Review of International Economics, 11(2), 253–267, 2003 Attitudes Towards Immigration: a Trade-Theoretic Approach Sanoussi Bilal, Jean-Marie Grether, and Jaime de Melo* Abstract The paper uses a three-factor (capital, low- and high-skill labor), two-household (low- and high-skill individuals), two-sector trade model to analyze the determinants of voter attitudes towards immigration under direct democracy, and to identify factors that would be coherent with both the observed increase in the skilled–unskilled wage differential and the stiffening attitudes towards low-skill capital-poor immigration. If the import-competing sector is intensive in the use of low-skill labor, and capital is the middle factor, an improvement in the terms of trade or neutral technical progress in the exporting sector leads nationals to oppose immigration of capital-poor low-skill households. An increase in income inequality is also likely to stiffen attitudes towards this type of capital-poor, low-skill immigration prevalent in Europe until recently. 1. Introduction Incentives for international migration have largely been studied as responses to factorreward differences with factors of production moving internationally to maximize income. The standard Heckscher–Ohlin model predicts that, on efficiency grounds, countries should be as open to the indirect inflow of factor services embodied in goods as to the direct flows of capital and labor. Yet, globalization of the world economy has revealed an asymmetry in policies as countries have become concurrently more open to flows of goods and capital and less open to direct flows of labor. Moreover, for countries receiving immigrants, restrictions have become more severe towards unskilled labor. At the same time there has been a widening disparity between the incomes of skilled and unskilled workers. This is in accordance with the Stolper–Samuelson predictions of increased globalization, suggesting that there is still some usefulness for the standard trade-theoretic framework when studying international factor migration, and that there may be causal links between changes in the external environment, the deterioration in the labor markets of many recipient countries, and the stiffening of attitudes towards immigration. To be sure, to understand attitudes towards the movement of people, one must go beyond the standard economic framework where factors of production are apersonal entities. Broadening the framework to take into account the attributes of individuals as in models of locational choice (Hillman, 1994) or cultural preferences (Schiff, 1997) is indispensable to understand attitudes towards immigration. But it is hard to dismiss * Bilal: International Economic Development Group, Overseas Development Institute, 111 Westminster Bridge Road, London SE1 7JD, UK. E-mail: s.bilal@odi.org.uk. Grether: University of Neuchâtel, Division Economique et Sociale, Pierre-à-Mazel 7, 2000 Neuchâtel, Switzerland. E-mail: jean-marie.grether@unine.ch. de Melo: University of Geneva, Département d’économie politique, 40 Boulevard du Pont-d’Arve, 1211 Genève 4, Switzerland. E-mail: jaime.demelo@ecopo.unige.ch. This is a revised version of a paper presented at the CEPR workshop “Globalization, Regional Integration and Development,” Venice, 31 January/1 February 1998 and at a seminar at the University of Geneva. We thank Riccardo Faini, Ronald Jones, Tobias Müller, Arvind Panagariya, Maurice Schiff, an Associate Editor, a referee, and seminar participants for comments and the FNRS for financial support under grant 12-42011.94. © Blackwell Publishing Ltd 2003, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA 254 Sanoussi Bilal, Jean-Marie Grether, and Jaime de Melo the view that, like others, immigration policies are usefully analyzed in a politicaleconomy setting in which they are endogenously explained by economic and political self-interest and the institutional mechanisms of collective choice.1 We study attitudes towards immigration in a model where attitudes are determined entirely by economic self-interest in a direct-democracy framework à la Mayer (1984), the simplest institutional mechanism of collective choice. We use a three-factor version of the two-good Heckscher–Ohlin model which allows us to consider two types of households (high-skill and low-skill) each owning varying degrees of capital.2 We can thus study the effects of changes in the distribution of income on household attitudes towards immigration. Section 2 presents the building blocks of the model: household factor ownership patterns and factor intensity conditions. Section 3 develops a simple graphical analysis that allows us to determine household attitudes towards immigration in terms of two parameters: a “similarity” index measuring differences in factor intensities and the household composition in the population. Section 4 carries out standard comparative statics exercises that establish how factor accumulation, terms-of-trade changes, and disembodied technical progress affect households’ attitudes towards immigration. Section 5 interprets model-predicted results about changes in attitudes towards immigration in terms of broad stylized facts on the evolution of product and factor markets in receiving countries. For example, we introduce skewness in capital ownership within each household group and show how an increase in income inequality can lead to a stiffening in attitude towards capital-poor immigrants. Section 6 concludes. 