Mergers and Acquisitions

Transcription

1 Migrant Networks, Greenfield FDI, and Cross-Border Mergers and Acquisitions Jackie M.L. Chan Huanhuan Zheng November, 2016 Abstract We examine the relationship between international migrant networks and two types of foreign direct investment (FDI): greenfield investment and cross-border mergers and acquisitions (M&A). Greenfield and brownfield investment (i.e., M&A) that differ in job creation and technology transfer face different implicit entrance barriers. We find that overseas Chinese migrant networks are more closely associated with M&A than greenfield investment. Moreover, the relation between overseas Chinese migrant networks and Chinese outward investment is more pronounced for investment from privately-own enterprises, in tertiary industries, and in host countries that are more culturally distant from China with lower trust, less efficient contract enforcement, and more regulations. The results are robust to various econometric specifications and estimation methods, including an instrumental variables approach that addresses potential endogeneity issues. Keywords: Overseas Chinese, migration, foreign direct investment JEL: F21, F22, F23 Jackie M.L. Chan, Department of Economics, 9/F, Esther Lee Building, Chinese University of Hong Kong, Shatin, N.T., Hong Kong, Huanhuan Zheng, Department of Economics, Lau Chor Tak Institute of Global Economics and Finance, Chinese University of Hong Kong, Shatin, N.T., Hong Kong, 1

2 1 Introduction Ethic networks overcome informal barriers by transmitting information that matches traders and investors with business opportunities, and by building trust that substitutes contract enforcement and deters opportunism (Rauch 2001). Existing studies document significant evidence on how ethic networks promote international trade (Gould 1994; Head and Ries 1998; Rauch and Trindade 2002) and cross-border investments (Burchardi, Chaney and Hassan 2016; Javorcik,Özden, Spatareanu and Neagu 2011). However, little is known about the role of migrant network on greenfield and brownfield investments, two distinct forms of foreign direct investment (FDI). Greenfield FDI refers to set up new companies while brownfield FDI refers to acquire existing firms in host countries. Greenfield FDI creates job opportunities by building up new plants while brownfield FDI tends to be associated with little job creation (if no job reduction, i.e. through replacing local employees with foreigners). Moreover, Greenfield FDI transfers new technology to host country, i.e. through spillover effect, while brownfield FDI tends to absorb technology from the acquired company. As job creation is the top priority for attracting FDI, followed by technology transfer (UCTAD Survey 2014), host countries generally welcome greenfield FDI more than brownfield FDI. Anecdotal evidence also shows that greenfield FDI receive more favorable treatments than brownfield FDI. Host countries constantly block brownfield FDI out of national security or protection of core technology. For example, in 2008, concerning national security risks, Committee on Foreign Investment in the U.S. blocked Chinese telecom giant Huawei from its proposed acquisitions of 3Com, which provides anti-hacking computer software to the U.S. military. Citing the same reason, Canada rejected the proposed acquisition of the Allstream division of Manitoba Telecom Services by Accelero Capital Holdings from Egypt. In contrast, a number of countries provide corporate tax reductions and/or subsidies to attract greenfield FDI that creates sufficient jobs or brings in new technology. For example, Czech Republic offers employment grants for qualified projects that invest at least 9.25 million euro and creates at least 2

3 250 jobs. South Korea offers tax incentive for FDI that meets the minimum investment amount and job-creation. (better examples?) The differential treatments suggests higher entrance barrier for brownfield FDI than for greenfield FDI. It therefore requires deeper knowledge on law, regulations and culture as well as more subtle skills on communications and execution to get official approval of brownfield FDI than Greenfield FDI. Besides, for brownfield FDI, before approval, there are several important steps, such as, identifying the target of acquisitions and negotiating the deals. Sealing a crossborder M&A deal thus requires sophisticated understanding of the capabilities and preferences of both the target s and the acquirer s business as well as home and host country characteristics. Migrants, who are familiar with both sides of law, regulation and culture, play important roles in overcoming these barriers. The differential treatments of host country on greenfield and brownfield FDI create incommensurable challenges for migration network to match investment opportunities. Existing studies focus on the aggregate FDI and generally find a positive impact of migrant networks on FDI. Given that brownfield FDI faces larger information asymmetry and relies more on contract enforcement than greenfield FDI, it is natural to expect a larger impact of migration network on brownfield FDI. However it remains unexplored how the distinct characteristics of greenfield and brownfield FDI can broaden our understanding on the role of migration networks. This paper studies how the relation between migrant networks and Chinese outward FDI (OFDI) varies with their entry modes. We find that overseas Chinese networks are more closely associated with brownfield FDI than greenfield FDI. In particular, adding 1% to the stock of overseas Chinese weighted by host country population is associated with a 3% and a 9% increase in the number of greenfield and brownfield investment projects, and a 4% and a 22% increase in the value of greenfield and brownfield OFDI. The result remains robust after taking into account the endogeneity issue arising from reverse causality, i.e. FDI flows affect migration decisions. However, we find no evidence that immigrants to China are related to Chinese OFDI to immigrants country of origins. This result is different from studies that use U.S data and find that U.S. immigrants promote U.S. FDI to their countries of origin (Burchardi, Chaney 3

