Immigration and International Trade: a Semiparametric Empirical Investigation Kusum Mundra1 Department of Economics San Diego State University San Diego, CA 92122
[email protected] September 2003
1I
will like to acknowledge the guidance and support of Aman Ullah, Prasanta K. Pattanaik, and R. Robert Russell. Helpful comments were received from Gloria GonzalezRivera and seminar participants at the University of California, Riverside. Thanks to David M. Gould for providing the data. All errors are naturally mine.
Abstract This paper examines the effect of immigration on the US trade flows. The model hypothesizes that immigration facilitates international trade with home countries by lowering transaction costs. Immigrants also demand products from their country of origin, and thus stimulate trade. Using a panel data set I estimate a dynamic, fixed-effect model. The immigrant stock, a proxy for transaction costs, enters the model non-parametrically, whereas other variables enter the model log-linearly, as implied by the gravity model of international trade. To estimate this semiparametric model, I develop a new instrumental variable estimator with desirable asymptotic properties. The results indicate that the immigration effect on imports is positive for both finished and intermediate goods, but the effect on exports is positive only for finished goods. Keywords: International Trade, Immigration, Semiparametric Dynamic Panel
I. Introduction A dramatic increase in the size of the immigration flow has rekindled the debate on the global effect of immigration. Major changes in the national origin and skill levels of the immigrants have made immigration very controversial for the US. Two questions often come up in recent debates regarding immigration. First, how does immigration affect trade patterns? And conversely, how does the trade policy affect immigration patterns?1 Using a new semiparametric (SP) dynamic panel data framework, this paper studies the effect of immigration into the US on the trade flows between the US and the country of origin of the immigrants. At the same time, this research also analyzes an important question in international trade: whether the flow of labor encourages the flow of goods or the flow of goods discourages the flow of labor, in other words whether immigration and trade flows are complements or substitutes. This is an unresolved question in international trade and makes an excellent case for an empirical work. There are several reasons as to why one expects migration of people from one country to another to affect the trade flows between the two countries. Immigration affects the factor supply in the two countries under consideration, this, in turn, affects the production patterns.2 Immigrants also bring in better information about their country of origin and this may lower the transactions costs involved in trade.3 The assimilation of immigrants and natives may also affect trade flows. One form of assimilation might be that people from different countries become fluent in each others’ language; this might lower costs of trading. Immigrants are also active entrepreneurs with a high propensity for risk taking: the Jews of New York, the Japanese from San Francisco, Los Angeles and New York, the Cubans of Miami and the Chinese of New York are a few examples one can think of in this context.4 Another form of 1
The affect of trade on the domestic labor market is examined in Borjas et. all (1997). They show that in the manufacturing sector workers most affected by imports are disproportionately immigrants, women, blacks and less educated persons, whereas those most affected by exports are disproportionately native-born, nonblacks and educated men. Trefler (1997) shows that in standard Ricardian and Heckscher-Ohlin trade models trade policy changes affect wage inequality. 2 This is the effect in the traditional literature on trade and immigration. See among others, Mundell (1957), Markusen (1983), Bhagwati and Srinivasan (1983), Wong (1986), Ethier (1986), Razin and Sadka (1992), and Trefler (1997). 3 See, empirical works by Gould (1994), Rauch (1996, 1999). 4 On various studies and evidences of immigrant assimilation, there is a vast literature, see Light
1
assimilation might be that the immigrants introduce the natives in their country of settlement and the people in their country of origin to new products. Another effect is demand or preference of immigrants for goods from their country of origin, that opens trade in new channels. Remittances sent by immigrants to their countries of origin may also affect the trade flows. A general review of trade and immigration shows that, at least in the case of the US, the magnitude of immigration and the volume of trade have moved in the same direction in the last few decades. During 1966-1988, the average annual influx of immigrants into the US was to the extent of 481,513 persons and in the year 1990 it had increased to about 700,000, an increase of 45%. Also, US exports of goods as a percentage of Gross Domestic Product (GDP) has increased from 3.89 in 1960 to 8.03 in 1995, an increase of 135% and US imports of goods as a percentage of GDP has increased from 2.89 in 1960 to 10.42 in 1995, an increase of 261%.5 Immigration from the Asian and Latin American countries has increased dramatically after the passing of the Immigration Act of 1965, which did away with the earlier immigration quotas with these countries, see Figure 1. It is of interest that the share of the US trade with the major Asian, North American and South American newly industrialized countries (NIC’s) is also on the rise; see Figure 2 and Figure 3.6 The question here is, do the immigrants play any role in the goods trade between their home-countries and the US? In this paper, I concentrate on the effects of immigration on trade flows through the following channels. First, because immigrants’ carrying home-country information or forming networks lower transactions cost, better trading contacts are formed. Immigrants affect the factor supply in production both in the receiving country they have entered and in the sending country of their origin. Likewise immigrants affect the demand for goods in their host country and their country of origin. In this paper, I empirically estimate the effect of immigration on the US trade flows in a SP dynamic panel data model. My results show that the direct effect of immigration in and Bonacich (1988), Light and Rosenstein (1995). Lazear (1995), discusses the culture and language effect on the domestic trade. 5 The percentage increase is calculated from the value of imports and exports (billions of 1982 dollars) taken from the Survey of Current Business. 6 Here North America includes Mexico and the Caribbean countries. The data for U S exports and imports are from the Canadian World Trade Data Statistics.
2
terms of lowering transactions cost to trade is more positive for finished goods than for intermediate goods. The paper is structured as follows. Section II talks about the econometric model, in Section III I present a conceptual discussion of the various links that may exist between immigration and trade flows. In Section IV, I describe the estimation strategy and the data used in the study. In Section V, I discuss the estimated results and I conclude in Section VI.
II. The Econometric Model The SP dynamic panel data model is given as follows,
yit = αi + yit−1 γ + xit β + m(zit ) + uit
(1)
i = 1, ..., n t = 1, ..., T In the above model i = 1, ..., n denotes the cross-section and t = 1, ..., T stands for the time period, xit and zit are of dimension p and q respectively, β is a p×1 unknown parameter vector and αi is the cross-sectional fixed-effect. In the fixed-effect dynamic SP model given by (1), xit and yit−1 enter the model linearly and it is not known how zit effects yit , making the model nonparametric in z. In the dynamic error component model Li and Ullah (1998) and Berg et al. (1999) used a demeaning transformation after taking conditional expectation to eliminate the nonparametric part. I take a Taylor’s expansion of the model in (1) around a point z, thus rewriting (1) as follows, yit = αi + yit−1 γ + xit β + m(z) + (zit − z)m0 (z) + uit + R
(2)
where R includes the higher order terms of the expansion that asymptotically goes to zero, see Appendix. The model given by equation (2) can be transformed to: Yit = Y−1,it γ + Xit0 β + Zit m0 (z) + Uit
(3)
where Yit = yit −y i., Y−1,it = y−1,it −y −1i., Xit = xit −xi., Zit = zit −z i., Uit = uit −ui., T T T T P P P P P y i. = yit /T, y −1i. = t y−1it /T, xi. = xit /T, z i. = zit /T, ui. = uit /T. By t
t
3
t
t
taking deviations of the variables across cross-sectional mean the model given in (2) is demeaned in (3). Taking conditional expectation of (3) with respect to Zit we get E(Yit |Zit ) = E (Y−1,it |Zit ) γ + E(Xit |Zit )0 β + Zit0 β(z)
(4)
and subtracting (4) from (3) we get
Yit − E(Yit |Zit ) = (Y−1,it − E (Y−1,it |Zit )) γ + (Xit − E(Xit |Zit ))0 β + Uit
(5)
¯ it β + Uit Y¯it = Y¯−1,it γ + X
(6)
˜ it ρ + Uit Y¯it = X ˜ = (Y¯−1 X ¯ ) and ρ = (γ β). To get consistent estimate for ρ given in where X 1 ¯ it−1 (6), I assume that there exists instrumental variables X (a one-period lag of the ¡ ¢ 1 1 ¯ it−1 = 0 explanatory variable X ) for Y¯−1,it such that E Uit | X ¡
¢ 1 ¯ it−1 ¯ it X , X ¡ 1 ¢ 1 ¯1 whereX = Xit−1 − E Xit−1 |Zit it−1 ¡ 1 ¢ ¯ = X ¯ ¯ thus, W X Let, Wit =
The feasible OLS instrumental variable estimator for ρ is given by ¡ 0 ¢−1 ¡ 0 ¢ ˆ ¯ X ¯ ¯ Y¯ ρ= W W
(7)
To estimate ρ we need to know the unknown conditional expectations E(Ait | Zit )
1 where Ait is Yit , Y−1,it , Xit−1 , and Xit in (5). Following Robinson (1988), these can
however be estimated by the nonparametric kernel estimators Aˆit =
XX j
³
Zit −Zjs a
Ajs Kit , js /
s
XX j
´
Kit ,js
(8)
s
, j = 1, · · · , n ; s = 1, · · · , T, is the kernel function q Q k(Zit , `), where k and a is the window width. I use the product kernel K(Zit ) = where Kit , js = K
`=1
4
is the univariate kernel and Zit , ` is the `th component of Zit . Replacing the unknown √ conditional expectations in (5) by (8) gives ρˆ and that is N consistent estimator, for the consistency and normality results see Appendix. ˆ
To get the semiparametric estimator of m0 (z) we substitute ρ in (3) and get ˜ it ρˆ = Zit m0 (z) + Uit Yˆit = Yit − X the Kernel weighted semiparametric estimator of m0 (z) as ¡ ¢ X X Yˆit Zit K zit −z h ¡ ¢ m ˆ (z) = ΣΣZit K zith−z 0
(9)
The asymptotics of (9) are in the Appendix.
