Evolution of the Mexico-U.S. Migration System: Insights from the Mexican Migration Project

Abstract

Since 1987 the Mexican Migration Project has collected and disseminated representative survey data on documented and undocumented migration to the United States. The MMP currently includes surveys of 161 communities, which together contain data on 27,113 households and 169,945 individuals, 26,446 of whom have U.S. migratory experience. These data are here used to trace the evolution of the Mexico-U.S. migration system from the late 19^th to the early 21^st century, revealing how shifts in U.S. immigration and border policies have been critical to the formation of different eras of migration characterized by distinctive patterns of migration, settlement, and return in different legal statuses. The present era is characterized by the repression of the large population of undocumented migrants and their U.S. citizen children by an ongoing regime of mass detention and deportation and the simultaneous recruitment of Mexican workers for exploitation on short term temporary visas. The future of Mexican migration to the United States will be revealed by subsequent waves of data collection by the Mexican Migration Project.

In 1982 we didn’t know we were laying the foundations for what would become the Mexican Migration Project. We were simply launching a pilot study to test a new approach for gathering data on Mexico-U.S. migration. Movement between the two countries was nothing new, of course. It had been going on for most of the 20^th century. After 1965, however, migration from Mexico increasingly occurred outside of legal channels, and thus was not captured by the usual statistical systems. The lack of information on Mexican migration had become a serious problem, given that “illegal migration” was a topic of keen interest to scholars, policy makers, and the public.

The pilot ultimately proved to be successful, yielding a series of refereed journal articles and a 1987 book describing in some detail the history, patterns, and processes of U.S. migration from four communities in West-Central Mexico. This success provided a proof-of-concept for the new data collection approach we advocated, and in 1987 we sought and received additional funding to apply the approach more widely. Our plan was to gather data in 4–6 Mexican communities each year over a period of five years in order to begin building up a database sufficient to study the evolution of the Mexico-U.S. migration system as it unfolded in real time. Thus was born the Mexican Migration Project, which has been funded continuously and has collected new data each year since the first award in 1987.

The MMP database has always been publicly available. At first, the data were disseminated to users on floppy disks, then on diskettes, then using email attachments, and finally via the project’s webpage, which currently has 5,012 registered data users. Over the years, the database has provided the foundation for some 46 books, 90 articles and book chapters, and 74 doctoral dissertations, though of course these counts are always changing. We ourselves have authored or coauthored many of these publications. Although our success in research and publication has been rewarding, it has also been painful as it enabled us to document in great detail deleterious consequences of U.S. immigration and border policies and the hardships they inflicted on millions of people beginning in 1986.

Here we outline the origins and evolution of the Mexican Migration Project as a binational research effort. We then draw upon MMP database to demonstrate how U.S. immigration and border policies have crucially shaped the evolution of the migratory system in very counterproductive ways. We conclude by reviewing research that documents the accumulating social and economic costs of U.S. policies and the needless human suffering caused by the current impasse on immigration reform.

Origins of the Mexican Migration Project

The new approach to data collection that we sought to test in 1982 combined ethnographic methods from anthropology with survey methods from sociology to create a hybrid design we labeled the “ethnosurvey” (Massey 1987). From survey research we borrowed the tools of random sampling, standardized data, and quantification, and from ethnography we took the techniques of in-depth personal interviewing, participant observation, and qualitative analysis. In an ethnosurvey, instead of reading verbatim through a scripted questionnaire of pre-determined queries and fixed responses, ethnographically-trained interviewers endeavor to undertake a naturalistic conversation with respondents to learn about the characteristics of the household they inhabit and the traits, actions, and behaviors of its members.

Although natural, the conversations are by no means free-form. Rather, they are guided by a series of topic-specific tables containing blank cells that are gradually filled in by the interviewer during course of a structured conversation. Unlike standard ethnographies, however, households in the MMP are randomly selected to represent specific communities of interest; and as the ethnosurvey proceeds, researchers simultaneously compile additional information from field observations, local archives, key informants, and other sources. Once fieldwork has been completed, investigators gather additional data from national and international statistical sources. The communities to be surveyed are strategically chosen to cover different segments of Mexico’s social, economic, and geographic structure.

The original ethnosurvey instrument was designed collaboratively at the Colegio de Michoacán in the summer of 1982 by a four-member research team. It was first applied during the winter of 1982–1983 in four Mexican communities: a traditional agrarian village, a more commercial agricultural town, an industrial town, and an urban working class neighborhood in Guadalajara. Within each community, 200 households were randomly selected and household heads were interviewed, and while interviewing proceeded qualitative fieldwork was undertaken. These data gathering activities quickly revealed where in the United States migrants from each community went, and during the summer of 1983 fieldworkers traveled to these U.S. destinations to survey 20 settled out-migrant households originating from each community.

