Surveys are studies conducted to obtain information on characteristics of a population. Well-designed surveys based on probability sampling can provide valuable information about a large population using just a small representative sample of subjects. Because sample surveys are less expensive than conducting a full census of the population of interest, they are quite attractive as an option for obtaining population measurements. A number of sample surveys that include questions on migration are conducted in both the United States and Mexico; these surveys are discussed in this chapter.
The theory of sampling begins with a target population of observations to be studied: for example, households or individuals about which inferences will be made. The sampling frame is a list of sampling units from which the sample to be actually observed is drawn. Probability sampling assumes that each unit in the population has a known, nonzero chance of selection into the sample, and a chance method is used to select the units in the sample. Estimates for the entire population of interest and measures of the uncertainty (or variance) in these estimates can be computed from the sample data and the sampling design.
The ideal of probability sampling is to construct a perfect frame for the target population and then select a sample from that frame in accordance with the probability sampling design. The survey then observes the true value for each study variable on each unit selected into the sample (the data collection) and processes the data without introducing errors. In practice, most surveys fall short of these ideal conditions. For example, in many surveys, not all of the selected sample units respond. In addition, coverage error exists since the sampling frame may not have included part
of the target population. Statistical inferences are affected by nonsampling errors, such as coverage, measurement, and non-response errors. Survey researchers often use weighting procedures to account for the features of the sampling design and make adjustments for non-response or coverage error.
For the purposes of this report, it is important to determine how well the existing surveys are designed, in order to assess their usefulness. The surveys described are conducted in the United States, in Mexico, or both, and the panel evaluated them to determine their effectiveness in estimating both the number of foreign nationals who attempt illegal entry across the U.S.–Mexico land border and the probability of apprehension of illegal entrants. Our evaluation also addresses the possibility of obtaining these estimates on a quarterly basis for particular regions of the U.S.–Mexico border.
This chapter outlines the features of the major surveys in the United States and Mexico (summarized in Table 3-1) that collect information about migration and border crossing; details about some of the specific questions asked by the surveys are provided in Appendix A. Chapter 4 discusses the usefulness and limitations of these surveys in estimating the number of foreign nationals who attempt illegal entry across the U.S.–Mexico land border.
AMERICAN COMMUNITY SURVEY
The American Community Survey (ACS) conducted by the U.S. Census Bureau is intended to collect data comparable with the Census 2000 long-form sample data, but to do so every year rather than every 10 years (Grieco and Rytina, 2011). Information from the ACS is used to administer federal and state programs and distribute federal funds. The survey asks about age, gender, race, family and relationships, income and benefits, health insurance, education, veteran status, disabilities, where people work and how they get there, where people live, and how much people pay for certain essentials. Fully implemented since 2005, the ACS collects annual data in twelve monthly samples. The target population of the ACS is the entire resident population of the United States and Puerto Rico. The sampling frame reflects this target population by identifying all addresses of households for the 2005 ACS and all addresses of both households and group quarters for the ACS since 2006. Data collection uses three modes that take place over a 3-month period: mail, telephone, and personal visit. The target population for the ACS only includes people who are deemed to be residents of the United States; short-term migrants, such as many undocumented Mexican workers, would not be part of the ACS universe.
The 2010 Public-Use Sample, which is a 1 percent sample of the U.S. population, included 1,204,000 households and 3,062,000 respondents. Of these, 145,000 households were immigrant households (35,000 Mexican
and 8,400 Central American) and 348,000 respondents were foreign-born (92,900 Mexican and 22,200 Central American). Each year’s ACS is weighted to the current population estimates for that year with a weighting methodology that controls for small areas, race/Hispanic origin, age, gender, and marital status/households (see U.S. Census Bureau, 2009, for details). The annual ACS aggregates the 12 monthly samples to yield annual data. The U.S. Census Bureau does not release information about the dates interviews were conducted and, therefore, there is no specific “reference date” within the year for the survey.1 Immigration-related questions asked by the ACS include nativity/citizenship (but not legal status), year of arrival in the United States,2 country of birth, year of naturalization, residence 12 months before interview, and language spoken at home.
The U.S. Department of Homeland Security (DHS) uses the ACS to estimate the size of the unauthorized immigrant population residing in the United States (see Box 3-1). DHS has published annual estimates for 2006-2011 using the ACS data for 2005-2010 (Hoefer et al., 2011).
CURRENT POPULATION SURVEY
The CPS is the primary source of labor force statistics for the U.S. population and is also the source of high-profile economic statistics such as the national unemployment rate. The CPS, which is sponsored by the U.S. Census Bureau and the U.S. Bureau of Labor Statistics, is administered monthly by the U.S. Census Bureau (Grieco and Rytina, 2011). This survey frequently includes specialized supplements, and the Annual Social and Economic Supplement (ASEC), conducted in February-April each year, has an expanded sample (a double sample of Hispanic households and an oversample of other minority households and households with children). Although monthly data are generally limited to labor force items and education, a broad range of social, economic, and demographic data are collected in the CPS ASEC (formerly known as the “March supplement”).