4 and Hassan 2016; Javorcik,Özden, Spatareanu and Neagu 2011). One possible reason is that, while U.S is among the top immigrants receiving country, immigrants to China are relatively small which limits their networking efficiency. This paper also documents new findings that differentiate migrant networks role on overcoming information asymmetry and enhancing contract enforcement. The positive relation between overseas Chinese networks and OFDI is found to be stronger in knowledge-intensive tertiary sector, which is typically associated with greater information barriers than other sectors. Such a relation is also found to be stronger in host countries with less efficient contract enforcement. These results provide direct evidence that overseas Chinese network function through both information and trust channels. In addition, overseas Chinese networks are found to be more closely related to OFDI from private-own enterprisers (POEs), which utilize overseas Chinese network more actively than state-own enterprises (SOE) according to CCPIT (2015). Literature Review The role of migration network in overcoming informal barriers is first documented in international trade literature. Gould (1994) find that the stock of immigrants in the U.S. promotes the bilateral import and export between U.S. and immigrants country of origin. Head and Ries (1998) find similar impacts of immigrants on Canadian bilateral trade. Using French plant-level data, Combes, Lafourcade and Mayer (2005) show that both immigrants in France and French emigrants improve French bilateral trade with the corresponding trading partners. These papers highlight the migrants role on reducing information asymmetry (information channel), which facilitates the matching of buyers and sellers. Focusing on overseas Chinese network, Rauch and Trindade (2002) further show that the information channel is more important for commodities without reference prices (more heterogeneity) than for commodities with them (more homogeneous), as ethic network provides thicker information to match the heterogeneous buyers and sellers in characteristic space. In this way, Rauch and Trindade singles out the migrants role in overcoming information asymmetry from their role in building trust that provides implicit guarantee for contract enforcement (trust channel). Javorcik, Özden, Spatare- 4

5 anu and Neagu (2011) evaluates the migrant networks in the context of FDI. They find that the stock of immigrants in the US increases the size of US FDI in these immigrants country of origin. Moreover, they find that better educated immigrants exert larger impacts on US FDI to their country of origins. We contribute to this strand of literature by differentiating the role of migrant networks on greenfield and brownfield FDI. Our empirical evidence suggests that migration networks matter more for brownfield FDI than for greenfield FDI because the former faces more information asymmetry and more challenge on contract enforcement. By documenting different barriers faced by greenfield and brownfield investments and offset to an extent by migrant networks, our papers also complements existing studies on the choices of foreign market entry modes. Existing literature focuses on how firms choose different entry modes to optimize operation and organization based on their own heterogeneous comparative advantage (Hennart and Park 1993; Nocke and Yeaple 2007). While firm heterogeneity is certainly important drivers of the entry modes, host country characteristics also affect firms choices between greenfield and brownfield FDI. Theoretical evidences suggest that multinational firms respond to FDI policy and trade liberation when choosing their entry modes (Qiu and Wang 2011; Stepanok 2015). Our paper provides explicit evidence that the same host country element can have different implications for greenfield and brownfield investments due to the distinct nature of these two forms of FDI. Moreover, we highlight migration networks as a new source underlying the choices between greenfield and brownfield investments. We also add to the literature by differentiating the overseas Chinese networks s role on OFDI through information and trust channels. Similar to Rauch and Trindade (2002) who documents a stronger impact of ethic networks on more differentiated trades (thus more information asymmetry), we find the overseas Chinese networks to be more closely associated with OFDI from more knolwedge-intensive industry. Moreover, we find that, in host countries with weak contract enforcement, the relation between overseas Chinese networks and Chinese OFDI is stronger. To the best of our knolwedge, we are the first to provide direct evidence that overseas Chinese networks enhance contract enforcement. 5

6 2 Data Transaction-level data on Chinese outward greenfield and brownfield (or cross-border M&As) FDI are respectively from fdi intelligence and Security Data Company (SDC). We limit our sample to year 2003 to 2014, because 2003 is the earliest year that fdi intelligence record greenfield investments. Our sample includes 3184 greenfield investment projects in 136 distinct administrative regions worthing 314 billion USD, and 1393 cross-border M&A deals in 90 distinct administrative regions with a reported transaction value of 218 billion USD. Table 1 summarizes the number of Chinese outward greenfield (#GF) and brownfield investment projects (#BF) in each host country from the year 2003 to In 81 countries where China practices both greenfield and brownfield investment in, the number of greenfield projects is much larger than the cross-border M&A deals with exceptions of Australia and Canada. U.S. ranks the top destination of Chinese outward investment in either #GF or #BF. There are 55 countries that only receive greenfield investments and 9 countries that only obtain brownfield investments. Greenfield investment appears to dominate Chinese outward investment globally. Figure 1 provides the time-series variations in the number and total value of greenfield and brownfield investments. The left panel shows that the number of greenfield investment projects #GF is consistently larger than the number of brownfield cross-border M&A deals #BF, with the former being 2-3 times as much as the latter. However, in terms of transaction value, the size of brownfield OFDI (BF) sometimes exceed that of greenfield OFDI (GF). It means that the brownfield project is typically larger in size than the greenfield project. For each investing (acquiring) company, we collects its ownership structure from ORBIS Bureau van Dijk. A firm is classified as a state-own enterprise (SOE) if at least 25.01% of its ownerships belong ultimately to Chinese governments, and a private-own enterprises (POE) otherwise. Decomposing OFDI by the ownership structure of the Chinese investors, we observe in Figure 2 similar patterns of larger number of greenfield projects in smaller size relative to brownfield investments, regardless of whether the OFDI originates from SOEs or POEs. Greenfield OFDI appears to be more popular for small investment projects than brownfield 6