Quantitative data from the ethnosurvey’s data tables were entered into the computer during the spring and summer of 1983 to create a series of four interlinked files. A person-level file contained basic social, economic, and demographic data on the household head, the spouse, and all children of the head (flagging those offspring no longer living in the household), as well as any additional people present at the time of the interview. In addition to information about each person’s age, sex, education, etc., the file included data about the first and last trips to the United States and first and last trips taken internally within Mexico. For each trip, interviewers recorded the destination, duration, occupation, and wages earned, and migrants to the United States also reported their legal status.

The second file was a household-level file that contained basic information on the socio-demographic composition of the household, the current dwelling and its contents, and key assets it owned (e.g. land, homes, businesses, autos, consumer goods, etc.). The final two files focused on the life experiences of household heads. A life history file compiled a year-by-year history for each head, from the point of entry into the labor force to the survey date. This account included complete histories of marriage, childbearing, employment, property ownership, migration, and border crossing. In addition, a focused migration file compiled information about social and economic aspects of the head’s most recent trip to the United States, including extensive information about jobs, earnings, spending, savings, and remittances as well as detailed data about the head’s connections to various people and institutions in the United States.

The foregoing data files provided the basis for a series of publications culminating in the book, Return to Aztlan, which offered a comprehensive comparative analysis of U.S. migration from the four communities sampled, detailing the role that migration played in the lives of community members at home and abroad (Massey et al. 1987: 285–286). Based on this analysis, we concluded that migration originated in structural transformations that altered the relations of production in sending and receiving societies, which simultaneously displaced workers in Mexico while creating a structural demand for low-skilled workers in the United States, a demand that was often satisfied by deliberate labor recruitment. Once migration had begun, social networks quickly emerged to make U.S. jobs ever more accessible throughout the community. Over time, the deployment of family members abroad for periods of wage was incorporated into household survival strategies and used tactically at different points in the life cycle to finance particular consumption and investment goals.

The growing reliance of individuals and households on U.S. wage labor in turn altered individual motivations, household incentives, and community structures in ways that fostered additional migration, yielding a process of cumulative causation. As migratory flows matured over time, some individuals and families inevitably acquired social and economic ties in the United States that promoted longer-term settlement. The relative number of settled migrants generally was not very large, however, and migrant networks continued to be sustained by the regular movement of people back and forth across the border, and at times even by the return home of formerly “settled” migrants.

Evolution of the Mexican Migration Project

Given the success of the pilot project and the apparent utility of the information it generated, we sought to institutionalize data collection over the longer term, in hopes of being able to provide other researchers with a valid and reliable source of data on documented and undocumented migration to the United States. Specifically, using the approach we tested and perfected on the pilot, we proposed annually to survey 4–6 Mexican communities and their U.S. branch settlements into the indefinite future. Over time, we would build a database of detailed information on Mexico-U.S. migration gathered from a large and diverse set origin communities located throughout Mexico and their branch settlements in the United States, moving well beyond the four original pilot communities.

Three of the original communities were located in the state of Jalisco and one was in Michoacán, two of the five states that comprised the historical heartland for migration to the United States. Moving forward, we initially concentrated on surveying communities in the other three heartland states (Guanajuato, San Luís Potosí, and Zacatecas) while continuing to add samples from other locations in Jalisco and Michoacán. We also endeavored to fill out the database with surveys of communities in smaller states in the heartland region in western Mexico (Aguascalientes, Colima, and Nayarit). As the geography of migration changed over time, however, we strove to broaden the sample by surveying communities new sending states located in Mexico’s south (Guerrero, Morelos, and Oaxaca), north (Baja California, Chihuahua, Durango, Nuevo Leon, and Sinaloa) and its central region (Puebla, Mexico, Tlaxcala, Querétaro), as well as communities in states along Mexico’s gulf coast (Tabasco, Veracruz, Yucatan).

At this writing, the MMP database includes surveys of 161 communities, which together contain data on 27,113 households and 169,945 individuals, 26,446 of whom have U.S. migratory experience. Over the years, additional files have been added to the database improve coverage and support a fuller research agenda. In 1993 we added a spouse life history file into the database to supplement the file that had long been compiled for heads. In 2005 we encountered the first migrants to Canada and added a Canadian migrant file to parallel the original U.S. migrant file. In 2007 we created an other migrant file, which compiled detailed information about the latest U.S. trip made by a household member other than the head in cases where the head was not a U.S. migrant.

Once a sufficient number of communities had been included in the database, we began to develop a set of contextual files to assess the milieu within which migratory decisions were being made. Thus we created a community level file containing basic information on the characteristics of each community from 1950 to the survey date, which we developed using data from the Mexican Census and the MMP’s own Community Data Questionnaire. This file was recently modified to include community homicide rates from 1990 to the present, obtained from Mexico’s National Institute of Statistics and Geography. A related migration prevalence file provides annual estimates of the relative number of U.S. migrants in each community at different points in time using a prevalence ratio developed by Massey, Goldring, and Durand (1994). Their measure used retrospective data to compute this ratio for each year moving backward from the survey date. The prevalence ratio is defined as the number of people with U.S. migratory experience divided by the total number of people alive in a given year of observation.

