Bilateral Migration and Multinationals - On the Welfare Effects of Firm and Labor Mobility

Bilateral Migration and Multinationals: On the Welfare Effects of Firm and Labor Mobility Abstract. This paper starts by observing two novel facts. First, bilateral migration flows are pervasive across OECD countries, both for high-skilled and low-skilled workers. This fact goes against the common belief that migration merely reallocates cheap labor from poor to rich countries. Second, multinational corporations tend to hire a large number of migrant workers. In this paper I develop a general equilibrium model that is able to reproduce these facts. In the model, migration is bilateral because of imperfect substitutability between native and foreign workers, and the operations of multinational corporations. I calibrate the model to match aggregate data on multinational production and migration stocks between the United States and Canada in 2000. The calibrated economy is a laboratory to run counterfactual experiments on the joint effects of economic policies on welfare. Opening to migration alone does not necessarily benefit native workers, especially the low-skilled ones, while the interaction between migration and multinational corporations results in net positive effect on welfare. Migration quotas, if they are reciprocal, have negative effects on native workers' welfare. The experiment results lend supports to the view that greater openness to migration can bring mutual welfare gains. 1 1. Introduction International labor migration manifests many different patterns. Among some countries, workers can mainly move in one direction, such as between Mexico and the United States. There were 46.5 million Mexican workers in the United States in 2000, while there were only 350 thousand US workers in Mexico. International movement of labor can also be more bilateral, such as between Canada and the United States, where there were 312 thousand Canadian workers in the United States and comparably 332 thousand US workers in Canada. Generally speaking, we see a bilateral pattern much more frequently for migrations among OECD countries than for migrations among the entire world. In this paper, I propose a general equilibrium model to explain high level of bilateral migration among OECD countries. This paper starts by observing two novel facts. First, bilateral migration flows are pervasive across OECD countries, both for high-skilled and low-skilled workers. Second, multinational corporations tend to hire a larger number of migrant workers than domestic firms. These facts challenge traditional perspectives of international migration. On the first point, the assumption that workers (native or migrant) are homogeneous within a skill group may not be suitable for modeling international migration because it does not explain bilaterality. On the second point, immigration can affect the local labor market not only from the supply side, but also from the demand side due to the operations of multinational corporations. Here, I formalize these two concepts in a general equilibrium framework and discuss its theoretical and quantitative implications. The model contains two key components – labor migration and multinational corporations (MNCs). Workers migrate to maximize their personal incomes given their skill levels. Workers are hired by local firms or foreign MNCs’ affiliates that operate in the 2 host country. If we assume that workers have higher marginal productivity when matching with firms that come from their origin country, then there are two consequences. First, foreign firms tend to hire a larger proportion of migrant workers. Second, an increase in the number of immigrants expands the market share of foreign companies because they are gaining competitive advantage from a higher ratio of migrant workers to native workers. As shown in Helpman et. al. (2004), average industry productivity is affected by the extent of multinational firms’ operations. Migration enhances the competitive advantage of foreign firms, thus indirectly affecting overall productivity. There are two forces delivering bilateral migration within skill groups. First, the production technology is described by a nested-CES function that allows for the possibility that native and migrant workers are not perfectly substitutable 2 . The heterogeneity between native and migrant workers motivates firms to diversify their workforce to reduce average production costs. The need for workforce diversification drives bilateral migration between two countries. However, I show this channel alone tends to generate extreme patterns (either extremely high or extremely low bilaterality). The extreme patterns are not consistent with the data, which is why the model features a second channel. The second channel stems from the assumption that the marginal productivity of workers varies according to the firms’ country of origin.3 This feature together with the presence of multinational corporations creates additional demand for migrants and makes the model flexible enough to generate the wide range of migration patterns observed in the data. Using the model, I consider several counterfactual policy experiments to understand the welfare implications of multinational firm and labor mobility. I calibrate the 2 This specification is gaining popularity in the labor literature that analyzes the effects of immigration. Examples include Borjas (2003), Ottaviano and Peri (2012), and Docquier et. al. (2012). 3 Many studies show that human capital is location-specific. Examples include Friedberg (2000), Krupka (2009), and Young (2013). 3 model to US and Canadian data4 to evaluate the impact of policy changes. First, if we shut down MNC, then bilateral migration flows are dominated by migrants from the U.S. to Canada due to country-size effect, where the smaller country has a higher marginal return on labor force growth. In this case, Canadian workers gain but US workers lose from the openness. On the other hand, with MNCs, the U.S. becomes the net receiving country of both high-skilled and low-skilled migrants and native workers in all countries benefiting from the openness to migration. Welfare improvements though are largely due to openness to MNCs, and migration enhances these gains. 5 This illustrates the important role of MNC’s on the welfare implication of migration. Theoretically, when there was no MNC, migration could be thought as merely the relocation of production factors without an external effect on overall productivity. In this case, only the countries receiving net human capital benefit from the relocation. With MNCs, migration is not only a process of relocating production factors across countries, but it also affects the aggregate productivity through the interaction with MNCs that causes the intra-industry reallocation of market shares to more productive firms. Furthermore, migration reduces variable costs of MNCs’ offshore establishments and results in more available product varieties to consumers. Both channels contribute to an increase in global production efficiency (measured by real GDP per capita). Therefore, in addition to the welfare gains due to the openness to MNCs, migration stimulates even further gains. 6 4 The United States and Canada are used here to demonstrate how migration interacts with MNCs to generate the observed pattern and the welfare implications. Similar exercise can be applied to any other pairs of OECD countries. 5 The gains in the real wages are ranging from 0.3% to 0.6% under the calibrated moving costs. Under free migration, the gains are ranging from 5% to 23%. 