The appellate review system is intended to serve as an efficient remedy for imperfect judicial decision making. However, it can fulfil this task only when appeals are ex ante unpredictable to the judge, and thus can be expected to occur primarily in case of a bad verdict. Using data from case records of a German trial court, we show that the probability of appeal can be predicted based on easily observable exogenous factors. Controlling for the complexity of a legal case, we find that judges also tend to decrease their effort when the ex ante probability of appeal is low. Thus, our empirical evidence indicates an inefficiency in the appellate review system because trial judges allocate their effort to cases not exclusively according to case complexity, but particularly according to the ex ante probability of being reviewed.