Immigration Data Needs
In preceding chapters, we reviewed the topics of immigration study that were covered at the workshop. In this chapter, we recapitulate the specific data needs previously discussed, with a focus on the recommendations for the improvement of current data. The data needs are grouped into several topics: the decennial census, the Current Population Survey, Immigration and Naturalization Service records, case studies, and data on nationality, race, and ethnicity. Chapter 7 discusses the pros and cons of developing a new longitudinal survey of immigrants.
It is important to review the role of the census, the cornerstone of the nation's statistical information on the population, in providing immigration data. The 1990 decennial census included several questions that are important for immigration research: on nativity (place of birth of respondents), date of entry to the United States for the foreign-born population, citizenship, language used at home, English language abilities, ethnic ancestry, and race and ethnicity. Workshop participants strongly endorsed the retention of these basic data in future censuses.
A question on parental nativity (place of birth of respondents' parents) is an important one for the decennial census. Although such information was collected on the 1970 and earlier censuses, it was not included on either the 1980 or the 1990 census.1 Parental nativity data provide the information required to examine
the social and economic characteristics of the sons and daughters of immigrants. Children of immigrants are a critical generation for study: they reflect the success and rapidity of adjustment to U.S. society.2 The children of immigrants are a pivotal, young subgroup of a national population increasingly affected by large-scale immigration.
Workshop participants noted that further enhancements of decennial census data would help to make the census more useful for immigration analysis. The key needed enhancements include the addition of a question on duration of permanent resident status and a question on date of naturalization. Census questions require a strong political mandate for inclusion; for the 1990 census, there was no adequate mandate for including parental nativity on the questionnaire. The Immigration Act of 1990 now provides a federal mandate for additional nativity information.
Recommendation 1. We urge that the Immigration and Naturalization Service work with other federal agencies and the Bureau of the Census, under the overall direction of the Office of Management and Budget, to include key immigration questions on future censuses, including a question on nativity and parental nativity, based on the requirements of the Immigration Act of 1990.
Participants noted the prime importance of decennial census data for the study of numerically small immigrant groups. National data on such population groups as refugees from Vietnam, Laos, Cambodia, Cuba, and El Salvador come only from the decennial census because there is no other cost-effective way to collect comparable national data for small and widely scattered population subgroups.
Discussion at the workshop included the topic of using optical scanning in taking the decennial census. The technology now exists for processing handwritten responses, and the Bureau of the Census is currently examining the possibility of using optical scanning for the 2000 census. Optical scanning would have several benefits: it would allow respondents to provide written responses instead of selecting precoded categories, and open-ended questions could be used more frequently. Participants were particularly interested in the usefulness for analysis of asking for the specific year of entry into the United States, rather than using precoded periods of time. Participants emphasized, however, that it is especially important to establish the scientific validity of open-ended census questions, if optical scanning makes it possible to have greater use of them. They also ac-
knowledged that census questions should be easily understood by respondents and that the form should be easy to process.
Mindful of these constraints, workshop participants strongly suggested that the Bureau of the Census should consider the optical scanning of handwritten responses for the decennial census. Participants pointed out that academic research groups have successfully processed handwritten records from the 1910 census. It would be disappointing if the 2000 census did not employ optical scanning technology.
An important source of decennial data for immigration research is the Public Use Microdata Sample (PUMS). The Bureau of the Census's 1990 PUMS files are 1 and 5 percent samples of the individual data from the decennial census. All individual identifications, including specific geographic residence, are deleted from the PUMS files in order to preserve individual privacy. The PUMS files are widely used by immigration researchers, particularly for the study of numerically small and widely scattered racial and ethnic groups. One valuable enhancement of the PUMS files would be to add such contextual data as local unemployment rates; the study of the impact of immigration on employment levels requires local-area data. Publicly available data from federal government agencies have limited geographic identification in order to protect respondent confidentiality. For example, data from the Current Population Survey reveal geographic identification of the respondent's residence for states and major metropolitan areas. Census PUMS files display geographic identification for counties or groups of counties (aggregating smaller populated counties). Researchers who wish to take the small-area context into account (such as the proportion of foreign-born people in medium-sized cities or rural counties) must have the macro data merged with the micro data prior to data release; they cannot create such data sets on their own. Workshop participants argued that such contextual data could be tabulated from the census and merged with individual PUMS data, at relatively low cost.
