Understanding Expert Choices Using Decision Time

Laboratory experiments show a robust relationship between decision time and the perceived value of the selected alternative. We ask how these findings translate to decision-making in the field when delay arises from an endogenous decision to collect further information. In a stylized model of choice between two alternatives, we show that: (i) the less commonly-chosen alternative is more likely to be selected after a longer delay; (ii) decision time peaks when the probability of either choice is 50 percent; and (iii) the ultimate quality of the chosen alternative may increase or decrease with decision time, depending on whether earlier or later signals are more informative. We test these predictions in three settings. The first is an editorial setting where we observe proxies for paper quality and some of the signals available to editors. We document that (i) the probability of a positive decision rises with decision time; (ii) average decision time has an inverse-U shaped relation to the predicted probability of a positive decision, with a peak near fifty percent; and (iii) paper quality is positively (negatively) related to decision time for papers with Reject (R&R) decisions. We present corroborating evidence from two other settings: one involving the decisions of loan officers; the other involving young adults’ responses to a survey question on marriage expectations.