The Delphi Method, III: Use of Self-Ratings To Improve Group Estimates

Abstract

An analysis of the validity of using self-rating as a technique for selecting more accurate subgroups in applications of the Delphi method for eliciting group judgments. A series of experiments was conducted using 16 groups of upper-class and graduate college students answering almanac-type questions (20 subjects per group and 20 questions per subject). The findings indicate that if the difference in average self-rating between the subgroups is substantial, and if the subgroups are held to reasonable size, both the degree of improvement and the total number of improvements are greater than when feedback alone is used. This study augments the results reported in RM-5888 and RM-5957.