Estimating Demographic Characteristics Using the Ratio-Correlation Method

Abstract

Studies of the relative accuracy of methods of estimating the population at the substate level have generally found that the greatest degree of accuracy is provided by the ratio-correlation, or regression, method. This paper reports on research aimed at evaluating the effectiveness of this method in estimating the age and race composition of populations at the substate level. In addition to the basic multiple regression equation, variants such as stratification and the averaging of estimates from simple regression equations are also tested. Surprisingly, for Virginia localities the most satisfactory results are generated by the nonstratified multiple regression equation.