Abstract
The error of closure in the 2000 census, or the difference between estimated and enumerated populations, poses special problems for public agencies that rely on census data. Nationally and at the state level, populations were only slightly underestimated, but underestimations were high in rapidly growing counties in the South and intermontaine West, as well as in New York City. Race-specific estimates proved far less reliable, with severe overestimates and underestimates of all racial groups in various counties nationwide. We offer explanations for the estimation error and discuss its impact on cancer rates and trends and its implications for cancer surveillance research.
Notes
1The same technically holds true for data by sex, but sex ratios within and between counties in a state over a decade are quite stable.
2Aside from estimates of such small populations being of dubious value, below about thirty persons, the flexible definition of “accurate” encountered limitations, as the distribution of estimates became skewed, with overestimates being more likely than underestimates. Below about fifteen persons, the moving standard deviation typically exceeded the estimate, making it impossible for a county to be underestimated. This was resulting in spurious patterns of overestimation in places like the Dakotas.
*The authors would like to thank James Weed, Dean Judson, Ben Hankey, Barnali Das, and the anonymous reviewers for helpful comments and suggestions.