Egypt's Criminal Judgements: Penalty Types by Offense
We checked which penalty type shows up in the ruling text across seven offense categories in Egypt's criminal corpus, and found that a reliable average sentence length isn't something we could responsibly extract yet.
- 1Armed robbery cases mention aggravated imprisonment more often than any other offense category checked (39.7%, 1,877 of 4,725), ahead of drug crimes (31.3%) and murder (28.2%).
- 2Murder cases mention the death penalty far more than any other category checked (9.6%, 578 of 6,036). Theft shows a surprisingly close rate (7.5%), which we traced to topic overlap rather than a real signal, see section 3.
- 3A standalone fine is rarely mentioned as the penalty in any offense category, never exceeding 7% anywhere we checked. Criminal appellate rulings in this corpus center on custodial and capital penalties, not monetary ones.
- 4We deliberately did not calculate an average sentence length in years. Section 1 explains why: two extraction traps (ancillary measures worded like prison terms, and appellate text reciting a lower court's sentence rather than stating a final one) made a naive numeric average unreliable enough that we'd rather publish nothing than publish a wrong number.
1. Why we didn't calculate average sentence length
Egyptian court rulings write sentence duration as Arabic number words inside free text ("بالسجن لمدة خمس عشرة سنة", imprisoned for fifteen years), not as a structured field, so getting a number out means parsing that text. It's technically possible, but two problems showed up immediately when we sampled it. First, phrasing for ancillary measures is nearly identical to phrasing for prison terms: one ruling read "أمرت بوضعهم تحت مراقبة الشرطة لمدة خمس سنين" (ordered them placed under police surveillance for five years), which a naive pattern match would miscount as a five-year sentence. Second, almost all of these records are Court of Cassation appellate rulings, meaning the text often recites the lower court's original sentence as procedural history while the appeal itself concerns something else entirely (admissibility, procedure, evidence), not a clean restatement of a final term. Publishing an average built on top of both of those failure modes would have looked precise without being accurate, so we stopped short of it.
Whether a penalty type (death, life hard labor, aggravated imprisonment, lesser detention, fine) is mentioned anywhere in the ruling text for a given offense category. That's a co-occurrence check, not a verdict extraction: it tells you how often a penalty type comes up in cases tagged with an offense, not what any single defendant was finally sentenced to.
2. Penalty type by offense category
Percentages are share of that offense category's records where the penalty term appears anywhere in the ruling text. Categories are drawn from free-text topic tags and aren't mutually exclusive (a single record can carry multiple topics), and a record can mention more than one penalty type, so rows don't sum to 100%.
| Offense | Records | Death | Life Hard Labor | Aggravated Imprisonment | Detention | Fine |
|---|---|---|---|---|---|---|
| Murder | 6,036 | 9.6% | 2.6% | 28.2% | 4.5% | 0.2% |
| Drug crimes | 12,204 | 1.2% | 1.6% | 31.3% | 3.5% | 0.4% |
| Theft | 3,166 | 7.5% | 1.9% | 26.1% | 6.9% | 0.2% |
| Armed robbery | 4,725 | 2.5% | 0.8% | 39.7% | 5.3% | 0.3% |
| Assault | 4,496 | 0.6% | 0.3% | 15.7% | 7.6% | 0.1% |
| Firearms | 523 | 1.1% | 0.2% | 25.8% | 7.1% | 1.0% |
| Fraud | 207 | 0.5% | 0.0% | 3.9% | 2.9% | 0.5% |
3. The theft/death-penalty overlap
Theft showing a 7.5% death-penalty mention rate looks off at first glance since simple theft isn't a capital offense under Egyptian law. Sampling those records directly explains it: several are armed robbery cases (السرقة بالإكراه) that also carry a general سرقة topic tag, and Egyptian law does allow capital punishment for robbery aggravated by resulting death. The rate isn't noise, but it's a reminder that these offense categories overlap rather than partition the corpus cleanly, and a "Theft" row here includes some records that are really armed robbery cases by another name.
4. How to read this data
Treat these percentages as a directional signal of penalty severity by offense type within this corpus, not as sentencing statistics you'd cite in isolation. They're useful for a retrieval system deciding how much weight to give an "aggravated imprisonment" query against "armed robbery" versus "fraud" records, for instance. They're not a substitute for a proper sentencing study, which would need trial-court original judgments with structured sentence fields (this corpus is almost entirely appellate review), or a purpose-built Arabic legal-NLP pipeline validated against a manually labeled sample.
5. Limitations
Offense categories come from free-text topic tags, not a fixed taxonomy, and overlap with each other (see section 3). Penalty-type counts are text co-occurrence via indexed phrase search, not verdict extraction, so they capture what's discussed or reviewed in a ruling, not necessarily the final confirmed sentence. OCR noise affects the underlying text at a rate we haven't separately measured for this subset. This report covers Egypt only.