Investments in Longitudinal Surveys, Databases, Advanced Statistical Research, and Computation Technology
Robert M. Hauser
University of Wisconsin
Dr. Hauser introduced his presentation as a “survey of surveys”—in particular, of major longitudinal surveys. First he described their features:
-
They represent real populations, either regional or national.
-
The number of cases is large, usually several thousand households or persons.
-
Their content covers a wide window in time, preferably decades rather than years.
-
They measure key variables in multiple content domains and thus invite multidisciplinary research.
-
The same variables are measured repeatedly across time, or retrospective histories are obtained.
The remainder of this “survey of [major longitudinal] surveys” would proceed in the following order: the big ones, epidemiological panel studies, some small (but good) ones, past or lost ones, offshore properties, and future prospects.
THE BIG ONES
These include the Health and Retirement Survey, the Longitudinal Studies of Aging, the National Long-Term Care Survey, the National Sur-
vey of Families and Households, Americans’ Changing Lives, the Panel Study of Income Dynamics, and the Wisconsin Longitudinal Study.
The Health and Retirement Survey is a prospective study, now including more than 20,000 adults. It started in 1992-1993, and it has basically become a continuous longitudinal survey of all cohorts in the United States over the age of 50. “This is just a terrific vehicle; it is the Cadillac of such surveys,” Dr. Hauser said. “It doesn’t attempt to do everything for everybody. There is a specific emphasis on policies affecting retirement, health insurance, savings, and economic well-being.” Some of its topics include health, disability and cognition, retirement plans, a variety of attitudes and preferences, family structure and transfers, employment status, job history and requirements, housing, income and net worth, health insurance, and pension plans.
The Longitudinal Studies of Aging were two six-year panel studies—the first in 1984, the second in 1994, with each followed biennially through four waves. The samples were drawn from the National Health Interview Survey, Dr. Hauser said, and the two studies ascertain a long list of self-reported social and health conditions such as housing characteristics, family structure and living arrangements, relationships and social contracts, and use of community services.
The National Long-Term Care Survey has gone through five panels from the early 1980s through 1999. Though it addresses the entire aged population—data are drawn from Medicare records—there is a focus on the functionally impaired. It provides very good data on trends in disability and mortality, he said.
The National Survey of Families and Households, which began in 1987-1988 and is now entering its third wave, “has really reinvented the sociology of the family in the United States,” Dr. Hauser said. The design study focuses on relationships between parents and children, spouses and exspouses, ex-spouses and children, cohabitors—“all of these possible role relationships that now exist in abundance among American families.”
Americans’ Changing Lives, another broad-based population survey that started with 3,600 cases in 1986, has gone through three waves and is about to enter its fourth. It focuses on “productive” social relationships and cross-cultural variations within them, as well as on stressful events, chronic strains, and their effects on health functioning and productive activity. This survey has a substantial oversample of black Americans, Dr. Hauser said, and a lot of the activity focuses on comparisons between blacks and whites.
The Panel Study of Income Dynamics is an annual household survey
that started with a poverty focus in 1968 and included 5,000 families. It has grown in size because “everybody gets followed, wherever they go, as they leave the original households and form new households,” Dr. Hauser said. Its content now involves a central focus on demographic and economic stability and change, but there are numerous supplements—for example, the National Institute on Aging has supported supplements on health, health expenditures, housing, savings, pensions, and retirement.
The Wisconsin Longitudinal Study started out in 1957 with 10,317 high school graduates who were followed in 1964, 1975, 1992, and “we hope, again soon,” Dr. Hauser said. The study has mostly focused on education, careers, and families and “is very rich in measures obtained in adolescence— including IQ, educational and occupational aspirations, social background, and other kinds of social influences” (such as school characteristics). Since 1992, he noted, “we have been picking up health, wealth, and well-being.”
EPIDEMIOLOGICAL PANEL STUDIES
Dr. Hauser mentioned two well-known examples: the Alameda County Health and Ways of Living Study, “which followed a very nice long study of a community sample from the mid-1960s to the mid-1990s,” and the Established Populations for Epidemiologic Studies of the Elderly, which covered 1981-1993.
SOME SMALL (BUT GOOD) ONES
The Terman study “is in some sense a model for everything that follows and in other ways absolutely horrible,” Dr. Hauser said. “They selected all of the participants in the study on the basis of the value of the one variable that they thought was important—mental ability. Then the study went forward from there. But they followed these people all the way—from childhood to middle childhood to maturity and beyond—and they did that very well indeed.”