2. The Model As did the early users of the “3 ¥ 2” model (Batra and Casas, 1976; Jones and Easton, 1983; Thompson, 1983), we assume a price-taking economy in international markets, and we take a long-run view by assuming that all factors are mobile across sectors. Two goods are produced, X1 (imported) and X2 (exported), with constant-returns-to-scale production functions using three factors, low-skill labor, high-skill labor, and capital (with endowments of L, H, and K, respectively).A one-household version of this model has been used by Davies and Wooton (1992) to study the effects of immigration on wages and the functional distribution of income. We discuss first assumptions about households, then turn to technology and determinants of factor prices. Households There are two types of households (v), each owning either one unit of low-skill labor (v = l) or one unit of high-skill labor (v = h) and a positive amount of capital (Kv). Households have identical and homothetic preferences. Household income depends on factor prices wK, wL, and wH. Thus, the incomes of low- and high-skill households are yl = wL + wKKl and yh = wH + wKKh, respectively. It is assumed that immigrants spend their income in the receiving country and only a fraction of them (initially set to one) are assimilated and vote. Suppose a direct democracy in which people vote on immigration policy. In the absence of voting costs, a voter will favor entry of new immigrants, which we consider as permanent, if it increases her utility. Define the indirect utility of voter v as Uv = Uv(p1, p2, yv), where pj is the domestic price of good j (j = 1, 2). Since goods’ prices are fixed, even in the general case where immigrants dM (dM = dL or dH) bring in capital © Blackwell Publishing Ltd 2003 ATTITUDES TOWARDS IMMIGRATION 255 dK, we have ∂p/∂M = ∂p/∂K = 0. Thus, the condition for household v to favor immigration (dUv ≥ 0) reduces to dyv = ∂ yv ∂ yv ÈÊ ∂ wv ∂ wv dK ˆ Ê ∂ wK ∂ wK dK ˆ ˘ dM + dK = dM Í + + + Kv ≥ 0. (1) ∂M ∂K ∂ K dM ¯ ˙˚ ÎË ∂ M ∂ K dM ¯ Ë ∂ M In expression (1), the first term in parentheses on the right-hand side gives the effect of immigration on labor income, the second the effect of immigration on capital income. It shows that a voter’s attitude towards immigration depends on her skills (v = h, l), on her capital ownership (Kv), on which type of immigrants come in (dM = dL, dH), on how they are endowed in capital (dK/dM), and on how factor prices react to changes in factor endowments. We now turn to this latter determinant, leaving the systematic discussion of the former factors to section 3. Factor Endowments and Factor Prices To complete the model, we must specify factor-intensity assumptions. Presumably, countries that receive immigrants are net importers of low-skill-labor intensive products. Whether they are exporters of capital or rather of skilled-labor intensive products is more a matter of debate. To simplify presentation, we assume here that they export skilled-labor intensive products, which leaves capital as the “middle” factor.3 If aij is the amount of factor i used in a unit of good j (with i = L, H, K and j = 1, 2), this means that the following factor extremity conditions are satisfied: aL1 aL 2 > aK 1 aK 2 > aH 1 aH 2 . (2) We know from Ruffin (1981) that in the 3 ¥ 2 model the two extreme factors are “enemies” (i.e., an increase in the endowment of one of the extreme factors reduces the reward of the other, while increasing the reward of the middle factor), whereas the middle factor is everybody’s “friend” (i.e., an increase in the endowment of the middle factor increases the wage of the extreme factors). Applied to our case, as both types of labor are enemies, this means that the middle factor, capital, whether it is owned by national households or brought in by immigrants, plays a (critical) compensatory role. This is straightforward from expression (1) on the voter’s side: if immigrants bring in no capital (dK/dM = 0), the increase in capital reward (dwk /dM > 0) mitigates and may overturn the decrease in wages (dwv /dM < 0). But it is also true on the immigrants’ side: as soon as they bring in capital, their impact on each source of households’ income becomes ambiguous (as dwi /dM and dwi /dK have opposite signs "i = K, L, H). To sort out these ambiguities, we need to extend Ruffin’s analysis to allow for generalized endowment changes. Total differentiation of the usual zero-profit (Si aijwi = pj) and full-employment (Sj aij xj = Vi) conditions leads to a five-equation system with unknowns wL, wK, wH, X1, and X2.As in Jones and Easton (1983, eq. 20), we solve implicitly for output changes from the extreme factor markets, which leads to the following reduced form for factor prices variations (using a circumflex to denote a relative change): q L1wˆ L + q K 1wˆ K + q H 1wˆ H = pˆ 1 , (3) q L 2wˆ L + q K 2wˆ K + q H 2wˆ H = pˆ 2 , xL wˆ L + xK wˆ K + x H wˆ H = Vˆ , (4) (5) © Blackwell Publishing Ltd 2003 256 Sanoussi Bilal, Jean-Marie Grether, and Jaime de Melo where qij is the share of factor i in sector j’s costs (qij ∫ wiaij/pj), xi is the (generalequilibrium) economy-wide elasticity4 of the use of the middle factor with respect to wi, and V̂ is given by the following expression: Vˆ = Kˆ - (a L Lˆ + a H Hˆ ), (6) a L ∫ (l K - l H ) (l L - l H ) , a K ∫ (l L - l K ) (l L - l H ) = 1 - a L , (7) where and li is the share of factor i used by sector 1. From (2), it is straightforward to show that lL > lK > lH, which means that aL (or aH) can be interpreted as the index of similarity between factor L (or H) and factor K. The value taken by this similarity index depends on factor shares. To fix likely orders of magnitude, suppose that capital is used with the same intensity by both sectors (qK1 = qK2). Then it can be shown5 that the similarity index becomes a L = q L (q L + q H ) , a H = q H (q L + q H ) , (7¢) where q i is the share of factor i in national income (e.g., qL ∫ wLL/Y, Y ∫ wLL + wHH + wkK). Thus, in this special case (used in the comparative statics of section 4), the similarity index is equal to the share of the particular labor skill in total income accruing to labor. The similarity