U.S. Department of Commerce

The Census Bureau Approach For Allocating
International Migration to States, Counties,
and Places: 1981-1991

David L. Word

Population Division
U.S. Bureau of the Census
Washington, D.C. 20233

October 1992

Population Division Working Paper No. 1

Line Divider

This working paper was presented at the 1993 annual meeting of the Population Association of America, Cincinnati, Ohio, April 1-3, 1993.

The views expressed are attributable to the author and do not necessarily reflect the views of the U. S. Bureau of the Census.

If you have any questions concerning this report, please e-mail a message to pop@census.gov. Include the name of this report and author in the body of the message.

October 1992

The Census Bureau Approach for Allocating International Migration
to States, Counties and Places: 1981-1991

Technical Contact: David L. Word
Population Division
(301) 457-2103


1.1 Overview: Because decennial census data were not sufficient for meeting an increased demand for current population information, the United States Census Bureau initiated a formal population estimates program in the 1940's (see Bureau of the Census, 1944). That early estimates program supplemented census data with updated statistics on births, deaths, immigration and emigration. Although today's population estimates programs are more sophisticated, the underlying concepts have not changed.

The general relation between a current population estimate, a previous census and estimated components of change is contained in the well known formula:

Pt+n = Pt + Bt,t+n - Dt,t+n + It,t+n - Et,t+n + Mt,t+n


Pt+n = the estimated population on the estimate date--n years after the most recent Decennial Census.
Pt = the enumerated population at the time (t) of the most recent census
Bt,t+n = the estimated number of births occurring during the interval (t,t+n)
Dt,t+n = the estimated number of deaths occurring during the interval (t,t+n)
It,t+n = the estimated number of arriving immigrants during the interval (t,t+n)
Et,t+n = the estimated number of departing emigrants during the interval (t,t+n), and
Mt,t+n = the estimated number of net domestic migrants during the interval (t,t+n).

Equation 1 allows the creation of a current population estimate Pt+n by using the previous census Pt as a bench mark and adding or subtracting numerical estimates of births, deaths, immigrants, and emigrants.

It is certainly possible for the domestic migration component, Mt,t+n, to have a sizeable impact when building a population estimate Pt+n for States, counties, and places; but when Pt+n references the entire United States Mt,t+n takes on a value of zero.

The data on births (Bt,t+n) and deaths (Dt,t+n) compiled by the National Center for Health Statistics (NCHS) are well reported for the United States as a whole and for individual States and counties. (There are, however, instances where births and deaths from NCHS are geographically misassigned within counties). Since this paper's focus is on the twin components of immigration (It,t+n) and emigration (Et,t+n), there will be no further discussion on births, deaths, or domestic migration.

1.2. Focus: The heart of this paper is the description of operations that measure immigration to and emigration from the United States. More to the point, it attempts to explain how these national estimates of immigration and emigration are apportioned to local areas (States, counties, and places).

The initial and most substantive section of the paper (Defining International Migration) delineates 9 separate and distinct categories of international migration, addresses the data sources used to generate the numbers within each migration category, and examines the assumptions needed to arrive at the international migration estimates. In order to better understand this text, we strongly encourage any reader to peruse Appendix Table A: "Estimating International Migration--A Synopsis" before proceeding further.

The short middle section (Estimating Domestic Migration Flows) provides background on the domestic migration estimates emanating from the Census Bureau's Administrative Records method. It explains how domestic and international migration may become intertwined and why the local area estimates program does not use each and every international migration data source that is required for the national population estimates program.

The final section (Allocating International Migration to Specific Locations) reports on the thought process and data base that is currently (1980 to 1991) used to distribute international migration to specific local areas. The illustrative attachments provided (tables 2 and 3) are designed for States; but the methodological principles are as applicable for counties and cities as they are for States.

1.3 General Comments on International Migration: International migration receives much of its prominence as a component of population because it is concentrated within a few States. Moreover, the recent immigrants are often clustered within a very few cities in those States. Table 3 at the end of this report indicates that one-third of all immigrants go to a single state--California, and one half go to either California or New York. About three-quarters of the immigrants end up in one of six states: California, New York, Texas, Florida, Illinois, or New Jersey.

Not only is immigration confined to a small number of areas, but recent (post 1965) immigration is largely Hispanic and Asian. From information appearing in column 7 of table 1, we note that 40 percent of the 942,000 net immigrants arriving in the United States between April 1, 1990 and July 1, 1991 came from Spanish speaking countries and an additional 30 percent of the immigrants came from nations where the overwhelming proportion of the population is Asian.

