Ethnic Chinese Networks and International Investment Evidence from Inward FDI in China

Transcription

1 Ethnic Chinese Networks and International Investment Evidence from Inward FDI in China Ting Gao Department of Economics University of Missouri Columbia, MO Phone: (573) September 2000 JEL Classification: F21, F23 Abstract This paper studies the role of business and social networks in international investment by examining the effects of ethnic Chinese networks on foreign direct investment (FDI) in China. After controlling for a variety of economic determinants of FDI, this study finds a significant positive role in inward FDI of ethnic Chinese networks proxied by the population share of ethnic Chinese in the investing country.

2 1. Introduction The role of business and social networks in promoting international trade has attracted increasing research interest in recent years. Rauch and Casella (1998), for example, provide a theoretical model in which information-sharing social or ethnic networks can improve the allocation of resources and increase the volume of trade in the international market under incomplete information. On the empirical front, Head, Ries and Wagner (1997) find that immigrants significantly increase trade between Canada and source countries; Rauch (1999) presents evidence that common language and colonial ties play an important role in international trade, more so in differentiated products than in homogeneous products because search barriers are higher to trade in the former; Rauch and Trindade (1999) shows that ethnic Chinese networks increase bilateral trade, again with a larger effect on trade in differential products. In this paper I continue this line of inquiry by examining the role of the Overseas Chinese in international investment. Specifically, I use data on inward foreign direct investment (FDI) in China from a cross section of source (investing) countries to examine the role of ethnic Chinese networks proxied by the population share of ethnic Chinese in the source country. In a recent study dealing with a similar issue, Baker and Benjamin (1997) look at the impact of Asia-Pacific immigration on foreign investment between Canada and source countries and report that there is virtually no significant link between foreign investment and the stock of Asia-Pacific immigrants in Canada. Conceivably, business and social networks promote international trade and investment through similar mechanisms: helping businesses overcome informal barriers. In the case of international trade, as argued by Rauch and Trindade (1999), ethnic Chinese networks increase trade primarily by two mechanisms. First, through formal and informal contact, ethnic Chinese networks facilitate information sharing that helps match buyers and sell- 1

3 ers in the international market. Second, ethnic Chinese groups help deter opportunistic behavior such as contract violation through enforcement of community sanctions. There is no doubt that information sharing and contract enforcement are also crucial in investment decisions. Information on labor and material costs, suppliers, distributors, and other market conditions is often necessary to assess the future profitability of an investment project. Before committing any capital, a potential investor also wants to make sure that agreements with suppliers and customers will not be violated in the future and his or her legal rights are protected. Thus, investment barriers can be high in countries where information exchange through official channels is limited and the legal environment is weak. Foreign investors may face additional difficulties if they are unfamiliar with the host country s regulations, language and custom. In this respect China appears to have been a country with high barriers to foreign investment. Foreign investment in China is only a recent phenomenon. Although China s regulatory framework for FDI has improved gradually over time since it opened up to international trade and investment (Fu (1999)), properties rights and contract laws are still weak and only enforced sporadically. Moreover, language and cultural barriers, corruption, and legal uncertainties also present additional difficulties to foreign (especially Western) investors, as lucidly illustrated by a number of real examples in Weidenbaum and Hughes (1996). This, however, can give an edge to overseas Chinese businesses. While many Western firms balk at seemingly insurmountable obstacles, some overseas Chinese firms have managed to get around and, in fact, turned these problems to their advantage through informal channels such as personal contact and family connections (or, the bamboo networks, in the terms of Weidenbaum and Hughes (1996)). Through these networks, trade and investment information is passed along, business introductions are made, and deals are done often on 2

4 handshakes. This paper sets out to investigate whether overseas ethnic Chinese play a positive role in FDI in China and how important this role is quantitatively. At first glance, it may seem rather obvious that much of China s inward FDI come from countries or regions with a large presence of Chinese or ethnic Chinese. Table 1 shows the ten largest FDI source countries over the period of Hong Kong and Macau combined are the single largest FDI source region for China, with cumulative FDI of US$ billion. 1 Taiwan ranks the fourth on the list. These two regions are almost 100% Chinese. Singapore and Thailand, which too have a large population of Ethnic Chinese, also make the top ten list. However, caution must be taken before one draws quick conclusions based on such a casual observation. Besides ethnicity and common language, many other factors may have contributed to the large FDI flows from Hong Kong and Taiwan into China. These two regions experienced rapid economic growth in recent decades. As they move into the rank of technologically advanced regions and experience dramatic increases in labor costs, it is natural for businesses in these two regions to turn to a neighbor such as mainland China, which offers both a virtually unlimited supply of cheap labor and a potentially large market. Geographic proximity, labor costs, and market potential can all play a role in the large inflows of FDI in China from Hong Kong and Taiwan. It is thus worthwhile to pursue this issue rigorously. I will attempt to take into account other factors that have been suggested by the empirical literature to have potential roles in determining FDI while focusing on the effect of ethnic Chinese networks. I will also check the robustness of results by excluding from the sample regions Hong Kong and Taiwan that have a large Chinese population. The rest of the paper is organized as follows. Some elaboration of ethnic Chinese 1 Hong Kong and Macau are no longer qualified to be foreign investors since Hong Kong was turned over to P.R. China in 1997 and Macau in

