Location Choice of China’s OFDI in “Belt and Road” Countries – Empirical Evidence from Big Data

With the continuous development of globalization, China has been devoted to promoting the degree of openness. The implementation of “the Belt and Road” initiative brings huge development prospects for China’s OFDI. Utilizing a panel data set consisting of a total of 51 countries along the “the Belt and Road” from 2007 to 2016, this paper establishes an empirical model and examines the influencing factors of the location choice on China’s OFDI in “the Belt and Road” countries. This paper collects comprehensive country-level data of 51 countries along “the Belt and Road” with Python. The number of investment enterprises in the host country published by the Open Directory of Record Results of Overseas Investment Enterprises (Institutions) is used as the explanatory variable. This paper conducts Poisson’s regression to study the influencing factors of location choice. According to the empirical results, bilateral trade agreements can help reduce the difficulties that China may face with in investing and facilitate China to access the host country for investment. In addition, Chinese enterprises may invest in countries where the degree of governance is moderate based on the negative impact on China’s investment in “the Belt and Road” countries. Also, China may invest in countries where the natural resources are poor considering the degree of governance even though natural resources have a positive effect on investment. Furthermore, the level of infrastructure construction in the host country has a positive effect on China’s direct investment in countries along “the Belt and Road”. Besides, good political environment can attract Chinese enterprises to invest in countries along “the Belt and Road”.


INTRODUCTION
The global economy has been in a downturn for a long time since the outbreak of the international financial crisis in 2008. Facing the complicated and ever-changing domestic and international economic situation in the new era, Chinese President Xi proposed "the Belt and Road" initiative during his visit to Central Asia and Southeast Asian countries from September to October 2013. Chinese companies have established nearly 60 economic and trade cooperation zones in more than 20 countries along the route, and the cumulative investment is about 18.5 billion dollars since the initiative was proposed. It can be seen that the countries along the "Belt and Road" have become an important strategic location for China's OFDI, and further promoting future investment in countries and regions along the route has become the key to China's opening up and economic development. Consequently, how to plan the location choice according to Chinese enterprises' need and the characteristics of countries and regions along "the Belt and Road" based on "the Belt and Road" strategy will support theory and policy guarantee for achieving China's industrial structural upgrade and Chinese enterprises' goal of "going out". Meanwhile, China is one of the representatives of emerging economies and this research has important implications for the investment path selection of other emerging developing countries.
Foreign direct investment is of great significance to economic growth and economic restructuring. And the location choice is the key to the Chinese company OFDI. At the same time, many factors in the host country have an important impact on the location choice of the company's OFDI. Therefore, from the perspective of the characteristics of the host country, this paper studies the issue of China's choice of direct investment location in the countries along "the Belt and Road" and further improves the research in this field under the background of implementation of "the Belt and Road". In addition, "the Belt and Road" Initiative involves a large number of countries and regions, and the political, cultural and economic development levels among these countries and regions are quite different, which has caused Chinese enterprises to encounter many challenges and risks in the process of direct investment location choice. This paper uses Poisson regression model to test how different characteristics of host countries affect China's location choice of direct investment in countries along "the Belt and Road", and provides empirical evidence for the influencing factors of China's foreign direct investment location choice. Meanwhile, it is also important for the path selection of emerging developing countries.

LITERATURE REVIEW
Foreign direct investment (OFDI) is of great significance for promoting a country's economic openness and economic development, and has attracted widespread attention from scholars at home and abroad. For investors, choosing the right country is crucial, that is, the location choice of foreign investment has an important impact on OFDI. There is no sound theoretical system for the OFDI location selection problem, but many studies have recognized the importance of location selection. Hymer 1 (1960) first described the location choice in the monopoly theory of traditional industrial organization theory. The main reason for the company's foreign investment is that it has a monopoly advantage over the local enterprises of the same type. Well 2 (1977) proposed "small-scale technology theory", arguing that for most developing countries, low-income countries with smaller market demand are more attractive. However, this theory does not explain the phenomenon that many developing countries invest in developed countries.
