Behavioral sunk cost. Subsequent choices with prior investment

Abstract

In this paper we address the sunk cost effect with behavioural investment. From an incentivized task in a survey with a built-in game of two parts. Part one is a game with a 2X2 design: High sunk cost condition, High and Low treatment group. And Relevant sunk cost condition, Relevant and Irrelevant treatment group. The game is where subjects put in time and effort, for which the attributes of the game differed depending on treatment group. Part two is a decision task equal for all treatment groups, where subjects choose between a risk-free alternative and a risky alternative involving a stake. We find no evidence from testing hypothesis 1, being more likely to accept a risky bet when experiencing high sunk costs. We do find evidence from testing hypothesis 2, being more likely to accept a risky bet when experiencing Relevant sunk costs. This evidence is consistent and show robust results at the 10% significance level, in both bivariate and multivariate analyses. Using a contingency table for the bivariate analysis and a logit model for the multivariate analysis.