When making postcensal population estimates of the Hispanic or Asian populations, the accuracy of the total population estimate of the Hispanic or Asian population is largely reliant upon the accuracy of the estimate of Hispanic and Asian international migration.


Before launching into a description of the nine individual categories of international migration, we are providing a definition of international migration that is consistent with the goals of the Census Bureau's Population Estimates Programs.

For purposes of the Census Bureau's population estimates programs, an individual is classified as an International Migrant if his or her crossing of the United States border would necessitate a change in the resident population of the United States as defined in Decennial Census operations.

Tourists (either American or foreign) and "Day Workers" are not international migrants under that definition. But military reservists who served abroad in the Desert Storm Operation would have been treated as international out-migrants when they departed from the United States and international in-migrants when they returned to the United States. American college students who study abroad and foreign students who matriculate at American universities would also be classified as international migrants for Census purposes.

The 9 categories of international migration listed below (see Appendix A) are extremely diverse. The first four entries (1 through 4) satisfy traditional images of international migration and these are the only immigration categories that are used explicitly when generating LOCAL Population Estimates. Migration categories 5 through 8 generally pertain to the international movement of United States citizens. Visitors (Category 9) theoretically have no affect on the decennial census but in practice they probably do.

Nine Categories of International Migration

  1. Legal Immigrants
  2. Undocumented Immigrants
  3. Emigrants
  4. Puerto Rican Migration
  5. Movement of Armed Forces and their Dependents
  6. Movement of Civilian Federal Employees and their Dependents
  7. Movement of non Federal Employees and their Dependents
  8. Students, temporary workers, and exchange visitors
  9. Visitors

The National Population Estimates model for calculating net residual migration certainly requires a compilation of categories 1, 2, 3 , and 4; but it also relies on numerical estimates for categories 5 and 6. Since the local estimates program does not explicitly consider persons falling in international migration categories 5 and 6 as immigrants, the sum of the international migration estimates for the individual states will not match the national estimates of net residual migration. (The procedures used in estimating local area migration appears in section 3--Estimating Domestic Migration Flows).

This lack of closure does not appeal to demographers who want population estimation models to follow a strict demographic accounts procedure (as shown in equation 1). Conceptually, it is possible to construct a place specific estimate of international migration for categories 5 and 6. But (1) this component is already subsumed as a domestic migration component--see page 11, (2) these two categories of international migration would be highly concentrated towards areas with large military populations, and (3) there is no explicit information for allocating this component to any individual military base.

Moreover, the migration of persons falling into international migration categories 5 and 6 is largely "temporary". Analysts interested in immigration data would not want estimates of "permanent" immigration contaminated with the inclusion of these two categories.

The measurement of immigration and emigration is much more complicated than the measure of births and deaths. Much of this complication arises because the Census Bureau definitions of the 9 subcategories of immigration and emigration presented do not resemble formal definitions employed by the Immigration and Naturalization Service (INS) nor do they match the working definition set out in bold print on the previous page.

The formal (INS) definition of immigration appears in section 2.1.1. INS does not define emigration because it does not collect statistics on emigration; however, the U.N. recommendation of emigration presupposes intent as well as a time span (one year).

2.1 Direct Migration Flows

The first four categories of international migration are considered to be "direct" because their movement cannot be traced by matching addresses on adjacent IRS returns. Thus it is imperative to design a system for assigning their immigration down to specific States, counties and places.

The administrative data for legal immigration (section 2.1.1 below) are thought to be as reliable as administrative data on births and deaths. Furthermore, legal immigration is numerically (600,000 per year) the most important of the international migration categories. Estimates of the remaining components of immigration require somewhat heroic assumptions which are not always satisfactory.

2.1.1 Category 1, Legal Immigration: INS defines an immigrant to be an alien who has been admitted to the United States for the purpose of legal permanent residence. All immigrants are issued an I-551 document from INS; the I-551 is better known as the Green Card.

Refugees are a specific class of "nonimmigrants" who are generally permitted to adjust to immigrant status at a later date. An immigrant may eventually become a citizen; but lack of citizenship does not cause a resident to be excluded from enumeration in a United States Decennial Census.

INS maintains exceedingly detailed data sets for persons meeting their definitions of an immigrant (see Immigration and Naturalization Service, 1991). The administrative numbers (approximately 600,000 per annum) from INS are assignable to specific geographic entries within the United States.