The target population of the CPS is the civilian non-institutional population living in the United States. The sampling frame is a list of housing addresses obtained from the most recent decennial census and updated with new housing units built after the census. The CPS ASEC also includes military personnel living in off-base housing with civilian adult household members. In 2011, the average monthly sample (from January to November) in the public-use sample included 54,000 households and
1 The ACS is weighted using population estimates for July 1 of the survey year.
2 The question about year of arrival can be subject to various interpretations for circular migrants, and estimates of migration based on year of arrival tend to be higher than those based on residence 1 year ago.
|ACS||CPS (ASEC)||Mexican Census (10% sample–long form)||ENOE|
|Target population||U.S. households||U.S. households||Mexican households||Mexican households|
|Frequency of survey||Annual||Annual||Every 10 years||Quarterly|
|Mexico and U.S. samples||No||No||No||No|
|Total sample size||~1,204,000 households||~97,000 households per month||~2.9 million dwellings||120,260 dwellings per quarter|
|INCLUDES QUESTIONS ABOUT:|
|Documentation status at crossing||No||No||No||No|
|Number of attempts||No||No||No||No|
|Reasons for migrating||No||No||No||Yes|
|Intention to migrate||No||No||No||Yes
aAsked of those not currently working.
bWeights are estimated; due to the nature of the survey design, their accuracy is difficult to quantify.
cEMIF-N is conducted continuously, and, beginning with 2012, data are released on a quarterly basis. From 2013 on, the intent is to reduce the delay in reporting each quarter to 2 months.
eAsked on U.S. side only.
|Mexican households||Migrants passing through Mexican border cities||Mexican households||Sending communities in Mexico||Sending communities in Mexico|
|1992, 1997, 2006, 2009||Quarterly
||2002, 2005, 2009||Annual||Annual|
|~40,000-100,000 dwellings||~14,000 respondents
||~35,000-40,000 respondents||~600-1,000 households||~700-1,000 respondents|
The residual methodology, which subtracts estimates of the legally resident foreign-born population from estimates of the total foreign-born population, is used to estimate the size of the unauthorized immigrant population residing in the United States. Warren and Passel (1987) applied residual methods using information from the 1980 Census. Edmonston and colleagues (1990) applied similar methods to data from the Current Population Survey (CPS)a in 1983 and 1986 and to data from the 1990 Census. With the regular collection of data on the foreign-born population in the CPS, beginning in 1994, more frequent estimates based on the residual method appeared (e.g., Passel and Clark, 1998; U.S. Immigration and Naturalization Service, 2001; Warren, 1994 [cited in U.S. Immigration and Naturalization Service, 2001]). An early residual-type estimate developed by Bean and colleagues (1997) for the U.S. Commission on Immigration Reform combined data from the 1996 CPS March supplement with data from Mexico to estimate the number of unauthorized Mexicans in the United States (with a correction for CPS omissions). Annual estimates of the unauthorized immigrant population have been published by DHS for 2006-2011 using the ACS (Hoefer et al., 2012) and by the Pew Hispanic Center for 2000-2010 using the CPS (Passel and Cohn, 2011).
Most residual estimates of the unauthorized immigrant population take the same broad approach:
1. An estimate of the number of legal foreign-born residents in the United States is developed using administrative data on legal admissions across a number of years. The data are combined using demographic techniques to allow for mortality and emigration after admission to the United States.
2. A survey (ACS, CPS, or Census) is used to generate tabulations of the number of immigrants found in the survey.
3. An initial estimate of the number of unauthorized immigrants appearing in the survey is derived by subtracting the results of step (1) from the results of step (2).
4. A final estimate of the total number of unauthorized immigrants in the country is derived by adjusting the results of step (3) for undercount in the survey.
The various estimates using this approach differ in terms of their specificity (e.g., countries of origin, states, and periods of entry) and in the assumptions made about various parameters of the estimation process. The DHS and Pew estimates
both assume that all immigrants who entered the United States before 1980 are legally present, so the construction of a legal foreign-born population estimate uses only administrative data on admissions for 1980 and later.
There are three main areas where the existing residual estimates rely on data and assumptions for which strong empirical support is lacking and which can affect the overall magnitude of the estimates. First, the United States lacks data on departures from the country, so emigration of legal residents must be estimated using a variety of methods. Second, the DHS and Pew estimates differ in how they handle legal temporary migrants with longer-term visas who appear in the ACS and CPS (such as foreign students and intracompany transfers). Third, firm estimates of coverage of unauthorized and legal immigrants in surveys such as the ACS and CPS are also lacking, so both estimates rely on information drawn from local surveys and on assumptions based on various external sources.
To place the sensitivity of the residual estimates in context, the average annual growth in the unauthorized population between 2000 and 2005 was more than 400,000 according to DHS estimates (Hoefer et al., 2011) and more than 500,000 according to the Pew estimates (Passel and Cohn, 2011). But both sets of estimates show peaks in 2007 followed by declines, so that the net annual growth between 2005 and 2010 was less than 100,000. Annual estimates of emigration of post-1980 legal immigrants in recent years have exceeded 200,000, and the estimated cumulative emigration since 1980 amounts to 3.6 million (Hoefer et al., 2011). So, relatively small variations in this hard-to-measure component could have a significant impact on the resulting estimate of unauthorized immigrants. The number of legal temporary migrants living in the United States in 2010 according to DHS estimates is 1.8 million (Hoefer et al., 2011), yet the Pew estimates find evidence in the CPS for fewer than 1 million.
Both the DHS and Pew estimates allow for undercount of legal and unauthorized immigrants amounting to about 1.5 million. Yet, these assumptions rely principally on a relatively small study done in Los Angeles after the 2000 Census (Marcelli and Ong, 2002). Clearly, better information on coverage of immigrants in the ACS and CPS would improve the precision of the residual estimates and their face validity. The importance of understanding survey coverage of immigrants has long been recognized, and the U.S. Government Accountability Office (GAO) has recommended that the U.S. Census Bureau devote time and resources to the topic (U.S. General Accounting Office, 1998). Little official government work has been done to fill this vital information gap, and the GAO recommendation is considered “open.”
a Described below.