III. Immigration and International Trade: Conceptual Frameworks In this section, I outline some of the ways in which immigration may affect the volume and the pattern of trade. Transactions Cost: Immigrants in the US tend to settle where there already are big immigrant populations. The initial economic pull of labor markets and the subsequent reinforcement of migrant concentration through migrant networks (chain migration) causes this concentration of immigrant population, see Massey (1988). International migration forges innovative networks, which then reinforce the very migration that produced them according to Light and Rosenstein (1995). Immigrants are not only laborers but they are also active and influential entrepreneurs, within vast closed ethnic enclaves or immigrant economies in the receiving country. This emergence of enclaves is facilitated by the concentration of immigrants of the same origin with business expertise, on availability of effective labor and on capital. Immigrant groups in these enclaves are also middle-man minorities, and provide ethnic occupational niches for the new arrivals. In these enclaves labor is often provided by the newly arrived immigrants who in turn form networks encouraging further immigration. These networks facilitate new immigrants to learn different trades and help them to make a place for themselves in the new country; this builds a stronger contact with the immigrants’ home-country. In other words, because of strong social networking and the presence of social capital in these groups, they have active and strong 5
ties with their country of origin. Immigrants involved in retail trade have first hand information about the markets of their home countries and the market and demand patterns of the US and they also have contacts with big wholesale traders. Today immigrants are also better connected to their home country due to cheaper transportation costs. All these factors greatly reduces the transactions cost to trade and thus facilitates bilateral trade between the US and their country of origin. In Gould (1994) trade links and trading networks are formed due to the presence of immigrant population in the US. Immigrants are the link between their country of origin and their host country. The immigrants know about the home country preferences, foreign market information becomes cheaper to obtain in the US and immigrant contacts lead to trust building that makes trade negotiations easier to conduct.7 In international trade there is increasing evidences that proximity and preexisting ties forming trading networks are facilitating trade between nations. Rauch (1999), discusses the importance of the links: caused by geographical proximity, common language and colonial ties in matching international buyers and sellers trading in differentiated goods where brand names are important.8 Immigrants adapt to their new country, assimilating into what was a foreign society, with the second generation of the migrants being more fluent in English (and more often bilingual) than the first, which greatly reduces communication problems. The extent of assimilation depends on the skill level and the nationality of the migrants. These migrants enable the people of the US and the country of origin to have better information about each other, greatly facilitating bilateral trade. The longer the immigrants live in the US, the more likely they are to adapt to the US society, changing their preferences and life styles, and weakening their contacts with their country of origin. There are studies looking into the impact of immigrant skill levels on the receiving country’s economy. Immigrants with higher skill are in a better position to take advantage of the information they possess. “For modern immigrants 7
Koreans exports to the US substantially increased since the early 1970s, when a massive influx of Koreans to the US began. By virtue of the advantages associated with their language and ethnic background, many Korean immigrants have been able to establish businesses importing merchandise from Korea, see Min (1989). 8 Rauch (1999) has an interesting discussion about organized exchange and reference prices, he talks about Lead (SITC 685)a homogeneous good that is listed in almost all organized exchanges. Whereas a good like footwear (SITC 851) are not listed on any organized exchanges, and it is in this kind of differentiated product links play an important role in trading.
6
...the homeland is no longer something to be forsaken, released into a mist of memory or nostalgia. As the world has grown smaller, the immigrant experience has inevitably changed. Unlike the Europeans who fled persecution and war in the first half of this century, few modern immigrants abandon their motherlands forever, shutting one door, opening another and never looking back. Instead, they straddle between two worlds, in varying degrees, depending on where they come from and what they can afford.”-New York Times (July 19, 1998). Factor Supply : Immigration changes the factor supplies in the sending and the receiving countries; the population and the labor supply rises in the receiving country, and that of the sending country falls. The effect of immigration is similar to the traditional role of labor mobility in international trade. One assumes that an increase in the labor supply will increase the production and demand for goods in the receiving country, and the opposite happens in the immigrant sending country. In international trade theory, there is no such consensus, as to whether the goods and labor flows are substitutes or complements. In fact, the role of goods trade has been emphasized as an alternative to labor migration: “Given the difficulties posed by the prospect of large-scale migration from East to West, and the risk that such large-scale migration would actually leave worse-off the remaining population in the East, we need to ask what alternatives are available. Ideal policy should try to bring good jobs to the East rather than Eastern workers to the West. International trade ... can act as a substitute for migration”, Layard et al. (1992). It is well known that in the traditional Heckscher-Ohlin (H-O) framework, if the two countries have different factor endowments, then goods and factor flows are substitutes. But, if the factor endowments are same, if all the other well known H-O assumptions hold, and if the technologies in the two countries are different, then the goods and the factor flows are complements.9 International migration is motivated by differences in both factor endowments and technologies across countries, thus trade theory does not tell us unambiguously whether the immigrants and the trade in goods act as substitutes or complements. Demand Effect: In the traditional trade models, if we do not restrict immigrants to have similar demand patterns as natives, then the change in the terms of trade 9
See, Bhagwati and Srinivasan (1983), Bhagwati (1987)
7
and the volume of trade due to immigration is ambiguous, Mundra (1999). Different immigrant groups with different occupations and settlement patterns come to the US not simply as laborers affecting the labor market but they come as human beings with diverse cultures, cuisine’s and lifestyles comprising “Immigrant America”. The present patterns of immigration into the US are very different than those seen in the early part of this century. Today immigrants are of varying skill levels, including high skilled computer professionals from Asian countries and large unskilled immigrants from Latin American countries, especially Mexico. Immigrant populations have different demand patterns than natives and demand products from their home countries, which influence US imports. Some immigrants might be manufacturing products in the US for exporting to their countries of origin effectively reversing the movement of goods. The immigrants also assimilate into the host country society over time. As they are introduced to new products of the host country and their demand patterns are influenced by those of the natives.10 What is the effect of demand patterns on trade flows is an empirical question, investigating the connection between the trade and goods mobility for the Atlantic economy between 1870 and 1940 Collins et all. (1997) have found evidence supporting complementary flows of goods and labor. Other Factors: History tells us that the migration patterns have followed migration from peripheral countries to core countries. Studies of migration into the US from Mexico, Puerto Rico, Cuba and Dominican Republic show that the indigenous populations who do not migrate are influenced by the migrant populations who adopt the culture of the US. Most of the migration studies take into account the effect of people who move and do not analyze the effect on the people who do not. By looking into the movement of bilateral trade between the US and those trading partners that are also the source countries of the immigrants, I am not only accounting for the effect of migration on the US but also the effect it has on the immigrants home countries. 11
The type and quality of migrants brings on the one hand varying short term ad10
This might stimulate trade based on Linder (1961) hypothesis. Remittances sent by immigrants to their home country might be influencing the demand and activities of the people who do not migrate, this might have second or third round effects on their home country’s trade. Remittances inflow constituted about 8.5 per cent of the total value of world trade in services estimated at 770 billion U. S. dollars in 1990, second only to the crude oil trade, see Russell and Teitelbaum (1993). The large magnitude of current remittances might be effecting the growth of the immigrant sending countries, thus influencing trade flows. Exploring the effect of remittances is a future work. 11
8
justment costs and loss of resources for their countries of origin Money (1998), but on the other hand there are long term benefits because of the flow of resources from the scarce to the abundant and more productive regions. Thus the study of immigration involves many complex factors not only economic, ethnic, and sociological but also political, which effect the international relations of the US with other countries. In the US networks established by the migrants become stronger as immigration continues. Immigration increases further due to widening of the trade gap with various countries, and its effect is enhanced further by the immigrant networking which goes a long way in forming trade links between the US and the immigrants countries of origin.
