The Stability of Behavioral (and Other) Measures

Title: The Stability of Behavioral (and Other) Measures

Abstract: We repeatedly administered an incentivized survey to a representative sample of the U.S. population over the course of a year. This survey contained incentivized choice tasks, qualitative assessments of economic preferences, and political and policy questions. The distribution of choices is very stable over time, although individuals’ choices may vary considerably. Most incentivized choice tasks exhibit moderate levels of stability—less than questions associated with identity, but more than survey-based measures ofbehavior. Principal components of incentivized choices align with those documented in our work on “Econographics” and are more stable than the individual choices they are drawn from. Qualitative assessments of economic preferences tend to be more stable than corresponding incentivized choice tasks, although answers to these two types of elicitations are largely uncorrelated. Repeated exposure to choice tasks improves the stability of choices. At an individual level, the single best predictor of stable choices is an in-survey IQ test. Simpler choice tasks can improve the stability of measured choices. Taken together our results suggest that preferences are relatively stable, and that economists can considerably improve their measurement.