6 This result does not take into account for the transition path between the two equilibria, which may incur welfare costs in the short-run. 4 Second, contrary to the traditional view that migration quotas preserve the career opportunities of native workers, the results of this analysis show that reciprocal migration quotas (where both countries implement similar rules to limit migration inflow) in general have negative effects on native workers’ real wages. Instead of preserving the career opportunities of native workers, reciprocal migration quotas protect early immigrants from the competition of potential latter immigrants that are limited by the quotas. This regulation hurts native workers since the quotas prevent the economy from achieving higher production efficiency though a similar mechanism, as argued in the first point. Third, the welfare implications of reducing the moving costs are complex. In general, reducing moving costs improves the welfare of native workers. However, the welfare improvements are not necessary evenly distributed among all skill groups of workers. My results show that there are cases where the benefits accrue to one group of native workers more than another due to substitution. For example, if we reduce the moving costs for only high-skilled migrants from both countries, Canadian low-skilled workers end up being worse off even when real per capita GDP is higher because they are substituted out by US high-skilled workers. These distributional effects of change in migration policy suggest that policy makers should be cautious about the side effect of unwanted inequality when introducing new migration policies to improve the total welfare. My research contributes to a growing body of literature that analyzes the welfare effects of international migration using calibrated models. An early contribution by Hamilton and Whalley (1984) indicates that large cross-country TFP differences could be a source of substantial gains from international migration. Klein and Ventura (2007, 2009) argue that the coexistence of barriers to labor mobility and cross-country TFP differences is the result of a misallocation of the world's labor force. They develop a two-location growth model and calibrate international differences in labor quality and TFP to evaluate 5 the welfare costs of barriers to international labor mobility. Benhabib and Jovanovic (2012) investigate the optimal level of migration using a calibrated one-sector model that assumes migration is the only redistributive tool. Recently Giovanni, Levchenko, and Ortega (2013) propose a quantitative multi-sector model that includes international TFP differences, trade, remittances, and a heterogeneous workforce to explore different channels that could benefit countries sending and receiving migrants. These studies focus on migration flows that are mainly driven by international TFP differences. My research, on the other hand, provides insights on migration flows and their welfare implications among comparably developed countries. In my model, migration is mainly driven by international workforce heterogeneity and the operations of multinational corporations. The quantitative analysis of this paper is closely related to Docquier, Ozden, and Peri (2012), which simulated the labor market effects of net immigration and emigration in OECD countries using an aggregate model, and featured nested-CES production function as in Ottaviano and Peri (2012). In their study, migration alters the industry-wise average productivity through schooling externality (as in Acemoglu and Angrist (2001)) and capital accumulation corresponding to different skill-compositions of the labor force. They find that emigration of high-skilled workers has a negative effect on less educated native workers, thus increasing inequality. My research is different from Docquier et. al. (2012) in two major ways. First, their paper separately discusses the effects on welfare of immigration and emigration to/from a country, while my research focuses on the general equilibrium resulting in the migration between two countries. Second, in my research, migration affects the average productivity of an economy through the channel of intraindustry reallocation as recognized in Melitz (2003) and Helpman et. al. (2004). Unlike the schooling externality in Decquier et. al. (2012), which always results in a decrease in the 6 average productivity due to emigration of high-skilled workers, my model can generate mutual productivity gains due to migration. From a broader perspective, my research complements a large body of literature that estimates gravity models of two-way migration. Mayda (2010) investigates the determinants of migration inflows into 14 OECD countries between 1980 and 1995 and analyzes the effect of migration on average income and income dispersion in destination and origin countries. Ortega and Peri (2011) jointly estimate the effects of trade and immigration on income with a gravity-based approach, as in Frankel and Romer (1999). On the selection and sorting issue, Grogger and Hanson (2012) argue that a simple model of income maximization can explain positive selection and sorting of immigrants to OECD countries. Beine et. al. (2012) discuss the effect of diaspora network on the selection of migrants. My analysis shares with these papers the emphasis on the underlying mechanisms of bilateral migration flows, but focuses on the general equilibrium perspective of the interaction between migration and multinational corporations. There is a small but growing empirical literature looking at the impact of migrants on FDI in their origin countries. Examples include Kugler and Rapoport (2007) and Javorcik et. al. (2011). My analysis provides an alternative view that migrants can enhance the competitive advantage of firms from their country of origin, and thus bring more foreign business activities to the destination country. This view is gaining support from empirical studies such as Buch et. al. (2006) and field studies such as Harzing (2001) and Barry (2004). Finally, I only model horizontal FDI (the form of FDI that aims to make sale in the host country). Brainard (1997) reports that more than 80 percent of US multinationals’ overseas production is used to serve foreign markets, horizontal FDI seems 7 to be the dominating form of multinationals’ operations. However, it is inarguable that searching for cheaper labor substitutes is also an important driving force for firms to establish offshore affiliates. The trade-off between offshoring and migration is not covered by the model. The rest of the paper is organized as follows. Section 2 describes the observation of bilateral migration and migrant-firm matching patterns. Section 3 presents the model and Section 4 discusses the equilibrium in the symmetric and asymmetric cases. Section 5 presents the calibration and counterfactual analysis, and section 6 concludes. 2. Stylized Facts I examine the bilateral labor migration data of 30 OECD countries7 in 2000 and finds that the migration of high-skilled labor8 tends to be more bilateral9 than that of lowskilled workers. Further, I examine how the interaction between income maximizing migrants and MNCs can forge the patterns observed. Connections between MNCs and immigrant workers from the same country of origin are not unexpected. For example, modern management practices usually require intensive team cooperation. People with the same cultural and language background can understand each other more easily, leading to more effective collaboration. Enterprises may also have their own proprietary production technology, which means it is generally more cost-efficient to hire expatriates for their foreign operations rather than train new employees abroad. I argue in this paper that the 7 OECD Stat DIOC database. Countries include Australia, Austria, Belgium, Canada, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Japan, Luxembourg, Mexico, Netherlands, New Zealand, Norway, Poland, Portugal, Slovak Republic, Spain, Sweden, Switzerland, Turkey, United Kingdom, and United States. 8 Defined as people who are of working age and hold at least two-year college degrees 9 Measured by how two-way migration flows are similar in sizes. 8 interaction does exist and it is theoretically important to consider both MNCs and migration together in evaluating immigration related policies. The stylized facts regarding migration patterns among OECD countries and the connection between multinational corporations and migration are presented herein. 2.1 Migration Pattern among OECD Countries OECD Statistics has collected rich datasets of bilateral migration stocks among OECD countries in 2000. The data sorts immigrants according to their duration of stay, country of origin, age, labor status, education attainment, and field of study. From this detailed data, we can examine whether people from different skill groups exhibit different migration patterns. Here I focus on two subgroups of the immigrants – high-skilled labor and low-skilled labor. I define high-skilled labor as people who are currently in the labor force and obtained at least two-year college degrees. Low-skilled labor are people who are in the labor force but do not have a college degree. Log-scaled scatter plots of the bilateral migrant stocks for high-skilled and lowskilled workers are shown in Figure 1 and Figure 2 respectively. There is a large proportion of data points closely align on the 45-degree line for both high-skilled and lowskilled workers. This implies that two-way migrations between these country pairs are similar in sizes. This observation is not consistent with the common thought of one-way migration from poor countries to rich countries. As we can see from the figures, distance (shown by the diameter of a dot) between two countries does not have strong relationship with migration bilaterality. Many country pairs that are relatively far away from each other are still demonstrating strong migration bilaterality (located closely to the 45-degree 9 line). The bilateral patterns shown in Figure 1 and Figure 2 actually are similar to the bilateral trade flows among OECD countries (Figure 3). The close analogy between trade and migration suggests us to think about the possibility that labor is not just simply a homogeneous factor of production, and there may be heterogeneities among workers from different countries that result in some international “trade” of talents. High-skilled Migrant Stocks 1000000 100000 10000 1000 100 10 1 1 10 100 1000 10000 100000 1000000 Figure 1: Log-scaled bilateral high-skilled labor migrant stocks among OECD countries (Each dot represents a country pair