Regarding the impact of bilateral investment agreements (BITs) on the location choice of OFDI, the empirical conclusions are different. Hallward-Drieweier 3 (2003) analyzed the OFDI data of 31 developing countries for 20 years, and concluded that BIT has little impact on developing countries' external investment. Nevertheless, there are some papers finding that BIT has a little positive impact on OFDI. Busse et al. 4 (2010) used foreign direct investment flows data of developing countries attracting developed countries. And the results show that BIT has a positive effect. Furthermore, Lu Minghong 5 (2000) pointed out that BIT promotes flow and stock of OFDI. According to these papers, the empirical results of BIT's impact on FDI are not consistent. So it is necessary that add BIT as one of the variables of this paper is important.
From the motivation of OFDI location selection, except for the internal motivations such as market, labor, natural resources or strategic needs, the political, economic and environmental aspects of the host country are also important factors affecting the location choice of the OFDI. Buckley et al. 6 (2007) used China's OFDI flows in 49 countries as explanatory variables and found that Chinese companies invest in countries with large market size, strong cultural similarity, and high political risk. Natural resources have no effect on location choice. On the contrary, Kolstad and Wiig 7 (2009) found that Chinese companies tend to choose countries with large natural resource endowments. Also, Song Weijia and Xu Hongwei 8 (2012) take China's 10-year panel data on direct investment in 51 countries as a sample. This study found that Chinese companies value natural resources, technology, infrastructure and other factors when investing abroad. Jiang Heng 9 (2015) proposed that there are many countries with more internal conflicts among countries along "the Belt and Road", which will bring risks to investments. Meng Qingqiang 10 (2016) used the data on Chinese companies' investment to countries along the route from 2003 to 2013 to explore the investment motivations and found that market, efficiency and natural resources are the main attractions, infrastructure and tariffs are also investment motivations. Besides, Wu Xianming and Huang Chuntao 11 (2016) subdivided the OFDI of Chinese enterprise into investment in developed and developing countries. The study found that Chinese companies all showed obvious market seeking motivation and natural resources seeking motivation. It is can be seen that variables used in their paper are not quite same and the selection of measurement indexes is not comprehensive. On the whole, there are few literatures that combine multiple influencing factors to study location choices.
Although there are many papers on the location choice of OFDI, there are also some shortcomings: (1) there are relatively few studies on OFDI of developing countries, especially China. (2) there is no uniform standard for selection of model and associated explanatory variables and the variables selected is less comprehensive. In view of this, this paper studies how the characteristics of host countries affect China's location choices for direct investment in countries along "the Belt and Road", which is aimed to guide the location selection of OFDI in the future with the deepening development of OFDI. The main innovations of this paper are as follows: (1) This paper consider China as an example of developing countries to study the location choice for countries along "the Belt and Road". (2) This paper uses Poisson regression to comprehensively examine the impact of BIT and other characteristics of host country on the location choice of direct investment. (3) This paper uses new proxy variables to examine the role of core variables which enriches the robustness test.

Econometric model
This paper uses a Poisson regression model. The Poisson distribution is a discrete probability distribution that describes the number of random events occurring in a unit of time. The application in this paper is: the probability of a foreign-invested enterprise investing in a host country i at a specific time t is independent, then the number of foreign-invested enterprises in the host country is regarded as a Poisson distribution: Eit ~Poisson(λit), where the density parameter λ that changes with the host country and time represents the mean and variance of the distribution (Wang Fangfang et al. 12 , 2011).
The regression equation in this paper is as follows (the argument is delayed backward by one period): The explained variable y(c)it indicates the number of enterprises in China directly investing in a host country in the year t. BIT it-1 indicates that China signs a bilateral investment agreement with a host country in year t-1. GOV it-1 is the government governance of a host country in year t-1. POL it-1 is the politic environment of a host country in year t-1. INF it-1 is the fundamental infrastructure of a host country in year t-1. NRS it-1 is the natural resource of a host country in year t-1. These five variables are core explanatory variables of this paper, and agent variables will be selected from different aspects for measurement. Denoting CONTROL it-1 as control variables which include other characteristics of the host country i and national and annual fixed effects. β 1 , β 2 , β 3 , β 4 , β 5 and β 0 are regression coefficients to be estimated. ε it is random disturbance item.

