Untitled-2 NOTA CRÍTICA The Political Determinants of Migration Control: A Quantitative Analysis Marc R. Rosenblum University of New Orleans Although most analysts agree that the fundamental causes of international migration are overwhelmingly eco- nomic and demographic—pushes, pulls, and social networks—the mi- gration policies of host states are fil- ters through which these factors operate. At a minimum, lower migra- tion quotas, more emphasis on en- forcement, and barriers to migrant integration reduce the expected ben- efits of migration and raise the ex- pected costs, which deters some migrants and causes others to choose alternate destinations. Thus, despite dramatically falling transportation costs and rising international inequal- ity, post-World War II migration flows have increased far more slowly than trade and financial flows; and most of the world’s poor still choose not to migrate. Not only do migration-control policies “matter,” but they vary widely and often in unexpected ways. Although “settler states,” like the United States, Canada, Austra- lia, and New Zealand, admit more migrants than Old World coun- tries, the differences are smaller than one might expect.1 Other mi- gration statistics belie national im- ages and conventional wisdom, with “non-immigrant” Germany and “effectively restrictionist” En- gland ranking first and third, re- spectively, in terms of relative migrant admissions (defined as the natural log of admissions divided by the natural log of host-state population). It is perhaps also sur- prising that the progressive Scan- dinavian democracies (Finland, Norway, Sweden) admit relatively fewer migrants than do their neigh- bors (France, Germany, Belgium, Switzerland, and the Netherlands). In accounting for variations in migration policy, political scientists have developed hypotheses focused on three areas: interest groups, po- litical institutions, and interna- tional factors. The most common of these hypotheses is based on the 1 Between 1962 and 1998, settler states admitted an annual average of 6.8 migrants per 1000 residents, compared to 5.0 per 1000 residents in non-settler states; but by 1998, the figures had converged to 5.1 for the former and 4.7 for the latter. 162 MIGRACIONES INTERNACIONALES role of interest groups. Because im- migration is characterized by cross- cutting cleavages, migration policy is rarely a clearly partisan issue. In- stead, it is argued, client politics tends to influence immigration policy, with owners of land and capital enjoying privileged-group status and seeking the concentrated benefits of lower wages that a more open policy would bring (Freeman, 1995; Joppke 1998). Immigrants also support a more open policy but tend to be less well organized (Hanson et al., 2001). This leaves unions and nationalists as the only significant groups opposing labor inflows. The former are often di- vided between the desire to block migration and the desire to orga- nize new constituents (Haus, 1995; Watts, 2000), and the latter have generally been only a latent politi- cal force (Hainsworth, 2000). In- terest-group theories thus attribute the “gap” observed between popu- lar demands for migration control and generous admissions policies to the existence of well-organized sup- porters and a latent or divided op- position (Cornelius, Martin, and Hollifield, 1995). A second set of arguments focuses on political institutions. Jeannette Money (1999) argues that demands for migration-policy changes are transmitted from gateway commu- nities to national policy-makers only as a result of closely contested elec- tions. Keith Fitzgerald (1996) em- phasizes the path-dependent nature of migration-enforcement institu- tions. More generally, several ana- lysts emphasize the vulnerability of courts and other liberal institutions to exploitation by pro-immigration actors, with the expectation that more “efficient” institutional sys- tems (that is, those with fewer ac- cess points) are more capable of enforcing effective migration con- trols (Hollifield, 1992; Joppke, 1998; Jacobson, 1996). Finally, at least two migration- policy arguments exist that are in- ternational in nature. Arguing from a variety of disciplinary approaches, Wayne Cornelius, Philip Martin, and James Hollifield (1995), Dou- glas Massey et al. (1998), and Saskia Sassen (1998) emphasize economic integration as both a migration push factor and a limitation on the will- ingness or ability of host states to control inflows. This “economic