, . The diameter of a dot is proportional to the distance between the pair of countries) 10 Low-skilled Migrant Stocks 1000000 100000 10000 1000 100 10 1 1 10 100 1000 10000 100000 1000000 Figure 2: Log-scaled bilateral low-skilled labor migrant stocks among OECD countries (Each dot represents a country pair
, . The diameter of a dot is proportional to the distance between the pair of countries) 11 Bilateral Trade Flows Millions 10000 1000 100 10 1 0.1 0.01 0.01 0.1 1 10 100 1000 10000 Millions Figure 3: Log-scaled bilateral trades of commodities10 in dollars between OECD countries in 2000 (Each dot represents a country pair
, . The diameter of a dot is proportional to the distance between the pair of countries) I construct an index to measure the bilaterality of different migration groups. Suppose we denote migration stocks between two countries as , ,  , , where , is the stock of migrants from country country in country and  , is the stock of migrants from in country . The bilaterality index is:   _, ,  , = 1 − 10 , −  ,  . , +  , (1) NBER-United Nations Trade Data. 12 Here we take the difference of two migration stocks divided by the sum. The index is 1 if , =  , . To the other end, if , is very different from  , then the index approaches to zero. This numerical measure allows us to summarize the average trends of the migration patterns of different groups of workers with a single number. I calculate index values for different migration groups, illustrated in Table 1. Table 1: Migration Bilaterality Indices for OECD Countries All OECD European GDP per capita > Union (EU) $20,000 High-skilled Simple Average 0.4214 (0.0173)** 0.4285 (0.0260) 0.4674 (0.0239) Weighted Average* 0.3947 0.5892 0.4451 Simple Average 0.3839 (0.0166) 0.4130 (0.0262) 0.4618 (0.0241) Weighted Average* 0.2104 0.4357 0.3986 Low-skilled * Weight by the total migrant stocks among the country pair ** Numbers in parentheses are standard errors Table 1 shows that both high-skilled and low-skilled migration stocks exhibit a certain degree of bilaterality. According to the simple averages for all OECD countries, the numbers are around 40%. This means that if we normalize total migrants between a pair of countries to 100, then the average migration stock in one country is 80 and in another country is 20. Although people have the tendency to move to one country rather than 13 another, the bilaterality is still significant and suggests that push/pull factors for international migration may not have the same effects on workers from different countries. The average values of bilaterality index for high-skilled migration are consistently higher than the average values of index for low-skilled migration, especially if we consider the weighted averages. This suggests that workers with different skill levels face different push/pull factors for migration. Finally, the average index values for the subset of countries in European Union (EU) are higher that the index values for all OECD countries. If we consider the weighted averages only, we can see that the index values for EU are much higher than the others. We know that EU has a highly integrated labor market. This would indicate that migration bilaterality is positively associated with economic integration. 2.2 Connection between Migration and MNCs Multinational corporations play an important role in globalization. Anecdotal evidence suggests the hypothesis that multinational corporations are active in creating migration opportunities. For example, the Boston Consulting Group (BCG), a multinational consulting company that manages 6,200 consultants in 43 countries 11 , reports that they constantly deploy about 20% of their employees as expatriates to support foreign offices. We can also detect this connection between multinationals and migration through the matching of migrants and MNCs. According to the Survey on Americans Overseas (Koppenfels (2012)), about half of American workers in for-profit private sectors abroad are working for international companies. Another survey by Taiwanese human resource 11 Data from BCG.com. Retrieved 2013-03-06. 14 agencies12 shows that in 2011 there are about 75% of Taiwanese workers who work in the mainland China were working in international firms. These matching patterns demonstrate the existence of special connections between multinational corporations and migration. Harzing (2001) conducted interviews with MNCs and found that they often employ expatriates to transfer management activities to foreign affiliates. Barry (2004) reports that Intel’s decision to invest in Ireland is promoted by the ability to hire engineers from the U.S. Buch et al. (2006) finds that FDIs and labor migration from the same country of origin are positively correlated in Germany’s states. Finally, I show in Table 2 the matching of immigrant workers and foreign firms compared with local firms in Brazil. The data includes all exporters in the linked employer-employee data for Brazil13 during the period 1995-2001 as described in Muendler and Rauch (2012). The definition of foreign firms is that they are FDI affiliates. 14 Table 2: Proportions of Immigrants Hired by Different Types of Firms Type of Firms Domestic Foreign Low-Skilled High-Skilled All Immigrants Immigrants Immigrants 0.0205 (0.0003)* 0.1964 (0.001) 0.0397 (0.0033) 0.0113 (0.001) 0.2973 (0.0059) 0.0557 (0.0019) * Numbers in parenthesis are standard errors 12 104 HR Agency, (2011), "Taiwan brain drain crisis survey" and 1111 HR Agency, (2011), "Taiwanese work abroad survey." 13 Code courtesy of S. Bazzi (Boston Univeristy) 14 FDI indicator by J. Poole (UC – Santa Cruz) 15 Table 2 shows that foreign firms hire a higher proportion of immigrant workers on average than domestic firms. This is mainly due to the fact that foreign firms hire a much higher proportion of high-skilled immigrants than local firms. Domestic firms on average hire a higher proportion of low-skilled immigrants than foreign firms, but the difference is small. This lends support to the argument that there is a close connection between migration and multinational firms. 2.3 Summary of Empirics This section illustrates two important observations. First, migrations between OECD countries in general exhibit significant bilaterality, and high-skilled migrations tend to be more bilateral then low-skilled migrations. Second, international firms tend to create migration opportunities and hire more migrants than local firms. This finding is supported by field studies and matching patterns between different types of workers and firms. The model I propose is aiming for reproducing these two key observations. 3. Model The model is a two-country general equilibrium model of FDI and labor migration. The model is based on Helpman, Melitz, and Yeaple (2004), which extends the Melitz’s trade model to incorporate horizontal FDI. It provides an important insight that firm heterogeneity plays a significant role on the determination of FDI flows. In this paper, I will show that firm heterogeneity is also a major component that determines bilateral migration patterns. According to the data, migration patterns are different for low-skilled and high-skilled workers. Thus, I specify two types of workers in each country by their 16 skill levels. Moreover, workers with the same skill level but from different countries are classified as different types of workers. 3.1 Consumer’s Preference and Demand A representative consumer’s preferences are given by a CES utility function over a continuum goods index by : $ "Max *+ (
) )∈1 ,
-./ -./  - 2 (2) # "3. . + 4
,
 = 5 , ! )∈1 where Ω is the set of available products, ,
 is the quantity of the product consumed by the representative consumer, 4
