1 Economic Theory and International Migration Author(s): George J. Borjas Source: International Migration Review, Vol. 23, No. 3, Special Silver Anniversary Issue: International Migration an Assessment for the 90's (Autumn, 1989), pp Published by: The Center for Migration Studies of New York, Inc. Stable URL: Accessed: 10/03/ :43 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact The Center for Migration Studies of New York, Inc. is collaborating with JSTOR to digitize, preserve and extend access to International Migration Review.
2 Economic International Theory Migration And George J. Borjas University of California, Santa Barbara and National Bureau of Economic Research The modern literature on the economics of immigration focuses on three related issues: 1) what determines the size and skill composition of immigrant flows to any particular host country; 2) how do the immigrants adapt to the host country's economy; and 3) what is the impact of immigrants on the host country's economy? This article reviews the theoretical framework and empirical evidence provided by the economics literature on these questions. It demonstrates that the economic approach, using the assumptions that individual migra? tion behavior is guided by the search for better economic opportunities and that the exchanges among the various players are regulated by an immigration market, leads to substantive insights into these issues. Economics studies the allocation of scarce resources among alternative uses. Labor is a scarce resource that maybe "allocated" to different labor markets. An economic theory of immigration analyzes the allocation of labor across international boundaries. The theory is based on the behavioral assumption that individuals migrate because it is in their benefit (either in terms of psychic satisfaction or income) to do so. Individual behavior, of course, is constrained by their wealth and by the existence of immigration policies that limit (or encourage) the entry of persons into particular geographic areas. In this framework, the economic approach to the theory of immigration addresses three questions: 1. What factors determine the direction, size and composition of the immigrant flow? That is, given any initial sorting of the population across countries, international differences in income opportunities, political conditions and immigration policies imply that incentives exist for some individuals to migrate to other countries. An economic This research was funded by a grant from the National Institute of Child Health and Human Development, Grant No. 1 R01 HD IMR Volume xxiii, No
3 458 INTERNATIONAL MIGRATION REVIEW theory of migration must provide some insights into which way the flows go, how large the flows are and which kinds of individuals become immigrants. 2. How do immigrants adapt to the host country? After migration takes place, immigrants find themselves in a foreign (and sometimes hostile) environment. A learning process about the host country's cultural, political and economic characteristics begins to take place and the immigrant begins to "assimilate." An economic theory of migration should describe the process by which this type of assimilation takes place and describe which factors make it more likely for successful assimilation to take place. 3. What is the impact of immigration on the economies of the sending and receiving countries? The large migration flows that occur across international boundaries will lead to significant changes in economic conditions in both the source and host countries. A theory of immigra? tion should describe the adjustments that take place in the various labor markets as the flows occur. No single, unified theory of immigration that simultaneously addresses all these issues yet exists. Instead, a number of theories or hypotheses have been developed to explore each (or a specific aspect) of the various questions individually. Thus we have theories (e.g., Hicks; 1939 and Sjaastad, 1962) which view migration as a human capital investment, and hence imply that migration is more likely the higher the returns and the lower the costs. Other theories explain the "brain drain" in terms of asymmetric information regarding the skill level of immigrants: the host country has more informa? tion about immigrant skills than the source country (Kwok and Leland, 1982). Finally, some theories conjecture that the impact of migration on natives in the receiving country is likely to be small because immigrants take (or are forced to take) jobs that natives refuse to accept (Piore, 1979). All these theories, of course, focus on extremely narrow topics within the economics of immigration and may not even be logically consistent with each other. However, recent analytical developments make it likely that a comprehensive economic theory of immigration can and will be developed. These recent models, based on the neoclassical principles of utility-maxi? mization for individuals and pro fit-maximization for employers, have already provided insightful interpretations of empirical observations. This study presents a survey of this recent literature. Before proceeding to a discussion of these models, it is instructive to begin with what are perhaps the earliest analytical studies of immigration in the economics literature. The question of factor (in particular labor) mobility, of course, forms an integral part of international trade theory. The standard
4 ECONOMIC THEORY AND INTERNATIONAL MIGRATION 459 model in trade theory, due to Hecksher, Ohlin and Samuelson, begins by assuming that labor is immobile across countries. In the absence of immigra? tion flows, neoclassical trade theory yields two fundamental theorems that are very relevant to the study of migration (Ethier, 1986; Jones, 1987): 1. The Hecksher-Ohlin Theorem: A country exports goods that make relatively intensive use of the country's relatively abundant factors. That is, if a country has a relatively large population (i.e., it is relatively abundant in labor), and if labor is intensively used in the production of, for example, textiles, the country will export textiles. 