CHANGES IN TRIP DURATION

Transcription

1 CHANGES IN TRIP DURATION FOR MEXICAN IMMIGRANTS TO THE UNITED STATES Belinda I. Reyes Introduction Mexican migration has been characterized by its temporary nature (Bean, Telles & Lowell, 1987; Calavitas, 1994; Cornelius, 1976; Hugo, 1982; Jenkings, 1977; Jones, 1982; Kossoudji, 1983; Massey, et. al., 1987; Mines & de Janvry, 1982; Ranney & Reyes, 2001; Roberts, 1995; White, Bean & Espenshade, 1990). Although not all immigrants move temporarily, most of the Mexican immigrants who enter the United States have been found to return to Mexico after a few months or years (Reicher & Massey, 1979; Mines, 1981; Lopez, 1986; Massey, et. al., 1987; Durand & Massey, 1992; Massey, Goldring & Durand, 1994; Massey & Singer, 1995). In recent years, however, the overall pattern of trip duration may have changed. Community studies find increases in trip duration and more settlement in the United States (Alarcón, 1995, 1995a; Cornelius, 2002), while quantitative studies find a shortening in trip duration after the passage of the Immigration Reform and Control Act (Durand, Massey & Zenteno, 2000). While, there is an extensive literature exploring the determinants of trip duration (Lindstrom, 1996; Massey and Espinosa, 1997; and Reyes, 2001, for example), this article explores the patterns of trip duration for Mexican immigrants to United States and the factors that led to the change in the patterns observed. For this analysis, I developed hierarchical discrete-time logit models of the probability of return using data from the Mexican Migration Project (MMP). Reasons for Changes in Trip Duration < Figure 1 Here > Figure 1 shows the rate of return of people who had made a trip to the United States in the five-year periods before 1992, 1997, and These data are generated from the 1992 and 1997 ENADID and the 2000 Mexican Census. In these surveys, informants were asked about every person who lived in the household at some point in the last five years who had made an international migration independent of their current residence, capturing international migrants who were connected to some household in Mexico. A smaller proportion of immigrants returned to Mexico in the later part of the 1990s (the period) than in prior periods. Fifty-four percent of the immigrants in the sample returned to Mexico by the time of the survey in the 1987 to 1992 period, whereas only 25 percent returned to Mexico between 1995 and Even those who did return to Mexico are staying longer in the United States, as shown in Figure 2. In the period, those who returned to Mexico lived on average 10 months in the United States, compared to those who returned to Mexico in the later part of the 1990s, who lived in the United States for almost 16 months before returning. < Figure 2 Here > 1

2 The pattern of trip duration could have changed for several reasons: a change in the characteristics of the migrants, an improvement in the networks available to migrants, changes in economic opportunities in either Mexico or the United States, or changes in U.S. immigration policy. The effect of each of these factors on trip duration will be explained below. If different people migrated to the United States in the 1990s than in prior period, this could lead to changes in trip duration. Female migrants, skilled workers, non-household heads, wealthier migrants, and migrants from less traditional sending regions have longer trips and are more likely to settle in the United States than their counterparts (Massey & Espinosa, 1997; Lindstrom, 1997; Reyes, 2001). Furthermore, Marcelli & Cornelius (2001) argue that in the last ten years, there has been a change in the types of migrants that move to the United States. They find greater female migration, migrants with higher levels of education, and more migrants from urban places and from nontraditional sending regions than in the past. If true, this should lead to longer duration for the overall cohort of migrants. U.S. experience and the expansion of networks in the United States, also affects migrants expectations and desires to return, leading to adjustments in duration or long-term settlement (Massey, et. al., 1987; Massey, 1987; Massey, 1990; Massey, Goldring, & Durand, 1994; Massey & Espinosa, 1997; Piore, 1979). With U.S. experience, migrants become accustom to life in the Unites States and may readjust their return plans. Moreover, as immigrants expand their networks and create communities in the United States, they increase opportunities for other migrants, leading to even greater settlement. As the migration process expands in particular communities, the expectation is that more and more migrants would become permanent settlers in the United States (Massey, et. al., 1987; Massey & Liang, 1989). Furthermore, the conditions in both Mexico and the United States could lead to changes in trip duration (Roberts, 1995; Massey & Espinosa, 1997; Massey et.al., 1987). As long as immigrants are able to support family members at home, they may choose to leave their families behind and engage in circular migration. However, if opportunities at home are no longer sufficient to support their families, they may choose to resettle their families and stay longer in the United States (Roberts, 1995). If opportunities in the United States improve, migrants may also be able to secure their targeted level of income quicker than otherwise expected, increasing the likelihood of return. U.S. immigration policy may also be an important factor. The legalization of almost 3 million previous unauthorized immigrants under the Immigration Reform and Control Act (IRCA) of 1986 could have led to longer duration as newly legalized migrants established households in the United States (Alarcón, 1995, 1995a). Massey and Espinosa (1997) argue that as males are legalized, their wives become more insistent on legalizing other family members and moving the whole family to the United States. On the other hand, a legalization may allow migrants to engage in back-and-forth migration without fear of detention, leading to more circular migration (Durand, Massey, & Zenteno, 2000). Without further analysis, it is unclear which of these two effects is stronger. More stringent border enforcement may also affect trip duration (Kossoudji, 1992; Massey & Espinosa, 1997; Orrenius, 2001; Cornelius, 1998 & 2002; Marcelli & Cornelius, 2001; Reyes, Johnson, Sweringen, 2002; Massey, Durand, & Nolan, 2002). As the difficulty and cost of crossing increases, the risk of not being able to reenter the United States also increases. This risk may encourage immigrant to remain in the United States rather than move back-and-forth, making their migration more permanent (Massey & Liang, 1989). Also with higher migration cost, it takes longer for immigrants to achieve their targeted level of income increasing the amount of time they have to stay in the United States. In this paper, I explore the effects of temporal changes in the characteristics of the migrant sample, networks or resources accumulation, changes in macro-economic conditions in Mexico and the United States, and U.S. immigration policy on the time-trend of trip duration in the United States. Data Source For this analysis, I used a sample of 71 communities generated by the MMP (for details on the sample, see Massey and Singer, 1995; and Lindstrom and Massey, 1994). Between two and five communities were surveyed in successive years to gather detailed information on each household s members, independent of their current place of residence. With these data, I can examine return rates 2

