Blending domain estimates from two victimization surveys with possible bias

Dual frame survey methods are popular for blending estimates from two independent surveys that measure the same quantities. The relative weights assigned to observations from the surveys may be used for direct small domain estimates. In this paper, we explore methods for small domain estimation from two surveys when one survey may be biased relative to the other. These methods are explored in the context of a proposed new companion survey for the US National Crime Victimization Survey. The Canadian Journal of Statistics 40: 679696; 2012 (C) 2012 Statistical Society of Canada