Spring 2001 Buehler-Martin Distinguished Lecture Series - April 16, 17
and 19, 2001
University of Minnesota
School of Statistics
College of Liberal Arts
Statistical Issues in Census 2000
David Freedman
Professor of Statistics
University of California, Berkeley
Monday, April 16, 2001
4:00 PM,
170
Physics
Minneapolis, East Bank Campus
Social at 3:30 PM,
300
Ford Hall
Abstract
I will discuss Census 2000, focusing on capture-recapture methods to correct
undercounts. Experience with such methods in 1980 and 1990 suggests that they
can add more error than they take out. Adjustment is strongly dependent on
somewhat arbitrary technical decisions, fieldwork is quite complicated, and
there is ample room for measurement errors of different kinds. In 1990, for
instance, the net undercount was estimated as 5.3 million, but 3 to 4 million
out of this total represented errors in the adjustment process rather than
errors in the census.
In 2000, preliminary indications from administrative records (like birth
certificates and death certificates) are that the Census overcounted the
population by a number in the range from 1 to 4 million. Proposed statistical
adjustments would have added another 3 million persons. If the administrative
records are right, adjustment is going in the wrong direction. The Census
Bureau recommended against adjustment, and this advice was accepted by the
Secretary of Commerce.