133,900 respondents; 6,400 households were immigrant (1,600 Mexican and 440 Central American), and 14,600 respondents were foreign-born (4,000 Mexican and 1,100 Central American). The 2011 CPS ASEC sample included about 97,000 households and 205,000 respondents.
Each year’s CPS is weighted to the year’s population estimates generated in the previous year. The weighting methodology has final controls for states by broad age group (and race,3 depending on sample size); national data by age, gender, and race; and national data by age, gender, and Hispanic origin. The CPS ASEC has an additional weighting step to ensure that both spouses (or both unmarried partners) have equal weights.
Immigration-related questions asked by the CPS include nativity/citizenship (but not legal status), period of arrival in the United States (released in public-use data at 2-3 year intervals),4 country of birth, residence 12 months before interview (ASEC only), and country of birth of parents.5 Due to the small sample size of the CPS, there are limitations for measuring year-to-year changes and date of arrival disaggregation.
The Pew Hispanic Center has used the CPS to estimate the size of the unauthorized immigrant population residing in the United States for 2000-2011.6 It does so using a residual methodology (see Box 3-1). Using additional data (such as survey information on date of arrival) and assumptions, inflows and outflows of unauthorized immigrants can also be estimated from the ACS- or CPS-based series of annual population estimates (Passel and Cohn 2009a, 2010); such estimates are extremely sensitive to underlying assumptions, as discussed in Box 3-1.
MEXICAN CENSUS OF HOUSING AND POPULATION (LONG AND SHORT QUESTIONNAIRES)
The Mexican Census of Housing and Population is funded by the Government of Mexico, with data collected, processed, and made publicly available by the National Institute of Statistics and Geography (known in Spanish as Instituto Nacional de Estadística y Geografía [INEGI]). This census is conducted every 10 years, at the beginning of each decade. The
3 With categories for blacks and all others.
4 As with the ACS, the question about year of arrival can be subject to various interpretations for circular migrants, and—even more so than in the ACS—CPS estimates of migration based on year of arrival are higher than estimates based on residence 1 year ago.
5 The CPS does not have data on language spoken at home. The CPS has asked monthly questions about place of birth, parental place of birth, U.S. citizenship status, and year of entry into the United States since 1994 (Grieco and Rytina, 2011).
6 The Pew Hispanic Center uses the CPS for historical reasons, and it plans to switch over to the ACS in the near future.
In both 2000 and 2010, two different questionnaires were administered. The short form is administered to the entire Mexican population and collects basic social-demographic information for all household members, as well as characteristics of the dwelling.7 It includes information about the place of residence 5 years prior to the interview (i.e., for 2010, the location of residence in 2005 was asked),8 which permits estimation of the number and basic social-demographic characteristics of the population that returned from the United States in the past 5 years.
A long questionnaire was administered to a random sample of about 10 percent of the total Mexican population in the 2000 and 2010 census. In 2010, more than 2.9 million dwellings were interviewed. Expansion weights9 were estimated for each dwelling based on the sampling weight and on an adjustment factor accounting for differential response rates. The sample provides estimates reflecting the population at the following levels: national, state, state with four community sizes predefined, municipality, and localities with more than 50,000 inhabitants (Instituto Nacional de Estadística y Geografía, 2011b, 2011c).
The long questionnaire includes the questions listed in the short questionnaire, plus a special section on the international migration experiences of household members during the prior 5 years, including those no longer in the household. In addition to the question on place of residence 5 years ago, which allows estimation of return migration between 2005 and 2010, the long questionnaire includes a set of questions on international migration. These questions refer to the migration of any person who is currently living in the dwelling or who lived in the dwelling between 2005 and 2010. Information is collected on the number of people, gender, age at migration, date of last migration, state of residence at the time of the migration, country of destination, country of current residence, date of return (for those who returned), and whether the person is currently living in the same household (Instituto Nacional de Estadística y Geografía, 2011a:36-42). The questionnaire also asks whether the household is currently receiving remittances. The 2010 census sample includes 83,757 households with at
7 The definition of “household” changed between the 2000 and the 2010 Census. In 2000, there could be more than one household per dwelling. In 2010, there was no question about the number of households in a dwelling. This change in the definition needs to be taken into account when comparing the international migration information from the two censuses.
8 This question is asked for all of the dwellers ages 5 years and older.
9 Expansion weights are sampling weights. The word “expansion” is used because it describes the expansion of that sampling unit to the population from which it was sampled.
least one return migrant,10 31,536 households with at least one circular migrant,11 and 89,601 households with at least one out-migrant.12 In total, 186,456 households reported a migration experience, and 149,000 households reported that they were receiving remittances.
Using the questions in the long questionnaire, the following analyses are possible:
• Estimates of the flow of out-migration, circular migration, and return migration in the past 5 years nationally and by state of residence at the time of the migration (see, e.g., Giorguli and Gutiérrez, in press);
• Gender and age composition of the above flows;
• Estimates of the return rates and the social-demographic characteristics of return migrants (and of circular migrants if currently living in the same household); and
• Geographical profile of the intensity of migration by municipality (based on the proportion of households where at least one member has migrated and/or received remittances).
