Migration between Mexico and the U.S. estimated from a border survey
Michael S. Rendall,* Emma Aguila,* Ricardo Basurto-Dávila,*
and Mark S. Handcock**
7 May 2009
Abstract
The Mexico-U.S. migration flow is one of the world’s largest, but is also among the more
difficult to quantify and capture in survey sources. In this context, the Survey of
Migration at the North Border of Mexico (EMIF) offers a unique source of information
on both regulated and unauthorized components of flows of Mexican-born migrants to
and from the U.S., from 1993 to the present. The survey is conducted using probabilistic
sampling methods at transit points of the eight main border-crossing cities of Mexico.
The EMIF has been used very little in the international scholarly literature, possibly in
part because its statistical properties are not well established. We evaluate the EMIF here
by comparison to alternative estimates of emigration and return migration from national
household surveys and censuses in Mexico and the U.S. We find the EMIF’s primary
strength is in capturing returning male migrants of working ages. A secondary strength is
in its capturing male emigrants at all but the younger working ages. Its estimates of male
emigrants are double those of U.S. data sources (in which they appear as immigrants).
We attribute this to better capture of unauthorized and circular migrants in the EMIF. Its
coverage of female emigrants and return migrants is less reliable, but appears to have
improved in the early 2000s. The EMIF represents reasonably well the geographic
origins and educational attainment of Mexico’s migrants to and from the U.S., but
captures less educated migrants better than more educated migrants.
* RAND Labor and Population program and Population Research Center. Please address
correspondence to the first author at mrendall@rand.org.
** Department of Statistics, University of Washington, Seattle
Acknowledgements:
This paper was prepared for presentation at the 2009 Annual Meeting of the Population
Association of America, Detroit, and the 2009 Metropolis British Columbia and Center
for Research and Analysis of Migration Workshop on the Economics of Immigration.
We gratefully acknowledge support from the National Institute of Aging under
investigator grant R21AG030170, and from the National Institute of Child Health and
Human Development research infrastructure grant R24-HD050906, and thank Sarah
Kups for her valuable research assistance.
I. Introduction
The challenges of quantification and substantive analysis of international migration flows
have increased in many countries as unauthorized migration has come to account for an
increasingly large share of migrants (e.g., Carling 2007). The total and unauthorized
Mexico-U.S. migration flows are among the world’s largest, but have long been difficult
to quantify and capture due to the large unauthorized component. Over the last 15 years,
U.S. data sources indicate that between 400,000 and 600,000 people per year migrated
from Mexico to the U.S. (Passel and Suro 2006). The majority of these migration events
are likely to be unauthorized border crossings (Passel and Cohn 2008). Estimates from
Mexican data sources (INEGI 2008), however, indicate recent migrant flows that are
approximately twice these magnitudes. Discrepancies between data sources at the origin
and destination countries are unfortunately very common in data sources on international
migration (Poulain 1993). The sizes of return U.S.-Mexico flows are also very large, and
estimates of these again are highly variable depending on data source used. They have
been estimated to be anything from almost as large as the Mexico-U.S. flow (Massey and
Singer 1995) to less than half as large (INEGI 2008).
Not only the numbers of migrants, but also their characteristics, migration
histories, and their behaviors in other domains including work and family are of interest.
Reflecting this interest, a number of specialist surveys have been fielded in Mexico
allowing in-depth insights about the Mexico-U.S. migration process and its relationship
to both Mexican and U.S. social and economic conditions. Only two surveys, however,
have been fielded with the frequency and regularity to allow for exploration of individual
and environmental processes behind changes in the migration process. The best known
1
of these, and the source of large amounts of scholarship, is the Mexican Migration Project
(MMP, Durand and Massey 2004a). The MMP is a probability survey of the main
migrant-sending areas of Mexico, expanding over time in its coverage. The frequency
and regularity of data collection in the MMP has allowed analyses such as changes in
duration of stay in the U.S. (Reyes 2004; Riosmena 2004). The second, and less well
known, of the two surveys of migrants that have been fielded with high frequency and
regularity is the Survey of Migration at the North Border of Mexico (EMIF, [Encuesta
sobre Migración en la Frontera Norte de México], CONAPO 2006).
The survey design and sampling method of the EMIF is similar to that for the
main survey used for estimating migration flows to and from the United Kingdom, the
International Passenger Survey (IPS, Office for National Statistics 2008). 1 In both the
EMIF and the IPS migrants are first distinguished from non-migrant travelers with
questions that allow the elimination of other flows such as tourists and residents of the
crossing border city whose trips to and from the U.S. do not involve a change of country
of usual residence. The sets of places sampled in the EMIF, however, is broader. While
the IPS samples only ports (air, sea, and bus/train), thereby restricting the sample
essentially to authorized crossing points and authorized migrants, the EMIF additionally
samples transit points for people arriving in the border towns and cities. This allows for
the final sample to include both authorized and unauthorized international migrants.
1
See Rendall, Tomassini, and Elliot (2003) for an evaluation of the IPS against Census and household
survey methods of collection, and Rendall, Wright, and Horsfield (2005) for additional evaluation of the
IPS emigrant flows against receiving countries’ administrative statistics on immigrants from the U.K.
2
Emigrants and return migrants in the EMIF are identified based on their
residential status in the country they left (the U.S.) or are leaving (Mexico) and their
intended period of stay in the U.S. if leaving and in Mexico if arriving. A questionnaire
about the current trip and about migration history and future intentions, and about
selected socio-demographic characteristics and labor-market behavior, is then
administered to the identified migrants. The EMIF excludes from its sample all people
born in the United States, but includes migrants born in Mexico and elsewhere, and
identifies migrants by country of birth (e.g., migrants from Central American countries
traversing Mexico on the way to and back from the U.S.). In the present study, we limit
analyses to Mexican-born migrants.
The EMIF’s strengths are its probabilistic sampling of migrants in both directions,
its coverage of both authorized and unauthorized migrants, its annual periodicity, and its
labor force and remittance questions. It additionally has the potential to estimate
migration over the life course due to its asking age at first labor migration to the U.S. of
individuals both intending to emigrate to the U.S. and of individuals returning from the
U.S. The EMIF has so far been used in studies published in Mexico (e.g., Anguiano
2003; Mendoza 2004), and has begun to enjoy some exposure in studies in the
international scholarly literature (see especially the studies of Amuedo-Dorantes and
colleagues for analyses of migrant remittances and of migration flow changes in response
to changes in border enforcement ---- Amuedo-Dorantes and Poza 2005, 2006; AmuedoDorantes and Bansak 2007). The population representativeness of the EMIF’s migration
3
flows is not, however, well established in the international scholarly literature. 2 Our
main goal in this study is to evaluate the representativeness of the EMIF of migrant flows,
both from Mexico to the U.S. and from U.S. to Mexico. We achieve this by comparing
the size and composition of migrant flows in the EMIF to emigration and return
migration data in national household surveys and censuses both in Mexico and in the U.S.
The remainder of the paper is structured as follows. In section II immediately
below, the strengths and weaknesses of other, household-based data sources available for
the estimation and analysis of Mexico-U.S. and U.S.-Mexico migration is discussed. In
section III, the EMIF data and international migrant definitions and the household census
and survey data sources used to evaluate the EMIF and migrant definitions are described.
In section IV, results of these comparisons are described, followed by conclusions in
section V.
II. Data sources for estimation of emigration and return migration between Mexico
and the U.S.
Reviews of international migration statistics and data sources on international migration
(e.g., Bilsborrow, Hugo, Oberai and Zlotnick 1997; United Nations 2002) note
administrative records, such as population registers and immigration permit data, and
population censuses as being the main sources of data on migration flows. Specialist
household surveys are considered as a further source for immigrant and immigration
analyses. With respect to administrative data on migration flows, immigration permit
2 Amuedo-Dorantes and colleagues, for example, cite only a 1998 study by the Mexican Department of
Labor and Social Welfare of the first, 1993-94 Wave of the EMIF.
4
data are of limited use in the Mexico-U.S. context due to the large number of
undocumented migrants (Passel and Cohn 2008). We review now three alternative
categories of data with potential use in estimating the total migration flows between
Mexico and the U.S: censuses and microcensuses, large-scale general-purpose household
surveys, and specialist household surveys.
Censuses and Microcensuses
Net and gross migration estimates, have been generated from both U.S. and Mexican
censuses and microcensuses. Net migration of the Mexican-born population can be
estimated by two consecutive censuses, as shown by Hill and Wong (2005) alternately
using the U.S. and Mexican Censuses of 1990 and 2000. Gross migration can be
estimated over a five-year interval in both the U.S. and Mexican censuses using the
“country of residence 5 years ago” questions in each country’s Census. Rendall and Torr
(2008) use this question to estimate gross migration rates between the U.S. and Mexico
for children born in the U.S. to Mexican-born mothers. In the U.S.-Mexico migration
context, however, estimation of migration in a shorter than a five-year interval is clearly
desirable given the high frequency of short-term and circular migration between Mexico
and the U.S. (Bean et al 2001).
The 2000 U.S. Census has a “year of arrival” question that identifies migration in
single years, including the 2000 Census year itself and the year 1999. There is debate
about the validity and reliability of responses to this question (Ellis and Wright 1998;
Redstone and Massey 2004), although this has mostly focused on years of arrival further
back from the census year. Passel and Suro (2006) find in the ACS that “year of arrival”
5
and “country of residence one year ago” questions generate similar estimates of
immigrant inflows.
Mexico conducts a microcensus (“Conteo”) in the half-way year between
censuses, including 1995 and 2005. It includes the same migration questions as the
decennial Census. In countries with high rates of emigration as is the case of Mexico,
reports by remaining household members may be used to count emigration events. Wong
Luna, Resano Peréz, and Martínez Matiñon (2006) use the 2000 Census and the 1995
Microcensus [Conteo Nacional de Población y Viviendas] to describe changes in
emigration and return migration. The Mexican Census Public Use Microdata Samples
(PUMS) do not contain information from these more detailed questions. The Mexican
Census PUMS data are therefore limited to the five-year migration intervals (Minnesota
Population Center 2006).
The U.S. has, since 2001, conducted an annual microcensus called the American
Community Survey (ACS, U.S. Census Bureau 2007). The ACS has a “year of arrival”
question that identifies migration in single years, as in the Census. Additionally, however,
the ACS has a question on place of residence one year before. Since 2005, when the
ACS first attained its full planned sample size, the U.S. Census Bureau have used the
ACS to estimate gross migration flows in its annual population estimates series (U.S.
Census Bureau 2009). A disadvantage of any U.S. data source, however, is its likely
undercoverage of unauthorized migrants (Lindstrom and Massey 1994). While, in the
best case, the 2000 U.S. Census is assumed by Warren (2003) to have only 10%
undercoverage of the stock of unauthorized migrants, it is likely that the most recent
migrants (for example, those that have arrived in the past year) will have the greatest
6
levels of undercoverage. A much greater undercoverage of immigrant flows versus of
immigrant migrant stocks is found, for example, in the large-scale European Labor Force
Survey (Marti and Rodriguez 2007).