IV. Data and Estimation Strategy A panel data model where the cross-section is the immigrants home-country, which is also the US trading partner is estimated. The model has a non-parametric transactions cost dependent on the immigrant stock; in other words no parametric functional form of how the immigrant stock affects the trade flows is specified. Other variables capturing different effects enter the model log-linearly. The theory behind the empirical model is derived from the gravity equation. Gravity equation has been very successful in empirical trade analysis but the theory behind the gravity equation is ambiguous. At the heart of the gravity equation is the idea that the volume of trade between two countries will be directly proportional to the Gross National Product (GNP) or the GDP of the two countries and inversely proportional to the geographical distance between the two.12 For the theory behind the gravity equation see, Bergstrand (1984, 1989, 1990), Deardorff (1995), Harrigan (1994). For empirical use of the gravity equation to test the trade theories, see Linneman (1966), Aitken (1973), Leamer (1974), Helpman (1987), Gould (1994), Rauch (1999).13 12
Yi and Yj is the GNP of country i and j. Let D be the distance between them, then the export (Y )a (Y )b
i j flow from country i to country j is, Xij = cci cj (1+eD f . ij ) 13 These works have used Heckscher—Ohlin models, monopolistic competition models in differentiated products and combination of the two type of models to provide the microfoundations behind the reduced form gravity equation. Deardorff (1995) has pointed out that gravity type equations are supported by many trade models and hence its empirical success is “mere fact of life”. Because of the different theoretical models behind the gravity equation, I think gravity equation works here, since we do not know which type of model holds between U S and the ith trading partner or the immigrant home-country.
9
I study the effect of immigration using a semiparametric fixed effect instrumental variable (SPFEIV) in a gravity-type model for bilateral trade where the dependent variable is US exports and US imports. In the SP case, there is a parametric functional form, the most common being linear, for some variables and no functional form assumption for others. The variables which enter the model linearly are the GDP, population, price deflators, and others, both for the US and the immigrant home country. In the SP estimation, I have assumed no parametric form for the transactions cost as a function of the immigrant stock.14 The variable used in the study is immigrant stock rather than immigration flow, this is crucial because the effect of immigration on the transactions cost is influenced not only by the current immigration flow but also by the past immigrant population. When the immigrants enter the US, they carry new information with them and there is a possibility of information not shared by the old and the new immigrants. Because of this the effect on trade due to a change in the immigrant stock maybe larger from a country which already has a vast pool of immigrant population, than from a country, with a smaller pool of immigrant population. The the invariant cross-sectional fixed effect for different immigrant sending countries captures the country specific effects like infrastructure, institutional character, language, culture and differences in factor endowment. The fixed-effect also captures the missing variables from the study, see Hsiao (1986). These country specific effects may affect bilateral trade flows and influence the trade from different countries in various ways. The empirical model is dynamic in nature, implying that one-period lagged US exports and US imports flows affect current trade flows. This is important for the analysis because it takes into account the lags and the adjustment in the international trade market. Dynamic model enables the past income and price levels to affect the current trade flows.
Motivation for a Semiparametric Framework I build a partial model to study the effect of immigrant stock on the US trade flows 14
In Gould (1994), the transaction costs as a function of the stock of immigrants from country i into the US is given by ZU S,i = Ae−ρ[Mi,US /(θ+Mi,U S )] ρ > 0, θ > 0, A > 0 where ZU S,i represents the transaction costs of trade associated with obtaining foreign market information about country i in the US, and Mi,U S represents the immigrant stock from country i into the US. The hypothesis Gould maintains is that immigrants bring with them information about the markets of their origin which decreases transaction costs.
10
for aggregate goods.15 The model is estimated both parametrically and nonparametrically. Based on the parametric functional form for the transaction cost given by Gould (1994), the non-linear parametric model for exports and imports is as follows: Exports: log EXU S,i = α1 (Mi,U S /(α2 + Mi,U S )) + εex
(10)
log IMi,U S = β 1 (Mi,U S /(β 2 + Mi,U S )) + εim
(11)
Imports:
The nonparametric partial model is given by yit = m(zit ) + uit i = 1, ..., N
(12) t = 1, ..., T
based on (10) in (12), y = log EXUS,j : log of exports of goods from the US to the home country i, z = Mi,U S : immigrant stock from the i − th country into the US and m(zit ) is specified as α1 (Mi,U S /(α2 + Mi,US )). Based on (11) in (12), yit =
log IMi,U S : log of imports of goods from the home country i to the US and where m(zit ) is specified as β 1 (Mi,U S /(β 2 + Mi,U S )) and t = 1, ...T. The models 10 - 13 are estimated as balanced panel models where the crosssection is the 47 trading partners of US, which are also the countries of origin of the immigrants. The time period under study is 1973-1980, hence T is eight years. The non-linear estimation of (10) and (11) is given in the following table16 :
Table 1 :
Parametric Non-Linear Estimation of Immigrant Stock Effect on US Exports and Imports
Immigration Information Variable Immigrant Sensitivity Variable
Aggregate Exports Aggregate Imports 3.1 3 (0.04) (0.06) 2290 3034 (212) (356)
15
Aggregate goods include both intermediate and final goods. The consistent test for a parametric regression model, in Li and Wang (1998) rejected the null, of the parametric functional form given in (10) and (11). The standard errors are given in the parenthesis. 16
11
The parametric partial model based on the parametric functional assumption of decreasing transaction cost due to immigrants carrying the home country information shows that the immigrants effect on bilateral trade flows is effective only when immigrant stock is small (Figures 4 and 5). If we plot the data on US exports and imports against immigrant stock, refer to Figure 4 and Figure 5 again, one sees that the immigrants’ effect is strong and steady at high immigrant stock level. This is also very intuitive, because the transactions cost and the demand effect will be increasing with the immigration levels. There is also no uniform positive change in the trade flows at all level of immigration from different 47 countries to US, as assumed in the parametric partial model. In the nonparametric estimator of (12), no functional form is assumed for the transaction cost, and m(zit ) is estimated by a kernel weighted local constant estimator, see Pagan and Ullah, (1999). This is a local averaging or smoothing estimator that fairly captures the movement in the trade flows. Nonparametric estimation shows that migrants do not always have a positive effect on trade. The transactions cost and the demand effect does not always hold, and when it does, it is greater for the aggregate US imports than for the aggregate US exports. Clearly, the migrants have other effects than the transactions cost effect on the host and home countries trade. For example, the US bilateral trade flows may be affected by the augmentation of labor supply resulting from immigration. Moreover, there are also other economic factors, such as income, prices, and population in the US and the trading partners. The ratio of skilled to unskilled workers in different immigrant populations, and their average length of stay needs to be accounted for in the empirical study. I assume that these other variables enter the model linearly, though for transactions costs I do not assume any specific functional form.
