bulletin, Vol. 33 No. 2, pp. 891-904. Heckman, J.J., Ichimura, H. and Todd, P.E. (1997), “Matching as an econometric evaluation estimator: evidence from evaluating a job training programme”, Review of Economic Studies, Vol. 64 No. 4, pp. 605-654. Helpman, E., Melitz, M. and Yeaple, S. (2004), “Export versus FDI with heterogeneous firms”, American Economic Review, Vol. 94 No. 1, pp. 300-316. Hijzen, A., Tomohiko, I. and Yasuyuki, T. (2007), “The effects of multinational production on domestic performance: evidence from Japanese firms”, Discussion Papers Series No. 07-E-006, The Research Institute of Economy, Trade and Industry, Tokyo. Huang, Y. and Zhang, Y. (2017), “How does outward foreign direct investment enhance firm productivity? A heterogeneous empirical analysis from Chinese manufacturing”, China Economic Review, Vol. 44 No. 3, pp. 1-15. Hymer, S. (1969), “Multinational corporation and international oligopoly: the non-American challenge”, Center Discussion Paper No. 76, Yale University, Economic Growth Center, New Haven, CT. Imbriani, C., Pittiglio, R. and Reganati, F. (2011), “Outward foreign direct investment and domestic performance: the Italian manufacturing and services sectors”, Atlantic Economic Journal, Vol. 39 No. 4, pp. 369-381. Jiang, G.H. and Jiang, D.C. (2014a), “Outward FDI of Chinese industrial enterprises and progress of enterprise productivity”, The Journal of World Economy, Vol. 9 No. 9, pp. 53-76 (in Chinese). Jiang, G.H. and Jiang, D.C. (2014b), “The ‘export effect’ of foreign direct investment by Chinese enterprises”, Economic Research Journal, Vol. 49 No. 5, pp. 160-173 (in Chinese). Jin, S., Guo, H., Delgado, M.S. and Wang, H.H. (2017), “Benefit or damage? The productivity effects of FDI in the Chinese food industry”, Food Policy, Vol. 68 No. 4, pp. 1-9. Jing, G.Z. and Li, P. (2016), “Has OFDI promoted the quality of export product in China”, Journal of International Trade, Vol. 8 No. 12, pp. 131-142 (in Chinese). Keller, W. and Yeaple, S. (2004), “Multinational enterprises, international trade, and productivity growth: firm-level evidence from the United States”, NBER Working Paper No. 9504, 1050 Massachusetts Avenue, Cambridge. Kimura, F. and Kiyota, K. (2006), “Exports, FDI, and productivity: dynamic evidence from Japanese firms”, Review of World Economics, Vol. 142 No. 4, pp. 695-719. Li, F.C. (2012), “The industrial upgrading effect of home countries’ outward foreign direct investment: an empirical study from provincial panel of China”, Journal of International Trade, Vol. 6 No. 10, pp. 124-134 (in Chinese). Li, X. and Shi, L. (2014), “An economics explanation for the crisis of trust in food industry”, Economic Research Journal, Vol. 49 No. 1, pp. 169-181 (in Chinese). Mao, Q.L. and Xu, J.Y. (2014), “Dose foreign direct investment of Chinese enterprises promote enterprise innovation”, The Journal of World Economy, Vol. 8 No. 8, pp. 98-125 (in Chinese). Mucchielli, J.L. and Soubaya, I. (2002), “Intra-firm trade and foreign investment: an empirical analysis of French firm”, in Lipsey, R. and Mucchielli, J.L. (Eds), Multinational Firms and Impacts on Employment, Trade and Technology, pp. 43-83. Olley, G.S. and Pakes, A. (1996), “The dynamics of productivity in the telecommunications equipment industry”, Econometrica, Vol. 64 No. 6, pp. 1263-1297. Pradhan, J.P. and Singh, N. (2008), “Outward FDI and knowledge flows: a study of the Indian automotive sector”, International Journal of Institutions and Economies, Vol. 1 No. 1, pp. 155-186. Rosenbaum, P.R. and Rubin, D.B. (1983), “The central role of the propensity score in observational studies for causal effects”, Biometrika, Vol. 70 No. 1, pp. 41-55. Ryuhei, W. and Takashi, N. (2012), “Productivity and FDI of Taiwan firms: a review from a nonparametric approach”, Discussion Papers No. 12-E -033, The Research Institute of Economy, Trade and Industry, Tokyo. Shen, J.X. and Ju, Y. (2016), “The effects of OFDI reverse technology spillover in China’s information technology industry: a quantile regression study”, International Business, Vol. 1 No. 8, pp. 79-87+149 (in Chinese). Syverson, C. (2010), “What determines productivity?”, NBER Working Papers, Vol. 49 No. 2, pp. 326-365. UNCTAD (2016), “World investment report 2015”, United Nations, WBG, GNI per capita, Atlas method (current US$), New York, NY, available at: https://data.worldbank.org/indicator/NY.GNP.PCAP.CD (accessed July 28, 2011).

Wen, H.W. (2017), “Does Chinese firm outward FDI relieve overcapacity-an empirical study based on Chinese industrial enterprises database”, Journal of International Trade, Vol. 4 No. 10, pp. 107-117 (in Chinese).

Xie, Q.L., Rawski, T.G. and Zhang, Y.F. (2008), “Productivity growth and convergence across China s industrial economy”, China Economic Quarterly, Vol. 7 No. 3, pp. 809-826 (in Chinese).

Yang, P.L. and Cao, Z.Y. (2017), “The effect of outward foreign direct investment on firm’ s profit rate: evidences from Chinese industrial firms”, Journal of Zhongnan University of Economics and Law, Vol. 1 No. 1, pp. 132-139+160 (in Chinese).

Yang, R.D. (2015), “Study on the total factor productivity of Chinese manufacturing enterprises”, Economic Research Journal, Vol. 50 No. 2, pp. 61-74 (in Chinese).

Yang, S.F., Chen, K.M. and Huang, T.H. (2013), “Outward foreign direct investment and technical efficiency: evidence from Taiwan’s manufacturing firms”, Journal of Asian Economics, Vol. 27 No. 27, pp. 7-17.

Ye, J. and Zhao, Y.P. (2016), “Outward foreign direct investment and reverse technology spillover: an empirical analysis based on company-level microdata”, Journal of International Trade, Vol. 1 No. 12, pp. 134-144 (in Chinese).