3 for 1970 until 1998 and incorporate person, household and community characteristics to the model of return. Return in this context is not necessarily permanent move back to Mexico, some immigrants may eventually move back to the United States. A limitation of the MMP sample is is that it is not a representative sample of the Mexican population, but a representative sample of the relevant communities in Mexico. A few studies tested the representativeness of the sample, and find that the data is consistent with national samples, but since it over samples immigrant-sending towns, it has more international migrants than national samples (Durand, Massey and Zenteno, 2000; Zenteno and Massey, 1998). Many of the immigrants were interviewed in the western part of Mexico, long the most important source region for Mexican migration to the United States (Cross & Sandos, 1981; Durand, Massey & Zenteno, 2000). As other regions were added to the migration flow (Marcelli & Cornelius, 2001), the MMP expanded its sample to include new sending states, such as Veracruz, Oaxaca, Zacatecas, Puebla, and Baja California in the mid-1990s. For this analysis, I restricted the sample to the communities surveyed after 1990s so that all communities have data following IRCA. However, because new sending regions were added to the MMP in the mid-1990s, this could affect the patterns observed. For this reason, I examined new sending regions separately from other communities. There are therefore two sets of models: one for all the communities surveyed after 1990 and another for only the communities surveyed after 1994 (referred to as the restricted sample). I also limited the analysis to males older than 15 and looked at the years between 16 and 35 years old to determine the probability of return before age 35. Furthermore, I looked at trip duration separately for legal and illegal immigrants, as the effect of each of these factors may vary by immigration status. The MMP interviewed 12,322 households, generating a sample of 83,527 people, 15,645 of whom had lived in the United States. The sample interviewed after 1990s, including all gender and age restrictions, generates 11,241 trips to the United States between 1970 and Methodology I ran a set of hierarchical discrete-time-logit models to determine whether the time-trend observed for trip duration is due to changes in the characteristics of the Mexican migrant sample, to networks or resources accumulation, to economic conditions in Mexico or the United States, or to U.S. immigration policy. The model is hierarchical because it more accurately disentangles the effect of temporal changes in the sample, the family, or the community from changes in macro conditions and U.S. immigration policy. Hierarchical Discrete-time Logit Models This model consist of two-equations: a discrete-time logit model of the probability of return and controls for individual, household and community characteristics (Heckman and Singer, 1984; Kalbfleisch and Prentice, 1980; Lancaster, 1990); and an OLS equation modeling differences in duration over time while controlling for the conditions of the United States and the Mexican economies and U.S. immigration policy. The model is described below: Yit = (PERSON)it + (NETWORK)it + (TIME)it + δ 1 ( YEAR)it + it, (1) δ 1 = ά + ά 1 (MEX) t + ά 2 (US) t + ά 3(IMM_POL) t + µ t (2) Yit is the dependent variable and measures the log odds that person i will return to Mexico, given that t years have elapsed. 1 To see if temporal changes in the characteristics of the migrant population have led to changes in the time trend, this equation controls for personal, household and community characteristics that have been found to affect people s duration of stay in the United States (Massey & 1 Yit =1 if the person return in that particular year; otherwise Yit =0. 3