Data Semiparametric fixed effect dynamic models with instrumental variable are used to estimate the effect of immigrants’ links to their home countries. The balanced pooled panel is estimated for 47 US trading partners over eight years from 1973-1980. Annual data on immigration stock, skill levels and length of stay is from the U.S. census and Immigration and Naturalization public-use data. Data on aggregate trade flows is taken from International Monetary Funds (IMF) Direction of Trade Statistics 12
and other variables from IMF International Financial Statistics, (for details on the data, see Appendix B).
Semiparametric Model The SP full model is given by the econometric model discussed in Section II as follows,
yit = αi + yit−1 γ + xit β + m(zit ) + uit
(13)
(i = 1, ..., n; t = 1, ..., T ) where in (13) for US exports, y = log EXus,i , y−1 = EXt−1 , x is a vector of:
2 Yus , Yi , P OPus , P OPi , Pus , Pi , P Xus , P Ii , SKU Kus,i , ST AYus,i , ST AYus,i . For US
imports in (13), y = log IMi,us , y−1 = IMt−1 , x is a vector of: Yus , Yi , P OPus , P OPi , 2 Pus , Pi , P Xi , P Ius , SKU Kus,i , ST AYus,i , ST AYus,i . Where:
log EXus,i : Log of Export of US to the i − th country
EXt−1 : Export lagged one year
log IMi,us : Log of Import of US from the i − th country
IMt−1 : Import lagged one year
Yus and Yi : the US and home-country GDP P OPus and P OPi : the US and home country population Pus and Pi : the US and home country GDP deflators P Xus and P Ii : the US export unit value index and the home country import unit value index P Ius and P Xi : the US import unit value index and home country export unit value index SKUKus,i : the ratio of skilled immigrants to unskilled immigrants from home country i into the US. ST AYus,i : the average length of stay of the immigrants in the US zi : Immigrant Stock from country i αi : the country specific or the fixed cross sectional effect. In the model given by (13) m(zit ) is not assumed to have any parametric functional form and the country specific characteristics such as infrastructure, or culture, that are omitted from the study but nonetheless affect the export and import flows are 13
a part of the fixed effect. These omitted variables captured by the fixed-effect αi , are also correlated with the other independent variables, which is not the case if one ¡ z −z ¢ 17 it assumesna random effect model. A normal kernel is chosen, given as K = h o ¡ ¢ 2 √1 exp − 1 zit −z , where h is the window-width, c is a constant, and s is the 2 h 2π
standard deviation for variable z; for details on the choice of h and K see H¨ardle −1
(1990) and Pagan and Ullah (1999). The estimate, m ˆ 0 (z) is (nT hq+1 )
consistent,
see Appendix A. The choice of instruments for the lag of the dependent variable in the empirical study was USGDP for the US imports and for US exports the instrument for the lagged dependent variable is the home country GDP.18
V. Estimated Results To study the effect of changes in the immigrant stock on the US exports and imports, elasticity estimates m ˆ 0 (z) at different immigrant stock level for all types of goods aggregate, intermediate and finished goods is calculated. An advantage of using nonparametric methodology is that the elasticity can be estimated at every data point. This shows the US bilateral trade with the i − th country brought about
by an additional immigrant from that country. On this basis I calculate the average dollar value change (averaged over eight years) in the value of the US bilateral trade P 0 P flows as: ave(m ˆ 0i (z)) × z¯i , where, ave(m ˆ 0i (z)) = m ˆ (z) /T and z¯i = z it /T is t
t
the average immigrant stock into the US from the i − th country.19 These values are presented in Figures 6-11 and in Tables 6-8, where we see that immigrant effect
is positive for US imports across all goods but that is not the case for US exports. Immigrant effect is positive for finished exports but not for intermediate goods. This result is in contrast to the result by Gould(1994), where he finds that immigrant 17
Random Effect estimation was done with the following model for the given data, but the Hausman Test, see Hausman (1978), Hsiao (1986), Baltagi (1995), rejected the null, favoring the fixed effect model. In cross-country study the individual, unobservable invariant cross-sectional effects are mostly assumed as fixed effects, often broadly interpreted as the tariff agreements, culture, and the country specific institutions. 18 For the details on aggregate, intermediate and finished data and the countries under study, see Appendix. The choice of instrument is tricky and many times there is some question about what is a good instrument. Here my choice is dictated both by theory and statistics. Theory tells us that the imports of any country would depend on the income of that country. The correlation between the US imports and the US GDP in the given data is 0.86, the correlation between the US exports and the home-country GDP is 0.83, one can say reasonably high, for the real data. 19 I also estimated the elasticity for every cross-section or country at the average immigrant stock level, the results are similar.
14
effect is positive for all US bilateral trade. The estimated results show that the immigrant-networking effect on trade was strong for finished products. Finished products are differentiated goods, where brand names are important for trading and for finished products the country specific information carried by immigrants encourage trade.20 It is in the consumer goods industries that immigrants are actively involved in business and trading. In 1975 in the Koreatown in Los Angeles, 31.3 percent of Korean firms were involved in retail trade. Small businesses have been very prevalent among the immigrant groups and immigrants are more hard working with higher propensity towards self employment, see Light (1990). Thus, we see that the direct effect of immigrants in terms of lowering the transactions cost to trade supports trade flows in finished goods, both for exports and imports. Finished goods are differentiated products with inelastic demand and country specific information is crucial in the trading of consumer goods. The empirical finding supports this hypothesis. The immigrants affect intermediate imports positively, but not intermediate exports. In fact, for intermediate exports, the immigrants are lowering the US exports to the immigrants’ home countries. This is puzzling and an interesting result, in contrast to the earlier finding of immigration affecting both US exports and US imports positively across all types of goods, Gould(1994). This shows that immigration is lowering the income of the home countries, in turn adversely affecting their imports.21 The US immigration policy during the seventies was tilted towards the immigration of skilled and professional immigrants (popularly called “brain drain”) from the less developed countries to the US. High-skilled immigration must be more adversely effecting the national income of the home country than unskilled immigration.
22
The Immigration Act of 1965, gave preference to people who had job skills needed 20
For example in 1978, South Korea accounted for 2.2 percent of total US imports a rise from almost zero in the 1970s. The most important products from South Korea were clothing, veneers, footwear, and electrical machinery, and each of these exceeded $100 million. During this period South Korea was the biggest supplier of travel goods, handbags, fur goods, plastic articles and miscellaneous manufactures, see Light and Bonacich (1988) 21 To provide a more rigorous support to this argument, a detailed analysis of the effect of immigration on the home country is required. 22 The decline of the home-country imports and hence US exports depends on the past immigration from the home-country to the US, thus this argument holds inspite of many empirical studies showing that the skills and wages of immigrants into the US has been declining for the period 1960-1990, Borjas (1999).