5 networks is provided in section 2. In section 3, empirical specifications are formulated, drawing upon a variety of models of multinational production and FDI. Section 4 describes the sources of data. Section 5 contains empirical results and discussions. Finally concluding remarks are collected in section More about Ethnic Chinese Networks According to the numbers collected in Yeung and Olds (2000) (1995 figures), there are approximately 55 million ethnic Chinese living outside mainland China. Among those, about 50 millions live in Asia, 3.4 millions in America, and the rest spread over across Europe, Africa, and Oceania. A large portion of overseas Chinese reside in East and South East Asia. The ethnic Chinese population share is very high in Taiwan (99%), Hong Kong (98%) and Singapore (76%). 2 The recent rise of East and Southeast Asia in the world economy has prompted renewed interest in overseas Chinese networks, partly because of the importance of businesses run by ethnic Chinese in those regions. Companies owned by ethnic Chinese in Singapore, Malaysia, Thailand, Indonesia, and the Phillipines account for about 70% of the private business sector in those countries (Weidenbaum and Hughes (1996)). From the business perspective, overseas Chinese networks refer to formal or informal associations of overseas ethnic Chinese businesspeople. Ethnic Chinese networks are often characterized as building on personal connections based on regional collegiality and kinship (Hamilton (1996)). Business transactions, ranging from financing to advance orders to material purchases, are often done through personal contact and recommendations by friends. It is observed that informal connections have played a large role in transactions among overseas Chinese businesses (Hodder (1996)). Exactly how efficient this type of networks compares to Western commerce, which is 2 The numbers here are for 1995, which are different from those in Table 1. 4

6 underpinned by strong market and legal institutions, is still a subject of debate. However, it is evident that ethnic Chinese networks (or broadly, Asian business networks) do display flexibility and efficiency in certain settings and have been successful in Southeast Asia where market institutions are deemed relatively weak. Ethnic Chinese Networks have been strengthened by the formation of large-scale formal overseas Chinese associations. Liu (2000) provides a detailed account of large gatherings of overseas ethnic Chinese and shows the growing trend of globalization in ethnic Chinese Networks in the last few decades. For example, since the 1960s about one hundred world conventions of overseas Chinese associations have been held. Some associations enjoy strong financial support from ethnic Chinese business leaders and political backing from their local communities. Often these associations serve as information exchange, providing commercial information for their members and businesspeople outside their communities. It is also worth pointing out that many overseas Chinese associations consider establishing connection between members and their ancestral hometowns in China as one of their major goals. Delegates from ancestral home regions in China are often invited to brief business information and promote trade and investment opportunities in China at trade fairs and investment talks arranged by these associations. In addition, these associations also serve as enforcement mechanisms of business trust. According to Liu (2000, p. 112), through formulating and sustaining proper codes of behavior and supporting Chinese cultural values, Chinese social organizations have played a part in fostering a sociocultural ethic that was conducive to business trust. More importantly, they have served as a third-party enforcement mechanism encouraging business trust. As previously discussed, mechanisms such as information sharing and business trust building, through either informal personal contact or formal overseas Chinese associations, 5

7 are likely to play a positive part in foreign investment in China, especially from overseas Chinese businesses. I now turn to the design of this study and empirical evidence. 3. Empirical Specifications The goal of this paper is to examine the role of ethnic Chinese networks in FDI in China. To achieve that, other factors that play a role in international investment in general and in China s FDI in particular must be controlled for in order to isolate the effects of ethnic Chinese networks. It is thus necessary to seek guidance from the literature on FDI and multinational production to identify these other factors. The literature distinguishes between two types of multinational firms (Caves (1996), Markusen (1995)). Vertical multinational firms refers to those whose vertically-integrated production units are located in different countries. One of the country characteristics driving vertical multinational production, as formulated in Helpman (1984) and Helpman and Krugman (1985), is relative factor endowment differences. When differences in relative factor endowments are large enough between two countries, factor proportions-based international trade theory implies that free trade in goods and services does not lead to factor price equalization. As a consequence, some firms may separate their production stages to take advantage of factor price differences. For example, a firm can locate its capital intensive production facilities in a capital abundant country where capital is relatively cheap, and locate its labor intensive production plants in a low labor-cost country. Since the separation of vertically-linked production stages entails intra-firm trade across regions, high trade costs are likely to discourage this type of multinational production Horizontal multinational production, on the other hand, occurs when a firm has several production facilities in different countries or regions, each producing identical or similar products. This is likely to happen when plant-level scale economies are small and trade costs across countries are high. This idea has been explicitly modeled in several recent 6

