Hungry Professors? Decision Biases Are Less Widespread than Previously Thought

TL;DR

This study challenges prior claims that decision biases, such as effects of scheduling and mental depletion, are widespread by analyzing a large natural experiment in law school oral exams, finding no evidence of such biases.

Contribution

The paper provides empirical evidence that decision biases related to scheduling are less prevalent than previously thought, using a large-scale natural experiment to test these effects.

Findings

No significant scheduling bias in exam grades

Previous biases may be overstated due to study limitations

Caution needed when generalizing earlier findings

Abstract

In many situations people make sequences of similar, but unrelated decisions. Such decision sequences are prevalent in many important contexts including judicial judgments, loan approvals, college admissions, and athletic competitions. A growing literature claims that decisions in such sequences may be severely biased because decision outcomes seem to be systematically affected by the scheduling. In particular, it has been argued that mental depletion leads to harsher decisions before food breaks and that the ``law of small numbers'' induces decisions to be negatively auto-correlated (i.e. favorable decisions are followed by unfavorable ones and vice versa). These findings have attracted much academic and media attention and it has been suspected that they may only represent the ``tip of the iceberg''. However, voices of caution point out that existing studies may suffer from serious limitations, because the decision order is not randomly determined, other influencing factors are hard to exclude, or direct evidence for the underlying mechanisms is not available. We exploit a large-scale natural experiment in a context in which the previous literature would predict the presence of scheduling biases. Specifically, we investigate whether the grades of randomly scheduled oral exams in Law School depend on the position of the exam in the sequence. Our rich data enables us to filter-out student, professor, day, and course-specific features. Our results contradict the previous findings and suggest that caution is advised when generalizing from previous studies for policy advice.