ETHNIC FIRMS, DIASPORAS AND INTERNATIONAL TRADE

ETHNIC FIRMS, DIASPORAS AND INTERNATIONAL TRADE Massimiliano Bratti Luca De Benedictis Gianluca Santoni 25 April 2016 Abstract Generally, the complementarity between migration and trade is rationalized using the concept of business and social network effect à la Rauch (2001). In this paper, we highlight a new complementary channel through which immigrants can foster international trade: ethnic firms (a.k.a. firms in the host country owned by foreign-borne entrepreneurs). Using data on immigrants location, on the presence of ethnic firms in the manufacturing sector, and on trade flows between Italian provinces (NUTS-3) and more than 200 foreign countries, we assess, though a double instrumental variable strategy, the causal relationship going from diasporas and ethnic firms toward trade flows. Both, the stock of immigrants and their entrepreneurship, have a positive, significant and economically meaningful effect on exports. The effect on imports is only due to diasporas. Keywords: Migration, Trade, Entrepreneurship. JEL Classification: F10, F14, F22, R10. DEMM, University of Milan; IZA (Bonn) and LdA (Milan). E-mail: massimiliano.bratti@unimi.it Corresponding Author. Rossi-Doria Center, University Roma Tre and DED - University of Macerata. E-mail: luca.debenedictis@unimc.it Cepii. E-mail: gianluca.santoni@cepii.fr A preliminary version of the paper has been presented at ITSG Cagliari (July, 2014). We thank Giuseppe de Arcangelis, Gianmarco I.P. Ottaviano and Hillel Rapoport for the insightful comments. We also thank Rosalia Alessi and Antonella Ciccarese for their generous help with the data. Luca De Benedictis wishes to thank EIEF and NHH for hosting him during his visit in 2015. The usual disclaimer applies. 1

1 Introduction The May, 2013 HSBC Global Connections bulletin by the Economist Intelligence Unit (2013) starts with this evocative story: When Tom Chin decided to follow his aunt from Malaysia to British Columbia in 1979 to attend Simon Fraser University, he had no idea that he d become so entranced by the Coastal Mountains that hed never leave. He also had no idea that 33 years later, hed be exporting organic health supplements back to his native country. As President and Founder of British Columbia based Organika Health Products Inc., he now distributes more than 400 products to Malaysia and to more than 22 other countries across the globe. Among the 100 million Generally, the widely reported complementarity between migration and exports (see Parsons and Winters (2014) for a recent review of the evidence) is rationalized using the concept introduced by Gould (1994) and popularized by Rauch (2001) under the name of business and social network effect: Immigrants foster bilateral trade between their area of residence and their country of origin because of their superior knowledge of, or preferential access to, market opportunities in their home-country. Immigrants can also help enforce international contracts with business partners of their homecountries especially if these same countries are characterized by low quality of institutions, corruption and poor implementation of property rights. Family or ethnic links can act as collaterals. Immigrants linguistic skills, their knowledge of bureaucratic loopholes or the familiarity with traditional cultural habits and the way to conduct business, can help to counterbalance the low level of trust that naturally corresponds to relations that by their own nature must take place between distant entities. Firms located in the area of immigrants settlement can take advantage of this incoming knowledge capital. If the acquisition of the necessary knowledge of a specific foreign market is a prerequisite for a fruitful international business, immigrants can help local firms reducing the fixed sunk cost associated with that acquisition. The more the process takes place outside market rules, the more immigrants knowledge can become a precious advantage in the hands of farsighted local entrepreneurs. More or less all contributions to this literature have the business and social network effect as a common stepping stone for the analysis. Differ- 2