2. The Factor-Price Equalization Theorem: Free trade in goods will equalize the prices of factors across countries. That is, even in the absence of migration flows, the existence of free trade in goods will lead to the equalization capital across countries. of the wage rates of labor and the price of These two theorems imply that since a labor abundant country is export? ing those goods that are relatively intensive in the production of labor, it is, in a sense, exporting labor. The export of labor intensive goods leads to the equalization of wage rates across countries even if labor itself is immobile. In other words, the trading of goods substitutes for the trading of people. The introduction of immigration into the Hecksher-Ohlin-Samuelson framework, therefore, does not fundamentally alter the results of the analysis since the international immigration of income-maximizing persons is simply another way of ensuring that factor prices are equalized across countries. This theoretical structure addresses some of the questions that were posed at the outset: which way do immigrant flows go; what happens to the economies of the sending and receiving countries; what will be the size of the immigrant flow? In addition, the Hecksher-Ohlin-Samuelson framework has the advantage that it treats migration flows and goods flows symmetrically. Hence it presents a systematic study of the "internationaliza? tion" of the world economy, whether this internationalization is caused by the trading of goods or by the trading of people. Unfortunately, the model becomes very complex when it is expanded beyond the simplest 2x2x2 framework (i.e., 2 countries, 2 goods, 2 factors of production). Hence, it becomes quite difficult to analyze such questions as the composition of the 9 The derivation of the two theorems requires a number of technical assumptions that have not been discussed in the text. These include assumptions about the preferences of consumers in both countries and about the technologies used in the production of goods. The introduction of migration into the Hecksher-Ohlin-Samuelson framework allows the derivation of the Factor- Price Equalization Theorem even if some of the technical assumptions do not hold (Ethier, 1987).
5 460 INTERNATIONAL MIGRATION REVIEW immigrant flow, the impact of changes in immigration policy, etc. Recent theoretical developments, therefore, ignore the international trade aspects of labor migration and focus solely on the study of migration flows. They have borrowed, however, one of the key insights of the inter? national migration literature: that there exists an "immigration market." Just as goods are traded across international boundaries in the international goods market, people are also "traded" across the same boundaries in the immigration market. The survey begins with a description and charac? terization of equilibrium in this marketplace. THE IMMIGRATION MARKET The key idea guiding recent theoretical research in the economics of immigration is that there exists an "immigration market" sorting im? migrants across potential host countries (See, Borjas, 1987c). Individuals residing in any source country consider the possibility of remaining there or of migrating to a number of potential host countries. Individuals make the migration decision by considering the values of the various alternatives, and choosing the option that best suits them given the financial and legal constraints that regulate the international migration process. These constraints include not only the individual's financial resources (which determine the extent to which the migration is feasible), but also include the legal environment imposed by both sending and receiving countries. For instance, host countries legislate immigration policies based on the immigrant's skills, wealth, occupation, political background and/or family relationships with residents of the host country. These various statutes generate variations in migration costs among different individuals or among individuals originating in different countries. In a sense, host countries "compete" through these immigration policies for the human and physical capital of the potential migrants. Host countries with a certain set of immigration regulations (for example, those that make it easy to migrate if the potential immigrant has a family member in the host country) attract different types of persons than host countries with other sets of regulations (for example, a point system based on the individual's skills or occupations). Similarly, source countries may regulate the departure of their residents. In some countries (e.g., the Soviet Union) emigration legislation imposes fines or penalties on potential migrants and makes it very difficult for residents to leave the source country. Source countries, therefore, enter the immigration market by setting emigration policies that regulate the types of individuals that can obtain exit visas. Neoclassical economic theory assumes that individuals maximize utility: individuals "search" for the country of residence that maximizes their well-being. As noted above, the search is constrained by the individual's
6 ECONOMIC THEORY AND INTERNATIONAL MIGRATION 461 financial resources, by the immigration regulations imposed by competing host countries and by the emigration regulations of the source country. In the immigration market the various pieces of information are exchanged and the various options are compared. In a sense, competing host countries make "migration offers" from which individuals compare and choose. The information gathered in this marketplace leads many individuals to con? clude that it is "profitable" to remain in their birthplace (i.e., they find it expensive to migrate to another country). Conversely, other individuals conclude that they are better off in some other country. The immigration market nonrandomly sorts these individuals across host countries. An im? portant contribution of economic theory equilibrium sorting that takes place in this marketplace. is to describe the kind of It is important to note that although the idea of an immigration market is somewhat novel in the immigration literature, the notion that different agents are considering the allocation of resources among alternative uses and that this allocation is guided by a market basically defines economics. Formally, there is little difference between the problem of allocating in? dividuals among countries and the problem of allocating individuals among jobs. In both problems, individuals consider a number of options to which they can allocate their time. Firms (or countries) offer different "employ? ment" contracts. Individuals compare these offers and nonrandomly sort themselves among the available jobs (countries). This approach to the