We later also constructed an environmental file at the state level to assess the ecological circumstances community members faced at various points in time. The file includes indicators of climate, land usage, environmental degradation, rainfall, trash production, and other measures. To summarize macroeconomic conditions prevailing in the United States and Mexico from 1965 to the present, we developed a national level file that includes indicators pertaining to immigration and border enforcement, Mexico’s economy and finances, binational trade with the United States, and conditions in the binational labor market.

Finally, a national historical file contains annual data from 1900 to the present on the number of legal Mexican immigrants entering the United States, the Mexican-born population of the United States, the number of Mexican naturalizations, legal border crossings, border apprehensions, the size and budget of the U.S. Border Patrol, the population of Mexico and the United States, and the balance of trade between the two nations. Although the various contextual files were added to the database at different points in time, we constructed files for communities surveyed at earlier points in time to include all communities in the contextual files.

In order to maintain continuity in the various data files, items are generally not dropped from the ethnosurvey instrument. Over the years, however, new items and modules have periodically been added as new topics and circumstances have arisen. In 1998, for example, we added items on legalization in the wake of the Immigration Reform and Control Act (IRCA), which authorized an amnesty for long-term undocumented residents of the United States and offered a special legalization program for agricultural workers. In the same year, we also added items on the head’s income and wages in Mexico to improve measurement of the household’s economic situation. In 2000, we included new items on migrants’ marital status at the time of migration as well as items pertaining to the application for and receipt of U.S. citizenship. In 2007, we added an entire new table of items to gather information on the health of the household head and spouse. Finally, in the wake of President Obama’s announcement of Deferred Action for Child Arrivals, we added items about the application for and receipt of DACA.

All MMP data files are available for free download from the project website at http://mmp.opr.princeton.edu/, along with additional information about the project, its personnel, publications, documentation, data alerts, results, and links to qualitative datasets containing oral histories and votive paintings concerning migration to the United States. Several of the quantitative files are longitudinal and even the cross-sectional files contain time-specific information, enabling analysts to reconstruct what was happening each year in the lives of individuals in terms of the conditions prevailing in their households, communities, states, and the binational political economy.

The analytic potential of the MMP database is suggested by Figure 1, which plots the number of person years observed for household heads and spouses in different five-year periods from the late 19^th to the early 21^st centuries. Among household heads, a small number of person years (98) are observed as far back as the late 1800s and the entire life history file contains 1,325,336 person years of information. The number of observed person years begins to rise rapidly after 1910, accelerates from the 1920s through the 1950s, slows down during the 1960s and 1970s and peaks in 1980–1984 when 129,596 person-years are observed. Thereafter the curve drops rapidly as it nears the latest period of 2015–2016.

Figure 1.

Person years observed for household heads and spouses in the MMP161 database

Even in this latter period, however, we have 3,624 person years of information available for analysis. Because the compilation of life histories for spouses began after that of household heads, spousal person years are first observed only in 1905–1909 and the curve rises more slowly and reaches its maximum at a later date (1990–1994) and at a lower level of just 85,151 person years. Nonetheless, even in the latest period we have 2,825 person years to work with, and in total the spouse file contains 768,887 person years of data.

In addition to observing specific person years in the lives of household heads and spouses, the MMP data also enable us to observe people on their first and last trips to the United States. As shown in Figure 2, the very first trip to the United States was recorded in the period 1905–1909, indicating the deep historical roots of the Mexico-U.S. migration system. The number of first trips remains small from 1910 through the 1930s, in large measure because most migrants from that era had died by the survey date and could not report their trips to MMP interviewers. Beginning in the 1940s, however, and continuing through the mid-1960s the curve rises very rapidly, peaking at 4,653 in 1985–1989 before declining to a mere 14 in the final period. The series for last U.S. trips begins in 1910–1914 and the number remains low through the early 1940s. Thereafter the curve rises at a modest pace through the late 1960s and then increases exponentially to peak at 5,428 trips in 1990–1994 before falling rapidly downward to 48 in 2015–2016.

Figure 2.

First and lates U.S. trips observed in the MMP161 database

Evolution of the Mexico-U.S. Migration System

The foregoing data files can be used to trace the evolution of the Mexico-U.S. migration system from the early 20^th century to the present time. In Return to Aztlan, we described the workings of a relatively stable and benign migration system based on the circulation of unauthorized workers, one in which well-developed interpersonal networks connected zones of labor surplus in Mexico to areas of labor scarcity in the United States, thus facilitating the circular movement of mostly male workers back and forth for seasonal wage labor. At that time, migrants generally sought to earn money in the United States to finance projects of consumer spending and productive investment at home. This migration system evolved over many years, changing in response to specific events occurring on both sides of the border.