2.1.2 Category 2, Undocumented Immigration: This component consists of aliens who do not have legal status in the United States; yet who are enumerated in a United States Census. The Decennial Census questionnaire is not constructed to determine legal status, and there are no explicit steps taken to exclude this group from the enumeration process. The majority of the undocumented immigration in this country are thought to arrive from Mexico and Central America. Moreover, a disproportionate share of the Hispanic undocumented population settles in California.

There are no regular administrative data sources for determining a precise number of undocumented residents in the United States. However, several coordinated research activities lead us to believe that this population is increasing by approximately 200,000 per year (see Passel and Woodrow, 1984).

We obtained the 200,000 figure cited above in several steps. First the 1980 Census count of recent (1975 to 1980) foreign born arrivals is assumed to be the sum of legal and undocumented immigrants arriving in this country over that time interval. In comparing 1980 Census figures of the foreign born who settled in this country after 1975 to INS sources of the "legally" resident foreign born who entered the United States between 1975 and 1980, it is "apparent" that nearly 1,000,000 net undocumented immigrants who entered this country between 1975 and 1980 were enumerated in the 1980 Census (see Warren and Passel 1987). The Census sample is sufficiently large to partition this total estimate of undocumented immigrants to individual countries. By this operation, we estimate that 60 percent of the recent arriving undocumented aliens come from Mexico. Column 5 of table 1 contains more specific information.

Concurrent analyses from Current Population Survey (CPS) data for the 1979-83 period also supported an estimate of 200,000 total illegal immigrants per annum; but that sample is not large enough to produce an independent estimate of undocumented immigrants by source country. If subsequent surveys had indicated that the original 200,000 estimate was no longer viable, we would have considered making a change to this component for the later years of the 1980 decade; but subsequent analyses of CPS data for 1986, 1988, and 1989 indicated that there was no evidence to support a change on the original 200,000 estimate (see Passel, 1986, for discussion surrounding the inclusion of the 200,000 figure for purposes of producing population estimates during the 1980's; and see Woodrow, 1991, for a review of the CPS estimates of the undocumented population).

The method for developing estimates of this enumerated undocumented population entering the United States between 1975 and 1980 relies heavily on independent estimates of emigration (described next) who left the United States in the same 1975-80 period.

2.1.3 Category 3, Emigration: Emigration is the process whereby a legal resident of the United States leaves this country and takes up permanent residence in a foreign nation. There are no current programmatic reasons for the United States Government to record emigration data, and the Government abandoned any pretext of collecting that data in 1957. It is necessary, however, to establish a numerical value for this component in a population estimates model.

Emigration estimates are constructed from the same analytical processes used in estimating the flow of undocumented aliens into the United States. Briefly, empirically based estimates of emigration were developed for the 1960-70 decade from a cohort analysis of 1960 and 1970 census data on the foreign born population (see Warren and Peck, 1980); a second set of estimates were developed from a systematic analysis of how the INS series of total alien registrants was affected via annual INS data on intended legal residents (see Warren and Kraley, 1985).

Both of these procedures suggest that the level of emigration for the 1975-80 period was close to 160,000 per year. For purposes of population estimation, we have assumed that emigration continued at that level throughout the 1980's. From the data appearing in columns 5 and 6 of table 1, one will note a striking difference between the countries favored by emigrants as opposed to the source country(s) of the incoming undocumented population.

For purposes of population estimates, we subdivide emigrants into two classes: native born emigrants and emigrants who were born abroad. We estimate the level of all emigrants to be 160,000 per year. It is assumed that 27,000 of these emigrants are native born while the remaining 133,000 are foreign born.

Conceivably one might wonder about the magnitude of the group--"emigrating" illegal aliens. The restricted definition of an emigrant (the exit of a legal resident of the United States) renders this concept moot. The estimate of undocumented immigration--200,000 per year (discussed in section 2.1.2) is a net measure that makes no explicit allowance for the departure of undocumented residents.

In the near future, we will be examining 1990 Census and Current Population Survey (CPS) results in conjunction with recent INS statistics in order to update the existing estimates of net undocumented immigration (200,000) as well as the current annual estimates of emigration (160,000). All tables included at the back of this paper were prepared using 1980 data based estimates.

2.1.4 Category 4, Puerto Rican Migration: Because Puerto Rico is a Commonwealth of the United States and its residents are citizens of the United States, there is no restriction on movement between the United States and Puerto Rico. However, the methodology used in estimating the resident population of the United States does require that we differentiate the population of Puerto Rico from that of the mainland United States. For lack of any better data, the calculation of movement between Puerto Rico and the United States mainland is derived from "differencing" airline passenger data into and out of the San Juan airport.

