Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
The Role of Nongovernmental Activities in Immigration Studies Many government agencies and organizations produce data about immigration, as detailed in the two preceding chapters. These data reflect a large, cumbersome, and poorly coordinated official immigration statistics "system." In addition to the government data, however, many unofficial data sources exist. Moreover, most of the current analysis of immigration data is done outside government, by university researchers, foundations, private firms, and others in the private sector. The immigration statistics system is thus even larger (and less systematic) than our discussion has so far indicated. "Unofficial" data consist mainly of those collected by individual researchers and not-for-pro£it institutions or for-profit firms. There are some grey definitional areas in this classification, however. In some cases, government agencies finance the collection of data by nongovernment organizations and individuals: for example, the General Social Survey is conducted by the National Opinion Research Center with support from the National Science Foundation, and its data have been used to compare immigrants with the native-born. Although such programs are funded by the government, the government usually disclaims responsibility for the data, so we classify them as unofficial. There are also data collected by foreign governments or quasi-governmental agencies, such as the United Nations High Commission for Refugees. These data, although official in some contexts, are unofficial as they are used within the United States. There are large numbers of small-scale data collection activities in the area of immigration carried out by individual researchers, substantial numbers of larger-scale activities carried out by institutions and companies, and large-scale operations carried out by foreign governments and international organizations. It is beyond the scope of this study to attempt to review the quality and relevance of all these unofficial data sets, even for a representative sample of them. Thus in this chapter we do not attempt to review in detail the information collected and the collection processes involved. Instead, we examine what the role of unofficial data should be in an overall system of immigration statistics: what sorts of data collection are best left to nongovernmental bodies, and how needed data collection can be stimulated and financed. A new departure in this chapter is the explicit discussion of data analysis and its implications for data collection. 101
102 Good analysis is an essential link in the chain from raw data to policy decision and is a topic suitable for this nongovernmental chapter because, with the exception of the Census Bureau's research activities, most analysis of migration data is conducted outside government. DATA PRODUCTION AND ANALYSIS Data production and data analysis are separate activities, although analysis may be constrained by the ways in which data are produced, and collection may be tailored to the intended analysis. Data production refers to the collection, compilation, coding, and storage of basic data concerning immigration. Data are generally produced intentionally, but the intentions of the producer may not coincide with the requirements of the analyst. To take an example, most of the data collected by the INS are collected for programmatic and administrative purposes rather than for policy analysis, but some of them can be used for analytical purposes, sometimes in ways never dreamt of by those establishing the collection process. The production of immigration statistics is very decentralized. As we have already noted, a number of federal agencies regularly produce data on immigrants or refugees, either as a primary objective or as a by-product of their other activities. Furthermore, the states provide at least some information on vital events occurring to immigrants, refugee program use, bilingual education, and other topics that vary by state. Outside government, there is, if anything, even greater diversity and decentralization. Unofficial data are produced by intensive ethnographic studies of immigrants and of communities of both origin and destination of migrants, local surveys and compilations of local data, reconstructions of historical series from existing but unsystematized data, and many other collection processes. In addition, as noted, there are also data from foreign governments and international agencies. However, data quantity does not compensate for data quality. Existing nongovernmental data sets are often inadequate in terms of coverage, validity, and reliability of the data they contain, or because they are impossible to compare or to integrate with other official or unofficial data. Analysts have tried to solve some of these problems with more creative uses of the existing data, but the root of the problem lies in the decentralized nature of this nonofficial data production system. Data analysis includes both primary analysis (that is, to examine the questions for which the data were collected) and secondary analysis (that is, to examine questions for which the data are relevant, though not the purpose for which they were collected). Though data are generally subjected to primary analysis, secondary analysis is even more widespread because of the high cos ts of data collection. The agenda for analysis is set by current intellectual issues, policy debates, theoretical developments in one or more of the social sciences, or simply the curiosity of an investigator. Such studies, usually published in professional journals, monographs, or the popular press, may not achieve their full potential impact because their audiences are specialized along political, interest, or disciplinary lines. To the public or the policy maker, the resulting debates and controversies, especially those centering on issues of data adequacy, may seem partisan
