Sampling and Census 2000-Methodological Issues:
Report to the Donner Foundation

Beth Osborne Daponte, Stephen E. Fienberg, Joseph B. Kadane, Duane L. Steffey


A national census is an extraordinarily complex and difficult undertaking, particularly in a nation as populous and diverse as the United States. Recent American censuses have been the focus of controversy concerning the accuracy of their population estimates and the viability of using sampling and statistical estimation to improve the estimates obtained via traditional methods. With Census Day 2000 less than a year away, this report attempts to provide a broad and useful perspective on current census-related activities and to propose a framework for evaluating alternative census methods in hopes of resolving ongoing technical disputes.

We begin with a brief historical discussion of the purposes and recent criticisms of the decennial census, as well as an overview of the key elements of modern census operations, to provide some background for the remainder of the report. Our primary mission, however, is to discuss key methodological issues in depth and to examine the implications of the 1990 census experience for Census 2000. Such evaluations, however, lead us to consider which criteria are appropriate in judging between more traditional and more innovative census designs. We also consider the relevant experience of countries with similar cultural traditions--namely, Canada, England, and Australia.

Here is the full postscript text for this technical report. It is 270596 bytes long.