ent authors concentrate on different aspects of the immigrant-firm link. On the immigrant side, researchers have mainly considered the role of immigrants educational levels and occupations. As for education, Felbermayr and Jung (2009) estimated positive effects on trade of low and high skilled immigrants, but not of medium skilled immigrants, 1 although the effects of the first two educational categories were not statistically different; more recent work, however, reports larger effects on trade of highly educated migrants (Felbermayr and Toubal, 2012, Muller and Tai, 2012, Sangita, 2013). As for immigrants occupations, Herander and Saavedra (2005) reported that immigrants in skilled occupations had a higher effect on trade; Aleksynska and Peri (2013) focused on immigrants which could play a pivotal role to establish business connections: those employed in managerial and sales jobs. They showed that immigrants in those occupations had an effect over and above the total amount of immigrants, but only on imports. However, by further splitting immigrants in managerial and sales jobs by educational level, the authors found a statistically significant positive effect of highly educated immigrants on both imports and exports. Consistent with the social network effects is also the evidence of Head and Ries (1998) that independent immigrants, who presumably kept stronger ties with their home country, had a larger impact on trade than family immigrants, and of Herander and Saavedra (2005) that older immigrants had a higher effect on trade. On the firm side, Bratti et al. (2014) give evidence that firms located in less developed provinces in Italy receive the greater advantage from the presence of immigrants. The same happens to small firms with respect to large firms. The evidence is however indirect, being based on provincial data and relying on aggregate information on the size distribution of firms at the provincial level. In any case, the evidence goes in favor of a stronger role of immigrants communities the smaller is the firm and the poorer is the geographical area where the community is located. On the contrary, the more productive the firm is or the more developed the province is, the less relevant the role of immigrants knowledge capital seems to be. In short, if local firms are easily able to pay, through market exchanges, the fixed cost of exporting, they will not have much of an advantage in taking immigrants on board as firms ambassadors. Complementary evidence is brought by Bastos and Silva (2012) that using firm-level data from Portugal matched 1 The authors argue that a potential explanation of this last finding is that mediumskilled workers may be predominantly employed in the non-tradeable sector. 3

with historically-determined emigration stocks show that larger the stock of Portuguese migrants in a given destination country the higher the likelihood of export participation of firms. This effect tends to be significantly larger for firms of the Northern regions of Portugal, that are more likely to have a larger number of contacts with emigrants communities abroad, since the majority of Portuguese emigrants come from the North of the country. Conditional on a firm serving a market, the presence of migrant networks appears to be correlated to the intensive margin of bilateral trade: how much the firm sells overseas. Recent work explores more direct evidence of the immigrant-firm link exploiting the information contained in employer-employee data. Hiller (2013) used Danish data and found that firms foreign employment rather than the local presence of foreigners spurs trade. 2 According to the latter micro-level evidence the firm seems to be the relevant place where immigrants exert their trade-enhancing role. In that case the immigrant-firm link can be less conceived as a positive externality, while instead it takes the form of a more traditional transaction where immigrants sell their knowledge capital to the local firm. A possibility which has been largely neglected by the empirical literature on the effect of migration on bilateral trade is that immigrants may exploit their superior knowledge of the home country by directly trading with it. An important omitted variable in the gravity models would be in this case the number (or share) of foreign firms present in a given region. This channel is potentially able also to explain both the evidence in Hiller (2013) since ethnic entrepreneurs could be more likely to also hire ethnic workers and export at the same time by exploiting their business connections with the origin country, and the findings of Aleksynska and Peri (2013) since highly educated foreign managers and sales persons are more likely to work for foreign firms. Controlling for both the stock of immigrants and the stock of immigrant firms in the gravity equation, we will be able to disentangle the two potential channels, the one going through social networks, that is knowledge flows from foreign individuals toward native firms (or foreign firms 2 Hatzigeorgiou and Lodefalk (2011) also investigated the relationship between immigrant employment and firm s exports (in Sweden) but did not control for regional immigration. They find a positive effect of foreign employees on exports. 4

of a different ethnicity ) from the effect of immigrant firms, which can be explained in terms of their direct exploitation of market knowledge or as (foreign) firms to (native) firms knowledge spillovers (business networks). A point in which the past literature is still not well developed is the identification of the exact channel through which immigrants affect trade. In particular, an interesting issue is to establish whether the social interactions and knowledge flows promoting trade take place inside or outside the firm. Our paper is related to Combes et al. (2005) who studying trade between French departément defined networks of firms using the number of plants owned by the same business group located in the origin and destination departément for trade flows, and found that migrants increased trade flows by a factor of two while networks of firms had an effect which was twice as large as that of migrants in some specifications. Similarly, Greaney (2005) found that Japanese affiliates in the US traded more than 130 times more with Japan than elsewhere. The main differences of our work with Combes et al. (2005) and Greaney (2005) are that we use in our paper a more comprehensive definition of business networks, considering the ethnic ownership of the firms and not their participation in the same business group, and unlike Combes et al. (2005) we consider foreign trade of Italian provinces rather than interregional trade within a country. In this paper, we highlight a new channel through which immigrants can foster exports from the area where they locate to their country of origin: immigrants can become entrepreneurs. Using data on immigrants location in Italian provinces (NUTS-3), on the formation of immigrants firms, and on export flows between Italian provinces and more than 200 foreign countries, we aim at assessing the causal relationship going from immigrants and immigrants firms toward export flows, and find a positive, significant and economically meaningful relationship between immigrants entrepreneurs and exports. The structure of the paper is as follows. The next section set the stage for the analysis introducing the main actor and its overall characteristics. Section 2 describes the main features of the data we use. Section 4 briefly summarizes the part of the literature on immigrants and trade which is more relevant for the current study. Section 6 describes the empirical strategy that we follow to investigate the effect of immigrants firms on trade, and comment on the main findings, while Section 7 concludes. 5