Recommendation 2. We recommend that the Bureau of the Census consider ways to add local-area contextual data to the Public Use Microdata Sample (PUMS) files. Contextual data on such variables as local employment, income, education, and racial and ethnic composition would measurably improve this important data set for academic and policy research on immigrants.
Current Population Survey
The Current Population Survey is the largest regular household survey conducted by the Bureau of the Census. In 1994, it covered interviews of about 60,000 households per month (about 140,000 to 150,000 individual observations), including the respondent and other household members. When a household is selected for the Current Population Survey and interviewed for the first
time, the respondent and all household members are asked a common set of household questions. Selected information is recorded for each individual, including age, sex, race/ethnicity, and relationship to household head. The respondent is then interviewed monthly for the first four months, not interviewed for eight months, and then interviewed monthly for four more months. The interview schedule is designed primarily for the best statistical estimate of monthly unemployment changes—the main purpose of the Current Population Survey. The monthly interviews include special supplements, such as one with detailed income questions.
The Current Population Survey, which produces a great deal of valuable data, is the key federal survey available for immigration analysis. For more than a decade prior to 1991, the Current Population Survey occasionally included questions on parental nativity and other immigration-related issues. Parental nativity—and such related questions as citizenship and year of immigration—are essential for immigration research and should be included as key questions as a regular part of the Current Population Survey.
Subsequent to the September 1992 workshop, and partially in response to discussions at the workshop, a group of federal agencies worked to place a nativity question on the Current Population Survey. As of 1994, the Current Population Survey collects nativity information for household members and parents of members, allowing researchers to distinguish among foreign-born, native-born of foreign-born parents, and native-born of native-born parents. In addition, the survey includes data on the year of entry for immigrants and citizenship status. The survey makes available basic information on immigration for all survey months and for all members of the household. It is notable that these data are available as well for all supplements to the Current Population Survey.
Recommendation 3. The committees applaud the introduction of key questions on nativity as a regular part of the Current Population Survey. Questions on nativity, parental nativity, citizenship, and year of entry into the United States provide information essential to the understanding of immigration in this country. We urge the Bureau of the Census to retain these key immigration-related topics in the Current Population Survey.
In addition to the information currently available, workshop participants noted that priority should be given to the periodic inclusion in the Current Population Survey of other immigration-related questions, such as duration of job search, more detail on immigration experiences, and duration of welfare receipt.
Recommendation 4. We recommend that the Bureau of the Census, in consultation with federal agencies and immigration researchers, review the possibility of adding special immigration questions to the Current
Population Survey. Additional, more detailed immigration-related questions would enhance the value of the Current Population Survey data for immigration policy research. Such questions might be included in the Current Population Survey on a special basis, perhaps on one of the special monthly supplements, or on a periodic basis, depending on the purpose and usefulness of the data.
Immigration is too often measured as an event rather than as a process. Measurement of an event implies different questionnaire design than measurement of a process. From the process perspective, if a question on date of entry is interpreted to refer to the most recent entry, then the immigrant's experience in the United States would be underestimated.3 If, however, the date of entry is interpreted to be the first entry, then their experience in the United States is overestimated. For studies of the immigration process, questions should be asked on both first and most recent dates of entry and the number of entries to the United States; such questions might be asked on a special basis in the Current Population Survey. Even better information could be gleaned from immigrant histories, although such data would be too lengthy and complicated for the Current Population Survey.
There was relatively limited discussion at the workshop about the advantages of enhancing the data on immigrant status on the Survey of Income and Program Participation (SIPP), the major federal household survey that collects information on individual and household participation in federal and state social services. It covers some topics of current interest for immigration policy, such as the use, amount, and duration of welfare. Workshop participants noted, without extended discussion, that the SIPP sample size is smaller than that of the Current Population Survey and would provide limited data for immigrants by country of origin, except for Mexicans. SIPP is a valuable data source, however, and the usefulness of SIPP for immigration research warrants further exploration.