4 Espinosa, 1997; Lindstrom, 1997; Reyes, 2001). PERSONit are the characteristics of i, his family, his community and his migration experience at t; NETWORKit captures the effect of family and community networks on the probability of return. TIMEit captures the effect of length of stay in the United States on the probability of return and YEARit are fixed effect for the calendar years (the years between 1970 and 1998), capturing the effect of temporal changes in socio-economic conditions, immigration policy, and other macro level factors that could lead to changes in return not control for by the other variables in model (1). Equation (2) then models the macro factors that could lead to the temporal changes captured by YEARit. MEXt measures the conditions of the Mexican economy at t; USt measures the conditions of the United States economy at t; and IMM_POLt captures U.S. immigration policy in terms of legal entries to the United States and border enforcement at year t (Table 1 described the variables used in the model). 0,..., 4, ά 0,..., ά 3 and δ 1 are the coefficients to be estimated and it and µt are the error terms. I assume that it follows a logistic distribution and µt follows a normal distribution. The model is specified per time unit that each individual was observed. < Table 1 here > To capture differences in the time trend, I first looked at the probability of return without any controls, Yit = 0 + (TIME)it + δ 1 ( YEAR)it + it, (3) and concentrate on the effect of YEARit, δ 1, on the probability of return, which captures the overall pattern of return. Next, I ran model (1). If most of the temporal trend observed disappears once we control for all the factors in (1), this would mean that most of the changes in duration could be explained by changes in the characteristics of the sample or networks and resources accumulation. Nevertheless, if differences over time persist, then other factors are at play. I then looked at the effect of macro conditions in the United States and the Mexican economies and U.S. immigration policy on the remaining time-trend. To incorporate these factors, I used the coefficient δ 1 generated in (1) as a dependent variable and run equation (2). This estimation technique is employed because incorporating the macro factors in the discrete-time logit model will lead to underestimated standard errors, making it almost impossible to reject the null that the variables in (2) have an effect on return. 2 Empirical Analysis Figure 3 shows the probability of a return to Mexico within the first year of migration for male migrants in the MMP sample. Some of the year-to-year changes are not statistically significant, but the overall time-trend is significant. Corresponding to the pattern described in Figures 1 and 2, return probabilities were higher in the post-irca periods, , than in the early 1990s, , and they were even lower in the later part of the 1990s. However, the variation in the probability of return over time has been more dramatic for legal immigrants than for unauthorized immigrants. After the passage of IRCA, return rates for legal immigrants increased dramatically (Durant, Massey, and Zenteno (2000) obtained similar findings). In 1986, only 13 percent of legal migrants returned within a year of migration; but by 1992, 42 percent returned within a year of migration. After 1992, however, the return rate of legal migrants began to decline, reaching 18 percent in < Figure 3 Here > Like legal migrants, unauthorized immigrants were more likely to return within a year of migration after IRCA than before. But return rates declined in the 1990s, reaching one of its lowest levels in the late 1990s. Cornelius (1998), Marcelli and Cornelius (2001) and Massey, Durand, and Nolan (2002) also found a decline in return rates in the 1990s. By 1997, only 15 percent of unauthorized immigrants returned within one year of migration. Although less dramatic than for legal immigrants, a 2 Furthermore, my estimates of the time-trend suffer from heteroskedasticity. I corrected the heteroskedasticity by weighting the second stage OLS equation by 1/sqr (N), where N is the number of observations available for each year. 4