2 Immigrants as entrepreneurs The 22nd of May, 2013 bulletin HSBC Global Connections, by the Economist Intelligence Unit starts with this evocative story: When Tom Chin decided to follow his aunt from Malaysia to British Columbia in 1979 to attend Simon Fraser University, he had no idea that hed become so entranced by the Coastal Mountains that hed never leave. He also had no idea that 33 years later, hed be exporting organic health supplements back to his native country. As President and Founder of British Columbia based Organika Health Products Inc., he now distributes more than 400 products to Malaysia and to more than 22 other countries across the globe. Among the 100 million 3 Data and descriptive statistics The empirical analysis is performed combining three publicly available datasets on province-level export flows, foreign born residents 3 and foreign entrepreneurs in manufacturing sectors for the period 2002-2011. Export flows report the value, originally recorded in euros, of custom transactions between Italian provinces and around 210 destination countries, while data on foreign born residents cover 187 ethnic groups. 4 Concerning foreign born entrepreneurs we use data produced by the National Chamber of Commerce (Infocamere). The responsibilities of the Chamber of Commerce are defined by the Law n. 580 of 1993. 5 The most relevant duty of Infocamere for our analysis is that provincial offices are in charge of producing and keeping the registry of all active firms in Italy. We focus on individually-owned firms (impresa individuale, individual firms or individual enterprises hereafter) a form of business that recognizes in a single individual the whole legal and financial firm representation. For individual firms, we can associate firm ownership to a unique person and nationality, and such data can be used to analyze bilateral trade using gravity models. Individual firms are the most common legal form of Italian firms. At the end of 2013 around 54% of all active firms in Italy were 3 Data are collected by the Italian National Institute of Statistics (ISTAT). 4 For a detailed description of this two data base see Bratti et al. (2014) that employ the same data up to 2009. 5 Further reformed on February 2010 by Law Decree DL23. 6

individual firms. 6 At the geographical level the distribution of individual firms is extremely highly correlated with the overall firms distribution see Figure 1 suggesting that our variable can be interpreted as a proxy for the overall entrepreneurial activity existing at province level. We maintain here this high correlation between the number of individual firms and the total number of firms will hold both on native and immigrants owned business. 7 The overall number of foreign owned firms increased substantially over the last decade at an annualized rate of 4.4% countervailing the decrease in domestically owned individual enterprises, -4.5% per year 8, and accounting for 8.9% of overall individual firms in 2011 (see Figure 2). From a closer look at the evolution of the time series, it emerges that the greatest contribution to the sharp rise in foreign entrepreneurs comes from Eastern European and exta-eu countries. Individual firms owned by 2002 EU residents decline at an annual pace of -2.45% whereas the same figure for extra-eu countries reports an annualized increase of 6.04% over the same period (2002-2011). Table 1 reports the evolution of the foreign presence both as residents and individual entrepreneurs for the first twenty nationalities (in 2011). As expected the evolution of the two time series is strongly correlated even if not perfectly the Pearson (unconditional) correlation among foreign individual firms and residents is 76%. Finally, Table 2 reports descriptive statistics of the estimation sample for our main variables of interests. 6 According to Infocamere. 7 We might expect nonetheless a larger prevalence of individual firms among immigrants than natives. 8 Annual changes are computed as compound annual growth rate. 7

Figure 1: Correlation among Individual and Overall Firms distribution, province level, estimation sample 2002-2011 Active Firms 9 10 11 12 13 6 7 8 9 10 Individual Firms Active Firms Fitted values Note: The figure plots the (log) of individual firms in manufacturing industries by province and year on the population of active firms, irrespective of the legal forms. Simply regressing the log of firm population by province and year on the log of individual firms returns an highly significant coefficient of 0.90 (SE 0.035) and a R 2 of 0.89. 8

Figure 2: Evolution of foreign owned firms (share of total firms) 0.02.04.06.08.1 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Whole Economy Manufacturing Note: The figure plots the share of the foreign owned individual firms on the total number of individual firms for the whole economy and the manufacturing sector. 9