globalization” argument is often complemented by a “liberal global- ization” argument, which holds that states’ integration within the inter- national system acts as a normative or an institutional constraint on their ability to restrict inflows (Jacobson, 1996; Soysal, 1994). Each of these arguments has re- ceived extensive attention in recent years; however, these competing hypotheses about immigration policy have rarely been tested com- paratively or using quantitative analysis (with the exceptions of Money, 1999, and O’Rourke and Williamson, 1999). Thus, this re- search note proposes a research de- sign to fill this gap, and it presents initial results of my analysis of the three hypotheses outlined above. NOTA CRÍTICA 163 Research Design I propose to test these competing models of immigration policy-mak- ing using a time-series, cross-sec- tional analysis of immigration to 15 OECD states from 1962 through 1998.2 The inclusion of host states ranging from low-flow cases, like Ja- pan and Finland, to high-flow cases, like the United States and Germany, and of periods before, during, and after the economic shocks of the 1970s insures a wide variation in immigration outcomes. This re- search design must resolve at least four sets of methodological issues. First, analysis of immigration policy confronts a fundamental problem in that no reliable compara- tive policy data exist: Each country has unique visa categories, enforce- ment mechanisms, integration rules, and citizenship procedures, among other policy dimensions.3 In the United States, for example, at least 59 types of non-immigrant visas and 100 types of permanent visas exist. Illicit undocumented flows further compound measurement problems. Thus, my analysis follows Money’s example by focusing on total legal immigration.4 Legal im- migration has the methodological advantage of being the one figure for which cross-national time-series data are readily available and com- parable: All states define legal per- manent immigrants in similar ways, and legal immigrants are easy to count. “Front-door” migration is also attractive theoretically because it is the largest category of inflows to developed states, and legal per- manent migrants promote chain migration. Both proponents and opponents of migration therefore recognize these flows as having high stakes, making this category a criti- cal test of interest-group hypoth- eses in particular. A second set of methodological issues relates to the analysis of time- series, cross-sectional data, which is likely to be characterized by con- temporaneous and panel-specific correlation of errors and complex dynamic effects. I minimize the former problem by including coun- try-specific dummy variables (fixed effects) and time-period variables (a dummy coded 1 during the rela- tively high flow period prior to 1974, and a second dummy coded 2 My sample includes Australia, Belgium, Canada, Denmark, Finland, France, Germany, Japan, the Netherlands, New Zealand, Norway, Sweden, Switzerland, the United Kingdom, and the United States. I thank Jeannette Money for making her (1999) immigration data available. I have supplemented her sample by adding an additional country (Switzerland), as well as nine additional years (1990-1998). 3 Kevin O’Rourke and Jeffrey Williamson (1999) seek to resolve this issue by creating an index of policy shifts from 1870 to 1920, but no such data exist for the current period. 4 Because both migration and population data are highly skewed in my sample, I transform these data by analyzing a ratio of immigration (logged) to host-state population (logged). My 555 possible country-years included 54 missing observations. I employed multiple imputation software to fill in these observations rather than risk model inefficiency and bias associated with list-wise deletion (King et al. 2001). Reported results employed Clarify software and Monte Carlo simulations to account for the uncertainty associated with these imputations. Tests indi- cate that these methodological choices did not substantially affect results. 