This paper is now published in the Journal of Economics and Statistics.

Abstract

A growing literature has shown how people sometimes prefer to randomize between two options. We study how prevalent this behavior is in an experiment using a novel and simple method. We allow subjects to randomize between options in a series of questions in which one of the alternatives is fixed and the other varies, capturing the range of values for which subjects want to randomize. We find that most subjects choose to randomize in most questions. Crucially, they do so for ranges of values are ‘very large’: for example, when comparing a fixed amount $x with a lottery that pays $20 or $0 with equal chances, subjects typically randomize for all xs between $5.3 and $12. Large ranges are found in other questions as well, showing how prevalent the desire to randomization is. We connect ranges to standard choices, Certainty-Bias, and non-Monotonicity.