7 Table 1: Number of Investment Projects by Country between 2003 and Both Greenfield (GF) and Bronfield (BF) Only Greenfield (GF) or Bronfield (BF) Country #GF #BF Country #GF #BF Country #GF Country #GF Argentina 14 5 Luxembourg 7 3 Afghanistan 3 Rwanda 4 Australia Macao, China 7 3 Algeria 12 Saudi Arabia 12 Austria 1 6 Malaysia Angola 1 Senegal 3 Azerbaijan 9 3 Mexico 33 4 Antigua and Barbuda 1 Serbia 2 Belarus 11 1 Mongolia 7 11 Armenia 1 Slovenia 1 Belgium 27 4 Mozambique 3 1 Bahrain 3 Sudan 4 Bolivia 1 2 Namibia 1 1 Bangladesh 6 Tajikistan 5 Brazil Netherlands 46 2 Bosnia and Herzegovina 2 Macedonia 1 Cambodia 14 1 New Zealand 8 11 Botswana 1 Turkmenistan 5 Canada Nigeria 11 2 Brunei Darussalam 3 Uganda 5 Cayman Islands 1 5 North Korea 6 1 Bulgaria 16 United Rep. of Tanzania 4 Chile 9 2 Norway 3 4 Ivory Coast 1 Uruguay 3 Colombia 9 4 Pakistan 28 4 Cameroon 3 Uzbekistan 13 Cyprus 2 1 Papua New Guinea 3 1 Chad 2 Venezuela 18 Czech Republic 13 3 Peru 14 5 Costa Rica 1 Yemen 1 Dem. Rep. of Congo 5 1 Philippines 36 1 Croatia 1 Denmark 24 7 Poland 33 1 Cuba 2 Ecuador 6 2 Portugal 5 3 Ethiopia 14 Egypt 17 3 Republic of Korea Fiji 1 Estonia 1 1 Romania 35 1 Greece 13 Finland 3 1 Rusia Federation Guyana 5 France Singapore Haiti 1 Gabon 2 4 Slovakia 4 1 Honduras 2 Georgia 2 1 South Africa 48 1 Iran, Islamic Republic of 12 Germany Spain Kenya 11 Ghana 9 1 Sri Lanka 3 2 Kuwait 4 Hong Kong, China Sweden 22 8 Lao 1 Hungary 29 8 Switzerland 13 6 Latvia 1 India Syrian Arab Republic 3 1 Madagascar 2 Indonesia Taiwan Micronesia 1 Iraq 1 2 Thailand Moldova, Republic of 1 Country #BF Ireland 11 2 Tunisia 2 1 Morocco 3 Aruba 1 Israel 6 7 Turkey 22 4 Myanmar 9 Barbados 4 Italy Ukraine 7 7 Nepal 2 British Virgin Islands 42 Jamaica 4 1 United Arab Emirates 43 1 Nicaragua 2 Dominican Republic 1 Japan United Kingdom Niger 2 Eritrea 1 Jordan 6 1 United States Oman 3 Isle of Man 1 Kazakhstan 14 9 Viet Nam 64 7 Panama 5 Republic of the Congo 1 Kyrgyzstan 5 2 Zambia 18 1 Paraguay 2 Sierra Leone 3 Liberia 1 1 Zimbabwe 7 2 Qatar 5 Trinidad and Tobago 2 Lithuania 7 1 OFDI among both SOEs and POEs. Our dependent variable includes the number of greenfield investment projects (#GF), the number of brownfield cross-border M&A deals (#BF), the total number of greenfield and brownfield projects (#All), the investment value of greenfield FDI (GF), the transaction value of cross-border M&A deals (BF), the sum of greenfield and brownfield FDI (All), and the aggregate FDI recorded by UNCTAD (UN FDI), in logarithmic form. Bilateral migration are obtained from United Nations (UN), Department of Economic and Social Affairs. The bilateral migration stocks are available every five years from 1990 to 2000, and for year 2013 and Our key explanatory variable is Emigrants it, the number of Chinese emigrants to host country i at year t normalized by the host country s population size, where the host country population size is from the World Development Indicators (WDI). For the years when the migration data are not available, we carry the value of Emigrants forward until the new data becomes available. For example, Emigrants i2006 to Emigrants i2009 are the 7

8 No. of Investment Projects Brownfield Greenfield Value of Investment (million USD) 7 x 104 Brownfield Greenfield Figure 1: Greenfield and Brownfield FDI same with Emigrants i2005. Another explanatory variable Immigrants it, the number of Chinese immigrants from host country i at year t normalized by the host country s population size, is processed in the same way. The control variables include a set of host country characteristics that are found to be important determinants of FDI. To account for the market-seeking and growth-seeking motives of FDI (Buckley, Clegg, Cross, Liu, Voss and Zheng 2007; Deng 2004), we control for the difference in the logarithm of real GDP per capita in USD between host country and China (log(gdp per capita)) and the difference in GDP growth between host country and China (GDP Growth). The difference in the financial development, measured by the private credit from deposit money banks and other financial institutions divided GDP, between host country and China (Financial Development), is included because it is found to facilitate international expansion of firms through FDI (Desbordes and Wei 2014). Since multinational enterprises seek for new technology development through FDI (Cantwell, Dunning and Janne 2004; Glass and Saggi 1998), we add log(patent), the log difference in the number of patent applications between host country and China. We use Trade Openness, the difference in exports plus imports divided by GDP between host country and China, to control for the degree of business inter- 8