Large-scale general-purpose household surveys
General-purpose household surveys are often considered to have serious deficiencies for
capturing migration, although this view is not universally held. On the negative side,
Bilsborrow et al (1997, p.239) argue that “[t]he desirability of concentrating on recent
migrants when analysing the causes or consequences of international migration implies
that general purpose surveys are usually not useful because of the small numbers of
recent migrants covered.” Added to this statistical efficiency problem are language
difficulties and legal-status reasons for new immigrants not responding to household
surveys (Martí and Ródenas 2007). This latter argument does not apply, however, to
returning migrants. Rendall, Tomassini, and Elliot (2003) find evidence for good capture
of returning migrants in some of the same European Labour Force Surveys that Martí and
Ródenas (2007) find to be inadequate for capturing overall immigrant inflows. We use
this observation in support of our use of a similar large-scale household survey in Mexico
to capture return migrants in the Mexican Employment Survey (ENE), and thereby to
evaluate return migration flows in the EMIF (discussed below).
The ENE is similar in purpose and structure to the U.S. Current Population
Survey (CPS). Both include a rotation group structure in which residences are
interviewed in successive quarters. The ENE’s successor survey from 2005 onwards, the
National Survey of Employment and Occupation (ENOE) has questions identifying both
7
individuals who arrived since the last quarter and who departed since the last quarter.
The Mexican National Institute of Statistics used these data to estimate recent trends in
emigration and immigration (INEGI 2008). Passel and Suro (2006) similarly use the U.S.
CPS to estimate immigration from the question on place of residence one year ago.
These estimates are generally lower estimates than those in the ACS and Census,
presumably due to greater undercoverage in the CPS. In a variant on the indirect
estimation of emigration from consecutive censuses, Van Hook, Passel, and Zhang
(2006) use the observation of the Mexican-born population in successive quarters of the
CPS both at the same residence and elsewhere in the U.S. to estimate emigration of the
foreign-born population, including specifically the Mexican-born population.
Specialist surveys
An additional group of surveys with national coverage in Mexico is found in specialist
surveys of migration and socio-economic and health behavior. Of these, the one with the
largest sample size is the occasional large-scale demographic survey called the National
Survey of Population Dynamics (ENADID). It was conducted in 1992, 1997, and 2006
and includes reports of emigration from and return migration to Mexico in the five years
before the survey. It dates the timing of the last emigration or return migration event
occurring to a current or former household member in this five-year window, and
provides a count of the number of emigrations inside the five-year period for each
migrant. The major limitations of the ENADID are its infrequent periodicity, and that it
captures the emigration only of individuals who have ongoing attachment to current
Mexican households. As Wong et al (2006, p.14) note, this does not represent the full
8
population especially of female Mexican emigrants to the U.S. Missed are the emigration
of individuals in single-person households, emigration of complete households, and
emigration that occurs to individuals in households that dissolved between the migration
event and the survey. Hill and Wong (2005) compare ENADID results with residual net
migration estimates from the 1990 and 2000 censuses alternately of Mexico or the US.
They find ratios of male to female emigrants in ENADID that are more than double those
of their residual estimates and conclude that the ENADID underestimates female
emigration due to the greater likelihood of women’s settling in the U.S. with their family,
and therefore being lost to the ENADID’s Mexican household sampling frame.
Smaller specialist surveys have been designed to oversample the main migrantsending areas of Mexico. Two recent examples of household surveys that, while being
national in scope, oversample the high migrant-sending regions, are the Mexican Health
and Aging Study (MHAS) of individuals aged 50 and above and the Mexican Family Life
Survey (MxFLS) of individuals of all ages. Aguila and Zissimopoulos (2008) use the
MHAS to analyze age patterns of return migration differentiating between those with and
without U.S. citizenship, permanent residence, or other legal status in the U.S.
Rubelcava et al (2008) use the MxFLS to analyze the relationship of health to emigration
from Mexico to the U.S.
In the domain of surveys that limit their sampling to major migrant sending areas,
and that therefore do not cover the national population in Mexico, the Mexican Migration
Project (MMP, Durand and Massey 2004a) stands out as being by far the most important
to the scholarly literature in the U.S. It has been a key source of data for estimation,
theoretical development and testing of explanations for Mexican-US migration over the
9
recent decades (e.g., Lindstrom 1996; Cerrutti and Massey 2001). Major contributions of
the MMP for migration flow estimation have been to show the high frequency of
temporary and repeat migration among both authorized and unauthorized Mexican
migrants, and to estimate and distinguish the flows of authorized and unauthorized
migrants (Riosmena 2004; Reyes 2004).
The MMP data, however, need to be used with caution for migrant flow
estimation due to their being only partly based on probability sampling. They are neither
a probability sample of the national population of Mexico nor of Mexicans in the U.S. In
Mexico, the MMP uses random sampling methods to select households within nonrandomly selected Mexican communities. Durand and Massey (2004a) report the MMP’s
representativeness to be very good when compared to a probability sample of returning
migrants in Mexico’s National Survey of Population Dynamics (ENADID). In one way
this is less reassuring than it might otherwise be, since the ENADID also samples
migrants from a universe consisting only of those who have returned to Mexico, or whose
migration between the U.S. and Mexico if they are not currently living in Mexico is
reported by family members remaining in Mexico. Its sample definition is therefore
similar to that of the MMP, but on a national scale. Both surveys deviate from a sample
definition that covers all migrants. Using the MMP, Kana’iaupuni (2000) reports that
married or cohabiting women were far more likely to migrate after their partner than
either before, or in the same year as, their partner. This gives reason for having greater
confidence in the ENADID, and so also the MMP, for estimating men’s than women’s
emigration and return migration.
10
The MMP is also more suited to the analysis of the return migration of those with
strong household attachments in Mexico than it is for the estimation of overall return
migration flows or rates. Because it is largely a Mexican sending community sample, the
MMP has stronger statistical properties for the estimation of Mexican emigration to the
U.S. than for the estimation of Mexican-born US-resident rates of return migration to
Mexico. Evidence for potentially large biases may be seen when examining the estimates
of Massey and Zenteno (1999) and the samples those estimates were derived from. Their
return migration rates to Mexico display a highly unusual shape, increasing
monotonically with age to levels as high as 40 return migrants per hundred annually for
Mexican-born men’s and women in their 60s. Such rates applied year after year to a
cohort would imply almost no aging in the U.S. among Mexican-born immigrants. They
derived these rates from a combination of a total of 2,961 migrants from probabilisticsampling households in the Mexican communities, of whom 96 percent were male, and
addtional U.S.-resident returning migrants from a snowball sample of 415 households (p.
5331-5332). The large gender imbalance in the community-survey sample may reflect a
bias against finding women due to their longer stays in the U.S.
III. EMIF Data Evaluation
The data we evaluate are from the 1993/94 to 2003/04 years of the Survey of Migration
in the North Border of Mexico (EMIF, CONAPO 2006) [Encuesta sobre Migración en la
Frontera Norte de México], the years for which data have been released publicly. For
tabulations of migration extending through 2006, see CONAPO (2008b). The data we
use to evaluate the EMIF are, in Mexico, (1) the 1997 National Survey of Demographic
11
Dynamics (ENADID, INEGI 2003) [Encuesta Nacional de la Dinamica Demografica];
(2) the 2000-2004 National Survey of Employment (INEGI 2000a) [Encuesta Nacional
de Empleo]; and in the U.S., (3) the 2000 Census Public Use Microdata Sample (PUMS,
Ruggles et al 2004), and (4) the 2000-2004 American Community Survey (ACS). These
are chosen as being the nationally representative household surveys with questions and
large enough sample sizes to allow for the estimation of migration flows in a single year.
Deriving accurate estimates in both the EMIF and Mexican comparison data
sources presents challenges in part due to uncertainties related to documentation of the
survey data sources and difficulties finding benchmark estimates of migration by the data
providers or other users of the data. For the EMIF itself, we were able to establish a
match of our estimates with estimates published by CONAPO (2006), and to understand
reasons for differences between our estimates and CONAPO’s due to different definitions
of “migrant.” ENADID 1997 estimates by Wong Luna et al (2006) can also be compared
to those we provide here. 3
We experienced some difficulty in establishing the correctness of our estimates of
return migration in the Mexican Employment Survey (ENE). We could not locate any
comparison estimates using the ENE by the data provider, INEGI, or from other
researchers. We were able to derive estimates of return migration in the ENE’s successor
the National Survey of Occupation and Employment (ENOE) that are very close to the
estimates published by INEGI (2008). Because of the similarity of the ENE and ENOE
questions, this gives us confidence in our conclusion that the EMIF captures return
migration much more fully than does the ENE. We provide a complete description of our
coding of migrants in the ENE and the other data sources below and in Appendix A.
3. A summary of this comparison will be provided in a subsequent version of the paper.
12
Encuesta de Migración en la Frontera Norte de México (EMIF)
The EMIF was created to understand the flow in and out of the U.S. from the Mexican
border by the Mexican National Population Council (CONAPO), College of the North
Border (COLEF), and the Mexican Ministry of Labor (CONAPO 2006). The survey
samples the following groups: a) individuals that have the intention to immigrate to the
US through the Mexican border, b) persons that return voluntarily to Mexico from the US,
c) those that are apprehended by the border patrol and returned to Mexico, and d) internal
migrants from the south to the northern border cities. In the present study, we evaluate
only data collected on groups a) and b), as group c) are assumed not to have taken up
residence in the U.S. and d) are internal, not international, migrants.
International migrants between Mexico and the U.S. that are not explicitly
included in the EMIF migrant universe are (1) people born in the United States, (2) those
that enter and leave the U.S. by airplane or ship without stopping in Mexican cities in the
U.S.-Mexico border, and (3) people that live in the border localities, even when they are
moving permanently to the U.S. For this third group of border-city residents, we are able
to estimate their contribution to the total migrant flows from Mexican household surveys,
and present results on this below. We know of no data source to check on how many
migrants enter the U.S. or return from the U.S. through airports and shipping ports
outside the border cities (the second excluded group), and are able only to address this in
the present study through comparisons of total migrants and migrants by regional origins
in Mexico. Throughout our study, we exclude not only U.S.-born international migrants,
but also international migrants born in other countries (e.g., those born in Central
13
American countries using Mexico as a transit country). The EMIF theoretically covers
all types and ages of migrants, but in practice it is oriented towards working-age migrants.
As we describe in more detail below, it is also more oriented towards identifying
migrants whose purpose of migration was or is for work.
EMIF Sampling methodology
The EMIF has been collected in “waves” each of a year’s duration at approximately
annual intervals since 1993/94. Data from the first nine waves are available for our study.
The first wave of this survey took place between March 28th 1993 and March 27th 1994,
the second from December 14th 1994 to December 13th 1995, the third from July 11th
1996 to July 10th 1997, the fourth from July 11th 1998 to July 10th 1999, the fifth form
July 11 1999 to July 10th 2000, the sixth from July 11th, 2000 to July 10th 2001, the
seventh from July 11th 2001 to July 10th 2002, the eighth from July 11th, 2002 to July
10th, 2003, and the ninth is from July 11th, 2003 to June 30th, 2004. Data collection in the
EMIF is ongoing according to this approximately annual periodicity. The average
sample in each year contains approximately 7,000 individuals traveling north (about 40
percent of them intend to migrate) and 5,000 migrants traveling south (about 50 percent
of them are return migrants).