The effect of different historical events on the likelihood of migration is clearly registered in the MMP’s person-year data file. Figure 3 draws on these data to show the estimated probability that household heads and spouses departed for the United States in years from 1900 through 2015. As seen in the figure, household heads first began to migrate northward in the period 1905–1907, when the 1907 “Gentlemen’s Agreement with Japan” curtailed migration from that country, causing labor shortages to emerge throughout the southwest, especially in agriculture. In response, private labor contractors followed rail lines southward into Mexico in search of replacement workers (Cardoso 1980; Durand and Arias 2000).

Figure 3.

Probability of U.S. migration for household heads and spouses in the MMP161 database

With the outbreak of the First World War in 1914 immigration from Europe came to a screeching halt and labor demand surged, not only in the southwest but also in industrial cities in the nation’s midsection, such as Chicago (Arias and Durand 2008). Thus we observe a clear uptick in heads’ migration probabilities during the period 1912–1916. Once the United States entered the war in 1917, labor demand intensified and the U.S. Congress responded by creating a government-sponsored temporary worker program for Mexicans (Reisler 1976) producing yet another jump in migration probabilities from 1917 to 1921. Although the likelihood of migration fell from 1921 to 1925 it surged once again in the late 1920s after immigration quotas imposed in 1921 and 1924 curtailed immigration from Europe during the tight labor markets of the “Roaring Twenties” (Cardoso 1980).

The onset of the Great Depression in 1929 led to the mass deportation of Mexicans, however, and the likelihood of U.S. migration remained low throughout the 1930s (Hoffman 1974). Although conditions in Mexico generally favored out-migration to the United States throughout the period from 1900 to 1940 (owing to poverty and landlessness before 1910, civil war from 1910 to 1920, and economic stagnation from 1920 to 1940) what really explained the ups and downs of Mexican migration during this period were policies and practices enacted in the United States, namely the initiation of private recruitment efforts circa 1907, followed by the creation of government-sponsored temporary worker programs a decade later, and culminating with deportation campaigns launched after 1929.

Despite the forced removal of Mexicans during the 1930s, as soon as the United States entered the Second World War, authorities approached Mexico to negotiate a binational treaty that arranged for the annual entry of legal workers for seasonal labor in U.S. agriculture (Galarza 1964; Calavita 1992). The resulting Bracero Program lasted from 1942 through 1964, and its effect on the likelihood of migration is readily apparent in Figure 3. Between 1940 and 1945 the probability of U.S. migration rose nearly seven times, going from 0.003 to 0.020 before leveling off briefly and then rising to new plateau of 0.029 from 1956 to 1959. In 1960 Congress began to phase out the Bracero Program, finally letting it expire at the end of 1964, bringing the probability of migration down to 0.017.

In the following year, Congress imposed the first-ever numerical limits on immigration from the Americas, setting a cap of 120,000 total visas for the Western Hemisphere beginning in 1968. Subsequent legislation tightened the limits and by 1976 all nations in the hemisphere were subject to an annual cap of just 20,000 visas. Conditions of labor supply and demand had not changed, however, and by 1965 many millions of Braceros had established ties to U.S. employers at traditional points of destination. As a result, Mexican migration did not stop when opportunities for legal entry evaporated (Massey and Pren 2012). Instead undocumented migration rose sharply, and the likelihood of a head’s departure climbed from 0.017 in 1966 to 0.040 in 1979.

Although the odds of departure dropped slightly during the height of Mexico’s oil boom in the early 1980s, it surged once again with the onset of the economic crisis, rising from 0.035 in 1982 to 0.050 in 1989. Thereafter it dropped steadily despite the tightening of U.S. labor markets during the boom of the 1990s, falling to 0.027 in 1997. Although the likelihood of migration ticked up slightly to reach 0.030 in 1999, thereafter it fell very sharply, reaching 0.008 in 2012, the lowest level since 1943. Although the migration probability rebounded somewhat between 1999 and 2016, it still remains well below the level observed back in 1943

Mexican migration to the United States historically has been led by male workers, with females generally following later on as husbands and fathers begin to spend longer durations north of the border (Massey and Cerrutti 2001). This pattern is clearly seen in the spousal migration probabilities, which remain well below those of household heads throughout the period. Prior to 1940 the probability of spousal migration hovered around zero, rising slightly during the 1940s, dropping back near zero during height of the Bracero Program, and then rising steadily during the 1970s and 1980s before reaching a peak of 0.005 in 1995. Since then the likelihood of migration has trended steadily downward, reaching zero in 2016. At present it is not entirely clear whether mass Mexican migration has finally come to an end, or whether it is simply entering a new phase, though even in many heartland communities the tradition of U.S. migration seems to have reached a state of collapse (Durand and Arias 2014).