economics of immigration makes it very clear that both host and source countries can have a major impact on the number and composition of the immigrant flow by altering immigration policies. Similarly, changes in the levels of economic activity in the various countries will also have a major impact on the size and composition of the immigrant flow since these changes basically alter the nature of the "offer" made by competing countries to potential migrants. It will be seen that this approach leads to a very clear? and empirically testable?categorization of the types of immigrant flows that arise in a world where individuals search for the "best" country. A MODEL OF IMMIGRATION The characteristics of the sorting generated by the immigration market are best understood in a simple model where there are only two countries, the source country and the host country (See, Borjas, 1988; Roy, 1951). The generalization to a larger number of countries complicates the technical aspects of the model without fundamentally changing the key characteristics of the sorting. In addition, it may be assumed that individual migration is guided by comparisons of incomes across countries. Obviously, incomemaximization is a very strong assumption since individuals also consider
7 462 INTERNATIONAL MIGRATION REVIEW other aspects of the countries in their migration decisions (e.g., weather, culture, the crime rate, etc.). It can be shown, however, that (under some assumptions) income-maximization is a necessary condition for utility max? imization (Hirshleifer, 1970). The income-maximization hypothesis has been used successfully in the human capital literature, and its use here is justifiable because the model then leads to an empirically testable charac? terization of the optimal sorting of immigrants across countries. Residents of the source country have (log) earnings which are charac? terized by the earnings function: logw0 = X(50 + e0 (1) where w0 are the individual's earnings in the source country (country 0), X is a vector of observable demographic characteristics (such as education and age) and e0 is a random variable which is assumed to be normally distributed with mean zero and variance a ^. In addition, the disturbance e0 is assumed to be uncorrelated with the socioeconomic variables X. It is useful to interpret e0 as the component of earnings associated with unobserved "ability" or "luck" among individuals with the same observable skills (i.e., with similar demographic variables X). It should also be noted that the assumption that e0 is normally distributed, although standard in the literature, is quite applicable for the problem at hand. Since the logarithm of earnings is assumed to be normally distributed, earnings will be log-nor? mally distributed and positively skewed (i.e., a long tail to the right of the earnings distribution). The assumed shape of the earnings distribution, therefore, is quite close to the actual shape of observed income distributions for many countries (Lydall, 1968). The earnings structure facing individuals participating in the labor market of the host country (Country 1) is given by: logw1 = X61 + e1 (2) where el is also assumed variance a ^ and el is also independent of X. to be normally distributed with mean zero and The random variables e0 and el have correlation coefficient/?. If/? is positive and near unity, the labor markets of the host country and the source country "value" unobserved ability in the same way. That is, persons who are able or "lucky" due to unobservable factors perform well in both countries. On the other hand, if/? is small or negative, persons who do relatively well in one country find that the other country does not value their skills as highly (and perhaps will attach a negative value to them). Unobserved skills are,
8 ECONOMIC THEORY AND INTERNATIONAL MIGRATION 463 in a sense, "specific" to the country of origin. It seems reasonable to suppose that for most pairings of source and host countries, the parameter/? is likely to be positive and large. There are, however, interesting cases where this assumption need not be true. A few technical points about equations (1) and (2) are worth noting. First, note that the variables w0 and wl do not necessarily have to represent the individual's earnings in any given year. Instead they can be defined in terms of the present value of the earnings profiles of the individual in each of the two potential countries of residence. This interpretation is quite sensible since it says that migration behavior is determined by a comparison of lifetime earnings across potential countries. Second, the parameter vectors <50 and d l indicate the price that the source and host countries are willing to pay for the individual's socioeconomic characteristics. For instance, suppose that one of the variables in the X vector is education. Then its coefficient in (1) indicates the percentage return to an additional year of education in the source country, or the "rate of return" to education in the source country, while the coefficient of education in (2) gives the rate of return to education in the host country. Finally, note that because the random terms e0 and el are assumed to have zero mean, the term Xc50 gives the expected earnings for an individual with demographic characteristics X in the source country, while the term Xdl gives the expected earnings for an individual with the same demographic characteristics in the host country. In other words, the terms Xc50 and Xd? give the mean earnings of a person with characteristics X selected at random from the population. This "random" person, by con? struction, has unobserved characteristics with value zero. Suppose that individuals considering the possibility of migrating from the source country to the host country face mobility costs C. Define an index function by: I = log w0 + C *[X<ai-<50)-jr] + (el-e0) (3) where Jt is a "time-equivalent" measure of the costs of migration (i.e., Jt = C/w0). The key behavioral assumption is that individuals compare earnings streams in the various countries and choose to reside in the country with the largest earnings, net of migration costs. The definition of the index function in (3) implies that individuals migrate to the host country when the index variable I is positive, and individuals remain in the source country when I is negative or zero. The second equality in (3) follows by assuming a linear approximation to the index function.
