Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 68
--> 7 Longitudinal Studies of Immigrants The value of longitudinal data to the study of immigration is a theme that was sounded throughout the workshop. In examining the potential of a longitudinal survey, it is helpful to consider a taxonomy for longitudinal data (see Table 2). One type is new prospective data, defined as a new data collection study in which a selection of respondents is surveyed and then followed regularly over time. It has a heavy respondent burden, is expensive, takes a long time to collect, and requires a comparison group. A second type is new retrospective data, defined as a data collection scheme in which a selection of respondents is interviewed and asked about changes in the past. A retrospective survey would select those residing now in the United States, would be cheaper, would provide data immediately, would have less respondent burden, and would show data on duration. A retrospective survey also has the value of providing baseline information, if respondent follow-up occurs. A third type is new prospective design with a retrospective baseline, defined as a data collection study that combines a retrospective study at the beginning with prospective follow-up of respondents over time. This type of design is preferable to the first design because it collects past information on the respondents at the very beginning of the study. A fourth type is a prior prospective design with newly matched administrative records. This design involves updating by linking the existing survey to recent administrative records. A fifth type is the study of synthetic cohorts in censuses and surveys. Smith's
OCR for page 69
--> TABLE 2 Taxonomy for Longitudinal Data Collection Approach Description Advantages Limitations Prospective Follow cohort over time Contemporary measurement Respondent burden, expensive, requires long time, needs comparison group Retrospective Select sample and obtain data on prior periods Cheaper, provides data immediately, less respondent burden, produces duration data Data is historical, limited by respondent recall, some previous objective measures impossible to collect, potential unknown selection bias from immigrants who have already emigrated Prospective with retrospective baseline Collect initial retrospective data and follow cohorts over time Provides past data for initial analysis High respondent burden, expensive Prior prospective linked with administrative records Link existing, earlier survey with current administrative records Cheaper, provides data for current analysis, low respondent burden Current data are limited to content of records Cohorts in censuses or surveys Compare cohorts (not individuals) in successive censuses or surveys Cheaper, low respondent burden, can provide analysis for long periods of time Limited to cohort comparisons, possible selection bias
OCR for page 70
--> work (1992) illustrates this type of design, in which birth or immigration cohorts are followed in successive censuses or surveys. Advantages and Disadvantages Longitudinal surveys using a prospective sample are expensive and complex. Before undertaking such a survey, a clear idea is needed of its appropriateness for the particular questions being addressed and its advantages over alternative research approaches. Specifically, three important factors are relevant. First, a longitudinal survey needs to focus on a topic that involves change over time. Second, longitudinal surveys need a comparison group for the study. Third, a longitudinal survey works best when it measures an objective behavior. The expense of a large longitudinal survey is another factor. For example, it costs about $2 million per year for the Department of Labor's longitudinal survey on labor force participation of 4,000 respondents. A longitudinal survey of immigrants would need to include the recruitment of a new sample of immigrants every two or three years, thereby increasing the costs beyond those of a longitudinal sample of a single cohort. A longitudinal design with new prospective data is only one approach to studying change over time. As mentioned previously, alternatives to longitudinal survey data can be collected using retrospective surveys, such as in a life history format, or by successive cross-sectional surveys. Both alternatives are cheaper than prospective longitudinal surveys, although both have limitations (a retrospective survey would not include past immigrants who have already emigrated, for instance). Moreover, prospective longitudinal surveys have special problems with the reinterviewing of the original group of respondents. For example, the Panel Survey of Income Dynamics originally did not seek to interview people who had been lost in a previous wave; the survey now tries to continue the interviewing even if a respondent was lost in a previous wave. The National Longitudinal Survey of Youth also tries to continue to interview the respondents, even those lost during a previous wave. This seems to be a good practice for longitudinal surveys. Without it and with loss of respondents over time, those who are reinterviewed are not randomly chosen, and this bias is compounded in successive waves. The bias of respondents lost in follow-up in longitudinal studies can be minimized but cannot be totally eliminated. Retrospective studies and comparative cross-sectional studies have the advantage of selecting a sample that is representative of the desired population at the time of the survey. An adequate comparison group is an important issue for prospective longitudinal surveys. What is the appropriate comparison group for a survey of immigrants? Depending on the purpose of the research, the appropriate comparison group might be long-term residents or citizens of the United States, the population of the country of origin, immigrants with different durations of residence in the United States from different countries of origin, or immigrants who entered