Figure 3: Individual firms for 1,000 inhabitants Natives Firms Migrants Firms [6.31, 10.22] [5.73, 6.30] [5.13, 5.63] [4.60, 5.12] [3.39, 4.57] [15.27, 98.23] [9.62, 15.13] [7.11, 9.58] [5.66, 6.96] [2.99, 5.54] For 1,000 Natives; mean values 2002-2011 For 1,000 Migrants; mean values 2002-2011 Note: The figure plots the incidence of individual firms on the total number of inhabitants, native (left panel) or migrants (right panel), by province. Foreign and Native firms distribution correlates positively at the province-by-country level. Regressing the log of foreign owned individual firms on the log of native firms given a significant coefficient of 0.053 (SE 0.012) and an R2 of.135, after controlling for origin-by-time and province fixed effects. 10

Table 1: Foreign Individual Firms and residence: top 20 origin countries in 2011 iso3 MF# MF pp Share M# M pp Share MF pp M pp CHN 14733 47.33 192867 4.87 12.35 7.60 ROM 1873 6.02 815197 20.60 30.63 19.43 CHE 1499 4.82 7720 0.20-2.09-1.32 MAR 1418 4.56 397448 10.05 7.32 9.34 DEU 1119 3.59 34075 0.86 0.70-1.40 ALB 1076 3.46 440365 11.13 8.53 11.29 TUN 682 2.19 81611 2.06 4.13 3.10 FRA 646 2.08 23366 0.59-0.18-3.64 ARG 522 1.68 7647 0.19-1.98-2.26 YUG 515 1.65 85306 2.16 5.24-0.52 EGY 440 1.41 64751 1.64 8.61 4.38 SEN 428 1.38 71688 1.81 7.91 3.46 BGD 403 1.29 79722 2.01 16.64 9.64 PAK 327 1.05 67206 1.70 13.47 8.86 VEN 308 0.99 4645 0.12 4.10-1.27 BRA 283 0.91 36484 0.92 8.30 1.12 BEL 268 0.86 4708 0.12 0.40-3.20 GBR 228 0.73 22105 0.56 1.99-2.37 NGA 221 0.71 47725 1.21 9.97 3.20 UKR 200 0.64 175383 4.43 53.88 17.78 Top 20 302556 87.35 2744355 67.23 9.50 4.16 Total 361440 100.00 3956454 100.00 10.27 0.78 Note. M refers to foreign born residents while M F to foreign owned individual firms. Countries are ordered according to the individual firms share for the top 20 nationalities in 2011. The last two columns report the average annual growth rates between 2002 and 2011. 11

Table 2: Descriptive Statistics Variable N >0 Mean p25 p50 p75 Max Exports Mln US$ 141453 28.31 0.10 0.98 9.07 6972.78 M 120519 261 3 15 74 153556 MF 72323 4 0 0 2 3567 F 200850 2793 1237 2035 3261 18301 Dist Km 200850 6068.71 2576.56 5125.04 8484.74 19017.43 Note. M refers to foreign born residents, MF to foreign owned individual firms while F to domestic owned individual firm, all variables are in level. N stands for the number of non-zero cells at province by country by year level. 4 Previous literature on immigrants and trade A large number of studies, comprehensively reviewed in Parsons and Winters (2014), has investigated the trade-enhancing effect of immigrants. Most studies have found a positive correlation between the stock of immigrants and both import and export flows; a few of them have also made attempts to qualify the association as causal. Identification of causal effects has been based on instrumental variables (IVs) estimation. Some studies used the instrument based on immigrant enclaves popularized by Altonji and Card (1991). 9 Examples include Peri and Requena-Silvente (2010) and Bratti et al. (2014), which both found positive effects of immigrants on trade for Spain and Italy, respectively. More recently, Parsons and Vézina (2014) have exploited a quasi-natural experiment (the Vietnamise boat people who established in the US) to identify the causal effect of immigrants on trade, confirming the results of earlier studies. Up to now, researchers have mainly considered the role of immigrants educational levels and occupations. As for education, Felbermayr and Jung (2009) estimated positive effects on trade of low and high skilled immigrants, but not of medium skilled immigrants, 10 although the effects of the first 9 The idea is that immigrants tend to follow early immigrants of the same nationality in their location choices. 10 The authors argue that a potential explanation of this last finding is that mediumskilled workers may be predominantly employed in the non-tradeable sector. 12

