02 Discerning-Discretion-Estimating-Prosecutor-Effects 2020
The paper investigates how individual prosecutors in North Carolina Superior Court differ in their impact on two core criminal‑justice outcomes: incarceration length and three‑year recidivism. Using a novel empirical design that compares prosecutors only within narrowly defined sentencing contexts (same crime type, geographic office, and five‑year period), the authors residualize case characteristics, correct for sampling error via split‑sample methods, and estimate each prosecutor’s causal effect on incarceration days, recidivism rates, and an internally‑consistent “horizontal equity” metric. They find substantial heterogeneity: a one‑standard‑deviation increase in sentencing severity adds about 12 days (≈15 % longer) to sentences, while a similar increase in recidivism raises three‑year re‑offense by 3.6 percentage points (≈6 %). Only 24 % of the variation in recidivism is explained by incarceration effects, implying that most prosecutors differ in how selectively they incarcerate high‑risk defendants – a component the authors label “skill.” Skill variation is comparable in magnitude to unconditional recidivism variation, and it correlates positively with the propensity to incarcerate defendants for more than six months but negatively with horizontal equity. The paper also documents balanced baseline characteristics across prosecutors, details extensive data restrictions, and outlines future extensions (racial equity, young‑defendant incarceration, tenure effects, and alternative frontier specifications).
Topics
Prosecutorial Discretion as a Multi‑Dimensional Policy Lever
Sets up the central research question: whether prosecutors systematically differ in incarceration severity, recidivism outcomes, and horizontal equity, and how these dimensions may trade off.
Within‑Context Empirical Design for Causal Prosecutor Effects
Describes the core identification strategy: defining sentencing contexts (crime × office × time), residualizing case characteristics, and estimating prosecutor‑specific averages relative to context means.
Data Construction, Restrictions, and Balance Checks
Details the North Carolina Superior Court dataset (1995‑2009), case exclusions (probation violations, severe offenses, drug trafficking, etc.), minimum case‑cell thresholds, and presents balance statistics showing near‑random assignment of cases to prosecutors within contexts.