Abstract
The ratio-correlation method is one of several methods currently being used in the United States for making postcensal estimates of subnational units such as states and counties. It seems to have been commonly presumed by those writing on the subject that the working of the ratiocorrelation method can be understood simply on the basis of the multiple regression theory. That this common notion may sometimes be untenable is demonstrated in this paper. It is pointed out that the ratio-correlation method of subnational population estimation has certain characteristic features that make its application fall sometimes outside the usual contexts in which the conventional multiple regression theory is applicable. A number of alternatives to the ratio-correlation method are suggested. Some of the alternatives suggested are shown to yield relatively more accurate results when used for estimating postcensal populations of North Carolina counties.