index aL (or aH) also represents the sensitivity of the demand for the middle factor following a change in the supply of factor L (or H) when factor and product prices are kept constant and output changes are such that extreme factor markets are cleared. Thus, in (5), V̂ can be interpreted as the difference between the growth in the capital stock and the increase in its demand induced by the growth of labor endowments (at constant prices and accommodating changes in output). If this difference is positive, factor growth will lead to an excess supply of capital, leading to a decrease in the middle factor’s reward and an increase in the reward of extreme factors (this according to the friend–enemy relationship and to the left-hand side of equation (5), which represents the necessary adjustment of capital demand through factor rewards’ changes). The reverse applies when V̂ < 0, while V̂ = 0 defines the set of factor growth rates that have no effect on factor prices. Setting product prices constant, equations (3)–(5) lead to the following relationships between changes in factor prices and changes in factor endowments: wˆ K = -(1 D)Vˆ , wˆ L = a L (1 D)(q K q L )Vˆ , wˆ H = a H (1 D)(q K q H )Vˆ , (8) where D is a positive constant reflecting factor substitutability.6 Note that the change in each extreme factor reward (in the reward of the middle factor) is a positive (negative) function of V̂ , a result which is consistent with the interpretation of V̂ as the change in the net supply of the middle factor. 3. Attitudes Towards Immigration How do these changes in factor rewards translate into changes in attitudes towards immigration? As noted in the previous section, that attitude depends on the voter’s skills, on capital endowment, and on the skills and capital endowment of immigrants. A useful property to start from is that, in this type of model, a marginal factor immigration leaves the income of residents unchanged.7 With capital evenly distributed within each household category, an assumption we make until section 5, this implies © Blackwell Publishing Ltd 2003 ATTITUDES TOWARDS IMMIGRATION 257 that, as in Benhabib (1996), high-skill and low-skill households will always adopt an opposite attitude towards immigration. In the following paragraphs we first show how the compensatory role of capital materializes in critical capital ownership levels for both residents and immigrants. Then we develop a graphical representation of how these capital-ownership levels and factor-intensity conditions interact to determine household attitudes following a marginal immigration. Critical Capital Ownership Levels for Residents and Immigrants Using (8), we can rewrite the change in each household income following immigration as8 K˘ K˘ È 1 ˘ È È 1 ˘ È dyl = Í- wK ˙ [Vˆ ] ÍKl - a L ˙, dyh = Í- wK ˙ [Vˆ ] ÍK h - a H ˙, L˚ H˚ Î D ˚ Î Î D ˚ Î (9) which shows that income changes result from the multiplication of three terms, the first one being negative, while the signs of the other two are ambiguous, depending on capital ownership levels. Consider first a Ruffin-type case, with an inflow of low-skill immigrants with no capital ownership. Then all wages will fall, as low-skill and high-skill labor are enemies. Does this imply that all households oppose immigration? Certainly not, because they own capital, and capital income increases. In fact we know, from the zero-sum property of aggregate income, that this overall increase in capital income will exactly match the decrease in labor income. However, apart from the very special case where capital is allocated over the two household categories in such a way as to keep their income unchanged, one group will lose, the other will gain. We call the former group “capitalpoor” households, who will oppose immigrants, the latter “capital-rich” households, who will welcome them. As high-skill and low-skill labor are enemies, the same logic applies to the case of high-skill immigrants. This is exactly what is conveyed by equation (9): as immigrants bring in no capital, V̂ is negative, which makes the product of the first two terms positive. Whether the sign of the whole expression remains positive or not depends on the sign of the last term between brackets: it is positive if the household is capital-rich, which will be so provided her capital ownership, Kv, is larger than a critical level. It turns out that lowskill (high-skill) households are capital-rich if their capital ownership is larger than aL(K/L)(aH(K/H)). However, with capital equally distributed within each household group, as K = KL + KH = LKl + HKh, it is easy to show that one condition excludes the other. If, say, low-skill households are capital-poor, high-skill households are capitalrich, so they will never share the same attitude towards immigrants (except in the very special case where effective capital ownership is equal to the critical level in both groups so that every household is indifferent towards immigration). Going beyond Ruffin-type immigration, what happens if immigrants also own capital? Again equation (9) tells us everything. As long as the capital brought by immigrants is below a critical level, V̂ remains negative, leading to the same attitudes as those just described. In this case immigrants are “capital-poor.” However, if the capital ownership of immigrants exceeds that critical level, they become “capital-rich” and the attitudes of residents are reversed because V̂ changes sign (which implies, by (8), that the whole pattern of factor rewards’ changes also reverses). It turns out that the critical capital ownership levels for immigrants are identical to those of the corresponding skill category of residents. © Blackwell Publishing Ltd 2003 258 Sanoussi Bilal, Jean-Marie Grether, and Jaime de Melo Table 1. Attitudes Towards Immigration and Critical