Projection Bias in Effort Choices

Abstract

Our tastes fluctuate, often rapidly: we grow tired and thirsty from running, and we savor food or crave coffee more the longer we go without it. Our perceptions of our tastes are biased toward our current tastes: we misperceive future tastes as being closer to our current tastes than they will be – a tendency labeled projection bias by (Loewenstein, O’Donoghue, and Rabin 2003). In this project, I estimate projection bias in real effort choices on Amazon Mechanical Turk (MTurk), using a design with three features: (i) I measure directly the size of projection bias (ii) without requiring functional form assumptions on the utility (iii) while ruling out learning as an explanation. In addition, I estimate whether projection bias decreases with repetitions; whether people project onto others and onto their past selves; and whether people expect others to behave in projection-biased ways.