8 studies (Brainard (1993), Horstman and Markusen (1992)) in which multinational production is a result of the trade-off between proximity to markets and scale economies from production concentration. Higher transport costs, or larger and more similar regional markets are likely to lead to more horizontal multinational activity. A newer model of multinational production, encompassing both vertical and horizontal, is proposed in Markusen (1997) and Markusen et al. (1996). The so-called knowledgecapital multinationals model, while emphasizing multinational firms ownership advantages arising from knowledge capital and knowledge-creating activities such as research and development, allows the possibilities of both vertical and multinational production. Knowledge capital can be applied to production in different geographic locations with small additional costs and is the source of scale economies at the firm level. The production of knowledge capital, as usually assumed, is relatively intensive in skilled labor. Thus, the relative endowment of skilled labor has strong implications for multinational activity. Other factors such as market size, interacting with relative skill endowments, determine the extent of vertical and horizontal multinational production. These models suggest a variety of country characteristics as determinants of FDI and multinational production: the sizes of the home and host country, trade costs, relative capital and skilled labor endowments, and the income level of each country all play a role in determining the scale of international investment between countries. Some of those are summarized in Table 2. Next I turn to FDI in China. China experienced tremendous growth in inward FDI during the last two decades or so. Prior to 1978 when the economic reforms started, FDI was prohibited in China. Since then annual realized FDI in China has grown to $45.3 billion by 1997 and total cumulative FDI has reached $220 billion. According UNCTAD (1999, p. 56), China is the single largest FDI recipient in the developing world. Table 3 clearly shows the growing trend of annual realized FDI inflows in China over the period of 7

9 What do the theories of FDI tell us about the causes of FDI in China? One of the location advantages China possesses is its large pool of unskilled workers. A few fast growing countries or regions such as Hong Kong, Taiwan, and South Korea, facing surging labor costs in their own countries, have been investing heavily in China to take advantage of low labor costs. Although the technological sophistication of production by foreign invested firms has improved over time, much FDI in China is still unskilled labor-intensive production aiming at the export markets in the developed world. It is then logical to hypothesize that a good portion of FDI into China in recent years goes under the category of vertical multinational production. Adding support to this hypothesis is the fact that China has been reluctant to open its domestic markets to foreign invested firms. In much of the 1980s, foreign invested firms were required to export a major portion of their output. Restrictions on foreign exchange also forced foreign invested companies to sell output outside China to earn hard currency in order to pay for imported inputs. Many domestic markets were totally closed to these firms. As a result, until 1991 almost all industrial output by foreign invested firms was exported (Naughton (1996)). Given the limited access to China s domestic markets, low labor costs, which often motivate vertical multinational production, should be an important factor explaining FDI in China. However, these were hardly the only reason why foreign firms set up production facilities in China. Indeed, since 1992 access to China s domestic markets has been eased. By 1994, foreign invested firms were able to sell roughly half of their output in China s domestic markets. Even in the early years of reform when China s domestic markets were mostly closed, many companies, eyeing on the large market potential, ventured into China to simply gain a foothold. The 1.2 billion consumer market of China itself becomes increasingly attractive to foreign investors as China intensifies its efforts to become a member of the 8

10 World Trade Organization. The motive to serve local markets, one of the reasons behind horizontal multinational production, should not be totally dismissed in the case of China. With that in mind, I consider economic factors behind both vertical and horizontal FDI. The first variable to be accounted for is relative factor endowment differences between the source country and China. This is an important determinant of vertical investment. Relative factor endowment differences are often proxied by the difference in per capita GDP (Helpman (1987), Ekholm (1998), Brainard (1997)). In a two-factor (capital and labor) world, this is a reasonably good proxy. In a multi-factor world, more direct measures such as the capital-labor ratio and/or skilled-unskilled labor ratio should be preferred. However, problems often arise as to how the capital stock variable is constructed, what types of labor go under each skill category, and whether other factors such as land and natural resources should also be considered. Regardless of what measures to be used, the models of vertical multinational production predict that, all else equal, the volume of multinational activity increases with differences in relative factor endowments. The effect of country size should be controlled for as well. A large country such as the United States is likely to have more FDI in China than its smaller neighbor Canada. In the horizontal multinationals model, country size has clear implications for FDI. Although the vertical multinationals model does not explicitly suggest a role of size, it is conceivable that size should matter for vertical multinational activity as well in the real world. To control for this effect, I use either the source country s GDP or the combined GDP of the source country and China. The horizontal multinational production model predicts that the combined size of the source and host countries should have a positive impact on the volume of multinational activity. It also implies that the volume of FDI increases with similarity in country size. In other words, given the combined size, the larger the size difference between the home 9