103 or merely arcane. Just as diverse motivations underlie the collection of data by unofficial sources, so too the analysis of data by unofficial analysts responds to many goals. A lack of consensus about important issues among analysts may be a sign not of factionalism, but rather of healthy diversity. Primary analyses of official data may be guided by official or quasi-official perceptions of what is required and of the frequency with which they are required. The Census Bureau is an example of a statistical agency that both produces data and publishes analyses of these data, many at regular intervals, others as occasional papers. Comparisons of official data sources may occasionally be undertaken to provide statistical benchmarks. However, analyses of immigration data under official auspices have been relatively rare. The intellectual division of labor has allocated this task, often by default, to indiviudal researchers outside government. A considerable volume of recent social science research about immigration has been secondary analysis of official, census-type survey data, most important the 1960 and 1970 census public-use samples and the 1976 Survey of Income and Education. Reliance on such data, however, greatly restricts the range of substantive issues that can be addressed and the analytical approaches that can be pursued. Given the limited social and economic variables available in census files and the virtual absence of cultural indicators, it is not surprising that most immigration studies relying on census data focus on the socioeconomic characteristics of the foreign-born population, usually differentiated by national origin and occasionally by period of arrival. Census micro data encourage the use of individuals or households as units of analysis, but analysis by aggregates such as area of residence could portray the macro dimensions of social phenomena. Multilevel analyses combining person and place variables are less common, but not entirely absent from the literature. The extensive reliance of researchers on official data is evident in the studies noted in the major bibliographies on immigration (see Appendix F.), although the coverage of small-scale and ethnographic studies by the bibliographies is not entirely complete. It can thus be seen that the analysis of immigration data has been profoundly affected by the collection of data. Researchers face the choice of either collecting their own data, which for reasons of cost will be restricted in general to surveys collecting extensive information from small numbers of people, or using official data generally collected for other purposes and thus of limited analytical potential. It should be noted, however, that the dearth of quality research studies in the field of immigration should not be blamed entirely on data deficiencies. Even the limited data available from official sources offer a potential for analytical study that has not been fully exploited. Even the analysis of large data sets is expensive, and funding for immigration research in the recent past has not been sufficient to support extensive analysis or to attract research analysts to the immigration field. TYPES OF UNOFFICIAL DATA An important distinction must be made between studies based on statistical data (e.g., social surveys) and those based on nonstatistical
104 data (e.g., ethnographic, archival, and historical material). The data collection processes involved should be viewed as dif ferent parts of a continuum, from the quick, extensive (in terms of population coverage) yet limited ~ in terms of topics covered and questionnaire length) collection process of the traditional survey to the slow, limited (in terms of population coverage) and intensive ~ in terms of interview length and topics covered) collection process of the ethnographic study. The survey typically collects information about respondent status, either at the time of the survey ~ age, marriage, income, and education, for example) or at some specified earlier time (place of residence five years before the survey, for example) for a statistically representative sample of the study population. The ethnographic survey, on the other hand, collects a wealth of additional info`Q.ation about opinions, motivations, and community context, often through the use of open-ended questions, but the survey population is generally selected purposively and cannot be taken as representative of any larger population of interest. Each data type has its relative strengths. However, the potential for complementarily among various types of unof f icia 1 data is af fee ted by the problems that arise when combining data from dif ferent sources and by the analytical approaches that have been used to address specific questions. Studies based on both ethnographic and survey data have generated useful insights into immigration as a social phenomenon. Both survey and ethnographic data sources can be tailored to specific substantive questions about immigration as a social process whose causes and consequences extend from individuals to communities of origin and destination. Their scope, depth, and the generalizability of the information differ considerably, however, as a result of the qualitative nature of ethnographic data and the quantitative.nature of survey data, as well as the manner of solic iting, recording, and coding informal ion . Ethnographic data benefit from greater respondent flexibility but often at the cost of generalizability; survey data often sacrifice the richness of open-ended responses to facilitate coding and to obtain a standardized data set. Ethnographic Data The main strengths of ethnographic data lie in their depth and comprehensive coverage of an immigrant community or a specific aspect of the immigration process. Ethnographic data provide a great deal more information about social processes and interactions that structure immigration flows than do conventional survey data. Of course, the comprehensiveness of information produced depends on the amount of time spent in the field and the number of field sites. Ethnographic data are generally not exchangeable between analysts, so they are not easily subjected to secondary analysis, integration with other data sets , or verif iciest ion except through restudy by dif ferent invest igators . However, the ethnographic practice of soliciting information from multiple sources--participant observation, key community informants, and respondent interviews--provides some basis for internal data consistency checks and for response validation. The major drawbacks of ethnographic information are the limits on exchange of information among researchers, the limited generalizability