 is the price of the product that is . Further, goods are imperfect substitutable, the elasticity of substitution 7 is larger than one. Finally, 5 is the aggregate expenditure. As in Melitz (2003), the optimal consumption for a product variety is: where : is the price index and ,∗
 = 9 : = 9+ 4
 ; : .- ∙ 5 , : 4
/.-  ; )∈1 / /.- (3) . (4) 3.2 Firms Every firm pays an entry cost = > at the time of entering the market. This entry cost includes all outlays for establishing a new firm such as production development and brand advertising. This paper does not explicitly discuss the structure of the entry cost. 17 Note that when the number of new entries is not constrained, the ex-ante expected profit for firms would be offset by this cost in equilibrium. Firms are characterized by productivity parameter ?. Firms draw their productivity levels from a distribution with the CDF @
? while entry. After entry, firms decide if they want to stay in market given their own productivity levels. Less competitive firms that cannot make profits exit market. In addition, I assume that there is a proportion A of firms that exit exogenously in each period. If a firm decides to stay operational, the next decision is that if it wants to sell in the domestic market only or enter the foreign market and become a multinational firm. In all cases, every firm pays a fixed cost = for its domestic production. Multinational firms pay fixed cost = B for their offshore operations. If a firm decides to become a multinational, then it would have two establishments. One produces domestically and serves only the home market. Another one produces and serves the foreign market. As in Helpman et. al. (2004), firms in equilibrium would not serve the foreign market without serving the home market. Furthermore, only highly competitive firms (with high enough productivity levels) become multinationals. This paper does not refer to trade because it would not alter the main implications of the model. The investigation remains focused on the relationship between migration and multinational firms. Throughout the following paper, the term “firm” is used to denote the entire company (including its home headquarter and foreign affiliate), and “establishment” is used to denote a production unit. An establishment could be a firm’s headquarter in the home country or a firm’s foreign affiliate. All establishments owned by the same firm have 18 the same productivity level and skill compatibility with different type of workers, but establishments make their own production decisions within their own markets. The marginal cost for an establishment of a firm with productivity level ? is: C 
? = 1 ∙D E , ? (5) where index denotes the country where the establishment is located and index denotes the country of the firm’s origin. Labor is the only input for production. D E  denotes the average wage rate for workers who work in the establishment. 3.3 Labor Endowments, Moving Costs, and Wage Rates The model assumes there are two countries (denoted by country 1 and country 2) and four types of labor, respectively high-skilled workers from country 1, high-skilled workers from country 2, low-skilled workers from country 1, and low-skilled workers from country 2. Each country has endowments for high-skilled and low-skilled workers, which E  and GH (for = I1,2K). are denoted by F Moving costs are separately specified according worker’s skill level, source country, and destination country15. The moving costs for high-skilled migrants from country to denoted by L , and for low-skilled migrants from country to is is denoted by  , . Here I assume that L , ,  , = 1 if = (there is no migration costs for native workers staying in their home country), and they are larger than one if ≠ N. The migration costs indicate that there are usually some extra outlays for employers to hire international workers. For example, in the U.S., employers need to pay H1-visa fees for international workers they 15 The migration costs need not to be symmetric between two countries. Although transportation costs might be similar for moving back and forth, each country may have its unique regulations on immigration and thus impose different costs. 19 hire. The research follows in Bojas’ specification (1987) that the moving costs are proportional to worker’s income. I use nested-CES production function to aggregate different types of workers. The first layer provides a CES aggregator for high-skilled and low-skilled workers. The second layer provides a CES aggregator for immigrant and native workers within each skill level. This specification is similar to Ottaviano and Peri (2012). The average wage rate for labor is defined as: D E = OP ∙ 3̅  The index Q  /.Q +
1 − P ∙ SH Q /  /.Q /.Q T . again denotes where the establishment is located and index (6) denotes the firm’s country of origin. 3̅  and SH are respectively the average wage rates for high-skilled workers and low-skilled workers. U is the elasticity of substitution between low-skilled labor and high-skilled labor. P ∈
0,1 is the output share of high-skilled workers. The next layer is aggregation of immigrant and native workers: 3̅  = 9W XY Y ∑[ X SH = 9W  ]Y Y ∑[ ] ,Y ∙3 [ ∙ L [ ∙  ,Y ,Y /.\ ∙S ; ,Y /.^ / /.\ (7) / /.^ (8) ; 3 ,Y is the wage rate for high-skilled workers who are from country N and work in country . Similarly, S,Y is the wage rate for low-skilled workers who are from country N and work in country . The parameters _ and Γ are elasticity of substitution between immigrant and native workers. The parameters X Y and ] Y represent the compatibility (so called the skill compatibility parameters) between firms and workers that may come from different 20 origins. I normalized these parameters to 1 if = N, and they are assumed to be less than one if otherwise. Workers from different countries often have different cultural backgrounds, languages, and educational trainings so they tend to have less compatibility with firms from other countries. 3.4 Demands for Migrant Workers By Shephard’s lemma, firms’ marginal demands for different types of workers are: ℎ,Y
? =  ,Y
? = XY 1 Q 1 \ ∙P ∙ ∙ E b c ∙ D ∑[ X [ L ,Y ? Q ]Y 1 1 ^ ∙
1 − PQ ∙ ∙ E b c ∙ D ∑[ ] [  ,Y ? ∙ 3̅  Q \.Q ∙ SH ∙ 3 ,Y ^.Q .\ ∙ S,Y (9) .\ . (10) We can see that marginal labor demands are decreasing functions in firm’s productivity level and relative cost to other production factors. Next, to find the aggregate labor demand, we first define the total demand of the representative establishment (with the average productivity of its kind). The total demand of the representative establishment that comes from country , 
?e ∙ ℎ,Y
?e, F ,Y = d  g , ?e ∙ ℎ,Y ?e g , G ,Y , 
?e ∙  ,Y
?e , =d  g , ?e ∙  ,Y ?e g , and operates in country is: =f = N =f ≠ N (11) =f = N =f ≠ N (12) Here F ,Y is the demand for high-skilled workers of the representative establishment and G,Y is the demand for low-skilled workers. ?e is the average productivity of local establishments and ?e g is the average productivity of foreign establishments. 21 Finally, the total demand for migrants from country N to is: ℎ hℎ_3N _ h3 ,Y = W i ∙ j  ∙ F ,Y fD_3N _ h3 ,Y = W i ∙ j  ∙ G,Y , where j  is the proportion of firms from country (13) (14) that have an establishment in country . 3.5 Migration Quota Migration quotas are also usually implemented separately for workers with different skill levels. For example, in the U.S. there are different work visas (e.g., H-1B and H-2B) for different type of workers and each type of work visa has its own limit cap. Therefore, we can model the quotas independently for each type of workers. I discuss here only the case in which migration quotas are effective (i.e., where the constraint is binding). In considering country implementing an effective migration quota to workers of type  from country , since the quota is effective, we can denote it as a percentage16 of the total number of immigrants when there was no constraint. I denote the percentage by kl ∈ m0,1n, where  = Iℎ, K denotes the skill level of workers, , implements the migration quota, and denotes the country that denotes the country of origin of the workers. kl = 1 if immigration is unconstrained, kl = 0 if legal immigration is totally banned, and , , kl ∈