5 small decline in the probability of return of unauthorized immigrants represents a large increase in the number of illegal immigrants in the United States. For instance, if a million unauthorized male migrant enter the United States every year, 280,000 of those who entered the United States in 1986 would still be in the United States after four years, compared to 414,000 of those who entered the United States in Temporal Changes in the Sample As discussed above, this pattern could be due to temporal changes in the characteristics of the sample. Table 2 shows changes in immigrant characteristics over time. The MMP data suggest changes in the characteristics of immigrants. The average education of immigrants has increased, a smaller proportion of them are household heads, fewer immigrants are moving to California, and more originate from households with more resources and migration experience. In addition, 15 percent of those who entered the United States illegally in the late 1990s changed their status, compared to only 6 percent of those who moved before IRCA. I also ran the same descriptives for the sample of communities surveyed after 1994 to determine whether these changes were due to a re-sampling in the MMP or if they also took place in communities surveyed after 1994 and I found changes in the characteristics of the sample even among those communities. < Table 3 Here > Table 3 shows some of the results of the discrete-time logit model. The table shows the odd ratios for each variable. A value greater than 1 means that the person with that characteristic has a greater probability of return than his counterpart. For example, an undocumented immigrant who owns a home in Mexico is more likely to return to Mexico than one without a home in Mexico, holding constant for other factors. The results are consistent with prior work on this subject (Massey & Espinosa, 1997; Lindstrom, 1997; Reyes, 2001). There is some variation by immigration status, but in general, household heads, wealthier immigrants (as measured by owning land and a home in Mexico), agricultural workers, and people with no connections in the United States are more likely to return than their counterparts are. Figure 4a and 4b show what happens to the time-trend once I control for all the factors in the model. Even after taking into account all the factors in the model, there is an increase in the probability of return after the passage of IRCA followed by a decline in the probability of return in the 1990s. However, the decline in the 1990s is not as dramatic as that shown in Figure 3, which suggests that some of the change in the probability of return could be due to changes in the characteristics of the sample. The changes in the characteristics of the sample described in Table 2 decrease the probability of return in the 1990s. For example, as more legal family members are added to the unauthorized immigrant sample and more non-household heads move to the United States to join other family members, duration of stay increases. < Figure 4a and 4b Here > Network and Resource Accumulation Also important is the effect of networks in the probability of return. Table 4 shows the changes in family networks, as measured by having another migrants in the household and the immigration status of other household members. The proportion of migrants who originate from a household with other migrants in the household, or who have another family members in the United States, increased over time. So did the proportion of migrants with legal migrants in their household. In the late 1990s, 51 percent of unauthorized immigrants had a legal migrant in the household, either because a greater proportion of Mexican households had legal immigrants or because a greater proportion of migrants originated from household with legal immigrants. Table 3 showed that holding constant for other factors, having another household member in the migration stream decreases the probability of return, especially if the other household member is a legal migrant. Figure 5 shows the time trend in the probability of return for documented immigrants depending on the migration experience of other household members and their legal status. These results are generated from a model that interacts the migration experience and legal status of other household members with the year dummies, capturing differences in the patterns of return for legal 5

6 immigrants with and without migration networks (the model is available upon request from the author). The top line shows the time trend for legal immigrants without other migrants in the household, as compared to legal immigrants with other migrants in the household. Having another family member with migration experience decreases the probability of return, even if the other household member is not legally in the United States. The legal status of the other family member did not have an independent effect on the probability of return until the 1990s, making legal migrants with other legal migrants in the households less likely to return to Mexico than others with non-legal migrants in their household. As argued by Massey & Espinosa (1997), as males are legalized, their wives become more insistent on legalizing other family members and moving the whole family to the United States. Nevertheless, this pattern became more important during the period of increasing anti-immigrant rhetoric in the 1990s and the largest enforcement build-up in United States history. Legalizing the family may have become more important for immigrants after the enforcement build-up, and the families may have then shifted from temporary to permanent migration, to protect themselves and their families against a growing anti-immigrant sentiment in the country (Massey & Liang, 1989). Although I presented only the results for legal migrants, the results are similar for unauthorized immigrants. However, there is less of a difference in the time trend by the immigration status of other household members, and the decline in the late 1990s is not as dramatic. < Figure 5 Here > Another indication of the importance of networks is the effect of the community of origin on the probability of return. As shown in Figure 4a and 4b, most of the IRCA effect was concentrated in the major sending regions, which had well established networks in the United States. When I restrict the sample to communities surveyed after 1994 (bottom line in Figures 4a and 4b), which over-represents new sending regions, return probabilities follow a consistent pattern of increase, but no change is observed after the passage of IRCA. This could be because most of the people legalized through IRCA were from the western part of Mexico or because they possess the networks to bring the information about U.S. immigration policy to the home community. They may be able to mobilize resources quickly to help people cross the U.S.-Mexico border and benefit from the legalization as well as respond to border controls. Macro-Economic Condition and U.S. Immigration Policy Figures 4a and 4b show that even after controlling for temporal changes in the characteristics of the sample and network and resource accumulation, most of the time trend in trip duration remains unexplained. Next, I estimated equation (2) to explore the effect of economic conditions in Mexico and the United States as well as changes in U.S. immigration policy on the probability of return. This second stage model has its limitation. I only have 28 years of data; hence, if a particular variable has a small effect on the probability of return, it may not be statistically significant because the sample size is so small. We can only capture the effect of those variables that by the time of the survey had a large effect on the probability of return. The first model examines the effect of conditions in Mexico and the United States without controlling for temporal changes in the sample, in other words, before controlling for the factors in model (1). The second model examines the effect of conditions in the United States and Mexico on the probability of return after controlling for all the factors in model (1). The last model shows the effect of economic conditions and immigration policy on the restricted sample of communities surveyed after < Table 4 Here > The only economic variable that seems to have a consistent effect on the probability of return for male migrants is the exchange rate -- the higher the exchange rate, the higher the probability of return. The more pesos the immigrant is able to exchange for his dollars in Mexico, the more value he gets for his savings. Massey & Espinosa (1997) discuss a couple that sold everything in the United States during the devaluation in the mid-1990s in Mexico because they were able to increase the value of their savings and make a better living in Mexico. Of all macro factors, U.S. immigration policy appears to have the strongest effect on the probability of return of Mexican male migrants. IRCA led to an increase in the number of legal and illegal immi- 6