Another three very small studies tend to get grouped together, he said, because their data are often analyzed together: the Berkeley Growth Study, the Berkeley Guidance Study, and the Oakland Growth Study. Then there is the Nun Study—a small survey of a select group. “Yet it has so far produced at least two absolutely astonishing and important findings,” Dr. Hauser said, “because the window of observation is so long—from late adolescence through senescence and death—and because the quality of
measurement is extraordinary across that span and because the content spans the social, the psychological, and the biological. That is where the payoff is, and what we have to do is get that kind of payoff from general population studies.” Finally, the MacArthur studies of successful aging, also relatively small, “have given us some wonderful and provocative findings,” he said, as mentioned earlier in this symposium.
PAST OR LOST ONES
Two examples are the now-defunct Project Talent—“a study of adolescents in 1960 that originally had some 400,000 high school kids enrolled, and quite possibly the worst longitudinal study ever conducted,” Dr. Hauser said, and the National Longitudinal Studies, which focused “on labor market experience and are pretty much in the past.” On the other hand, a historical reconstruction survey by the University of Chicago’s Robert Fogel and collaborators—based on records of men who were in the Union Army during the Civil War—is “amazingly wonderful”—an “important historical record that . . . connects the medical and the social.”
OFFSHORE PROPERTIES
“The British have done this better than anybody else,” Dr. Hauser said, “with birth cohorts of 1946, 1958, and 1970 that cover all of the right stuff.” Other nations—Germany and Australia, for example—have done excellent work as well. And the Indonesian and Malaysian Family Life Surveys, each spanning about a dozen years, “show that you can do very high-quality longitudinal research, again combining the social and the biomedical, in places where we wouldn’t think of it as being so easy to do that kind of work.”
FUTURE PROSPECTS
“The real point here is that we have to start thinking now about very young people who, with any luck at all, are going to get to be old,” Dr. Hauser said. He referred to several school-based national cohort studies that ought to be followed up. Similarly with regard to household-based national cohort studies—such as the National Longitudinal Study of Youth, from 1979, which “has been immensely successful throughout its life.” In the new panel of 1997, he said, “people should be thinking now about
what we would like to know [about those kids] 50 or 60 or even 70 years from now.”
STRENGTHS AND WEAKNESSES
“The real strength of longitudinal studies,” Hauser said, “is that they give you ‘the big picture,’ a true description of the life course in real populations.” Moreover, he added, “they permit quantitative analysis of social, economic, psychological, and biological processes as they occur. Studies that are relational or multilevel in design—e.g., including spouses, siblings, schools, employers, and localities—can provide important findings about social interaction and environmental or contextual influences on the life course.” A final advantage, he said, is that, “because of the large number of participants in such studies, it is sometimes possible to study what happens in rare or narrowly defined populations—e.g., parents who have experienced the death or mental illness of a child.”
“Most of the studies that I have described to you are of short duration—at least are still of short duration,” Dr. Hauser added. “So we have to have the commitment to keep them going. We can’t get tired of them any time too soon.”
He also pointed out some of the challenges inherent in such studies. Many “are short on bioindicators. . . . With most of them, the only bioindicator we have is death.” Also, “rich as they are, they can’t really help us to solve questions of the difference between causation and correlation.” They are expensive—the Health and Retirement Survey alone now costs about $9 million a year. There are real problems of coverage; many of these studies “have very serious problems of attrition. And some of them actually have rather poor coverage on the first round.”
In the final phase of his talk, Dr. Hauser enumerated the goals for a system of longitudinal studies as follows:
-
Multidisciplinary. “We need multidisciplinary integration in design, measurement, and analysis.”
-
Multiplicity. “We shouldn’t put all of our eggs in one basket.”
-
Overlap and complementarity. We need “different studies specializing in different things, not having every [study] try to do exactly the same thing.”
-
Comparability among studies.
-
Data publicly available. Among researchers and scholars in the social
-
and economic area, “nobody owns any data. And that is the way it ought to be across the board.”
-
Continuity. “We need to have planning so that we are always getting new data. We are only beginning at the beginning of people’s lives, and we will need to be able to reap the harvest regularly and not on the basis of occasional one-shot expenditures.”