Whereas the Current Population Survey is the key survey for use by immigration researchers, there have been discussions in recent decades about a joint survey in Mexico and the United States for immigration study.4 Such surveys would have value for policy studies in both countries. They could explore potential immigration, immigration before departure and after arrival, and return mi-
grants. Joint surveys have been discussed before in general terms; there may be a real opportunity for them at this time.
Recommendation 5. We recommend that U.S. federal statistical agencies meet with counterpart institutions in Mexico to discuss the potential for establishing joint surveys on immigration. Such a meeting should include key immigration researchers from both countries.
Immigration and Naturalization Service Records
Although some valuable changes have occurred in the Immigration and Naturalization Service statistical system in the past decade, changes since 1985—for the Immigration and Naturalization Service and other agencies collecting immigration data—also reveal that weaknesses in the data persist: the Current Population Survey questions on nativity have not been asked regularly, data on emigration and illegal aliens remain poor, and little is known about foreign students and new citizens. Some obvious opportunities for improving immigration data remain. Workshop discussion did not include evaluation of all INS administrative records (see Jasso and Rosenzweig, 1990), but focused on three broad areas for improvement in Immigration and Naturalization Service Records:
- Cooperative efforts are needed to improve immigration statistics. The matching of administrative records, especially records on program and welfare use, offers many possibilities for better immigration data. It would be useful, for example, to match welfare records to a sample of recent immigrants to make comparisons with the native-born. The results of such an effort would be beneficial for both the study of immigration and social program use.
- Coordination between the Immigration and Naturalization Service and the Social Security Administration could provide improved data and should be encouraged. Matching records from the Immigration and Naturalization Service and the Social Security Administration could provide longitudinal data on earnings for immigrants, an important but neglected topic of policy study. One possibility for exploration would be to assign a social security number to immigrants as part of the Immigration and Naturalization Service's administrative procedures. Such a step was recommended before—most recently by the Department of Health and Human Services in 1992—but to date has not been accepted by the Immigration and Naturalization Service and the Social Security Administration. A second possibility would be to take a sample of records on entering immigrants from past periods and link them to the Social Security Administration's records to provide data on earnings since arrival in the United States. With Immigration and Naturalization Service data linked to earnings information on individuals, dependents, and relatives, meaningful information on family processes would also be available for study.
- Conducting sample surveys of immigrants, using Immigration and Naturalization Service or other records, would add valuable information about newly arriving immigrants. There are special problems in surveying immigrants because they are widely dispersed in the U.S. population and are few in number in many areas. One possible survey design would be to use a sample of approved applications for a permanent resident visa—known as a green card. A relatively accurate address is provided by green card applicants. It would be possible to obtain information by enclosing a mail questionnaire, perhaps using several different languages, when the green card is delivered.
Workshop participants made suggestions for the Immigration and Naturalization Service to explore the collection of new data, including improving data on nonimmigrants, adding information on immigrant adjustment to information on previous nonimmigrant status, matching administrative records, conducting a longitudinal survey of immigrants, doing special surveys, and sponsoring case studies. Survey data on new immigrants, in particular, would offer useful additional data for immigration policy research.
Recommendation 6. We recommend that the Immigration and Naturalization Service establish the design and usefulness of a survey of green card applicants. A survey of new immigrants appears to be feasible, using the relatively accurate addresses that are provided by immigrants in order to receive their permanent resident visa.
A survey of new immigrants would provide cross-sectional data on legal new entrants into the United States. An ongoing survey, perhaps conducted annually or every few years, would also provide baseline data for longitudinal data collection. The workshop did not include discussion of specific proposals for such longitudinal data collection, although future study could weigh the merits and design for such a proposal. For example, should the survey of approved applicants be a cross-sectional survey to supplement administrative data on immigrants (e.g., collecting information on education, occupation in native country, support networks)? Should it be a longitudinal survey?
The workshop also included discussion of changes in Immigration and Naturalization Service administrative records that would lead to improvement. These records are collected primarily for processing applications for visas, visa adjustments, legalization programs, and naturalization. Although it is difficult to add questions for research purposes to administrative records, some simple changes would greatly enhance their value, including coding to permit: (1) easier linkage of the sponsor to the immigrants, (2) easier sampling of records, (3) easier linkage of information from the applicant and the sponsor, and (4) linkage of immigrant cohort records with other Immigration and Naturalization Service records for individuals.
