For the last three decades, assessments of the future of survey data collection have had a significant focus on longitudinal surveys. Among the three major surveys sponsored by NSF, the PSID
has always been known as longitudinal survey – and for good reason. It tracked respondents coming from the Survey of Economic Opportunity plus an augmentation sample, for decades.
Using yearly, and later biennial, surveysthe PSID collects data on the household and its evolution through time. The other two main NSF-sponsored surveys - the ANES and the GSS are primarily thought of as cross-sectional, but longitudinal elements have been present in these two surveys for decades. Increasingly, survey efforts that most observers identify as cross-sectional include panel aspects, sometimes unintentionally. For example, the IPUMS project at the University of Minnesota takes the quintessential cross-sectional survey – the U.S. Decennial Census – and links records together to create a panel. The American Community Survey (ACS) is considered a cross sectional survey2 , but it can be used to screen for people with certain characteristics for more detailed follow-up. For example, the ACS has been proposed as the basis for the National Science Foundation’s data collection efforts on science and engineering education and advanced degree recipients. For certain respondents, the ACS would effectively morph into a longitudinal survey. New efforts to re-purpose data from a cross sectional effort so they become longitudinal reinforce the common impression that longitudinal surveys are central to the future of survey research. The second-highest profile cross-sectional data set is the Current Population Survey (CPS), however, even it is a panel study with rotation groups that cycle in and out of the survey. As a Web site from HHS says “Aggregation of the monthly data to obtain a longitudinal data set would require the expertise of a skilled statistician”. Such efforts generate analytic dividends and periodically users re-invent a longitudinal file from the CPS.3 The major reservation about longitudinal surveys centers on attrition. This paper seeks to change the way we think about attrition as well as summarize two propositions on the avoidance of attrition.