Pedro Le Mattre de Carvalho’s research while affiliated with ISPA - Instituto Universitário and other places

According to the normative model of intertemporal choice, the Discounted Utility Model (Samuelson, 1937), the utility of a future outcome is discounted exponentially as a function of its delay. There is ample evidence that people do not comply with this proposition. In this study, 280 subjects performed a matching task in order to verify a new anomaly to exponential discounting, as well as four well known anomalies in the literature: The delay effect, the magnitude effect, the sign effect, and the delay-speedup asymmetry. The new anomaly is an interaction between the framing, sign, and delay to the outcomes: In gains, the delay effect is more pronounced in the delay frame than in the speedup frame; in losses, the reverse occurs. All anomalies were verified. We advance an explanation of the interaction effect in terms of diminishing sensitivity to delays.

Intertemporal choice has a direct relation with phenomena that pertain to every-day life, such as saving, consumption, and investment behavior. This study advances our understanding of intertemporal choice, in that it considers triadic choice contexts (three options) and not only dyadic ones (two options), to which most empirical studies are limited. For the first time, this study tries to understand how the preference between options is influenced by other options in the choice context. Current models of intertemporal choice, the so-called discounting models, do not consider such influences, because each option is evaluated independently from the others. We focus on polarization effects, induced by the introduction of a third option to the choice set and the framing of the third option as the default option. The results confirmed these effects. We discuss how discounting models must be replaced by other models, in which people make direct comparisons between the options. The results also confirmed phenomena that until today have been accommodated by discounting models: The delay, magnitude, and sign effects.