This differencing procedure ostensibly enables us to separate tourists, who number in the millions from immigrants/emigrants who number in the thousands. However, the amount of error in the 1990 population estimate of the Commonwealth of Puerto Rico vis a vis its 1990 Census count is approximately 200,000 or about 6 percent of Puerto Rico's population. An error of this magnitude demonstrates that the passenger data is not an acceptable source for estimating migration between Puerto Rico and the mainland United States.

Based on the passenger data series, the estimated level of annual "net" movement between Puerto Rico and the mainland of the United States has been as high as 40,000 in either direction. Although 40,000 is not a large component in terms of United States population change, it takes on added importance because the "permanent" migration to and from Puerto Rico is considered to be entirely Hispanic. Moreover, Puerto Rican migration is almost entirely limited to the East Coast of the United States.

Theoretically, we should treat Guam, American Samoa, the Virgin Islands and any other Trust Territory in a manner analogous to Puerto Rico. As a practical matter, the numbers involved are so small and the potential effect on the U.S. population is so slight, that these remaining outlying area flows are ignored in the population estimates model.

2.2 Abstract Migration Flows: Unlike the "direct" migration flows discussed in section 2.1, international migration categories 5 through 9 (below) deal with a concept of an "abstract net migration flow" as opposed to explicit data on individual movement across the boundaries of the United States. The calculations of these abstract net migration flows are determined by performing successive subtractions on annual stock (or group) data.

2.2.1 Category 5, Armed Forces: Armed Forces personnel serving overseas would not normally be enumerated in a decennial census. Throughout the 1980's, the size of the overseas Armed Forces population hovered around 500,000. In March, 1991, our nation's participation in Desert Storm caused the overseas Armed Forces population to soar to 700,000. But by the middle of 1992 the overseas military population had dipped below 400,000. Thus in the space of only two years, the United States had an Armed Forces out-migration of 200,000 which was then countered by an in-migration of 300,000.

Administrative counts of the overseas military population are furnished on a monthly basis by the Department of Defense (DOD). In addition to these monthly estimates of the overseas military population, (DOD) also provides data relating to the size of the overseas Armed Forces Dependent population.

The administrative counts of the overseas Armed Forces and their dependents, along with the administrative counts of the overseas civilian Federal employees (Category 7, section 2.2.3) and their dependents as of April 1, 1990 were appended to the 1990 resident Census for computing the apportionment population. The total supplementary overseas population on April 1, 1990 was 922,819.

2.2.2 Category 6, Civilian Federal Employees: The Office of Personnel Management (OPM) supplies monthly data on the number of Civilian Federal Employees who are working overseas at any time. That number is relatively constant and at this time (1992) is about 60,000. For purposes of population estimation we assume that each federal overseas employee has on average 1.2 dependents.

2.2.3 Category 7, Non Federal Employees: There are no reliable data on the number of non federal employees and/or their dependents who are overseas at any point in time. These persons would not normally be expected to be enumerated in a decennial census. The size of this population (like students--category 8 in section 2.2.4 below) would only affect the size of the United States resident population if it were subject to large fluctuations.

2.2.4 Category 8, Students, Temporary Workers, and Exchange Visitors: Persons fitting this category are not strictly defined as immigrants because they do not have immigrant status (see international migration category 1 in section 2.1.1). But citizens of foreign countries residing in the United States with these qualifications would normally be enumerated in a decennial census at their American address. Likewise, American students who are studying abroad at Census time should not be included in our census.

It is certainly possible to tabulate the number of foreign students by location within this country at the time of our decennial census; but the decennial census provides no information on the numbers of American students who are studying abroad. The Census Bureau does not estimate the numbers of foreign students in this country or American students overseas on an ongoing basis. However, an annual survey conducted by the Institute of International Education indicates that there were (in 1991) approximately 400,000 foreign students in the United States as opposed to 75,000 Amercians who are studying abroad. 50 percent of the foreign students are citizens of various Asian countries.

In terms of the postcensal estimates program, it is not critical to have a "fix" on the size of either student group provided that the numbers (race/origin/age/sex) are relatively constant over time.

2.2.5 Category 9, Visitors: Visitors from foreign nations are not supposed to be enumerated in the United States Decennial Census, and Americans who are temporarily in another country on Census Day will be enumerated at their usual place of residence within the United States (see 1990 Census of Population and Housing, 1991). We make no attempt at estimating the size of either group because it has no effect on a population estimate.