11 and source country, the smaller the scale of multinational activity between them. So a third variable, the absolute difference in GDP between the home and source country, is to be included in the empirical specification. Trade costs, however, have opposite implications for vertical and horizontal multinational production. By definition vertical multinational production leads to intra-firm trade, and therefore is negatively affected by trade costs. Horizontal multinational production is likely to increase as trade costs across regions are higher, since part of the reason for this type of production is to economize on trade costs. However, in a more realistic scenario, horizontal multinational firms often ship materials across plants in different countries. Moreover, the farther apart two plants are, the more difficult coordination and communication are. If both vertical and horizontal multinational production are present, it is not all clear how trade costs affect multinational activity. I also take into account other factors such as whether the source and host country share the same language. These factors are typically controlled for in studies examining cross-country trade or multinational activity. The measure of multinational activity in this study is bilateral cumulative FDI in China over As argued by some authors (e.g., Brainard (1997)), a more appropriate measure of multinational production is production or sales by foreign invested firms in the host country, since it can be meaningfully compared to exports, which is a different mode of serving foreign markets. In studies on U.S. or Swedish multinational firms this measure has been widely used because of the availability of the information. Employment by affiliates of foreign multinational firms can be used as an alternative measure (Ekholm (1998)). The problem is that for the majority of countries, especially developing countries, systematic collection of foreign affiliate production has not been done. The lack of direct measures of production or sales often forces researchers to use the FDI measure based on 10

12 the balance of payments accounts. There are some drawbacks with this measure. For example, FDI numbers neglect reinvestment funded by local income of foreign affiliates or through funds raised in the host country. Nonetheless, this is the best measure of multinational activity available in the case of China. The basic empirical specification is the following F DI ij = α + β 1 T GDP ij + β 2 GDP Diff ij + β 3 P GDP Diff ij +β 4 CP op i + ΓX ij + ɛ ij, (1) where subscripts i and j indicate the source country and China respectively, F DI ij is cumulative FDI (or the FDI stock) from i to j, T GDP ij = GDP i + GDP j is the combined GDP of the home and source country, GDP Diff ij = GDP i GDP j is the absolute value of the difference in GDP between i and j, P GDP Diff ij is the difference in per capita GDP between i and j, CP op i is the ethnic Chinese population share in the source country, and finally X ij is a vector of variables including distance between i and j and a dummy indicating whether i and j share the same language. As in Rauch and Trindade (1999), CP op i is the variable representing the strength of ethnic Chinese Networks in the source country. In the regressions reported below, F DI ij, T GDP ij, GDP Diff ij and P GDP Diff ij are all in logarithm. In an alternative specification I also use two direct measures of relative factor endowment differences between the source country and China. The first one (KLDiff ij ) is the difference in capital-labor ratio, which indicates the relative abundance of capital. The other (SKDiff ij ) is the difference in the relative endowment of skilled labor. These two replace the variable measuring the difference in per capita GDP (P GDP Diff ij ) in equation (1). F DI ij = α + β 1 T GDP ij + β 2 GDP Diff ij + β 3 KLDiff ij +β 4 SKDiff ij + β 5 CP op i + ΓX ij + µ ij. (2) 11

13 How KLDiff ij and SKDiff ij are constructed is detailed in section 4. In the knowledge capital model, the difference in the abundance of skilled workers (KLDiff ij ) is an important determining factor of multinational production. Carr, Markusen and Maskus (1998) also include in their explanatory variables the interaction between the difference in skill endowments KLDiff ij and the difference in country size GDP Diff ij. It is to capture the model implication that a smaller country with rich skill endowments is more likely to become an FDI source country. In results that are not reported in this paper, the interaction term is not significant and does not improve the fit. So it is omitted from the specifications. 4. Data Information about ethnic Chinese population across countries is mainly taken from Rauch and Trindade (1999), where two sets of numbers on ethnic Chinese population, for 1980 and 1990 respectively, are shown. For the purpose of this paper, I choose to use the 1990 information (available for 57 countries excluding China). This information is supplemented with numbers from Segal (1993) to compile a list of 68 countries in our sample which have reported FDI in China. The numbers from Segal (1993) are mostly for The table in the Appendix describes the information on ethnic Chinese population in the sample countries listed in the alphabetic order. GDP, foreign exchange rates and total country population are from the International Financial Statistics (IFS). Annual GDP figures in domestic currencies are converted into U.S. dollars using market exchange rates. Numbers for Hong Kong, Macau and Taiwan are taken from the Statistical Yearbook of China ( ). Annual realized (or actual) FDI in China by source country over is gathered from four main sources: China Foreign Economic Statistics, China Foreign Economic Statistical Yearbook, The Almanac of China s Foreign Economic Relations and Trade, and The Statistical Yearbook 12

