Effects of Imperfect Unit Size Information on Complex Sample Designs and Estimators

Randall K. Powers and John L. Eltinge

Abstract

Work with sample surveys often makes extensive use of measures of size. Two prominent examples are the use of “probability proportional to size“ sampling; and use of size measures in adjustment of survey weights through, e.g., ratio estimation, poststratification or calibration weighting. However, many survey applications use size variables that are imperfect approximations to the idealized size measures that would produce optimal efficiency results. This paper explores the effects that alternative size measures may have on the efficiency of some standard design-estimator pairs. Principal emphasis is placed on numerical results of a simulation study that uses size measures and economic variables available through the Quarterly Census of Employment and Wages of the Bureau of Labor Statistics.