0,1 when immigration is allowed and an effective migration quota is implemented. , 16 In reality, quotas are usually implemented as absolute numerical caps. However, if quotas are binding, then we can always find a one-to-one mapping between the percentage value and the absolute cap. These two denotations are equivalent in the context of the model. 22 4. General Equilibrium The equilibrium is defined by the set of variables oI?H K, p?H g, q, Ij K, Ii K, I3 , K, IS , Kr, where = I1,2K denotes the country of origin = I1,2K denotes the destination country. ?H is cutoff productivity for local firms, which is the lowest productivity level that firms with productivity levels lower than this threshold would exit the market. ?H, is the cutoff g productivity for foreign firms form country and operating in country proportion of firms from country that are multinational. 3 , skilled workers from country and working in country , and S . j is the is the wage rate of high, is the wage rate of low- skilled workers. s u In addition, I assume that the productivity distribution is Pareto and @
? = 1 − OtT =f ? ≥ , where  is the scale parameter, and w is the shape parameter. 4.1 Equilibrium Conditions The following conditions determine the equilibrium. The derivation of the equilibrium is in Appendix A. (1) Zero-cutoff Profit Condition Since the production function is assumed to be increasing return to scale, the profit of a firm is an increasing function in productivity. There exist productivity levels ?H and ?H, such that firms with productivity levels less than ?H close down and only firms with g productivity level higher than ?H choose to become multinational firms. The zero cutoff g profits condition can be expressed as a set of equations: x
?H  = 0 (15) 23 x O?H, T = 0 g (2) (16) Free Entry Condition The ex-ante expected profit for a new entrant firm is: y> } = z {W
1 − A| ∙ x − = > € = |~ 1 − @
?H  ∙ xH − = > . A (17) Free entry of new firms drives the ex-ante expected profits to zero. Therefore, we have y> = 0 =f ∀ . As in Melitz (2003) and Helpman et. al. (2004), we can use the condition g (1) and (2) to solve I?H K, o?H, r, and Ij K. (3) Labor Market Clearing and Migration Incentive Compatibility If there is no migration quota workers are assumed to be able to move across countries by paying the moving costs. Therefore, in equilibrium, we must have the real incomes (adjusted for the moving costs) for workers who migrate to foreign countries equal to the real incomes for the same type of workers who stay in their home countries. Otherwise, workers would keep moving to the country where they can earn higher real wages. Further, the total labor demand should equal to total labor supply in all countries. We can use this condition to solve Ii K, p3 (4) ,Y q, and pS ,Y q. Migration Quota Notice that the equilibrium condition (3) holds true only if we do not have an effective migration quota in existence. If there are effective migration quotas, the countries that implement them would have excess demands for foreign workers. In this case, we have: 24 3 , 3 , > : : = kƒ < 1 , S, S, > : : For workers who are form country =k , (18) <1 (19) that are effectively regulated by migration quota in country . By our notation of the migration quota, the new supply functions of foreign workers are: F †,, = kƒ ∙ F >,, G†,, = k , , (20) ∙ G>,, , where F>,, and G>,, are amounts of immigration from country (21) to in the unconstraint equilibrium solved by using equilibrium condition (3). We can use these new supply functions to look for the equilibrium with binding migration quotas. 4.2 Analysis of the Equilibrium 4.2.1 International Labor Heterogeneity and Migration Pattern An important channel in the model that generates bilateral migration flow is heterogeneity in workers from different countries. Since workers from different countries are not perfectly substitutable, firms are motivated to hire foreign workers to reduce the average production costs. However, I show here that this channel alone tend to generate extreme migration patterns. 25 We start by considering only migration of high-skilled workers, and a similar analysis can be applied to migration of low-skilled workers. If we shut down MNC (i.e., j = 0 ∀ ), then we can write the number of migrants from country  h3 , = {* ‡ , = P Q ∙ O 3  L , ∙ 3 /.\ 2 + X € \ /.\ 7 .5 T ∙ i ∙ /.- ∙ 7−1 : to : ∙ ‡ , Dℎ , X 1 + X / ./ /.\ (22) ∙ ?e-./ ∙ D E Q.- ∙ 3̅ .Q Here 3  denotes the real wage of high-skilled workers from country . We can see that the number of high-skilled migrants from country to is decreasing in the wage (adjusted for the moving costs) of migrant workers, relative to the average cost for hiring high-skilled workers of firms from country . The sensitivity of the number of migrants to the relative difference in migrants’ wage, to the average wage, largely depends on the elasticity of substitution between workers from country and country (i.e., _). Now we consider the ratio of the number of high-skilled migrants from country to to the number of high-skilled migrants from country to , which is: [ˆ‰Šl|‹ [ˆ‰Šl|‹ Œ, ,Œ =Ž “ ‘’“ Œ  ‘’“ b Œ,  c ”•Œ  ∙ ‘’“  b ,Œ Œ c ”•Œ  ∙ – — Œ, ∙ ,Œ — (23) . First, if two countries are symmetric, then the ratio is always 1. This means that between two symmetric countries, the migration flow is perfectly bilateral. Second, suppose we have two asymmetric countries but they are similar enough so that — Œ, — ,Œ equilibrium we have 3 > 3  , then the number of migrants from country to than the number of migrants from country to (i.e., [ˆ‰Šl|‹ [ˆ‰Šl|‹ Œ, ,Œ ≈ 1 and in is smaller < 1) and the magnitude 26 of the difference depends on relative wages and the elasticity of substitution between workers from country and country . The migration bilaterality is getting lower while the elasticity of substitution is getting larger. Consider the fact that foreign and native high-skilled workers are usually seen as close substitutes (Ottaviano and Peri (2012) estimate this elasticity in the U.S. is about 33). A small difference in wage rates could be augmented to a large difference in the amount of migration stocks in two countries. 4.2.2 Migration and Multinational Corporations In 4.2.1, I show that international labor heterogeneity is an important driving factor of bilateral migration, but that alone tends to generate extreme migration patterns. Here I discuss how the operations of multinational corporations can balance this out, allowing the model to be flexible enough to generate a wide range of migration patterns. Again, the discussion focuses on the migration of high-skilled workers and a similar analysis can be applied to migration of low-skilled workers. Consider the case with MNCs. The number of high-skilled migrant from country to country is:  h3 , = d{* ‡ , = P Q ∙ O z , 3 2 , L ∙3 /.\ \ + X € ∙ ‡, ™ + d{X  ⋅ * 7 .5 T ∙ i ∙ /.- ∙ 7−1 : X / /.\ ./ 1 + X .7 5 1  = PQ ∙ O T ∙ i ∙ j ∙ /.- ∙ 7−1 : 1 + X 3 2 , L ∙3 /.\ ∙ ?e-./ ∙ D E / /.\./ ∙ ?e \ + 1€ ∙ z , ™ , Dℎ Q.- g -./ ∙ 3̅ ∙ D E (24) .Q Q.- ∙ 3̅  .Q Compared to (22), we have an additional component (in the second large bracket), which is due to demand of foreign establishments in country . Notice that the positions of the 27 compatibility parameter ( ξ ) are different in the components. This is because international firms have higher compatibility with workers from their origin country than with native workers in the destination country. We compare the bilateral migration flow between the two countries that is due to operations of multinational corporations with a ratio:  f = Ž Œ •Œ ⋅b Œ,  c  ∙   •Œ ⋅b ,Œ Œ c  ∙ ‘’“ ‘’“ ”/ ”/ – . The ratio compares the second components of  h3 (24). z , and z , \ , (25) and  h3 , as defined in are dropped since they are dominated when _ is large. The ratio gives us a sense of the relative contribution of multinational corporations to migration flows in different countries. Assuming that two countries are asymmetric and in equilibrium we have 3 > 3  as in 4.2.1. Further, for simplicity, we assume that the moving costs and the skill compatibility are symmetric between two countries (i.e., L , = L , and X  = X ). With these assumptions, the ratio in (57) is larger than 1 and increasing in _. The result indicates that multinational corporations contribute much more to migration flow from country to than the flow from country to . Notice that this trend is the opposite of the one mentioned in 4.2.1. In 4.2.1 we see that with international wage difference, workers tend to move from low-income country to high-income country and cause low migration bilaterality. However, multinational corporations provide another channel to balance this trend by providing an extra demand for migrant workers from high-income country to low-income country. 28 4.2.3 Migration and Productivity The cutoff productivity for domestic and foreign establishments according equilibrium condition (1) and (2) are: / = + j, ∙ = B 7−1 u ?H = 9 ∙ ; ∙ A ∙ => 1+w−7 ?H, g (26) / D E :  5 = /.= ?H ∙  ∙ ∙ *  ∙ B 2 , 5 = D E : (27) where ?H is the cutoff productivity for domestic establishments in country , ?H, is the g from country , and j, is the cutoff productivity for foreign establishments in country proportion of firms from country that have foreign establishments in country . This proportion is: j, D E / .u /.