In practice, the visitor from overseas who stays only in commercial establishments would have little opportunity to be erroneously enumerated in our Census. However, if the "visit" is one of long duration to a housing unit, the visitor might well be enumerated in our Census. By the same token, an American resident who is between residences and who is on an extended visit overseas might not be enumerated in our Decennial Census.


3.1 Basic Administrative Records Procedures: Twenty years ago, the Census Bureau unveiled its Administrative Records method for making population estimates. This method (which can be used concurrently for States, counties and places) begins with the most recent census. To the census count, we add births and subtract deaths occurring in the interval (as shown in equation 1). The net migration estimate is derived from tracking mailing addresses on consecutive federal income tax returns.

When an individual files a tax return in consecutive years (as evidenced by the presence of the same Social Security Number--SSN) the mailing addresses in the two years are linked. If the addresses are identical, it is assumed that movement did not occur during the interval. When the addresses differ, we need to determine whether the movement indicates a crossing of place, county, or state boundaries. A sizeable portion of detected movement will never affect a population estimate because many moves are extremely local in nature. (e.g. movement between units in the same apartment complex or even movement across town within the same city).

Even though mailing address and residential address might not coincide, the basic premise of the Administrative Records estimating procedure is that migration can only have occurred when there is positive evidence (different mailing addresses in consecutive years) that movement has taken place during the interval between the two tax filings.

3.2 Use of Administrative Records Procedures for Measuring International Migration: The Administrative Records process will not capture the initial entrance into the United States of immigrants and refugees, nor will it detect movement of newly entering undocumented immigrants. Also, the paucity of overseas tax returns without military designations (APO and FPO addresses) supports the position that the Administrative Records method captures only a small proportion of persons fitting into international migration category 3--emigrants from the United States.

Residents of Puerto Rico who move (temporarily or permanently) to the United States mainland pose a special problem. Puerto Rican residents are subject to a Puerto Rico Commonwealth tax, but they are not required to file a United States Federal tax return unless they earned income in the mainland. As a result, the current formulation of the Administrative Records method does not trace migration flows from Puerto Rico to the mainland United States or from the United States to the Island of Puerto Rico.

Although it is not generally well known, the Administrative Records method does provide a measure of "immigration and emigration" for those persons who do fit into international migration categories 5 and 6. Every year approximately 900,000 tax returns containing nearly 1.8 million exemptions are filed from an overseas address. Approximately 25 percent of these persons (exemptions) have a U.S address in one year of the match and an overseas address in the other year. The Administrative Records method can and will pinpoint a precise location within the United States for both the "emigrating" and "immigrating" tax paying population.

The very magnitude of the number of overseas tax returns (900,000) suggests that a substantial portion of the international migration flows described under section 2.2 (Abstract Migration Flows) are captured through the use of linked tax records. We believe that Administrative Records almost certainly detects "immigration" and "emigration" for international migration categories (5)--Armed Forces and their dependents, (6)--Federal Workers and their dependents, as well as (7)--non Federal employees and their dependents. The matched tax return process may also detect a portion of international migration category (8)--American students studying abroad.

These assertions appear to be substantiated through an informal inspection of overseas tax returns that are not affiliated with the military or U.S government. In cases where the surname is "ethnic", there is a strong correlation between surname and the foreign country where the United States federal tax return was filed.

Since it is apparent that the Administrative Records procedure does detect "temporary" overseas migration and additionally provides these migrants an exact location within the United States, we feel that it is logical to treat the overseas tax returns in the same manner as tax returns filed in the United States. As a consequence, the local estimates program uses all tax based data (including returns filed from overseas locations) to measure "internal" migration, and does not rely on any national controls for estimating the net overseas migration of Armed Forces, federal employees and their dependents.


The Administrative Records method for estimating population of local areas imitates the national population model to as great an extent as possible. In other words, population change for an area is based on births, deaths immigration and emigration as well as domestic migration. (see equation 1) To accomplish that goal, we must determine a location within the United States for the net international migration allowance appearing in the far right column of table 1. Based on prior discussions, the Immigration and Naturalization Service (INS) collects precise addresses within the United States for the first three alien immigration groups appearing in table 1. But the remaining international migration groups (see Appendix A) which are necessary to the integrity of the national estimates operation are generally subsumed in the Administrative Records process.