OCR for page 71
--> the United States under different immigration categories. The design of a longitudinal survey needs to take the comparison group into account. Successive recruitment of new cohorts of immigrants may be a better way to ensure that a longitudinal immigrant survey does not become dated. New immigrants could be added to an immigrant study population and then followed along with those already in the longitudinal survey. With data on individuals, a longitudinal study would provide insights into emigration. Emigration of immigrants from the United States is an important concern (a substantial proportion, as much as one-third, of immigrants subsequently depart), yet emigration data are scarce. Following individuals in a longitudinal survey would provide data on the selection of emigration and factors related to emigration. The contexts in which respondents live are important to consider in a longitudinal survey. Since immigrants experience new cultural milieus, it would be wise to ask them about their personal experiences and attitudes. Some retrospective questions, however, are difficult to ask on a survey—for example morally sanctioned behavior (bribery, embezzlement), deviant behavior, and the cultural aspects of behavior. Retrospective questions may also suffer from recall errors. It is important that surveys include questions on both behaviors and attitudes of immigrants in such sensitive areas as fertility and criminality. It would be rash to attempt a longitudinal study of a prospective sample of immigrants without a clear view of the potential benefits. For short-term policy planning, there is little need for a longitudinal design. However, if the study focuses on change in attitudes over time or change in behaviors over time in both the United States and the country of origin, a longitudinal design is the most appropriate approach. Alternatives to a Longitudinal Survey of Immigrants Several alternatives to developing an entirely new longitudinal survey of immigrants exist. One could augment the samples of ongoing surveys. The foreign-born population is a special group, and one could oversample this group in an existing survey. For example, both the Panel Survey on Income Dynamics and the Health and Nutrition Examination Survey have added Hispanic samples to their data collection in recent years. Alternatively, one could alter the survey instrument in a current survey. For example, one could collect information on prior occupation, visa type, and migration history. One could also develop a new sample and a new instrument for the special study of immigration. In thinking about alternatives, workshop participants noted six issues in designing a study of immigrants. First, panel data would be helpful in studying the adaptation of immigrants using a prospective design. The specific merits of panel data are described in the section below on special-purpose surveys.
OCR for page 72
--> Second, it is important to think about the sampling frame for studies of immigrants. Immigration and Naturalization Service administrative records might provide a sampling frame, but there are problems (such as the exclusion of illegal immigrants), and a comparison group would still be needed. Linking Immigration and Naturalization Service data is a possibility, but it might be legally troublesome. And, of course, the foreign-born population stock differs appreciably from recent flows; immigration flows shift year by year. Although recent flows have been predominantly from Latin America and Asia, such countries as Germany and Canada would offer useful comparisons for a survey of the foreign-born population. Third, the social and economic status of immigrations is important and, in order to understand it, it may be necessary to oversample some groups in an immigrant survey. Some groups enter in relatively small numbers and may warrant oversampling in order to ensure sufficient sample size for analysis. For example, a survey may want to oversample entrepreneurial immigrants because they are few in number but enter under an important new visa category. Fourth, data collection by the Immigration and Naturalization Service involves some special issues. It would be difficult for the government to ask survey questions that might have legal repercussions. Also, subjective questions are easier to ask in surveys sponsored by universities or businesses, since they do not necessarily require government approval of the questionnaire design. Certain information is often not released in government data in order to fulfill confidentiality requirements—another special issue. Data on contextual variables in government surveys are usually absent or minimal. Fifth, data on duration of stay are important for some immigrant analyses. Surveys can obtain parallel lifelines (retrospective questions on migration history, fertility history, and so on). But it is a special challenge to understanding how to collect duration data on the migration process. Sixth, ethnographic data are important to consider.1 Studies using participant or direct observation can offer useful additional insights into immigrant adjustment. The major attraction of ethnographic field research is the comprehensiveness of the inquiry. By going directly to the social phenomenon under study and observing it as completely as possible, ethnographic observations can develop a fuller understanding of changes. This technique may be useful for topics that seem to defy quantification or in the use of self-reported survey questionnaires, for which nuances of attitude or behavior are critical and observation in the natural setting is paramount. Ethnographic studies may provide im- 1 Ethnographic data are collected by personal observation by the researcher or research staff of people and families. Ethnographic studies differ from the case studies discussed earlier, which are of a specific research site (such as a factory or city) but might use survey data collection techniques.
