Sequential Deliberation

We present a dynamic model of deliberation in which `jurors' decide every period whether to continue deliberation, which generates costly information, or stop and take a binding vote yielding a decision. For homogeneous juries, the model is a reinterpretation of the classic Wald (1947) sequential testing of statistical hypotheses. In heterogeneous juries, the resources spent on deliberation depend on the jury's preference profile. We show that voting rules at the decision stage are inconsequential when either information collection is very cheap or deliberation agendas are strict enough. Furthermore, wider preference distributions, more stringent deliberation agendas, or more unanimous decision voting rules, lead to greater deliberation times and more accurate decisions.