Disparate-impact labor statistician
Expert class library · Employment · Plaintiff side
Uses statistical analysis to show that an employment practice produced significant disparities in hiring, pay, or promotion by protected group. Builds regression models that control for legitimate factors.
What this expert is retained to answer
- Are there statistically significant disparities in pay, promotion, or selection by protected group?
- Do the disparities persist after controlling for legitimate factors such as job, tenure, and experience?
- Can data be pooled across divisions or decision-makers, or must it be analyzed separately?
- Which employment practice, if any, is associated with the disparity?
Methods
- Multiple regression of pay or promotion
- Standard deviation analysis
- Applicant flow analysis
- Availability benchmarking
How the testimony is attacked
- Improper pooling across units. Defendants argue data should be analyzed by division or decision-maker, and cite a failed Chow test. Chen-Oster held that failing a Chow test is not an absolute bar to pooling.
- Tainted or omitted variables. Controlling for variables that may themselves reflect bias, such as performance ratings, can understate a disparity, while omitting legitimate factors can overstate it. Chen-Oster and McReynolds both addressed these specification fights.
- Multiple comparisons. Testing many subgroups raises the chance of a spurious significant result. In Karlo the trial court excluded the analysis for lacking a Bonferroni correction, and the Third Circuit held that this was not an automatic ground for exclusion.
- Errors in underlying data. Coding errors and unreliable inputs are a common line of attack. McReynolds treated alleged errors as a battle of the experts where both sides used similar methods.
- Lack of fit to the legal claim. Statistics must match a claim the law recognizes. Karlo rejected a fit objection once it held that subgroup age claims are cognizable.
What the public record shows
A deliberately narrow CourtListener search, "disparate impact" AND "regression analysis" AND (Daubert OR "Rule 702"), returned 6 opinions filed since 2015, as of October 2, 2026; broader searches return more. Three that show how courts handle this class of testimony:
| Outcome | Case | Court | Why |
|---|---|---|---|
| Admitted | Chen-Oster v. Goldman, Sachs & Co.114 F. Supp. 3d 110 | S.D.N.Y. 2015 | The court denied the motion to exclude the plaintiffs' labor economist, holding that a failed Chow test does not automatically forbid pooling across divisions and that leaving out arguably tainted variables can be appropriate; in the same ruling it excluded the defense statistician's matched-pair comparisons because they offered fact-level comparisons better presented through fact witnesses than with an expert's imprimatur. |
| Admitted | McReynolds v. Sodexho Marriott Services, Inc.349 F. Supp. 2d 30 | D.D.C. 2004 | The court denied the employer's motion to exclude the plaintiffs' statistical expert on promotion disparities, treating claimed errors and omitted factors as a battle between competent experts using similar methods. |
| Exclusion reversed | Karlo v. Pittsburgh Glass Works, LLC849 F.3d 61 | 3d Cir. 2017 | Reversing the trial court, the Third Circuit vacated the exclusion of the plaintiffs' statistician, holding that the missing Bonferroni correction and the other stated grounds did not justify exclusion, and remanded for further Daubert proceedings. |
Each case links to the free opinion text on CourtListener.
Under amended Rule 702
Since December 1, 2023, Rule 702 says expressly that the party offering an expert must show the court it is more likely than not that the testimony meets each requirement: that it rests on sufficient facts or data, uses reliable methods, and reflects a reliable application of those methods to the case. Questions about the basis of an opinion are no longer automatically matters of weight for the jury. For how the circuits have applied the amendment, see the Rule 702 tracker, which follows each court of appeals; for what that means for preparing or attacking this class of expert, see the guide on amended Rule 702.
Related classes
Guides
Simulating this class on a matter
Supreme Mind simulates a class of expert, never a named individual, on your fact pattern and returns the likely opinion, the ranked cross-examination weaknesses, the Rule 702 attack surface and what it means for settlement. Read how it works, or book a demo.
Last reviewed October 2, 2026. How this page is built: rulings are found by searching court opinions on CourtListener, and each one is read in the opinion before it is summarised here; outcomes are labelled by what the court did with the expert's testimony. No individual expert is named. This page summarises public decisions for orientation and is not legal advice; read the opinion before relying on any ruling.