15
in the US, or who had families in the US, or were refugees from some countries. Rising demands for medical care led to the enactment of Medicare and Medicaid Act of 1965. This created a sharp demand for physicians, nurses and other medical personnel. Other highly educated immigrants found a market for their skills as mathematicians, computer experts, physicists, chemists, and engineers. During the 1960s and 1970s, American universities had increasing numbers of foreign graduate students(mostly Asians), and few returned home. Most of them will find research and teaching positions in the US, where they had better working conditions than in their home countries, see Reimers (1996). This influx might have adversely affected the national income of the country of origin of the immigrants and negatively influencing their demand for imports of intermediate good and hence US intermediate exports. This is not seen for finished exports because the transaction cost effect of immigrants is increasing the trade flow and the demand for the finished products is more inelastic than the intermediate goods.23 The US government was recognizing more and more that the country’s future will critically depend on the skill level of the immigrant population. The advocates of the immigration of skilled workers proclaim that the political costs of integration of new immigrants into the social structure of the country are less than that for unskilled immigrants. Their ability to adapt to the local language and customs is greater. The skilled immigrants also draw less on the welfare state and thus make fewer demands on the public resources, see Borjas (1995). Critics of this view argue that to maximize the beneficial impact of immigration on those who are already the natives and residents of the US, the social marginal product of the immigrants should exceed their own private returns, Bhagwati (1996). This exodus of highly skilled populations might be hampering the growth of the immigrant home-countries, adversely affecting its imports, and hence US exports.24 The skilled immigrants are able to assimilate into the US society faster by simply breaking their ties or links with their home countries, rather than networking for trade links. This is what the results show in the aggregate 23
During the energy price hike of 1970’s, US export prices increased, making the US exports less competitive in the world market,. This is captured by the US GDP deflator and US export unit value index in the empirical analysis. 24 Immense work has been done in analyzing the role of immigration in the host country of the immigrants, though very little is done in terms of analyzing the effect of immigration on the home country of the immigrants.
16
and intermediate goods trade flows, though the immigrants had a positive effect on the US bilateral trade in finished goods. The estimated coefficient of other variables (included in the linear part of the model in equation 13) is reported in Table 2, Table 3 and Table 4, for aggregate, intermediate and finished goods, respectively. The higher the skilled-unskilled ratio of the immigrants, the higher is the effect of the immigrants on the trade flows between the US and the immigrant sending countries. One can think of various intuitive reasons for this result. With higher skill levels (here measured in terms of the level of education), immigrants have the human capital to carry better information about their country and to use the information effectively. This increases the flow of goods between the US and their home-countries. One can argue that higher skilled immigrants might have more entrepreneurial zeal and more access to social capital. Immigrants earning higher income prefer goods from their home countries, thus opening trade in new channels through the demand effect. Also, skilled workers are likely to be involved in import substituting production activities, because they carry the technology and the know-how to start production in the US; that is reflected in the negative coefficients for the US aggregate imports. The estimated results show that the length of the immigrants’ stay in the US does not favour trade. As the length of stay increases, trade with the country of origin falls and this fall is significant at one percent level for the import of finished goods. This might be because the country-specific information that immigrants carry becomes obsolete over time. The new waves of immigrants enter the country with fresh information and better contacts with the countries of their origin, but as they stay longer in the US, their contacts and information about the trading country declines. One can also argue that immigrants with higher skills are more scattered in the US than the lower skilled immigrants who are concentrated in large settlements. This scattering may be helping the immigrants with higher skill to assimilate better, but it also might be weakening their contacts with their home country. Thus, higher skills affect the trade flows in two ways, which work against each other and make the effect-of-stay variable indeterminate. The primary objective of this study is not to analyze the effect of price and income variables: the effect of the price and income variables depends on the relative mag-
17
nitudes of demand and supply elasticities between the two countries, the US and its trading partner or the immigrant home country.25 For example, the USGDP deflator affects the intermediate and aggregate exports negatively, but affects the finished exports positively. These are interpreted as the demand elasticities of substitution for the immigrant country of origin exceeding one for intermediate imports but not for the finished goods, in support of the fact that the imports of finished goods are more dependent on the country of origin than the imports of intermediate goods. Homecountry import unit value index affects the trade positively; this implies that the elasticity of substitution among importables exceeds that between domestic and imported products for the home-country. This contention has been supported by earlier works demonstrating that importables substitute more closely with each other than they do with domestic output. Also, say for the US imports, in the manufacturing sector for both the intermediate goods and the consumer goods, the US import unit value index affects the imports positively; this also holds for the trading partners. The demand elasticity of substitution among imports exceeds the overall elasticity between domestic and imported products for all the countries. From the results of the income and price variables, the findings are: First both for the US and all the trading partners included in the study, the substitutability across countries in imports is restricted, especially for consumer imports where country specific information plays an important part. Second, for all the countries, importables are closer substitutes to each other than they are with domestic output.
VI. Conclusion Using the new semiparametric instrumental variable technique in this empirical study, we see that the presence of immigrants in the US has an effect on the US bilateral trade flows with the immigrant home-countries. This panel data study shows that immigrants from different countries bring different magnitudes of effects on the value of trade. There is a positive effect on all US imports(aggregate, intermediate and finished) but not on the intermediate exports, though the effect on the finished goods exports is positive. This work shows that during the period 1973-1980, the immi25
For the discussion on different elasticities explaining the price and income coefficients for the trading partners, under the CES utility functions and the transformation function in production see Bergstrand (1984).
18
grants, by bringing better information and trading contacts with their home country, are supporting US bilateral trade in finished goods. The networking effect is strong for finished goods where country-specific information is important in trading, but that is not the case with intermediate goods. This result is in contrast to the earlier parametric empirical work by Gould (1994). Gould assumed a parametric functional form for the transaction cost as a function of the immigrant stock, with the results showing that the immigrants’ networking had a positive effect on the trade across all types goods, and the “immigrant-link effect” was stronger for US exports than imports. In this semiparametric data driven methodology, there is strong immigrant effect on the differentiated finished products but not on the homogeneous intermediate goods; this supports the hypothesis that immigrant networking is stronger for finished goods, Rauch (1999). The ratio of skilled to unskilled workers is significantly increasing trade in final goods. Together with other effects of immigration, the effect of immigrants on trade flows needs to be considered when formulating immigration policies. Further studies with bigger time periods to look across the effect of immigration on trade flows across business cycles is required; also intensive studies focussing on the big immigrant sending countries like Mexico and China will be very interesting and crucial for the US immigration policy with these countries.
19
Appendix A The assumptions that are needed for the consistency and asymptotic normal0
ity of ρˆ and m ˆ (z) are as follows. Following Robinson (1988) let Gλµ denote the class of functions such that if g ² Gλµ , then g is µ times differentiable; g and its derivatives (up to order µ) are all bounded by some function that has λ − th or-
der finite moments. Also, K2 denotes the class of non-negative kernel functions R k : satisfying k(v) v m dv = δom for m = 0 , 1(δom is the Kronecker’s delta), R k(v)vv 0 dv = Ck I(I > 0 , and k(u) = O((1+ | u |3+η )−1 ) for some η > 0. Further, R we denote k 2 (v)vv 0 dv = Dk .I. We now state the following assumptions: (A1) (i) for all t,
(yit , xit , zit , wit ) are i.i.d. across i and zit admits a density
4 function f ² G∞ µ−1 , E(x | z), E(y | z) and E(w | z) ∈ Gµ for some positive integer
µ > 2 (ii) E(uit | xit , zit , wit ) = 0 , E(u2it | xit , zit , ) = σ 2 (xit , zit ) is continuous in xit
and zit , and uit , ηit = xit − E(xit | zit ), ξ it = (wit − E(wit | xit ) have finite (4 + δ)th
moment for some δ > 0. (A2) k¯ ∈ Kλ ; as n → ∞, a → 0, na4λ → 0 and namax(2q−4, q) → ∞.
(A3) k ∈ K2 and k(v) ≥ 0 ; as n → ∞ , h → 0 , nhq+2 → ∞ and nhq+4 → 0.