OCR for page 73
--> portant insights, suggestive rather than definitive, although they do suffer from problems of validation, reliability, and generalizability. Workshop participants suggested augmenting a large, ongoing longitudinal survey—the Panel Study on Income Dynamics—with a sample of the foreign-born population. A topical module could also be added to the first wave in order to collect pertinent retrospective information. This would enhance a current survey, albeit at moderate cost, and would initiate a longitudinal follow-up of immigrants with baseline data on such topics as past immigration, labor force, and family experience. Other participants expressed reservations about adding immigrant subsamples to currently ongoing longitudinal surveys. The problem, they noted, is that existing longitudinal surveys generally have not supplemented their samples with recent immigrants. Therefore, as time goes by, they remain representative of the population resident at the time the survey began. Depending on the level of immigration, they become less and less representative of the current population. For example, although the Panel Survey of Income Dynamics is useful for analyzing black-white differences in income dynamics, its sample excludes immigrants since 1968. Adding a sample of current immigrants would not compensate for the exclusion of immigrants in prior years and would therefore permit comparisons only of current immigrants with those who arrived prior to 1968. A supplemental strategy would need to add immigrants since 1968 and replenish with new immigrants on a regular basis. Similar arguments apply to the National Longitudinal Survey of Youth, which does not include Asians. Any new survey initiative should adequately represent the racial and ethnic diversity of the U.S. population. A key point of contention is whether existing longitudinal surveys, with potential cost savings derived from building on an existing survey, could be redesigned to offer an adequate comparison population for an immigrant subsample, or whether it would be more cost effective to begin an entirely new longitudinal survey. Longitudinal data takes considerable time to accumulate for questions involving years of data. Another suggestion by workshop participants was to take a sample from records of earlier admitted permanent residents (INS or other administrative records might be used, or an earlier survey might provide a baseline data set) and interview the respondents in the present. This approach would miss undocumented immigrants, but it may be a second-best, less expensive approach that would provide useful information as soon as the data are collected and analyzed. It might also provide valuable information on groups of legal immigrants, for which there are important questions about adjustment in economic status, education, and social change. Assuming for the moment the desirability of a longitudinal survey of immigrants, workshop participants discussed some important technical issues that need attention:
OCR for page 74
--> Sample size and coverage. There is a trade-off between including more groups and collecting larger samples for major groups. Hispanics, the largest collective ethnic group in the foreign-born population, do not constitute a sufficiently large subsample in most major surveys. But a sufficiently large sample for specific ethnic groups would result in an extremely large overall sample size. Visa status at the time of entry to the United States. Status at entry is a desirable measure to collect, but it is extremely problematic. Using self-reported interviews, it is difficult to collect information on the various visa categories, participation in various legalization programs, and unauthorized entry. Much of this information can be found in the Immigration and Naturalization Service's administrative records. Even so, the use of administrative records would not include residents who entered illegally and had not participated in a legalization program. Local labor markets. Many studies on immigrant behavior, particularly those involving labor force participation, require information about the respondent's local labor market. For the analysis of such data, researchers need to have access to local-area geographic codes that would allow them to measure labor markets and other local context variables; alternatively, these contextual data need to be collected as part of the survey operation. Government agencies ordinarily suppress local geographic identification in the data they collect, which limits the analysis by eliminating contextual information. Knowledge about some local-area contexts is critical for questions about the economic adjustment of immigrants, but the confidentiality restrictions on this type of data prohibit their collection and dissemination by federal statistical agencies. Sample survey design for reinterviews. A neglected aspect of the design of longitudinal surveys is the sample scheme for periodic reinterviews. For a longitudinal sample, a more careful design is warranted than simply the notion of contacting the sample respondents every year or "every so often." The frequency of reinterview should be based on the assumption of stochastic processes of change over time and on how to sample from those processes. In general, a survey should sample more frequently for processes that are changing more rapidly. Thus, for example, for a study of immigrants, one