9 Panel A: Greenfield and Brownfield FDI by SOE Brownfield Greenfield 4 x 104 Brownfield Greenfield 3.5 No. of Investment Projects Value of Investment (million USD) Panel B: Greenfield and Brownfield FDI by POE Brownfield Greenfield 4.5 x 104 Brownfield Greenfield 4 No. of Investment Projects Value of Investment (million USD) Figure 2: Greenfield and Brownfield FDI by Ownership 9

10 actions with the rest of the world following Blonigen, Davies, Waddell and Naughton (2007). As regulatory, bureaucratic and legal quality affects the cross-border investments (Baker, Foley and Wurgler 2009; Rossi and Volpin 2004), we follow Bekaert, Harvey and Lundblad (2005) to include Institutional Quality, which is proxied by the difference in the sum of of corruption, law and order, and bureaucratic quality from International Country Risk Guide (ICRG), between host country and China. To capture the regional effect, we include Distance, the log distance between host country and China weighted by population density. To control for cultural similarity that makes cross-border investments more likely, we add Common Language, a dummy, that equals to 1 if at least 9% of the population in the host country speaks the same language with China, and Common Legal System, a dummy that equals to 1 if the host country shares the same legal framework with China. All the time-varying macro control variables are lagged by one year. The definition and sources of variables used in this paper are summarized in Appendix Table 1. 3 Results 3.1 Baseline Results We study the impact of migration network on different types of FDI using the following baseline specification lnofdi it = c+βmigrant it + γx it 1 +Year t + ε it (1) The dependent variable OFDI refers to both count and value measures of Chinese OFDI in host country i at year t. It includes the number of greenfield investment projects (#GF), the number of brownfield cross-border M&A deals (#BF), number of greenfield and brownfield projects (#All), the value of greenfield FDI (GF), the transaction value of cross-border M&A deals (BF), the sum of greenfield and brownfield FDI (All), and the aggregate FDI recorded by UNCTAD (UN FDI). The key explanatory variable is Migrant. We use both Emigrants it, 10

11 defined as the number of overseas Chinese in host country i as a ratio of the total population in country i at period t, and Immigrants it, measured by the number of immigrants present in China from host country i weighted by the total population in country i at period t. The controlling set of host country characteristics X it 1 defined in the previous section. All these macro variables are lagged by one year. We also year fixed effects in the model. Standard errors of the estimated coefficients allow for clustering of observations by by host country. In the case of sectoral-level regression, one digit SIC industry fixed effects are added. Table 2 reports baseline results based on an OLS estimation of Eq.(1). We find that Emigrants is positively and significantly associated with different measures of OFDI. Economically, holding other variables constant, adding 1% Emigrants in level raises the count measures #GF, #BF and #All by 2.3%, 5.7% and 3.4%, respectively, and increases the investments value measures GF, BF, All and UN FDI by 2.7%, 13.7%, 5.7% and 8.5%, respectively. The coefficient of Emigrants in the regression of #BF (BF) is larger than that in the regression of #GF (GF), and the difference is significantly at 1% level with a χ 2 -statistic of 40 (30), It suggests overseas Chinese stocks are more closely related to brownfield OFDI than to greenfield OFDI. Intuitively, the cross border M&A requires intensive knowledge of the capabilities and preferences of the acquiring and target firms that are from different cultural backgrounds, which increases its reliance on overseas Chinese that bridge the gap. The coefficients of control variables are generally consistent with historical observations. Recall that these control variables are measured as the difference between host country and China, our results show that countries with lower GDP per capita than China receives more #GF and GF (column 1 and 4) but less BF from China. This finding is consistent with the Chinese multinationals strategy of going abroad encircled the cities from the rural areas, a pragmatic business strategy to build capacity and accumulate wealth in marke with low competition (rural areas) before marching to developed markets to undertake competition (cities) 1. 1 Encircled the cities from the rural areas is initially a military strategy developed by Mao Zedong, the founding Chairman of China. Guided by this strategy, Communist party established revolutionary bases in rural areas that are largely ignored by the Kuomintang party and gradually accumulate arm forces and wealth to fight with Kuomintang party in cities. This strategy is thought to be crucial for the victory of Communist party in the domestic war. The strategy provides important guidelines for business practice, see for example 11

12 Markets with higher growth rate, greater financial development and better institutional quality generally attract more Chinese OFDI. Countries that are farther away from or sharing the same legal system with China generally attract more Chinese OFDI. There is no evidence that countries that shares the common language with China receives more Chinese OFDI. Replacing the explanatory variable Emigrants with Immigrants, the result summarized in Table 3 shows no evidence that immigrants present in China help direct Chinese OFDI in any forms to their country of origins. This is different from Javorcik,Özden, Spatareanu and Neagu (2011) who find that immigrants in US increase the US FDI to their countries of origin. The difference can due to the fact that, unlike the U.S., immigrants to China are considerably small in size, which limits their networking efficiency and therefore their role in overcoming informal barriers. For this reason, in the following, we will focus our analysis on overseas Chinese. We perform a number of additional tests to ensure our results are robust to alternative specifications. Appendix Table 2 summarizes the estimated coefficients of Emigrants but omit the control variables for brevity (included but not reported). Our result that overseas Chinese networks are more closely related to brownfield OFDI than to greenfield OFDI remains robust across the following regressions: (A) replacing the bilateral OFDI with sectoral-level OFDI based on one-digit SIC industry classification and controlling for both year and industry fix effect to verify whether the results hold at industry level; (B and C) running cross-sectional regression for the year 2010 and 2013 when migration stocks are available to mitigate the concern on carrying over previous migration data 2 ; (D) replacing Migrant it with Migrant it 5 in both panel and cross-sectional regression to address the issue that it takes longer for the emigrants network to take effect; (E) replacing lnofdi it with ln(1+ofdi it ) to differentiate zeros with missing observations; and (F) performing Poisson pseudo-maximum-likelihood (PPML) regression to further tackle with the issues of many zeros and/or missing observations in the measures of FDI. 2 For year 2005 when the migration stocks data are also available, the results are dropped because the number of observations for #BF and BF are too small for regressions. 12

