Variability in Complex Constructs: Inferring Risk Preference and Temporal Discounting

Abstract

We examine the prevalence and extent of variability across measurements of supposedly stable behavioral economic traits. We begin by reviewing how these traits are conceptualized in behavioral economics, and how different instruments for eliciting them lead to variability in their measurements. We then consider factors such as experiment structure, affect, and context, known to influence or correlate in some way with the inferred values of these constructs: from domain or “subject-level” influences to local influences. We introduce the idea that an important – and cognitively meaningful – source of potential variation in experimentally-inferred measures may come from temporal sequence or the influence of trial order. Finally, we discuss how some of these sources of variation may not be ultimately all be brought under experimental or analytical control, and propose that they should instead be exposed and considered for their predictive value in different settings.