Criminal sentences

763 pandemic-fraud sentences, and how long they ran

From a working ledger of 1,133 pandemic relief-fraud records, 763 carry an actual custodial sentence that can be placed on a chart. Their median term is 37 months; the mean is longer, 53 months, pulled up by 77 terms of ten years or more.

2026-09-17

The cleaning rule (1,133 → 763)

Keep only rows with an actual imposed custodial term (has_numeric_sentence).
Keep only analysis-grade rows (high_confidence_for_analysis); drop medium- and low-confidence extractions.
Keep only a real, plausible term of 0 < months ≤ 1200 (up to 100 years); this drops zero/unknown terms and 9999 placeholder artifacts.
Keep only rows assignable to a fraud family, excluding the two program-sponsor / gatekeeper rows that are not comparable to individual-defendant sentences.

The result is 763 plottable observations.

Sentence length, by fraud family and by band

Two views of the same 763 sentences. By family, the bar shows how many cleaned sentences fall in each fraud category and its median term. By band, it shows how the terms spread from under a year to twenty-plus.

What a sentence here is. Each observation is a custodial term imposed in a pandemic relief-fraud matter, read from the sentencing record. The families are our classifications of what the fraud targeted. Every term on the charts is shown in months; the years figure in the top stats is the median converted (37 ÷ 12).

In the archive

Sources. All figures recompute from the criminal-sentences dataset (1,133 rows) under the cleaning rule stated above, which yields 763 plottable observations (median 37 months, mean 53.1). The published extract is pandemic-fraud-sentences.csv, with its schema on the release page. Sentences are custodial terms read from the sentencing record. Build date 2026-09-17.