Internal Seminar: Filip Fidanoski

Title: "Can a single model account for both risky choices and inter-temporal choices? Testing the assumptions underlying models of risky intertemporal choice: A conceptual replication"

  • Date: 07 December 2022 from 13:00 to 14:00

  • Event location: Seminar Room - Piazza Scaravilli, 2 + Microsoft Teams Meeting

Abstract

Luckman et al. (2018) experimentally tested the conjecture that a single model of risky intertemporal choice can account for both risky and intertemporal choices, and under the conditions of their experiment, found evidence supporting it. Given the existing literature, that is a remarkable result which warrants (conceptual) replication. A key reason to be sceptical about the result is that Luckman et al. (2018), following a well-established tradition in psychology, employed first-year psychology students that participated for course credit as subjects. Proper incentivization is a long-standing bone of contention among experimentally working economists and psychologists and it is widely accepted among economists (and even many psychologists) that the elicitation of risk preferences and time preferences is very much a function of the way you incentivize the choices. Another important reason to be sceptical about the result (which interacts with the first) is that the experiment was not properly powered up; hence the no-difference results reported by the authors might be spurious. We find indeed a significant difference between the risky and intertemporal choices at the group level and individual level. Interestingly, we find that there are no significant differences in the choices made by subjects recruited through the ORSEE pool and those from the Prolific pool. We find further that order effects matter for intertemporal choices, but not for risky choices, and that there is no significant difference between choices made by subjects paid a flat incentive vis-à-vis choices made by subjects paid under the random incentive scheme. At the individual level, we find evidence in favour of the model that assumes common value function, but separate choice function. This result is robust across the subject pools, order of presentation, and incentive systems, but it is sensitive on different prior distributions. Other semi-nested models are sensitive to the different orders of presentation as well as incentive systems.