NATIONAL SURVEY OF OCCUPATION AND EMPLOYMENT (ENOE)
Mexico’s National Survey of Occupation and Employment (known in Spanish as Encuesta Nacional de Ocupación y Empleo [ENOE]) is funded by the Government of Mexico. INEGI is responsible for collecting and processing the ENOE data and making them publicly available on its website. The main objective of this national probability survey is to capture short-term changes in the occupation and labor force situation of the working age population in Mexico. It also includes detailed information on employment characteristics, access to social benefits, hours worked per week, income, social-demographic characteristics, and residential status. The first round of ENOE was conducted in 2005, although antecedents of the survey date back to 1972.
The target population is the working age population of Mexico (individuals 12 years and older). The sampling frame is INEGI’s National Households Frame 2002, which is based on cartographic and demographic
10 “Return migrant” refers to a person born in Mexico who lived in the United States before 2005 and returned any time between 2005 and 2010.
11 “Circular” migrant refers to a person born in Mexico who moved to the United States after 2005 and returned before 2010. There is some overlap between circular migrants and return migrants (Passel et al., 2012).
12 “Out-migrant” refers to a person born in Mexico who left for the United States between 2005 and 2010 and has not returned.
information from the General Population and Housing Census 2000. ENOE collects information from approximately 120,000 dwellings every quarter and is based on a probabilistic, two-staged, stratified cluster sampling design. The sampling frame is stratified by socioeconomic status and is designed to reflect national, state, city, and community-size levels.13 Households within dwellings are interviewed five times within a period of 15 months. Each quarter, 20 percent of the dwellings leave the sample after completing the fifth round and are replaced by new entries, which are randomly selected with unequal probabilities of selection from the stratified sampling frame. Expansion factors include a sampling weight, a weight to reflect differential non-response rates, and calibration weights to the official projections of the population for a given year.
ENOE asks those not currently working whether they have tried to look for a job in another country or whether they are preparing to cross the border. ENOE also has information on whether individuals receive remittances, though without specifying the amount and regularity. Other migration information is captured in the household roster, which asks about the residential status of all individuals living in the dwelling at the time of the prior round of interviews. If a person is no longer living in the household, the respondent is asked about the reasons why he/she moved out (work, study, health problems, family reasons, among others) and his/her current place of residence (the same state, another state, or outside of the country14). This information can be complemented with the basic social-demographic variables obtained in the prior interview. The household roster also captures information on new arrivals. Aside from the social-demographic profile of the new members of the household, it asks the reasons for moving in (similar to the responses for those moving out) and which state or country the respondent was leaving before arriving to the household. In the 2010 ENOE, an average of 239 households per round reported out-migrants and 185 households reported returned migrants moving in from outside of Mexico. An average of 367 households per round had either a new arrival or somebody leaving for another country. (As noted above, each round includes more than 120,000 dwellings.) Researchers state that ENOE can capture short-term changes15 in Mexico–U.S. migration trends, and it has been used to estimate out-migration and in-migration flows from 2006 to the present (Bustos, 2011; Zenteno, 2011).
13 The four size categories for communities are less than 2,500 inhabitants, 2,500 to 14,999 inhabitants, 15,000 to 99,999 inhabitants, and 100,000 or more inhabitants.
14 ENOE does not specify the country of origin or destination of international migrants. Nevertheless, the vast majority of all international movements—above 95 percent—are known to occur between Mexico and the United States (Galindo and Ramos, 2009:114).
15 That is, quarterly migration or any movement between the first and last round (15 months).
Mexico’s National Survey of Population Dynamics (known in Spanish as Encuesta Nacional de la Dinámica Demográfica [ENADID]) is funded by the National Council on Population (known in Spanish as Consejo Nacional de Población [CONAPO]) and INEGI. The purpose of ENADID is to provide information about the different components of population dynamics (e.g., fertility, mortality, internal and international migration). INEGI is responsible for questionnaire design, operational planning, survey execution, and data design, while CONAPO (along with a group of academic experts) participates in a conceptual review and validates the results before publication. ENADID was first fielded in 1992, with subsequent cross-sections in 1997, 2006, and 2009. In its origins, ENADID was one of the few national probability-based surveys that had instruments to measure international out- and in-migration.16
The target population of ENADID is the population permanently residing in private homes in Mexico. The sampling frame is INEGI’s National Households Frame 2002, based on cartographic and demographic information from the General Population and Housing Census 2000. Data from each of ENADID’s cross-sections come from samples aimed at reflecting characteristics of the whole country, each of the 31 states and the Federal District, and urban and rural areas (Instituto Nacional de Estadística y Geografía, undated-a, undated-b; Instituto Nacional de Salud Pública, 2008). Each cross-section of the ENADID aims to represent all household members at the time of the survey (including both Mexican- and foreign-born individuals) and all individuals who were members of the household roughly 5 years prior to the survey and who left the household to move to another country during this period, whether they returned to the sampled household or not.17 Sample sizes for the ENADID surveys have ranged between 40,000 and 100,000 households.
ENADID includes questions to measure international out- and in-migration. In the case of in-migration, all of the ENADID cross-sections use the relatively conventional retrospective question asking if people living in the household were living in another state or country 5 years prior to
16 Before 2000, the Mexican Census only measured in-migration; ENOE’s predecessors only occasionally included migration questions; the Survey of Migration at the Northern Border (see next section) had only started in 1993 and was not continuous (and it also had a different scope and use than the ENADID in terms of understanding international migration flows); and more detailed, non-representative surveys like the Mexican Migration Project (MMP) were only starting and had (more) limited coverage and sample sizes.