The Eras of Mexican Migration

Looking back over more than three decades, we now realize that Return to Aztlan described migration patterns that prevailed at the very end of what became known as the undocumented era, which coincided with the high tide of Import Substitution Industrialization in Mexico modest enforcement efforts along Mexico-U.S. border (see Massey, Durand, and Malone 2002). Although we did not know it at the time, the migration system described in Return to Aztlan was about to change quite dramatically. In 1986 Mexico entered the General Agreement on Tariffs and Trade (GATT), opening its economy to international trade and investment and inaugurating a neoliberal era of state downsizing, privatization, and market-oriented reforms. In the same year the U.S. Congress enacted the Immigration Reform and Control Act (IRCA), which legalized two million of former undocumented migrants but also began a decades-long build-up of border enforcement while also criminalizing the hiring of undocumented migrants by U.S. employers.

The neoliberal wave crested in 1994, when Mexico joined Canada and the United States in the North American Free Trade Agreement (NAFTA), a treaty designed to promote the free cross-border movement of goods, capital, commodities, information, and services—but crucially not labor (Massey, Durand, and Malone 2002). To underscore its reluctance to integrate North American labor markets, in the ensuing years the United States steadily intensified border enforcement, rolling out Operation Blockade in El Paso in 1993 and Operation Gatekeeper in 1994 and continuing thereafter to build up the Border Patrol’s personnel and budget through legislation passed in 1990, 1996, 2001, and 2010 (Massey and Pren 2012).

These shifts in U.S. immigration and border policy had profound effects on patterns of movement between Mexico and the United States. As early as 1999, Durand, Massey, and Parrado (1999: 518) had heralded a “new era of Mexican migration to the United States” in which U.S. policies transformed a “formerly temporary and regionally concentrated flow of undocumented, predominantly male, and disproportionately rural workers into a settled, urbanized population of immigrant families dispersed throughout the nation.” In that same year, Phillips and Massey (1999) documented the existence of a “new labor market” where undocumented migrants experienced a significant earnings penalty that had not existed prior to 1986. Durand, Massey and Charvet (2000:1) subsequently described the “changing geography of Mexican immigration,” as migrant destinations shifted away from California, Texas, and Illinois to new destination states in the rest of the nation; and Durand, Massey, and Zenteno (2001) identified a new geography of emigration within Mexico, with the origins of migrants shifting out of the west-central region toward new sending communities in Mexico’s north, central and southern regions. Finally, a decade into the new millennium Massey and Brown (2011) detected the emergence of “new migration streams between Mexico and Canada.”

The foregoing shifts in patterns and processes of Mexico-U.S. migration were generally accompanied by changes in the legal composition of the migratory flows, which we have hinted at but not explicitly described up to this point. In Figure 4 we do so by examining the likelihood of departure by household heads in three statuses: documented, undocumented, and temporary. Documented migrants include persons holding a permanent resident visa or naturalized U.S. citizenship; undocumented migrants include those who lack a valid visa for work or residence in the United States and entered without authorization or violated the terms of a tourist visa; and temporary migrants include those holding a visa that allows temporary employment in the U.S. (e.g. Bracero visas prior to 1965 and E, H, L, O, or TN visas today).

Figure 4.

Probability of U.S. migration in by legal status among Household heads in the MMP161 database

We can see from this graph that the very first migrants who entered the United States from Mexico were undocumented. In the early 20^th century, the border was largely unmarked and undefended and most migrants simply walked across without any sort of inspection. Indeed, the Border Patrol was not formed until 1924, and then only with 450 officers. Between 1910 and 1917 migrants in all three categories—documented, undocumented, and temporary—entered the United States in significant numbers, a state of affairs that also prevailed during the period from 1918 to 1919.

The burst of migration during the 1920s was terminated by the deportation campaigns of the 1930s. In the early 1940s, however, both temporary and undocumented migration began to rise once again, though temporary legal migration rose most rapidly as the Bracero Program expanded. When Bracero visas were scaled back from 62,000 in 1944 to 20,000 in 1947, the likelihood of undocumented migration spiked briefly and the probability of temporary legal migration fell. However, once the supply of Bracero visas expanded to enable the entry from 200,000 migrants in 1952 to 450,000 in 1956, the likelihood of undocumented migration fell and the probability of temporary migration rose, first to 0.018 at the former date and to 0.022 at the latter date.

As the Bracero Program was phased out during the early 1960s, many growers sponsored former Braceros for permanent resident visas, enabling their regular circulation back and forth for seasonal work and thus assuring a stable labor supply (Massey et al. 1987). From 1955 to 1969 the probability of documented migration rose from 0.001 to 0.006, but with restrictive caps placed on permanent resident visas, first across the hemisphere and then upon each nation, the likelihood of documented migration stalled from 1970 to 1985. The end result of this curtailment of access to legal visas was a sharp increase in the likelihood of undocumented migration, which jumped from 0.002 to 0.032 in 1979. From 1979 to 1982 the likelihood of undocumented migration dropped to 0.028 during the oil boom, but it then spiked to a maximum of 0.036 in 1985 at the height of the ensuing economic crisis.

