The item count technique, used often to investigate illegal or socially undesirable behaviours, requires respondents to indicate merely the number of applicable items from among a list. However, the number of applicable items indicated via the item count question tends to be smaller than when it is calculated from the direct `applies/does not apply' responses to each item. Because this inconsistency, which we refer to as the underreporting effect, often disturbs proper item count estimates, the causes of this effect are explored in this paper. Web survey results revealed that the order of the response alternatives is irrelevant to the underreporting effect, and that the underreporting effect is caused by the response format in which the item count question requests merely the number of applicable items and not the number of non-applicable items. It is also shown that the magnitude of the underreporting effect decreases when the respondents are asked to indicate the numbers of both applicable and non-applicable items, which we refer to as elaborate item count questioning.