Capital Ownership Levels Immigrants Low-skill households Capital-poora Capital-richb High-skill households capital-poor Kl < aLK/L opposed favorable favorable opposed capital-rich Kh > aHK/H capital-rich Kl > aLK/L favorable opposed opposed favorable capital-poor Kh < aHK/H Kv (KM v ) is the effective capital ownership of household (immigrant) v. a Either KlM < aLK/L or KhM < aHK/H. b Either KlM > aLK/L or KhM > aHK/H. The complete pattern of attitudes towards immigration, along with the different levels of critical capital ownership, are summarized in Table 1, with the position of high-skill residents in italics to distinguish them from low-skill ones. Note that attitudes are basically “capital-concerned”: capital-poor households oppose capital-poor immigrants and vice versa, while the skill level of residents or immigrants only matters for the level of critical capital ownership. Moreover, inspection of critical capital ownership values reveals the following alternative interpretation of the similarity index: Proposition 1. When capital ownership is evenly distributed within each group of households, aL (aH = 1 - aL) is the share of total capital that must accrue to low-skill (high-skill) households to leave them just indifferent to immigration. This proposition underlines the graphical analysis of the next section. A Graphical Representation of Attitudes Towards Immigration It is now natural to compare the critical shares of capital ownership just identified above with the effective shares of capital owned by each category of households. Let us define these effective shares by sL ∫ KL /K, sH ∫ KH /K and introduce two additional ratios: t ∫ (KL L) (K L + H ) , (10) 艎 ∫ L (L + H ) . (11) In the above expressions, t is an index of intergroup disparity of the distribution of capital (t = 1 when there is no interhousehold disparity in the distribution of capital and 0 when all capital ownership goes to the high-skill household group) while 艎 is the share of low-skill households in total population. Simple algebra shows that sL = t 艎, s H = 1 - t 艎. (12) Thus, attitudes towards immigration depend on the straightforward comparison between t艎 and aL: if t艎 = sL < aL, low-skill households are capital-poor, but as aL = 1 - aH and sL = 1 - sH, this implies that sH > aH, so that high-skill households are capitalrich, which confirms again the opposite opinion held by the two household categories. © Blackwell Publishing Ltd 2003 ATTITUDES TOWARDS IMMIGRATION 259 Figure 1. Determinants of National Attitude Towards Immigration The national attitude towards immigration can now be depicted graphically as in Figure 1, on the basis of t, 艎, and aL values. Indifference for all households is reached along the aL = t艎 line, which coincides with the diagonal when t = 1. For all points above (below) this line, low-skill households are capital-poor (capital-rich), the reverse being true for high-skill households. Thus, if immigrants are capital-poor, all points above the indifference line correspond to cases where low-skill households oppose immigration, while positions are reversed below the indifference line. As national attitude will match the attitude of the majority group, in the clear areas 1 and 3 (the shaded areas 2 and 4), the economy favors (opposes) capital-poor immigration. If immigrants were capital-rich, all positions would be reversed. Finally, an increase in intergroup capital distribution disparity (a lower t), which is represented by a clockwise rotation of the indifference line, would be associated with a widening of the range of (艎, aL) values where low-skill households are capital-poor, leading to new shaded areas (not indicated in the figure). In Figure 1, suppose that the share of low-skill labor is 艎0 < 0.5 and that the similarity index takes the value a 0L = 艎0. Suppose also that there is no intergroup disparity in capital ownership (t = 1) and that immigrants are capital-poor in the sense defined above. In this case, poor households are a minority, but since both groups are indifferent to immigration, this is a borderline situation. Suppose now that aL > a 0L. Then, high-skill households will favor immigration (area 1), and since they are the majority, this implies that the political process, as modeled here, would be favorable to immigration of capital-poor immigrants. Conversely, if aL < a 0L, high-skill households (who are the majority) would oppose immigration and this would reflect the national attitude towards immigration. Finally, suppose that aL = a 0L, but that there is intergroup disparity in capital ownership (t < 1). Then, high-skill households that own more capital would be in favor of immigration (the intersection of the two dashed lines would now be in a clear area) and the economy would be favorable to immigration. The assumption that all immigrants are assimilated and vote can easily be relaxed. Assume that only a proportion g of resident low-skill households vote (0 £ g £ 1). Then, remembering that 艎 ∫ L/(L + H), the condition for voting low-skill households to become the majority is 艎 > 艎min = (g + 1)-1(0.5 £ 艎min £ 1). In terms of Figure 1, this © Blackwell Publishing Ltd 2003 260 Sanoussi Bilal, Jean-Marie Grether, and Jaime de Melo means that the vertical line beyond which there is a reversal of attitude towards immigration is shifted to the right to 艎min, leading to new shaded areas which are more akin to the attitude of high-skill households. Finally, Figure 1 also illustrates how attitudes towards immigration would be handled in special cases. Suppose first that low-skill households own no capital (t = 0). Then, the attitude towards immigration will depend only on the share of low-skill households in the total population. Take next the 2 ¥ 2 model. If high-skill (low-skill) labor and capital are used with the same