two educational categories were not statistically different; more recent work, however, reports larger effects on trade of highly educated migrants (Felbermayr and Toubal, 2012, Muller and Tai, 2012, Sangita, 2013). As for immigrants occupations, Herander and Saavedra (2005) reported that immigrants in skilled occupations had a higher effect on trade; Aleksynska and Peri (2013) focused on immigrants which could play a pivotal role to establish business connections: those employed in managerial and sales jobs. They showed that immigrants in those occupations had an effect over and above the total amount of immigrants, but only on imports. However, by further splitting immigrants in managerial and sales jobs by educational level, the authors found a statistically significant positive effect of highly educated immigrants on both imports and exports. Consistent with the social network effects is also the evidence of Head and Ries (1998) that independent immigrants, who presumably kept stronger ties with their home country, had a larger impact on trade than family immigrants, and of Herander and Saavedra (2005) that older immigrants had a higher effect on trade. A point in which the past literature is still not well developed is the identification of the exact channel through which immigrants affect trade. In particular, an interesting issue is to establish whether the social interactions and knowledge flows promoting trade take place inside or outside the firm. Recent work has exploited employer-employee data to give an answer to this important question. Hiller (2013) used Danish data and found that firms foreign employment rather than the local presence of foreigners spurs trade. 11 According to the latter micro-level evidence the firm seems to be the relevant place where immigrants exert their trade-enhancing role. A possibility which has been largely neglected by the migration and trade gravity literature is that immigrants may exploit their superior knowledge of the home country by directly trading with it. An important omitted variable in the gravity models would be in this case the number (or share) of foreign firms present in a given region. This channel is potentially able also to explain both the evidence in Hiller (2013) since ethnic entrepreneurs could be more likely to also hire ethnic workers and export at the same time by exploiting their business connections with the origin country, and the findings of Aleksynska and Peri (2013) since highly educated foreign managers and sales 11 Hatzigeorgiou and Lodefalk (2011) also investigated the relationship between immigrant employment and firm s exports (in Sweden) but did not control for regional immigration. They find a positive effect of foreign employees on exports. 13

persons are more likely to work for foreign firms. Controlling for both the stock of immigrants and the stock of immigrant firms in the gravity equation, we will be able to disentangle the two potential channels, the one going through social networks, that is knowledge flows from foreign individuals toward native firms (or foreign firms of a different ethnicity ) from the effect of immigrant firms, which can be explained in terms of their direct exploitation of market knowledge or as (foreign) firms to (native) firms knowledge spillovers (business networks). Our paper is related to Combes et al. (2005) who studying trade between French departément defined networks of firms using the number of plants owned by the same business group located in the origin and destination departément for trade flows, and found that migrants increased trade flows by a factor of two while networks of firms had an effect which was twice as large as that of migrants in some specifications. Similarly, Greaney (2005) found that Japanese affiliates in the US traded more than 130 times more with Japan than elsewhere. The main differences of our work with Combes et al. (2005) and Greaney (2005) are that we use in our paper a more comprehensive definition of business networks, considering the ethnic ownership of the firms and not their participation in the same business group, and unlike Combes et al. (2005) we consider foreign trade of Italian provinces rather than interregional trade within a country. 5 Identification of the effect of migrants and ethnic firms on trade A well known problem in the trade-migration literature is that unobservable variables affecting migrants location choices may be correlated with those influencing trade, determining an endogeneity bias. The common solution to this problem consists in exploiting a presumably exogenous source of variation in immigrants locations using an IVs strategy. When the independent variable of interest is the stock of migrants, this variation is generally provided by migrants enclaves. The idea is to use the geographical distribution of migrants by ethnicity in the past to apportion nationwide current flows of migrants to different regions. This was the instrument proposed back in the past by Altonji and Card (1991) in their study of the effect of immigrants on natives labor market outcomes, which has already been widely used in 14

trade-migration studies (see, for instance, Peri and Requena-Silvente, 2010, Bratti et al., 2014). 12 The underlying idea is that the presence in Italy of individuals coming from the same foreign country may provide useful information about the host country to the potential migrant, reduce relocation costs and increase the potential benefits of migration. In our case, identification is complicated by the presence of another potentially endogenous variable: the stock of migrant-owned firms (which we will refer to as ethnic firms for brevity). To tackle the endogeneity of this variable we build an instrument as follows. The idea is to exploit comparative advantage by ethnicity deriving by several factors (e.g., culture, the country of origin s level of development) in starting certain types of activities. The focus of our analysis is on manufacturing firms. Let us define Ser ijt the number of firms in services in province i owned by citizens of country j at time t, while the number of the corresponding manufacturing firms is defined as Man ijt. We take as a proxy of the relative advantage in starting manufacturing vs. service businesses for ethnicity j in province i in a base year back in the past (e.g., 1997), i.e. RA ij97 = Man ij97 /Ser ij97. The total stock of active service firms owned by nationals of country j in year t in Italy is defined as Ser jt = i Ser ijt. The relative weight of province i in the total service businesses owned by ethnic group j in Italy in 1997 is defined as w ij97 = Ser ij97 /Ser j97. Then using the usual shift and share mechanism, we build the predicted number of manufacturing business as: P redman ijt = (Man ij97 /Ser ij97 ) (Ser jt w ij97 ). (1) How does the instrument work? Say that in year 1997 in Italy Indians owned 0.4 businesses in Manufacturing for every businesses in Services. Then given that in year t Indians owned 1000 service businesses in the whole Italy, and that Milan accounted for 15% of overall Indian s service businesses in Italy in 1997, our instrument takes the value of 0.4*1000*0.15=60, i.e. the number of manufacturing firms we should expect in Milan if Indians create Manufacturing businesses according to the 1997 relative rate. The relative 12 Some recent studies exploit instead quasi-natural experiments provided by the random allocation of refugees across US states (Parsons and Vézina, 2014, Steingress, 2015). Even if considering specific episodes of migration generally allows for a convincing identification, refugees are a minority of migrants and often very different from economic migrants, and their effect on trade may not be easily generalizable. 15

