DO REGIONAL ASYMMETRIES MATTER FOR THE TRADE- MIGRATION LINK?

DO REGIONAL ASYMMETRIES MATTER FOR THE TRADE- MIGRATION LINK? José V. BLANES, Joan MARTÍN-MONTANER and Guadalupe SERRANO FIRST DRAFT Abstract Empirical evidence has shown that the arrival of foreign workers tend to enhance the bilateral trade between the source and the receiving country. However, the fact that immigrants are not uniformly distributed across regions in the host country has not been taken into account until very recently. Therefore, this increase in bilateral trade should be asymmetrical at the regional level. Under this assumption, we analyse the trade of the Spanish regions with the main source countries of immigration to Spain. In our study, we consider immigrant characteristics (such as the educational level and the type of obs occupied), together with variables regarding country (and region) characteristics. Keywords: international trade, migration, regional economics JEL codes: F14, F22 CORRESPONDING AUTHOR: Joan Martín-Montaner Department of Economics Universitat Jaume I de Castelló C/ Vicente Sos Baynat, s/n 12071 Castelló Tel. +34 964 72 86 09 Fax. +34 964 72 85 91 e-mail: martin@eco.ui.es

I. INTRODUCTION In the last few years, empirical evidence has been provided on the existence of a link between immigration and trade. According to the research by Gould (1994), Head and Ries (1998) or Girma and Yu (2002), the presence of immigrant population has a positive effect on trade flows between the host and the home countries. The theory of international trade, most of all in a monopolistic competition framework, provides strong arguments for such a complementary relationship to exist. Notwithstanding this, it has been in the network literature where most of the research on this topic has leaned on as the theoretical framework (see Rauch (1999)). Thus, the higher bilateral trade between the receiving and the source country could be caused by the bias in the immigrants preferences towards home-country products or by reductions in transaction costs derived from the higher knowledge about their native markets. The former (demand effect) implies more imports from the native country, whereas the latter (network effect) could imply more imports, more exports or both. That means that depending on which mechanism has a bigger effect on trade flows, immigration can induce either inter-industry or intra-industry adustment of host country production. Both effects mentioned above will be likely affected by how immigrants are located in the host country. The key point is to find out whether higher concentration of workers coming from the same country in one region involves that the effect on this region s trade flows surpasses the effect at a national level. Thus, this likely regional bias in the trade adustments (and the relocation of economic activities that they involve) is an additional and key point to be taken into account in any policy design focused on the reduction of regional inequalities. This regional approach can also be found in some recent papers such as Wagner, Head and Ries (2002) and Coombes, Lafourcade and Mayer (2005). We analyse these issues in the case of Spain. Its recent role as final destination for immigrants together with its distribution of immigrant s home-countries and its intermediate development level between developing countries and the most developed countries makes it an ideal candidate to our purposes. Moreover as it is a small number of regions which concentrate the foreign population. Therefore, we want to analyse the

effect of immigration on the type of regional trade adustments. To do this, we use data regarding bilateral trade of Spanish regions with a number of non-eu countries. The set of explanatory variables combines immigrant and product characteristics, together with several variables to control for country and regional characteristics. The relationship between immigration and trade adustment has been analysed for the Spanish case by Blanes (2005) and Blanes and Martín-Montaner (2006). In these papers trade data were explained at both bilateral and industry level, considering individual characteristics of immigrants related to the role they play in the Spanish labour market. This allowed us to deeper explain not only the existence of a link between immigration and trade but the mechanisms behind this link. However, since the Spanish economy is not homogeneous but rather it differs among its regions, the introduction of the space dimension in the study will allow us to determine until what point these effects depend on the different concentration of the immigrants of determined nationality or, on the contrary, if this concentration is irrelevant for this nexus between immigration and trade appears. II. SOME THEORETICAL CONSIDERATIONS There is a debate concerning whether labour movements and trade are complements or substitutes 1. The Factor Prize Equalization Theorem provides a strong inference that trade and immigration are substitutes, as far as we are working under the Heckscher-Ohlin assumptions. Thus, as people moves from the source country to the host country, their relative factor endowments become more similar (provided there are no changes in the capital stocks). This means that there will be no room for trade based upon comparative advantages. However, things can be rather different if the bilateral trade is intra-industry type, and it depends on the existence of scale economies and product differentiation. In this context, trade could appear even when the trade partners have exactly the same factor endowments, as it is no comparative advantage-induced trade. Therefore, it is likely to find a complementary relationship between international migration and trade. 1 See, for example, Faini et al. (1999).

