Biases in forecasting future product utility: Two essays

This dissertation examines consumers' ability to accurately anticipate the hedonic value they will derive from consuming a product in the future. Two common types of such forecasting problems are studied: consumer valuation of new attributes of a next-generation product at the time of purchase, and predicting preferences for the whole product on the basis of product fragments. In the new attribute valuation problem, consumers need to decide whether it is worthwhile to acquire the new attribute. Optimal decisions require consumers to compute the expected value of acquiring the new attribute based on their beliefs about their future utilization. Drawing on the research on affective forecasting, we predict that consumers will fail to fully incorporate their future utilization into a priori valuation and purchase decisions. This forecasting failure results in an overvaluation bias, whereby people exhibit excessive willingness-to-pay for the new attribute relative to its ex post actual utilization and benefit. In the second type of forecasting problems, consumers do not have the opportunity to observe the final product (e.g., the appearance of a room after a new wallpaper is applied) at the time of purchase and thus need to extrapolate their preferences for the whole through visualization. A process model is proposed which predicts qualitatively different decision strategies depending on features of the judgment context, which differentially affect the accuracy of hedonic forecasts. Hypotheses about behavior in both contexts are tested in five laboratory experiments. The data provide consistent support for most of the predicted biases. For example, in the first essay we report evidence supporting the idea that consumers may overvalue new product attributes relative to subsequent utilization, and that this bias is not mitigated by allowing decision makers a brief trial ownership. Likewise, in the second essay we report support for the idea that the process of mental extrapolation can be sensitive to such cues as the configurality of whole objects which are shown as examples, and that forecasts of liking based on fragments tend to be more extreme than post-hoc judgments of completed images. The implications of these results for research in consumer decision making are reported.