weight of each province remains fixed at 1997 (w ij97 ) to avoid the dynamic of trade in the estimation period potentially impacting on it. The potential advantages of building the instrument following the expression (1) rather than simply reallocating manufacturing businesses as in the enclave instrument used for migrants are that: (i) the time variant ethnic specific component of the instrument Ser jt does not refer to manufacturing but to services, whose dynamic is probably more exogenous to trade in manufacturing, even at the nationwide level, than that of manufacturing firms; (ii) by taking the ratio Man ij97 /Ser ij97 we partly neutralize factors which are province-ethnicity (i.e. trading-pair) specific. Assume that Man ij97 /Ser ij97 = (θ ij f(.))/(θ ij g(.)) where f(.) and g(.) are some unknown functions of exogenous variables while θ ij is a measure of country j nationals ability of opening business in province i, depending for instance on cultural affinity or oppeness of the native population of province i towards ethnic group j. θ ij may also be correlated with nationals i ability of doing trade between province i and their country. Natives in i may, for instance, be more prone to buy products coming from j or sell products to this country since the corresponding ethnic group is well integrated and trusted. Taking the ratio, θ ij would cancel out in this multiplicative form. Also in non-multiplicative forms, taking the ratio should attenuate the impact of trading-pair unobservables. 6 The econometric model In order to investigate the role of the stock of immigrants and the number of ethnic firms on export flows we employ a gravity equation à la Anderson and van Wincoop (2003) 13 X ijt = δ it +θ jt + α 1 ln(m i,j,t ) + α 2 ln(mf i,j,t )+ + α 3 ln(d ij ) + α 4 Border i,j,t + ɛ i,j,t (2) where i is the subscript for Italian provinces (NUTS-3), j indicates the foreign country (i.e., the country of origin of immigrants), and t stands for time. δ i is 13 The Anderson and van Wincoop (2003) specification of the gravity equation can be derived from micro-foundations, and results from an expenditure function that takes into account the fundamental role of general equilibrium effects in trade: aka, the multilateral resistance index. See De Benedictis and Taglioni (2011), Anderson (2011) and Head and Mayer (2015) on the theoretical foundation of the gravity equation. 16

a province fixed effect, θ j is a foreign country fixed effects, and φ t time fixed effects. X i,j,t is trade (exports or imports) between province i and country j at time t (including zero-trade observations). Y i,t and Y j,t are province and foreign countries GDPs at time t 1, respectively, and M i,j,t and MF i,j,t are the stocks of immigrants and immigrant firms, respectively, from country j located in province i, acting as a trade-enhancing force in contrast to d ij, which is the great-circle distance between province i and country j. C i,j,t is a vector of additional controls, including a a dummy for contiguity between the Italian province i and the foreign country j, a border dummy, which are included to take into account possible nonlinearities in distance, and a timevarying dichotomous indicator for participation in the EU or EFTA. We also include in the gravity model the total number of firms in province i at time t, excluding ethnic firms from country j (so as this covariate has province, country and time variation). The role of this regressor is to avoid that the number of immigrants firms captures some unobserved factor a the province level, such as easyness of doing business or entrepreneurship, which may spur trade. Finally, ɛ ijt is an error term clustered at the province-country level. Data cover the 2002-2010 period. Table 3: Baseline Estimation, sample 2002-2011 OLS PPML Dep Variable: ln(exp ijt ) > 0 ln(imp ijt ) > 0 X ijt 0 X ijt 0 (1) (2) (3) (4) ln(mf i,j,t ) 0.075*** 0.162*** 0.049*** 0.033 (0.015) (0.022) (0.019) (0.024) ln(m i,j,t ) 0.102*** 0.096*** 0.070*** 0.198*** (0.009) (0.016) (0.020) (0.032) ln(d ij ) -1.129*** -1.609*** -0.721*** -0.958*** (0.065) (0.081) (0.121) (0.183) Border ij 0.405** 0.107 0.220** 0.236 (0.198) (0.190) (0.109) (0.184) Observations 141,091 94,084 199,614 194,773 R2 0.796 0.673 0.890 0.921 FEs it; jt it; jt it; jt it; jt Standard errors in parentheses clustered at province-by-country level, *** p<0.01, ** p<0.05, * p<0.1. All variables are at time t (no lagged covariates). Since the log-log version of the gravity model has been recently subject 17

