Multi-Jurisdiction Replication of VPRAI Racial Detention-Recommendation Gaps

What the evidence says

In the federal court system, Black defendants are 8.7 percentage points more likely than white defendants to receive a custodial sentence, and the average custodial sentence for Black defendants is almost double (~2x) that of white defendants. This gap does not trace back to earlier stages of case processing: Black and white individuals show similar progression from arrest through to sentencing in federal courts, and there is no evidence of sample selection bias affecting the federal sentencing-gap estimate. Because neither differential funnel progression nor selection into the sentenced sample explains the disparity, the source attributes federal sentencing disparities to decisions made at the sentencing stage itself, describing federal judges as "largely responsible" for the gap.

The picture differs in state court systems. In New Orleans state courts, the source finds sample selection bias: the unobserved factors that drive whether a case progresses through each stage are positively correlated with the severity of the eventual sentence. Failing to account for this understates the true Black-white sentencing gap — an uncorrected OLS estimate underestimates the gap by 15%.

In state courts across the 75 largest counties, the pattern by racial/ethnic group diverges. Black arrestees, as in the federal system, show no differential stadial progression, so a selection model adds no gain over a simple comparison for the Black-white gap. Hispanic arrestees, by contrast, show significant differential progression through case stages. That differential progression combines with significant sample selection specifically for prison sentencing to produce a 13% underestimate in the OLS estimate of the Hispanic-white prison sentencing gap; selection bias is not present for jail sentencing in this same setting.

Taken together across the three jurisdictions examined, sample selection bias is not a uniform feature of court processing: it is absent in federal courts, present in New Orleans state courts, and present only for prison (not jail) sentencing in the 75-largest-counties state court sample. The sources here do not address any risk-assessment or detention-recommendation instrument (such as VPRAI-type tools) or whether racial gaps in pretrial detention recommendations replicate across jurisdictions — the evidence base above concerns post-arrest sentencing disparities in court processing, not multi-jurisdiction replication of an actuarial detention-recommendation tool's racial gaps. That specific question is unaddressed by the current sources.

Where the evidence splits

No contradictions in the current sources.

Thresholds and dose-response

No threshold established in the current sources.

Open empirical questions

• Whether the federal, New Orleans, and 75-largest-counties sentencing-gap patterns generalize to court systems outside those three settings.
• What accounts for the divergence in differential stadial progression between Black arrestees (none measured) and Hispanic arrestees (significant) in the 75-largest-counties sample — no mechanism is established in the sources.
• Whether the jail/prison split in selection-bias presence extends to other sentencing outcomes (for example, probation or diversion) not covered here.
• Whether any actuarial pretrial risk-assessment or detention-recommendation instrument shows racial gaps that replicate across jurisdictions — unaddressed by the sources on this page.