164 MIGRACIONES INTERNACIONALES 1 for the 1974-1986 period of re- duced migration flows). I also con- trol for heteroscedasticity by employing panel-corrected stan- dard errors (Beck and Katz, 1995). I address time-series dynamics of my data, some of which are nonstationary, by employing an “er- ror correction” model that regresses the first difference of the dependent variable on the lagged dependent variable as well as on lags and first differences of the independent vari- ables (Davidson et al., 1978). Al- though the resulting model has minor collinearity problems, this approach effectively models short- term and long-term dynamic effects and controls serial correlation of er- rors, as confirmed by reported Lagrange multiplier statistics (see tables below). Fourth, in order to focus on the political determinants of migration control, I control for underlying economic and demographic moti- vations. Country fixed effects and lagged migration data capture ef- fects of previous migration, and as economic controls, I include receiv- ing-state data on unemployment and GNP. Due to data limitations, I do not control for emigration pushes. This omission should not be problematic, however, because the economic and demographic pushes in the developing world are so extreme in the period under con- sideration that it is reasonable to as- sume, as Money does, that “variation in flows is determined almost exclusively by government policy in the host state rather than by the supply of migrants” (1999, 23). Finally, I developed 10 operational measures for the hypotheses identi- fied above, and I predicted their re- lationships to immigration inflows (see Table 1). The following section discusses these variables and the pre- liminary results of my analysis. Table 1. Independent Variables and Predicted Relationships. Hypothesis Variable Definition Source Prediction Percent of workers Golden et al., Interest groups Union density with union representation 1997; ILO - Migrant Manufacturing as share employers of total exports World Bank, 2001 + Left-right 0: Left government partisanship 1: Center government 2: Right government Beck et al., 2001 + Ethnic Effective number homogeneity of ethnic groups US CIA, 2002 - Settler state Coded 1 for Australia, Canada, New Zealand, United States + Institutions Partisan unity -1 * Rae index of fractionalization Banks, 2002 - Institutional unity -1 * Federal-Unitary Index Lijphart, 1999 - Partisan-institutional Polconv variable (probability unity of policy change based on veto player analysis) Henisz, 2002 - International Trade Trade as proportion of GNP World Bank, 2001 + Regime membership Proportion of existing UN human rights treaties ratified UNHCR, 2002 + NOTA CRÍTICA 165 Analysis of Results Table 2 presents the results of my control model (excluding country and time-period coefficients) and of five tests of the interest-group hypothesis.5 Four of these variables are significantly related to immigra- tion inflows in the expected direc- tion: unionization rates are associ- ated with less immigration (short-term and long-term); right- wing governments are associated with more migration and left-wing governments with less (long-term only); ethnic homogeneity is asso- 5 Lagged independent variables in the error-correction model measure the long-term equi- librium relationship between independent and dependent variables and should be multi- plied by -1 times the coefficient on the lagged dependent variable prior to substantive interpretation. First-differenced variables measure the short-term relationship between a change in independent variables and the dependent variable. Fixed effects also measure long-term relations but are interpreted directly (see Davidson et al., 1978). Table 2. Political Determinants of Migration Flows, 15 OECD States.a Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Dependent -.230*** -.242*** -.232*** -.237*** -.230*** -.230*** Lags variable (.035) (.036) (.035) (.034) (.035) (.035) -.185*** -.195*** -.184*** -.198*** -.185*** -.185*** Unemp. (.054) (.053) (.053) (.054) (.054) (.054) 1.26*** 1.40*** 1.33*** 1.36*** 1.26*** 1.26*** GDP (.466) (.460) (.462) (.461) (.466) (.466) Union -0.33** density -- (.014) -- -- -- -- Migrant -0.31* employers -- -- (.016) -- -- -- Left-right -.273** partisanship -- -- -- (.121) -- -- -.511*** -.512*** -.518*** -.512*** -.511*** -.511*** Differences Unemp. (.116) (.114) (.115) (.115) (.116) (.116) .686 .699 .842 .759 .686 .686 GDP (.977) (.950) (.979) (.952) (.977) (.977) Union -.074*** density -- (.026) -- -- -- -- Migrant -.018 employers -- -- (.041) -- -- -- Left-right .039 partisanship -- -- -- (.155) -- -- Fixed Ethnic -.090*** effects homogeneity -- -- -- -- (.025) -- 1.25* Settler state -- -- -- -- -- (.770) Adjusted R2 .138 .156 .144 .145 .141 .141 .087 .073 .089 .034 .088 .088 LM1b (.106) (.107) (.106) (.101) (.106) (.106) -.077 -.091 -.078 -.119 -.077 -.077 LM2b (.097) (.099) (.097) (.091) (.097) (.097) aOrdinary least squares coefficients with panel-corrected standard errors in parentheses. Constants and country fixed effects omitted from table. bLagrange Multiplier coefficients on first and second lagged residuals, regressed on residuals. Insignificant results indicate no significant serial correlation of errors. *** P<.01; **P<.05; *P<.10, in two-tailed t-tests. 