Sunk cost in investment decisions

We experimentally investigate the effect of sunk cost in a two-stage investment continuation task. After an initial investment, participants have to decide whether or not to continue the project with an additional investment. We do not find a standard sunk cost bias, but observe a robust reverse sunk cost effect: the larger the initial investment, the lower the likelihood to continue investing. This holds despite the fact that we replicate the standard sunk cost bias in hypothetical scenarios. We argue that both, risk aversion together with narrow choice bracketing and loss aversion can account for the reverse sunk cost effect.