How to Read Federal Sentencing Data
A practical guide to the statistics on PlainSentencing and what they mean.
The Data Source: USSC Annual Datafiles
PlainSentencing aggregates data from the United States Sentencing Commission's public datafiles, which contain case-level information on every federal offender sentenced during each fiscal year (October 1 – September 30). The FY2015–FY2024 dataset covers over 660,000 federal sentencing records.
Each record includes variables for: district, offense type (guideline), sentence length, guideline range min/max, whether the sentence was within/above/below the range, plea type, and demographic variables.
Key Statistics Defined
Case Count
The number of offenders sentenced in a given district, offense type, or fiscal year. Case counts reflect prosecution and charging patterns as much as crime rates — districts with more federal prosecutors and enforcement resources process more cases.
Average Sentence (Months)
The mean sentence length in months across all cases in a category. This includes prison sentences and probationary sentences (coded as 0 months in the USSC data). When a district's average seems low, it may be because many defendants received probation rather than prison.
Within Guidelines (%)
The percentage of sentences that fell within the guideline range calculated for that defendant. A high within-guidelines rate suggests judges in that district closely follow the Guidelines; a lower rate reflects more frequent use of departure and variance authority.
Above Guidelines (%)
Sentences above the guideline range — typically resulting from government-initiated upward departures or judicial findings of aggravating factors not adequately captured in the standard calculation.
Below Guidelines (%)
Sentences below the guideline range. Most below-guidelines sentences result from government cooperation motions (§5K1.1) or other government-sponsored departures. Downward variances under § 3553(a) are also included.
Booker Variance
In USSC data, "Booker variance" specifically refers to cases coded as outside the guidelines range where the court cited § 3553(a) factors rather than a specific guideline departure provision. These are included in the total departure rate but tracked separately.
Guilty Plea Rate (%)
The percentage of defendants who pled guilty rather than went to trial. Federal guilty plea rates are consistently high (90%+) nationally because plea agreements typically result in substantially lower sentences and because the government's conviction rate at trial is very high.
Prison Rate (%)
The percentage of defendants sentenced to imprisonment (as opposed to probation or a split sentence). Prison rates vary by offense — most drug trafficking and violent crime convictions result in prison; some fraud and regulatory offenses more often result in probation.
Disparity Score
PlainSentencing calculates a disparity score for each district and fiscal year: the average percentage by which sentences in that district differ from the national average for the same offense types. A score of +20% means sentences are, on average, 20% longer than national norms for comparable offenses.
Important limitations:
- The score controls for offense type but not for the full range of case-specific factors (drug quantity, weapon involvement, cooperation, etc.).
- Districts with few cases in a given offense type may show volatile scores due to small samples.
- The score is a relative measure — it tells you where a district sits in the distribution, not whether any particular sentence was appropriate.
Reading Trend Data
Year-to-year changes in sentencing statistics reflect:
- Guideline amendments — the USSC regularly updates offense levels, which changes the baseline.
- Prosecution priorities — shifts in DOJ enforcement priorities change the case mix.
- Judicial turnover — new judges in a district bring different sentencing philosophies.
- Case-specific factors — a single high-profile case can significantly move a district's average in a small district.
What This Data Cannot Tell You
PlainSentencing data is descriptive and aggregate. It cannot:
- Predict how a specific case will be sentenced.
- Tell you whether any particular sentence was correct or appropriate.
- Explain the specific facts behind aggregate patterns.
- Substitute for legal advice about an actual case.
For case-specific guidance, consult a qualified federal defense attorney.
Worked example: building a year-over-year comparison
A common analyst task is comparing district-level sentencing trends across years. For the Southern District of New York (SDNY), FY2021 saw 1,138 sentences with a median of 38 months; FY2022 saw 1,094 sentences with a median of 42 months; FY2023 saw 1,167 sentences with a median of 40 months. Raw, this looks like SDNY sentences ticked up then receded. Decompose by offense category and the picture shifts: drug-trafficking median rose from 60 to 78 months over the three years (a 30% increase), while immigration median fell from 24 to 19 months (a 21% decrease). The overall median is stable only because the offense mix shifted toward more drug cases. A district-level claim about "tougher sentencing" requires offense-controlled comparisons or it collapses into measurement noise.
Variable reference for common analyses
| Variable | Type | Common analysis |
|---|---|---|
| SENTTOT | Continuous (months) | Median sentence length |
| OFFTYPE2 | Categorical | Offense-mix decomposition |
| DEPART | Categorical | Variance/departure rates |
| XFOLSOR | Continuous (1-43) | Final offense level |
| CRHISCAT | Ordinal (I-VI) | Criminal history control |
| DISTRICT | Categorical | Geographic comparisons |
| SEX, NEWRACE | Categorical | Demographic stratification |
Where the data is silent
The USSC datafile captures the outcome of federal sentencing but not the inputs to plea negotiation. Charge decisions, declination patterns, and informal cooperation arrangements all happen upstream and produce visible effects only through their impact on offense type and 5K1.1 frequency. Similarly, the datafile records final sentence but not appellate disposition — a district median of 40 months may include some sentences later reduced on appeal to 30 months, but those are not reflected. For appellate outcomes, the Administrative Office of US Courts (AOUSC) and the Bureau of Justice Statistics (BJS) provide separate datasets. Cross-walking USSC sentencing data with AOUSC docket data is technically possible via case-number matching but is rarely done because the AOUSC dataset is harder to access than the USSC public-use files.