9 464 INTERNATIONAL MIGRATION REVIEW It is easier to illustrate the key results of the model when it is assumed that the "time-equivalent" measure of mobility costs are constant across individuals in the source country. This assumption does not say dollar mobility costs are constant (since it is the ratio of C to w0 that is assumed constant). Instead, it is basically assuming that individuals with higher earnings capacities find it more expensive to migrate. This correlation may be caused by the higher foregone earnings associated with unemployment spells in the host country while the individual is searching for job, higher costs of moving household goods, etc. The relaxation of this assumption does not change the fundamental results of the analysis in any way. Three important questions are easily addressed within this framework. First, what factors determine the size of the migration flow generated by the income-maximization hypothesis? Second, what types of selection in the observed characteristics X are created by the endogenous migration decision? In other words, are the individuals who migrate characterized by high or low levels of X? Third, what types of selection in the unobserved characteristics e are created by the endogenous migration decision? In other words, are immigrants characterized by high levels of ability (high e) or by low levels of ability (low e)? The Size of the Migration Flow Individuals migrate when the index variable I is positive. The probability that an individual (with characteristics X) born in the source country migrates to the host country is given by: P = Pr[v > -X((5L - <50) + Ji] = 1 - O(z) (4) where v = el - e0, z = [-X^ - d0) + ji] I ay-, and <J> is the cumulative distribution function for a normal random variable. Inspection of equation (4) reveals several properties of the emigration rate in an economy popu? lated by income-maximizing individuals:? The emigration rate (for persons with characteristics X) is higher the greater the mean income in the host country? The emigration rate (for persons with characteristics X) greater the mean income in the source country? The emigration rate (for persons with characteristics X) greater the level of migration costs is lower the is lower the? The emigration rate (for persons with characteristics X) is higher the greater the payoff to the observed demographic variables X in the host country relative to the payoff in the source country.
10 ECONOMIC THEORY AND INTERNATIONAL MIGRATION 465 The first three of these predictions are well known and have been widely tested in both the immigration literature and in the literature that analyzes internal migration patterns in the United States (See, for example, Jasso and Rosenzweig, 1985; and the survey by Greenwood, 1975). These predictions date back to the analysis of Hicks (1939) and form the theoretical core of the human capital approach to the economics of migration (Sjaastad, 1962). The fourth prediction, although it has not been stressed in the literature, also follows from the model: Individuals with a certain level of schooling are more likely to migrate the higher the rate of return to schooling in the host country relative to the rate of return to schooling in the source country. There is a sense in which these theoretical predictions are unsatisfactory. They simply are too obvious. The insight that, given the assumption that individuals are income-maximizers, persons migrate away from low income areas to high income areas when mobility costs are low is tautological. Although historically these have been the insights stressed by the literature, the income-maximization model summarized by equations (1 )-(3) has much more to say about the types of sorting generated by the immigration market. Selection in Observed Characteristics For simplicity, suppose that the observed vector of demographic variables, X, is composed of only a single factor? education. Suppose also that education is normally distributed in the population of the source country, so that the educational attainment of the population can be described by the equation: X=jux + ex (5) where jux is the mean of education in the source country and ex is a normal random variable describing the heterogeneity of educational attainment in the source country. It is assumed that ex has mean zero and variance a ^. It is worth stressing that the assumptions underlying the analysis in this section? that X only has a single characteristic and that X is normally distributed?are made for purely pedagogical reasons. The analysis can be generalized to any number of demographic variables. Moreover, the assumption of normality only simplifies the mathematics, without changing the nature of the results. Individuals migrate to the host country when the index variable I in equation (3) is greater than zero. It is, therefore, possible to calculate the conditional mean of educational attainment among persons who find it
11 466 INTERNATIONAL MIGRATION REVIEW profitable to migrate, E(X I>0). Using the normality assumptions, it can be shown that: 3 where k is a positive number. E(X\I>0)=fix + k(dl-d0) (6) The conditional expectation in (6) is composed of two terms. The first is the mean of educational attainment in the population in the source country, jux, while the second is the "selectivity bias" generated by the fact that persons who migrate are not randomly selected from the population. Equation (6) reveals that the mean educational attainment of migrants will differ from the mean in the source country as long as (dl - d0) ^0. The analysis thus shows that the mean schooling level of migrants will be greater or less than the mean schooling level in the country of origin depending on which of the two countries (i.e., the source or the host country) values schooling more. Define the migrant flow to be positively selected in terms of educational attainment if migrants are more likely to be composed of highly educated individuals. Equation (6) shows that this type of positive selection will occur when - (dl d0) > 0, so that the labor market in the host country attaches a higher value to schooling than the labor market in the source country. Define the migrant flow to be negatively selected in terms of educational attainment if migrants are more likely to be composed of individuals with little schooling. Equation (6) shows that negative selection is generated when the host country pays a lower "price" for educational skills than the source country (i.e., (dl - d0) < 0). Put succintly, educated persons migrate to the country that values educated labor the most. Therefore, international labor flows are no different from the flows of goods implied by international trade theory. Workers, like goods, flow to the country that is willing to pay the most for them. Selection in Unobserved Characteristics As equations (1) and (2) show, individual earnings can be decomposed in terms of the returns to observable skills (X<5) and the "returns" to unobservable ability (e). We have seen that selection on the basis of the observable demographic variables X is guided solely by differences in the coefficient vectors d0 and d?. This result, however, says nothing about which types of individuals in terms of the unobserved characteristics e are the most likely to migrate. 3 Heckman (1979, 1987) presents a detailed discussion of the mathematical and economic structures of models of self-selection.