One common feature to all the situations above is that migration means a change in the labour supply in both the source and the host country, without any reference to the composition of the labour supply. As workers are assumed to have the same preferences and there is perfect information in all cases, it does not make any difference whether the change in the number of worker happens because of migratory flows or because variations in the birth rates. However, introducing market imperfections, such as information asymmetries or incomplete markets, provides a rationale for international trade being enhanced by immigration because the latter changes the national composition of the labour supplies. From the existing literature, we can easily postulate two ways this enhancing effect might act through. First, the potential bias of the consumption preferences of the foreign workers towards their native country products could give raise to a demand effect. The obvious consequence would be an increase in the imports of the host country. The second (and more interesting) effect relates trade with the level and quality of information about foreign markets. Thus, immigrants are assumed to display a higher knowledge of their native markets which might reduce transaction costs of bilateral trade. Girma and Yu (2002) stand out two types of mechanisms: individual-specific and non individual-specific. The former implies an active performance of immigrants by means of business connections, whereas the latter originates from a more general knowledge of the social institutions of the home country. In both cases, a simultaneous increase in imports and exports of the host country could be expected. In terms of the usual international trade terminology, we could express the effects above saying that the demand effect would translate into higher inter-industry trade, whereas the network effect could (although not necessarily) imply more intraindustry trade. This perspective allows to directly linking the effects of immigration with the literature focused on intra-industry trade. Both fields can be related quite straightforwardly: on the one hand, both theoretical and empirical research has showed that intra-industry trade is mostly caused by the exchange of differentiated products; on the other hand, product differentiation is likely a key issue to allow migration effects on trade to appear. First, the demand effect seems to fit more naturally in those cases where there are different varieties of the same product, so that foreigners have a preference (in the lancasterian sense) for those produced in their home country. Second,

in regard of the network effect, recent papers have shown that differentiated products are more sensible to reductions in transaction costs than homogeneous products (see Rauch, 1999). Several issues must be taken into account regarding the likely effect of migration on trade, concerning both country and migrant's characteristics. First, the influence of newcomers on the connections with the source country will be less noticeable as long as there are already strong commercial connections. In other words, the volume of trade between both countries should reduce the effect of the arrival of immigrants. Secondly, those migrants characteristics which could affect their effectiveness in exploiting their home country linkages should also be taken into account. These characteristics are mostly referred to the quality of the ob and/or the human capital endowments of foreign workers: the higher they are, the more likely they will have (and exploit) connections in their home country. III. THE ECONOMETRIC MODEL In this paper we test first for the existence and relevance of a positive effect of immigrants in the Spanish regions on their bilateral trade with their home countries. Following the previous literature, we use an augmented gravity equation for trade to test the link between immigration and bilateral trade. The basic gravity equation relates positively the volume of trade to the mass of the two countries (region and country in our case) and negatively to the trade costs between them. Our analysis is conducted over the1995-2004 sample period. Trade data have been obtained from the Dirección General de Aduanas; the original data following the TARIC classification were reorganised into consumption goods and production goods according to the BEC system. We use information from the Spanish Encuesta de Población Activa (EPA) for the number of foreign workers. Thus, data on the Spanish region where they work, their educational level and the sector where they work are available. With regard to the macro variables we use regional data and national data. For each Spanish region, GDP at current market prices referred to 1995 base year comes

from the Regional Accounts data base, and population corresponds to the population proection referred to the first of July. Both variables are available in the National Statistics Institute of Spain (www.ine.es). For the rest of the world countries, GDP in current US dollars, exchange rates euro/us dollar, population and gross fixed capital formation in current US dollars are from the United Nations National Accounts Main Aggregates data base. For some of the countries, information of any of these variables in the year 2004 was not available. In those cases, we estimate those missing data by using the correspondent growth rates from the World Bank s Key Indicators Data and Statistics. Capital stocks are calculated from gross fixed capital formation, I, based on the perpetual inventory model: K t = (1-δ) K t-1 + I t where δ is the depreciation rate which is assumed to be of 5% for all Spanish regions and countries 2. Initial capital stock is calculated following the Griliches (1980) procedure: K 0 =I 0 /(δ+g) where g is the averaged annual logarithmic growth of I over the whole available period 1970-2003, and the zero subindex indicates the firs year for which I data is available, that is, 1970. Related to data on gross fixed capital formation in Spanish regions, we use regional estimations from the Ministry of Economy BD-MORES data base, available until 2000 year. We obtain BD-Mores regionalization shares, which are very stable over the sample and forecast those for the 2001-2004 periods by means of movil averages. Then we apply such shares to Spanish Gross fixed capital formation from UN statistics and obtain the gross fixed capital formation for each Spanish region. We use the product of each Spanish region i and trade partner in year t relative to the World GDP in the same year to measure the size of the two countries (gdprel it ): 2 Our assumption of common depreciation rates among regions and countries is considered a first step to enhance international comparability of capital stock estimates (Groote et al., 1996). We use the average depreciation rate estimated for Spanish regions, which implies a 20 years service life of capital. But, as Blades (1993, p. 404) remarks the "use of erroneous service lives does not introduce any systematic bias into capital stock growth rates" and thus, would not affect our international and temporal comparisons.