The methodology of the EMIF has been developed to collect probability samples
with which to estimate the periodical, seasonal, or cyclical journeys of migrants both to
and from the United States (CONAPO 2008). The principle of the sampling plan is based
on probabilistic sampling of times in places of arrival in, or departure from, each of eight
border cities. The first (1993/94) wave of the EMIF was of 23 border localities. Of these
14
23, eight of them included 94% of all migrants. From the second (1994 to 1995) wave
onwards, therefore, only these eight major migration cities were included: Tijuana and
Mexicali in the State of Baja California, Nogales in the State of Sonora, Ciudad Juárez in
the State of Chihuahua, Piedras Negras in the State of Coahuila, and Nuevo Laredo,
Reynosa, and Matamoros in the State of Tamaulipas.
[FIGURE 1 ABOUT HERE]
The EMIF uses a multistage sampling design: in the first stage geographic units are
selected (regions, cities, zones, and sampling points); the second stage involves selecting
temporal sampling units: every quarter, day of the week, and hour; finally, at the selected
place and time, people answer a screening instrument, from which migratory flows are
identified and the respective questionnaire is applied. Probability factors are assigned at
every stage, proportional to the flows of people passing through the geographic unit at the
selected time, with the following exceptions. Regions and quarters receive a probability
factor of one (i.e., flows are sampled in all eight regions and four quarters every year).
Survey days are randomly sampled from the times and days of arrivals and departures of
flights or buses or transportation. For example in bus stations, a sample day consists of an
interval of 24 hours. At the airports the sample day consists of time intervals around the
times for departures and arrivals of planes.
Identification of migrants from among all travelers
15
In the EMIF, migrants are distinguished from non-migrant travelers with questions that
allow the elimination of other flows such as tourists and residents of the crossing border
city whose trips to and from the U.S. do not involve a change of country of usual
residence. The EMIF excludes and identifies migrants born in countries other than
Mexico (e.g., Central American countries traversing Mexico on the way to and back from
the U.S.).
Beyond the above limitations and restrictions of individuals who could possibly
be identified in the EMIF, we try as much as possible to be inclusive of anyone that the
EMIF’s filter questions identify as “migrants.” Our intention is to capture everyone for
whom the full questionnaire about their background and migration experience and
intentions is applied. In general, this means anyone who lived or intends to live in the
U.S. for more than one month, with specific exclusions of some people going to, or have
been in, the U.S. for more than one month but for non-work purposes. We describe the
operation of these filter questions of EMIF for identifying intending emigrants and
returning migrants in the paragraphs immediately below. Except for a relatively small
change in the filter questions between the 1996/97 wave and the 1997/98 wave, the same
questions were used in each EMIF year. Details of the exact migrant filter questions are
given in Appendix A.
In the EMIF sample of individuals traveling “north”, we classified as “intending
emigrants” (intenders) those respondents who expressed an intention to cross into the
United States during their current trip for one of the following reasons: work, look for
work, reunite with their family, or reunite with friends. In addition, we also classified as
intenders those who responded to be traveling to the United States on vacation or to shop,
16
but planned to stay in the US for at least one year. The intending emigrant is identified by
means of a set of five to six questions that allows the recognition in a systematic way, and
that also allows us to eliminate other subjects in the flow including tourists and other
non-migrant travelers, people born in the United States and other countries outside
Mexico. Residents of the border city are also eliminated with these filter questions,
irrespective of whether they live in one side of the border and work on the other side or if
in fact they are moving to the U.S. as migrants.
In EMIF’s sample of individuals traveling from the United States to Mexico, we
classified as “returning migrants” (returnees) those who gave a reason other than vacation
or visiting family to be traveling to Mexico. From those who said to be traveling to
Mexico on vacation or to visit family, we also classified as returnees those who said they
were planning to work while in Mexico, and those who said they were not planning to
return to the US. In addition, we did not classify as returnees those who said they were
in the United States as students, tourists, or shopping and did not work or look for work
during their stay.
Encuesta Nacional de la Dinámica Demográfica (ENADID)
We noted earlier that the National Survey of Population Dynamics (ENADID) was
conducted in 1992, 1997, and 2006. Of these years, only 1997 coincides with our EMIF
sample. We use it to estimate emigration and return migration in the 12 months before
the ENADID 1997 to compare to the EMIF’s 1996/97 year. The ENADID 1997 was
carried out by Mexico’s National Institute of Statistics, Geography, and Informatics
(INEGI) between the months of September and December, 1997 (INEGI 2003).
17
Information was collected on housing characteristics, socio-demographic characteristics
of household members and, among other demographic events, domestic and international
migration of all current and some former household members. These include for labor
migrants only educational attainment and documented versus undocumented work status
in the U.S. 4
Because the ENADID’s questions about migration timing are concentrated on the
“last” migration event, we are able to identify at most one emigration and one return
migration event in the year up to the survey interview. Three sets of questions addressed
the topic of international migration, respectively found in Sections IV, VI, and VII of the
ENADID. We use all three sections in our ENADID comparison estimates to the
1996/97 EMIF migration estimates. We provide in the Appendix section A.3 the
questions that create the universe of individuals eligible for migration questions. Section
IV “eligibles” are those individuals living at the surveyed residence at the time of the
survey or “who normally live [at the surveyed residence],” as identified in Section III (as
“usual residents”). This section is the most inclusive of types of migration events: all
migration events, no matter for what purpose, are included in the Section IV questions
identifying time of last migration.
Section VI “eligibles” are those individuals of age 12 or older, living at the
surveyed residence at the time of the survey or who “normally lived” at the surveyed
residence, even if they were not living there at the time of the survey. Only their
migration events to the U.S. “for work or look for work,” however, are recorded in this
4 In future analyses we plan to compare the EMIF’s composition of documented versus undocumented
emigrants and return migrants with that estimated from the ENADID.
18
section. Section VII “eligibles” are those individuals living at the surveyed residence at
the time of the survey, those who “normally lived” at the residence, and those who had
lived in the residence over the previous five years, even if they did not live there and
were not considered “regular” residents anymore. Their recorded migration events are
explicitly limited in the questionnaire to those “for work or study (to live)” over the
1992-1997 period, although the interviewer manual specifies that they were interested in
capturing anyone who came to live in the U.S. between 1992 and 1997.
We describe now in more detail the three sections’ questions both on migration
and other characteristics relevant to our comparisons of migrants between the ENADID
and EMIF. First, in ENADID’s Section IV, questions were asked about the Mexican
state or foreign country in which a current household member (see question 3.6 in
Appendix A.3 for how this is defined) lived immediately before her current residence.
This section contains information on the length of time the individual lived in her
previous place of residence and the length of time since she moved to her residence at the
time of the interview. In both cases, no information is collected on specific year and
month of migration or return migration; rather, the length of time is measured as number
of years—or number of months if the period was less than one year. This is sufficient to
allow us to code return migration events occurring in the 12 months before the survey to
all individuals living in the household at survey date. Coding of emigration events
occurring in the 12 months before the survey can be done by adding the length of time
the individual lived in the previous place of residence and the length of time since he or
she moved to the current residence (e.g., if a woman moved to her current residence three
19
months before the survey and lived in her previous residence for two months, we can
conclude that she emigrated five months before the survey).
There are two important limitations to using ENADID’s Section IV to estimate
migration flows: (a) these questions are asked only from individuals who lived in the
household at the time of the interview, thus missing return migrants who have migrated
again; and (b) because information was collected only about the most recent episode of
either domestic or international migration, an international migrant who returned to
Mexico and subsequently moved to another Mexican state is captured only as domestic
migrant, not as an international migrant. Furthermore, any returning migrant in the last
year who subsequently re-emigrated to the U.S. will be omitted from the count of return
migration events.
The second set of international migration questions was asked in ENADID’s
Section VI. These questions focus on labor migration to the United States. No migration
events other than “for work or to look for work” are identified in this section. The month
and year of the last emigration event and the month and year of the last return migration
event are both asked. Additionally asked in Section VI, however, are the individual’s
educational attainment and whether the individual had documents authorizing him or her
to work in the United States.
Finally, ENADID’s Section VII collects information on international migrants in
two parts. The first part considers only those individuals listed in Section III as members
of the household (“who normally live here”), including those who in that Section were
identified as living elsewhere at the time of the interview or “living elsewhere even if
currently staying here.” The second part identifies international migrants nowhere else
20
identified in the ENADID. These are individuals who lived in the household sometime in
the last five years but were no longer considered household members at the time of the
interview. For both groups, the questions are restricted to those who emigrated after
January 1992. Their emigration events coded, moreover, are not restricted on reason for
migration. The questions in this section contain information on gender, age at the time of
the last migration, month and year of last migration, country of destination, number of
migration episodes, and month and year of return (if any).
One important limitation for return migration flow estimates from this section is
that it only captures return migration flows of individuals who last emigrated after 1992
(thus, a person who migrated in, say, 1990 and returned in 1997 will not be captured in
estimates of return flows during 1997 that use this section). Also, for the group of
individuals who are no longer considered members of the household, the only
information available from them is that collected in these questions (not including, for
example, their educational attainment).
To provide the best possible estimates of emigration and return migration flows
between Mexico and the United States derivable from the ENADID, we combined data
from all three sections, using the following procedure:
1.
For individuals captured in all three sections, only in sections VI and VII, or only
in section VII, the dates of last migration and last return migration (if any) were taken
directly from section VII, as it provides the specific month and year for these events. 5
5. Both sections VI and VII of ENADID contain specific information on month and year of migration and
return migration, so for individuals captured in both of these sections either of both dates could have been
used. For the great majority of them, however, the dates were identical and thus the choice does not result
in meaningfully different estimates.
21
Age at the time of migration and return migration was calculated using the information in
this section as well.
2.
For respondents captured only in sections IV and VI, or only in section VI, the
dates of last migration and last return migration were taken from Section VI, which also
provides specific month and year of migration and return migration. Age at the time of
migration and return migration were calculated using these dates.
3.
Finally, for those individuals who only appear as international migrants in section
IV, the dates of last migration and last return were computed using the information in that
section. Since specific month and year are not available in the data, they were estimated
using the information provided on the length of the last migration spell and the length of
time since the individual last returned to Mexico.
Since combining all three sections results in a sample with no information on the
education for some individuals (those in section VII who are no longer household
members), a similar procedure was conducted to combine only sections IV and VI, in
order to obtain estimates of the distribution of education for migrants and return migrants
that can be compared to similar estimates from EMIF.
Encuesta Nacional de Empleo (ENE) Survey
We use the 2000 to 2004 Mexican Employment Surveys (ENE) to compare the EMIF’s
return migrant estimates. The ENE were, until the introduction of the ENOE in 2005
(INEGI 2005), conducted separately in the main urban areas (ENEU, INEGI 2000b) and
in the rest of the country (ENE, INEGI 2000a). Beginning in 1998, the two surveys were
designed so that their aggregation could produce nationally-representative population
22
estimates (INEGI 2000b). Since the third quarter of 2000, the survey has been conducted
quarterly. The sample sizes are very large. The ENE 2000, for example, includes the
results of 163,838 households and 436,344 individual observations. The response rate of
the employment surveys is high, on average 86.8% since 1995.