- : 5 = = Ž  ∙ ∙ *  ∙ B2 : 5 = D E  – . (28) In this paper, migration affects overall productivity through its indirect effect on intra-industry reallocation of market shares among firms with different productivity. To illustrate this point, this paper posits a special case where only migration from country to is allowed. In comparing this case to the case of disallowing migration, we can derive that the ratio  EŒ œ E ŒŒ œ is higher when migration is allowed due to international labor heterogeneity and higher skill compatibility between firms and workers from the same 29 origin. We can see from (57) that a higher ratio  EŒ œ E ŒŒ œ leads to a higher proportion of multinational firms, which in (26) increases the cutoff productivity17. Migration from country to foreign establishment from country country also tends to increase the cutoff productivity for to ( ?H, ) because foreign establishments from g are gaining competitive advantage (with a larger  EŒ œ E ŒŒ œ ) due to migration. The tougher competition brought by highly productive foreign firms causes intra-industry reallocation of market shares to more productive firms and increases aggregate productivity in country . In general, the model shows that the movement of both immigrants and emigrants leads to productivity gains. Immigrants bring more foreign business activities, which increase local competition and uplift overall productivity. On the other hand, emigrants enhance offshore business opportunities that attract more new entrants in the origin country. 5. Quantitative Analysis In this section I consider counterfactual experiments in evaluating quantitatively the impact of policy changes in regard to international migration. Here the quantitative analysis targeting migration and multinational corporations between the United States and Canada. Similar analysis can be extended to any pair of countries when addressing the impact of immigration policy changes between two countries. 17 As well as aggregate productivity since it is proportional to the cutoff productivity as recognized in Melitz (2003). 30 5.1 Calibration and Sensitivity Analysis 5.1.1 Calibration I sort the parameters into two broad categories. The first category contains parameters that pertain to industry characteristics, which includes the elasticity of substitution over products 7 , the entry cost f > , the fixed production cost for local establishments f, the fixed production cost for foreign establishments f B , and the shape parameter κ and the scale parameter B of the Pareto productivity distribution. The second category contains parameters that pertain to the labor market, including labor E  and GH , migration costs L , and  , , skill compatibilities X  and ]  , the endowments F share parameter of high-skilled labor α, and elasticities of substitution among different types of labor U, γ, and Γ. The elasticity of substitution over products and the shape parameter of the productivity distribution are calibrated according to Luttmer (2007) and Broda and Weinstein (2004). Broda and Weinstein report that the median elasticity of their estimation for sectors at the 5-digit SITC level in the U.S. is 2.7. I use this number to calibrate the parameter σ. Luttmer reports that to match the tail shape of the firm size distribution in the U.S., the ratio £ -./ should equal to 1.06, which implies that κ = 1.8. According to the Business Dynamic Statistics by the U.S. Census Bureau, the exit rate in 2000 of firms that are older than five years is about 9% and the exit rate of firms in their first year is 23.9%. I use 9% as the exogenous exit rate of firms (i.e., δ = 0.09) and 23.9% as the endogenous survival rate of the newly established firms. The endogenous s u t survival rate implied by the model is ℙ
? ≥ ?H = O E T = ¨ g”©Œ ∙gª -./ ∙ ­ «∙g¬ /”u.- ./ . By 31 normalizing f and B both to 1, we can use this formula to calibrate f > to match the observed survival rate of new firms. The proportion of multinational firms in the U.S. according to Bernard (2009) is about 1%. I use this to calibrate f B since the proportion of multinational firms is given by χ = {  EŒ œ E ŒŒ œ ∙ ¯Œ ¯ ∙ ‘ °  g ‘’± O °Œ ∙ gª T € .u . To calibrate labor endowments, I normalize the total population in the U.S. to 10 and adjust the population of Canada by the relative country size. The actual number of high-skilled and low-skilled workers in each country depends on the ratio of collegegraduates to non-college-graduates. According to the OECD Stat country profiles, we have E / = 3.58, H E µ = 0.4367, LH/ = 6.42, and LHµ = 0.6743. H The elasticities of substitution over different types of labor are calibrated according to the estimation by Ottaviano and Peri (2012), where we have ρ = 2, γ = 33, and Γ = 11.1. The share parameter of high-skilled labor α is calibrated to match the income distribution of the U.S. in 2000. According to the UNU-WIDER World Income Inequality Database V2.0c, incomes paid to the top 30% of wage earners is 58.24% of the total wage payments and to the top 40% is 68.36% of the total payments. Since the high-skilled endowment in the U.S. is calibrated to be 35.8% of the total labor force, I use linear interpolation to calculate the percentage of total income payments to high-skilled workers in the U.S., which is 64%. I calibrate α to so that the percentage of total income payments to high-skilled workers in the U.S. match this number. Finally, the migration costs L , and  , and the skill compatibility parameters are calibrated to best the migration pattern between the U.S. and Canada. According to the OECD Stat DIOC database, the high-skilled migrants from the U.S. to Canada is 0.24% of 32 its high-skilled labor force, high-skilled migrants from Canada to the U.S. is 5.55%, lowskilled migrants from the U.S. to Canada is 0.24% of the low-skilled labor force, and lowskilled migrants from Canada to the U.S. is 4.14%. Further, the bilaterality index for highskilled workers is 0.53 and for low-skilled workers is 0.38. These are the targeting numbers to match. The calibration result is summarized in Table 3. 33 Table 3: Calibration Result (Country 1 denotes the U.S. and country 2 denotes Canada) Parameter E/ IF Eµ IF H/ K ,G , GHµ K L/,µ L µ,/  /,µ  µ,/ Xµ/ X/µ ]µ/ ]/µ U _ Γ 7 Calibrated Value Description Source I0.4367,0.6744K Labor endowments OECD Stat Moving costs of high-skilled migrants To match migration pattern Moving Costs of low-skilled migrants To match migration pattern 0.8483 Skill compatibility parameters for high- To match migration pattern 2 skilled workers I3.58,6.42K 1.1896 1.0984 1.5585 1.5835 0.9373 0.7741 0.9027 33 11.1 2.7 A 0.07  1 w P = =B => 1.8 0.7 1 76 325.27 skilled workers Skill compatibility parameters for lowElasticity of substitution between high- To match migration pattern Ottaviano and Peri (2012) skilled and low-skilled workers Elasticity of substitution between native and foreign high-skilled workers Elasticity of substitution between native and foreign low-skilled workers Elasticity of substitution over different Broda and Weinstein (2004) products Exogenous firm exit rate Business Dynamic Statistics by the U.S. Census Bureau Parameters of productivity distribution Luttmer (2007) Normalization Output share of high-skilled workers UNU-WIDER World Income Inequality Database Fixed production costs Normalization Fixed production costs - FDI Bernard (2009) Entry cost Business Dynamic Statistics 34 Table 4: Percentage of Immigrants Hired by Different Type of Firms USA High-Skilled Low-Skilled With MNCs Local Firms 0.61 0.36 Foreign Firms 0.76 0.52 Without MNC Local Firms 0.04 0.04 With MNCs Local Firms 1.88 0.84 Foreign Firms 2.35 1.19 Without MNC Local Firms 22.40 6.44 Canada High-Skilled Low-Skilled Table 4 shows that the percentages of immigrant workers that different types of firms hire compared to their native workforces. We can see that the percentages of highskilled immigrants are higher than low-skilled immigrants for all types of firms. Foreign firms tend to hire more immigrants than local firms. The ranges of the percentage of immigrant workers hired are 0.6 to 2.35 for high-skilled workers, and 0.36 to 1.19 for lowskilled workers. These patterns are roughly consistent with the pattern we observed in the Brazil data. If we consider the case that MNC is not allowed, then the matching patterns are very different in two countries. This is due to the fact that migration pattern is nearly unilateral without MNC. 5.1.2 Sensitivity Analysis To show that how different sets of parameters affects migration pattern, I discuss in this subsection the sensitivity analysis of selected key parameters. These key parameters include elasticity of substitution between native and foreign workers, migration costs, and compatibility parameters between different types of firm and worker. 