12 ECONOMIC THEORY AND INTERNATIONAL MIGRATION 467 There is one reason why selectivity in unobserved ability is likely to be very important. It is well known that the explanatory power of earnings functions like (1) and (2) is relatively small, regardless of the specification of the vector X. It is not uncommon, for example, for earnings function of this type to have an R2 of less than 20 or 30 percent. In other words, an exhaustive set of variables X "explains" less than a third of the variance in earnings among individuals. Unobserved differences among individuals either in 'skills/ 'ability' or Tuck' dominate earnings distribution data. Consider the conditional expectations E(e0 X, I>0) and E(e1 X, I>0). The first expectation gives the average "residual" in the source country for persons who decided to emigrate from that country. The second expectation gives the average residual of the immigrants in the country of destination. Note that these conditional means "hold constant" the vector of charac? teristics X, so that they are asking the following question: given X, what kinds of persons migrate? The normality assumptions imply that these conditional expectations are given by: Qo = E<eo X,I>0) Of) <7l Of) = A -^(/o--?) (7) ov?r Q1 = E<ei X,I>0) = On <7i <7i -^(^-/>) A (8) where A = <p(z) / (1- <X>(z)) and <p is the density function of the standard normal. The variables Q0 and Q1? therefore, measure the labor market skills of the migrant flow across the two countries in terms of the unobserved characteristics of the immigrant pool. Note that if persons who emigrated from the source country are of "average" unmeasured skills or ability, one would expect that Qo=0. In addition, if these immigrants have an ability level equal to that of the native population in the host country, one would expect that Qi=0. Non-zero values of Q0 and Q1? therefore, indicate the extent to which the self- selection of the immigrant pool leads to a foreign born population that is not of average unmeasured skills. Inspection of the definitions of Qq and Qj reveals that three cases are possible. Positive Selection: Qjq>0 and Qj>0 By definition, positive selection exists when immigrants have above average abilities in the source and host countries. In other words, the earnings of persons emigrating to the source country are above average (among in? dividuals with characteristics X) and the earnings of these immigrants in the host country are also above average (compared to the earnings of natives
13 468 INTERNATIONAL MIGRATION REVIEW with characteristics X). Inspection of equations (7) and (8) reveals that a necessary and sufficient condition for this selection to occur are that: p >p and &i>o0 (9) (a0 a A wherep is a positive constant (the parameter/? is defined by min?,? ). If the correlation coefficient in the earnings across the two countries is sufficiently high and if income is more dispersed in the host country than in the country of origin, immigrants arriving in the host country will be selected from the upper tail of the source country's income distribution, and will outperform demographically similar natives upon arrival to the host country. Intuitively, this occurs because the source country, in a sense, is "taxing" high ability workers and "insuring" low ability workers against poor labor market outcomes. These taxes and subsidies are, of course, reflected in the fact that the host country's income distribution has more inequality than the source country's income distribution. Since high income workers benefit relatively more than low income workers from migration to the host country, a "brain drain" is generated. The host country, with its greater degree of inequality in earnings opportunities, becomes a magnet for persons who are likely to do well in the labor market.4 Negative Selection: Qq<0 and Qj<0 This type of selection arises when the host country draws persons who have below average incomes in the source country and who, holding charac? teristics X constant, perform poorly in the host country's labor market. The necessary and sufficient conditions for negative selection to occur are: p >p and ol <a0 (10) Negative selection requires that the correlation in earnings across the two countries /? be sufficiently positive, but that the source country exhibits more income inequality than the host country. Intuitively, negative selection arises when the host country "taxes" high-income workers relatively more than the source country and provides better insurance for low income workers against poor labor market outcomes. This opportunity set leads to large incentives for low ability persons to migrate since they can improve It is of interest to note that the means of the income distributions or the level of mobility costs do not enter the condition in equation (9). Mean income levels, therefore, do not affect the type of selection that is generated. Note, however, that the variables Qq and Q^ depend on A. Since A is an inverse function of the probability of migration, mean income levels do affect the "intensity" of the selection.
