When Bail Is Not Posted

Among New York local-court misdemeanors with a recorded first cash-bail amount above $1, not posting bail at arraignment goes with a plea disposition rate about 7.6 percentage points higher within 30 days.

That figure compares cases with the same charge, court, year, criminal-history category and bail band, on the 4,778 not-posted and 977 posted cases with a match. It holds at +7.6, +7.7 and +7.6 under three sets of controls on that one fixed group.

1From a million rows to the comparison

The source is the Office of Court Administration's pretrial release extract. Each row is one defendant on one docket. Dockets from the same arrest can be linked, but one person cannot be followed across arrests.

Felonies are set aside because transfers to Supreme or County Court count as dispositions and cannot be followed. Cases after September 2025 are set aside as a conservative choice, because the extract's cutoff date is not documented.

2What the fields actually mean

• Timing. Dates are month and year only, so the elapsed-days field is used, never subtracted dates. Same-day dispositions are recorded as 1, and results are reported at both 30 and 31.
• The $1 code. The data dictionary says $1 is often set to credit time served when a person is remanded on another case. Of 43,803 misdemeanor bail cases, 24,913 carry exactly $1. That is why they are kept apart. The $1 amount is an imperfect marker of another hold, and an amount above $1 does not exclude one.
• Criminal history. A blank means no matching record, not a clean record, so there are three categories: positive, all zero, unknown.
• Plea date. Disposition is when all charges were resolved. The outcome is a plea disposition, not the day a plea was entered.
• Later posting. Among not-posted cases above $1, 13,050 of 15,876 show no posting observed. That is not the same as never posted.

3Plea disposition within 30 days, by status at arraignment

Rates exclude cases with unknown timing: 40 of the not-posted and 6 of the posted cases above $1. Assigning those to either extreme keeps the not-posted rate within 15.9% to 16.1% and the posted rate within 5.6% to 6.0%. Transfers within 30 days are 2.8% and 4.4%.

4The contrast inside matched cells

The unadjusted gap on the fixed group is 10.8 points. Two sets of 1,000 draws, with exponential weights on arrest cycles and posted weights frozen, give sensitivity ranges of 6.7 to 8.5 and 6.8 to 8.6. In 462 of 645 cells there is one posted case, so this understates uncertainty and is not a confidence interval.

5A judge check that does not support the design

The natural next step is to compare stricter and more lenient judges. The extract gives the judge but only the month, so judges can be compared only within a court and month, across 596,710 arraignments and 426 judges.

A judge one standard deviation stricter also hears a caseload 3.26 points more likely to be a class A misdemeanor, 1.39 points more likely to be assault and 1.07 points more likely to involve a Black defendant, and cases are 2.57 points more likely to continue past arraignment. These checks cannot separate judge assignment from other causes, so no causal judge result is reported.

§Limits

• People who post bail differ: 53% of posted cases have a positive criminal history, against 74% of not-posted cases.
• Bail amounts reflect a judge's view of the case, so they are not a dose of financial pressure.
• The extract has no measure of wealth, work or obligations such as rent.
• Court date and shift data would improve assignment controls, but would not establish a causal design on their own.

§Sources

• New York State Unified Court System, Pretrial Release Data, with documentation and data dictionary.
• Dobbie, Goldin and Yang (2018), on pretrial detention, conviction and guilty pleas.
• Stevenson (2018), on inability to pay bail.
• Gupta, Hansman and Frenchman (2016), on money bail and credit constraints.
• Leslie and Pope (2017), on New York City arraignments.