to some criticism, we use the Poisson Pseudo Maximum Likelihood (PPML) estimator proposed by Silva and Tenreyro (2006). 14 A first set of results is reported in Table 3. All specifications in this table control for separate province, country and time fixed effects. Column (1) excludes the number of firms in province i (excluding those of ethnicity j). This column shows statistically significant positive elasticities of export flows with respect to both the stock of immigrants (0.050) and the stock of immigrants firms (0.073). The estimated elasticities do not change after including the number of firms in column (2). In column (3), we report the estimates of the gravity model excluding the two major Italian trade hubs (Milan and Rome), for which trade flows are more likely to be affected by measurement errors (e.g., for trade made through wholesalers) and the results are practically unvaried. 15 Table 4: IV Estimation, sample 2002-2011 2SLS O IV-PPML Dep Variable: ln(exp ijt ) > 0 ln(imp ijt ) > 0 X ijt 0 X ijt 0 (1) (2) (3) (4) ln(mf i,j,t ) 0.059** 0.164*** 0.065** 0.019 (0.025) (0.038) (0.032) (0.036) ln(m i,j,t ) 0.162*** 0.177*** 0.105*** 0.273*** (0.018) (0.031) (0.033) (0.048) ln(d ij ) -1.103*** -1.583*** -0.706*** -0.912*** (0.064) (0.082) (0.124) (0.183) Border ij 0.326* -0.019 0.171 0.182 (0.192) (0.182) (0.109) (0.183) Observations 141,091 94,084 199,614 200,850 F-Test 1981 1546 FEs it; jt it; jt it; jt it; jt Standard errors in parentheses clustered at province-by-country level, *** p<0.01, ** p<0.05, * p<0.1. All variables are at time t (no lagged covariates). In Table 4 we go one step further toward saturating the gravity model. In 14 The debate on the most appropriate nonlinear estimator to be applied when zeros are a relevant proportion of trade flows is still very open. See De Benedictis and Taglioni (2011), Baltagi et al. (2014) and Head and Mayer (2015) on this specific point of the gravity literature. If the stock of immigrants and immigrants firms are included in logarithm in the graivity equation, their coefficients can be interpreted as elasticities. 15 Province fixed effects are likely to partly address this issue. 18

particular, we replace the time-invariant province and country fixed effects with three-year (time variant) province and country fixed effects to take better account of multilateral trade resistance factors. In analogy with the previous table, column (1) does not include F i, j,t. In this specification both immigrants firms and immigrants stocks are statistically significant, with estimated elasticities of 0.092 and 0.061, respectively. Including F i, j,t in column (2) does not change the overall picture. When we allow for the effect of distance to be nonlinear, by including dummies for quantiles of log-distance, the estimated elasticities are largely unaffected (column (3)). Excluding Rome and Milan, in this last specification, slightly increases the elasticities of immigrants and reduces that of migrant firms (column (4)). The last column of Table 4 reports the log-log specification estimated only on the positive trade flows. In this specification the estimated elasticities are larger, 0.11 for migrant firms and 0.08 for the stock of immigrants, and always statistically significant at the 1% level. Gianluca, please comment on Table?? Overall, the results in this section indicate that both the stock of immigrants and the stock of immigrant firms significantly contribute to spur foreign trade of Italian provinces. As ethnic firms of country j are also more likely to hire workers from country j, and under the assumption that an entrepreneur (or firm owner) of ethnicity j has little to learn, in terms of business knowledge relevant to trade with country j, from her co-nationals, we tend to interpret the effect of immigrants as the role exerted either on natives firms (i.e. firms with Italian ownership) or firms of ethnicity different from j. As we anticipated, given the province level of our analysis, we are not able to say whether MF i,j,t is capturing the fact that ethnic entrepreneurs directly exploit their information advantage by exporting, or whether the effect is partly due to spillovers between firms of ethnicity j and either natives or other ethnic firms. We can however safely conclude, at this stage, that ethnic entrepreneurship is positively associated with Italian exports. 16 7 Concluding remarks 16 In the future, we will make an attempt to identify causal effects using an instrumental variables (IVs) strategy. 19