7 grants entering the United States (Johnson, 1996; Warren, 2000), but it also led to an increase in the probability of return. 3 In 1989, 450,000 more immigrants were granted legal permanent status than in 1988; this model predicts that this increase alone would have led to an increase in the proportion of immigrants returning within the first year of migration: from 15 to 28 percent increase for legal immigrants and from 18 to 26 percent increase for illegal immigrants. This increase in return may have been due to an increase in circular migration because the legalization allowed immigrant to move back-and-forth between Mexico and the United States without fear of detention (Durand, Massey & Zenteno, 2000). In addition, people may have crossed the U.S.- Mexico border temporarily to visit their newly legalized family members or to try to legalize their status, especially with the Special Agricultural Workers (SAW) provision of IRCA. This would explain the unexpectedly high number of immigrants legalized under IRCA, as hypothesized by other authors (Martin, 1994). In all the models, the build-up at the U.S.-Mexican border has a negative effect on the probability of return, indicating that increases in the number of man-hours at the border increases duration of stay in the United States. However, it is only statistically significant for legal immigrants and for unauthorized immigrants in the model without controls (Massey and Espinosa (1997) generated similar findings with fewer communities). An increase in the number of agents at the border increases legal immigrant trip duration. This may be because the border build-up makes it more important for legal immigrants to stay in the United States, to try to legalize other family members and bring them to the United States (Massey & Liang, 1989). For unauthorized immigrants, the effect may be small, or it may not have substantially affected duration for unauthorized immigrants as of But as shown in Figure 1 and 2, it is possible that the effect became stronger in the late 1990s. Conclusion The findings from the MMP sample indicate a dramatic increase in the probability of return after IRCA but a decline in the probability of return in the 1990s. These patterns corroborate those observed using the Mexican 2000 Census and the 1992 and 1997 ENADID, which show a decline in the probability of return, especially in the late 1990s. Changes in the characteristics of the sample partly explain the trend for unauthorized immigrants but other factors appear to be more important. This study reiterates the importance of social networks for migration. Only the communities in the western part of Mexico experienced a change in migration patterns during IRCA, perhaps because the networks in these communities provided the information and connections that influenced people s migration as a response to policy changes in the United States. Moreover, the presence of other migrants in the household had a strong effect on return probabilities. Immigrants from families with migration experience are less likely to return to the Mexico. They may have better knowledge and resources to do well in the United States, especially if the other migrant is legal. And as more households send people to the United States and more of them become legalized, the patterns of duration changes towards more settlement, especially in periods of strong border controls and anti-immigrant rhetoric. This study points to important unintended consequences of U.S. immigration policy. IRCA may have led to an increase in immigration, but many of those moving in the few years after IRCA were moving for short periods of time. In 1989, 20 percent of those who moved illegally and 35 percent of those who moved legally returned to Mexico within the first year after migration. However, this pattern changed in the 1990s, and the build-up may have been a part of the explanation for the change. The analysis of the MMP sample shows no statistically significant effect of the build-up on the probability of return of unauthorized immigrants and a negative and significant effect of the build-up on the probability of return of legal migrants. However, analysis of national data indicates that most of the decline in return probabilities took place after It is possible that with more data the MMP analysis would have uncovered a strong effect of border enforcement on duration of stay for unauthorized immigrants, as well as for legal immigrants. An increase in trip duration in the late 1990s is consistent with the unexpected high number of unauthorized immigrants found in the 2000 U.S. Census. 3 My proxy for the IRCA effect is the number of legal residents admitted to the United States in a particular year. 7