may want to sample the newly arrived and/or younger immigrants more often because processes of labor and geographic mobility and education achievement are changing relatively more rapidly for them. Respondents lost to follow-up. Not all persons in the original sample can be followed successfully over time in longitudinal surveys. The loss in follow-up surveys presents both a financial cost (data are collected for which there is no subsequent information) and a statistical challenge (there may be a bias in those who are lost to follow-up). Although the Annual Survey of Refugees obtains respectable overall response rates and reasonable reinterview rates of those originally surveyed, there is some drop-off in the follow-up of individuals and families, apparently as they begin to move and lose contact with their original refugee
OCR for page 75
--> community. A special concern in a longitudinal survey of immigrants is emigration: it is crucial for the follow-up to distinguish return migrants who can no longer be contacted because they have departed from the United States. Special-Purpose Immigrant Surveys Although there have been few surveys of selective immigrant populations, special-purpose immigrant surveys can provide worthwhile information at a reasonable cost. For example, the East-West Population Institute has been supporting a special longitudinal survey of immigrants for several years. In 1985, the project began looking at Koreans and Filipinos. Korea and the Philippine Islands were among the largest single source of Asian immigration in the 1980s.2 The researchers took an initial sample of visa applicants in 1985 in Korea and the Philippines. By 1988, 53 percent of the visa applicants resided in the United States and were contacted by mail. Telephone surveys were conducted in 1988–1989 and 1991–1992, and about 13 percent of the original visa applicants were still in the survey. This longitudinal survey looked at employment and self-employment information, gathered as retrospective data at the first interview. This type of prospective study of immigrants poses several challenges: it is very time-consuming, with long periods before data are available to analyze. It requires a strong central staff in order to continue the data collection. Attrition from the original group of respondents can limit the usefulness of the data. Special-purpose immigrant surveys are not as expensive as large national surveys and can allow experimentation with new survey techniques or limit attention to a specific population group or topic of study. They are especially valuable for studies with a local focus. Workshop participants encouraged the development of new special-purpose immigration surveys. Need for Longitudinal Data Workshop participants considered whether a special longitudinal survey of immigrants should be pursued, recognizing that a new prospective survey would be expensive yet would yield valuable data. They also acknowledged that it would take many years before data on temporal changes are available, yet researchers and policy makers in the year 2000 would be grateful if data collection were to begin now. 2 Immigration from Korea has declined considerably from the level of the mid-1980s. The Philippines remain of one at the largest single-country sources of Asian immigration. In the 1990s, China, India, and Vietnam provided the largest sources of Asian immigrants—although the annual number of entrants fluctuates a great deal by country.
OCR for page 76
--> TABLE 3 Existing Longitudinal Surveys with an Immigrant Component Survey Description Survey Dates Original Panel Sample Size Foreign-Born in Samplea Survey of Income and Program Participation (SIPP) A continuing household survey on people's economic well-being and the receipt of assistance from government programs designed to measure short spells of need and program use. Supplemental questions on the 2nd interview include migration history 1983 to present; panel interviewed every 4 months for 32 months. Rotating panel design 21,000 households 1991 topical migration module: 10.6% Panel Study of Income Dynamics (PSID) A continuing intergenerational survey of families on demographic, economic, family structure, composition and sources of income, housing, and health issues information (sample of Latinos added in 1990) 1968 to present 16,000 individuals 1990 sample added 7,500 Latinos to the original sample: many were native born National Longitudinal Survey of Youth (NLSY) National random sample of youth on a wide range of variables, including labor market experience, marital status, education, and health. Oversamples Hispanic, black, and economically disadvantaged 1979 to present 12,686 individuals who were ages 14–21 in the year 1979 874 in 1979:644 in 1990, when immigration module was added National Education Longitudinal Study (NELS) A continuing survey that provides trend data about critical transition experiences by young people as they develop, attend school, and embark on careers. The survey furnishes information on how school policies, teacher practices, and family involvement affect student educational outcomes 1988 (8th grade) 1990 (10th grade) 1992 (12th grade) 1988:29,000 persons 1990:25,000 persons 1992:21,000 persons Foreign born: 6.6 % plus 0.3 % Puerto Ricans National Maternal and Infant Health Survey Broad-spectrum survey aimed at augmenting birth certificate information. Collected data covered income, entitlement, social support, depression, prenatal care, hospital/health care providers, child health, child care, and use of federal programs. 