13 Table 2: Overseas Chinese and Chinese outward direct investment (OFDI). The sample period is from year 2003 to The dependent variables are the number of greenfield investment projects (#GF), the number of brownfield cross-border M&A deals (#BF), the total number of greenfield and brownfield projects (#All), the value of greenfield FDI (GF), the transaction value of cross-border M&A deals (BF), the sum of greenfield and brownfield FDI (All), and the aggregate FDI recorded by UNCTAD (UN FDI) in logarithmic form. Emigrants is the number of overseas Chinese weighted by host country population. log(gdp per capita) and GDP Growth are respectively the logarithmic GDP per capita and GDP growth. Financial Development is the private credit divided by GDP. log(patent) is the number of patent applications. Trade Openness is the sum of export and import divided by GDP. Institutional Quality is the sum of of corruption, law and order, and bureaucratic quality. Distance is the log distance between host country and China weighted by population density. Common Language is a dummy that equals to 1 if at least 9% of the population in the host country speaks the same language with China. Common Legal System is a dummy that equals to 1 if host country shares the same legal system with China. All explanatory variables except for Emigrants, Distance, Common Language, Common Legal System are calculated as the difference between the host country and China and lagged by one year. All regressions control for year fix effect. Robust t-statistics in parenthesis are corrected for correlation across observations for a given country. The symbol ***, **, * denote significance level at 1%, 5% and 10% respectively. (1) (2) (3) (4) (5) (6) (7) VARIABLES ln(#gf) ln(#bf) ln(#all) ln(gf) ln(bf) ln(all) ln(un FDI) 13 Emigrants 3.173*** 9.339*** 6.393*** 3.814* *** *** *** (4.471) (8.922) (6.969) (1.957) (4.988) (4.186) (6.760) log(gdp per Capita) *** *** *** 1.147* *** (-3.581) (-0.525) (-3.232) (-3.961) (1.932) (-3.005) (-0.559) GDP Growth * *** (-0.228) (0.172) (-0.040) (1.699) (0.086) (1.084) (2.916) Financial Development (0.842) (0.402) (1.326) (0.013) (-1.430) (0.068) (1.420) Trade Openness (1.117) (-0.766) (0.413) (-0.124) (-1.068) (-0.519) (0.407) log(patent) 0.339*** 0.230*** 0.369*** 0.456*** *** 0.303** (7.293) (3.426) (7.832) (5.279) (0.587) (5.809) (2.247) Institutional Quality * (1.571) (1.930) (1.452) (0.625) (-0.777) (0.584) (-0.281) Distance 0.274* 0.277* 0.277* *** (1.816) (1.834) (1.733) (1.167) (3.575) (1.607) (-0.777) Common Language (-0.061) (1.298) (1.093) (1.444) (-0.125) (1.295) (-0.331) Common Legal System * * (1.063) (1.783) (0.935) (1.766) (1.097) (1.552) (-0.317) Observations R-squared

14 Table 3: Chinese immigrants and outward direct investment (OFDI). OLS. The sample period is from year 2003 to Immigrants it is the number of immigrants present in China from host country i weighted by the total population in country i at period t. Variable descriptions are the same as Table 2. All regressions control for year fix effect. Robust t-statistics in parenthesis are corrected for correlation across observations for a given country. The symbol ***, **, * denote significance level at 1%, 5% and 10% respectively. (1) (2) (3) (4) (5) (6) (7) VARIABLES ln(#gf) ln(#bf) ln(#all) ln(gf) ln(bf) ln(all) ln(un FDI) 14 Immigrants (1.631) (0.641) (1.107) (0.312) (0.655) (0.825) (0.589) log(gdp per Capita) *** * * 2.179*** (-3.670) (0.896) (-1.859) (-1.940) (2.988) (-0.768) (1.352) GDP Growth (-1.241) (-0.362) (-0.972) (-0.817) (-0.220) (-0.934) (-0.591) Financial Development * * * (0.396) (-0.442) (0.002) (-1.890) (-0.882) (-1.901) (-1.846) Trade Openness 0.390* 0.770** 0.572** ** (1.955) (2.498) (2.425) (1.460) (0.828) (1.648) (2.677) log(patent) 0.393*** *** 0.450*** *** 0.345** (5.433) (1.263) (4.727) (3.468) (0.123) (3.186) (2.482) Institutional Quality ** 0.112** *** (1.598) (2.184) (2.686) (1.290) (-1.245) (3.125) (1.575) Distance 0.522*** 0.498** 0.548*** 0.859*** 1.284*** 0.906*** (3.543) (2.261) (3.029) (3.452) (3.952) (4.157) (0.824) Common Language (0.030) (-0.164) (0.075) (0.119) (-0.654) (-0.523) (-1.230) Common Legal System 0.384* ** * (1.741) (-0.344) (1.340) (2.398) (-0.347) (1.811) (1.298) Observations R-squared

