CoChat

The DOI matches its Crossref record; 115 of the 116 references it cites were found in a registry or an index; 13 of the 33 sources its claims cite were checked for support.

Administrative Records Mask Racially Biased Policing

Open access
KNOX, DEAN, LOWE, WILL, MUMMOLO, JONATHAN · 2020 · American Political Science Review

Abstract

Researchers often lack the necessary data to credibly estimate racial discrimination in policing. In particular, police administrative records lack information on civilians police observe but do not investigate. In this article, we show that if police racially discriminate when choosing whom to investigate, analyses using administrative records to estimate racial discrimination in police behavior are statistically biased, and many quantities of interest are unidentified—even among investigated individuals—absent strong and untestable assumptions. Using principal stratification in a causal mediation framework, we derive the exact form of the statistical bias that results from traditional estimation. We develop a bias-correction procedure and nonparametric sharp bounds for race effects, replicate published findings, and show the traditional estimator can severely underestimate levels of racially biased policing or mask discrimination entirely. We conclude by outlining a general and feasible design for future studies that is robust to this inferential snare.

Overview

CoChat Summary

## TL;DR

The paper argues that analyses using police administrative records can systematically misstate racial bias because the data only include people police chose to stop, making the observed sample itself a product of potentially discriminatory selection. Using principal stratification and a causal mediation framework, the authors derive the bias, propose bounds and a bias-correction strategy, and show in a reanalysis of NYPD data that conventional estimates can dramatically understate racially discriminatory use of force.

## Key findings

  • If police discriminate when deciding whom to stop, then estimates based only on stopped individuals are statistically biased; the paper says the bias is always negative, so naïve estimates understate racially discriminatory police violence.
  • In their reanalysis of Fryer (2019) using NYPD Stop, Question and Frisk data from 2003–13, the paper says the naïve approach suggested only about 10,000 black and Hispanic civilians were handcuffed because of racial bias, while the bias-corrected estimate is approximately 56,000.
  • For uses of force against black and Hispanic civilians in those data, the paper says the naïve estimate implied 10% were discriminatory, while the bias-corrected estimate is 39%.
  • In one example, the naïve approach implied 2.4 instances per 1,000 encounters of pushing Hispanic suspects to a wall due to suspect race, while the revised estimate was at least 26 per 1,000 encounters.

## Methods

  • The paper uses a principal stratification / causal mediation framework to define quantities of interest, derive the exact form of bias from conventional estimation, and distinguish estimates among stopped encounters from the overall causal effect of race.
  • It proposes nonparametric sharp bounds for the average treatment effect among stops and the average treatment effect on the treated, and a bias-correction procedure based on an estimated share of racially discriminatory stops.
  • It reanalyzes and extends Fryer (2019) using the NYPD Stop, Question and Frisk database (2003–13), which the paper says contains roughly 5 million pedestrian stop records, and compares naïve versus bias-corrected estimates across force thresholds.

## Limitations

  • The paper explicitly says its bias-correction and bounding methods still rely on weaker assumptions that can be difficult to verify, and that some key quantities remain only partially identified without additional data.
  • It states that point identification of the ATE requires additional data on unreported encounters; police administrative records alone are “virtually uninformative” about the unobserved proportion of discriminatory stops.
  • Evident from the text: the reanalysis depends on assumptions about the proportion of discriminatory stops, including a conservative estimate from prior work, so corrected estimates still depend on an externally supplied sensitivity parameter.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1017/s0003055420000039

Cites

115 of 116 references verified0 citation errors · 1 not checkable

Statistics

Nothing to recompute.

Figures

Image 1
From the PDF
Image 2
From the PDF
Image 3
From the PDF
FIGURE 4 Bounds for Racially Discriminatory Use of Force, any Severity
FIGURE 4 Bounds for Racially Discriminatory Use of Force, any SeverityFrom the PDF
FIGURE 5 Estimated Number of Racially Discriminatory Uses of Force against Black and Hispanic Civilians, Divided by Total Observed Uses of Force among Those Groups Using Na¨ ı ve (Red Dot) and Bias-Corrected (Blue Triangle) Estimators of the ATTM 5 1
FIGURE 5 Estimated Number of Racially Discriminatory Uses of Force against Black and Hispanic Civilians, Divided by Total Observed Uses of Force among Those Groups Using Na¨ ı ve (Red Dot) and Bias-Corrected (Blue Triangle) Estimators of the ATTM 5 1From the PDF
Image 6
From the PDF

Verification

Claims the paper makes

11 Valid · 5 Not valid · 15 Not checked · 114 Own results and interpretations

60 more claims

Create a free account to see all 145 claims and to verify the 4 references still open.

See all claims — free account
  • It may state two separate claims as one. 33
  • Probably not a substantial claim. 54
  • The quote only partly states this claim. 2

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Self-check results appear here once an independent check has confirmed them.

Cited works that were retracted

Works this paper cites that were later retracted, from the retraction notices publishers deposit. Major when the notice came out before this paper was published.

1 of 1

Citations

Cites

1 retracted

Show all 116 references

Research works printed but not linked: 1

  1. Kocieniewski, 2002, Study Suggests Racial Gap in Speeding in New Jersey

Cited by

See all 314

Metadata

Authors
KNOX, D., LOWE, W., & MUMMOLO, J.
Affiliations
  • Princeton University (US)
  • Hertie School (DE)
Corresponding author
  • KNOX, DEAN
  • LOWE, WILL
  • MUMMOLO, JONATHAN
Venue
American Political Science Review
Published
2020-05-21
Volume
114
Issue
3
Pages
619-637
ISSN
  • 0003-0554 (ISSN-L)
  • 1537-5943
OpenAlex
W3122864926
Cited by
314
References
115
Citations by year
2026: 39 · 2025: 39 · 2024: 47 · 2023: 68 · 2022: 48 · 2021: 34 · 2020: 33
FWCI
161.39
Type
journal article
Language
en
Open access
Locations
  • American Political Science Review (publishedVersion)
  • SSRN Electronic Journal (acceptedVersion)
  • PhilPapers (PhilPapers Foundation) (submittedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • OPUS 4 (Zuse Institute Berlin) (submittedVersion)
Topics
  • Social Sciences › Political Science and International Relations › Policing Practices and Perceptions
  • Physical Sciences › Mathematics › Statistics and Probability › Advanced Causal Inference Techniques
  • Social Sciences › Sociology and Political Science › Criminal Justice and Corrections Analysis
Keywords
racial discrimination in policing, administrative records, statistical bias, principal stratification, causal mediation analysis, bias correction, sharp bounds, unobserved heterogeneity

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-09-29, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-09-29 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Administrative Records Mask Racially Biased Policing. https://app.cochat.ai/papers/10.1017/s0003055420000039
The paper
KNOX, D., LOWE, W., & MUMMOLO, J. (2020). Administrative Records Mask Racially Biased Policing. American Political Science Review. https://doi.org/10.1017/s0003055420000039

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.