摘要：Two ways of administering the standard Juster scale were used to predict the rate at which consumers would buy frequently purchased items. Respondents were asked to estimate their households' purchases of seven frequently purchased items over a four week period. Half of the respondents were asked for separate Juster probabilities for each level of purchase; half were asked to allocate tokens over the levels of purchase in proportion to their Juster probabilities. Respondents were reinterviewed four weeks later to determine actual purchases of each item. The method using tokens was superior in all measures of predictive accuracy. For all items, this method slightly overestimated purchases while the other produced considerable underestimates