15 3.2 Instrument Variable Approach So far we have documented a significant and positive relation between overseas Chinese network and Chinese outward brownfield investments in either the number of projects or value of transactions. Moreover, such a linkage is stronger than that between overseas Chinese network and Chinese outward greenfield investments. Nonetheless one may be concerned of the endogeneity issue i.e. the networks can be created by FDI as it transfers employees to host countries. To address this concern, we use the instrumental variable (IV) approach. The first instrument is Emigrant other it, defined as the average emigrant stocks from any countries other than China to the host country i at year t normalized by the host country s total population. As the variations in Emigrant other reflect the migration policy and other implicit host country factors, we expect Emigrant other to resemble Emigrants if the pushing factors affect all migrations alike and to move in the opposite direction with Emigrants if there is crowding out effect i.e. due to the immigration quotas in host country. Therefore, we expect Emigrant other to satisfy the relevance criteria. Emigrants from countries other than China to the host country may have good knowledge on the host market but they lack of understanding on Chinese market, which prevent them from directly matching Chinese outward investments with host market opportunities or overcoming other informal barriers. Therefore Emigrant other is likely to satisfy the exogenous criteria. Table 4 reports the two-stage least squares (2SLS) estimation results. The first stage regressing in column 1 shows that Emigrant other is significantly and negatively related to Emigrants, suggesting the crowding out effects among emigrants targeting the same country dominates. The other tests of the instrument s quality are presented at the bottom of column 2-8. The endogeneity tests generally reject the null hypothesis that the instrumented variable Emigrants could be treated as exogenous except when the dependent variable is #BF or #All. The instrument Emigrant other also passes underidentification and weak identification tests for each of the seven dependent variables. The underidentification tests report the Kleibergen-Paap Lagrange multiplier statistic, which range from 12 to 26 and rejects the null hypothesis that the excluded instruments are not relevant at 5% significance level. The weak identification tests 15

16 report the Cragg-Donald Wald F statistics, which are all higher than the Stock-Yogo s weak identification test critical values at 5% maximal IV relative bias, These results suggests that the instrument Emigrant other is relevant and not weak. The second stage results for each of the seven OFDI variables are pretend in column 2-8. Similar to the OLS results, Emigrants is significantly and positively related to brownfield OFDI (both #BF and BF ) and such a linkage is stronger than that between Emigrants and greenfield FDI (both #GF and GF). For example, with every 1% more Emigrants, the number of crossborder M&As #BF increases by 6.4%, more than the increase in the number of greenfield projects (3.4%). The difference is statistically significant at 5% level. The relation between Emigrants and GF becomes insignificant after we taking into account of the endogeneity issue. The results remain the same when using alternative methods such as general method of moments or limited information likelyhood (not reported). 16

17 Table 4: Instrumental variable approach. This table reports the first and second-stage estimation results based on two-stage least square method. The excluded instrument Emigrant other it is the average emigrant stocks from any countries other than China to the host country i at year t normalized by the host country s population. All regressions control for the same set of control variables described in Table 2 that are not reported. Robust t-statistics in parenthesis are corrected for correlation across observations for a given country. The Endogeneity test report the C statistic and the corresponding p value under the null hypothesis that Emigrants can be treated as exogenous. The Underidentification test reports the Kleibergen-Paap Lagrange multiplier statistic and corresponding p value. The weak identification test reports the Cragg-Donald Wald F statistics. The Stock-Yogo s weak identification test critical values at 5% maximal IV relative bias is The symbol ***, **, * denote significance level at 1%, 5% and 10% respectively. (1) (2) (3) (4) (5) (6) (7) (8) First Stage Second Stage 17 Emigrants ln(#gf) ln(#bf) ln(#all) ln(gf) ln(bf) ln(all) ln(un FDI) Emigrants 3.444* 6.448*** 5.657*** * (1.869) (4.735) (2.942) (-1.432) (1.728) (-0.675) (0.652) Emigrant other *** (-3.446) log(gdp per Capita) 0.003** *** *** *** 1.273*** *** (2.523) (-5.862) (-0.774) (-5.638) (-5.083) (2.690) (-3.609) (-0.965) GDP Growth * *** (-0.695) (-0.250) (0.163) (-0.077) (1.826) (-0.047) (1.041) (3.035) Financial Development ** *** (-0.787) (1.564) (0.994) (2.575) (0.141) (-1.041) (0.232) (2.595) Trade Openness 0.020*** * (4.014) (1.376) (-0.543) (0.922) (1.096) (-0.463) (0.838) (1.738) log(patent) 0.001** 0.339*** 0.239*** 0.371*** 0.482*** *** 0.326*** (2.309) (14.425) (6.767) (16.170) (8.274) (0.982) (9.032) (5.263) Institutional Quality *** 0.064*** 0.088*** 0.062*** (-3.048) (2.950) (2.934) (2.817) (0.377) (-1.570) (0.226) (-0.809) Distance *** 0.277*** 0.246*** 0.271*** *** ** (-3.982) (3.797) (2.874) (3.649) (1.212) (2.677) (1.557) (-2.179) Common Language 0.092*** * * ** (3.751) (-0.120) (1.939) (1.496) (1.910) (0.339) (2.094) (0.706) Common Legal System *** 0.199* 0.460*** 0.169* 0.482* (-3.308) (1.953) (2.735) (1.703) (1.954) (0.965) (1.467) (-1.266) Observations R-squared Endogeneity test p value Underidentification p value Weak Identification