When the Census Bureau set up its existing system for allotting international migration in 1981, INS had been incurring problems in processing their data. As a result, the Census Bureau was forced to abandon INS as a primary source for local area immigration data at that time. Although INS's processing problems had been corrected by 1982, the Census Bureau decided to continue allocating international migration by the "indirect distribution" procedure that was already in place.

4.1 Development of Country Groupings

The basic premise for allocating international immigration to precise locations within the United States was that the geographic distribution (e.g. States) for immigrants arriving between 1975 and 1980 in each of 16 country groupings would continue throughout the 1980's (see tables 1 and 2). In making national population estimates it is imperative to decompose individual migration components (legal immigrants, undocumented aliens and emigrants) because the legal flows are subject to known changes within a decade. For local estimates, however, decennial census data does not differentiate undocumented immigrants from legal immigrants. Moreover, the measurement of emigration streams from the United States would only be ascertained from censuses of foreign countries.

To a certain extent the individual countries or country groupings appearing on the attached tables are arbitrary. They were originally selected with the goal of separating the four major race/ethnic categories: (1) Hispanic, (2) non Hispanic White, (3) Black and (4) Asian. But it was also necessary to ensure that the direction of the migration flows within the four race/ethnic categories would be basically homogeneous. The information appearing in Table 2 aptly illustrates why it was necessary to further subdivide the four race/ethnic categories as illustrated below.

Mexico, Cuba, and the Dominican Republic are undeniably Hispanic nations, and each of these countries sends (or has sent) many of its citizens to the United States. However, the immigrants to the United States from those three nations do not head toward the same areas within the United States.

A cursory examination of some cell values appearing in table 2 reveals the following information on immigration patterns for the 1975-80 period.


  Mexico Cuba Dominican
Total 100.0 100.0 100.0
California 59.8 6.0 0.5
Florida 0.7 69.7 4.1
Illinois 8.9 1.5 0.4
New Jersey 0.1 9.4 10.4
New York 0.5 5.6 77.5
Texas 20.7 2.7 0.2
All Other States 9.3 5.1 6.9

Source: Table 2 following explanatory text

Together, the states of California and Texas received 80 percent of Mexico's emigrant pool for the 1975-80 period. (80 percent of all Americans responding to the 1980 Census sample questionnaire with Mexico as the answer to country of birth--and 1975 to 1980 as the answer to year of entry, lived in either California or Texas). On the other hand, almost all of the recent Cuban and Dominican immigrants were living in Atlantic seaboard States.

Conceivably, we could have chosen more than or fewer than 16 country groups, but the choice of 16 appeared to be a happy compromise--large enough to separate countries that had truly different patterns of immigration to the United States, but small enough to avoid the problems inherent in using proportions from small samples.

The first six of the 16 source countries appearing in tables 1, 2, and 3 are customarily considered to be "Spanish". (Spain is included in tabulations for South America, while Portuguese speaking Brazil is included in the category Southern Europe). The 7th to 10th rows of table 1 and the 7th to 10th columns in tables 2 and 3 contain countries whose previous citizenry was basically white and non Hispanic. Rows 11 and 12 of table 1 and columns 11 and 12 of tables 2 and 3 include those nations that are majority black. The final four country groupings in both tables contain various Asian combinations.

Table 1 provides April 1, 1990 to July 1, 1991 national estimates of that portion of international migration that cannot be captured by Administrative Records methodology. The total net international migration figure appearing in column 7 is 942,491. Mexico, (206,066) and Eastern Asia, (120,669) contributed the largest numbers of immigrants during the 1990-1991 period while the estimated number of immigrants from Southern Europe was almost nonexistent.

The distribution of immigration by foreign country group (table 2) is not necessarily operative for each and every one of the six immigration categories appearing in table 1. Research, based on 1980 census data, indicates that California and Texas receive roughly equal proportions of legal immigrants from Mexico; but that California receives a far greater proportion of the pool of undocumented Mexican emigrants (see Passel and Woodrow; 1984).

4.2 Derivation of Immigration Estimates

The estimates of international immigration to States (table 3) were developed by applying the rates in the table 2 matrix to the values appearing in column 7, of table 1. The resultant numbers in table 3 which are expressed in thousands have been rounded to the nearest one hundred. As expected, the states with the largest numbers of estimated immigrants are: California (312,000), Florida (52,000), Illinois (58,000), New Jersey (39,000), New York (148,000), and Texas (76,000). Maryland, Massachusetts, Michigan, Ohio, Virginia and Washington are each estimated to have received 15,000 immigrants in the 15 month period April 1, 1990 to July 1, 1991.