Workshop participants stressed the importance of designing records so that samples can be taken from them. The technique of sampling has advantages because of the difficulty of adding information to administrative records. As a more cost-effective alternative, a sample of records could be selected for more intensive effort. From a research perspective, Immigration and Naturalization Service records are sometimes missing items for some variables. It is important to be able to sample from the records for follow-up surveys or linkage to other records to compensate for the missing information.
Working with Immigration and Naturalization Service records entails significant confidentiality requirements. Overall, the Immigration and Naturalization Service cannot release individual data for research outside the agency. Individual data on immigrants are not released, but the agency does release special administrative data, such as the names of newly naturalized citizens, as a matter of public record. For recent surveys of the population of illegal aliens who applied for legalization, it was necessary for the Immigration and Naturalization Service to obtain a special exemption for analysis by outside contractors. In addition, the Immigration and Naturalization Service cannot grant anonymity to illegal aliens who report their residency status on an Immigration and Naturalization Service survey. These special confidentiality conditions limit possible research activities by the agency itself and by others using its records; they need to be taken into account when future data collection is proposed. Nevertheless, it is important to exploit the inherent value of Immigration and Naturalization Service administrative records and to enhance them for better information for immigration policy.
There is a long-standing debate about the usefulness of macro-level and micro-level studies in immigration research. Macro-level data—data for aggregates such as states or national averages—is useful for studying overall trends. But national averages are not always useful for policy research. Public programs and policies affect individuals more than they change averages. Some researchers therefore argue for the need for studies using micro-level data, information on such specific units as individuals, families, and companies. One problem with micro-level studies is that the subjects may be minor in the overall pattern. In the best-designed research, the results of macro-level and micro-level studies complement each other.
Within the domain of micro-level studies, there is special debate among social scientists about the value of case studies. In immigration research, case studies make specific inquiries about selected aspects of immigration. They are often directed to such topics as industries or areas in which problems are presumed to exist. A methodological challenge is how to interpret a case study: If
the study area or industry was picked because of an existing problem, how representative are the results?
National studies often report findings different from those of case studies. In particular, case studies often find negative effects from immigration, in large part because they tend to select for negative effects. Moreover, it is easier to find negative immigration effects in local studies because positive effects often exist only for the most mobile immigrants, who are not concentrated in one place or industry. The more diffuse positive effects, workshop participants argued, are more difficult to detect in research data and hence may be a factor in the likelihood of case studies reporting negative immigration effects.
Case studies, with detailed descriptions of specific processes and changes, can be the point of departure for the study of new topics. Some researchers regard descriptive case studies as an early stage of work, and there is some value to this point of view. Case studies do not provide conclusions that apply beyond the subject matter described. Rather, they can provide ideas for further research to study and generalize. Nonetheless, good case studies can serve a broader role than merely the initial stage of research. A case study can have important scientific value in its own right, providing information valuable for specific purposes, even if the findings do not apply to all areas or industries. In arguing for a role for case studies, it should be emphasized that good scientific principles are essential to good case studies.
Currently available case studies, as mentioned above, have tended to deal with problems or negative aspects of immigration. Some of the physical and human capital benefits of immigration are often not mentioned in case studies. For example, U.S. residents get cheaper meals, better maintained lawns and gardens, less expensive child care, improved rural medical care, and larger numbers of skilled engineers and scientists because of immigrant labor. Some case studies focus on immigrant entrepreneurs. It would be helpful for public discussion if there were a better balance of positive and negative inquiries in case studies.
Data on Nationality, Race, and Ethnicity
Race, ethnicity, and national origin are major topics for immigration research. Studies of immigration by national origin (either country of birth or country of ancestral birth) often differ from studies of the ''average'' immigrant. It may make sense, for many studies, to focus on nationality groups, rather than to examine all immigrants combined. Adding Asian Indians and Chinese and Cubans together, even if they are in similar occupations or areas or social conditions, can cause problems for study and may not lead to sensible policy conclusions. Nationality groups have different origins, different experiences, and different conditions in the United States after entry. It is seldom reasonable to
combine nationality groups for immigration policy research. Similar arguments can be made for race and ethnicity groups.