18 3.3 Further Analysis State and Private Ownership There are significant difference between SOE and POE in their business practice. According to a survey on Chinese ODI by China Council for the Promotion of International Trade (CCPIT, 2015, Figure 6.11), 51.1% POEs turn to overseas Chinese chamber of commerce in dealing with investment risk, while this number is only 38.6% for SOEs. The survey also shows that 93.5% POEs highly values opinions from their overseas employees while this number is 80.7% for SOEs (CCPIT, 2015, Figure 6.8). These evidence suggests POEs are more open to utilizing overseas Chinese networks than SOEs. We therefore conjecture that the overseas Chinese network to be associated more closely with POEs investment than with SOEs investment. To explore the role of ownership, we classify the greenfield and brownfield projects according to the ultimate owners of their investing companies. An investing (acquiring) company is considered to be a SOE if at least 25.01% of its ownerships belong ultimately to the Chinese governments as recorded by ORBIS Bureau van Dijk, and a POE otherwise. We then calculate the bilateral #GF, #BF, GF and BF from Chiense SOEs and POEs separately. Table 5 presents the OLS estimation results for the greenfield and brownfield investments originated from SOEs and POEs in Panel A and B respectively. For OFDI measures #GF, #BF and BF, the estimated coefficient of Emigrants is larger in the regression of SOEs investments than in the regression of POEs investments. Panel C reports the χ 2 test of the null hypothesis that the estimated coefficient Emigrants is the same in the regression of SOEs and POEs related OFDI measures. The reported χ 2 statistics reject the null hypothesis at 5% (1%) significance level for #GF and #BF. The results suggests that the overseas Chinese networks are more closely related to the number of both greenfield and brownfield OFDI projects originated from POEs than that from SOEs. Regardless the origins of OFDI, the coefficient of Emigrants is larger in the regression of #BF (BF ) than in the regression of #GF (GF ) and the difference is statistically significant at 1% level (not reported). Our previous result that overseas Chinese networks are more closely related to brownfield investments than greenfield investments remains robust for both POEs 18

19 and SOEs investments. Table 5: State versus Private OFDI. The dependent variables in Panel A are the number of greenfield investment projects (#GF), the number of brownfield cross-border M&A deals (#BF), the value of greenfield FDI (GF), the transaction value of cross-border M&A deals (BF) from Chinese state-own enterprises (SOEs) in logarithmic form. The dependent variables in Panel B are the same measures of outward foreign direct investments (OFDI) from Chinese private-own enterprises (POEs). χ 2 and prob>χ 2 report the statistics and p value under the null hypothesis that the estimated coefficient of Emigrants is the same in the regressions of SOEs and POEs OFDI measures. All regressions control for the same set of control variables described in Table 2. Robust t-statistics in parenthesis are corrected for correlation across observations for a given country. The symbol ***, **, * denote significance level at 1%, 5% and 10% respectively. Panel A: OFDI from SOEs ln(#gf) ln(#bf) ln(gf) ln(bf) Emigrants 1.436** 4.309*** *** (2.554) (6.178) (1.439) (3.638) Observations R-squared Panel B: OFDI from POEs ln(#gf) ln(#bf) ln(gf) ln(bf) Emigrants 3.048*** 9.299*** *** (4.253) (10.096) (1.644) (5.731) Observations R-squared χ prob> χ Heterogeneity across Industries The problem of asymmetric information is severe for knowledge-intensive business such as information technology, engineering, research and development (Coff 1999). With their knowhow and comparative advantage to combine resources from both sides efficiently, overseas Chinese are expected to add more value to more knowledge-intensive industry. To test the hypothesis, we calculate the total number of greenfield and brownfield projects (#GF and #BF) as well as their values (GF and BF) in each of the primary (SIC two digit code 01-14), secondary (SIC two digit code 15-39) and tertiary industry (SIC two digit code 40-99). We then create a dummy variable D Tertiary that equals to one if the observation is from 19

20 tertiary industry and zero otherwise. As tertiary industry requires more intensive knowledge than the other industries, we expect the positive relation between overseas Chinese networks and Chinese outward investments to be stronger in this industry. We estimate Eq.(1) expanded with an interaction between Emigrants and D Tertiary and present the results in Table 6. The interaction term is positive and statistically significant for all four measures of Chinese OFDI, which support a stronger relation between Emigrants and Chinese OFDI in the tertiary industry. Moreover, consistent with our previous finding, we find that the overseas Chinese networks are more closely related to the brownfield OFDI than to greenfield OFDI after controlling for the sectoral difference. We also estimate Eq.(1) for each of the sector separately and report them in Appendix Table 3. The estimated coefficient of Emigrants is positive and statistically significant across different measures of OFDI in the secondary and tertiary sector but not in the primary sector. Moreover, for the same measure of OFDI, the coefficient of Emigrants is greater in the tertiary sector than in the secondary sector, and the difference is significant at 1% level. It suggests that positive relation between overseas Chinese networks and Chinese OFDI are mostly driven by the investments in tertiary industry followed by those in the secondary industry. Table 6: Cross-Industry Difference. The dummy variable D Tertiary equals to one if the Chinese outward foreign direct investment is in tertiary industry and zero otherwise. All regressions control for the same set of control variables described in Table 2 that are not reported. Robust t-statistics in parenthesis are corrected for correlation across observations for a given country. The symbol ***, **, * denote significance level at 1%, 5% and 10% respectively. (1) (2) (3) (4) ln(#gf) ln(#bf) ln(gf) ln(bf) Emigrants *** *** (0.823) (3.001) (-0.081) (5.910) Emigrants D T ertiary 4.274*** 5.627*** 7.121*** 5.420*** (4.870) (5.885) (3.301) (2.970) Observations R-squared