There are two key questions that provide data on return migrants from the U.S.
(see Appendix A): “residence status” (Condición de residencia) and “migrant origin”
(Migración). One of the codes for residence status is “new migrant resident” (Residente
nuevo inmigrante). “New” is in relation to the previous wave of the survey. Each
residence receives five quarterly interviews. In the second through fifth interviews, any
new member of the residence is coded as either a “new migrant resident” or a “new nonmigrant resident” according to whether the move was from outside the area (e.g., city or
metropolitan area), including moves from outside the country. To reduce the likelihood
that such moves are missed in the survey, interviewers are instructed to code “new
migrant resident” in the first wave that the individual is picked up by the survey in the
case that the individual had been omitted in error after having arrived as a new resident in
the previous wave (see Manual for Interviewers INEGI 2000c, p. 37 and p. 87). We
estimated return migrants in the ENE as follows: classify as return migrants if “condicion
de residencia” is “new resident, immigrant” and “migracion” code is one of the three U.S.
location codes, 334, 335, or 339.
Weighting the ENE estimates of return migrants to obtain annual estimates for the
population is complicated by the problem that in any given quarter only for households in
their second through fifth interviews are we able to identify migrants. The ENE’s sample
weights, meanwhile, are constructed to estimate population totals for households in their
23
first through fifth interviews. Because the groups of households in their first through
fifth interviews are, in general, equal sized, the solution is simply to scale up the
aggregates to account for the missing fifth by multiplying all ENE-weighted aggregates
of migrants by 1.25.
Like the EMIF, and unlike the ENADID, the ENE potentially counts more than
one return migration event per year for a given individual. In practice, however, this is
made unlikely given the criteria for identifying return migrants in the ENE as having
been resident elsewhere to be considered a “new resident” in a given quarter. We found
this to be a rare occurrence too when we used the ENE’s successor, the ENOE of 2005, to
count how frequently the same individual appears in two different quarters as a “new
immigrant” coming from the U.S. (results not shown).
U.S. Census and American Community Survey (ACS)
Both the Census and ACS have been used previously to estimate migration flows from
Mexico (e.g., Passel and Suro 2006). We use the 2000 U.S. Census PUMS (U.S. Census
Bureau 2006; Ruggles et al 2004) questions on place of birth and year of arrival (see
Appendix A for an extract of the questions). The Census was conducted as at April 1
2000. We approximate the annual immigrant flow with those Mexican-born U.S.
residents at Census time whose year of arrival was 1999.
The American Community Survey (ACS) has been conducted since 2000. In 2005,
the ACS became fully implemented as a household microcensus and now samples three
million addresses. The Census Bureau has used the ACSs of 2005 and later years to
estimate annual migration using the question on place of residence one year ago (U.S.
24
Census Bureau 2009). To match the EMIF years, we use the ACS years 2001-2004,
when its sample size was somewhat smaller. We identify Mexican-born immigrants to
the U.S. in the ACS Public Use Microdata Sample (IPUMS USA, 2009) by their place of
birth and place of residence the previous year. We found little difference between
estimates from this definition and from using the year of arrival variable similar to that in
the 2000 Census (as seen also in Passel and Suro 2005).
The analyses of Passel and Suro (2006) suggest that coverage of the Census is
more complete than the ACS, while both are more complete in their coverage than is the
CPS. Both the Census and ACS will, however, undercount Mexican migrants to some
degree because of factors including their high prevalence of unauthorized statuses.
IV. Results
We present comparative estimates of total Mexico-U.S. annual “emigrant” flows of
Mexican-born migrants aged 15 and over between: (1) the 1996/97 EMIF and 1996/97
ENADID (men only); (2) the 1999/2000 EMIF and the 2000 U.S. Census; and (3) the
1999/2000 to 2003/04 EMIF and the 2000 to 2004 ACS. All these comparisons involve
probability samples (including the Census PUMS). We present comparative estimates of
total U.S.-Mexico annual “return migration” flows of Mexican-born migrants aged 15
and over in: (1) the 1996/97 EMIF and 1996/97 ENADID; and (2) the 1999/2000 to
2003/2004 EMIF and the 1999/2000 to 2003/2004 ENE.
We use the appropriate sampling weights provided with each data source. In the
present version of the paper, no confidence intervals or statistical tests of difference
between the EMIF and comparison data source are included in the main results. These
25
will be included in a subsequent version of the paper. The most difficult problem is
estimating sampling variance around the EMIF migration estimates. A method for doing
this is proposed in Appendix B, and example estimates of variance about 1996/97 EMIF
return migration are given there. Sampling error about migrant totals in household
surveys in the receiving country have been found elsewhere (Rendall et al 2003) to be
much higher than under simple random sampling, due to clustering of new immigrants in
households. Following Rendall et al, we plan to use resampling methods to estimate the
true sampling error around the household survey estimates (ACS, ENE, and ENADID).
A. 1996/97 EMIF comparisons to the 1997 ENADID
[TABLE 1 ABOUT HERE]
1996/97 EMIF versus ENADID Estimates of Return Migration Flows
We first examine how EMIF’s estimates of return migration compare to those from
ENADID. From both surveys, we obtain estimates by gender of total migrant flows,
average age at the time of return, age groups, education groups, and region of destination
in Mexico. The results of these estimations for the periods 1996-1997 are shown in
Tables 1. The EMIF estimate of the total return migration flow between 1996 and 1997
for males is remarkably similar to those from ENADID (276,901 vs. 293,492), while the
EMIF estimate for females is grossly underestimated relative to that found in the
ENADID (13,352 versus 85,670). These weighted estimates are derived from almost
identical unweighted counts of return migrants for men (1,172 EMIF and 1,173
26
ENADID), but from a count of only 52 return migrant women in the EMIF in the
1996/97 year (versus 316 in the ENADID).
[FIGURE 2 ABOUT HERE]
The age composition of return male migrants is very similar between the EMIF and
ENADID (see also Figure 2). The mean age of returning male migrants is 32.2 in the
EMIF and 31.8 in the ENADID. As many as 43.2% and 43.7% respectively of all return
migrants are aged 20-29 respectively in the EMIF and ENADID, and 28.7% and 29.8%
respectively are aged 30-39.
When comparing the distribution of Mexican-born migrant origins, we use the
standard definition of areas, including the “traditional” migrant sending states.6 We also
divide the “North” region into the EMIF cities (localities) and all other localities in the
Northern status. These “traditional” migrant sending states account for approximately
half of all returning migrants in both the EMIF (52.1%) and the ENADID (50.7%). The
EMIF appears to slightly overestimate males returning to Mexico’s north region outside
the EMIF cities (22.6% versus 16.4% of the total), and underestimates those traveling to
the central region (12.7% versus 18.4%). As expected, the EMIF underestimates return
6. Regional groupings of Mexican states are as follows: North: (BC, BC Sur, Coah, Chih,
NLeon, Sin, Son, Tams); Traditional: (Ags, Col, Dur, Guan, Jal, Mich, Nay, SLP, Zac);
Central: DF, Hgo, Mex, Mor, Pue, Qro, Tlax); South/SE: Cam, Chis, Guerr, Oax, QRoo,
Tab, Ver, Yuc). 27
migrants to EMIF sampling cities due to the design of the survey to intentionally exclude
the residents of these border cities.
The educational composition of returning male EMIF migrants is more oriented
towards lower-education migrants. While 37.8% of EMIF returners did not complete
primary school, this group comprises only 28.9% of the ENADID’s returners.
Conversely, while 13.7% of ENADID’s returning male migrants had completed high
school, only 8.5% of EMIF returners had this level of education. At least two potential
explanations should be considered to explain the lower educational attainment of EMIF
than ENADID migrants. First, the ENADID’s return migrant sample is limited to those
migrating to the U.S. for work. Approximately 30,000 (10%) of the ENADID return
migrant population were in the U.S. other than for work. Analysis of their age
composition might indicate whether these were student (high educated) or retired (more
likely to be low-educated). Second, the ENADID includes migrants returning by air to
other than the border cities, and these may be more likely to be more educated.
[FIGURE 3 ABOUT HERE]
Concerning female return migrants, these are severely underestimated in the EMIF at
every age (see also Figure 3). Regarding regional distributions, only for the North region
(excluding EMIF sample cities) is the EMIF estimate greater than one third of the total
estimated in the ENADID (6,247 in the EMIF versus 16,491 in the ENADID). This is
half (50.5%) of all returning migrant females in the EMIF. This may indicate a much
greater likelihood of returning female migrants in other than the Northern region of
28
Mexico to use air travel to central and southern parts of the country. It also seems likely,
however, that the EMIF has other biases against identifying female return migrants. The
orientation of the survey towards labor migrants may be one factor here.
In sum, EMIF appears to capture returning migration flows of Mexican-born men
well, particularly those with the lowest levels of education, when we consider ENADID
as a benchmark. The 1996/97 EMIF, however, severely undercounts female return
migrants.
1996/97 EMIF versus ENADID Estimates of Male Emigration Flows
The results of comparing male emigration between the EMIF and ENADID are reported
in Table 2. Over the one-year period 1996-97, the EMIF’s male emigration estimates are
33% lower than the ENADID’s (423,834 vs. 632,647). Unlike return migration, average
age and the distribution of age groups varies substantially between the EMIF and the
ENADID. Male emigrants in the EMIF were on average 32.0 years old, while those in
the ENADID were of average age 29.4. Examining the distribution of age groups
suggests that difference arises because EMIF captures a higher proportion of male
migrants in the age groups 30-39 and 50-59, and lower in age groups 15-19 and 20-29.
The differences in estimates of education categories are less pronounced. The EMIF and
ENADID capture similar proportions of male migrants with mid-level education, and the
differences in proportions of low- and high-education migrants are not large.
[TABLE 2 ABOUT HERE]
29
As for the distribution of the regional origins of male migrants, the EMIF and ENADID
survey similar percentages of migrants from the north and south/south east regions, but
the proportion of migrants from traditional migrant-sending states is larger in EMIF than
in the ENADID, while the opposite is true for migrants from the central region. In terms
of the EMIF sampling cities, the number of individuals intending to emigrate to the U.S.
is underestimated due to the design of the survey, intended to eliminate border-city
residents. The ENADID shows that the proportion of emigrants from the EMIF sampling
cities is only 2.6%. Even while the EMIF underestimates emigrants from its sampling
border cities, moreover, the difference is small.
[FIGURE 4 ABOUT HERE]
The distribution of ages among male emigrants to the U.S. is similar in shape between the
EMIF and ENADID as Figure 4 shows. There are underestimates, however, seen in the
EMIF up to age 60. The estimates among the oldest groups of intending emigrants are
similar between the EMIF and ENADID.
B. EMIF emigration compared to emigration in the 2000 U.S. Census and the
American Community Survey (ACS)
In this section we compare emigrants from EMIF 1999/00 with the 2000 U.S. Census, the
EMIF 2000/01 with the ACS 2001, the EMIF 2001/02 with the ACS 2002, the EMIF
2002/03 with ACS, 2003, and the EMIF 2003/04 with the ACS 2004 (see Table 3). The
EMIF estimates a larger total number of emigrants from 1999/00 to 2003/04 in
30
comparison the estimates of the 2000 to 2004 ACSs. We observe a decline in the number
of total emigrants in EMIF from 1999 to 2001 and a sharp increase in 2002 and 2003.