Under the assumptions (A1) and (A2), the asymptotic distributions of the semi-
parametric estimator ρˆ follow from Li and Stengos (1996), Li (1996) and Li and Ullah (1998), Ullah and Mundra (1998). This is given by X √ nT (ˆρ − ρ) ∼ N(0 , σ 2
−1
)
P P −1 where = E(ξ 01 η 1 /T ) η0i = (η i1 , . . . , ηiT ). A consistent estimators for is P P P P P c −1 where c = 1 ˆ it )(Xit − X ˆ it )0 = 1 ˆ i )0 (Xi − X ˆ i ). (Wit − W (Wi − W nT nT i
i
T
The semiparametric estimators ρˆ depend upon the kernel estimators which may have random denominator problem. This can be avoided by weighting (2.5) by P P the kernel density estimator fˆit = fˆ(Zit ) = nT1aq j s Kit , js . This gives ρ ˆsp =
S −1 ˆ fˆ , (W −W ˆ )fˆ, (Y −Yˆ ) fˆ.. Finally under the assumptions (A1) to (A3) and (W −Wˆ )fˆ, (X−X) noting that (nT hq+2 )1/2 (ˆρ − ρ) = op (1), it follows from Kneisner and Li (1996) that for n → ∞ where
P
1
0
(nT hq+2 )−1 (m ˆ (z) − m0 (z)) ∼ N (0 , =
σ2 (z) f (z)
X
1)
Ck−1 Dk Ck−1 , Ck and Dk are defined above. 20
Appendix B Aggregate Trade Data on Exports and imports are constructed from the International Monetary Fund (IMF), Direction of Trade Statistics. Trade data on consumer and producer manufactured imports and exports are derived from the Organization for Economic Cooperation and Development (OECD) statistics on trade in manufactured goods. The 1980 U. S. census and the Immigration and Naturalization Service (INS) public-use data on early immigration provides annual information on the stock of immigrants in the U. S. and their skill levels and length of stay. Income, prices, and population are extracted form the IMF’s International Financial Statistics. List of countries included in the study Australia Brazil Canada Colombia Cyprus Denmark El Salvador Ethiopia
Singapore Spain Sri Lanka Sweden Switzerland Syria Tanzania Thailand
Finland
Trinidad
France
Tunisia
Greece
Turkey
Hungary Iceland India Ireland
United Kingdom West Germany Yugoslavia Zimbabwe
Israel Italy Japan Jordan Kenya Malaysia Malta 21
Morocco Netherlands New Zealand Nicaragua Norway Pakistan Philippines South Africa South Korea Consumer Manufactured Products ISIC Code
Description
3111
Preserved meat products
3112
Dairy products
3113
Canned fruits and vegetables
3114
Canned and preserved fish
3117
Bakery Products
3119
Cocoa and sugar confectioneries
3132
Wine products
3133
Malt liquors and malt
3134
Soft drinks and carbonated waters
3140
Tobacco manufactures
3212
Textiles excluding wearing apparel
3220
Wearing apparel
3232
Leather product excluding footwear
3240
Leather footwear
3312
Cane containers and small cane ware
3320
Wood furniture and fixtures
3523
Soap and cosmetics
3610
Pottery, china, and earthware
3620
Glass products
3811
Cutlery, and general hardware
3812
Metal furniture and fixtures
22
3819
Metal products excluding machinery and equipment
3832
Radio and TV equipment and apparatus
3833
Electric appliances and housewares
3843
Motor vehicles
3844
Motorcycles and bicycles
3852
Photographic and optical goods
3853
Watches and clocks
3901
Jewelry
3902
Musical instruments
3903
Sporting goods Producer Manufactured Products
ISIC Code
Description
3122
Prepared animal feeds
3211
Weaving and finishing textiles
3213
Knitting mills
3215
Rope and twine industries
3232
Fur dressing and dyeing industries
3313
Sawmills, planning, and other wood mills
3411
Pulp and paper
3412
Paper boxes, and paperboard
3511
Basic industrial chemicals
3512
Fertilizers and pesticides
3513
Synthetic resins, and plastic
3521
Paints, varnishes, and lacquers
3522
Drugs and medicines
3530
Petroleum products
3540
Miscellaneous petroleum and coal products
3551
Tire and tubes industries
3692
Cement, lime, and plaster
3699
Nonmetallic mineral products
3710
Iron and steel basic industries
3801
Metal scrap
23
3813
Structural metal products
3821
Engines and turbines
3822
Agricultural machinery
3823
Metal and woodworking machinery
3824
Industry machinery, excluding 3823
3825
Office, computing and accounting machinery
3831
Electric industrial machinery
3841
Shipbuilding and repairing
3842
Railroad equipment
3845
Aircraft
3849
Transportation and equipment
3851
Scientific equipment
24
References Aitken, N. D. (1973), ‘The Effect of EEC and EFTA on Euoropean Trade: A Temporal Cross-Section Analysis’, American Economic Review, Vol. 63, pp. 881-92. Baltagi, B. H. (1995). Econometric Analysis of Panel Data, John Wiley, New York. Bergstrand, J. H. (1984). ‘The Gravity Equation in International Trade: Some Microeconomic Foundations and Empirical Evidence’, Review of Economic and Statistics, Vol. 67, pp. 474-81. Bergstrand, J. H. (1989), ‘The Generalized Gravity Equation Monopolistic Competition, and the Factor-Proportions Theory in International Trade’, Review of Economics and Statistics Vol. 71, pp. 143-53. Bergstrand, J. H. (1990). ‘The Heckscher-Ohlin-Samuelson Model, the Linder Hypothesis and the Determinants of Bilateral Intra-Industry Trade’, Economic Journal, Vol. 100, pp. 1216-229. Berg, M. D., Q. Li and A. Ullah (1999). ‘Instrumental Variable Estimation of Semiparametric Dynamic Panel Data Models: Monte Carlo Results on Several New and Existing Estimators’, Working Paper, University of California, Riverside. Bhagwati, J. N. (1987). International Trade: Selected Readings, edited by J. Bhagwati. Bhagwati, J. N. (1996). US Immigration Policy, in Immigrants and Immigration Policy: Individual Skills, Family Ties, And Group Identities, edited by H. O. Duleep and P. V. Wunnava, Jai Press Inc. Bhagwati, J. N. and T. N. Srinivasan (1983). Lectures on International Trade, MIT Press. Borjas, G. J. (1995). ‘The Economic Benefits from Immigration’, Journal of Economic Perspectives, Vol. 9(2), pp. 3-22. Borjas, G. J., R. Freeman and K. F. Lawrence (1997). ‘How much do Immigration and Trade Affect Labor Market Outcomes?’, Brookings Papers on Eonomic Activity. Borjas, G. J. (1999). ‘The Economic Analysis of Immigration’, in O. Ashenfelter and D. Card (eds) Handbook of Labor Economics, Vol. 3, pp. 1697-1790. Collins, W. J., K. H. O’Rourke and J. G. Williamson (1997). ‘Were Trade and Factor Mobility Substitutes in History’, Discussion Paper, Centre for Economic Policy 25
Research. Deardorff, A. V. (1995). ‘Determinants of the Bilateral Trade: Does gravity Work in a Neoclassical World?’, NBER Working Paper Series, No. 5377. Ethier, W. J. (1986). International Trade Theory and International Migration, in Research in Human Capital and Development, Edited by O. Stark. Gould, D, M. (1994). ‘Immigrant Links to the Home Country: Empirical Implications for US Bilateral Trade Flows’, Review of Economics and Statistics, Vol. 76, pp. 302-16. H¨ardle, W. (1990). Applied Nonparametric Regression, Cambridge University Press. Harrigan, J. (1994). ‘Scale Economies and the Volume of Trade’, Review of Economics and Statistics, Vol. 76, 321-28. Hausman, J. A. (1978). ‘Specification Tests in Econometrics’, Econometrica, Vol. 46, pp. 1251-71. Helpman, E. (1987). ‘Imperfect Competition and International Trade: Evidence from Fourteen Industrial Countries’, Journal of the Japanese and International Economics, Vol. 1, pp. 62-81. Hsiao, C. (1986). Analysis of Panel Data, Econometric Society Monographs, Cambridge University Press. Kniesner, T. and Q. Li, (1994). ‘Semiparametric Panel data Models with Dynamic Adjustment: Theoretical Considerations and an Application to Labor Supply’, manuscript, University of Guelph. Layard, R., O. Blanchard, R. Dornbusch and P. Krugman (1992). East-West Migration, MIT Press. Lazear, E. P. (1995). ‘Culture and Language’, NBER Working Paper Series No.5249. Leamer, E. (1974). ‘The Commodity Composition of International Trade in Manufactures: An Empirical Analysis’, Oxford Economic Papers, 350-374. Li, Q. (1996). ‘On the Root-N Consistent Semiparametric Estimation of Partially Linear Models’, Economic Letters Vol. 51, pp. 277-85. Li, Q. and T. Stengos (1996). ‘Semiparametric Estimation of Partially Linear Regression Model’, Journal of Econometrics, Vol. 71, pp. 389-97.