105 of the results beyond the population or locality of study, and the difficulty of subjecting the data to rigorous hypothesis testing using multivariate statistical techniques. The major advantages of ethnographic data are the depth of study they permit into the reasons for and process of migration; survey data on the social and economic characteristics of immigrants can provide only a partial, and possibly misleading, view of such reasons and processes. Survey Data The strengths of unofficial survey data complement the weaknesses of ethnographic data and vice versa. The primary strengths of unofficial surveys relative to official sources are their flexibility in selecting the number and scope of topics to be included and their ability to ask sensitive questions; relative to ethnographic studies, their strengths are that the population universe can be clearly and explicitly defined, that they adhere to statistical standards permitting the evaluation of possible sampling errors, and that they are amenable to rigorous empirical analysis. None of these strengths is absolute, for compromises in scope, representativeness, population coverage, and quantifiability are imposed by cost and time considerations. Moreover, some aspects of immigration are difficult to capture using random sampling survey techniques. The obvious example to date is that of illegal immigration. Nonrandom sampling methods, such as network samples, whereby one member of the study population is asked to provide names and addresses of other members of the population, who are then in turn interviewed and asked to provide more names, have been used with limited success, but the process of statistical inference is seriously impaired. (Such samples are sometimes also called snowball samples.) Other concerns not easily pursued with survey--or ethnographic--data are the macro-structural properties of the immigration process, including the changing nature, direction, and composition of aggregate flows; fortunately, official data sources are especially well suited to such issues. Despite the many virtues of survey data for addressing questions about immigration, these unofficial data sets suffer from several drawbacks. Most surveys give a one-time static snapshot of social and economic status that provides only limited information about the process of arriving at that status. Thus, most cross-sectional surveys of immigrants are limited in their ability to address questions about process or to establish clearly causal relationships. The exceptions are those few surveys that have collected retrospective histories of the timing of various events, such as migration, employment, and childbearing. The dating of changes in social and demographic status permits a more effective study of process, although events in the more distant past may not be representative of their time period, since the sample is representative of the present, and event intervals in the period shortly before the survey may be affected by censoring and truncation biases. An alternative to the use of retrospective questions in sample surveys involves reinterviewing a sample of respondents several times, for example every six months or every year. This longitudinal, panel approach may be preferable to that of repeated cross-sections for making
106 inferences about process because it permits control for previous events in a sequence without being subject to major event-dating errors that al feet life-history reports. However, cost factors have inhibited individual researchers from undertaking truly longitudinal surveys of immigrants. Moreover, immigrants often have a high propensity to move, which usually increases sample attrition and can, over time, impair the representativeness of the sample. Complexity and cost factors aside, it is noteworthy that there does not currently exist a nationally representative longitudinal study of recent immigrants. Longitudinal studies of the general population do not include a suf ficient number of immigrants to permit separate analysis even at the aggregate level and still less for nationality or other subgroups. The strategy used to define a universe and devise a sampling scheme may limit the usefulness of multipurpose surveys for studying immigrants: for example, the General Social Survey includes only the English-speaking population over age 18. Furthermore, even leaving aside the special problem of studying the illegal immigrant population, it is. not obvious how to design a survey to study the determinants and consequences of migration for the community and country of both origin and destination. With few exceptions, most sample surveys of immigrants have defined the universe on the basis of those who actually move across international boundaries and settle in a specific locality or who cross at a specific time. Such strategies for limiting the universe are appropriate for addressing questions about the experiences of immigrants in the destination country, but these samples limit interpretations of the causes and consequences of international migration in at least two important ways. First, by excluding those who decide not to emigrate, studies based on samples of individuals who have migrated across international boundaries distort our understanding of the determinants of migration and lead to potentially erroneous conclusions about the nature of migrant selectivity. Second, universes defined by time and locality, especially the latter, exclude an unknown number of immigrants who may have returned to their place of origin or moved on to another destination. This latter problem can be partly resolved by inquiring about past migration history, intended moves, and the existence of friends and relatives in other localities, but it introduces selection problems of unknown magnitude in the statistical analysis of the survey data and may ultimately distort conclusions about the individual, familial, and locational structure of aggregate flows. Survey design may al so at fee t the potent ial to study impact, since it is c [early neces sary in such a case to have information not only on migrants themselves but also on the rest of the community and on other communities. To summarize, the main advantages of survey data reside in their generalizability, their amenability to rigorous statistical analysis, and their high degree of exchangeability among researchers. In practice, however, the access to and distribution of data from specialized surveys about immigrants is not extensive, and there is no central clearinghouse that receives, classifies, and distributes data sets containing information about immigrants to interested researchers. The generalizability, substantive content, and amenability of unofficial surveys to rigorous secondary analyses of immigration issues varies with the objectives and design of the original data collection. Although cross-sectional surveys can be designed to deal with the timing of events