ACS shows a decreasing trend in the total number of emigrants from 1999 to 2002 and
captures an increase in the number of emigrants until 2003. The average age in EMIF is
higher than in ACS estimates. The latter is due to EMIF undercounting the youngest
groups from 15 to 19 years old, as was seen also in comparison to the 1996/97 ENADID.
[TABLE 3 ABOUT HERE]
As for the EMIF to ENADID comparison, the total number of female emigrants in the
EMIF is low in comparison to ACS. The difference between the EMIF and ACS is much
less, however, in part due to the substantially greater numbers of EMIF female emigrants
in the early 2000s than in the 1996/97 period. The EMIF shows a moderate increase in
the total number of female emigrants from 1999 to 2002 but a sharper increase in 2003.
The ACS shows a moderate increase in female emigrants from 1999 to 2000, a moderate
decline from 2001 to 2002, and a slightly sharper increase in 2003. The ACS also
captures an almost constant trend in the number of female emigrants but a sharper
increase in 2003 as in EMIF. For females also the EMIF undercounts the youngest
groups 15 to 19 years old in comparison to the ACS. The average age for female
emigrants in EMIF is considerably higher than in the ACS, between 4.2 and 11 years
higher.
31
C. EMIF return migration, 2000-2003, compared to the Mexican Survey of
Employment (ENE)
We next compare EMIF return migrant flows in the years 2000 to 2003 to those
estimated from estimates of people (re)joining Employment Survey households after
living in the U.S from 2000 to 2003. These ENE estimates sum over return migration
events in the four quarters of each year. Unfortunately, the 2000-2003 ENE appears to be
a much worse benchmark for comparison than is the 1996/97 ENADID. Returning male
migrants in the EMIF are approximately twice as numerous as they are in the weighted
ENE estimates.
[TABLE 4 ABOUT HERE]
Table 4 shows the EMIF captures a higher number of male and female return migrants
than does the ENE from 2000 to 2003. The ENE shows an increasing trend in a higher
proportion of persons 60 years old or more returning to Mexico between 2000 and 2003.
The EMIF shows an increasing proportion of men 30 to 39 years old from 2000 to 2002.
[FIGURES 5 AND 6 ABOUT HERE]
For returning female migrants, the average age is higher in the EMIF than in the ENE, by
between 1.3 years and 6.9 years (see also Figure 6). The profile of male return migrants
in the EMIF has shifted to the right (older ages) as compared to in 1996/97, as can be
seen observe in Figure 5. With the total numbers of return migrants underestimated by so
32
much in the ENE, however, it is not clear that the shift to the right in the age distribution
in the EMIF indicates a problem in that survey. Recall that the fit of the age distribution
of male return migrants between the EMIF and ENADID was very close (see again
Figure 2). More generally in Figures 5 and Figure 6, it can be seen that the age
distributions of male and female return migrants have both similarities and differences
between EMIF and ENE. The overall shapes are somewhat similar. In particular, there is
evidence of a “retirement” hump for women but not for men in both the EMIF and the
ENE.
V. Conclusions
We evaluated the EMIF here by comparison to alternative estimates of emigration and
return migration from national household surveys and censuses in Mexico and the U.S.
The best benchmark survey we found to evaluate the completeness and composition of
male emigrant and male and female return migrant estimates in the EMIF is that of the
1997 Mexican Survey of Demographic Dynamics (ENADID). Unlike the Mexican
census and microcensuses (‘Conteo’), and unlike the 2006 ENADID, the 1997 ENADID
includes among last year’s return migrants those who had emigrated more than five years
before. In comparison to the ENADID, the EMIF captures accurately male return
migrants and their age distribution but underestimates the number of female return
migrants. For emigrants to the U.S., the EMIF and ENADID show similar patterns for
age groups 30 to 59 years old but EMIF seems to underestimate the flow of emigrants for
those 15 to 29 years old. The 1996/97 EMIF and ENADID male return migrant totals
and distribution by age are both very similar. This indicates that a primary strength of the
33
EMIF is its capturing returning of male return migrants of working ages. A secondary
strength of the EMIF is in its capturing well male emigrants at all but the younger
working ages.
The EMIF’s good capture of male migrants returning to Mexico is a major
advantage given that men still dominate migrant flows between Mexico and the U.S., and
that it is not possible to track return migration directly with U.S. sources of information.
Indirect estimation methods (Hill and Wong 2005; Van Hook et al 2006) will always be
second best to direct estimation in the case that the data used in the direct estimates are of
good quality. The EMIF’s return migration data cover a substantially larger proportion of
return migrants, moreover, than the best nationally-representative household survey
source for this purpose collected on an annual basis, the Mexican Employment Survey
(ENE). We estimated return migration levels in the EMIF that were double the estimates
of U.S.-Mexico return migration in the ENE.
The EMIF’s estimates of female return migrants in 1996/97, however, are much
too low at only a fifth of those estimated in the ENADID. Estimates of female
emigration in the early 2000s appear to be somewhat better, approaching those estimated
in the ACS. The poorer capture of female than male migrants in the EMIF may be due to
a greater likelihood of returning female migrants in other than the Northern region of
Mexico to use air travel to central and southern parts of the country. This would be
consistent with longer durations in the U.S., possibly associated with higher proportions
of authorized migrant women than men. It may also be the case that the EMIF has other
biases against identifying female return migrants. The EMIF’s orientation towards
identifying labor migrants over other types of migrants may be one such biasing factor.
34
On male emigration from Mexico to the U.S., our comparison between the EMIF
and ENADID emigration in 1996/97 suggests incomplete coverage in the EMIF on the
scale of 30 to 40%. The EMIF’s estimates of emigration in the early 2000s are
nevertheless much higher than those in U.S. data sources (where they are recorded as
immigrant flows). In the prime male migration ages of 20 to 39, in particular, the EMIF
estimates of the early 2000s are approximately double those of the U.S. census and
microcensus (ACS) sources for Mexico-U.S. migration. It seems likely that the EMIF’s
coverage of unauthorized migration streams is likely to be a very important factor in the
much greater estimates of Mexico-U.S. migration. Findings from comparisons of the
1990 U.S. Census and Mexican Migration Project data (Lindstrom and Massey 1994) are
consistent with this conclusion. Those authors found that lower-education migrants were
disproportionately likely to be missed in the 1990 Census.
The ENADID’s higher estimates of male emigration than in the EMIF are
significant due to the ENADID’s identifying no more than one emigration event per
individual per year. This argues against that the possibility that the EMIF counts multiple
emigration events by the same person in a given year as an explanation for the higher
estimates of Mexico-U.S. migration events in the EMIF than in the U.S. Census and ACS
in the early 2000s. The possibility that those U.S. data sources miss immigration events
because the individual has already returned to Mexico by survey time, however, cannot
be discounted. Further analysis could examine this possibility in the ENADID or other
Mexican survey data source in which both emigration and return migration events to a
given individual in a given year are captured.
35
The EMIF is not only less biased than other annually collected data sources on
both sides of the border. It also includes significantly larger sample sizes. This will be
especially important for estimating distributions of migrant characteristics and for
multivariate analyses. The EMIF’s sampled emigrants are approximately 8 times as
many as those sampled in the ACS, and their return migrants are approximately 4 times
as many as those sampled in the ENE. The EMIF represents reasonably well the
geographic origins of Mexico’s migrants to and from the U.S., but is somewhat biased
toward capturing less educated migrants and against capturing more educated migrants.
A potential explanation here is that the EMIF disproportionately misses more educated
migrants flying from, or back to, points south of the border cities.
VI. References
Aguila, E., and Zissimopoulos, J. (2008). “Labor market and immigration behavior of
middle-aged and elderly Mexicans”, Michigan Retirement Center Working Paper,
University of Michigan.
Amuedo-Dorantes C., and S. Poza (2005) On the use of different Money transmission
methods by Mexican immigrants International Migration Review 39(3):554-576.
Amuedo-Dorantes C., and S. Poza (2006) Remittances as insurance: Evidence from
Mexican immigrants Journal of Population Economics 19:227-254.
Amuedo-Dorantes, C., and C. Bansak (2007) “The effectiveness of border enforcement in
deterring and postponing repetitive illegal crossings attempts.” Manuscript.
Anguiano, M. E. (2003) Emigrantes Indocumentados y Deportados Residentes en el
Estado de México Papeles de Población 36:133-160.
36
Bilsborrow, R.E., G. Hugo, A.S. Oberai, and H. Zlotnick (1997) International Migration
Statistics: Guidelines for Improving Data Collection Systems ILO, Geneva.
Bean, F.D., R. Corona, R. Tuirán, and K.A. Woodrow-Lafield (2001) Circular, Invisible,
and Ambiguous Migrants: Components of Difference in Estimates of the Number
of Unauthorized Mexican Migrants in the United States Demography 38(3):411422.
Carling, J. (2007) Unauthorized migration from Morocco to Spain International
Migration 45(4):3-36.
Cerrutti, M., and D.S. Massey (2001) On the auspices of female migration from Mexico
to the US Demography 38(2):187-200.
CONAPO (2006) Encuesta sobre Migración en la Frontera Norte de México 2003-2004.
National Population Council, Mexico City, pp. 107.
CONAPO (2008a) EMIF, Aspectos Metodológicos www.conapo.gob.mx Accessed
9/22/08.
CONAPO (2008b) Series Sobre Migración Internacional www.conapo.gob.mx.
Durand, J. and D.S. Massey (2004a) “Appendix: The Mexican Migration Project”
pp.321-336 in J. Durand and D.S. Massey (eds) Crossing the Border: Research
from the Mexican Migration Project New York: Russell Sage Foundation.
Durand, J. and D.S. Massey (2004b) “What we have learned from the Mexican Migration
Project” pp.1-14 in J. Durand and D.S. Massey (eds) Crossing the Border:
Research from the Mexican Migration Project New York: Russell Sage
Foundation.
37
Ellis, M., and R. Wright (1998) When immigrants are not migrants: Counting arrivals of
the foreign-born using the U.S. Census International Migration Review
32(1):127-144.
Hill, K., and R. Wong (2005) Mexico-U.S. migration: Views from both sides of the
border Population and Development Review 31(1):1-18.
Horvitz, D.G., and D.J.Thompson (1952) A generalization of sampling without
replacement from a finite universe Journal of the American Stastistical
Association 47:663-685.
INEGI (2000a) Aspectos Metodológicos de la Encuesta Nacional de la Encuesta
Nacional de Empleo Urbano (ENEU) [Methodology of the National Survey of
Urban Employment] www.inegi.gob.mx.
INEGI (2000b) Manual del Critico-Codificador de la Encuesta Nacional de Empleo
Urbano 2000 (ENEU) [Coding Instructions for the National Survey of Urban
Employment] www.inegi.gob.mx.
INEGI (2000c) Manual del Entrevistador de la Encuesta Nacional de Empleo Urbano
(ENEU) [Manual for Interviewers for the National Survey of Urban Employment]
www.inegi.gob.mx.