intensities in both sectors, aL = 0 (aL = 1), then households are indifferent to immigration as marginal changes in endowments have no impact on factor rewards. Finally take the specific factor model, assuming that capital is the mobile factor. Then expressions (7) simplify (lL = 1, lH = 0) and the similarity index aL (aH) collapses to the share of the mobile factor used in sector 1 (2). Results so far can be summarized as follows.9 Proposition 2. When capital ownership is evenly distributed within each group of households: (a) low-skill and high-skill households always have opposite attitudes towards immigration; (b) national attitude towards immigration is mainly “capital concerned,” as households’ attitudes are independent of the type of immigrants (but for the critical capital-ownership level of immigrants); (c) when only a fraction of the low-skill households vote, immigration has an effect on aggregate voters’ income and attitudes are more akin to those of high-skill households. 4. Comparative Statics How do changes in the economic environment affect attitudes towards immigration? We consider three exogenous changes. First there is growth through factor accumulation. In this case, critical-capital ownership levels for national households (see equation (9)) will be affected both directly, through factor endowment changes, and indirectly, through changes in the similarity index. This indirect effect is induced by variations in factor rewards (equation (8)), which derive from the comparative statics expressions (3) to (5). Second, we take up changes in relative prices. It is widely perceived that a reduction in tariffs and other barriers to trade (such as the cost of doing business) have been important components of globalization that could have affected attitudes towards immigration. Here we proxy the effects of globalization by an exogenous rise in the relative price of the exporting sector (X2).10 Third, there is the impact of technical progress. It is believed that technical progress has contributed, partially at least, to the increasing wage gap between workers of different skills. We deal with disembodied technical progress, either neutral or lowskill-labor saving. The relevant comparative statics expressions are equations (3) to (5) with p̂1= 0, p̂2 > 0, V̂ = 0 in the case of relative price changes and p̂1 = -v1, p̂2 = -v2, V̂ = -h in the case of technical progress (where vj = Siqijâijex, h = hk - (aLhL + aHhH), hi = li1âi1ex + li2âi2ex and âijex denotes the exogenous component of the rate of change of aij). From the above, it is clear that neutral technical progress in sector 2 would be equivalent to an increase in the price of that sector. Hence, after dealing with the © Blackwell Publishing Ltd 2003 ATTITUDES TOWARDS IMMIGRATION 261 effects of a reduction of protection, we only treat across-the-board low-skill laborsaving technical progress in the following section. To save space, we only summarize the results; the derivations can be found in the appendix to Bilal et al. (1998). Growth in Factor Endowments To get qualitative results, we make two simplifying assumptions. First, we consider cases where only one factor’s endowment changes at a time, namely a capital increase or an increase in the endowment of high-skill labor. Second, we neutralize for the direct effect of endowment changes. Thus, in the case of capital growth, we assume that it is spread evenly across households. In the case of labor growth, we assume that the initial capital endowment of the group is spread evenly across individuals. Then, at constant factor prices, the critical level of capital ownership would change proportionately to the individual ownership of capital, which neutralizes the direct effect of endowment changes. Variations in the attitude towards immigration are thus driven only by the indirect effect, through changes in the similarity index aL. It turns out that, in the general case, the sign of â L is ambiguous, depending on factor intensities and substitutability/complementarity between factors (reflected in the value taken by D in (8)). To resolve this ambiguity, we take the “symmetric” case mentioned above (i.e., qK1 = qK2) and consider a one-level CES functional form for technology. Then, simple conditions can be derived under which an increase in factor endowment leads to changes in the similarity index and hence to changes in the attitudes towards immigration. These added restrictions also allow us to sign the effects of the other shocks on the value of the similarity index. In the case of capital accumulation, aL will fall if the elasticity of substitution is higher in sector 2, a plausible condition as this sector is the exporting one. As aL falls, the critical-capital ownership level of high(low)-skill households increases (decreases), which means that this household category becomes more opposed (favorable) to immigration. To interpret this result, return to Figure 1 and suppose that initially t = 1, a = a 0L, and 艎 = 艎0, which leads to indifference towards immigration, high-skill households being the majority. The increase in capital will shift the economy’s point downwards, to shaded area 4, meaning opposition to (capital-poor) immigration, which reflects the new position of the majority group. The impact of an increase in high-skill labor endowment is apparently more complex. On the one hand, it implies a decrease in 艎 which, starting from the same initial indifference point, would lead to a favorable national attitude towards (capitalpoor) immigration. But on the other hand, it also leads to a change in aL, which could run against the previous effect. However, it turns out that, in the “realistic” case where the elasticity of substitution is higher in the exporting sector, although aL may eventually fall, it will never reverse the first