8 Bibliography: Alarcón, R. (1995). Immigrants or Transitional Workers? The Settlement Process Among Mexicans in Rural California, The California Institute for Rural Studies, Davis, California. (1995a). Transnational Communities, Regional Development, and the Future of Mexican Immigration, Berkeley Planning Journal, 10: Bean, F.D., E.E. Telles, and B.L. Lowel (1987). Undocumented Migration to the United States: Perspective and Evidence. Population and Development Review, 13: Berg, E.J. (1961). Backward-Sloping Labor Supply Functions in Dual Economies: The African Case, Quarterly Journal of Economics, 75: Calavita, Kitty (1994). U.S. Immigration and Policy Responses: The Limitations of Legislation, pp , in Cornelius, Wayne A., Philip L Martin and James F. Hollifield (eds.), Controlling Illegal Immigration: A Global Perspective, Stanford California: Stanford University Press. Cornelius, W.A. (1976). Outmigration from Rural Mexican Communities. Interdisciplinary Communication Program Occasional Monograph Series 5, no. 2:1-39. Washington, D.C.: Smithsonian Institute. (1998). The Structure Embeddedness of Demand for Mexican Immigrant Labor: New evidence from California, pp , in Marcelo Suarez-Orozco (ed.) Crossings: Mexican Immigration in Interdisciplinary Perspective. Cambridge, MA: Harvard University Press/David Rockefeller Center for Latin American Studies. (2002) Death at the Border: Unintended Consequences of U.S. Immigration Control Policy, Population and Development Review, Vol. 27 (4): Cross, H.E. and J.A. Sandos (1981). Across the Border: Rural Development in Mexico and Recent Migration to the United States, Institute of Governmental Studies, University of California, Berkeley. Djajic, S. and R. Milbourne (1988). A General Equilibrium Model of Guess-worker Migration: A Source-Country Perspective, Journal of International Economics, 25: Durand, Jorge, and Douglas Massey (1992). Mexican Migration to the United States: A Critical Review, Latin American Research Review 27: Douglas S. Massey and Rene M. Zenteno (2000). Mexican Immigration to the United States: Continuities and Change, Latin American Research Review 34: Heckman, James and Burton Singer (1984). Econometric Duration Analysis, Journal of Econometrics 24: Hill, J.K. (1987). Immigrant Decisions Concerning Duration of Stay and Migratory Frequency, Journal of Development Economics, 25: Hugo, G.J. (1982). Circular Migration in Indonesia, Population and Development Review, 8(1): Jenkins, J.C. (1977). Push-Pull in Recent Mexican Migration to the U.S., International Migration Review, 11: Johnson, Hans P. (1996). Undocumented Immigration to California: San Francisco, CA: Public Policy Institute of California. Jones, R.C. (1982). Undocumented Migration from Mexico: Some Geographic Questions, Annals of the Association of American Geographers, 72: Kalbfleinsch, J.D. and R.L Prentice (1980). The Statistical Analysis of Failure Time Data. New York: Wiley and Sons. Kossoudji, Sherrie A. (1992). Playing Cat and Mouse at the U.S.-Mexican Border, Demography, 29(2): Lancaster, Tony (1990). The Econometric Analysis of Transition Data: Cambridge University Press Economics, UC San Diego, updated. 8

9 Lindstrom, D.P (1996). Economic Opportunities in Mexico and Return Migration from the United States, Demography, 33(3): and D.S. Massey (1994). Selective Emigration, Cohort Quality, and Models of Immigrant Assimilation, Social Science Research, 23: López, Gustavo (1986). La Casa Dividida: Un estudio de Caso Sobre Migración a Estados Unidos en un Pueblo Michoacano, Zamora, Michoacan, México: Colegio de Michoacan. Marcelli, Enrico and Wayne Cornelius (2001). The Changing profile of Mexican migrants to the United States: New evidence from California and Mexico, Latin American Research Review 36(3): Martin, Philip L. (1994). Good Intentions Gone Awry: IRCA and U.S. Agriculture, The Annals of the American Academy, 534: Massey, Douglas. (1987). Understanding Mexican Migration to the United States, American Journal of Sociology 92: (1990). Social Structure, Household Strategy, and the Cumulative Causation of Migration, Population Index, Vol. 56: 3-26., R. Alarcón, J. Durand, and H. Gonzalez (1987). Return to Aztlan: the Social Process of International Migration from Western Mexico. Berkeley and Los Angeles: University of California Press. and Audrey Singer. (1995). New Estimates of Undocumented Mexican Migration and the Probability of Apprehension, Demography, 32: , J. Durand, Jorge Durand and Nolan J. Malone (2002). Beyond Smoke and Mirrors: Mexican Immigration in an Era of Economic Integration, New York: Russell Sage Foundation. and Kristin Espinosa (1997). What s Driving Mexico-U.S. Migration? A Theoretical, Empirical and Policy Analysis, American Journal of Sociology, 102: Luin P. Goldring and Jorge Durand (1994). Continuities in Transnational Migration: An Analysis of 19 Communities, American Journal of Sociology, 99: and Zai Lian. (1989). The Long-term Consequences of a Temporary Worker Program: The US Bracero Experience, Population Research and Policy Review 8: Mines, Richard (1981). Developing a Community Tradition of Migration, Monograph. University of California, San Diego, La Jolla: Program in United States Mexico Studies. and A. de Janvry (1982). Migration to the United States and Mexican Rural Development: A Case Study, American Journal of Agricultural Economics, 64: Orrenius, Pia M. (2001). Does Increased Border Enforcement Trap Illegal Immigrants inside the United States? Working Paper. Piore, M.J. (1979). Birds of Passage: Migrant Workers and Industrial Society, New York: Cambridge University Press. Ranney, S.I and S.A. Kossoudji (1983). Profile of Temporary Mexican Labor Migrants to the U.S., Population and Development Review, 9: Reichert, Josh S. and Douglas Massey (1979). Patterns of U.S. Migration from a Mexican Sending Community: A Comparison of Legal and Illegal, International Migration Review 13(4): Reyes, Belinda I. (2001). Immigrant Trip Duration: The Case of Immigrants from Western Mexico, International Migration Review, 35 (4): , Hans P. Johnson and Richard Van Sweringer (2002). Holding the Line? The Effect of the Recent Border Build-up on Unauthorized Immigration, San Francisco, California, Public Policy Institute of California Report. Roberts, Barry R. (1995). Socially Expected Durations and the Economic Adjustment of Immigrants, pp. **-**, in Alejandro Portes (ed.), The Economic Sociology of Immigration: Essays on Networks, Ethnicity and Entrepreneurship, New York: Russell Sage Foundation. 9