14 ECONOMIC THEORY AND INTERNATIONAL MIGRATION 469 their situation in the host country. Conversely, high ability persons have lower incentives to migrate since income opportunities in the home country are relatively good. Refugee Sorting: Qjj<0 and Qj>0 This kind of selection occurs when emigrants have below average ability (in the source country), but immigrants perform quite well in the host country's labor market. The necessary and sufficient condition for this type of selec? tion to occur is: P<P (11) If the correlation coefficient in earnings across the two countries p is negative or "small," the composition of the migrant pool is likely to resemble a refugee population. For instance, it is likely thatp is negative for countries that have recently experienced a Communist takeover. The change from a market economy to a Communist system is often accompanied by structural changes in the income distribution and by confiscation of entrepreneurial assets and redistribution to other persons. In essence, the income distribu? tion in the source country becomes a mirror image of its pre-revolution income distribution? persons who did well prior to the political upheavals see their assets vanish and given to persons who were not able to perform well in a market economy. The model thus predicts that immigrants from such systems will be in the lower tail of the "revolutionary" income distribu? tion, but will outperform the average worker in the host country since the immigrant has characteristics that are well suited for the market economy in the host country. This simple model, therefore, provides a useful categorization of the factors that determine the ability distribution of the migrant pool. Several important implications are generated which give some insight into a number of empirical findings in the literature. For example, many studies have documented the fact that refugee populations perform quite well in the U.S. labor market when compared to native workers of similar demographic characteristics. These empirical results can be understood in terms of the income-maximization hypothesis: prior to the political changes which led to a worsening of their economic status, refugees were relatively well off in their country of origin. Changes in political regimes devalue their skills, but ability is once again valuable when they migrate to a market economy. There is, therefore, no reason to resort to the arbitrary distinctions between "economic" and "non- economic" migration to explain the refugee experience. The theoretical analysis also provides an interesting explanation for the empirical finding that the "quality" of migrants to the United States has
15 470 INTERNATIONAL MIGRATION REVIEW declined in the postwar period (where quality is defined as the wage differential between migrants and natives with the same demographic characteristics). Prior to the 1965 Amendments to the Immigration and Nationality Act, immigration to the United States was regulated by numeri? cal quotas. These quotas were based on the ethnic population of the United States in 1920 and thus encouraged immigration from Western European countries and restricted migration from other continents, particularly Asia. The favored countries in the pre-1965 period have one important charac? teristic: their income distributions are probably much less dispersed than those of countries in Latin America or Asia. The 1965 Amendments abolished the restrictions on immigration from non-european countries, established a 20,000 numerical limit for legal migrants from any single country and led to a substantial increase in the number of immigrants originating in Asia and Latin America. The new flow of migrants thus originates in countries that are much more likely to have greater income inequality than the United States. Therefore, it would not be surprising to find that the standardized earnings of immigrants declined as a result of the 1965 Amendments. In addition, the 1965 Amendments led to a fundamental shift in the mechanism by which visas were allocated among potential migrants: the role played by observable skills and occupational characteristics was deem- phasized and the overwhelming number of visas began to be allocated according to the types of kinship relationships existing between potential migrants and persons currently residing in the United States. The economic model of immigration also suggests that this change in the statutes will lead to a substantial decline in immigrant quality. In particular, the family of the migrant that resides in the United States provides a "safety net" that insures the immigrant against poor labor market outcomes and unemployment periods in the months after migration. Low skilled persons who could not migrate without family connections to the United States, and hence without that insurance, will now find it worthwhile to do so. In effect, the kinship regulations in the immigration law create a lower bound in the income levels that low skilled immigrants can attain in the United States, and hence make it more likely that immigrants are negatively selected from the population. It is remarkable that the conditions determining selection in terms of unobserved ability have nothing whatsoever to do with the conditions determining selection in terms of observed demographic characteristics. Selection on the basis of unobserved ability depends entirely on the extent of income inequality in the host and source countries, and in the correlation between earnings in the two countries. Selection on the basis of observed demographic characteristics, on the other hand, depends on parameters measuring the price attached by each country's labor market to that par-
