Testing the Speed of Adjustment of the Reference Point in Models of Expectation-Based Reference-Dependent Preferences

We build on Koszegi and Rabin (2009) and propose a model in which the reference point consists of two periods lagged beliefs about consumption outcomes. As opposed to the original model in which the reference point instantaneously adjusts to new information, our formulation renders the reference point somewhat sticky and allows sensations of gain and losses to persist for more than one period. Using a controlled experiment, we assess the relative importance of t-1 lagged beliefs with respect to t-2 lagged beliefs. Our results suggest that the speed of adjustment of the reference point might depend on the size of the stakes at play. Under low stakes, subjects remain unaffected by our manipulation. We interpret this result as evidence of a very fast adjustment of the reference point. A follow-up study in which the stakes are multiplied threefold documents a context-dependent house money effect, a phenomenon which can be rationalized by multiple-lagged beliefs reference-dependent preferences. Overall, our results help understanding the conditions under which past beliefs and emotions should be expected to linger and to affect subsequent behavior.