1988 (at birth) follow-up in 1991; planning a 1997–98 survey follow-up 9,953 live births 5,332 fetal deaths 3,309 infant deaths None. Children were born in the United States; some mothers were foreign-born National Survey of Families and Households Survey at primary respondents 19 years or over, living in a household, with a self-administered portion for spouse or partner. The survey covered a broad range of family domains: composition, living arrangements, growing up, leaving home, education, employment, marriage/divorce, cohabitation, fertility, relations with elder parents, etc. 1987–1988, reinterviewing 1993 13,017 primary respondents (does not include spouses or partners) 910 foreign born plus 120 Puerto Ricans Health and Retirement Survey Data collection focused on retirement decisions, recognizing roles played by: pension and social security retirement incentives, job demands and worker capacities, health and longevity, family responsibilities, husband/wife career choices, economic status, and housing and location choices Survey started 1992, scheduled for every other year for 12+ years 12,656 persons (7,600 households) Approximately 10% National Agricultural Workers Survey (NAWS) Survey of farmworkers that provides data on the occupational activities of agricultural workers and families, job history matrices, descriptions of employment conditions (pay, health care, equipment, pesticide use), and information on social service use and work authorization. 1988 to present, but longitudinal aspect ended 1992 Approximately 2,500 persons Approximately 60% National Survey of College Graduates (NSCG) Survey of college graduates (identified on 1990 census) to gather information on education and employment history, merged with 1990 long form census data. Oversampled scientists and engineers. 1993 through 2001 151,000 individuals Approximately 30,000
OCR for page 77
Complete table found on previous page.
OCR for page 78
--> Survey Description Survey Dates Original Panel Sample Size Foreign-Born in Sample Survey of Doctorate Recipients Ongoing survey of recipients of Ph.D. degrees in the fields of engineering, sciences, and humanities on topics of income and employment. In 1991, discontinued questions on foreign-born doctorates. Every two years, 1973 to present (two follow-ups each survey year) 60,000 persons Approximately 20% Comparative Longitudinal Asian Immigration (CLAIM) Project Survey of Korean and Filipino adult (ages 18–69) visa holders (at time visa application made) asking for employment and self-employment information before immigration and in the United States. Survey started in 1986–1987, resurveyed in 1988–1989 and 1991–1992 Seoul: 1,834 individuals All Manila: 2,077 individuals All Children of Immigrants: The Adaptation Process of the Second Generation Study of 8th and 9th grade children of immigrants (offspring with at least one immigrant parent) in Miami, Fort Lauderdale, and San Diego, either born in the U.S. or having 5 years residence. The survey looks at models of assimilation among minorities, including language, ethnic self-identification, socioeconomic status, aspirations, social networks, and school performance 1992; Follow-up planned for 1995 5,267 individuals 3,194 National Survey of Recent College Graduates A survey of 1990, 1991, and 1992 bachelor's and master's graduates in science and engineering to collect information on long-term education and career progress of persons in science and engineering. The survey is based on the graduation lists of a sample of the U.S. institutions Survey started 1992; every other year through 2001 28,000 persons Approximately 10% Legalized Population Surveys (LPS1 and LPS2) Survey of population legalized under Section 245A of the Immigration Reform and Control Act to provide Congress with a profile of aliens at the time of legalization and 5 years after legalization. Both surveys collect information on language proficiency, immigration and migration, employment, family composition, health insurance and child care, social services, education, income, and remittances. LPS1 also collects employment (prior to U.S. entry, an entry, and at time of legalization application), and health expenditures. LPS2 also collects employment (work history from January 1991 through the survey date, benefits (including health insurance), assets, and household expenses (to determine food stamp eligibility) LPS1: winter/spring 1989; LPS2: spring/summer 1992. No further surveys planned LPS1:7,000 sampled; 6,193 interviewed. LPS2:5,000 sampled; 4,012 interviewed All Annual Survey of Refugees Survey of current employment or job search activity, past work history, income, English ability, household composition, social services received, and household use of welfare. The Refugee Act of 1980 requires that the annual report of the Refugee Resettlement Program contain an updated profile of employment and labor force statistics for refugees who entered the United States within a period of five years immediately preceding the year of the report. Survey started 1984; conducted annually in September–October. Rotating panel design with plans to continue indefinitely Approximately 150 of each year's arrivals. All a This column also lists Puerto Ricans in the sample, event though they are U.S. citizens.
