How should we measure consumer confidence?

Research on consumer confidence has mainly sought to evaluate the power of available data to predict economic outcomes. In contrast, this article considers how best to measure consumer confidence. We analyze the responses to eight questions that have appeared recently on the Michigan Survey of Consumers; four elicit expectations in the traditional qualitative manner and four use a newer "percent chance" format. Examination of the responses suggests three implications. It makes more sense to ask for expectations of events directly relevant to individual economic decisions than for predictions of general business conditions. Surveys should shift away from qualitative questions in favor of ones eliciting subjective probability judgments. While aggregating responses into an index of consumer confidence may provide simple summary statistics, results should also be presented on a question-by-question basis for different subgroups of the population.