17 That is, they could remain in the United States or live elsewhere in Mexico, but only those moving abroad during the prior 5 years would be recorded in these questions.
the survey.18 For the measurement of out-migration, the main respondent of the survey is asked if any person who currently lives or previously lived in the household left for the United States about 5 years prior to the survey (the period of reference varies slightly by cross-section), the times of out-migration and (if applicable) return, and the basic social-demographic profile of the emigrants. In 2006 and 2009, the ENADID questionnaire also included items aimed at measuring the documentation or visa held by people at the time of out-migration, including residence permits (other than a green card), work permits, green card, tourist visa, student visa, U.S. citizenship, other, and no documentation.
ENADID has been used to study international migration, mostly to describe the general social-demographic profile of migrants (Canales, 1999; Durand and Massey, 2003; Durand et al., 2001; Riosmena and Massey, 2012). It has also been used to complement analyses using convenience samples, such as the MMP (McKenzie and Rapoport, 2007) and as a criterion to evaluate the validity of the MMP (Massey and Capoferro, 2004; Zenteno and Massey, 1999), the Survey of Migration at the Northern Border of Mexico (Rendall et al., 2009), and ENOE (for measuring return migration) (Rendall et al., 2011). Estimates of international out- and in-migration using ENADID (along with many other data sources) are provided in Galindo and Ramos (2008).
SURVEY OF MIGRATION AT THE NORTHERN BORDER OF MEXICO
The Survey of Migration at the Northern Border of Mexico is known by the acronym EMIF-Norte (or simply EMIF-N) for its Spanish name, Encuesta sobre Migración en la Frontera Norte de México. EMIF-N is a multistage probability sample of flows across the U.S.–Mexico border and has been used to estimate migration flows (Rendall et al., 2009). It is sponsored by the Mexican government’s Ministry of Labor and Social Welfare, CONAPO, National Migration Institute (Instituto Nacional de Migración, INM), and Ministry of Foreign Relations. Funding has also been provided by the World Bank (Secretaría de Gobernación et al., 2010, 2011). The survey is carried out by El Colegio de la Frontera Norte (COLEF) and has been fielded annually since 1993. Microdata for the period from 1995 to the first quarter of 2012 are presently available on the COLEF website (http://www.colef.mx/emif) for public download. Data and documentation are also available on the CONAPO website (http://www.conapo.gob.mx/es/CONAPO/Encuestas).
18 For a discussion of the advantages and limitations of these questions, see Rogers and colleagues (2003).
The EMIF-N survey design is based on methods originally developed by biologists for sampling migratory populations (Bustamante, 1998; Santibáñez, 1999) and is similar to the United Kingdom’s International Passenger Survey (Economic and Social Data Service, 2008; Office for National Statistics, 2007). Like the International Passenger Survey, EMIF-N samples passengers at airports, bus depots, and train stations, but it also samples passengers at ports of entry, international bridges, and Mexican Customs inspection points (Bustamante et al., 1998; Consejo Nacional de Población, undated; Santibáñez 1997, 1999; Secretaría de Gobernación et al., 2010, 2011; Secretaría de Trabajo y Previsión Social, 1997, 1998, 1999).
EMIF-N is made up of four subsamples of migrants entering or leaving border cities: first, migrants entering from the interior (i.e., southern Mexico), who are identified by the survey as intending to cross the U.S.–Mexico border; second, apprehended migrants returned to the Mexican side of the border by U.S. authorities; third, migrants leaving the U.S. border region for the Mexican interior; and fourth, migrants returning to the interior from northern Mexican border cities (not of interest to the current study). Within each of the specific localities, multi-stage probability sampling is used and incorporates geographic and temporal stages of selection (Secretaría de Gobernación et al., 2010, 2011).
The target population is all individuals 15 and older: (1) who are not residents in a border city or the United States, not born in the United States, and arriving in a Mexican border region;19 and (2) whose travel is due to a job or job search; change of residence; being in transit to the United States; or other reasons such as study, tourism, or visiting family or friends; and (3) who have neither a fixed return date nor employment in their place of origin. The sampling frame is a list of all zones (such as airports and bus stations) and points (such as airport gates and entries to bus terminals from unloading areas) identified by EMIF-N investigators in preselected border cities.
In the data collected in 2008, the size of the subsample of migrants from the interior (i.e., southern Mexico) was 13,792, of which 8,075 reported intentions of crossing the border within the next 7 days. The size of the subsample of migrants returning from the United States was 7,729, and the size of the subsample of detained migrants returned by DHS was 6,989. Response dispositions for each of the four subsamples are not clearly documented, although information on refusers is collected. EMIF-N’s weights currently presume that refusers are nonmigrant travelers; the weights are not adjusted for survey non-response. Differential non-response is therefore a potential source of bias in EMIF-N.
19 With the exception of the subsample of migrants returning voluntarily from the United States, which includes migrants reporting that they reside in the United States.
EMIF-N has an adaptive time-location sampling design. The set of locations in the sampling frame evolves in response to the EMIF-N team’s perception of changes in the geographical and temporal distributions of migratory flows, with the intent of covering the total flow through the border region. Fieldwork prior to the first (1993-1994) wave of EMIF-N indicated that 23 border localities (cities or towns) “constituted practically the [entire] universe of crossing locations for the labor flows to and from the United States” (Secretaría de Trabajo y Previsión Social, 1998:20).20 The first wave surveyed in 18 of these localities and found that 8 cities accounted for 94 percent of the migrant flows between the United States and Mexico (Secretaría de Trabajo y Previsión Social, 1997, 1998). Although the sampling frame is continually updated, these eight cities (Cuidad Juárez, Matamoros, Mexicali, Nogales, Nuevo Laredo, Piedras Negras, Reynosa, and Tijuana) have been surveyed in every year of EMIF-N’s subsamples of northbound and southbound migrants.
