The Future of Survey Sampling

Abstract

The twentieth century saw a dramatic change in the way infor- mation was generated as probability sampling replaced full enumeration. We examine some key events of the past and issues being addressed today to gain perspective on the requirements that might steer changes in the di- rection of survey sampling in the future. The potential for online Web pan- els and other methods that do not use probability sampling are considered. Finally, we conclude with some thoughts on how the future of survey sam- pling might be shaped by unpredictable factors. Contemplating the potential directions that survey sampling may take in the future requires considering the current state of affairs and how we arrived at this point. Despite occasional issues and controversies, it seems fair to say that survey sampling, and probability sampling in particular, has helped transform our view of society and the issues it faces. Perhaps even more importantly, sam- pling has changed how we think about obtaining information from a large pop- ulation. We have seen the transition from full enumeration as the only acceptable method of learning about a population that was the standard in the late nineteenth century to ''partial investigations'' and finally to a full theory of sampling. Kruskal and Mosteller (1980) describe this transformation. This change has made it possible to study a wide range of topics about our society in a cost-effective and timely manner. The findings from surveys are generally accepted in popular, legal, and technical areas, although there are still debates about methods of implementation. Election surveys, both those in the run-up to the election and exit polls, demonstrate both the general acceptance and concerns about methods. Election surveys are very public and their accu- racy is easily judged; the results of these surveys are highly reported and the J. MICHAEL BRICK is a Vice President, Associate Director, and senior statistician at Westat, Rock- ville, MD, USA, and a research professor of the Joint Program in Survey Methodology at the Uni- versity of Maryland, College Park, MD, USA. He would like to thank the editors for the invitation to prepare this article and for their suggestions and comments. He would also like to thank Pat Dean Brick, Jill Montaquila, and Bruce Allen of Westat for their constructive comments on an earlier draft and their sharing of thoughts on new directions for survey research. *Address correspondence to J. Michael Brick, Westat, 1600 Research Boulevard, Rockville, MD 20850, USA; e-mail: MikeBrick@ westat.com.