Picking Pennies Bias: Exposing Oneself To Tail Risk
University of New South Wales; Financial Research Network (FIRN)
June 10, 2016
This study reveals through lab experiments a hitherto-undocumented form of gambling: Participants in a decision-making task exposed themselves to extreme levels of tail risk to pick-up additional pennies. The data offer suggestive evidence that the root cause of this behavior is greed — participants could not resist the temptation to seek more dollars despite knowing the risks entailed. The findings of follow-up experiments provide a hurdle to other explanations and a means of validating the greed hypothesis. These findings are surprising both due to the pervasiveness of the picking pennies behavior in the lab and its robustness.
Picking Pennies Bias: Exposing Oneself To Tail Risk – Introduction
In this paper I report on a novel behavioral bias which consists of exposing oneself to mindless levels of tail risk to win additional dollars—henceforth, “the picking pennies bias.” I build on recent results obtained in a lab experiment designed to study learning under tail risk (Payzan-LeNestour, 2016). In that experiment, task participants were to choose to bet in Normally-distributed reward prospects, which by design had positive expected value, while avoiding negatively-skewed ones, which yielded steady streams of good outcomes but eventually inflicted a major loss that annihilated all
previous gains. The task comprised several sessions of 20 trials each. In some of the sessions the prospect was of the Normal kind, in others it was negatively-skewed. At the beginning of each session the participants did not know which kind of prospect they were facing, however they could learn about it through observing the value generated by the underlying prospect in each of the 20 trials of the session. In the process of studying participants’ learning in the experiment, I discovered a novel behavioral bias: a group of participants bet after observing an extreme value whose realization made it clear that the underlying prospect was of the bad kind.
In this study I begin by analyzing in depth this behavior as it emerged in the data used in Payzan-LeNestour (2016). I show that in view of both those data and the existing neurobiological evidence from prior work, the most plausible explanation for the picking pennies bias is that task participants bet despite knowing the risks entailed, because they could not resist the temptation to seek more dollars through betting—“the greed hypothesis.”
While the analysis offers suggestive evidence supporting the greed hypothesis, it can only partially rule out the possibility that the picking pennies bias reflects instead learning failures (erroneous beliefs about the prospect kind) or risk-loving preferences in the subjects, as the original experiment was not specifically designed to flesh out the root cause of the bias. The current study sets out to do this. I
ran two follow-up experiments. The first allows me to exclude learning failures and risk-loving preferences as potential explanations, by neatly disentangling beliefs and preferences. The second constitutes a litmus test of the greed hypothesis in that it directly tests one of its key implications, namely, the idea that those subjects who picked pennies in the task could not help betting, i.e., their behavior was at variance with deliberation.
The first experiment replicates the conditions of the original one except that just before each session began, the subjects were told the nature of the prospect in the session, whereby there was no learning involved in that experiment. I find that the picking pennies bias prevails to the exact same extent as in the original experiment. It cannot, therefore, be attributed to learning failures. This finding also fully rules out the risk-loving explanation (I elaborate in Section 3.1).
The second experiment slightly augments the original task in that the subjects in each trial were also given the option to either bet or not bet (“skip”) in all of the remaining trials of the session. I find that the subjects made extensive use of the option to skip for all trials. Thus, when betting was suboptimal during the task, subjects effectively bound themselves to not betting. Such “self-binding” is reminiscent of Ulysses’ reliance on an externally imposed constraint to resist the call of the sirens (Elster, 1979). I further find that in that treatment, the prevalence of the picking pennies bias is significantly reduced. This result compellingly suggests that task participants were drawn to the picking pennies bias because of greed coupled with limited self-control. As such, it considerably strengthens the evidence for the greed hypothesis.
Finally, I find that the picking pennies bias cannot be ascribed to limited liability in the lab—the fact that task participants are not exposed to actual potential losses. Limited liability is a general concern in all experimental designs involving potential losses, as the rules imposed by the ethics committees are very strict in terms of not harming participants, so inflicting actual losses to participants is usually not authorized within our standard experimental procedures. I devised a new payment procedure (explained in detail in Section 4.3) that allowed me to expose subjects to real losses while complying with the ethics rules. In a third and final follow-up experiment that used that procedure, the picking pennies bias prevailed to the same extent as in the previous experimental sessions. It thus appears to be surprisingly robust.
See full PDF below.