Qualitative fieldwork is carried out on an ongoing basis using informants in localities (i.e., cities or towns in Mexico along the U.S.–Mexico border) who are knowledgeable about trends and geographic shifts in migrant flows. Based on this research, the sampling frame of localities (and sampling zones and points within localities) is updated annually to adjust for shifts in flows and to maintain high levels of coverage.21 Documentation of the localities sampled in each wave are available on the COLEF website for EMIF-N (http://www.colef.mx/emif).
Within localities included in the sampling frame, EMIF-N survey workers sample at zones (such as bus and train stations and airports) and at points within these zones through which migrants must pass (such as airport gangway doors or the points at which passengers disembark from buses). The probability of sampling a particular point and zone is based on estimates of the share of migrant flows passing through that zone during a particular time period (usually one of three 8-hour shifts each 24 hours). Since at least 2009, these estimates of the share of flows of migrants through sampling zones are based on periods during which EMIF researchers observe each sampling zone simultaneously for 24 hours a day for 7 consecutive days, enumerating all people passing through the zone, and administering to as many as possible the filter used to distinguish migrants
20 Translated from Spanish by the panel.
21 In the 2012 wave, northbound migrants were surveyed in the eight cities listed above and four additional cities (Agua Prieta, Altar, Cananea, and Ciudad Acuña). The subsample of apprehended migrants returned by the U.S. immigration authorities has also been consistently surveyed in the same eight core cities, with the exception of Piedras Negras, which was dropped from the two most recent waves. The 2012 apprehended subsample also included migrants surveyed in Ciudad Acuña and San Luis Río Colorado.
eligible for the survey from others, such as local travelers.22 The information thus gathered is used on an annual basis to update the sampling frame with regard to the share of migrant flows passing through each zone during each time period. The sampling frame is further updated on a quarterly basis based on less exhaustive ongoing fieldwork identifying shifts in migrant flows between years (Secretaría de Gobernación et al., 2011).
Full updates (based on 24-hour observations on 7 consecutive days) of the sampling frame have been carried out in 2009, 2011, and 2012. In each of these years, the update observations were carried out in the first quarter (during February or March) in order to measure flows during seasonal high levels of migration from Mexico to the United States. In 2013, EMIF-N investigators plan to extend these intensive week-long enumerations of flows to 20 localities, which will include all currently sampled localities plus those included in the first wave and dropped in the second wave. This plan will allow for updated information on the share of flows passing through the localities currently in the sampling frame.
The EMIF-N investigators have made an operational decision to focus resources on sampling as exhaustively as possible within localities in the sampling frame, but to exclude from the frame those localities that account for very small shares of migrant flows. The EMIF-N weights do not attempt to inflate estimates to account for the small share of migrants passing through localities (and within localities’ points and zones) not in the frame. Strictly speaking, estimates from EMIF-N using the weights produced by the EMIF-N team are estimates of the flows passing through sampling zones used in the survey, although, as noted, much effort is expended to make sure that sampling localities, zones, and points are selected and updated to adjust for any shifts in migrant flows and to maintain a high level of coverage for the population that migrates through a city. The design implicitly assumes that the flows through the localities not in the sampling frame at a given time are zero. Thus, the weights for the localities at times they are not in the sampling frame are assumed to be zero.
The sampling localities and sites for the EMIF-N subsample of apprehended migrants returned to Mexico by the U.S. authorities are selected based on data that the EMIF-N researchers receive from the Mexican immigration authorities (INM). For sampling frame updating, data are used on apprehended migrants returned during the same week as EMIF-N’s enumeration activities. In the past, the INM annual data on apprehended migrants returned have reported fewer migrants than data reported by the U.S. immigration authorities. INM data count only returned migrants
22 These estimates are based partially on direct observation (the periodic process whereby EMIF-N researchers observe each sampling zone continuously for a week) and partially on records of the operating hours and managers’ records for the zones in the sample.
received by INM from U.S. authorities under the framework of binational agreements. Differences between U.S. and Mexican numbers can be attributed to some combination of error in the statistical systems of one or both countries and “deficiencies and omissions in the application of the relevant binational procedures and protocols” (Sandoval et al., 2011:5).23 At present, EMIF-N researchers adjust their survey weights for migrants from localities based on INM counts.24 Lack of information sharing and coordination between DHS and INM can cause INM to undercount the flow of returned apprehended migrants and, therefore, downwardly bias EMIF-N’s estimates of such migrant flows.25
All subsamples gather information on the demographic characteristics and educational attainment of the respondent. For respondents intending to cross, there are survey items asking about the reason for the trip, the possession of valid documents, the location (city) of planned crossing, the reason for selecting this crossing site, the intended U.S. destination, whether the respondent has already arranged employment in the United States, the intended sector of employment, how long the migrant expects to stay in the United States, whether a smuggler has or will be hired, where the smuggler was/will be hired, and how much the migrant expects to pay the smuggler. There are also survey items regarding the number of previous trips to the United States and details of the most recent trip, including questions regarding previous use of a smuggler to cross, Mexican city of crossing, U.S. city in which respondent stayed the longest, whether the respondent had family or friends in that U.S. city, whether family or friends provided assistance, and employment and earnings in the United States on the last trip.