Data and variables selection
This paper selects the names of domestic investors, overseas investment enterprises (institutions) and investment countries (regions) of China's foreign direct investment from 2007 to 2016. Comprehensive countrylevel data are obtained with Python. This paper excludes the panel data of 12,971 enterprises in 51 countries along "the Belt and Road" with a serious data loss and a total investment enterprise of less than 10 in the recent ten years.
This paper selects the number of enterprises that China invests in the host country as the explained variables. The main factors affecting China's investment in host countries are: bilateral trade agreements (BIT), government governance (GOV), political environment (POL), infrastructure (INF), and natural resources (NRS). In order to consider the influencing factors as comprehensively as possible, this paper selects the agent variables from the above five aspects. All variable descriptions and descriptive statistics are shown in Table  1, Table 2 If it is signed in year t or before, BIT value will be 1，otherwise will be 0 Government governance(GOV) The average of the corruption index, the law and order index, and the democratic accountability index as proxy variables Political environment(POL) The value is between 0-9. The greater the score, the more serious the internal contradiction in the country or region.  Source of information: According to the data sources mentioned above, the data sources are available.

Empirical test
Before the regression analysis, this paper first uses the correlation analysis method to detect the multicollinearity that may exist in multiple regressions. In general, a correlation coefficient value between variables greater than 0.8 indicates multi-collinearity. According to the test results, there is no obvious linear correlation between the variables. The correlation coefficient of each variable is lower than 0.8, and most of them are lower than 0.5, that is, there is no serious multicollinearity problem. The Pearson correlation test found that there was no obvious linear correlation among the variables. The correlation coefficient of each variable was lower than 0.8, and most of them were lower than 0.5, that is, there was no serious multicollinearity problem. In this paper, the degree of fit between the Deviance goodness of fit and the Pearson goodness-of-fit test data and the model were calculated before the regression. The results show that both p values are 0, which is consistent with the characteristics of Poisson regression. Table 3 shows the regression results of China's choice of direct investment location for countries along "the Belt and Road". In table 3，Add the core variables of this paper one by one to test the impact of different core variables on the explanatory variables: Regression (1) includes only bilateral investment agreements. Regression (2) includes bilateral investment agreements and government governance. Regression (3) includes bilateral investment agreements and political environment. Regression (4) includes bilateral investment agreements and fundamental infrastructure. Regression (5) includes bilateral investment agreements and natural resources. Regression (6) includes all five core variables. Regression (7), (8) and (9) add all control variables except the geographic distance to the core variables, where regression (7) and (8) only add government governance or political environment respectively. And regression (9) adds these two variables. In this way, it can be observed whether there is certain multicollinearity between government governance and the political environment and affect the results. Regression (10) is the result that adds Geographic distance. It can be seen from Table 3 that in the process of adding explanatory variables one by one, the coefficients and significance of each core explanatory variable basically not change much, and the consistency is consistent, which also indicates that the test of the core explanatory variables is reliable. With the addition of explanatory variables one by one, the R2 of the model is gradually increased, indicating that the interpretation of the explanatory variables in the model is gradually enhanced, and the goodness of fit of the model is improved. At the same time, the p-value of the F-test is less than 0.01, indicating that the variables of the model are overall significant.