166 MIGRACIONES INTERNACIONALES ciated with less migration (as a fixed effect); and settler states are associ- ated with more migration (fixed effect). The importance of migrant employers was also related to im- migration in the short-term, but the sign was negative, the opposite from what the interest-group hy- pothesis predicts. Tests of five institutional and in- ternational hypotheses had mixed results (see Table 3). The significant and positive sign associated with Arend Lijphart’s (1999) index of institutional unity suggests that more federalized states (Switzerland, Germany) admit more migrants than more unitary states (England, Table 3. Institutional and International Determinants of Migration Flows, 15 OECD states.a Variable Model 7 Model 8 Model 9 Model 10 Model 11 Dependent -.230*** -.253*** -.239*** -.232*** -.232*** Lags variable (.035) (.037) (.035) (.035) (.035) -.185*** -.191*** -.181*** -.166*** -.171*** Unemp. (.054) (.054) (.053) (.056) (.055) 1.26*** 1.60*** 1.34*** 1.32*** 1.44*** GDP (.466) (.486) (.463) (.472) (.478) Partisan 5.27*** density -- (2.06) -- -- -- Partisan- -2.96 institutional unity -- -- (2.67) -- -- -.012 Trade -- -- -- (.015) -- Regime -.501 membership -- -- -- -- (.550) -.511*** -.484*** -.490*** -.533*** -.498*** Differences Unemp. (.116) (.114) (.115) (.118) (.115) .686 .927 .820 .336 .911 GDP (.977) (.964) (.974) (1.03) (.969) Partisan 5.13* unity -- (2.85) -- -- -- Institutional 9.62 unity -- -- (6.04) -- -- -.029 Trade -- -- -- (.024) -- Regime -1.45* membership -- -- -- -- (.834) Fixed Institutional 1.24* effects unity (.723) -- -- -- -- Adjusted R2 .141 .153 .147 .141 .144 .088 .116 .055 .075 .071 LM1b (.106) (.102) (.102) (.105) (.105) -.077 -.056 -.106 -.093 -.080 LM2b (.097) (.092) (.094) (.098) (.098) aOrdinary least squares coefficients with panel-corrected standard errors in parentheses. Constants and country fixed effects omitted from table. bLagrange Multiplier coefficients on first and second lagged residuals, regressed on residuals. Insignificant results indicate no significant serial correlation of errors. *** P<.01; **P<.05; *P<.10, in two-tailed t-tests. NOTA CRÍTICA 167 New Zealand). Similarly, the exist- ence of fewer parties in parliament positively correlates with inflows (short- and long-term). Addition- ally, overall institutional unity, which combines the number and ideological dispersal of veto players into a single index, is positively cor- related with admissions (only in the short-term, and only with a P-value of .11). Finally, with regard to in- ternational factors, only the first dif- ference of regime membership is statistically significant, but the sign is negative, suggesting that as coun- tries join human-rights regimes, their migrant admissions decline. Although space prevents further discussion of these findings, the re- sults reported as significant here are highly robust to alternative specifi- cations. Three points therefore bear emphasis. First, my findings clearly suggest that interest groups matter, though not always in the expected ways. On one hand, my evidence that unions, ethnic groups, and par- tisanship influence migration admis- sions in predictable ways strongly supports the overall validity of my approach. My counterintuitive find- ings about employer strength are somewhat troubling but probably re- flect operational problems with this variable (manufacturing as a share of exports to measure migrant-em- ployer strength is certainly the least valid of my indicators). Second, my finding that more pluralized institu- tional and partisan structures are consistently associated with less mi- gration directly conflicts with the predictions of the “liberal” hypoth- esis. My interpretation is that in the pursuit of economic growth without inflation, all states prefer higher mi- gration inflows; and institutional iso- lation enhances the ability of states to pursue this goal despite popular opposition. That is, contrary to ex- isting domestic institutional argu- ments, more veto players imply more access points for immigration oppo- nents, who are less well organized