To guard against the possibility of an unwanted result, the raw 1980 Census numbers used to construct the proportions occurring in table 2 were truncated to meet the requirement of a minimum of 2 sample immigrants from 2 separate sample households for each country group. (Most States could not meet this requirement for Cuban, Dominican, or Southern European migration; hence the preponderance of blank cells appearing in the appropriate columns of table 2. Furthermore, recent (1975-80) immigrants that were enumerated in group quarters--mainly college dorms--were also excluded from the "predicting" distribution. Treating the foreign born students in dormitories in this fashion is consistent with the national estimates premise that foreign students would eventually leave their campuses and return to their native country.

The procedure for assigning immigrants to counties and cities within states is analogous to that for States. But for counties and cities, we took one further step to ensure against a sampling fluke. We generated annual estimates of international migration for all counties and cities, but we established the principle of setting any estimate of international migration back to zero that failed to meet a threshold of 50. The purpose of the threshold was to ensure that immigration would only be allotted to communities whose numbers of sample 1975-80 entrants were large enough to be statistically valid. (Counties and places did not have to meet the prerequisite of 2 sample immigrants from 2 sample households for each country grouping).

4.3 Validating Immigration Estimates Using Alternative Methodology

We must continue to caution that the figures in columns (5) and (6) of table 1 and all of the cell values appearing in table 2 represent patterns of immigration settlement for the years 1975 to 1980. That data is now approximately 15 years old.

However, we did compare the 1986 INS direct estimates of immigration against 1986 distribution based estimates. The differences were generally small. We repeated that operation with 15 month direct data (April 1990 through June 1991) vis a vis distribution based estimates of immigration for the same time period. Those results appear in table 4.

As noted previously, the lack of direct administrative data from INS forced us to use an "indirect" method for distributing international migrants to subnational areas within the United States. Fortunately, (see table 4) the differences between the direct and indirect method of assigning immigration destinations within the United States are relatively minor.

Table 4 compares the amount of immigration that would be credited to each State for the 15 month period ending July 1, 1991 by a direct assignation and the indirect method that the Census Bureau has been using for the past 12 years. For the most part, the differences are small. California, New York, Florida, Texas, New Jersey and Illinois are assigned the greatest number of immigrants under both systems; while North Dakota, Montana, South Dakota and Wyoming receive few immigrants under either method of assignment.

Fewer than half the States receive 5,000 immigrants under either system for the 15 month interval ending on July 1, 1991. For some states, the percent difference in the immigration estimate may be large (See Wyoming in last row), but the sizes of the numbers involved cause only a negligible effect on the population estimate of the two competing proposals. In only 3 States (Hawaii, New Jersey, and Oklahoma) would the 1991 estimate be affected by as much as 0.1 percent if the direct method of allocating immigrants was substituted for the method now used.

Which method is preferable? On the surface, the direct estimate has to be given the edge simply because it is the aggregation of administrative records from INS which rely on current rather than historical data which is 15 years old. On the other hand, if we had access to distributions from 1990 rather than the 1980 Census, the time lag would be less important. There is one factor that might suggest that distributional data from a recent census is superior. Census Data distributions allow for the fact that a new immigrant's initial location within the United States would not be captured unless a tax return is filed from that initial location.

4.4 Interpreting Immigration Estimates

For purposes of small area population estimates it is necessary to allocate the 200,000 national net undocumented immigration estimate by source country. Nearly 60 percent of the total estimated undocumented immigration between 1975 and 1980 emanated from Mexico, and each of the four later CPS studies confirmed that the bulk of the undocumented immigration to the United States emanated from Mexico.

Estimates of international immigration provided in table 3 cover a span of 15 months rather than the traditional one year period. But many researchers are interested in the cumulative amount of immigration over a longer period. The apparent simplest expedient would be to sum immigrant estimates over a period of years. But an accumulation procedure is not wholly adequate because immigrants themselves are subject to internal migration.

Today's population estimates methodology cannot "track" international migrants after their initial arrival. Thus, the possibility exists that a person may be classified as an international immigrant in the early part of the interval (t,t+n) and be reclassified as a domestic migrant at a later date within the same interval. In other words, our estimation model can only estimate aggregate levels of incoming international immigrants to an area within a given interval of time; it cannot nor will not provide an estimate of the number of international immigrants who reside in an area at a particular moment in time.

