Sentencing Guidelines and Judicial Discretion: Quasi-experimental Evidence from Human Calculation Errors

There is a debate about whether advisory non-binding sentencing guidelines affect the sentences outcomes of individuals convicted in jurisdictions with this sentencing framework. Identifying the impact of sentencing guidelines is a difficult empirical problem because court actors may have preferences for sentencing severity that are correlated with the preferences that are outlined in the guidelines. But, in Maryland, ten percent of the recommended sentences computed in the guideline worksheets contain calculation errors. We use this unique source of quasi-experimental variation to quantify the extent to which sentencing guidelines influence policy outcomes. Among drug offenses, we find that the direct impact of the guidelines is roughly ½ the size of the overall correlation between recommendations and outcomes. For violent offenses, we find the same ½ discount for sentence recommendations that are higher than they should have been, but more responsiveness to recommendations that are too low. We find no evidence that the guidelines themselves directly affect discretion for property offenders, perhaps because judges generally have substantial experience with property cases and therefore do not rely on the errant information. Sentences are more sensitive to both accurate and inaccurate recommendations for crimes that occur less frequently and have more complicated sentencing. This suggests that when the court has more experience, the recommendations have less influence. More tentative findings suggest that, further down the decision chain, parole boards counteract the remaining influence of the guidelines.