than supporters of migration. Third, controlling for other factors, I find no evidence that international inte- gration or international institutional membership significantly affects states’ migration policies. Although null findings are hardest to verify, the absence of expected international effects was robust to all model speci- fications. Of course, the findings re- ported here are the first step in what appears to be a promising research agenda; and additional analysis of partisan and institutional interactive effects, change over time, and other issues is clearly in order. References Beck, Nathaniel, and Jonathan N. Katz, “What To Do (and Not To Do) with Time-Series Cross-Section Data”, American Political Science Re- view, 89, 1995, pp. 634-48. 168 MIGRACIONES INTERNACIONALES Cornelius, Wayne A., Philip L. Martin, and James F. Hollifield (eds.), Controlling Immigration: A Global Perspective, Stanford (Calif.), Stanford University Press, 1995. Davidson, James, David Hendry, Frank Srba, and Stephen Yeo, “Econo- metric Modeling of the Aggregate Time-Series Relationship between Consumers’ Expenditure and Income in the United Kingdom”, The Economic Journal, 352, 1978, pp. 661-92. Fitzgerald, Keith, The Face of the Nation: Immigration, the State, and the National Identity, Stanford (Calif.), Stanford University Press, 1996. Freeman, Gary, “Modes of Immigration Politics in Liberal Democratic States”, International Migration Review, 39, 1995, pp. 881-902. Golden, Miriam, Peter Lange, and Michael Wallerstein, Union Central- ization Among Advanced Industrial Societies: An Empirical Study, 1997, dataset available at http://www.shelley.polisci.ucla.edu/data. Version dated on February 11, 1998. Joppke, Christian, “Why Liberal States Accept Unwanted Immigration”, World Politics, 50, 1998, pp. 266-93. Hanson, Gordon, Kenneth Scheve, Matthew Slaughter, and Antonio Spilimbergo, “Immigration and the U.S. Economy: Labor Market Impacts, Illegal Entry, and Policy Choices”, report to the Fondazione Rodolfo Debenedetti (Milan, Italy), 2001. Hainsworth, Paul (ed.), The Politics of the Extreme Right, New York, Pinter, 2000. Haus, Leah, “Openings in the Wall: Transnational Migrants, Labor Unions, and U.S. Immigration Policy”, International Organization, 49, 1995, pp. 285-313. Henisz, W. J., “The Institutional Environment for Infrastructure Invest- ment”, Industrial and Corporate Change, 11, 2002. Hollifield, James, Immigrants, Markets, and States: The Political Economy of Postwar Europe, Cambridge, Harvard University Press, 1992. Jacobson, David, Rights Across Borders: Immigration and the Decline of Citizenship, Baltimore (Md.), Johns Hopkins University Press, 1996. King, Gary, James Honaker, Anne Joseph, and Kenneth Scheve, “Analyzing Incomplete Political Science Data: An Alternative Algorithm for Multiple Imputation”, American Political Science Review, 95, 2001, pp. 49-69. Lijphart, Arend, Patterns of Democracy: Government Forms and Performance in Thirty-six Countries, New Haven (Conn.), Yale University Press, 1999. Massey, Douglas, Joaquín Arango, Graeme Hugo, Ali Kouaouci, Adela Pellegrino, and J. Edward Taylor, Worlds in Motion: International Migra- tion at the End of the Millennium, Oxford, Oxford University Press, 1998. Money, Jeannette, Fences and Neighbors: The Political Geography of Im- migration Control in Advanced Market Economy Countries, Ithaca (N.Y.), Cornell University Press, 1999. NOTA CRÍTICA 169 O’Rourke, Kevin, and Jeffrey Williamson, Globalization and History: The Evolution of a Nineteenth Century Atlantic Economy, Cambridge, MIT Press, 1999. Sassen, Saskia, Globalization and Its Discontents, New York, The New Press, 1998. Soysal, Yasemin N., Limits of Citizenship: Migrants and Postnational Membership in Europe, Chicago, University of Chicago Press, 1994. United Nations High Commission for Refugees (UNHCR), Status of Rati- fications of the Principal International Human Rights Treaties, 2002. Available at http://www.unhchr.ch/pdf/report.pdf. Accessed on April 22, 2003. United States Central Intelligence Agency (US CIA), The World Fact Book, 2002. Available at http://www.odci.gov/cia/publications/ factbook/. Accessed April 22, 2003. Watts, Julie, “An Unconventional Brotherhood: Union Support for Lib- eralized Immigration in Europe”, Monograph #1, Center for Com- parative Immigration Studies-University of California, San Diego, La Jolla, Calif., 2000. World Bank, World Development Indicators [cd-rom], Washington, D.C., The World Bank, 2001.