14 of China. Data in these publications all originally come from the State Statistical Bureau. I have carefully checked the consistency of the data and found no problems. The Penn World Table (Mark 5.6) provides information on non-residential capital stock per worker. This is used as the measure of the abundance of capital. The 1990 numbers are used in this study because there are more missing values for other years. The abundance of skilled workers is derived from the Yearbook of Labor Statistics. Following Markusen and Maskus (1999), I use the sum of the numbers of workers in category 0/1 (professional, technical, and kindred workers) and category 2 (administrative workers) divided by a country s total employment as the measure of skill abundance for that country. The 1992 numbers are used if they are available. When they are not, I substitute those with the 1990 or 1991 numbers. Information on the distance between a source country and China and whether they share the same language is available on a web site ( Trade.Resources/ TradeData.html) maintained by Jon Haveman at Purdue University. 5. Results I first experiment with regressions using the gravity equation, which has been shown to be quite successful in explaining bilateral trade in the literature. In a typical gravity equation regression, the explanatory variables include the GDP s of the source and host country as well as the distance between the two countries. The GDP of the host country (China) in this case is a constant. The gravity equation also forms the basis for identifying the role of ethnic Chinese networks in international trade in Rauch and Trindade (1999). It often works reasonably well in explaining multinational production (Ekholm (1998), Brainard (1997)). Results based on the gravity equation are collected in Table 4. The dependent variable is the log of cumulative inward FDI over the period of Since there was very little 13

15 FDI into China prior to 1983, the dependent variable can be reasonably thought of as the FDI stock. The GDP of the source country in 1992 and distance between the country and China (both in logarithm) are among the explanatory variables. In column (1), cumulative FDI is found to be positively related to the source country s size (GDP) and negatively related to distance. Both the GDP and distance variables are statistically significant at the one percent level. The numbers in parentheses are heteroskedasticity-consistent t-statistics. The negative relationship between multinational activity and geographic distance is consistent with findings in many other studies (e.g., Ekholm (1998), Markusen and Maskus (1999)). In column (2), the population share of ethnic Chinese in the source country (the ethnic Chinese networks variable, hereinafter) is included in the regression. The coefficient on this variable is positive and statistically significant at the one percent level, indicating that the FDI stock is larger the greater is the ethnic Chinese population share in the source country. The inclusion of this variable increases the explanatory power of the equation significantly. Two dummy variables are also added to the regression in column (3). One is a language dummy that takes the value of 1 if the source country s primary language is Chinese. In the sample, Hong Kong and Taiwan are the only two regions that share the same language with China. The other dummy variable (E&SE Asia) indicates whether the source country is in the East or Southeast Asia region. I speculate that, due to their similar cultural backgrounds, companies in these regions may share some common ways of doing business that facilitate transactions such as cross-border investment. The coefficient on the ethnic Chinese networks variable continues to be positive and significant at the one percent level after the addition of the two dummy variables. The magnitude of the coefficient on the ethnic Chinese networks variable increases slightly. Column (3) also shows that the E&SE Asia variable is significantly positive. The language 14

16 dummy, however, is negative and not statistically significant. Thus, the regressions based on the gravity equation clearly show a positive relationship between the FDI stock and the presence of ethnic Chinese in the source country. I also extend the gravity equation to include the log of the absolute population of ethnic Chinese in the source country. A possible justification for this exercise is that the size of absolute population may matter since formal ethnic Chinese associations are likely to form when the population of ethnic Chinese exceeds certain minimum level. In columns (4)-(5), as it turns out, this variable is positive and significant when the E&SE Asia dummy is absent, but becomes insignificant when the E&SE Asia dummy is included. This may be because the population of ethnic Chinese in the East and Southeast Asian countries is relatively large, so the E&SE Asia dummy also captures part of the effect of the absolute size of ethnic Chinese. Table 5 reports regression results that are more closely based on equation 1. The difference in per capita GDP between the source country and China, which is a proxy for differences in relative factor endowments, is now included as one of the determinants of FDI. This variable is set to zero if the source country s per capita GDP is lower than that of China in 1992, because the vertical multinational production model does not predict any FDI from the source country in that case. In column (1) of Table 5, the coefficient on this variable is positive and statistically significant. Indeed, it indicates that the larger cumulative FDI in China is the greater is the difference in per capita GDP. This result is consistent with the prediction of the vertical multinational production model. Recall that the horizontal multinational production model predicts that FDI is positively related to combined GDP and negatively to the difference in GDP. This is confirmed in column (2), where the coefficient on the difference in total GDP is negative, and the coefficient on the combined GDP of the source country and China is positive. Both are significant at the one percent level. Alternatively, in a regression that is not reported here, 15

17 if the combined GDP is replaced by the GDP of the source country, the latter is also positive and significant. Since the vertical multinational production model has little to say about the role of the difference in absolute size, this result seems to be consistent with the horizontal multinational production model. Taken together, this set of regressions suggests that both vertical and horizontal multinational investment are significantly present in China. This is confirmed in column (3), where determinants of both vertical and horizontal investment are included at the same time. In addition, the distance variable continues to be negative, and the ethnic Chinese networks variable be positive and significant at the one percent level. In column (4) after the dummy variables are included, the ethnic Chinese networks variable is positive and significant, the distance variable is negative, and the E&SE Asia dummy is positive and significant, similar to the findings in the gravity equation regressions. Using the coefficient on the ethnic Chinese networks variable in column (4), the FDI stock increases by 6.4% when the population share of ethnic Chinese in the source country increases by one percentage point. The size of the effect also appears to be economically significant. It is possible that the result on the ethnic Chinese networks variable is driven by a small number of countries or regions that have a large presence of ethnic Chinese and also have a large FDI stock in China. This can potentially cause complication when interpreting the results, due to the fact that China did provide various preferential treatments to investments from Hong Kong and Taiwan at different points in time during the sample period. Therefore, it is possible that some uncontrolled factors other than the large populations of ethnic Chinese contributed to the large inflows of FDI from these two regions. The language dummy should be able to pick up these uncontrolled factors specific to Hong Kong and Taiwan. However, to make sure the effect of ethnic Chinese in the 16