OCR for page 79
Complete table found on previous page.
OCR for page 80
--> Several longitudinal data sets may already yield useful information on immigration. Following the workshop, National Research Council staff collected information on existing longitudinal surveys with an immigrant component. Table 3 displays information on 15 longitudinal surveys. Some of the surveys have general information (especially the Panel Study of Income Dynamics), and others have a special focus (such as the National Longitudinal Survey of Youth and the Health and Retirement Survey). Most of the surveys include a general sample of the foreign-horn population, although a few are limited to a specific pan-ethnic group or to one or two nationality groups. Despite the fact that the sample sizes of the surveys are often adequate for general analysis of the foreign-born population, most existing surveys are not adequate for analysis of recent immigration by nationality group. Also, it should be emphasized that existing surveys do not include the special data required for immigration policy analysis. The Panel Study of Income Dynamics (PSID) is one of the most valuable social science data sets. In 1990 it added a Hispanic supplement of more than 2,000 people (about two-thirds of whom were foreign-born) and continues to follow up these Hispanic families. This survey offers a pilot test of longitudinal data collection for the immigrant groups included in the Hispanic supplement. Only a small fraction of the Hispanic supplement were recent immigrants; most of the Hispanics in the 1990 supplement were survivors of past immigration or descendants of earlier immigrants. For a longitudinal survey of immigrants, it may be more useful to look selectively at very recent immigrants in order to provide data on current immigration policy. In data collection for its 1990 Hispanic supplement, the PSID did not collect retrospective information. It is especially valuable when beginning a prospective survey, particularly for the study of immigration, to collect migration histories and data on important past events (employment, marriage and childbearing, education). When there is an ongoing longitudinal survey, such as the PSID, with a highly qualified staff and a data dissemination program, it would be cheaper and more feasible to build on such a survey for additional longitudinal data on immigrants. For a major longitudinal survey of immigrants to be started, it may be cost-effective to consider building it onto a major ongoing longitudinal data collection program. But, as noted earlier, some existing longitudinal surveys do not offer an adequate representation of the ethnic and racial diversity of the U.S. population. It would cost considerably more to redesign an existing survey to include an adequate comparison group of the contemporary U.S. population, in addition to the costs of an immigrant subsample. Discussion of current major longitudinal surveys illustrates two dilemmas faced by immigration researchers. First, the PSID now includes a Hispanic supplement but does not collect data on new immigrants. The study design does not recruit immigrants into its cohort data collection. A longitudinal survey of Immigrants would require some ways to replenish the study with new entrants.
OCR for page 81
--> New immigrants might be added each year, or perhaps a special immigrant supplement of recent immigrants could be added every 3 to 5 years. Second, the Survey of Income and Program Participation (SIPP) currently recruits a new sample each year, following the study population for a 32-month period. In 1996, the Bureau of the Census is proposing to change the SIPP survey design and recruit a new and larger sample every 4 years, following the cohort for 4 years. SIPP, however, does not include a sufficiently large sample of immigrants for analysis. The SIPP survey design could oversample for some geographic areas or for foreign-born respondents, providing sufficient numbers of recent immigrants for separate analysis. Oversampling of recent immigrants in SIPP would provide valuable data, but it would also increase the survey complexity and cost. This report has dealt with a number of improvements other than a new longitudinal survey of immigrants that would provide immigration researchers with better data. Although a longitudinal survey of immigrants would be extremely useful, workshop participants recognized that developing and funding such a survey would require a strong and committed coalition of academic and public policy researchers.
Representative terms from entire chapter: