Beyond the Mexican Case: A Theoretical, Empirical and Policy Analysis of Central American Migration to the United States Center for Jack DeWaard & Keuntae Kim Demography and Ecology University of Wisconsin Madison

PROJECT DESCRIPTION KEY MEASURES KEY FINDINGS The aim of this paper is to both test and extend the existing body of Sex………………………………… Sex of respondent (reference category: female) theoretical and empirical work on the determinants of international 1. We more or less replicated the work of Massey and Espinosa (1997) in our additive models, but of Country of origin………………….. Country of origin (reference category: Nicaragua) migration. About a decade has passed since Massey and Espinosa’s (1997) U.S. Inflows from Mexico: 1946-2004 course applied to the countries of Costa Rica, Guatemala and Nicaragua. In Model 4, for instance, Age 15-49…………………………. Between ages 15 & 49 (reference category: 50+) detailed account of Mexico-U.S. migration, wherein they found support for 1,000,000 the signs for each of the theoretical predictors – expected wage ratio, real interest rate, foreign 900,000 social capital formation, human capital formation, and market consolidation 800,000 Labor force experience……………. Number of years actual labor force experience 700,000 liabilities, and migrant siblings – are all significant and in the expected direction. in explaining first and subsequent migration from Mexico to the United 600,000 Education…………………………. Number of years of school completed 500,000 Mexico 2. Turning to our interactive models, we begin to see the picture get considerably more complicated. States. Their work spawned subsequent efforts among scholars to further 400,000 300,000 Visa availability……………………. Legal immigration divided by sum of legal and gross entries elucidate the nuances of Mexico-U.S. migration, including the dynamics of 200,000 The interactions between analysis time and each country of origin dummy are significant. This is as 100,000 Expected wage ratio……………….. Predicted ratio from data on home & US wages origin communities and unique migration streams (Fussell 2004; Lindstrom 0 it should be. 6 1 5 46 61 76 90 00 19 1951 195 19 1966 197 19 1980 198 19 1995 20 Real interest rate…………………… Interest rate minus inflation rate and Lauster 2001). While the abundance of scholarship on Mexico-U.S. 3. We then experiment with the expected wage ratio. Massey and Espinosa (1997) picked up a weak migration is no doubt impressive, it remains to be seen whether the Foreign liabilities…………………… Rate of change in foreign liabilities of monetary authority positive effect at best. However, two- and three-way interaction terms show that the expected wage conclusions that have emerged from these studies can be said to hold beyond Migrant siblings……………………. Number of siblings with US migration experience ratio continues to be at play and, moreover, that the effects differ by undocumented and the Mexico-U.S. case. Period 1965-1980…………………... Year 1965-1980 (reference category: 1981-2000) documented migration. With respect to the former, we see that the effect of the expected wage Using data from the Latin American Migration Project (LAMP), we analyze *Various interactions (see results from interactive models below) ratio has declined over time in both Costa Rica and Guatemala relative to Nicaragua. The opposite U.S. Inflows from Central America: 1946-2004 first migration to the United States from Costa Rica, Guatemala, and is the case where documented migration is concerned. Nicaragua over the period 1965-2000. Like Massey and Espinosa (1997), we 35,000 30,000 RESULTS: ADDITIVE MODELS 5. Finally, Model 8 brings in a simple period effect, something we discussion in the conclusion of this employ a rich set of theoretical predictors and multinomial discrete time 25,000 Costa Rica 20,000 poster. event history models. We likewise extend the work of Massey and Espinosa Guatemala 15,000 Nicaragua (1997) in two respects. First, we do not restrict our sample to men; our doing 10,000 so represents our attempt to highlight international migration as a highly 5,000 0 PREDICTED HAZARDS

gendered process (Hondagneu-Sotelo 1994). Second, we provide an explicit 51 56 71 76 85 90 1946 19 19 1961 1966 19 19 1980 19 19 1995 2000 treatment of duration dependence and show that researchers must be duly Predicted Hazards: concerned with both the changes in their predictors and the changes in the Data Source: United Nations (2005) First Undocumented and Documented US Migration effects of their predictors over time. In our final models, we provide an 0.04000 example that nicely illustrates this distinction. 0.03000 Undocumented 0.02000 Documented