INEGI (2003) Sintesis Metodológicos de la Encuesta Nacional de la Dinamica
Demografica www.inegi.gob.mx
INEGI (2005) Encuesta Nacional de Ocupacion y Empleo (ENOE) [National Survey of
Occupation and Employment] www.inegi.gob.mx.
INEGI (2008) Ejercicio Estadistico del INEGI Confirma el Descenco en la Salida de
Mexicanos al Estranjero Communicado 240/08. www.inegi.gob.mx.
38
IPUMS USA (2009) User’s Guide http://usa.ipums.org/usa/doc.shtml
Kana’iaupuni, S.M. (2000) Reframing the migration question: An analysis of men,
women, and gender in Mexico Social Forces 78(4):1311-1348.
Lindstrom, D.P. (1996) Economic opportunity in Mexico and return migration from the
United States Demography 33(3):357-374.
Lindstrom, D.P., and D.S. Massey (1994) Selective emigration, cohort quality, and
models of immigrant assimilation Social Science Research 23:315-349.
Lumley, T. (2004) Analysis of complex survey samples Journal of Statistical Software
9(1):1-19.
Lumley, T. (2008) Survey: analysis of complex survey samples. R package version 3.111.
Martí, M., and C. Ródenas (2007) Migration estimation based on the Labour Force
Survey: An EU-15 perspective International Migration Review 41(1):101-126.
Massey, D.S., and A. Singer (1995) New estimates of undocumented Mexican migration
and the probability of apprehension Demography 32(2):203-213.
Massey, D.S., and R.M. Zenteno (1999) The dynamics of mass migration Proceedings of
the National Academy of Sciences 96:5328-5335.
Mendoza, C. (2004) Circuitos y espacios transnacionales en la migración entre México y
Estados Unidos: aportes de una encuesta de flujos. Migraciones Internacionales
Colegio de la Frontera, 2-003: 83-109.
Minnesota Population Center. (2006). Integrated Public Use Microdata SeriesInternational: Version 2.0. Minneapolis: University of Minnesota.
http://international.ipums.org.
39
Office for National Statistics (2008) The International Passenger Survey
www.statistics.gov.uk.
Passel, J.S., and D. Cohn (2008) “Rise, Peak, and Decline: Trends in U.S. Migration
1992-2004.” Pew Hispanic Center Report.
http://pewhispanic.org/files/reports/53.pdf
Passel, J.S., and R. Suro (2006) “Rise, Peak, and Decline: Trends in U.S. Migration
1992-2004.” Pew Hispanic Center Report.
http://pewhispanic.org/files/reports/94.pdf
Poulain M. (1993) Confronting the Statistics On Inter-European Migration: Towards A
Greater Harmonization European Journal Of Population 9(4): 353-381.
Redstone, I., and D.S. Massey (2004) Coming to stay: An analysis of the U.S. Census
question on immigrants’ year of arrival Demography 41(4):721-738.
R Development Core Team (2009) R: A language and environment for statistical
computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3900051-07-0, URL http://www.R-project.org.
Rendall, M.S., C. Tomassini, and D. Elliot (2003) Estimation of annual immigration from
the Labour Force Surveys of the United Kingdom and continental Europe
Statistical Journal of the UNECE 20(3-4):219-234.
40
Rendall, M.S., and B.M. Torr (2008) Emigration and schooling among second-generation
Mexican-American children International Migration Review 42(3):729-738.
Rendall, M.S., E. Wright, and G. Horsfield (2005) “Immigration to and Emigration from
the United Kingdom Estimated from the European Union Labour Force Surveys.”
Paper presented at the British Society for Population Studies.
Reyes, B.I. (2004) Changes in trip duration for Mexican immigrants to the United States
Population Research and Policy Review 23(3):235-257.
Riosmena, F. (2004) “Return versus settlement among undocumented Mexican migrants,
1980 to 1996,” pp.265-280 in Crossing the Border: Research from the Mexican
Migration Project J. Durand and D.S. Massey (eds), New York: Russell Sage
Foundation.
Rubelcava, L.N., G.M. Teruel, D. Thomas, and N. Goldman (2008) The healthy migrant
effect: New findings from the Mexican Family Life Survey American Journal of
Public Health 98(1):78-83.
Ruggles, S., Sobek, M., Alexander, T., Fitch, C. A., Goeken, R., Kelly Hall, P., King, M.
& Ronnander, C. (2004) Integrated Public Use Microdata Series, Version 3.0.
Minneapolis, MN: Minnesota Population Center. http://usa.ipums.org/usa.
Warren, R. (2003) Estimates of the unauthorized immigrant population residing in the
United States: 1990 to 2000. Washington: INS.
Wong Luna, R., E. Resano Peréz, and L. Martínez Matiñon (2006) Una constante
cambiate: La migración de la población mexicana hacia Estatos Unidos de
Norteamerica Zapopan, Jalisco: Universidad de Guadalajara.
41
United Nations (2002) “Measuring International Migration: Many Questions, Few
Answers”, in International Migration Report 2002 New York: United Nations.
U.S. Census Bureau (2007) The American Community Survey www.census.gov/acs.
U.S. Census Bureau (2006) “Design and Methodology American Community Survey”,
Technical Paper 67, http://www.census.gov/acs/www/Downloads/tp67.pdf.
U.S. Census Bureau (2009) Population Estimates: Methodology.
www.census.gov/popest/topics/methodology. Accessed 4/22/09.
Van Hook, J., J.S. Passel, and W. Zhang (2006) New approaches and estimates of foreign
born emigration Demography 43(2):361-382.
42
Appendix A: Questions in Mexico survey data sources to identify Mexico-U.S.
emigrants (EMIF, ENADID), and U.S.-Mexico returning migrants (EMIF, ENADID,
ENE), and to identify Mexico-U.S. immigrants in the U.S. Census and ACS.
In this appendix, we present the questions from each of the four data sources that are used
to identify Mexico-U.S. and U.S.-Mexico migrants. Only Mexican data sources are used
to identify U.S.-Mexican migrants. We refer to these as “return migrant” flows. MexicoU.S. migrants are identified in both U.S. and Mexican data sources. We describe them as
“immigants” when referring to the U.S. data sources and “emigrants” (or “intending
emigrants”) when referring to them in the Mexican data sources.
A.1 Filter questions in the EMIF to identify (intending) emigrants
A small change in one of the filter questions was made between the 1996/97 wave and
the 1997/98 wave. We first describe the filter questions as they were from 1993 through
1997, and then describe the change to one of the questions applied from 1997/98 onwards.
(1) 1993/94 to 1996/97 Waves
These filter questions apply to wave 1 (1993-1994), wave 2 (1995-1996), and wave 3
(1996-1997).
1. Where were you born?
Locality, municipality, and State or Country
If the person was born in the United States, end interview
2. Do you live in this locality or in the United States?
1 Yes – end interview
2 No
3. For which of the following reasons did you visit the border area?
43
1 Studies- continue to question 4.
2 Tourism, shopping, travel- continue to question 4.
3 Visit family or friends- continue to question 4.
4 Business- continue to question 4.
5 Work or look for work - Apply the emigrant questionnaire
6 Change of residence- Apply the emigrant questionnaire
7 in transit to the North (U.S.) - continue to question 4.
8 Work motives- continue to question 4.
4. Do you have work in the place where you are coming from or a specific date to return?
1 Yes- end interview
2 No- apply the emigrant questionnaire
(2) 1997/98 to 2003/04 Waves
In wave 4 (1997-1998), wave 5 (1999-2000), wave 6 (2000-2001), wave 7 (2001-2002),
wave 8 (2002-2003), and wave 9 (2003-2004), filtering was as for waves 1 to 3 with the
exception that in question 3, the order of possibly responses was changed slightly and
that five instead of two responses led directly to the application of the emigrant
questionnaire, without first asking the additional filter question 4.
3. For which of the following reasons did you visit the border area?
1 Studies- continue to question 4.
2 Tourism, shopping, travel- continue to question 4.
3 Visit family or friends- continue to question 4.
4 in transit to the North (U.S.) - Apply the emigrant questionnaire
5 Work motives - Apply the emigrant questionnaire
6 Business - Apply the emigrant questionnaire
7 Work or look for work - Apply the emigrant questionnaire
8 Change of residence - Apply the emigrant questionnaire
Together this will have made it slightly more likely for an EMIF survey respondent to
have been identified as an emigrant from the 1997/98 wave onwards. Of these changes,
the rationale for allowing those answering “Business” to proceed straight to the main
questionnaire without first having them answer question 4 is unclear, and potentially
could lead to falsely classifying some as migrants who will not take up residence in the
44
U.S. This is unlikely to be a major problem, however, since very few individuals who
answered “Business” are classified as migrants.
A.2 Filter questions in the EMIF to identify return migrants
As for the “intending emigrant” questionnaire, small changes to the filter questions of the
“return migrant” questionnaire were made between the 1996/97 wave and the 1997/98
wave. We first describe the filter questions as they were from 1993 through 1997, and
then describe the change to two of the questions applied from 1997/98 onwards.
(1) 1993/94 to 1996/97 Waves
These filter questions apply to wave 1 (1993-1994), wave 2 (1995-1996), and wave 3
(1996-1997).
1. Where were you born?
Locality, municipality, and State or Country
If the person was born in the United States, end interview
2. Do you live in this locality or in the United States?
1 this locality – end interview
2 United States
3 Other
3. For which of the following reasons did you visit the border region, or the U.S. city you
are coming from?
1 Studies- continue to question 4.
2 Tourism, shopping, travel- continue to question 4.
3 Visit family or friends- continue to question 4.
6 Work motives- continue to question 4.
4 Business- continue to question 5.
5 Work or look for work - continue to question 5.
4. Were you in this place [Mexican border city or U.S.] more than a month?
1 Yes- continue to question 5
2 No- end interview
5. In which city did you spend most of the time?
45
1 City in the U.S.-apply the returning migrant questionnaire
2 North border-apply the internal migrant questionnaire
In wave 4 (1997-1998), wave 5 (1999-2000), wave 6 (2000-2001), wave 7 (2001-2002),
wave 8 (2002-2003), and wave 9 (2003-2004) the second of the filter questions was split
in two, asking whether the person lives in the locality or the United States separately.
Also, filter question 3 include additional possible responses for visiting the border: “work
and look for work” and “change of residence”. Finally, the option “work motives” passes
the respondent to filter question 5 directly, skipping the question 4 filter asked in the first
three waves.
2. Do you live in this locality?
1 Yes – end interview
2 No
2a. Do you live in the United States?
1 Yes – apply the returning migrant questionnaire
2 No
3. For which of the following reasons did you visit the border region or the U.S. city you
are coming from?
1 Studies- continue to question 4.
2 Tourism, shopping, travel- continue to question 4.
3 Visit family or friends- continue to question 4.
5 Work motives- continue to question 5.
6 Business- continue to question 5.
7 Work or look for work - continue to question 5.
8 Change of residence- continue to question 5.
A.3 Migrant filter questions in the ENADID
Section III, q. 3.2. Please, tell me the name of each person who normally lives here,
beginning with the head of the household; do not forget to include infants and older
people (include as well any servants who live here).