10 Warren, Robert (2000). Annual Estimates of the Unauthorized Population Residing in the United States and Components of Changes: 1987 to 1997, Washington, DC: US Immigration and Naturalization Service. White, M.J., F.D. Bean, and T.J. Espenshade. (1990). The U.S Immigration Reform and Control Act and Undocumented Migration to the United States, Population Research and Policy Review 9: Zenteno, Rene and Douglas S. Massey. (1998). Especificidad versus Representatividad: Enfoques Metodológicos para el Estudio de la Migración Internacional, Estudios Demográficos y Urbanos 40:

11 Figure 1: Return Migration Rates 60% 50% 40% 54% 42% 30% 25% 20% 10% 0% Source: 1992 and 1997 ENADID and Mexican 2000 Census Note: These proportions have been standardized by years since departure (within the past five years). Non-standardized patterns are similar, with the proportion of migrants having returned to Mexico at 46 percent for the period, 35 percent for the period, and 23 percent for the period. 11

12 Figure 2: Average Length of Stay in the U.S. Months Mean Median Source: Mexican 2000 Census, 1992 and 1997 ENADID 12

13 Table 1: Definition of the Variables Used in the Model of the Probability of Return Variable Description Duration Variables One Five Dummy variables for the number of years in the U.S.(more than 5 years in the U.S. is the left years out category). Personal, Household and Community Characteristics Age age at year t Agesq Square of age Edyrsi Educational attainment at year t Edyrsi2 The square of education Land =1if the family owns land at year t; otherwise 0 Ownhome = 1 if the family own their home at year t; otherwise is 0 Small size If the population at the community of origin was less than 5,000 people at year t (left out category) Medium size If the population in the community of origin was between 5,000 and 50,000 people at year t Large size If the population at the community of origin was larger than 50,000 people at year t Mexican Dummies for Mexican states:guanajuato, Nayarit, Jalisco, Zacatecas, Guerrero, San Luis States Potosi, Colima, Sinaloa, Oaxaca, Puebla, Aguascalientes (left out Michoacan) Men in ag The proportion of men in the community s of origin who work in agriculture Migration Experience Agriculture = 1 if working in agriculture while in the US; otherwise 0 (left out category) Skilled = 1 if skilled worker while in the US; otherwise 0 Unskilled = 1 if unskilled worker while in the US; otherwise 0 Unemployed = 1 if unemployed while in the US; otherwise 0 US Destination states (left out category is Los Angeles) Dummies for US destinations: Los Angeles, other California, Texas, Illinois, and Other US Change status=1 if the person changed status before or at year t; otherwise is 0 Legal After =1 if the person was legalized after IRCA; otherwise is 0 IRCA Trips Number of trips made to the U.S. Household Networks Family in US = 1 if a family member in U.S. at t or has been in the U.S. within t-10 years; otherwise is 0 Legal = 1 if someone in the family was legalized before t; otherwise is 0 Year Dummy variable for years between 1970 and 1998 (left out 1989) Macro Variables for Second Stage Model Mexican This is the GDP per capita for a particular year and its squared term GDP Exchange Official exchange rate Pesos per dollars Rate US Unemp. Annual unemployment rate in the United States Rate Legal Admissions Total number of legal admission to the U.S. in a particular year and its squared Line watch Total number of hours spent by the INS guarding the U.S.-Mexico border and its squared hours 13

