Individual versus aggregate measures of digit preference

Abstract

Digit preference (DP) has been recognized by a disproportionate number of responses ending in preferred digits (e.g., 0 or 5). Individuals have been defined as exhibiting DP on a particular variable if their responses end in these preferred digits. Since 20% of all numbers end in 0 or 5, however, this individual‐based definition of DP is biased. In this paper, an alternative aggregate measure of digit preference (ADP_0or5) is defined and contrasted with the individual‐based measure (IDP_0or5). Data for this presentation were derived from a recent article (Vaske, Beaman, Manfredo, Covey, & Knox, 1996) addressing the influence of response strategy (e.g., record keeping, guessing) and recall frame (short versus long) on digit preference. The aggregate measure of DP shows almost no digit preference (5%) for those who were recalling short time frames and who kept records of their angling participation. The individual‐based measure indicated that 15% of the sample exhibited DP. For the long recall frame and respondents who guessed at their participation, aggregate digit preference = .64, while individual digit preference = .72. By moving to an aggregate measure of DP and by examining the structure of stated value frequencies, the paper provides the foundation for a general definition of DP, and the formulation of statistical tests and models that allow the magnitude of DP biases to be investigated.

Keywords:

• Heaping
• digit preference measures
• bias