18 regressions is not mainly derived from a small number of outliers, I rerun some of the previous regressions excluding Hong Kong, Taiwan, and Singapore, which each have more than 65% of their population being ethnic Chinese. The regression results are reported in columns (5)-(6) in Table 5, where the positive effect of ethnic Chinese on FDI is still clearly present and significant. In fact, the coefficient on the ethnic Chinese networks variable becomes bigger. In Table 6, regression results based on equation (2) are shown where direct measures of relative factor endowments are used to replace the difference in per capita GDP. One direct measure is the employment share of skilled workers, which indicates the skill abundance of a country. Strictly, what should be used in the regression is the difference in the relative skill abundance between the source and host country. But this information on China is not available. Here I instead use the source country s skill abundance directly in the regression. Due to missing values, the number of observations is reduced to 49. In column (1) of Table 6, the variable measuring skilled labor abundance of the host country is positive and significant at the one percent level, suggesting that FDI is positively related to the skill abundance of the source country, consistent with the knowledge capital model in Carr, Markusen and Maskus (1998). In column (2), with the reduced number of observations, combined GDP is positive, and the difference in GDP is still negative. Both are statistically significant, similar to the results in Table 5. Column (3) reports the regression that includes all explanatory variables in columns (1)-(2), the ethnic Chinese networks variable, and the dummy variables. In all three regressions (columns (1)-(3)), the ethnic Chinese networks variable is found to be positively related to the FDI stock, and the magnitude and significance of the coefficient are comparable to those in the previous regressions. In column (4) capital per worker from the Penn World Table is used as the measure of relative capital abundance. Recall that the year chosen for this measure is 1990 because 17

19 there are fewer missing observations for this particular year in the Penn World Table. Still, as a consequence, the number of observation reduces to 50 countries due to missing values. Without information on capital per worker in China, I use the log of the host country s capital per worker directly instead of the log of the difference in capital per worker between the source country and China. Numbers in column (4) show that the capital per worker variable is positive and significant at the one percent level. That is, FDI in China is positively related to the relative abundance of capital in the source country, a result consistent with the prediction of the vertical multinationals model. Again in column (5) I check the coefficients on the combined size of the source and host country and the difference in country size. The results are similar to those in column (2). In all columns (4)-(6), the ethnic Chinese networks variable is positive and significant at the one percent level. Column 7 in Table 6 reports a regression that includes both the relative skill abundance variable and the capital per worker variable. Only 40 observations remain after those with missing values are deleted, so the degree of freedom is reduced significantly. As the results of this regression show, the skill abundance variable is still positive and significant, but the capital abundance variable is no longer statistically significant and the coefficient is much smaller compared to column (6). The results seem to be in favor of the knowledge-capital model of multinational production, which stresses the role of skilled labor endowments. With all these alternative specifications in Tables 4-6, the ethnic Chinese networks variable is consistently positive and statistically significant at the one percent level, showing the robustness of the result. Therefore, I conclude with confidence that there is a robust positive relationship between the presence of ethnic Chinese in a source country and inward FDI from that country to China. Strictly speaking, the theories I reviewed in section 3 to formulate the empirical specifications apply to the overall activity of multinational firms. Without information on 18

20 multinational firms production, sales, or employment, accumulate FDI (or the FDI stock) is probably the next best proxy for multinational activity. There is no strong justification for applying equations (1)-(2) to annual FDI flows because annual flows only partly reflect increases in multinational activity. However, I go ahead to apply equation (1) to annual FDI to see if ethnic Chinese networks also contribute positively to FDI flows. The results for the most recent four years ( ) are reported in Table 7. As Table 7 shows, equation (1) in fact has reasonably good explanatory power for annual FDI as well, with the adjusted R-squared exceeding 0.65 each of the four years. The same sets of independent variables also explain much of the variation in annual FDI across countries. Similar to the regressions on the FDI stock, combined GDP here is positive, the difference in GDP is negative, and the difference in per capita GDP is positive. In all four years, the ethnic Chinese networks variable is positive and significant at the one percent level. This shows that overseas ethnic Chinese also contribute positively to increases in FDI in China. 6. Conclusions In this paper I present evidence that FDI in China is positively related to the population share of ethnic Chinese in the source country, and interpret this as showing the significant role of ethnic Chinese networks in FDI in China. This result is robust to a variety of specifications. A one percentage point increase in the ethnic Chinese population share leads to a 3.8% or higher increase in cumulative FDI in China, according to the regression results in this paper. This finding adds to the recent strand of literature on the role of group ties and social networks in overcoming informal barriers to international trade and investment. Such a role has often been suggested but not adequately accounted for in the formal theory of international trade and investment. Careful empirical investigations are also rare. I hope 19

