Bias in Mail Surveys Cannot Be Controlled by One Mailing

Abstract

Experimental data fail to show that a test for random order of response can detect bias after one mailing, as has been suggested. Indeed, use of such a test may lead to a conclusion of “no bias” when a large bias does in fact exist. Dr. Ford is a member of the statisticians Division of the American Telephone and Telegraph Company. Dr. Zeisel, author of Say It with Figures and co-author of Marienthal, is associate Director of Research of McCann-Erickson and lecturer at the New School of Social Research.