After 1986, IRCA’s amnesty and legalization programs provided legal residence documents to millions of people, and the probability of documented migration accordingly shot up from 0.006 in 1985 to 0.023 in 1989 while the likelihood of undocumented migration dropped from 0.036 to 0.023. The surge of legalizations played itself out during the first two-thirds of the 1990s and by 1997 the probability of documented migration had fallen back to 0.007, about where it had been on the eve of IRCA’s legalizations. Thereafter the likelihood of documented migration sagged further to bottom out at 0.003 in 2007. The likelihood of undocumented migration, meanwhile, fell from its maximum of .036 in 1985 to 0.017 in 1993 before stabilizing, rising abruptly to 0.021 during the boom years between 1997 and 1999, and then falling just as abruptly to 0.015 in 2004 before plummeting to a record low of 0.001 in 2016, a level not seen since the depths of the Great Depression in the 1930s.

After 1976 Congress quietly began increasing the number of temporary work visas available to Mexicans and the likelihood of temporary migration rose slowly from a value of 0.001 in that year to 0.003 in 2010, when the curve for temporary migrants going up crossed the curves for undocumented and documented migrants going down. By 2016 the probability of migration with a temporary visa had risen to 0.009, a likelihood not reached since the mid-1960s. At present, therefore, temporary migration on temporary work visas dominates the migratory flow into the United States.

The last facet of migration we consider is the likelihood of return migration to Mexico, which we compute for household heads in Figure 5, defining it as the probability of returning to Mexico within 12 months of entering the United States. Prior to 1940 the number of trips was too small to compute return probabilities by year, so in the graph we aggregate all data from 1900 to 1940 to compute a single average likelihood of 0.491. Thereafter the probability of return jumps sharply upward to reach a maximum of 0.815 in 1952. During the Bracero years from 1942 through 1964, the average likelihood returning within 12 months was 0.751, indicating the circular nature of migration during the Bracero years. After 1964 the probability of return migration fell back into the range of 0.400–0.500 from 1965 to 1985, and thereafter it dropped steadily to a record low of 0.175 in 2005 before rising back upward 0.450 in 2015.

Figure 5.

Probability of return migration from the United States by household heads in the MMP161 database

Figure 4 and 5 allow us to discern five distinct eras of Mexico U.S. migration, which we have labeled in the figures, each one primarily triggered by policy actions taken in the United States. The first era, from 1900 to 1941, is defined by bookend events of recruitment and expulsion. The era begins with deliberate recruitment efforts by the United States, first through the often exploitive efforts of private labor contractors (Durand and Arias 2000) and later through a temporary worker program established by the U.S. government (Reisler 1976). The era ends with the mass deportation campaigns of the 1930s, which were organized by the federal government in cooperation with local authorities (Balderrama and Rodríguez 1995). Between 1930 and 1936, some 450,000 Mexicans were forcibly expelled from the United States (Hoffman 1974).

In-between these two bookends a large wave of migration crested during the 1920s, one that scholars have labeled the “floodtide of Mexican immigration” (Cardoso 1980). Official statistics reveal that from 1920 to 1929, nearly 490,000 documented migrants entered from Mexico along with some 150,000 temporary workers and an unknown number traveling without documents (Massey, Durand, and Malone 2002). The imprint left in the ethnosurvey data does not reflect the actual size of the 1920s “floodtide” because, as noted earlier, most of the migrants who participated in it were dead by the time we began interviewing in the 1980s.

The next era, from 1942 through 1964, constitutes the era of “Bracero Circulation,” when the U.S. government again established a temporary labor program. During the 22 years of the Bracero Program’s existence, some 4.6 million Mexican men entered the United States with temporary visas for seasonal labor in agriculture and food processing (Massey, Durand, and Malone 2002). The era ends with the unilateral withdrawal of the United States from the accord, which Congress let expire at the stroke of midnight on December 31^st of 1964, and the enactment of restrictive amendments to the Immigration and Nationality Act in 1965, cutting off the supply of legal visas.

During the ensuing era of “Undocumented Circulation” from 1965 to 1985 the seasonal flows established during the Bracero Era continued and expanded under undocumented auspices, supported and sustained by well-developed migrant networks (Massey and Pren 2012). The era of undocumented circulation came to an end with the passage of the Immigration Reform and Control Act of 1986, which began the transformation of what had been a circular flow of male workers going to three states into a settled population of families living in 50 states—hence the label “Militarization and Settlement.” It accomplished this feat in two ways: granting legal permanent residence to around two million former undocumented migrants and launching what would prove to be a multi-decade militarization of the Mexico-U.S. border.

IRCA’s legalization program had two components: an amnesty for migrants who could document at least five years of U.S. residence and an additional special program for migrants who had worked at least 90 days in agriculture during each of the three years leading up to May 1, 1986. The former were already settled, of course, but most of the latter had been circulating back and forth and once IRCA passed they stopped circulating and settled down to work through the bureaucratic process of legalization (Durand, Massey, and Parrado 1999).