107 through the use of retrospective questions or through repeated surveys, and longitudinal surveys can be designed to examine processes, each strategy poses different problems of cost, sample attrition, recall error, and analytical limitation. Immigrant Case Records One relatively unexplored avenue for immigrant studies is the use of case records collected by private voluntary agencies that assist immigrants or refugees. These files provide a basis for following immigrants for a period of time and for noting their adaptation to life in the United States. Assuming that the necessary standards for confidentiality could be met, such data would offer many of the advantages of a longitudinal survey at a fraction of the cost. Potential Complementarities Among Unofficial Data Sources Although there are a number of possible combinations of data types, three particular combinations are promising for research. We term these three combinations multilevel studies, multimethod studies, and multi-data-set validations. A multilevel study uses combinations of data aggregated at different levels to establish a finding: for example, individual or household data might be used to confirm or enhance conclusions based on aggregate data. A multimethod study combines fundamentally different types of data: an example is the way ethnographic and sample survey data are used to complement each other in the study of Mexican migration to California undertaken by Massey (see Appendix C). Such studies frequently combine official and unofficial data, exploiting the relative strengths of each. The third category involves the cross-validation of a finding using different data sets that cover the same population or variable of interest. For example, U.S. estimates of immigration from a country might be compared with that country's estimated emigration to the United States. Such studies also typical ly require combinations of official and unof ficial data. Even given current data production systems, these strategies appear underexploited and offer scope for useful research effort. OBSTACLES TO DATA ANALYSIS The bulk of analysis of immigration data, whether collected under official or unofficial auspices, is done by the private sector, but accessible data can be analyzed. Except for the U.S. Census Bureau, agencies that produce official data either have been largely unaware of or unresponsive to the data needs of the research community. Problems of accessibility and ease of use represent an obstacle to data analysis. Many data sources remain inaccessible for reasons that cannot be explained by privacy or confidentiality concerns alone. Even for the data that are accessible, documentation is often sketchy or unavailable. Coding protocols are not explained, so that the effects of coding practices that differ from one source to another, or even within sources,
108 may be overlooked. With better access to existing data, the research community could produce more relevant and higher-quality analyses. The existence of such analyses is essential to better policy formation, since data per se, in the absence of any examination of implications, give no guidance for policy. The analysis is impossible without the data, but the data, to be useful, must be analyzed. Greater interaction with ache research community would provide a mechanism for improving the policy relevance of of ficially produced data. Given an understanding of analytical needs, data could be produced in more convenient forms . The expert advisory panel and the fellows program, recommended in Chapter 4, would provide the INS with an important source of expertise and feedback in improving its data collection. Contact with users need not be expensive and does not necessarily require the establishment of a permanent users service. Annual meetings of the professional associations of the research disciplines offer an opportunity to disseminate information about data products and services. The foundations and journals active in the immigration field can serve a similar function. Regular INS publications could also provide information about data availability and changes in data produc t ion prac tices . A second obstacle to analysis is the shortage of funding for immigration studies. This shortage has been particularly severe for unofficial data collection, which is generally expensive, but has also restricted the analysis of official data and professional interest in the field. Skepticism about data quality may have made immigration studies less attractive to such major grant-giving agencies as the National Science Foundation or the National Institutes of Health, even though data evaluation alone would represent a worthwhile outcome. Once again, the problems of data production haunt data analysis, although indirectly in this case. It should be noted, however, that the National Institutes of Health, through the National Institute of Child Health and Human Development, have been making efforts recently to encourage the submission of research proposals in the immigration field and to increase the funding allotted to it. The INS also could support a program for immigration studies channeled through the conventional funding agencies, which would apply their usual peer review and grant procedures. This approach would provide a mechanism for the contract research program already recommended in Chapter 4. The Office of Refugee Resettlement and other agencies concerned with refugees might enter into similar agreements to support research on refugees. SUMM,9RY AND RECOMMENDATIONS Unofficial data complement official data in important ways. Furthermore, most studies of immigration are now carried out by nongovernment researchers. However, problems of accessibility and quality of official data, and shortage of funds for unofficial data collection and analysis in general have severely limited the contribution of the nongovernment sector to the policy formation process. To improve this contribution, the panel recommends:
109 o Insofar as is feasible, official government data on immigrants and refugees should be made available to researchers outside the government; o The proposed Division of Immigration Statistics in the INS should establish and maintain contacts with the research community and keep it informed about the availability of data and changes in procedures. This recommendation also applies to all other agencies that produce immigrat ion data; and o Government agencies that provide funds for research should be encouraged to stimulate the submission of research proposals in the immigration field and to give particular attention to sound proposals for relevant research studies in the area.