16 ECONOMIC THEORY AND INTERNATIONAL MIGRATION 471 ticular skill. Any permutation of selection mechanisms in unobserved and observed characteristics is, therefore, theoretically possible. That is, nega? tive selection in unmeasured skill (or ability) may be occurring simultaneously with positive selection in education, or vice versa. Thus, the immigrant flow maybe composed of relatively highly educated persons, but these highly educated persons do not do well in the host country and did not do well in the source country (compared to other highly educated persons) prior to their migration. An important insight, therefore, is that the empiri? cal observation that the migrant flow to any given host country is composed of highly educated individuals does not imply that these highly educated persons are the most productive highly educated origin. persons in the country of This implication of the economic theory of immigration reveals that little can be learned from comparisons of unstandardized average earnings between migrants and natives in any host country. These comparisons incorporate the differences in both observed and unobserved characteristics and confound the two types of selections that characterize the immigrant flow. Simply because the average migrant earns more than the average native does not imply a positive selection of the immigrant population. This observation is consistent with a positive selection in observed characteristics (such as education) and a (weak) negative selection in ability. Similarly, the observation that immigrants have lower earnings than natives does not by itself imply that the migrant population is negatively selected. This empiri? cal observation is consistent with negative selection in the observed characteristics (the migrants may have little education), but weak positive selection in the unobserved characteristics. The analysis, therefore, provides an important theoretical reason for focusing on the study of standardized earnings comparisons between immigrants and natives. These standardized comparisons, which hold constant differences in education, age, etc. be? tween natives and immigrants, reveal the types of selections in unobserved characteristics and can be interpreted within the framework of the economic theory of immigration. This simple model of the immigration decision, therefore, indicates that the nonrandom sorting generated by the immigration market can be predicted in terms of a small number of primitive parameters. There is no theoretical reason for presuming, as many do, that immigrant flows are always composed of "the best and the brightest." Under some conditions, it is likely that the opposite will happen. There is also no theoretical justifica? tion for assuming that the composition of immigrant flows will be constant over time or across countries. As economic and political conditions change, economic theory predicts that the size, direction and composition of im? migrant flows will also change. In other words, there is no universal "law"
17 472 INTERNATIONAL MIGRATION REVIEW that must characterize all immigrant flows. The model opens up a large number of new research possibilities. example, it is no longer necessary to provide ad hoc explanations of why the U.S. earnings of immigrants from different countries tend to exhibit so much variation. The economic theory of immigration suggests that this variance can be "explained" in terms of the economic and political condi? tions that guided the nonrandom sorting of persons across countries at the time of migration. In fact, recent empirical work by Borjas (1987b, 1988) suggests that some of the model's implications are supported by the empiri? cal evidence in the United States. For THE ADAPTATION OF IMMIGRANTS Beginning with Chiswick (1978), the question of how immigrants "adapt" to labor market conditions in the host country has received intensive study. The economic model justifying much of this research is a simple application of human capital theory (Becker, 1975). When immigrants arrive in the host country, they lack many of the skills that are valuable in the host country (e.g., language, knowledge about the location of jobs, etc.). Hence, the earnings of foreign born persons immediately upon arrival are likely to be lower than the earnings of comparable natives. Over time, however, since immigrants have relatively lower earnings, they also have relatively higher incentives to invest in human capital than natives (because the costs of human capital investments are primarily composed of foregone earnings). This investment process includes the accumulation of language skills and the internal migration of foreign born persons within the host country. Immigrant earnings can be expected to rise relatively fast (compared to natives) as the returns to human capital investments are realized. The key prediction of human capital theory, therefore, is that the age/earnings profile of immigrants will be steeper than the age/earnings profile of natives. This "catching-up" of earnings profiles reflects the "assimilation" or adaptation of immigrants to the host country's labor market. Human capital theory has also been used to explain the difference in investment behavior between so-called "economic" and "noneconomic" immigrants (Borjas, 1982). In particular, the returns to human capital investments depend on the immigrant's perception of how permanent his move to the host country is likely to be. Political refugees are often characterized by the fact that there is little chance that political conditions in the source country will change sufficiently to allow for their return. Hence, political refugees have significantly lower probabilities of return migration than economic immigrants. The incentives to invest in human capital are affected by the length of the time horizon over which the returns to these investments are collected.
18 ECONOMIC THEORY AND INTERNATIONAL MIGRATION 473 For instance, it has been shown that women's incentives to invest in human capital are adversely affected by the fact that they spend a smaller fraction of their adult life in the labor market than do men (Mincer and Polachek, 1974). This result implies that refugees, since they are basically "stuck" in the host country, have much greater incentives to adapt to the host country than other immigrants who have higher probabilities of return migration. In other words, immigrant flows characterized by high return migration propensities will "assimilate" faster than immigrant flows characterized by low return migration propensities. Human capital theory, therefore, generates important predictions about the process through which immigrants adapt to the host country's labor market. Unfortunately, little systematic modeling of this process has been undertaken. For example, it would seem that the types of investments chosen by immigrants (who usually arrive in the host country as young adults) would not be the same as the types of investments chosen by natives. Similarly, most of the empirical research ignores the direct implication of human capital theory that measures of the human capital stock (e.g., school? ing, internal migration, job turnover, on-the-job training) will have different lifetime paths for natives and immigrants. Instead, the literature has focused almost exclusively on measuring the extent to which the age/earnings profile of immigrants differs from the age/earnings profile of comparable natives (i.e., natives with the same demographic characteristics). This econometric literature is based on the cross-section regression model: logwj = Xjy+aIj+y3ljyj (12) where w- is the wage of person j; X- is a vector of socioeconomic charac? teristics (e.g., education, age, region of residence, etc.); L is a dummy variable indicating if person j is an immigrant; and y. is the number of years that person j has resided in the host country.5 Since the individual's age is being held constant, the coefficient /? traces out the impact of residence in the host country on the relative performance of immigrants in the labor market. The coefficient /?, by definition, provides a measure of the extent to which immigrants adapt or assimilate to the host country's labor market. It is worth repeating that this definition of assimilation is quite narrow since it focuses solely on the extent to which the slope of age/earnings profiles of immigrants differs from the slope of the age/earnings profile of natives. The model in equation (12) has been estimated using a large number of 5 In order to simplify the discussion, equation (12) does not include higher-order terms in the years-since-migration variable. The results summarized in this section are not affected by the inclusion of these additional variables.
19 474 INTERNATIONAL MIGRATION REVIEW alternative data sets in several host countries (See, for example, Chiswick, 1978; Carliner, 1980; DeFreitas, 1980; Chiswick and Miller, 1985; Bloom and Gunderson, 1987). Much of the conventional wisdom that immigrants do quite well, and are positively selected, arises from the empirical evidence obtained from the cross-section estimates of equation (12). Recent econometric studies stress the fact that such inferences may be completely erroneous (Borjas, 1985). The key insight is that cross-section estimates of (12) provide biased measures of the adaptation process ex? perienced by immigrants in the host country. The bias is due to the well known problem that a single cross-section regression cannot differentiate between aging and cohort effects. In this context, aging effects capture the change in earnings experienced by a particular immigrant (or a particular immigrant cohort) as he ages in the host country; while cohort effects capture productivity (or ability) differences across different immigrant cohorts.6 In other words, the positive correlation between earnings and years-since-migration documented by cross-section regressions may arise either because immigrants do experience higher earnings growth than comparable natives, or because more recent immigrant cohorts have lower productivities (or are more likely to be negatively selected) than immigrants from earlier waves. The importance of this insight is illustrated in Figure 1, which presents age/earnings profiles for three separate immigrant cohorts. The earliest cohort arrives, say, in 1960 and is assumed to have the highest productivity level? at? any age in the foreign born population. The most recent cohort arrives in 1980 and is assumed to have the lowest productivity level in the immigrant population. Finally, the middle cohort arrives in 1970 and is assumed to have exactly the same level of unobserved skills as natives and, therefore, the age/earnings profiles of these two groups exactly coincide over the working life cycle. As drawn in Figure 1, the age/earnings profiles of all the immigrant cohorts are parallel to the age/earnings profile of the native population. This implies that immigrants do not experience any assimilation (in the sense that their earnings do not "catch up" to the earnings of natives). The true impact of years-since-migration on earnings, after controlling for age, is zero. Suppose a cross-section of data is collected in Consider the data available to a researcher analyzing this single data set. A cross-section allows the observer to identify only one point in each of the immigrant age/earn? ings profiles. In particular, a point at the age of entry (say 20) is observed The fact that a single cross-section of data cannot separately identify aging and cohort effects is well known in the demography literature but was ignored by the early studies of immigrant earnings. A recent statement of the technical problems is given by Heckman and Robb (1983).
20 ECONOMIC THEORY AND INTERNATIONAL MIGRATION 475 FIGURE 1 EARNINGS 1960 Cohort CO CD 1970 Cohort and natives or < 1980 Cohort for the cohort that most recently arrived (i.e., the 1980 cohort). The earnings at age 30 are observed for the immigrants that arrived in 1970, while the earnings at age 40 are observed for the 1960 arrivals. The cross-section earnings profile thus generated is given by CC. This cross-section regression line has two important characteristics. First, it is substantially steeper than the native earnings profile. Hence a cross-section regression will make it seem as if there is labor market assimilation when, by assumption, there is none (i.e., a cross-section regression leads to a non-zero estimate of /?). Second, the cross-section regression line crosses the native age/earnings profile at age 30. This gives the appearance that immigrants "overtake" the earnings of the native population when, in fact, no such event is ever experienced by any of the immigrant cohorts.