Judge-level disparity

Seventh Circuit · 2020–2025 · 1,375 modeled decisions

Decision relief

11.1%

Judge-model universe · 153 of 1,375

Routine3.4%
Doctrine19.8%
Counts & definition

Any reversal, vacatur, remand, or comparable appellate relief; the broader outcome used by the thesis judge analysis.

All decisions: 153 relief / 1,375 decisions.

Routine: 25 relief / 728 decisions.

Doctrine: 128 relief / 647 decisions.

Sentence relief

12.2%

Sentencing-at-issue decisions · 91 of 748

Routine3.6%
Doctrine21.4%
Counts & definition

Relief tied specifically to the sentence or further sentencing proceedings.

All decisions: 91 relief / 748 decisions; 653 unknown and excluded.

Routine: 14 relief / 388 decisions; 361 unknown and excluded.

Doctrine: 77 relief / 360 decisions; 292 unknown and excluded.

Judge comparisons

Decision relief · 1,375 decisions

Raw Routine, Overall, and Doctrine rates with one modeled Overall posterior mean. Rows are ordered by modeled Overall; the horizontal connector shows the model adjustment, and reference lines are raw circuit averages.

RawRoutineOverallDoctrineModeledOverallModeled higherModeled lower
0%10%20%30%40%Circuit Routine: 3.43% of unique decisionsRoutine 3.4%Circuit Overall: 11.13% of unique decisionsOverall 11.1%Circuit Doctrine: 19.78% of unique decisionsDoctrine 19.8%Flaum: raw Routine 12.5%, raw Overall 25.0%, raw Doctrine 28.6%; modeled Overall posterior mean 20.0%Flaum25.0% → 20.0%Manion: raw Routine 4.5%, raw Overall 21.1%, raw Doctrine 28.6%; modeled Overall posterior mean 16.6%Manion21.1% → 16.6%Ripple: raw Routine 5.9%, raw Overall 13.9%, raw Doctrine 15.3%; modeled Overall posterior mean 16.1%Ripple13.9% → 16.1%Barrett: raw Routine 7.1%, raw Overall 16.2%, raw Doctrine 21.7%; modeled Overall posterior mean 14.2%Barrett16.2% → 14.2%Wood: raw Routine 4.0%, raw Overall 14.4%, raw Doctrine 28.8%; modeled Overall posterior mean 13.1%Wood14.4% → 13.1%Jackson-Akiwumi: raw Routine 3.6%, raw Overall 11.7%, raw Doctrine 21.1%; modeled Overall posterior mean 12.0%Jackson-Akiwumi11.7% → 12.0%Rovner: raw Routine 3.2%, raw Overall 11.8%, raw Doctrine 19.8%; modeled Overall posterior mean 12.0%Rovner11.8% → 12.0%Kanne: raw Routine 6.7%, raw Overall 13.6%, raw Doctrine 20.8%; modeled Overall posterior mean 11.9%Kanne13.6% → 11.9%St. Eve: raw Routine 3.8%, raw Overall 11.6%, raw Doctrine 20.2%; modeled Overall posterior mean 11.9%St. Eve11.6% → 11.9%Hamilton: raw Routine 2.2%, raw Overall 11.3%, raw Doctrine 19.2%; modeled Overall posterior mean 11.8%Hamilton11.3% → 11.8%Lee: raw Routine 2.8%, raw Overall 12.1%, raw Doctrine 26.9%; modeled Overall posterior mean 11.6%Lee12.1% → 11.6%Bauer: raw Routine 3.8%, raw Overall 15.9%, raw Doctrine 33.3%; modeled Overall posterior mean 11.5%Bauer15.9% → 11.5%Brennan: raw Routine 2.0%, raw Overall 9.6%, raw Doctrine 17.7%; modeled Overall posterior mean 10.9%Brennan9.6% → 10.9%Easterbrook: raw Routine 3.7%, raw Overall 10.4%, raw Doctrine 18.7%; modeled Overall posterior mean 10.7%Easterbrook10.4% → 10.7%Sykes: raw Routine 3.4%, raw Overall 9.9%, raw Doctrine 18.4%; modeled Overall posterior mean 10.4%Sykes9.9% → 10.4%Pryor: raw Routine 3.4%, raw Overall 9.8%, raw Doctrine 22.7%; modeled Overall posterior mean 9.7%Pryor9.8% → 9.7%Maldonado: raw Routine 2.9%, raw Overall 5.5%, raw Doctrine 9.5%; modeled Overall posterior mean 9.6%Maldonado5.5% → 9.6%Scudder: raw Routine 2.8%, raw Overall 7.7%, raw Doctrine 14.1%; modeled Overall posterior mean 9.2%Scudder7.7% → 9.2%Kolar: raw Routine 1.7%, raw Overall 5.5%, raw Doctrine 12.1%; modeled Overall posterior mean 9.1%Kolar5.5% → 9.1%Kirsch: raw Routine 3.6%, raw Overall 7.7%, raw Doctrine 13.6%; modeled Overall posterior mean 9.0%Kirsch7.7% → 9.0%

How the comparison is made

  1. One row per decisionSeventh Circuit, 2020–2025
  2. Case characteristicsPosture, offense & publication track
  3. Actual panelUnique members share equal weight
  4. Joint modelRoutine baseline & Doctrine difference
  5. Common docketsFocal judge with average circuit peers
Analysis details

This refit counts each disposition once in the likelihood: 1,375 decisions and 4,152 unique panel memberships. The decision-relief chart uses this same universe. Sentence relief is shown for its separate 748-decision sentencing subset; Research detail models that outcome separately.

Each decision’s panel contribution is the average of its actual unique members—one-third per judge on a normal three-judge panel. Publication track is explicit; posture and offense are the other case-level controls. The model standardizes both publication tracks internally, then displays only their recombined Overall posterior mean for a focal judge with two average circuit colleagues.

Raw overall is the observed rate for decisions associated with that judge. Modeled overall recombines the two modeled track rates using that judge’s observed Routine and Doctrine counts. Judges are ordered by modeled Overall relief.

The likelihood counts each disposition once and averages the actual participating panel. The comparison is observational and does not identify an individual vote or causal effect.

Relief requires a material change in the party’s legal position. Clerical corrections, duplicate orders, corrected captions, and non-substantive judgment modifications are not decision relief or sentence relief.