Confounding in Intertemporal Choice Designs: New Sampling Methods for Exponential, Arithmetic, and Hyperbolic Discounting

Abstract

We examine the confounding between models of intertemporal choice. Critical outputs from hyperbolic, exponential and arithmetic discounting are all highly multicolinear in commonly used research designs. This confounding means that if one model defines a participant as impulsive, they all will: not because the models are intrinsically related, but for lack of design in the stimuli to detect their differences. Also, confounding makes it difficult to detect which model(s) people actually use, and in what circumstances. We develop methods to overcome the problem, and provide extensive tables of over 8,000 choice questions generated according to these principles, so that the degree of separation between models can be controlled. Researchers may copy and paste the questions out of the .pdf into Word, say, for further processing. We give instructions how best to sample the database of questions and/or construct alternative sampling designs.