35 Figure 4 shows the changes in migration stocks and migration bilaterality corresponds to different elasticity of substitution between US and Canadian high-skilled workers. The magnitude of bilateral migration stocks are decreasing in the elasticity of substitution. This result indicates three facts: 1. The magnitude of migration is decreasing in the elasticity of substitution. 2. The relationship between the elasticity of substitution and bilaterality of migration is not linear, the bilaterality first decreasing and then increasing as the elasticity of substitution increasing. 3. Elasticity of substitution between US and Canadian high-skilled workers does not only affect the migration pattern of highskilled workers, but also affects the migration pattern of low-skilled workers. This links to the complementarity between migration workers and multinational firms from the same country. Less high-skilled migration affects the multinationals’ activities and in turn affects the migration pattern of low-skilled migration. 0.9 1200 Bilaterality, High-skilled Bilaterality, Low-skilled 0.8 1000 0.7 800 0.5 600 0.4 0.3 400 0.2 Migrants from Canada to USA Migrants from USA to Canada 0.1 0 0 10 20 30 40 Elasticity of Substitution 50 60 200 0 70 Figure 4: Sensitivity of elasticity of substitution between high-skilled workers from different countries (the red line marks the benchmark value » = ¼¼) 36 1,000 People Bilaterality Index 0.6 Next, I present the change of migration magnitude and pattern regards to the change of the moving costs. Figure 5 shows the result of high-skilled migration between the U.S. and Canada if we change the calibrated moving costs for US emigrants. It is not surprising that the magnitude of high-skilled migration from the U.S. to Canada is decreasing in the moving costs. Notice that because in the calibrated economy (and also in the data), there are more migrants from Canada to the U.S. than migrants from the U.S. to Canada. Therefore, in the range of the moving costs we are presenting in Figure 5, bilaterality of high-skilled migration is decreasing in the moving costs of high-skilled US emigration to Canada. Once again, we also see that the magnitude and pattern of lowskilled migration is affected by the change due to the interaction between migrants and multinational firms. Figure 6 depicts high-skilled migration in the case that the moving costs of Canadian emigration to the U.S. are changing. The result is mostly symmetric to the case shown in Figure 5. One notable difference between Figure 5 and Figure 6 is that the bilaterality is decreasing as the moving costs decreasing in Figure 6 rather than increasing as in Figure 5. The bilaterality keeps increasing in the moving costs until it reaches the perfect bilaterality and starts to decrease in the moving costs. This indicates that decrease in the moving costs for US high-skilled emigration to Canada or increase the moving costs for Canadian high-skilled emigration to the U.S. in a certain range can improve the imbalance between US high-skilled emigrants to Canada and Canadian high-skilled emigrants to the U.S. 37 1 100 Bilaterality, High-skilled Bilaterality, Low-skilled 0.9 0.7 60 0.6 50 0.5 40 0.4 30 0.3 20 0.2 0.1 -50 1,000 People Migrants from Canada to USA 80 Migrants from USA to Canada 70 0.8 Bilaterality Index 90 10 -40 -30 -20 -10 0 10 20 30 40 % Change in Moving Costs for Migrantion from the US 0 50 Figure 5: Sensitivity of the moving costs for high-skilled migrants from the U.S. to Canada 1 0.9 0.8 350 Bilaterality, High-skilled Bilaterality, Low-skilled 300 Migrants from Canada to USA Migrants from USA to Canada 250 0.6 200 0.5 150 1,000 People Bilaterality Index 0.7 0.4 100 0.3 50 0.2 0.1 -50 -40 -30 -20 -10 0 10 20 30 40 % Change in Moving Costs for Migrantion from the Canada 0 50 Figure 6: Sensitivity of the moving costs for high-skilled migrants from Canada to the U.S. 38 I model the matching pattern between different types of workers and firms by the compatibility parameters. Figure 6 shows that simulation result of changing in the compatibility between high-skilled Canadian workers and US firms. We see that the magnitude of high-skilled migration does not change much by changing the compatibility parameter. The proportion of immigrant workers work in the firms from their home country increases in the compatibility for the both countries. Notice that here we just increase the compatibility between high-skilled Canadian workers and US firms but not the compatibility between high-skilled US workers and Canadian firms. However, not only the proportion of Canadian high-skilled immigrants in the U.S. work in firms from their home country increases, but also the proportion of US high-skilled immigrants in the Canada work in firms from their home country increases. The increase in compatibility between Canadian high-skilled workers and US firms also leads to increasing demand of US multinational firms in Canada for Canadian workers, and bid up Canadian workers’ wage. Therefore, Canadian firms also tend to hire more US immigrant workers as substitute. 39 55 65 54 60 53 55 52 50 51 45 Migrants from Canada to USA Migrants from USA to Canada 50 40 Migrant workers in local firms, USA Migrant workers in local firms, Canada 49 35 48 30 47 25 46 20 -20 -15 -10 -5 % change in ξ 0 5 10 Proportion (%) 1,000 People 70 45 15 Canada US Figure 7: Sensitivity of skill compatibility between US high-skilled workers and Canadian firms 5.2 Counterfactual Experiments In this subsection we consider counterfactual experiments in evaluating quantitatively the impact of policy changes in regard to international migration between the U.S. and Canada. 5.2.1 Openness to Migration and MNCs Four scenarios are compared: (1) Autarky, (2) Openness to migration only, (3) Openness to MNC only, and (4) Openness to both migration and MNCs. The result is summarized in Table 5, which includes the equilibrium real wages, details of migration flow, masses of firms, productivity, and real GDP per capita for each case. 40 Table 5: Simulation Result for Static Comparison Scenarios Autarky Migration MNCs Only Only Migration & MNCs Real Wage Low-skilled, USA 1* 0.9819 1.3705 1.3760 Low-skilled, Canada 0.8676 1.2503 1.4459 1.4502 High-skilled, USA 3.1950 3.1825 4.3930 4.4111 High-skilled, Canada 2.5624 3.3731 4.2817 4.3062 Low-skilled, USA to Canada - 0.0759 - 0.0065 Low-skilled, Canada to USA - 0.0017 - 0.0278 Bilaterality Index, Low-skilled - 0.0438 - 0.379 High-skilled, USA to Canada - 0.1835 - 0.0088 High-skilled, Canada to USA - 0.0009 - 0.0242 Bilaterality Index, High-skilled - 0.0098 - 0.5333 0.5836 0.5695 0.3311 0.3328 Foreign, USA - - 0.0033 0.0034 Local, Canada 0.4194 0.7109 0.3337 0.3361 - - 0.0033 0.0034 0.8502 0.8502 1.1662 1.1666 Foreign Cutoff, USA - - 15.0300 15.0334 Local Cutoff, Canada 0.8502 0.8502 1.1639 1.1638 Foreign Cutoff, Canada - - 14.9928 14.9860 Aggregate, USA Aggregate, Canada Real GDP per Capita USA Canada 4.6550 4.6550 4.6550 4.6550 8.9042 8.8675 8.9100 8.8633 100** 85.89 98.31 123.44 137.34 143.37 138.17 144.41 Migration Mass of Firms Local, USA Foreign, Canada Productivity Local Cutoff, USA * Normalized to 1 ** Normalized to 100 41 Starting from the economy under autarky, opening to FDI increases real wages for all types of workers. In addition, opening up to migration enhances the welfare gains. However, if we only have migration alone, then US workers end up with lower wages compared to autarky. On productivity, migration alone does not change cutoff productivity compare to autarky. If we have MNCs, then migration begins to impacts on cutoff productivity by interacting with MNCs, that causing intra-industry reallocation as shown in Section 4. Migration patterns are almost unilateral when we disallow MNCs (as shown by the bilaterality indexes, which are 0.0438 for high-skilled migrants and 0.0098 for low-skilled migrants). Migration from the U.S. to Canada dominates the total bilateral migration flow due to the country-size effect that a smaller country has a higher return to population increase. When we allow for MNCs, migration bilaterality is comparatively much higher than before; the bilaterality indexes in this case are 0.5333 for high-skilled migrants and 0.379 for low-skilled migrants. This illustrates that MNCs are an important driving factor for bilateral migration. Notice that compared to the MNCs Only case, in the Migration & MNCs case, the overall productivity in Canada is actually decreasing, whereas, Canadian real per capita GDP and the real wages for Canadian workers in all skill levels are still higher. Moreover, the mass of firms in both countries is higher in the Migration & MNCs case than in the MNC Only case. This illustrates that migration increases global production efficiency (characterized by higher real per capita GDPs in both countries) through two channels productivity improvement and increase in product varieties. This is similar to the extensive margin and intensive margin of gains due to openness to trade as discussed in Melitz (2003). Here the U.S. is gaining from both channels by opening to migration. Canada is gaining from increasing product varieties and losing productivity. For both 42 countries, the net result is that they have higher real per capita GDPs and real wages for all workers. Lastly, the benefits of openness in general accrue more to high-skilled workers than low-skilled workers. This distributional effect of migration is due to substitutions, since the output share of low-skilled workers is smaller than the output share of high-skilled workers as calibrated. Low-skilled workers are more vulnerable to foreign substitutes and thus obtain less gain from openness. 