A Theory of Judicial Deference

In many instances, appellate courts defer to lower courts and administrative agencies; the appellate court allows the other agent’s decision to stand even though the appellate court has strong reason to believe that decision incorrect. We provide a model in which such deference is rational. Our model is set in a two-dimensional case space. One dimension reflects "global" facts that are known to everyone; the other dimension reflects "local" facts that are known only to the trial court or agency. The appellate court has a preferred legal rule or partition of the case space into two sets of cases: those decided “1" or “plaintiff prevails” and those decided “0" or "defendant prevails." The initial decision maker may, with some probability, be biased in the sense that it thinks a different partition is best. A one-period model identifies the basic tradeoff the appellate court faces between granting discretion to a potentially biased trial court or agency and ruling as a matter of law. Deference risks that a biased agent will decide the case wrongly from the appellate court’s perspective. Ruling as a matter of law on the basis of global facts only ignores local facts and also risks error. The degree of deference granted balances these two risks. Intuitively, the model shows that cases where the global facts are inconclusive are the ones where the grant of discretion is most valuable. We then extend the model to two periods and investigate the relationship between trial court or agency reputation and the standard of deference. Counter-intuitively, we find that close or hard cases in period 1 are the ones that reveal the most about the underlying preferences of the trial court or agency. As a result, these cases are the most valuable to