gdprel it GDPit * GDPt = GDP WORLD, t We also include as explanatory variables the bilateral differences in market size (as measured by the GDP) and factor endowments. We have built up the Balassa indicator of relative inequality between regions and each trade partner dify it Yi = 1 + Yi + Y Yi ln Yi + Y Y + Yi + Y Y i ln Yi + Y defining therefore the variables difgdp it and difkl it. We also include a set of control variables in the form of dummy variables to capture particular characteristics of the partner countries which could influence the bilateral trade flows. Those variables encompass dummies for EU and NAFTA membership (eu i and nafta i, respectively) and for the partner being in the Magreb, Sub-saharian Africa and Latin America (magreb i, afrsub i and latinoam i, respectively). Fixed effects for each Spanish region (DR i ) and each year in the sample (DT t ) are also included. trade it = α + α gdprel 0 + α latinoam 7 1 + δ imm it + it i + α difgdp + α nafta β DR + i 8 2 i t it + α difkl + α farest χ DT t 9 3 t it + μit + α eu + α 10 4 proxor + α magreb 5 + α ocean 11 + α afrsub 6 + α noeu 11 IV. PRELIMINARY RESULTS Our first results are displayed in Table I. Different dependent variables have been considered, total bilateral trade (first column) and exports and imports separately (second and fourth columns). In each case, an additional distinction has been included, considering trade in intermediate and consumption goods for both exports and imports (third and fifth columns, respectively). The dummies for geographical regions, are not displayed, the same as the fixed effects for each region and each year, although they have been included in all the estimations.

As expected, the relative size of trade partners have an significant and positive effect on the volume of trade, whatever the trade flow used as the dependent variable. The differences in GDP are significant in the case of imports of intermediate goods (and in total imports), whereas differences in relative factor endowments are significant in the case of exports (but only when intermediate and consumption goods are taken into account separately). Finally, the explanatory variable immig, which represents the total number of foreign workers in each region coming from the different trade partners are always significant but in one case, which should confirm our main hypothesis. Table I. Total Number of Foreign Workers Total Exports Exports Exports Trade Interm Consump Interm Consump gdprel 2.037 2.713 3.164 3.118 3.343 4.585 3.710 difgdp 0.179 [0.664] -0.306 [0.539] -0.008 [0.988] 0.850 [0.118] 0.994 [0.089] 2.397-0.236 [0.707] difkl -0.027 [0.293] 0.024 [0.344] 0.047 [0.073] 0.057 [0.029] 0.027 [0.320] 0.030 [0.325] -0.016 [0.581] immig 0.006 [0.031] 0.001 [0.843] 0.015 [0.002] 0.013 [0.008] 0.022 0.024 0.047 constant -4.170-13.897-18.830-21.180-20.077-36.115-24.440 R 2 0.32 0.40 0.38 0.45 0.39 0.46 0.40 N. obs. 13906 13906 13906 13906 13906 13906 13906 NOTE: Figures in brackets are p-values. All variables but dummies are in natural logarithms. Our preliminary results suggest that the presence of foreign workers enhance bilateral trade between their region of residence and their country of origin. Next, we replicate the estimations above to include several characteristics of the foreign workers such as their educational level (Table II), and the activity they are developing locally (Table III). Some surprising results can be found. Thus, the higher the number of qualified foreigners (immhigh) goes with lower levels of trade, which is a surprising outcome, as qualified workers are supposed to exploit more efficiently the advantages of existing networks. Further research must be done on this field, probably distinguishing whether

this qualified workers come mostly from countries with high levels of bilateral trade with the Spanish regions or not. Medium educated workers (immedium) have no significant effects on the volume of imports, and a negative effect in the case of the bilateral exports. Only in the case of non qualified workers (immlow), the expected positive effect on trade is achieved. Table II. Foreign Workers by educational level Total Exports Exports Exports Trade Interm Consump Interm Consump gdprel 2.073 2.761 3.219 3.195 3.405 4.665 3.744 difgdp 0.166 [0.687] -0.320 [0.522] -0.022 [0.967] 0.833 [0.125] 0.978 [0.094] 2.376-0.270 [0.667] difkl -0.029 [0.249] 0.021 [0.421] 0.044 [0.098] 0.052 [0.047] 0.023 [0.400] 0.025 [0.419] -0.018 [0.533] immlow 0.015 0.016 0.021 0.026 0.025 0.032 0.045 immedium -0.004 [0.125] -0.010 [0.015] -0.003 [0.556] -0.014 [0.008] 0.002 [0.645] -0.006 [0.406] 0.010 [0.171] immhigh -0.019-0.028-0.024-0.030-0.028-0.030-0.007 [0.261] Constant -4.809-14.095-21.239-22.572-21.105-37.115-24.646 R 2 0.33 0.40 0.43 0.45 0.39 0.46 0.40 N. obs. 13906 13906 13906 13906 13906 13906 13906 NOTE: Figures in brackets are p-values. All variables but dummies are in natural logarithms.