21 3.3.3 Host Country Characteristics Emigrants networks substitute for trust when the contract enforcement is weak by deterring opportunitism (Rauch 2001). In countries where the enforcement of contract is less efficient, the emigrants networks are expected to be more closely related to OFDI. We employ three variables to measure the efficiency of contract enforcement in the host country. The first is Law it, the log difference between the percentile rank of the rule of law in host country i and China at period t. The higher the percentile rank of the rule of law from World Governance Indicators (WGI), the more reliable the rules of society, and the higher the quality of contract enforcement, property rights, the police, and the courts. The second measure is Cost it, the log difference in the cost (i.e. court fees and attorney fees) of resolving a commercial dispute through a local first-instance court as a percentage of the debt value, between host country and China at period t. The greater the cost of contract enforcement from World Bank (WB) Doing Business Data, the less efficient the contract enforcement is. The third measure is Trust it, the log difference in the social trust between host country and China at period t. Following Pevzner, Xie and Xin (2015), the social trust is measured by the number of individual answering most people can be trusted to the survey question in World Value Survey (WVS) Generally speaking, would you say that most people can be trusted or that you have to be very careful in dealing with people?, weighted by the total number of respondents in that country. In a more trust worthy society, the quality of contract enforcement is expected to be higher. We take into account of the host country characteristics that affect the efficiency of contract enforcement and present the OLS estimation results in Table 7. Panel A shows that the the interaction between Emigrants and Law is negative and statistically significant for all four measures of OFDI. It means that the association between emigrants networks and OFDI is weaker in countries with more reliable rule of law. Panel B shows that the interaction between Emigrants and Cost is positive but not statistically significant, for all four measures of OFDI. It suggests that relation between emigrants network and OFDI is stronger in countries with higher cost of resolving disputes. The evidence is however not significant. In Panel C, we observe that the interaction between Emigrants and Trust is positive for greenfield investment 21

22 (GF and #GF) but negative for brownfield investment (BF and #BF). The estimated coefficient of the interaction term is however only statistically significant for #BF. In a more trust-worthy society, the emigrants network is more closely associated with the number of brownfield deals. To summarize, our empirical results provide evidence that there is a stronger relation between overseas Chinese network and OFDI in countries with less efficient contract enforcement. 4 Conclusion This paper distinguishes the role of overseas Chinese networks on greenfield and brownfield Chinese OFDI from both information and trust channels. Empirical evidence suggest that overseas Chinese networks are more closely associated with brownfield OFDI than greenfield FDI because the former faces greater information asymmetry and requires more intensive contract enforcement. The results are robust across various model specifications, estimation methodologies, measures of OFDI and ownership structure. The positive relation between overseas Chinese networks and OFDI is found to be more pronounced in knowledge-intensive tertiary industry, and in host countries with weak contract enforcement. The results provide direct evidence that the overseas Chinese networks function through both overcoming information barriers and substituting for trust that enhance contract enforcement. We also document a stronger relation between overseas Chinese networks and OFDI from POEs relative to that from SOEs. 22

23 Table 7: The role of host country characteristics. The variable Law is measured by the log difference between the percentile rank of the rule of law in host country i and China, Cost is the log difference in the cost of resolving a commercial dispute through a local first-instance court between host country and China, and Trust is the log difference in the social trust between host country and China. All regressions control for the same set of control variables described in Table 2. Robust t-statistics in parenthesis are corrected for correlation across observations for a given country. The symbol ***, **, * denote significance level at 1%, 5% and 10% respectively. Panel A: Rule of Law ln(#gf) ln(#bf) ln(gf) ln(bf) Emigrants *** *** *** *** (6.340) (6.120) (4.612) (3.123) Emigrants Law *** *** *** * (-5.641) (-2.754) (-3.872) (-1.729) Law ** (0.298) (0.341) (-0.929) (-2.445) Observations R-squared Panel B: Cost of Contract Enforcement ln(#gf) ln(#bf) ln(gf) ln(bf) Emigrants *** *** (1.541) (6.287) (0.683) (3.943) Emigrants Cost (0.352) (0.840) (0.859) (0.133) Cost (-0.097) (0.240) (0.222) (0.735) Observations R-squared Panel C: Social Trust ln(#gf) ln(#bf) ln(gf) ln(bf) Emigrants *** *** (0.178) (6.307) (-0.297) (3.729) Emigrants Trust ** (0.297) (-2.208) (0.937) (-0.984) Trust ** 0.605** (1.569) (2.409) (2.090) (1.090) Observations R-squared