RESEARCH QUESTIONS Pr(T=t|T>=t,x) 0.01000

1. What’s driving Central American migration to the United States? 0.00000 2. Are the same factors driving Central American Migration to the United States that Massey and Analysis 5 1015202530 Time Espinosa found to be driving Mexico-U.S. migration, namely - social capital formation, human capital

formation, and market consolidation? Predicted Hazards: Predicted Hazards: 3. Where and how does gender fit in? Is there a constant gender effect over time? Or does impact of First Undocumented US Migration by Sex First Documented US Migration by Sex gender vary by, say, the legal status of the migrant (i.e., undocumented or documented)? 0.04000 0.04000 4. How do the relevant driving factors play out with respect to one another and with time? Does it make 0.03000 0.03000 Male sense that their effects should be considered constant? Or is their evidence for accelerated and/or Male 0.02000 0.02000 Female diminishing returns? Female Pr(T=t|T>=t,x) 5. The above questions considered, what is “left over” so to speak? Might our residuals be a further Pr(T=t|T>=t,x) 0.01000 0.01000 indication as to the importance of place when considering the process of international migration? 0.00000 0.00000 Analysis 5 1015202530 Analysis 5 1015202530 Time Time

DATA & METHODS Predicted Hazards: First Undocumented US Migration by Predicted Hazards: First Undocumented US Migration by Selected Country of Origin Selected Country of Origin * Latin American Migration Project (LAMP): http://lamp.opr.princeton.edu * 3,681 men and women ages 15+ from Costa Rica, Guatemala and Nicaragua between 1965-2000. 0.00200 0.00200 ** 2,763 males (74.83%) and 918 females (24.94%). RESULTS: INTERACTIVE MODELS 0.00150 0.00150 Costa Rica Costa Rica 0.00100 0.00100 ** 1,408 (35.51%) Costa Ricans, 508 (10.66%) Guatemalans and 1,765 (53.83%) Nicaraguans. Guatemala Nicaragua (note: percentages are weighted) Pr(T=t|T>=t,x) * 93,614 person years. Pr(T=t|T>=t,x) 0.00050 0.00050 ** 67,819 (72.13%) male and 25,795 (27.87%) female person years 0.00000 0.00000 Analysis 5 1015202530 Analysis 5 1015202530 ** 35,447 (35.66%) Costa Rican, 12,317 (10.29%) Guatemalan and 45,850 (54.04%) Nicaraguan Time Time person years. (note: percentages are weighted) Predicted Hazards: First Undocumented US Migration from Predicted Hazards: First Undocumented US Migration from * Outcome variable: Event of first migration to the United States Costa Rica by Sex Costa Rica by Sex ** Competing risks of: 0.00200 *** Undocumented migration vs. no migration 0.00200 0.00150 0.00150 *** Documented migration vs. no migration Male Male 0.00100 0.00100 Female * Undocumented first migration: Female Pr(T=t|T>=t,x) 0.00050 ** Log rank tests reveal: Pr(T=t|T>=t,x) 0.00050 *** Statistically significant difference in the baseline hazard functions for men and women. 0.00000 0.00000 Analysis 5 1015202530 Analysis 5 1015202530 *** Statistically significant differences in the baseline hazard functions for Costa Ricans, Time Time

Guatemalans and Nicaraguans. * Documented first migration ** Log rank tests reveal: DISCUSSION, CHALLENGES & NEXT STEPS *** No statistically significant difference in the baseline hazard functions for men and 1. The period effect in Model 8 seemingly warns us that our models are incomplete. This simple women. period effect is intended to be a rough catch for the periods of war and unrest in both Guatemala *** Statistically significant differences in the baseline hazard functions for Costa Ricans, and Nicaragua. The Guatemalan Civil War ran from 1960-1996. Nicaragua experienced profound Guatemalans and Nicaraguans. political changes with the Sandinista Revolution which arguably reached its apex in 1979. * Method: 2. As a methodological issue, the models developed in this poster are premised on a non-traditional ** Discrete time event history models for competing risks. risk set. While we “start the clock” in 1965 for those ages 15 or older, we also allow additional ** Also known as multinomial event history models; proportional odds models. persons to enter the risk set when they turn 15. While the notion of both increments and ** See Singer and Willet (2003) and Yamaguchi (1991). decrements is more realistic, it is also more difficult to model, especially that of duration dependence.