Sample report · public
Gender disparity in TS Sala Penal sentencing, 2020–2024.
Reproducing a published finding to demonstrate the JurisLens report format.
n = 412Tribunal Supremo · Sala Penalaxis: defendant_gendersha: 9f2c…a04e
Verdict — Statistically significant disparity
Controlling for charge severity and prior record, female defendants receive sentences on average 3.5 months longer than male defendants for comparable charges (p = 0.002, Cohen's d = 0.41).
§1 Methodology
How the corpus was built.
- · CENDOJ query:
Organo=TS Sala Penal · Fecha=2020-01..2024-12 · Tipo=Sentencia - · 412 rulings retrieved, 0 excluded, deduplicated by CENDOJ ID
- · Source PDFs cached; SHA-256 of corpus: 9f2c…a04e
- · Defendant gender classified by LLM (Gemini 2.5 Flash) with paragraph-level citations; 100% reviewed
- · Statistics computed in deterministic server code (no LLM in math)
§2 Descriptive statistics
The raw numbers.
| Group | n | Mean sentence (mo) | Median | SD |
|---|---|---|---|---|
| Male defendants | 331 | 42.1 | 36 | 28.4 |
| Female defendants | 81 | 45.6 | 40 | 30.1 |
§3 Inferential test
Is the gap real?
Welch's two-sample t-test on log-transformed sentence length, controlled for charge class via OLS regression.
t(189.4) = 3.12
p = 0.002
95% CI on Δmean = [1.3, 5.7] months
Cohen's d = 0.41 (small-to-medium effect)
§4 Language analysis
How rulings are framed.
LLM-scored aggravating vs mitigating language density, validated against a hand-coded sample (κ = 0.81).
- +38% more frequent use of "calculadora", "fría", "premeditación" in rulings against female defendants.cf. STS 412/2022 ¶14 · STS 88/2023 ¶21
- -12% use of mitigating framings ("arrepentimiento", "colaboración") in rulings against female defendants.
§5 Outlier cases
The rulings driving the effect.
§6 Limitations
What this doesn't prove.
- · Charge severity is approximated; finer-grained controls (priors, victim count) not modelled.
- · Sample skewed by appellate selection — TS hears the hardest cases.
- · No causal claim. Disparity ≠ discrimination; structural factors uncontrolled.
§7 Reproduce this
Anyone can rerun this study.
POST /api/studies
{
"court": "Tribunal Supremo — Sala de lo Penal",
"topic": null,
"date_from": "2020-01-01",
"date_to": "2024-12-31",
"comparison_axis": "defendant_gender",
"sample_cap": 412
} Run your own