effect. In sum, starting from indifference, an increase in high-skill labor endowment is likely to favor (capital-poor) immigration. Increase in the Relative Price of the Exporting Sector In the general case, an increase in the relative price of the exporting sector has an ambiguous impact on factor returns. However, if factors are substitutes, then as shown by Jones and Easton (1983), one obtains Stolper–Samuelson effects on the extreme factor returns, while the real return accruing to the middle factor depends on factor intensities and substitutability. In any case, to find out the effect on attitudes towards © Blackwell Publishing Ltd 2003 262 Sanoussi Bilal, Jean-Marie Grether, and Jaime de Melo immigration, we need again to identify the effect of the shock on the value of the similarity index aL. With CES production functions, whatever the values of the elasticities of substitution, aL unambiguously decreases following the increase in the relative price of the exporting sector (or a neutral technical progress in the same sector). Therefore, starting from the usual initial indifference point in Figure 1, this means that an increase in the relative price of the exporting sector (i.e., a reduction of protection) would generate an opposition to (capital-poor) immigration. Low-Skill Labor-Saving Technical Progress Finally, the effect of across-the-board labor-saving technical progress on factor prices is generally ambiguous. If factors are substitutes, although the change in the low-skill labor wage rate remains ambiguous, the wage rate of high-skill individuals falls while the return to capital increases. In the simple symmetric case of a CES production technology, if the elasticity of substitution is higher in the exporting sector, aL increases unambiguously. Thus, starting from the usual indifference point in Figure 1, across-theboard low-skill labor-saving technical progress leads the economy to favor (capitalpoor) immigration. Results of the comparative statics can be summarized as follows. Proposition 3. Suppose that: high-skill individuals are the majority; there is neither intergroup nor intragroup disparity in capital distribution; the capital factor share in total cost is the same in both sectors; technology is of the CES type with a higher elasticity of substitution in the exporting sector. Then, starting from an initial indifference, the national attitude towards (capital-poor) immigration will become: (a) favorable in the case of an increase in high-skill labor or a low-skill labor-saving technical progress in both sectors; (b) opposed in the case of an increase in the capital stock or an improvement in the terms of trade (which is equivalent to a neutral technical progress in the exporting sector). These results would be reversed if either low-skill individuals were the majority or immigrants were capital-rich. 5. Discussion Consider the following often-cited stylized facts pertinent for the evolution of the labor market in recipient countries: (i) a decrease in the relative price of import-competing activities as a result of the globalization that has reduced protection; (ii) an increase in the skilled–unskilled wage gap; (iii) an increase in income (and capital) inequality; and (iv) technical progress, perhaps of the labor-saving variety.11 According to the results in section 4, via Stolper–Samuelson effects, globalization is consistent with the observed increase in the wage gap. So is neutral technical progress in the exporting sector. However, across-the-board low-skill labor-saving technical progress would lead to a reduction of the wage gap along the lines shown by Findlay and Grubert (1959) in the 2 ¥ 2 case. In the symmetric case, endowment changes would leave the wage gap unchanged. In short, the dominant effect is the Stolper–Samuelson one. Now, if Stolper–Samuelson effects are dominant, as shown above, this would lead to a stiffening in attitudes towards capital-poor immigration. However, even in this © Blackwell Publishing Ltd 2003 ATTITUDES TOWARDS IMMIGRATION 263 simple symmetric case, once one allows for an increase in the share of high-skill households in the total population (which has no effect on the wage gap in this symmetric case), the attitude towards immigration is ambiguous since the effect of high-skill labor growth is favorable towards capital-poor immigration. Assume now that capital ownership is not necessarily evenly distributed within households but follows a beta distribution. The probability density function for the capital ownership of household v is f (Kv ) = G(b v + g v ) Ê Kv ˆ G(b v )G(g v ) Ë cv ¯ bv -1 Kv ˆ Ê 1Ë cv ¯ g v -1 Ê 1ˆ , Ë cv ¯ (13) where Kv belongs to the interval [0, cv], cv > 0, and G is the gamma function. For each household group, the upper-bound value cv is obtained from the mean: E(Kv) = cvbv/(bv + gv) = Kv /V, V = H, L. This distribution is symmetric (skewed to the left) if bv = gv(bv < gv). For simplicity, suppose that parameters describing the distribution of capital within each household group are the same for both groups (i.e., bl = bh, gl = gh).12 Start with the case where the distribution is symmetric (b = g ) and there is no intergroup disparity in the distribution of capital (t = 1). In this case, we get exactly the same results as in the case of no intragroup disparity of capital ownership. This is because the number of people in each household group having the attitude of the other group’s majority is the same (hence, Figure 2(a) is identical to Figure 1). Introduce now intergroup disparity in capital distribution (t < 1), maintaining the same symmetrical distribution for each household group (Figure 2(b)). Then, one is approximating the case where low-skill households have no capital and become systematically opposed to capital-poor immigrants.13 Introduce now skewness