21 that this study, along with Rauch (1999), Rauch and Trindade (1999) and others, will stimulate further research interest in this direction. The empirical study in this paper also takes into account other determinants of FDI suggested by recent models of multinational production. I find evidence of both vertical and horizontal FDI in China. Cumulative FDI is positively related to the difference in per capita GDP between the source country and China, or positively related to differences in relative abundance of capital and skilled labor. This confirms the implication of the vertical multinational production model that stresses the importance of differences in relative factor endowments. In the meantime, cumulative FDI is also positively related to the combined size of the source country and China, and negatively related to the size difference, which is predicted by the horizontal multinational production model. Intuitively, these results sit well with the current economic conditions in China. The positive role of overseas ethnic Chinese networks in FDI also implies that there exist significant informal barriers to foreign investment in China, and suggests that there is much room to improve in China s efforts to attract foreign investment. The empirical findings in this study say nothing specific about what informal barriers lead to the role of ethnic Chinese networks, even though problems with information sharing and the legal and regulatory environments are used to motivate this study. A more detailed examination in this regard is certainly worth future efforts. 20

22 References Becker, Michael and Dwayne Benjamin (1997), Asia Pacific Immigration and the Canadian Economy, in Richard G. Harris (ed.), The Asia Pacific Region in the Global Economy, p , University of Calgary Press. Brainard, S. Lael (1993), A Simple Theory of Multinational Corporations and Trade with a Trade-Off between Proximity and Concentration, NBER Working Paper No Brainard, S. Lael (1997), An Empirical Assessment of the Proximity-Concentration Trade-off Between Multinational Sales and Trade, American Economic Review 87(4): Carr, David L., James R. Markusen, and keith E. Maskus (1998), Estimating the Knowledge-Capital Model of the Multinational Enterprise, NBER Working Paper No Forthcoming in American Economic Review. Caves, Richard (1996), Multinational Enterprise and Economic Analysis, Cambridge University Press. China, Almanac of China s Foreign Economic Relations and Trade, Various Issues, China Economics Publishing House. China (1994), China Foreign Economic Statistical Yearbook, China Statistical Publishing House. China, China Foreign Economic Statistics ( ), China Statistical Information and Consultancy Service Center. China, Statistical Yearbook of China, Various Issues, China Statistical Publishing House. 21

23 Ekholm, Karolina (1998), Proximity Advantages, Scale Economies, and the Location of Production, in Pontus Braunerhjelm and Karolina Ekholm ed. The Geography of Multinational Firms, Kluwer Academic Publishers, Boston. Fu, Jun (1999), Institutions and Investment, manuscript, forthcoming, University of Michigan Press. Hamilton, Gary G. (1996), The Organization Foundations of Western and Chinese Commerce: A Historical and Comparative Analysis, in Gary G. Hamilton (ed.), Asia Business Networks, Walter de Gruyter, Berlin and New York. Head, Keith, John Ries, and Donald Wagner (1997), Immigrants as Trade Catalysts, in A.E. Safarian and Wendy Dobson (ed.) The People Link: Human Resource Linkages across the Pacific, University of Toronto Press. Helpman, Elhanan (1984), A Simple Theory of Trade with Multinational Corporations, Journal of International Economics 92: Helpman, Elhanan (1987), Imperfect Competition and International Trade: Evidence from Fourteen Industrial Countries, Journal of the Japanese and International Economies 1: Helpman, Elhanan and Paul Krugman (1985), Market Structure and Foreign Trade, The MIT Press, Cambridge. Hodder, Rupert (1996), Merchant Princes of the East, John Wiley&Sons. Horstmann, Ignatius J., and James R. Markusen (1992), Endogenous Market Structures in International Trade, Journal of International Economics 32, International Labor Office (1998), Yearbook of Labor Statistics, Geneva. Liu, Hong (2000), Globalization, Institutionalization and the Social Foundation of Chinese Business Networks in Yeung, Henry Wai-chung and kris Olds (ed.), Globalization of Chinese Business Firms, St. Martin s Press, New York. Markusen, James R. (1995), The Boundaries of Multinational Enterprises and the 22

