No. 1108

Title:

The effect of item uncertainty on item count responses

Author(s):

Takahiro, Tsuchiya (The Institute of Statistical Mathematics);
Yoko, Hirai (Tokyo Metropolitan University)

Key words:

indirect questioning technique, item count technique, undercounting effect, order effect, web survey

Abstract:

    The item count technique is a promising indirect questioning technique that is assumed to yield better results than the conventional direct questioning technique, particularly in cases of illegal or socially undesirable behaviors. The technique requires one group of respondents to indicate merely the number of applicable items from among the target key and other non-key items. Another group of respondents is asked to indicate the number of applicable items from among only the non-key items. The proportion of people who engaged in the keyed behavior is estimated by the difference in the mean numbers of the applicable items between the two respondent groups. However, it is known that the number of applicable items tends to be smaller when they are directly indicated via the item count technique than when it is indicated by the direct responses to each item. In our paper, this inconsistency is referred to as the undercounting effect, and the causes of the undercounting effect are explored on the basis of two hypotheses: uncertainty in judging the applicable items and order effect of the response alternatives. Web survey results revealed that the order of the response alternatives is irrelevant to the effect, but the ambiguity of the item lists principally affects the occurrence of the undercounting effect. Although the survey results suggest that a careful scrutiny of the item lists is indispensable in reducing the undercounting effect, it is insufficient in thoroughly inhibiting the undercounting effect.