Using Ethnographic Data to Evaluate Dual System Estimates
Sally W. Thurston
KEY WORDS: triple system estimation, movers, census undercount, correlation bias, administrative lists, ethnographic enumeration
ABSTRACT
Administrative or ethnographic lists provide a third source of names and addresses which can be used to expand the 2 x 2 table underlying the dual system estimate into a 3-way table (2 x 2 x 2) in which only one cell is unknown. Use of these lists makes it possible to check the independence assumption underlying the dual system estimate, and to estimate the correlation bias if this assumption is not met. This paper discusses ways in which knowledge of people’s mover status can be incorporated into triple system estimates, and the relative merits of using administrative or ethnographic lists for population estimations. Population estimates and coverage rates for two sites in which independent enumerations were conducted by ethnographers during the 1990 census are compared with and without use of the ethnographers’ lists. The statistical dependency of the census and the Post Enumeration Survey in these two sites is evaluated.
The author concludes that use of a triple system estimator is likely to lead to more accurate estimates of population and of census coverage rates than a dual system estimator, and that estimates using ethnographers’ lists have the potential to be more accurate than administrative lists, in part because the ethnographers’ lists refer to a more relevant time frame.
CITATION: Sally W. Thurston (1995) Using Ethnographic Data to Evaluate Dual System Estimates. Report prepared in partial fulfillment of Joint Statistical Agreement 91-31 with Harvard University. January 4, 1995.