24 Theory of International Trade, Journal of Economic Perspectives 9(2), Spring. Markusen, James R. (1997), Trade versus Investment Liberalization, NBER Working Paper Markusen, James R., Anthony J. Venables, Denise Eby Konan, and Kevin Zhang (1996), A Unified Treatment of Horizontal Direct Investment, Vertical Direct Investment, and the Pattern of Trade in Goods and Services, NBER Working Paper Markusen, James R. and Keith E. Maskus (2000), Discriminating among Alternative Theories of the Multinational Enterprises, in Blomstrom, M and L. Goldberg (ed.), Topics in Empirical International Economics, University of Chicago Press. Naughton, Barry (1996), China s Emergence and Prospects as a Trading Nation, Brookings Papers on Economic Activity 2: Rauch, James (1999), Networks versus Markets in International Trade, Journal of International Economics 48: Rauch, James and Alessandra Casella (1998), Overcoming Informational Barriers to International Resource Allocation: Prices and Group Ties, NBER Working Paper No Rauch, James and Vitor Trindade (1999), Ethnic Chinese Networks and International Trade, NBER Working Paper No Segal, Aaron (1993), An Atlas of International Migration, Hans Zell Publishers, London. UNCTAD (1999), World Development Report, United Nations, New York and Geneva. Weidenbaum, Murray and Samuel Hughes (1996), The Bamboo Networks, Martin Kessler Books, the Free Press, New York. Yeung, Henry Wai-chung and Kris Olds (2000), Globalizing Chinese Business Firms: Where are They Coming From, Where are They Heading? in Yeung, Henry Waichung and Kris Olds (ed.), Globalization of Chinese Business Firms, St. Martin s 23

25 Press, New York. 24

26 Table 1 Ten Largest FDI Source Regions: Region Cumulative FDI Ethnic Chinese in Total (Bil US$, Current Prices) Population (1990) (Percent) Hong Kong and Macau Japan United States Taiwan Singapore South Korea United Kingdom Germany France Thailand Note: Author s calculations based on data from China s State Statistical Bureau and Table 1 in Rauch and Trindade (1999).

27 Table 2 Factors Determining Multinational Activity (Theoretical Predictions) Variable Vertical Horizontal Knowledge Production Production Capital Combined GDP of Source and Host Country Difference in GDP 0 Difference in Per Capita GDP + Difference in Capital Per Worker +? Difference in Relative Skill Endowment +? + Distance??? Source: Markusen and Maskus (1999, Table 1), Ekholm (1997). The interactive terms in Markusen and Maskus (1999, Table 1) are omitted here.

28 Table 3 Foreign Direct Investment in China Year Actual FDI (Bil US$) Total Source: Statistical Yearbook of China (1998).

29 Table 4 FDI and Ethnic Chinese Networks: Gravity Equation Dependent Variable: Log of Cumulative FDI in China (1) (2) (3) (4) (5) Constant (2.736) (1.202) (0.466) (0.918) (0.050) Log(Source country s GDP in 1992) (6.229) (6.961) (7.412) (5.635) (5.611) Log(Distance) (3.540) (1.907) (0.096) (1.855) (0.577) Share of Ethnic Chinese in Population (4.783) (5.005) (3.006) (4.348) log(absolute Population of Ethnic Chinese) (2.780) (1.937) E&SE Asia Dummy (2.332) (1.240) Lang Dummy (0.578) (0.657) Adj. R No. of Observations Note: Heteroskedasticity-consistent t-statistics are in the parentheses.

30 Table 5 Ethnic Chinese Networks and FDI Dependent Variable: Log of Cumulative FDI in China (1) (2) (3) (4) (5) (6) Constant (7.549) (1.039) (6.664) (3.308) (2.421) (3.322) Log(Combined GDP in 1992) (5.372) (4.769) (5.503) (5.148) (5.169) Log(Difference in GDP in 1992) (5.692) (1.197) (1.563) (1.474) (1.595) Log(Difference in Per Capita GDP in 1992) (5.853) (4.443) (4.224) (4.215) (4.202) Log(Distance) (6.570) (3.770) (3.631) (1.531) (3.456) (1.581) Share of Ethnic Chinese in Population (4.645) (3.775) (3.691) (2.650) E&SE Asia Dummy (2.026) (1.331) Lang Dummy (0.766) Adj. R No. of Observations Note: Heteroskedasticity-consistent t-statistics are in the parentheses.

31 Table 6 Ethnic Chinese Networks and FDI Dependent Variable: Log of Cumulative FDI in China (1) (2) (3) (4) (5) (6) (7) Constant (5.975) (2.925) (3.496) (2.601) (2.154) (3.682) (1.728) Log(Combined GDP in 1992) (7.217) (7.079) (5.261) (3.786) (6.124) Log(Difference in GDP in 1992) (2.148) (2.654) (1.433) (1.055) (1.979) Difference in Skilled Labor Abundance (3.522) (4.554) (2.322) Log(Difference in Capital Abundance (4.661) (2.921) (0.805) Log(Distance) (4.944) (3.286) (1.141) (3.876) (2.911) (0.679) (1.320) Share of Ethnic Chinese in Population (5.239) (6.647) (4.569) (2.606) (3.638) (6.259) (4.409) E&SE Asia Dummy (1.954) (1.384) (0.826) Adj. R No. of Observations Note: Heteroskedasticity-consistent t-statistics are in the parentheses.