14 Figure 3: Proportion of Immigrants who Returned to Mexico Within the First Year of Migration 50% 45% unauthorized legal 40% 35% 30% 25% 20% 15% 10% 5% 0% Source: Author s estimates from the MMP data. Results presented in 3 years moving averages. 14

15 Table 2: Mean Value of the Independent Variables Variables Unauthorized Documented Before IRCA After IRCA Late90s BF IRCA ( ) AF IRCA ( ) LATE90s ( ) BF IRCA ( ) AF IRCA ( ) LATE90s ( ) Immigrants Characteristics and Experience Age Education Head Skilled Unskilled Unemployed Unknown Occ Trip Change status Other_CA Texas Illinois Other_state Source: Author s estimates from the MMP data. 15

16 Table 3: Odd Ratios and Standard Errors Discrete-Time Logit Model Documented Undocumented Odd Ratio Standard Errors Odd Ratio Standard Errors Personal and Household Characteristics Age 1.0* *** Age *** Edyrs 1.0*** * Edyrs Head 1.65*** *** Land *** Ownhome *** Migration Experience Legal after IRCA 0.27*** 0.1 Change status Trip Skilled 0.36*** *** Unskilled 0.93*** *** Unemployed *** Other_CA 0.78** Texas *** Illinois Other_state Household Networks Family in US 0.49*** *** before Legal Family 0.71*** Community of Origin Characteristics Men in ag *** Medium size 0.66*** Large size 0.36*** *** ***p<.001 ** p<.05 * p<.10 16

17 Table 4: Mean Value of Family Networks Variables Unauthorized Documented BF IRCA ( ) AF IRCA ( ) LATE90s ( ) BF IRCA ( ) AF IRCA ( ) LATE90s ( ) Household s Networks In US Before In US Legal

18 Figure 4a: Simulation of the Probability of Return Within One Year of Migration Figure 4b: Simulation of the Probability of Return Within One Year of Migration 45% 45% 40% 40% no controls full model restricted sample Unauthorized Immigrants Documented Immigrants 35% 35% 30% 30% 25% 25% 20% 20% 15% 15% 10% 10% 5% 5% 0% 0% Source: Author s estimates from the MMP data. Results presented in 3 years moving average 18

19 Figure 5: Simulation of the Probability of Return Within One Year of Migration by Migration Experience of Other Household Members and Their Legal Status Documented Immigrants Source: Author s simulations from time-logit models using the MMP data. 45% 40% No other Migrant Illegal Family Member Legal Family Member 35% 30% 25% 20% 15% 10% 5% 0%

20 Table 5: Parameter Estimates and Standard Errors for Second Stage OLS Equation 4 Documented Exchange Rate.056 (.10) Conditions of the US Economy US Unemployment Rate (.09) -.09 US Immigration Policy.20*** (.10) -.05 (.09) No Controls With Controls Intercept (.84) (.83) Conditions of the Mexican Economy Mexico GDP per Capita (.0006) (.0006) Mexico GDP Squared (.0009) (.0006) Restricted Sample 3.4* (.99) (.0006).0009 (.0009).27** (.11).11 (.10) Undocumented No Controls.53 (.48).0003 (.0003) (.0005).027 (.06) -.06 (.05) With Controls -1.03* (.48).0003 (.0003) (.0005).08 (.06) -.05 (.05) Restrict Sample -.91 (.54) (.0003) (.0005) -.05 (.06) -.05 (.06) Legal Admissions.58** (.21).39*** (.21).02 (.23).23*** (.13).30** (.13).17 (.13) Legal Adm. -.02** ** Squares (.009) (.008) (.01) (.005) (.005) (.005) Line Watch -1.81* -1.91* -1.57** -.78** Hours (.62) (.61) (.67) (.36) (.36) (.38) Line Watch.19**.189**.16** Squared (.07) (.07) (.07) (.04) (.04) (.04) R 2 67% 72% 49% 53% 67% 38% Adjusted R 53% 61% 29% 34% 54% 13% Notes: *** significant at a 1% level, ** significant at a 5% level, * significant at a 10% level. Referencia electrónica: 4 During the period of analysis, one average, the GDP per capita in Mexico was $2,643 pesos per person, the Mexican economy grow by 4 percent, 700,000 people were granted legally permanent status to the United States, the unemployment rate in the United States was 6.7 percent, and the INS spent on average 2.6 million men-hours guarding the U.S.-Mexican border. The GDP is normalized to $100 pesos, the legal admissions to 100,000 admissions and the number or hours are normalized at 1,000,000 hours. 20