Abstract: In criminal courts, prosecutors have considerable discretion over defendant’s sentencing, suggesting skilled prosecutors may be able to reduce both incarceration and future crime. Leveraging the quasi-random assignment of low-level felonies in North Carolina Superior Court, we find that prosecutors vary in their effects on both incarceration and re-offense. Since differences across prosecutors in their re-offense effects cannot be fully explained by their incarceration effects, prosecutors vary in their “skill” — the degree to which they selectively incarcerate those defendants most likely to re-offend. Indeed, prosecutors who are one standard deviation above the mean achieve a 2pp (8%) lower rate of re-offense than one would expect given their incarceration effect.