It is also interesting to test the effect of foreign workers on trade taking into account the branch of activity where they work. Obviously only those who were occupied when the survey was obtained are taken into account in this case. Table III. Foreign Workers by labour market dependence Total Exports Exports Exports Trade Interm Consump Interm Consump gdprel 2.097 2.789 3.258 3.250 3.458 4.730 3.819 difgdp 0.250 [0.544] -0.195 [0.695] 0.108 [0.843] 1.016 [0.061] 1.141 [0.051] 2.595-0.115 [0.854] difkl -0.030 [0.242] 0.020 [0.425] 0.043 [0.106] 0.050 [0.054] 0.020 [0.447] 0.021 [0.495] -0.018 [0.533] agric 0.028 0.040 0.037 0.056 0.050 0.053 0.084 extrac 0.021 [0.031] 0.030 [0.012] 0.036 [0.017] 0.051 [0.001] 0.023 [0.145] 0.060 [0.003] 0.033 [0.080] manuf -0.012 [0.001] -0.036-0.043-0.048-0.007 [0.215] -0.020 [0.021] 0.002 [0.080] build 0.002 [0.506] 0.004 [0.428] 0.005 [0.445] 0.006 [0.299] 0.016 [0.021] 0.016 [0.095] 0.030 [0.001] commerce -0.010 [0.001] -0.019-0.019 [0.005] -0.031-0.016 [0.009] -0.028 [0.002] -0.003 [0.662] hotels -0.004 [0.197] 0.012 [0.005] 0.025 0.016 [0.009] -0.024-0.018 [0.035] -0.016 [0.029] transport -0.021-0.025-0.026 [0.001] -0.036-0.061-0.080-0.058 another 0.001 [0.683] -0.005 [0.187] 0.003 [0.572] -0.001 [0.918] 0.018 0.024 [0.003] 0.026 [0.001] constant -4.826-13.966-21.171-22.540-21.843-38.081-24.267 R 2 0.33 0.40 0.43 0.46 0.39 0.46 0.40 N. obs. 13906 13906 13906 13906 13906 13906 13906 NOTE: Figures in brackets are p-values. All variables but dummies are in natural logarithms.

On the one hand, immigrants working in agriculture (agric) and extractive (extrac) activities seem to promote bilateral trade, both exports and imports. On the other hand, the effect is negative if they occupy positions in manufactures (manuf), commerce activities (commerce) or transports (transport). The most peculiar case corresponds to hotel trade (hotels), with a positive effect on exports and a negative one on imports. Finally, people working in building (build) and other services (another) display a positive effect, only significant in the case of imports. Although this latter effect, together with the results for agriculture and extractive activities seem to hint at the existence of a demand effect, the existence of some negative effects deserves further research. REFERENCES Blades, D. (1993). Comparing capital stocks. In Adam Szirmai, bar Van Ark, and Dirk Pilat (eds.) Explaining economic growth: essays in honor of Angus Maddison, Amsterdam: North-Holland. Blanes, J.V. (2005). Does immigration help to explain intra-industry trade? Evidence for Spain. Review of World Economics, vol. 141 (2), pp. 244-270. Blanes-Cristóbal, J.V. and J. Martín-Montaner (2006). Migration flows and intraindustry trade adustment. Review of World Economics, forthcoming. Combes, P.P., M. Lafourcade and T. Mayer (2005). The trade-creating effects of business and social networks: evidence from France. Journal of International Economics, 66, pp. 1-29 Girma, S. and Z. Yu (2002). The link between immigration and trade: evidence from the United Kingdom, Weltwirtschaftliches Archiv, vol. 138, pp. 115-130. Gould, D.M. (1994). Immigrant links to the home country: empirical implications for U.S. bilateral trade flows. The Review of Economics and Statistics, vol. 76; pp. 302-316.

Griliches, Z. (1980). R&D and the productivity slowdown. American Economic Review, 70. Papers and Proceedings, 343-48. Groote, P., R. Albers, and H. de Jong (1996). A Standardised Time Series of the Stock of Fixed Capital in the Netherlands, 1900-1995. Research Memorandum, University of Groningen. Head, K. and J. Ries (1998): Immigration and trade creation: Econometric evidence from Canada, Canadian Journal of Economics, vol. 31, pp.47-62. Rauch, J. (1999). Networks versus markets in international trade, Journal of International Economics, vol. 48, pp. 7-35. Wagner, D., K. Head and J. Ries (2002). Immigration and the trade of provinces, Scottish Journal of Political Economics, 49(5), pp. 507-525