Acknowledgements. Preliminary and incomplete, comments are welcome. Not for quotation or circulation. Comments are welcome. The usual disclaimer applies. 20

References Aleksynska M. and Peri G. (2013) Isolating the network effect of immigrants on trade, The World Economy, Early View. Altonji J.G. and Card D. (1991) The Effects of Immigration on the Labor Market Outcomes of Less-skilled Natives, in: Immigration, Trade and the Labor Market, National Bureau of Economic Research, Inc, NBER Chapters, 201 234. Anderson J.E. (2011) The Gravity Model, Annual Review of Economics, 3, 133 160. Anderson J.E. and van Wincoop E. (2003) Gravity with Gravitas: A Solution to the Border Puzzle, American Economic Review, 93, 1, 170 192. Baltagi B.H., Egger P. and Pfaffermayr M. (2014) Panel data gravity models of international trade, Technical Report 4616, CESifo Working Paper. Bastos P. and Silva J. (2012) Networks, firms, and trade, Journal of International Economics, 87, 2, 352 364. Bratti M., Benedictis L.D. and Santoni G. (2014) On the pro-trade effects of immigrants, Review of World Economics (Weltwirtschaftliches Archiv), 150, 3, 557 594. Combes P.P., Lafourcade M. and Mayer T. (2005) The trade-creating effects of business and social networks: evidence from France, Journal of International Economics, 66, 1, 1 29. De Benedictis L. and Taglioni D. (2011) The Gravity Model of International Trade, in: The Trade Impact of European Union Preferential policies: An Analysis through Gravity Models, De Benedictis L. and Salvatici L., eds., Springer, chapter 4, forthcoming. Economist Intelligence Unit (2013) How immigrants spur cross-border trade, Technical report, The Economist, https://globalconnections.hsbc.com/us/en/articles/how-immigrantsspur-cross-border-trade. Felbermayr G.J. and Jung B. (2009) The pro-trade effect of the brain drain: Sorting out confounding factors, Economics Letters, 104, 2, 72 75. 21

Felbermayr G.J. and Toubal F. (2012) Revisiting the trade-migration nexus: Evidence from new oecd data, World Development, 40, 5, 928 937. Gould D.M. (1994) Immigrant links to the home country: Empirical implications for U.S. bilateral trade flows, The Review of Economics and Statistics, 76, 2, 302 16. Greaney T.M. (2005) Measuring network effects on trade: Are japanese affiliates distinctive?, Journal of the Japanese and International Economies, 19, 2, 194 214. Hatzigeorgiou A. and Lodefalk M. (2011) Trade and Migration: Firm-Level Evidence, Working Papers 2011:6, Örebro University, School of Business. Head K. and Mayer T. (2015) Gravity equations: Workhorse, toolkit, and cookbook, in: Handbook of international economics, Vol.4, Gopinath, Helpman and Rogoff, eds., Elsevier. Head K. and Ries J. (1998) Immigration and Trade Creation: Econometric Evidence from Canada, Canadian Journal of Economics, 31, 1, 47 62. Herander M.G. and Saavedra L.A. (2005) Exports and the structure of immigrant-based networks: the role of geographic proximity, Review of Economics and Statistics, 87, 2, 323 335. Hiller S. (2013) Does immigrant employment matter for export sales? Evidence from Denmark, Review of World Economics (Weltwirtschaftliches Archiv), 149, 2, 369 394. Muller T. and Tai S.H.T. (2012) The role of migration networks in international trade a sectoral approach,, conference paper, 5th International Conference on Migration and Development, Paris 2012. Parsons C. and Vézina P.L. (2014) Migrant Networks and Trade: The Vietnamese Boat People as a Natural Experiment, Economics Series Working Papers 705, University of Oxford, Department of Economics. Parsons C.R. and Winters L.A. (2014) International migration, trade and aid: A survey, International Migration Institute, University of Oxford, WP-90-2014. 22

Peri G. and Requena-Silvente F. (2010) The trade creation effect of immigrants: evidence from the remarkable case of Spain, Canadian Journal of Economics, 43, 4, 1433 1459. Rauch J.E. (2001) Business and Social Networks in International Trade, Journal of Economic Literature, 39, 4, 1177 1203. Sangita S. (2013) The effect of diasporic business networks on international trade flows, Review of International Economics, 21, 2, 266 280. Silva J.S. and Tenreyro S. (2006) The Log of Gravity, Review of Economics and Statistics, 88, 4, 641 658. Steingress W. (2015) The Causal Impact of Migration on US Trade: Evidence from a Natural Experiment, IZA Discussion Papers 9058, Institute for the Study of Labor (IZA). 23