in the distribution of capital (with no intergroup disparity). Again suppose identical standard forms for each group (Figure 2(c)). Then, along the diagonal, although the mean voter would be indifferent to immigration, the median voter has lower capital in both groups, leading to a majority opposed to capital-poor immigration. As Figure 2(c) confirms, to reverse this unfavorable attitude towards immigration one must either alter the composition of the population or the value of the similarity index. Finally, Figure 2(d) combines intergroup disparity and withingroup skewness by combining the parameters used in Figures 2(b) and 2(c). Capital ownership distribution being highly skewed (see Wolff, 1996), Figures 2(c) or 2(d) are probably better guides of the roles of households composition and factor intensity differences in determining attitudes towards immigration. Also, taking a probabilistic approach to reflect ignorance about the economy’s parameter values, note that, starting from Figure 2(a), where the shaded area is 1/4, an increase in intergroup and/or intragroup disparity increases the importance of the shaded area, which approximately could reach 5/8, reflecting a higher probability of opposition to capital-poor immigration.14 After all, widening income disparities would appear to be coherent with the generalized move towards introducing capital requirements in immigration. 6. Conclusions This paper has developed a model in which preferences towards immigration are determined by natives’ perception of the impact of immigration on factor rewards. While noneconomic factors are likely to affect individual preferences (e.g., individual beliefs about civil rights and expectations regarding the cultural impact of immigrants), survey-based results indicate that individual preferences over immigration © Blackwell Publishing Ltd 2003 264 Sanoussi Bilal, Jean-Marie Grether, and Jaime de Melo Figure 2. Capital Ownership Distribution and Attitudes Towards Immigration policy in the United States are primarily determined by wage effects (see Scheve and Slaughter, 1999). These findings justify the approach taken here. In spite of the complexity of the three-factor two-household model, we were able to show that so long as household capital-ownership disparities are intergroup, then attitudes towards immigration are determined by two parameters: an index of factor-use similarity (whose values can be approximated by readily available data on factor shares in export- and import-competing sectors) and the household composition of the population. At the same time, the model does not allow factor flows to have a net impact on the aggregate income of residents, as it fails in particular to take into account the possibility that immigration brings with it a number of externalities due to social and/or cultural differences. Also, an important dimension of globalization is the decline of the jobs “sheltered” from international competition. This is probably affecting © Blackwell Publishing Ltd 2003 ATTITUDES TOWARDS IMMIGRATION 265 attitudes towards immigration, an effect that could be captured in this framework only if it were extended to include a nontraded sheltered sector. Nonetheless, the model provides a rich array of determinants of attitudes towards immigration. First, it highlights the importance of factor intensities in shaping attitudes towards immigration, in a model that introduces explicitly the role of capitalownership and of the skill-level of potential immigrants, which have been at the center of the recent stiffening of attitudes towards low-skill (and/or capital-poor) immigrants. In this framework, we show that the model’s predictions are coherent with the adverse impact on the attitudes towards immigration that might have occurred as a result of recent changes in OECD labor markets. For example, it shows that an increase in the capital stock and greater trade openness would both be consistent with a stiffening of attitudes towards capital-poor immigrants. Moreover, it suggests that the increasing income inequality (that may have been driven by the globalization process) could have reinforced the generalized move towards the introduction of capital requirements for immigrants. References Batra, R. and F. Casas, “A Synthesis of the Heckscher–Ohlin and the Neoclassical Models of International Trade,” Journal of International Economics 6 (1976):21–38. Benhabib, J., “On the Political Economy of Immigration,” European Economic Review 40 (1996):1737–43. Bilal, S., J.-M. Grether, and J. de Melo, “Determinant of Attitudes towards Immigration: a TradeTheoretic Approach,” discussion paper 1877, CEPR, London (1998). Davies, J. and I. Wooton, “Income Inequality and International Migration,” Economic Journal 102 (1992):789–802. Ethier, W., “Bundled International Factor Mobility,” in H. Kierzkowski (ed.), Protection and Competition in International Trade: Essays in Honor of W. M. Corden, Oxford: Blackwell (1987). Findlay, R. and H. Grubert, “Factor Intensities, Technical Progress, and the Terms of Trade,” Oxford Economic Papers 11 (1959):111–21. Freeman, G.P., “Modes of Immigration Politics in Liberal Democratic States,” International Migration Review 29 (1995):881–902. Hillman, A., “The Political Economy of Migration Policy,” in H. Siebert (ed.), Migration: a Challenge for Europe: Symposium 1993, Tübingen: Mohr (1994). Hillman, A. and A. Weiss, “Beyond International Factor Movements: Cultural Preferences, Endogenous Policies and the Migration of People: an Overview,” in R. Faini, J. de Melo, and K. Zimmermann (eds.), Trade and Migration: the Issues and the Evidence, Cambridge: Cambridge University Press (1999). Jones, R. W. and S. T. Easton, “Factor Intensities and Factor Substitution in General Equilibrium,” Journal of International Economics 15 (1983):65–99. Jones, R. W. and J. A. Scheinkman, “The Relevance of the Two-Sector Production Model in Trade Theory,” Journal of Political Economy 85 (1977):909–35. Mayer, W., “Endogenous Tariff Formation,” American Economic Review 74 (1984):970–85. Müller, T., “Migration Policy in a Small Open Economy with a Dual Labour Market,” Review of International Economics 11 (2003):130–43. Schiff, M., “Migration et echanges: aspects positifs et normatifs,” in J. de Melo and P. Guillaumont (eds.), Commerce nord-sud, migration et délocalisation, Economica, Paris (1997):33–54. Ruffin, R. J., “Trade and Factor Movements with Three Factors and Two Goods,” Economics Letters 7 (1981):177–82. Scheve, K. F. and M. J. Slaughter, “Labor-Market Competition and Individual Preferences over Immigration Policy,” NBER working paper6946 (1999). © Blackwell Publishing Ltd 2003 266 Sanoussi Bilal, Jean-Marie Grether, and Jaime de Melo Thompson, H., “Factor Migration and Income Distribution in International Trade,” Keio Economic Studies 20 (1983):65–70. Thompson, H. and D. Clark, “Factor Movements with Three Factors and Two Goods in the US Economy,” Economics Letters 12 (1983):53–60. Wellisch, D. and U. Walz, “Why do Rich Countries Prefer Free Trade Over Free Migration: the Role of the Modern Welfare State,” European Economic Review 42 (1998):1595–612. Wolff, E., “International Comparisons of Wealth Inequality,” Review of Income and Wealth 42 (1996):433–51. Wood, A., North–South Trade, Employment and Inequality, Oxford: Clarendon Press (1994). Notes 1. Relying on the interest-group model, Freeman (1995) argues that benefits from immigration are concentrated among producers (who enjoy lower wages and face a higher demand) whereas the costs are diffused among the working population group (which earns lower wages). Hence, organized interests get more attention from political authorities than the general public does, and immigration policies appear more liberal than those that would be wished by the majority of the population. For a critical survey of the political-economy literature on immigration, see Hillman and Weiss (1999). 2. Benhabib (1996) also uses a direct-democracy model, but with one factor (and one sector). His model shares some of the properties of the model developed here, namely that an immigration policy that increases (lowers) the economy-wide average capital–labor ratio will be defeated if the median voter’s capital–labor ratio is above (below) a critical capital ownership level. See Ethier (1987) for a treatment of bundled factor mobility in the m ¥ n generalization of the Heckscher–Ohlin model. 3. This has the advantage of focusing on the most interesting case in terms of the variety of results. Davies and Wooton (1992) argue that the broad evidence suggests that the US belongs to this case, but Thompson and Clark (1983) suggest that low-skill labor may be the middle factor. The alternative case where high-skill labor is the middle factor is treated in Bilal et al. (1998). 4. It can be shown that xi = s ki - (aLsLi + aHsHi ), where s ei is the economy-wide elasticity of the use of factor e (e = K, L, H) with respect to wi (i = K, L, H), under the assumption that each industry’s output is kept constant and aL, aH are defined in equation (7). 5. Multiplying the numerator and denominator of equation (7) by Y/(p1X1) leads to aL = (qL/qK)(qK1qH - qH1q K)/(qL1qH - qH1qL). This expression can be further simplified under the symmetry assumption, which implies that qK1 = qK2 = q K and that qL1 + qH1 = qL2 + qH2 = qL + qH. 6. D = (aLxL/q L + aHxH/q H - xK/qK); see Jones and Scheinkman (1977). 7. By the envelope theorem, the marginal impact of a unitary increase in any factor endowment on national income is equal to this factor reward, which means that total income of the incumbent factors remains unchanged. This zero-sum property of total income to marginal changes in immigration comes from the following assumptions: immigration is marginal, prices of goods are fixed, there is no redistribution policy (Wellisch and Walz, 1998), and the immigrants get paid the going wage, rather than some other wage as in, for example, the efficiency wage model with a dual labor market of Müller (2003). 8. From (9), remembering that aL + aH = 1 it is straightforward to check that dY = Sldyl + Shdyh = 0. 9. In the alternative case where high-skill labor is the middle factor, national attitude towards immigration becomes mainly “skill concerned”: as in this case capital and low-skill labor are enemies, attitudes are independent on the capital ownership of households (although they do depend on the capital ownership of high-skill immigrants). 10. To be rigorous, one should take into account the redistributive effects of a reduction in trade barriers (e.g., who gets the resulting tariff revenue). This is ignored here. 11. Some have proposed (Wood, 1994) that globalization has induced “defensive” labor-saving technical progress in the labor-intensive import-competing sectors. © Blackwell Publishing Ltd 2003 ATTITUDES TOWARDS IMMIGRATION 267 12. This implies identical standard forms for both groups (the standard form for the beta is obtained through the change of variable that restricts the domain to the interval [0, 1]). 13. For values of aL approximating zero, the critical capital ownership level of high(low)-skill households is very high (low). But as high-skill households receive a higher proportion of total capital, a sufficient number of them go along with low-skill households to create a majority favorable to capital-poor immigration. 14. Of course different values would be obtained with other functional forms to represent the distribution of capital. We have used the beta since it makes it easy to reproduce the base case of no intragroup capital ownership disparity. © Blackwell Publishing Ltd 2003