The militarization of the border entailed a massive increase in the Border Patrol’s personnel and budget. From 1986 to 2016 the agency went from 3,700 to 20,000 officers and its budget rose from $151 million to $3.6 billion (Massey, Durand, and Malone 2002). Naturally, these resources were initially targeted at the two busiest crossing sectors, a policy that was formalized with the launching of Operation Blockade in El Paso in 1993 and Operation Gatekeeper in San Diego in 1994. Prior to that time, these two sectors together accounted for around three quarters of all undocumented entries, with two thirds coming through California alone (Massey, Durand, and Pren 2016). In the end, this massive militarization did not deter would-be undocumented migrants from heading northward. Instead, it diverted migrants away from militarized segments of the border into the Sonoran Desert towards new crossing sites along the Arizona border, where few Border Patrol agents were stationed. It thus had little effect on the likelihood of apprehension or the odds of gaining entry over a series of attempted crossings (Massey, Durand, and Pren (2014).

Although border militarization did not have its intended effects, it did have profound consequences the behavior of undocumented migrants. First, the shift into rugged desert terrain induced them to make greater use of border smugglers while increasing the cost of their services. Even with these inputs, however, the new crossing routes carried a higher risk of injury or death (Massey, Durand, and Pren 2016). Second, the rising costs and risks of clandestine crossing, in turn, caused migrants to minimize border-crossing—not by staying home but by remaining longer in the United States once they had achieved entry, thereby driving down the rate of return migration and increasing the net inflow (Massey, Durand, and Pren 2015). As a result, from 1988 to 2008 the number of undocumented U.S. residents grew from two million to 12 million (Wasem 2011).

Third, once they had been diverted away from California, the migrants kept on going to new destinations throughout the country. According to U.S. census data, 63% of Mexicans who arrived between 1985 and 1990 went to California; but the share was only 28% among those who arrived between 1995 and 2000 (Massey and Capoferro 2008). Finally, as male workers stayed longer north of the border, they began sending for spouses and children, with the latter ultimately becoming today’s population of “dreamers.” Unmarried male migrants typically formed new relationships with other U.S. residents, at times citizens or legal permanent residents but most often with other undocumented migrants. Once reunited or newly paired, these couples began having U.S. citizen children, thereby creating millions of mixed-status families. Around half of all undocumented households currently have children present within them, 79% of whom are U.S.-born citizens (Passel and Taylor 2010; Capps, Fix, and Zong 2016).

The final era of migration dates from 2008 when the Great Recession struck the United States and brought about the definitive end of undocumented migration. Between 2008 and 2009 the undocumented population fell by a million persons and the number of undocumented Mexican residents continued to decline thereafter, dropping from 6.9 in 2008 to 5.6 million in 2015 (Passel and Cohn 2017). After 66 years of continuous movement, undocumented migration thus ended in 2008, not because of border enforcement but because of Mexico’s demographic transition.

In 1970 Mexico’s total fertility rate was 6.8 children per woman whereas today it stands around 2.2, slightly above replacement level. As a result, Mexico has become an aging society, with the median age rising from around 17 in 1970 to 29 today. According to MMP data, the average age of household heads who are in the labor force but have not yet migrated to the United States, which was 23 in 1970, is now 46 (Massey, Durand, and Pren 2016). Like most demographic events, migration is highly age dependent, rising sharply during the late teens to peak at around age 23 and the declining rapidly to low levels beyond age 30. If people do not begin migrating between the ages of 18 and 30 they are very unlikely ever to migrate at all; and as the average age has risen, the share of Mexicans in this age range has steadily dwindled. As a result, the likelihood of taking a first undocumented trip has dropped nearly to zero.

Some, though not all, of the net outflow of undocumented Mexicans stems from the huge increase in deportations, which began with the passage of the 1996 Anti-Terrorism and Effective Death Penalty Act, accelerated with the enactment of the 2001 USA Patriot Act, and skyrocketed under President Obama. Whereas only 35,000 Mexicans were deported in 1995, by 2000 the number had reached 150,000, and during the last year of the Bush Administration the number stood at 247,000. The number of deportations to Mexico peaked under Obama at 309,000 in 2013, with a total 2.2 million deported during his eight years in office. The current era is thus one of intense repression directed particularly at immigrants from Latin America, especially Mexicans. Of those migrants deported from the U.S. during 2016, 87% were from Latin America, with 64% coming from Mexico alone and another 20% from Central America.

However, the current era is not defined by repression alone, for even as undocumented migration has ceased, temporary migration on legal work visas is rising rapidly. The number of temporary work visas issued to Mexicans increased from 45,000 in 1997 to 252,000 in 2017. Over the same period, the annual number of entries by temporary Mexican workers climbed from 36,000 to 772,900. Consistent with these statistics, the MMP data plotted earlier in Figure 4 demonstrated that the probability of taking a trip with a temporary visa tripled among household heads.