26
Li, Q. and A. Ullah (1998). ‘Estimating Partially Linear Panel Data Models with One-Way Error Components’, Econometric Reviews, Vol 17(2), pp. 145-66. Li, Q. and S. Wang (1998). ‘A Simple Consistent Test for a Parametric Regression Function’, Journal of Econometrics, Vol. 87, pp. 145-165. Light, I (1990). ‘Local Economy and Ethnic Entrepreneurs’, ISSR Working Papers in the Social Sciences, Vol. 4(13). Light, I. and E. Bonacich (1988). Emmigrant Entrepreneurs, Berkley and Los Angeles: University of California Press. Light, I. and C. Rosenstein (1995). Race, Ethnicity and Entrepreneurship in Urban America, Aldine De Gruyter, New York. Linder, S. B. (1961). An Essay on Trade and Transformation, John Wiley and Sons, New York. Linneman, H. (1966). An Econometric Study of International Trade Flows, North Holland, Amsterdam. Markusen, J. R. (1983). ‘Factor Movements and Commodity Trade as Complements’, Journal of International Economics Vol. 14, pp. 341-56. Massey, D. S. (1988). ‘Economic Development and International Migration in Comparative Perspective’, Population and Development Review, Vol. 14, pp. 383413. Min, P. G. (1989). ‘Some Positive Functions of Ethnic Business for an Immigrant Community: Koreans in Los Angeles’, Final Report Submitted to National Science Foundation. Money, J. (1998). ‘The Management of International Migration: Short-Term Dislocations Versus Long-Term Benefits’, IGCC Policy Paper No. 34. Mundell, R. (1957). ‘International Trade and Factor Mobility’, American Economic Review Vol. 67, pp. 321-35. New York Times July 19, (1998). The New Immigrant Tide: A Shuttle Between Worlds- Holding on to the Homeland. Pagan, A. and A. Ullah (1999). Nonparametric Econometrics, Cambridge University Press. Rauch, J. E. (1996), ‘Trade and Search: Social Capital, Sogo Shosha, and Spillovers’, NBER Working Paper Series No.5618.
27
Rauch, J. E (1999). ‘Networks Versus Markets in International Trade’, Journal of International Economics, Vol. 48, 7-35. Razin, A. and E. Sadka (1992). ‘International Migration and International Trade’, NBER Working Paper 4230. Reimers, D. M. (1996). Third World Immigration to the United States, in Immigrants and Immigration Policy: Individual Skills, Family Ties, and Group Identities, edited by H. O. Duleep and P. V. Wunnava, JAI Press Inc. Robinson P. M. (1988). ‘Root-N Consistent Semiparametric Regression’, Econometrica, Vol. 56, pp. 931-54. Russell S. S. and M. S. Teitelbaum (1993). ‘International Migration and International Trade’, World Bank Discussion Papers 160. Trefler, D. (1997). ‘Immigrants and Natives in General Equilibrium Models’, NBER Working Paper Series No. 6209. Ullah, A. and K. Mundra (2001). ‘Semiparametric Panel Data Estimation: An Approach to Immigrant Homelink Effect on U.S. Producer Trade Flows’, in Handbook of Applied Econometrics and Statistical Inferences, Marcel Dekker Wong, K. (1986). ‘Are International Trade and Factor Mobility Substitutes?’, Journal of International Economics, Vol. 21, pp. 25-43.
28
FIGURE 1: IMMIGRATION INTO THE U. S. BY REGION, 1975-1995 1400000
NUMBER OF IMMIGRANTS (THOUSANDS)
1200000
1000000
800000
600000
400000
200000
0 75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
YEAR
ASIA
NORTH AMERICA
SOUTH AMERICA
EUROPE
AFRICA
95
FIGURE 2: U. S. EXPORTS TO DIFFERENT REGIONS AS A PROPORTION OF U. S. GNP 0.035
0.03
EXPORT/GNP
0.025
0.02
0.015
0.01
0.005
0 70
71
72
73
ASIA
74
75
76
77
78
NORTH AMERICA
79
80 81 YEAR
82
83
SOUTH AMERICA
84
85
86
EUROPE
87
88
89
AFRICA
90
FIGURE 3: U. S. IMPORTS FROM DIFFERENT REGIONS AS A PROPORTION OF U. S. GNP 0.06
0.05
IMPORT/GNP
0.04
0.03
0.02
0.01
0 70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
YEAR
ASIA
NORTH AMERICA
SOUTH AMERICA
EUROPE
AFRICA
90
Table 2: Bilateral Aggregate Trade Flows Between the U. S. and the Immigrant Home Countries - SPFEIV Estimates
Dependent Variable Lag Dependent Variable
U.S Export Unit Value Index
U. S. Aggregate Exports
U. S. Aggregate Imports
-0.42
0.29
(0.28)
(1.06)
1.01 (0.65)
Home-country Import Unit Value Index
0.13 (0.12)
U. S. Import Unit
-0.33
Value Index
(0.32)
Home-country Export Unit Value
-0.04
Index
(0.15)
Immigrant stay
Immigrant stay (squared)
Immigrant Skilled-Unskilled ratio
U.S GDP Deflator
Home-country GDP Deflator
0.04
-0.14
(0.05)
(0.10)
-0.003
0.006
(0.003)
(0.002)a
0.007
-0.008
(0.02)
(0.02)
-0.26
10.19
(5.24)
(2.46)a
-0.35
0.03 a
U.S GDP
(0.07)
(0.12)
-5.14
5.33 a
Home-country GDP
(2.15)
(3.18)c
0.78
0.26 a
U.S. Population
Home-country Population
(0.20)
(0.13)b
58.12
-116.47
(44.61)
(55.43)a
-0.03
-0.85
(0.59)
(0.20)
Standard error in parentheses. a Significant at one percent level. b Significant at five percent level. c Significant at ten percent level
Table 3: Bilateral Manufactured Intermediate Trade Flows Between the U. S. and the Immigrant Home Countries: SPFEIV Estimates
Dependent Variable Lag Dependent Variable
U. S. Producer Exports 0.56
-0.46 c
(0.31) U.S Export Unit Value Index
U. S. Producer Imports
(0.62)
1.96 (0.35)a
Home-country Import Unit Value Index
0.26 (0.10)a
U. S. Import Unit
0.76
Value Index
(0.66)
Home-country Export Unit Value
0.32
Index Immigrant stay
Immigrant stay (squared)
Immigrant skill-unskill ratio
U.S GDP Deflator
Home-country GDP Deflator
U.S GDP
Home-country GDP
(0.21) -0.04
-0.20
(0.04)
(0.08)a
0.004
0.009
(0.002)c
(0.005)c
0.008
0.04
(0.01)
(0.04)
-10.51
6.79
(3.21)
(6.08)
-0.10
-0.20
