Presents a method for estimating the total national number of doubly missing people (missing from Census counts and adjusted counts as well) and their distribution by race and sex. Application to the 1990 U.S. Census yields an estimate of three million doubly-missing people. (SLD)

Argues that the loss-function analysis used to adjust the census of 1990 depends on estimating probable errors in census and adjustment and that data from field-validation projects does not provide estimates accurate enough for the purpose. The loss-function analysis is actually driven by assumptions, not data. (SLD)

Advocates of the use of ecological regression as an indicator of ethnic voting behavior have not demonstrated its validity. The neighborhood model, which assumes that within each precinct Hispanic Americans and non-Hispanic Americans vote alike, is supported as a better way to draw conclusions about voting behavior. (SLD)

Techniques for adjusting census figures are discussed, with a focus on sampling error, uncertainty of estimates resulting from the luck of sample choice. Computer simulations illustrate the ways in which the smoothing algorithm may make adjustments less, rather than more, accurate. (SLD)

The use of ecological regression in voting rights cases is discussed in the context of a lawsuit against Los Angeles County (California) in 1990. Ecological regression assumes that systematic voting differences between precincts are explained by ethnic differences. An alternative neighborhood model is shown to lead to different conclusions. (SLD)