Methodology

Coverage

Fiscal years 2018 to 2025, 524,421 cases. Coverage begins in FY2018 because the Commission's current offence-type classification (OFFGUIDE), range-status categories (SENTRNGE) and capped sentence-length variables (SENTTCAP) all begin with the FY2018 file; earlier files use different definitions (OFFTYPSB and BOOKERCD) and are not mixed in. Every case in a year's file with a valid district code is counted; rows without one are reported on the year's page.

Entities

• District: the Commission's CIRCDIST code, 94 districts in Sourcebook order. Names are written in full; the citation form (E.D. Pa.) follows the Bluebook.
• Circuit: MONCIRC, 0 for the District of Columbia Circuit and 1 to 11. The district-to-circuit mapping is checked against MONCIRC on every row when a file is processed.
• Offence type: OFFGUIDE, the Commission's 30 primary offence types, assigned "mainly from the primary guideline the court applied at sentencing and then the count of conviction with the highest statutory maximum" (codebook).
• Guideline: GDLINEHI, the Chapter Two guideline of the computation that produced the offence level the court used. Pages exist for guidelines with at least 200 cases over the period.
• District × offence: pages exist where the district had at least 30 cases of the offence type in the last five fiscal years (2021–2025); pages with fewer than 50 such cases are built but marked not to be indexed.

Sentence length

Prison months are the Commission's SENTTCAP: total prison time imposed, including time served and §5G1.3 credit, excluding months of alternative confinement, with sentences of 470 months or longer and life recorded as 470 months. Cases with no prison term (probation, alternatives only, or a fine only) have no SENTTCAP value and are excluded from medians, means and the histogram; their share is shown separately as "cases with a prison term". Time-served sentences of unknown length are recorded by the Commission as one day (0.03 months) and round to zero here. Months are rounded to whole numbers before aggregation.

The median is the middle value (the mean of the two middle values when the count is even). Quartiles are the 25th and 75th percentiles. The mean counts life as 470 months, as the Commission does. Histogram bins: under 6 mo, 6–11 mo, 1–2 yr, 2–3 yr, 3–5 yr, 5–10 yr, 10–15 yr, 15–20 yr, 20–30 yr, 30 yr+, life.

Position relative to the guideline range

SENTRNGE, the Commission's post-Booker categories, grouped as the Sourcebook groups them: within range (0); above range (upward departure, 1; above-range variance, 6); below range, government sponsored (§5K1.1 substantial assistance, 2; §5K3.1 early disposition, 3; other government-sponsored departure, 4; government-sponsored variance, 7); below range, other (downward departure, 5; below-range variance, 8). Shares are of cases with a determinable status; the missing count is shown but never enters a percentage. Guideline pages show the nine categories separately.

Other variables

Criminal history category is XCRHISSR (I–VI, as determined by the court); final offence level is XFOLSOR; type of sentence is SENTIMP. All are used only as counts or medians over groups.

Suppression

Any group with fewer than 10 cases is hidden: counts print as "<10", and no median, share or histogram bar is drawn from it. In offence-mix tables such groups are pooled into one row. Medians require at least 10 cases with a prison term. The rule applies to every table cell, histogram bin and comparison figure. Because the site holds only group counts and month histograms, nothing on it can be traced to a case.

Comparisons and normalisation

Each entity is compared with its parent (district with circuit and nation; district × offence with the district, the circuit's cases of that offence type and the nation's), and caseloads are normalised as cases per fiscal year and as a share of the national total. Comparison words in the summaries ("longer than", "close to") use a 15% ratio band for months and a 5-point band for percentages; they are descriptions of aggregates, not judgements.

What differs from the Sourcebook

The Commission's Sourcebook reports mean and median sentence length for cases with a prison term using the same capped variable, so totals should match its Table 15 and district tables closely. Small differences arise from rounding months to whole numbers before taking the median and from the Sourcebook's occasional exclusion of cases with missing information in a given table.

Pipeline

Each year's CSV (about 1.5 GB) is streamed once and folded into a compact aggregate: counts by district, offence type, criminal history category and range status, each with a histogram of whole prison months. Those snapshots are committed with the site; builds read them and check the Commission's site once for a fiscal year that has no snapshot yet. Sentences on entity pages are assembled from fixed templates with the figures inserted; nothing is generated by a language model.