(0.08)
(0.17)
-2.13
7.01
(1.48)
(4.54)
0.12
0.07 c
(0.07) U.S. Population
76.63
-108.42 a
Home-country Population
(0.24)
(18.35)
(83.58)
0.99
-1.55 a
(0.35)
(1.16)
Standard error in parentheses. a Significant at one percent level. bSignificant at five percent level. c Significant at ten percent level
Table 4: Bilateral Manufactured Finished Trade Flows Between the U. S. and the Immigrant Home Countries - SPFEIV Estimates
Dependent Variable Lag Dependent Variable
U.S Export Unit Value Index
U. S. Consumer Exports
U. S. Consumer Imports
0.17
0.17
(0.46)
(0.21)
0.46 (1.05)
Home-country Import Unit Value Index
0.19 (0.17)
Home-country Export Unit Value
0.46
Index
(1.20)
U. S. Import Unit
0.19 (0.03)a
Value Index Immigrant stay
Immigrant stay (squared)
Immigrant Skilled-Unskilled ratio
U.S GDP Deflator
Home-country GDP Deflator
-0.11
-0.11
(0.09)
(0.008)a
0.005
0.005
(0.006)
(0.00003)a
0.004
0.004
(0.03)
(0.0007)a
4.82
4.82
(9.18)
(84.24)
-0.29
-0.29 a
U.S GDP
Home-country GDP
U.S. Population
Home-country Population
(0.09)
(0.008)a
-2.57
-2.571
(3.12)
(9.74)
0.29
0.29
(0.22)
(0.05)a
-3.46
-3.46
(71.95)
(5177)
-1.55
-1.55
(0.81)
(0.66)a
Standard error in parentheses. a Significant at one percent level. bSignificant at five percent level. c Significant at ten percent level
Table 5: Estimated Average Dollar Value Change in the U. S. Bilateral Aggregate Trade Between 1973-1980 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
COUNTRY AUSTRALIA AUSTRIA BRAZIL CANADA COLOMBIA CYPRUS DENMARK EL SALVADOR ETHIOPIA FINLAND FRANCE GREECE HUNGARY ICELAND INDIA IRELAND ISRAEL ITALY JAPAN JORDAN KENYA MALAYSIA MALTA MOROCCO NETHERLANDS NEW ZEALAND NICARAGUA NORWAY PAKISTAN PHILIPPINES S. AFRICA S. KOREA SINGAPORE SPAIN SRI LANKA SWEDEN SWITZERLAND SYRIA TANZANIA THAILAND TRINIDAD TUNISIA TURKEY U. K. W. GERMANY YUGOSLAVIA ZIMBABWE
AGGREGATE EXPORTS -17787.3 -131899 -17819.8 -638207 -36678.1 -3661.39 -30782.8 -16740.9 -3525.42 -21181.5 -80291 -117599 -131816 -2618.31 -39110.8 -191142 -23534.8 -993344 -71059.3 -8154.2 -2541.5 -3425.5 -5981.4 -4578.5 -78920.8 -5857.2 -15194.2 -53093.3 -9950 -129175 -5864.5 -20434.4 -1916.3 -44778.1 -2192.7 -67115.3 -27355 -9447.1 -1139.8 -18438.5 -23395.7 -1791 -26119.4 -574596 -831465 -106827 -1376.32
AGGREGATE IMPORTS 131087 495155 132411 1070635 174569 30926 202225 110900 30345 150459 344810 369866 493284 22090.8 133198 560241 150998 1408478 210956 68334.7 22310.2 28926.9 48898.8 38542.8 371186 49090.6 101402 297133 76851.9 173694 50410 137854 16639.7 258755 18951.2 343831 184924 77077.9 9817.7 114208 162019 15143.1 174552 593933 1340906 419197 11898.8
Table 6: Estimated Dollar Value Change in the U. S. Bilateral Intermediate Trade Between 1973-1980 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46
COUNTRY AUSTRALIA AUSTRIA BRAZIL CANADA COLOMBIA CYPRUS DENMARK EL SALVADOR ETHIOPIA FINLAND FRANCE GREECE HUNGARY ICELAND INDIA IRELAND ISRAEL ITALY JAPAN JORDAN KENYA MALAYSIA MALTA MOROCCO NETHERLANDS NEW ZEALAND NICARAGUA NORWAY PAKISTAN PHILIPPINES S. AFRICA S. KOREA SINGAPORE SPAIN SRI LANKA SWEDEN SWITZERLAND SYRIA TANZANIA THAILAND TRINIDAD TUNISIA TURKEY U. K. W. GERMANY YUGOSLAVIA
PRODUCER EXPORTS -32509.9 -107550 -32307.2 -434914 -38789.1 -8122.5 -50091.3 -27159 -7907.29 -38014.1 -74928.2 -59096 -100547 -5851.2 -30404.4 -106130 -36632.3 -576858 -46314.2 -17173.8 -5805.4 -7602.49 -12821.7 -10065.4 -81581.4 -12706.2 -26836.2 -71006.9 -19702.5 -41984.4 -12814 -32124.5 -4373.2 -60010.1 -4980.64 -79293.1 -45493 -19610.8 -2605.04 -30928.1 -35346.2 -4031.8 -42714.5 -264442 -541845 -76586.3
PRODUCER IMPORTS 221791.4 888320.4 222660.9 1538429 301829.3 52577.1 346592.4 188930.9 51364.2 257169.7 603547.7 634359 869802.1 37664.7 245517.7 1019169 257629.3 2467604 396748.9 114588.3 37676.7 49185.3 83351.9 65425.128 642411.5 83121.3 1177076.8 514030.1 130732.9 403570.1 584640 230839.2 28221.1 441513 32155.5 595827 315165.2 130125.7 16716 202750.6 265160.8 25857.5 297336.3 1480497 2024962 711211.6
47
ZIMBABWE
-3146.3
20229.2
Table 7: Estimated Dollar Value Change in the U. S. Bilateral Finished Trade Flows Between 1973-1980. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45
COUNTRY AUSTRALIA AUSTRIA BRAZIL CANADA COLOMBIA CYPRUS DENMARK EL SALVADOR ETHIOPIA FINLAND FRANCE GREECE HUNGARY ICELAND INDIA IRELAND ISRAEL ITALY JAPAN JORDAN KENYA MALAYSIA MALTA MOROCCO NETHERLANDS NEW ZEALAND NICARAGUA NORWAY PAKISTAN PHILIPPINES S. AFRICA S. KOREA SINGAPORE SPAIN SRI LANKA SWEDEN SWITZERLAND SYRIA TANZANIA THAILAND TRINIDAD TUNISIA TURKEY U. K. W. GERMANY
CONSUMER EXPORTS 104974.9 379468.1 106546.4 187431 137278.9 24666.8 160339.6 88255.6 24288.3 119529.9 269218.9 293528.2 383590 17578.6 98706.6 425664.1 120091.3 221486.2 154323.2 55099.5 17889.7 23068 38920.5 30781 292461.4 39283.1 79102 233962.8 61258 88784.3 40605.3 111248.8 13297.2 205970 15139.1 270493.5 147271.2 61826.6 7820.9 87852.5 133060.3 12035.2 139052.4 93202.3 232718.7
CONSUMER IMPORTS 125384.9 452770.6 127026.9 354642.1 163329.7 29581.6 191870.7 105441 29094.3 143147.3 320314.3 341527.5 454721.3 21101.1 118633.3 503278.6 143427.5 462836.1 185304 65772.7 21419.9 27665.9 46691.2 36892.4 347765.7 47040.8 95323.8 279720.9 73369.1 112145.3 48511.1 132200.4 15941 245319.9 18150.4 322721.9 176012.3 73930.9 9386.9 106319.7 157063.7 14454.8 166123.7 202102.9 447964.9
46 47
YUGOSLAVIA ZIMBABWE
335818.7 9490.1
393743.4 11385.3
