Prospective gain-loss utility: Ordered versus separated comparison

Koszegi and Rabin (2006, 2007) develop a model of expectations-based reference-dependent preferences in which the agent experiences “gain-loss utility” by comparing his actual consumption outcome to all of his previously expected consumption outcomes. Koszegi and Rabin (2009) then generalize the static model to a dynamic setting by assuming that the agent experiences both contemporaneous gain-loss utility over present consumption and “prospective” gain-loss utility by comparing his updated to his previous expectations about future consumption. Moreover, Koszegi and Rabin (2009) generalize the outcome-wise “static comparison” of gain-loss utility to a percentile-wise “ordered comparison,” in which the agent compares expected consumption outcomes at each percentile. This paper generalizes the static comparison slightly differently, to what I call a “separated comparison.” In the separated comparison, the agent compares each expected consumption outcome but experiences gain-loss utility only over uncertainty that has been realized, by effectively separating it from remaining future uncertainty. This paper shows that the separated comparison yields simple, tractable, and well-behaved equilibria in a broad class of economic models. Moreover, the equilibria are easy to solve for because the separated comparison preserves an outcome-wise linear structure.