Measured sentencing gaps do not reveal which group gets the unfair treatment

Evidence

One important point that needs to be highlighted is that these estimated effects reveal unfairness in the treatment of two (groups of) defendants who have otherwise similar cases, but they do not reveal whether one group is treated harshly or the other group is treated favorably. For example, the result that Black defendants receive longer sentences than Whites may imply that Whites were assigned sentences of “optimal” length but Blacks received sentences that were longer than optimal. Alternatively, it could imply that Blacks were assigned sentences that were correct, but that Whites were treated leniently and received shorter than optimal-length sentences. — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law, 752

Categorization

• DEEPTHINKING: A methodologically foundational ambiguity: disparity statistics cannot localize which side deviates from the ideal baseline, a point that destabilizes naive readings of bias data.
• XREF: Directly parallels the reader’s conviction-vs-measured-bias theme — an observed outcome gap between groups does not by itself tell you where the bias sits or whether professed neutrality hides it.
• corroborated: 2 independent passages support this claim — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law: A Definitive Look at…, 752