Similar questions are asked of migrants returning from the United States. The subsample of migrants returned by U.S. immigration authorities corresponds to DHS administrative data on apprehended aliens but provides a richer detail of respondents’ characteristics, history, and crossing experience. A question on how many times they had been captured by the U.S. Border Patrol and returned to Mexico allows the calculation of a probability of apprehension based on a repeated trials model (discussed in more detail in Chapter 5). EMIF-N has been used to estimate elements of unauthorized migration (Escobar-Latapí, 1999; Reyes et al., 2002), examine
23 Translated from Spanish by the panel.
24 If the INM counts for returnees for a given locality are not within the 95 percent confidence interval of the EMIF-N estimate (based on the original weights), the EMIF-N weights are post-stratified to the INM counts.
25 In addition to the adjustment of weights, EMIF-N researchers told panel members of a situation in 2010 in which DHS began returning migrants to San Luis Rio Colorado, a locality not previously used for such returns. Because the EMIF-N team only learned of this shift after it began and did not have operations in place in San Luis Rio Colorado, 3 months passed before the EMIF-N was able to begin collecting data at the new site.
routes of unauthorized migration (Anguiano Téllez, 2007), study legal status as a factor in migrant remittances (Amuedo-Dorantes and Pozo, 2005), and look at differences in wages by temporary/permanent status among unauthorized immigrants (Brownell, 2010).
MEXICAN FAMILY LIFE SURVEY
The Mexican Family Life Survey (MxFLS) is a longitudinal survey of individuals, households, families, and communities in Mexico. The survey is jointly administered by Luis Rubalcava (Centro de Investigaciones Demograficas y Economicas, Mexico), Graciela Teurel (Universidad Iberoamericana, Mexico), and Duncan Thomas (Duke University), with funding from the Ford Foundation and other public and private sources in Mexico and the United States. The main objective of the survey is to gather information on the dynamics of socioeconomic indicators, demographic and health indicators, and population migration over a period of at least 10 years (see, e.g., Genoni et al., 2011). The MxFLS is the first Mexican survey with a longitudinal structure that attempts to track participants regardless of migration decisions. The baseline sample of participants selected in 2002 was a probability sample selected from the national population in Mexico at that time.
The MxFLS consists—thus far—of three waves. The first wave (MxFLS-1) or baseline was administered in 2002. The target population was all households in Mexico in 2002. The sample for MxFLS-1 was selected by INEGI and was a multi-stage, stratified probability sample of the Mexican population in 2002. The baseline sample consisted of approximately 35,000 individuals living in approximately 8,440 households.
The second wave of the survey (MxFLS-2), which was conducted in 2005 and 2006, was designed to follow all baseline participants and their children born after 2002 (“panel respondents”). In addition, all other people living with panel respondents at the time of the re-interview were included in the survey. The re-interview rate was about 90 percent. After adding new children and co-residents, the sample in MxFLS-2 included approximately 37,000 respondents. A salient aspect of the MxFLS is that it attempts to follow panel respondents regardless of their migration decisions. Accordingly, survey administrators tried to contact those panel respondents thought to have migrated to the United States between waves 1 and 2. Of those, 91 percent (or 774 people) were located and re-interviewed in the United States.
The third wave of the survey (MxFLS-3) went into the field in 2009 and was still ongoing as of spring 2012. The intent is to re-interview as many of the original panel respondents as possible. To date, the re-interview rate is approximately 85 percent; after including children born after the first
re-interview (MxFLS-2), the current number of respondents in MxFLS-3 is approximately 38,000. Attempts to track down all panel respondents who have not yet been located continue. MxFLS-3 participants include individuals who never left Mexico, those who migrated to the United States before MxFLS-2 and are still in the United States, those who migrated to the United States after MxFLS-2, and those who migrated to the United States sometime between MxFLS-1 and MxFLS-3 but have since decided to return to Mexico.
A unique aspect of the MxFLS is that it gathers the same type of social-demographic information (including health, reproductive decisions, education, and other data) on Mexicans who have stayed in Mexico and those who have decided to migrate. To date, about 1,179 panel respondents have been re-interviewed in the United States. It is estimated that they constitute about 80 percent of the panel respondents who migrated and still reside in the United States. Approximately half of the panel respondents who were in the United States at the time of MxFLS-2 had returned to Mexico by MxFLS-3.
Every panel respondent is asked several questions that attempt to capture migration history at the personal level. Respondents are asked about their place of birth and about their place of residence at age 12. After age 12, respondents are asked to provide more detailed information about moves within Mexico and between Mexico and other countries. For each move that lasts more than 1 year, respondents are asked for the migration date, destination (locality, municipality, state, and country), reason for leaving, people moving with the respondent, and source of funds. If the respondent moved to the United States, the type of documentation (none, visa, green card, citizenship) carried by the respondent is also recorded. For the 2-year period preceding a survey wave, respondents are asked to report the same information, but only for moves that last a month or longer. Finally, respondents are also asked prospective questions about their intentions to move. If they plan on moving sometime in the future, the likely destination, reason for the move, and the existing networks in their destination are recorded.
MEXICAN MIGRATION PROJECT
The MMP is a binational research effort that is codirected by Jorge Durand (Professor of Social Anthropology at the University of Guadalajara) and Douglas S. Massey (Professor of Sociology and Public Affairs at Princeton University).26 Since its creation in 1982, the MMP has focused
26 The description of MMP that follows is based largely on information from the public website http://mmp.opr.princeton.edu/research/design-en.aspx.
on gathering social, demographic, and economic information on Mexican households and their U.S. migration experience. The MMP is supported by grants from the National Institute of Child Health and Human Development and the William and Flora Hewlett Foundation. MMP data are made available to the public through a website housed at Princeton University.