Thus the combination of a militarized border and renewed guest worker migration has led to the paradoxical situation where 5.6 million undocumented Mexicans are trapped north of the border living under increasing pressure from immigration authorities while hundreds of thousands of Mexicans with temporary and even permanent resident visas circulate back and forth (Massey, Durand, and Pren 2015). We therefore label the current period one of “Repression and Legal Circulation.”

Undocumented Mexicans in an Age of Repression

Over the course of the 20^th and 21^st centuries we have witnessed a steady progression across successive eras of Mexico-U.S. migration, each shaped by changing circumstances in Mexico to be sure, but fundamentally triggered at each juncture by dramatic shifts in U.S. policies and practices. In 2017 Donald Trump assumed the presidency on an explicit platform of white nationalism, one that singled out Mexicans as “rapists” and sought to curtail migration from “shithole countries” by constructing a $25 billion-dollar wall along the Mexico-U.S. border while also expanding the nation’s internal system of detention and deportation. The wall has yet to be built and its construction would have little effect on undocumented Mexican migration, which effectively ended a decade ago. In contrast, the immigration detention system processed nearly 324,000 persons in 2017 while Immigration and Customs Enforcement carried out 226,000 removals. A total of some 2.3 million migrants are currently under supervision by the Department of Homeland Security, reporting regularly to immigration officials under the ever-present threat of removal (Rinaldi 2017).

Today’s population of 11 million unauthorized residents are under pressure as never before and constitute the largest population lacking social, economic, and civil rights since the days of slavery, when the slave population was just under 4 million. At present some 5.1 million children live in the United States with an undocumented parent, and nearly 81% are U.S. citizens or legal permanent residents. These children are more likely than other children of immigrants to live in poverty under conditions of linguistic isolation, and are less likely to experience socioeconomic improvement over time (Capps, Fix, and Zong 2016). They also have less access to health care (Gelatt 2016) and are in rather poor health, both physically (Vargas and Ybarra 2017) and mentally (Delva et al. 2013).

Perhaps even more unfortunate than U.S. citizens living with undocumented parents in the United States are the more than 600,000 U.S. citizen children living in Mexico with their deported parents, where they are themselves paradoxically undocumented, lack the legal identification to prove legal residence and ensure access to Mexican social services (U.S. Embassy and Consulates in Mexico 2018). Recent research using MMP data indicates that adult undocumented migrants living in Mexico have worse health than their non-migrant neighbors and that the longer their total exposure to the United States and the more trips they have taken, the worse their health is upon returning home (Cheong and Massey 2018). Unsurprisingly, undocumented adults living in the U.S. likewise suffer serious health difficulties, especially with respect to mental health (Sullivan and Rehm 2005).

As noted earlier, before the passage of IRCA undocumented migrants earned the same wages as documented migrants, other things equal. Since then, however, wages and working conditions have deteriorated for Mexican migrants generally, especially for those lacking legal documents. From 1970 to 2010 the human capital endowments of Mexican immigrant workers increased but the returns to human capital fell, bringing about a stagnation of wages (see Massey and Gelatt 2010). Work by Massey and Gentsch (2014) using MMP data suggests that this stagnation of wages occurred through the joint influence of IRCA’s criminalization of undocumented hiring and the growth in the number of exploitable undocumented workers in U.S. labor markets.

Undocumented migrants earned 8.3% less in hourly wages than those with documents, and wages decreased by an additional 1.2% with each additional point in the percentage undocumented in the labor market. Durand, Massey, and Pren (2016) have further shown that undocumented migrants experience a “double disadvantage” in U.S. labor markets, with the lack of legal status pushing them into the secondary sector jobs where wages are 12.5% lower than in the primary sector while working within that sector they experience and additional 11.5% wage penalty, putting them in a very precarious economic position (Massey, Durand, and Pren 2016).

At this writing, the future of Mexican migration to the United States is unclear. On the one hand, a large number of undocumented migrants remain in the United States where they face a bleak existence under the presidency of Donald Trump, whose repressive policies seem destined to further marginalize them socially and economically within U.S. society, compromising not only their health and wellbeing, but that of their children, the large majority of whom are U.S. citizens and the nation’s future workers and taxpayers. On the other hand, legal migration is expanding and as the share traveling on temporary work visas rises this movement is increasingly circular. Thus we now inhabit what Durand (2016) calls a “bipolar” age, one characterized the ongoing repression of undocumented migrants north of the border combined with the ongoing exploitation of legal but temporary migrants who crisscross a heavily militarized border to work in an increasingly integrated North American labor market. How this bipolarity plays out in the future will be documented by succeeding waves of data collected by the Mexican Migration Project.

Figure 6.

Estimated size of the undocumented population in thousands