Does early release decrease or increase the probability that ex-convicts will return to prison? We exploit unique data from Israeli courts, where appearance before the judge throughout the day has an arbitrary component. We first show that judges more often deny parole requests of prisoners appearing further from the judges' last break. We then use this variation in instrumental variable estimations and find that early releases reduce the propensity of prisoners to return to prison. Robustness checks suggest that later and earlier cases are largely comparable and that potential selection is unlikely to explain the results.