Once the new immigrant is established in this country, he/she will begin to earn money, pay income taxes, and possibly even move to another geographic area within the United States. The Social Security number (SSN) is the sole link for establishing internal migration and there is no way of connecting an alien identification number from INS to a tax identifier--the SSN. Without that link it is impossible to estimate or even approximate secondary domestic migration flows for the recent immigrant population.


The international migration component is difficult to estimate because there is no guarantee that the extreme geographic concentration of immigrants will hold for the next ten year period. Although each immigrant group has its own favorite destination(s) within the United States, the distributional approach for assigning immigration will not be satisfactory when new immigration flows arrive from unexpected sources. Even though we have 1990 Census data on the location of persons born in the Soviet Union (as it existed through 1991), we can't discern the relative proportions of those hailing from the "new" nations of Russia, Ukraine, Georgia etc. The same principle holds for persons who provided Yugoslavia as their place of birth in the 1990 Census, but who might have used the terms Croatia, Serbia, or Macedonia if the 1990 Census were retaken.

When we revise the methods of assigning international migration in the 1990's, we plan to reexamine the possibility of using INS data for the legal immigration component and to rely on Census data for estimating the flows of undocumented immigrants and emigrants to specific areas within the United states. INS data on alien immigration (international migration category 1) appears to have face validity and it is available in micro record form. Estimating undocumented immigration and emigration is always more problematic; and the fact that Alien Registration Records are no longer available compounds that problem. It is possible that CPS data might be enriched for the purpose of estimating undocumented immigration (see Woodrow, 1990); but there is no guarantee that that request will be honored in budgetary hearings.

If we do use INS data to estimate legal immigration (by source country), we cannot automatically assume that the geographic distribution of the undocumented portion of immigration associated with a particular country mirrors the geographic distribution of the legal immigrants. It is possible that we may resort to applying a legal immigrant destination matrix to an estimated undocumented flow for any country where those numbers are small. But we will not be able to use that approach for those countries or groups of countries where undocumented immigration is large. This last point is especially salient for the distribution of undocumented immigrants from Mexico.

The choices of methods for estimating international migration in the 1990's will eventually result from the combined efforts of three separate areas within the Population Estimation area. Each of the areas has its own concerns; but ideally, the estimates of the international migration components will be vertically integrated. Satisfying both internal interests within the Population Division and outside customers is a large undertaking which will require a very deliberate approach.


Bureau of the Census, 1944, "Estimated Population in the Continental United States, by Regions, Divisions, and States:" July 1, 1940, 1941, and 1942", Population - Special Reports, Series P-44, No.11, April 14, 1944.

Immigration and Naturalization Service, 1991. Statistical Yearbook of the Immigration and Naturalization Service. Washington, D. C. U.S. Government Printing Office.

Passel, Jeffrey S., 1986, "Changes in the Estimation Procedures in the Current Population Survey." Employment and Earnings, Vol. 33, No. 2, February 1986: 7-10

Passel, Jeffrey S. and Karen A. Woodrow. "Geographic Distribution of Undocumented Immigrants: Estimates of Undocumented Aliens Counted in the 1980 Census by State." International Migration Review 18 (1984): 642-671

Warren, Robert and Ellen Percy Kraly, 1985, "The Elusive Exodous: Emigration from the United States." Population Reference Bureau, Population Trends, (1985)

Warren, Robert and Jennifer Marks Peck, 1980, "Foreign-born Emigration from the United States: 1960-1970." Demography 17 (1980): 71-84

Warren, Robert and Jeffrey S. Passel., 1987, "A Count of the Uncountable: Estimates of Undocumented Aliens Counted in the 1980 Census." Demography 24 (1987): 375-393

Woodrow, Karen A., 1990, "Emigration From the United States Using Multiplicity Surveys." Presented at the annual meeting of the Population Association of America, Toronto, Ontario, Canada

Woodrow, Karen A. "A Consideration of the Effect of Immigration Reform on the Number of Undocumented Residents in the United States." Population Research and Policy (forthcoming)

1990 Census of Population and Housing, 1991, Summary Population and Housing Characteristics, 1990 CPH-1, Appendix D "Collection and Processing Procedures" Washington D. C. U.S. Government Printing Office

Table 1 Estimated International Net Migration Subdivided by Migration Category and Source Country: April 1, 1990 to July 1, 1991
Table 2 Percent Distribution of International Migration by Source Country: 1975 to 1980
Table 3 Estimated Net International Migration by Source Country: April 1, 1990 to July 1, 1991
Table 4 Comparing Direct and Indirect Methods of Assigning International Immigration
Appendix A: Estimating International Migration--A Synopsis

Population Division Working Papers