Table3. The Influence of Core Variables and Control Variables on the Location Selection of OFDI in Chinese Enterprises
Based on the result, this part analyzes the impact of variables on OFDI location choice. Firstly, the coefficient of BIT is positive and significant at the 1% significance level. It indicates that the host country along "the Belt and Road" which sign BIT with China will attract Chinese enterprises to invest in this country. BIT is a kind of agreement which can protect the interest of both parties and can restrict on the wars, disputes and compensations that may arise in the investment process. It also can ensure the risk and economic loss of Chinese enterprises when investing, and play a significant role in promoting foreign direct investment. Secondly, the coefficient of government governance is negative and significant at 1% significance level in all regression equation containing this variable. The score of GOV is higher, the situation of governance is better. It can be believed that Chinese enterprises attend to invest in countries where government governance is poor. Generally, countries with better governance level can provide more stable environment and reduce the risk that results from Government Issue. This result is obviously different from previous study. The explanation of this result may be there are some convenient conditions in countries with poor government governance, which can attract Chinese enterprises to invest in these countries. Since the core explanatory variable consists of three subindicators, the paper analyzes specifically whether each sub-variable affects the result or each sub-variant affects other explanatory variables to cause the regression result. Thirdly, the coefficient of government environmental factors in the regression is mostly negative and significant at the 1% significance level. The results show that the country's internal political environment is also a major factor affecting China's foreign investment location choice. In the face of unstable investment country environment, the operation of the investment country will be affected, increasing investment costs and investment losses. The higher score of internal political environment indicates the lager internal conflict of the country, so the negative value indicates that China has fewer investment in "the Belt and Road" countries with greater internal conflicts, and is more inclined to be in "the Belt and Road" countries with a relatively stable internal political environment. In addition, for the infrastructure, the regression equation containing the core explanatory variable shows a positive coefficient and is significant at the 1% significance level. The regression results show that the level of infrastructure in the host country is directly proportional to the number of enterprises invested by China in "the Belt and Road" countries. That is, Chinese companies have more choices for countries and regions with better infrastructure. The level of infrastructure in the host country is also an important factor affecting China's choice of investment location in "the Belt and Road" countries. Finally, from the natural resource variable, the coefficient of the variable is positive in regression (5) and is significant at the significant level of 1%. In the other regression equations containing the variable, the coefficient of the variable is negative, and the regression (6) results show a significant level at 5%. Although the core explanatory variables are not significantly stable, and the sign of coefficients changed after adding different variables, natural resources are negative in the regression with government governance variables while others are positive according to the regression results. Therefore, companies generally prefer to invest in countries with more natural resources, enterprises, however, should also consider the constraints of other factors such as local political governance factors when choosing to invest in resource-rich countries, consequently choosing relatively poor natural resources.
Meanwhile, control variables also have a significant explanatory effect on the location choice of China's direct investment in "the Belt and Road" countries. The market size (GDP) is positive in all regression equations, that is, the market size of countries along "the Belt and Road" has a positive role in promoting foreign investment in China. In the choice of location, Chinese companies will choose to seek more markets and invest in countries and regions with larger market sizes. The unemployment rate (UNE) is significantly negative in the regression, and the host countries with lower unemployment are more attractive to Chinese companies for direct investment. The coefficient of the Tariff Level (TAR) variable is positively significant. According to the data above, the higher the average tariff rate of the host country, the lower the score of the variable, indicating that China invests more in countries with lower tariff rates. The bilateral trade relationship (BILT) is positive, indicating that the import and export trade between the host country and China has played a role in promoting China's direct investment in "the Belt and Road" countries. The level of financial development (FIN) is unstable in different regressions, and it is negative in the final regression results, indicating that the proportion of credit in the financial sector will affect China's Investment in the host country. The level of urbanization (URB) is significantly unstable, indicating that this has not been considered too much when investing in countries along "the Belt and Road". The coefficient of geographical distance (GEO_dis) is negatively significant, indicating that China has more choices for host countries that are closer to the direct investment in "the Belt and Road" countries.