5.2.2 Changes in Migration Costs Here I consider the impact of bilateral change in moving costs. The change in moving costs could come from countries adopting new regulations on immigration, such as new standards for migrant’s background checks or different tax treatments for foreign workers. Figure 818 shows the labor market outcome of bilateral changes in migration costs to all type of migrants. We notice from Figure 8a that in general real income is increasing as moving costs are decreasing, with the exception of low-skilled Canadian workers (illustrated by a hump-shaped curve around the original equilibrium point). The general gains are due to improvements in global production efficiency. As shown in Figure 9d, the real per capita GDPs of both countries are increasing as moving costs are decreasing. However, the gains are not necessarily evenly distributed among all workers. Since the output share of low-skilled workers is much less than the output share of high-skilled workers, low-skilled workers are more likely to be substituted out by foreign workers. We can see from Figure 8a that low-skilled workers gain less on their real wages from lower moving costs compared to their high-skilled counterparts. 18 See Appendix B. 43 The gains in production efficiency come from two channels. First, we see from Figure 9a that the mass of all types of establishment is increasing as moving costs are decreasing. This is similar to the extensive margin gain of trade as recognized in Melitz (2003). Second, we can see in Figure 9c that aggregate productivity is shifting by changing the moving costs. The United States is gaining in productivity with lower moving costs, while Canada is losing productivity with lower moving costs. This is due to intra-industry reallocation of market shares due to interaction between migration and MNCs. This is similar to the intensive margin gain of trade but here we could also have intensive margin loss due to decreasing the moving costs. Overall, the first force dominates so that production efficiency (as measured by GDPs per capita) increases. Figure 10 shows that the simulation results for the moving costs change only for high-skilled migrants. In this case, the negative effect on aggregate productivity in Canada due to reducing the moving costs dominates and the real per capita GDP in Canada decreases. In terms of real wage, Canadian low-skilled workers lose from the deduction of moving costs, while other types of workers gain from the deduction. This shows that greater openness to migration may not always be beneficial to native workers. Here we have the counter example: Since the relative moving costs are even higher for Canadian low-skilled workers to move across the border, they are more likely to be substituted by other types of workers. Further, Canadian firms are relatively less compatible with lowskilled US workers (compared to compatibility between US firms and Canadian low-skilled workers), so the fact that Canadian low-skilled workers become relatively less mobile causes Canadian multinationals to lose their competitive advantages in US market. This in turn causes aggregate productivity in Canada to fall rapidly, reducing the real GDP per capita in Canada. 44 On the other hand, Figure 11 shows that if we only reduce the moving costs for low-skilled workers, we have mutual gains to all type of workers in both countries. The real per capita GDPs in the two countries are increasing as the moving costs are being reduced. Global production efficiency is improved due to greater openness to migration in this case. Comparing the three different scenarios above - the moving costs are reduced for all type of migrants, the moving costs are reduced for high-skilled migrants only, and the moving costs are reduced for low-skilled migrants only - we can see that policies that aiming for greater migration could potentially be mutual beneficial, but may benefit one country and hurt another. The key is whether the mobility of less mobile workers is improved. If the moving costs for less mobile workers are reduced, then we can achieve welfare gains for all type of workers in both countries. Otherwise, if the policy induces further relative immobility, then the immobile workers are negatively impacted by migration. 5.2.3 Migration Quota In this subsection, I consider bilateral migration quotas in three scenarios – 1. Migration quotas are applied to all types of migrants, 2. Migration quotas are applied to high-skilled migrants only, and 3. Migration quotas are applied to low-skilled migrants only. The results of counterfactual experiments with regard to bilateral migration quotas are presented in Figure 12 - Figure 14. Notice that in general migration quotas increase real wages for immigrant workers but reduce real wages for native workers who stay in their home country. We see some exceptions, for example, if migration quotas are only applied to high-skilled migrants, then Canadian low-skilled workers who stay in Canada 45 gain (as shown in Figure 13). Also, if migration quotas are only applied to low-skilled migrants, then Canadian high-skilled workers who stay in Canada gain (as shown in Figure 14). However, other types of workers who stay in their home country lose due to the quotas. In general, we do not see migration quotas achieving mutual gains for two countries, or even gains to all native workers within one country. Figure 15 shows real per capita GDPs in two countries for each scenario. Except for the real per capita GDP in Canada increasing as the migration quotas are applied to high-skilled migrants, real GDP per capita in general is decreasing as there are more constrains on migration. This reiterates the point that migration quotas are reducing international production efficiency. Even if in some cases a country may gain from quotas, we see from our simulation that the gains accrue more to immigrant workers than native workers who stay in their home country. 6. Conclusion This paper established a general equilibrium model to discuss the interaction between migration and multinational corporations and the welfare implications. Theoretically, I illustrated that the operations of multinational corporations are important for forming bilateral migration patterns. Second, migration can affect aggregate productivity through multinationals’ operations. Without MNCs, migration does not have any effect on aggregate productivity. The impact on productivity (and therefore welfare) of interaction between migration and multinationals should be something policy makers are aware of. 46 I calibrated the model to US and Canadian data to make explicit references on several migration policy changes. I considered scenarios when bilateral moving costs are reduced and researched the effects of bilateral migration quotas. I found that reducing bilateral moving costs in order to increase mobility of relatively less mobile workers (usually low-skilled workers) can improve welfare (measured by real income) for all types of workers in both countries. It improves foreign business opportunities for multinationals, which, in turn improves international production efficiency (measured by GDP per capita). On the other hand, migration quotas tend to reduce international production efficiency and hurt native workers who stay in their home country. The results lend supports to the view that greater openness to migration can bring mutual welfare gains. 47