Each year during December and January (when seasonal and other migrants typically return home), the MMP randomly samples 200 households in several communities located throughout Mexico (for a total of 600 to 1,000 households interviewed per year). MMP communities are not randomly selected but are chosen to reflect a broad range of sizes, locations, urbanicity, and migration prevalence. Once communities have been chosen, frames are constructed by the enumeration of all households in smaller communities or all households in a neighborhood in larger metropolitan areas. Households are randomly selected within communities from among eligible households.
After gathering social, demographic, and economic information on the household and its members, interviewers collect basic information on each migrant’s first and last trip to the United States. From household heads who have migrated in the past, they compile a year-by-year retrospective history of U.S. migration and administer a detailed series of questions about the last trip northward, focusing on employment, earnings, and use of U.S. social services.
Following completion of the Mexican surveys, U.S.-based samples are gathered in some cases. Using tips from community contacts where available, interviewers travel to destination areas in the United States the following summer to administer identical questionnaires to migrants from the same communities sampled in Mexico; the exact number of migrants sampled (typically 10-20) depends on how many, if any, can be identified. These are permanent migrants who have settled north of the border and no longer return home. U.S.-based sampling is intended to generate a binational sample that is representative of permanent and return migrants. However, while most of the MMP communities sampled before 2000 have a U.S. counterpart, the majority sampled in 2000 and later do not. Because the U.S. sample is very small and limited to certain years, the MMP disproportionately represents return migrants and underrepresents permanent emigrants.
The data include community-based weights that reflect the community population. In the Mexican sample, the weight is calculated as the inverse of the sampling fraction where the number of households interviewed is divided by the number of eligible households in the predefined survey area from which the 200 surveyed households were drawn. The U.S. weight is also the inverse of the sampling fraction; the population size is estimated by comparing the number of adult children among surveyed household heads
who settled in the United States with the number who stayed in Mexico. Once the data are weighted, they are representative only of the population in the predetermined sampled communities, not of Mexico or of any well-defined geographic area in Mexico.27
The MMP contains detailed information on U.S. migration, including timing and location of illegal border crossings; number of failed attempts; smuggler usage and cost; past trips; U.S. destination, occupation, wages, and social ties; and duration of trip. It includes data on nonmigrant households as well, which is useful in modeling the migration decision. MMP data have been thoroughly documented and are widely used in the migration literature (see, e.g., Durand and Massey, 2004), but they are almost never used to measure migration flows.
THE MEXICAN MIGRATION FIELD RESEARCH PROGRAM
The Mexican Migration Field Research Program (MMFRP) was established in 2004 at the Center for Comparative Immigration Studies (CCIS) at the University of California, San Diego, which remains a cosponsor. It has been funded through multi-year grants from the Ford Foundation, the University of California Office of the President, and smaller extramural grants. By following migrants in rural Mexico and the United States over time, the MMFRP seeks to document and explain changes in their migration and settlement behavior.
The MMFRP follows three small, rural Mexican communities and their U.S. satellite communities over time.28 One community is interviewed per year and is then re-interviewed at two- or three-year intervals. Since 2004, the study has rotated between the three cities: San Miguel Tlacotepec, Oaxaca; Tlacuitapa, Jalisco; and Tunkas, Yucatan.29 The three small communities of fewer than 3,000 people were chosen to be “broadly representative of high-emigration communities in west-central and southern Mexico” (Cornelius et al., 2009:x). Similar to the MMP, the MMFRP administers household surveys to residents in the small towns at times when seasonal migrants are most likely to be present, typically in the months of January or February. The U.S. part of the survey is administered using a “snowball” technique, where participants in Mexico give contact information for their friends and family in the United States so that they may also answer
27 See The Mexican Migration Project Weights on the MMP website: http://mmp.opr.princeton.edu/databases/studydesign-en.aspx.
28 One community, Las Animas, Zacatecas, was interviewed just once in 2005.
29 Tlacuitapa was interviewed in 2005, 2007, and 2010; Tunkas in 2006, 2009, and 2012; and Tlacotepec in 2007-2008 and 2011.
The size of the combined Mexico- and U.S.-based sample is typically based on 700 to 1,000 interviews per year. The survey usually targets household heads, but it also interviews other adults (ages 15-65) in the household. Since a census is essentially taken within the community, there is no sampling error and hence there are no sample weights. Refusal rates tend to be less than 10 percent.
The focus of the interviews and, hence, the survey questionnaire changes every year. This allows the MMFRP to address the most recent developments, such as the impact of the U.S. recession on migration or the deterrent effect of border enforcement. Although the annual themes differ, the surveys consistently include information similar to the MMP on social, demographic, and economic variables. In addition, it asks about intended and actual migration, networks, remittances, perceptions of border enforcement and other immigration laws, and legal status. For the undocumented, it asks about how the border crossing was made, coyote usage and price, and any abuse or mistreatment, if applicable (for first and last trip, as in the MMP). The MMFRP also asks migrants why they return to Mexico or why they stay in the United States. Other information varies by survey year. At one time or another, questions have included welfare program usage in Mexico, race relations and ethnic identity, politics and civic participation, religion, future plans, and more.
The MMFRP data and documentation have recently been made available on the CCIS website. Researchers who participate in the MMFRP program in a given year write up their results as chapters in a book that focuses on that year’s survey theme. (See, e.g., Cornelius et al., 2010, which examines how the U.S. economic recession that began in 2007 has affected flows of Mexican immigrants to and from the United States.30)