Section III, q. 3.6. [Living condition, asked about each person identified in 3.2]
1 Lives in other place because he/she is working, studying, or for other reasons?
2 Lives in other place, although he/she is here at the moment?
3 Lives here normally?
4 Lives here, although he/she is elsewhere at the moment?
46
5 Lives here temporarily because he/she has no other place where to live?
Section IV, q. 4.2. Have you lived in another State of Mexico o country, even when it was
for a short period?
1 Yes- continue to question 4.3 and 4.4.
2 No
Section IV, q. 4.4. In which State of Mexico or country did you live before living in [the
place of interview]?_________________________
Section VI, q. 6.11. Have you ever moved to work or to look for work to the United
States?
1 Yes
2 No
Section VII, q. 7.3. During the last 5 years, from January 1992 up to date, did you move
to work or study in another country?
1 Yes
2 No
Section VII, q. 7.11. Apart from the persons we just made note, during the last 5 years,
from January 1992 up to date, did someone else that used to live or currently live in this
household move to another country?
1 Yes-continue to question 7.13
2 No
Section VII, q. 7.13. Please tell me the name each person that moved to another country
during the last 5 years?_______________________
A.4 Filter questions of ENE Survey
Household roster, Section for individuals that moved out of the household, q. 21. To
which State of Mexico or country did that person move?
1 Same State in Mexico
2 Other State in Mexico
3 Other country
Household roster, Section for individuals that are new residents, q. 23. From which State
of Mexico or country did that person come from?
1 Same State in Mexico
2 Other State in Mexico
3 Other country
A.5 Migrant identification questions in the U.S. Census
12. Where was this person born?
47
1 In the United States - Print name of the State
2 Outside the United States – Print the name of foreign country, or Puerto Rico, Guam,
etc.
14. When did this person come to live in the United States?
15. Did this person live in this house or apartment 5 years ago (on April 1, 1995)?
1 Person is under 5 years old
2 Yes, this house
3 No, outside the United States – Print name of foreign country, or Puerto Rico, Guam,
etc.
4 No, different house in the United States
A.5 Migrant identification questions in the ACS
12. Where was this person born?
1 In the United States - Print name of the State
2 Outside the United States – Print the name of foreign country, or Puerto Rico, Guam,
etc.
14. When did this person come to live in the United States?
15. Did this person live in this house or apartment 1 year ago?
1 Person is under 1 year old
2 Yes, this house
3 No, outside the United States – Print name of foreign country, or Puerto Rico, Guam,
etc.
4 No, different house in the United States
48
Appendix B. Variance estimation for EMIF migration totals
We use the Horvitz–Thompson (H-T) estimator (Horvitz and Thompson 1952) of a
population total to estimate variance about total migration or any quantity describing
migrants. Crucially, use of the H-T estimator means that as analysts we do not need any
information on non-migrant individuals that are sampled but filtered out before a full
EMIF questionnaire is applied (see Appendix A above). The probability sampling
weights included with the survey data are all the information needed to estimate
variances about migrant quantities, including their total number, under an assumption of
independent sampling. The EMIF’s procedure of random sampling within location and
time further means that the weights themselves provide the needed information on sample
clustering, as we show below.
Suppose yi is a variable on person i = 1,..., N I in a population of all migrants,
N I . In the simple case of just counting migrants, yi = 1. Denote the sample of identified
migrants in the EMIF numbering nI by the indices i = 1,..., nI . Let wi be the weight
associated with observation i . Because the weights are constructed to provide
population estimates, we assume that π i = 1 / wi is the probability that a migrant i is
N
sampled. The H-T estimator of the population total
∑ y is
i
i =1
n
ŷ = ∑ wi yi
(1)
i =1
The variance estimate is
n
n
V̂[ ŷ] = ∑ ∑ yi y j (wi w j − wij )
i =1 j =1
49
(2)
where wij = 1 / π ij and π ij is the probability of observing both i and j in the survey. In
the case that the individuals are sampled independently, that is, wij ≈ wi w j , i ≠ j , and
wij = wi , i = j, the formula simplifies to
n
V̂ [ ŷ] ≈ ∑ yi2 wi (wi − 1)
(3)
i =1
n
Let nI = ∑ yi be the number of migrants in the sample. Then to estimate the population
i =1
number of migrants, yi = 1 for each migrant identified in the sample and the population
nI
estimate can be written as ŷ = ∑ wi and the variance as:
i =1
nI
V̂ [ ŷ] ≈ ∑ wi (wi − 1) .
(4)
i =1
Now suppose that the individuals are clustered (e.g., by bus station and time), and
that individuals within the same cluster have the same weight. Suppose that the
correlation between two individuals from the same cluster being in the sample is ρ and
that this is common across all clusters and individuals. This is equivalent to:
⎧ (1 − ρ )π 2 + ρπ
i
i
1 ⎪⎪
πi
π ij =
=⎨
wi ⎪
π iπ j
⎪⎩
i and j in the same cluster, i ≠ j
i= j
otherwise
With this π ij , equation (1) can be used to compute V̂ [ ŷ] for clustered data. Note that
because ρ ≤ 1, independent of this assumption, an upper bound for V̂ [ ŷ] is the choice:
50
⎧ πi
1 ⎪
π ij = = ⎨
wi ⎪
ππ
⎩ i j
i and j in the same cluster
otherwise
Example: standard error calculation for total return migrants in 1996/97
Applying the above formulas to the 1,224 identified return migrants in the 1996/97 EMIF
sample with weights {w1 , w2 ,..., w1224 } , ignoring the clustering we get ŷ = 290,253 and
SE[ ŷ] = V̂[ ŷ] = 17,179 . While the data provider does not supply cluster identifiers with
the data made available on their website, the EMIF’s random sampling within sampling
time and location means that weights will be identical within clusters and, in general,
different across clusters. We then assume that all individuals with the same weight
belong to the same cluster, and that those with different weights are not in the same
cluster. When the clustering identified this way is taken into account,
SE[ ŷ] = V̂[ ŷ] ≤ 21,046 , or about 20% greater than the standard error under assumed
independence. Note that even a small amount of intra-cluster correlation inflates the
standard error as ρ = 0.01 produces SE[ ŷ] = 20,785 . These estimates may be computed
in standard survey software such as the survey package (Lumley 2004, 2008) in R (R
Development Core Team).
51
Table 1: Mexican-born 1996-1997 U.S.-Mexico Return Migrants aged 15 plus, estimated in the EMIF and ENADID 1997
Males
EMIF 1996-1997
Total
Percentage
Total
Average age
Age group
15-19
20-29
30-39
40-49
50-59
60+
Education
None/Incomplete primary
Primary/Inc. HS
High school +
Region
North (excl. EMIF cities)
Traditional
Central
South/South East
EMIF Cities
Observations
276,901
32.2
Females
ENADID 1997
Total
Percentage
293,492
31.8
EMIF 1996-1997
Total
Percentage
ENADID 1997
Total
Percentage
13,352
31.9
85,670
34.4
18,570
119,586
79,394
34,406
18,792
6,153
6.7%
43.2%
28.7%
12.4%
6.8%
2.2%
22,816
128,315
87,482
31,713
16,278
6,888
7.8%
43.7%
29.8%
10.8%
5.5%
2.3%
416
5,740
5,236
1,297
140
523
3.1%
43.0%
39.2%
9.7%
1.0%
3.9%
10,798
33,863
15,870
10,409
5,841
8,889
12.6%
39.5%
18.5%
12.2%
6.8%
10.4%
104,602
148,682
23,617
37.8%
53.7%
8.5%
77,953
154,434
36,930
28.9%
57.3%
13.7%
3,340
7,970
2,043
25.0%
59.7%
15.3%
15,053
52,760
9,477
19.5%
68.3%
12.3%
59,089
136,109
33,122
25,640
7,198
1,172
22.6%
52.1%
12.7%
9.8%
2.8%
48,245
148,905
53,945
25,445
17,402
1,173
16.4%
50.7%
18.4%
8.7%
5.9%
6,267
3,146
2,547
247
199
52
50.5%
25.4%
20.5%
2.0%
1.6%
16,491
41,019
9,339
8,997
9,824
316
19.2%
47.9%
10.9%
10.5%
11.5%
NOTES: Education totals and distributions for ENADID were calculated using a smaller sample of returnees for which information
on education achievement was available. Therefore, returnee numbers by education category do not add up to the total in the
top row.
Data Sources:
EMIF = Survey of Migration at the North Border.
ENADID = National Survey of Demographic Dynamics.
1
Table 2: 1996-1997 Mexico-US Male Emigrants
EMIF 1996-1997
Total
Percentage
Total
Average age
Age group
15-19
20-29
30-39
40-49
50-59
60+
Education
None/Incomplete primary
Primary/Inc. HS
High school +
Region
North (excl. EMIF cities)
Traditional
Central
South/South East
EMIF cities
Observations
423,834
32.0
ENADID 1997
Total
Percentage
632,647
29.4
38,950
155,820
143,289
47,186
32,539
6,049
9.2%
36.8%
33.8%
11.1%
7.7%
1.4%
102,807
268,849
158,600
63,893
29,016
9,482
16.3%
42.5%
25.1%
10.1%
4.6%
1.5%
160,076
237,953
25,806
37.8%
56.1%
6.1%
103,746
181,976
28,687
33.0%
57.9%
9.1%
81,313
245,196
51,125
45,694
169
1,899
19.2%
57.9%
12.1%
10.8%
0.0%
94,860
331,389
110,995
75,503
19,900
2,460
15.0%
52.4%
17.5%
11.9%
3.1%
NOTE: Education totals and distributions for ENADID were calculated using a smaller sample of migrants for which information on
their education achievement was available. Therefore, migrant numbers by education category do not add up to the total in the
top row. Female emigrant comparisons are omitted due to known underestimation in the ENADID.
Data Sources:
EMIF = Survey of Migration at the North Border.
ENADID = National Survey of Demographic Dynamics.