In this period of high net immigration, the U.S. population could include as many as 14 percent foreign-born in the first decades of the 21st century, due to a combination of low natural increase in the native-born population and heavy immigration (Edmonston and Passel, 1994). At the very time in which the United States requires more information on immigration and the foreign-born population, the 1980 and 1990 censuses did not collect detailed information on nativity and ancestry. Previous censuses asked about the country of birth of respondents and their parents, which provided information on immigrants, the sons and daughters of immigrants, and the native-born population to native-born parents. By contrast, the 1990 census asked only about the place of birth of respondents. This limits the 1990 information to examination of the foreign-born and native-born populations. We have discussed the need for more detailed parental nativity questions in the decennial census.
Race and ethnicity data are important for immigration studies. As immigration from Europe continues at low levels, the white (non-Hispanic) population is increasingly becoming a native-born population with native-born parents (Edmonston and Passel, 1994). The black population is experiencing its first substantial immigration in almost two centuries. The Asian population consists predominantly of recent immigrants, more than two-thirds of its members being first or second generation U.S. immigrants, and will continue to be immigrant-centered for the next several decades. The Hispanic population has more than half of its members in the first and second generations; this situation will continue at similar levels for the next 50 years, assuming current levels of immigration, fertility, and mortality.
The United States collects considerable general-purpose race and ethnicity data on its population. Basic information is collected on the Hispanic population in many large federal household surveys; limited information is collected on the Asian population. But there is little basic data collected on the foreign-born population by ethnicity. Indeed, the decennial census is the single important source of race and ethnic information for the foreign-born population. Mortality statistics on the foreign-born are believed to be poor. Fertility information on births to foreign-born women from vital statistics is limited; childbearing levels and patterns of foreign-born women can be studied, however, using "own-children" techniques with census data.
For basic information on the foreign-born population, the Current Population Survey is the best baseline data system. With the introduction in 1994 of questions on country of birth of parents of respondents and their parents, the Current Population Survey has the potential to provide essential data on immigration. Moreover, as discussed earlier, the Current Population Survey could also include special-purpose supplements in order to collect additional information about immigrants and their children.
Workshop discussion suggested two basic improvements to be made in current demographic data collection. One is the reintroduction of detailed parental nativity questions in the decennial census. The other is the maintenance of detailed nativity questions in the Current Population Survey, including expanded use of special supplements on topics of importance to immigration studies.
The significance of data on race and ethnicity is illustrated by studies of Caribbean immigrants, most of whom report that they are black in censuses and surveys. Studies of the black population in New York City reveals that the Caribbean-origin population accounts for one-third to one-half of the black population, depending on the local area of study. Many Caribbean parents have children now in New York City schools. It is important to be able to compare second-generation Caribbean black children to long-term native-born black children. The two groups of children have distinctive origins. Comparisons require data on the race and nativity of the respondent and their parents.
There are special challenges to the collection of self-reported race and ethnicity data for government statistics. One is that self-definition data depends heavily on people's understanding of the social conventions of racial definition in the United States. Immigration research provides evidence that immigrants may have self-perceptions and acquaintance with racial definitions that differ from those of the United States. Brazilian immigrants to the United States may wonder, for example, how to report themselves because the racial definitions in Brazil are not the same as those in the United States. Immigrants from the former Soviet Union have primarily an ethnic identity and often are members of ethnic minorities that are not specifically listed on U.S. government forms. For immigrants from the former Soviet Union and other European countries, it would more useful to collect specific ethnic identification.
Another challenge is that race and ethnicity are not factors in the selection of immigrants into the United States. The Immigration and Naturalization Service has no administrative reason for knowing the race or ethnicity of immigrants and, indeed, would be reluctant to give any impression that the race or ethnicity of applicants might be a factor in selection. Data on race and ethnicity of immigrants must rely on such other data sources, such as the Current Population Survey, the decennial census, and sample surveys.