Robustness Test
In order to ensure the validity of the conclusions of the model and whether the conclusions are different according to the choice of variables, this paper conducts robustness test. This part will use Corruption, Law and Order and Democratic Accountability Index as an alternative to the core variables to test the robustness of the benchmark regression. Furthermore, in order to test the impact resulting from multicollinearity between the government's governance sub-variables and the political environment, the regression of robustness test will be divided into the regression which add the political environment and not to add the political environment when the robustness test is conducted. Table 4 shows the results of robustness test using three different political governance sub-indicators. The coefficient of corruption in the regression is positive, and is significant at the level of significance of 1% and 5%, respectively. It indicates that the degree of corruption of the host government plays a role in promoting Chinese enterprises' direct investment in countries along "the Belt and Road". Law and order index and the democratic accountability index are both negative and significant at a significant level of 1%. This shows that the better the degree of legal and order system and democratic accountability of the host society has no positive impact on Chinese foreign direct investment. Although corruption, law and order, and democratic accountability are variables that measure the degree of government governance, the test results show the opposite sign of coefficient. This is precisely the particularity of China's location choice of investing directly in countries along "the Belt and Road": the Law and Order Index and the Democratic Accountability Index are about the protection and constraints in country's political system while the corruption index is the quality of the government's administrative execution. It is precisely because of the negative effects of law and order and democratic accountability on the number of investments that the coefficient of government governance in the benchmark regression of Table 3 is negatively significant. For most of China's enterprises, they have developed under the unsound system at the beginning. Therefore, this paper carefully speculates that Chinese enterprises pay more attention to the constraints of enterprises in the administrative execution of the government when they choose the host country for foreign investment. Additionally, from the results of the robustness test in table 4, it can be seen that core variables such as bilateral trade agreements, political environment, infrastructure coefficient, natural resources significant level and symbols are basically same. Moreover, the significance of market size, unemployment rate, tariff level, bilateral trade relations, financial development level, urbanization level, geographic distance and sign of coefficient are also the same as the previous benchmark regression. From the results of robustness test, the estimation results of each variable do not deviate much from the changes in the previous section. Overall, the empirical analysis of this paper is robust. and bilateral investment agreements can reduce the potential investment risks. Secondly, the government governance of the host country is also an important factor affecting China's location choice. Good government governance in the host country will impose high constraints on enterprises, so Chinese enterprises tend to choose countries with less good political governance. Thirdly, the government environment is also an important factor affecting the investment of countries along "the Belt and Road". Countries with fewer internal conflicts and more stable society are more likely to attract Chinese companies to invest in. Besides, infrastructure construction is positively affecting China's location choices. China is more inclined to choose a host country with better infrastructure to reduce the cost of local investment. Finally, natural resources are still a significant factor in China's direct investment in countries along "the Belt and Road". In general, China has the motivation to find a country with abundant natural resources when investing abroad, but China also considers the restrictions of other conditions in other host countries and chooses other countries with fewer natural resources to invest. Based on the results of empirical research and the current state of China's investment in countries along "the Belt and Road", this paper proposes three policy recommendations. First of all，China should sign bilateral investment agreements with countries along "the Belt and Road" and improve the specific provisions under the agreement of both parties. Although China has already signed bilateral investment agreements with most countries along "the Belt and Road", the standards of terms and conditions that were signed at that time have changed over time. In order to continue to promote China's direct investment in countries along "the Belt and Road" and to protect the interests of both parties, China should further improve the terms of the bilateral investment agreements already signed and add new terms.
For countries that have not signed BIT, China should prevent the negative impacts of the political environmental risks in the host countries and establish corresponding protection mechanisms. Chinese companies should learn from the experience of foreign direct investment in developed countries, and strengthen investment in countries with potential government risks along "the Belt and Road" from multiple levels, such as strengthening risk warnings for foreign investment and setting up investment risk loss reserves. It can Balance the distribution of location selections. Next, it is necessary to further promote the interconnection of China's infrastructure with the countries along "the Belt and Road". As China has a higher propensity for countries with sound infrastructure, strengthening the infrastructure construction of transportation, energy and communication in countries along "the Belt and Road" will attract Chinese investment. At the same time, cross-border connectivity with host countries, especially in transportation infrastructure, can help China promote cooperation and share achievement on the basis of using local infrastructure. In the end ， China should stimulate opportunities for investment in natural resource-rich countries along "the Belt and Road". Although the empirical results show that countries with abundant natural resources will attract Chinese companies to invest in them, the constraints imposed by the host government will also cause China to reduce its investment. Therefore, China should optimize its cooperative investment relationship with these countries and promote mutual benefit through communication between the two sides.