2
Table 3: Mexican-born Mexico-U.S. emigrants aged 15 plus, 1999-2003, estimated from the EMIF
compared to Mexico-U.S. immigrants in the U.S. Census and American Community Survey (ACS)
1999
2000
2001
EMIF
Total
Percent
Census 2000
Total
Percent
2000 ACS
Total
Percent
EMIF
Total
Percent
2001 ACS
Total
Percent
EMIF
Total
Percent
2002 ACS
Total
Percent
402,297
33.0
291,837
26.1
225,309
27.4
336,780
35.3
191,104
27.2
266,325
36.4
172,764
28.3
MALES
Total
Average age
Age group
15-19
20-29
30-39
40-49
50-59
60+
Observations
35,614
139,614
125,619
62,316
30,683
8,451
8.9%
34.7%
31.2%
15.5%
7.6%
2.1%
86,019
130,383
47,555
16,736
6,485
4,659
29.5%
44.7%
16.3%
5.7%
2.2%
1.6%
73,653
85,809
32,917
18,741
7,242
6,947
32.7%
38.1%
14.6%
8.3%
3.2%
3.1%
19,424
90,497
133,416
55,224
18,160
20,059
5.8%
26.9%
39.6%
16.4%
5.4%
6.0%
49,808
77,969
40,626
13,888
6,086
2,727
26.1%
40.8%
21.3%
7.3%
3.2%
1.4%
10,420
69,052
94,304
55,381
22,322
14,846
3.9%
25.9%
35.4%
20.8%
8.4%
5.6%
33,675
75,719
43,286
13,613
3,212
3,259
1,928
13,271
158
2,268
417
2,025
320
69,794
38.4
178,170
28.0
117,695
30.1
74,532
41.1
120,083
29.0
73,545
40.8
99,542
30.3
19.5%
43.8%
25.1%
7.9%
1.9%
1.9%
FEMALES
Total
Average age
Age group
15-19
20-29
30-39
40-49
50-59
60+
Observations
5,244
19,913
16,886
9,464
11,168
7,119
295
7.5%
28.5%
24.2%
13.6%
16.0%
10.2%
43,749
79,043
29,692
12,664
7,189
5,833
8,381
24.6%
44.4%
16.7%
7.1%
4.0%
3.3%
26,787
47,774
20,205
6,383
8,817
7,729
22.8%
40.6%
17.2%
5.4%
7.5%
6.6%
3,709
16,806
18,480
12,629
14,498
8,410
92
534
3
5.0%
22.5%
24.8%
16.9%
19.5%
11.3%
28,907
47,387
24,623
8,036
7,200
3,930
292
24.1%
39.5%
20.5%
6.7%
6.0%
3.3%
7,538
12,380
16,672
12,832
12,246
11,877
463
10.2%
16.8%
22.7%
17.4%
16.7%
16.1%
17,391
41,992
20,329
11,786
3,938
4,106
205
17.5%
42.2%
20.4%
11.8%
4.0%
4.1%
Table 3 (Cont.)
2002
2003
EMIF
Total
Percentage
2003 ACS
Total
Percentage
EMIF
Total
Percentage
2004 ACS
Total
Percentage
575,542
34.8
147,698
28.6
530,251
32.3
205,720
28.1
MALES
Total
Average age
Age group
15-19
20-29
30-39
40-49
50-59
60+
Observations
38,856
171,367
179,458
131,476
38,000
16,385
3,393
6.8%
29.8%
31.2%
22.8%
6.6%
2.8%
36,319
65,318
23,350
11,611
5,835
5,265
347
24.6%
44.2%
15.8%
7.9%
4.0%
3.6%
43,451
207,922
165,073
70,662
29,106
14,037
4,368
8.2%
39.2%
31.1%
13.3%
5.5%
2.6%
47,287
85,904
47,572
14,651
6,329
3,977
420
23.0%
41.8%
23.1%
7.1%
3.1%
1.9%
FEMALES
Total
Average age
Age group
15-19
20-29
30-39
40-49
50-59
60+
Observations
67,541
32.2
8,419
28,909
11,183
7,904
8,748
2,378
493
95,835
31.4
12.5%
42.8%
16.6%
11.7%
13.0%
3.5%
17,775
38,653
17,337
8,609
8,439
5,022
240
83,815
32.6
18.5%
40.3%
18.1%
9.0%
8.8%
5.2%
14,316
29,866
15,578
13,436
7,496
3,123
626
110,259
31.6
17.1%
35.6%
18.6%
16.0%
8.9%
3.7%
21,098
39,356
22,431
15,306
5,619
6,449
247
19.1%
35.7%
20.3%
13.9%
5.1%
5.8%
NOTES: 2000 Census estimates were computed from those who were born in Mexico and whose year of arrival was 1999. American
Community Survey estimates were computed from those respondents who were born in Mexico and were living in Mexico one year
before the survey. EMIF estimates were computed using all individuals intending to enter the US.
4
Table 4: Mexican-born U.S.-Mexico Return Migrants aged 15 plus, estimated from the EMIF and ENE surveys in Mexico
2000
EMIF
Total
Percent
2001
ENE
Total
Percent
EMIF
Total
Percent
2002
ENE
Total
Percent
EMIF
Total
Percent
2003
ENE
Total
Percent
EMIF
Total
Percent
ENE
Total
Percent
MALES
Total
Average age
Age group
15-19
20-29
30-39
40-49
50-59
60+
Observations
338,249
35.2
17,722
127,080
88,111
53,961
24,144
27,231
2,128
145,628
30.6
5.2%
37.6%
26.0%
16.0%
7.1%
8.1%
13,505
68,235
35,598
19,134
7,580
1,576
518
506,800
34.7
9.3%
46.9%
24.4%
13.1%
5.2%
1.1%
23,567
181,858
165,473
90,224
24,546
21,132
2,406
335,074
32.2
4.7%
35.9%
32.7%
17.8%
4.8%
4.2%
516,931
33.3
33,371
135,275
91,508
46,291
18,819
9,810
1,090
10.0%
40.4%
27.3%
13.8%
5.6%
2.9%
23,686
174,788
201,574
81,733
22,184
12,966
2,780
213,471
33.0
4.6%
33.8%
39.0%
15.8%
4.3%
2.5%
19,286
80,663
60,828
33,440
9,833
9,423
697
300,154
33.3
9.0%
37.8%
28.5%
15.7%
4.6%
4.4%
16,469
114,439
90,073
53,858
14,214
11,101
2,276
172,339
34.5
5.5%
38.1%
30.0%
17.9%
4.7%
3.7%
13,620
54,923
49,039
33,745
14,294
6,719
573
7.9%
31.9%
28.5%
19.6%
8.3%
3.9%
FEMALES
Total
Average age
Age group
15-19
20-29
30-39
40-49
50-59
60+
Observations
55,689
43.1
2,478
7,792
16,523
10,725
6,785
11,386
346
25,425
41.8
4.4%
14.0%
29.7%
19.3%
12.2%
20.4%
2,399
6,610
2,443
3,368
5,140
5,466
119
69,805
40.2
9.4%
26.0%
9.6%
13.2%
20.2%
21.5%
1,698
17,022
17,184
14,539
11,367
7,995
315
56,630
33.3
2.4%
24.4%
24.6%
20.8%
16.3%
11.5%
37,418
39.5
12,630
20,854
6,215
3,564
7,178
6,190
241
22.3%
36.8%
11.0%
6.3%
12.7%
10.9%
6,045
6,863
8,479
4,931
4,394
6,706
271
44,953
35.8
16.2%
18.3%
22.7%
13.2%
11.7%
17.9%
5,929
15,591
7,303
6,608
4,541
4,981
168
49,833
37.9
13.2%
34.7%
16.2%
14.7%
10.1%
11.1%
3,156
13,446
13,653
8,248
4,873
6,457
390
33,401
33.7
6.3%
27.0%
27.4%
16.6%
9.8%
13.0%
5,468
14,019
4,881
2,079
3,233
3,723
130
16.4%
42.0%
14.6%
6.2%
9.7%
11.1%
NOTE: Encuesta Nacional de Empleo (ENE) estimates were computed from survey respondents who were classified as “new
immigrant resident” and had returned from the United States during each respective year. EMIF estimates were computed using all
individuals returning to Mexico for reasons other than to visit their families or vacationing.
Data Sources:
EMIF = Survey of Migration at the North Border.
ENE = National Survey of Employment.
5
Figure 1: EMIF Geographic Sampling Regions
CALIFORNIA
ARIZONA
NEW MEXICO
Mexicali-Calexico
!
!
Tijuana-San Diego
!
BAJA CALIFORNIA Nogales-Nogales
SONORA
!
Ciudad Juarez-El Paso
TEXAS
Piedras Negras-Eagle Pass
CHIHUAHUA
!
COAHUILA Nuevo Laredo-Laredo
!
Reynosa-Mc Allen
! !
Matamoros-Brownsville
TAMAULIPAS
6
Figure 2: Age Distribution 1996-1997 EMIF and ENADID Male Return Migrants
12,000
Male Return Migrants
10,000
8,000
6,000
4,000
2,000
0
20
40
60
Age
EMIF 1996-1997
80
100
ENADID 1997
NOTE: The distributions are estimated using univariate kernel density methods. The ENADID graph includes all individuals who
said to have returned from the United States between 1996 and 1997, except for those sampled (approximately) during the months
when EMIF was not collected. The EMIF sample is limited to individuals returning to Mexico for reasons other than to visit their
families or vacationing.
Data Sources:
EMIF = Survey of Migration at the North Border; ENADID = National Survey of Demographic Dynamics.
7
Figure 3: Age Distribution 1996-1997 EMIF and ENADID Mexican-born Female Return Migrants
Female Return Migrants
3,000
2,500
2,000
1,500
1,000
500
0
20
40
60
Age
EMIF 1996-1997
80
100
ENADID 1997
NOTE: The distributions are estimated using univariate kernel density methods. The ENADID graph includes all individuals who
said to have returned from the United States between 1996 and 1997, except for those sampled (approximately) during the months
when EMIF was not collected. The EMIF sample is limited to individuals returning to Mexico for reasons other than to visit their
families or vacationing.
Data Sources:
EMIF = Survey of Migration at the North Border; ENADID = National Survey of Demographic Dynamics.
8
Figure 4: Age Distribution 1996-1997 EMIF and ENADID Mexican-born Male Emigrants
25,000
Male Emigrants
20,000
15,000
10,000
5,000
0
20
40
60
Age
EMIF 1996-1997
80
100
ENADID 1997
NOTE: The distributions are estimated using univariate kernel density methods. The ENADID graph includes all individuals who
said to have returned from the United States between 1996 and 1997, except for those sampled (approximately) during the months
when EMIF was not collected. The EMIF sample is limited to individuals returning to Mexico for reasons other than to visit their
families or vacationing.
Data Sources:
EMIF = Survey of Migration at the North Border; ENADID = National Survey of Demographic Dynamics.
9
Figure 5: Age Distribution 2000-2004 EMIF and ENE 2000-2004 Mexican-born Male Return Migrants
60,000
Male Return Migrants
50,000
40,000
30,000
20,000
10,000
0
20
40
60
Age
EMIF 2000-2004
80
100
ENE 2000-2004
NOTE: The distributions are estimated using univariate kernel density methods. The ENE graph includes all individuals who were
classified as “new immigrant resident” over these years and had returned from the United States. The EMIF sample is limited to
individuals returning to Mexico for reasons other than to visit their families or vacationing. For both surveys, the year 2004 consists
of observations sampled only during the first two quarters.
Data Sources:
EMIF = Survey of Migration at the North Border; ENE = National Survey of Employment.
10
Figure 6: Age Distribution 2000-2004 EMIF and ENE 2000-2004 Mexican-born Female Return Migrants
Female Return Migrants
5,000
4,000
3,000
2,000
1,000
0
20
40
60
Age
EMIF 2000-2004
80
100
ENE 2000-2004
NOTE: The distributions are estimated using univariate kernel density methods. The ENE graph includes all individuals who were
classified as “new immigrant resident” over these years and had returned from the United States. The EMIF sample is limited to
individuals returning to Mexico for reasons other than to visit their families or vacationing. For both surveys, the year 2004 consists
of observations sampled only during the first two quarters.
Data Sources:
EMIF = Survey of Migration at the North Border; ENE = National Survey of Employment.
11