How do researchers calculate police killing rates per 100,000 by race?
Executive summary
Researchers calculate police-killing rates per 100,000 by taking counts of people killed by police (from databases such as The Washington Post or Mapping Police Violence) and dividing those counts by the population of the relevant racial group, then scaling to 100,000 (or per million) to produce comparable rates across races; for example, The Washington Post’s fatal-shootings series reports rates like 6.2 per million for Black Americans vs. 2.4 per million for white Americans (reported via Statista) [1]. Different studies adjust that basic approach by restricting to age-sex subgroups, using life‑table methods to estimate lifetime risk, or controlling for exposure to police or crime rates—methods described in PNAS and Harvard analyses [2] [3].
1. What the raw calculation is — simple per‑capita rates
The straightforward rate researchers report is: (number of people of a given race killed by police in a time period) ÷ (population of that race in that time period) × scaling factor (100,000 or 1,000,000). Public trackers such as The Washington Post and Mapping Police Violence provide the numerator (counts by race) and census or other demographic sources provide the denominator; Statista reproduces Post figures as rates per million (e.g., 6.2 per million for Black Americans, 2.4 per million for white Americans) [1] [4].
2. Why researchers sometimes use age‑ and sex‑specific rates
Fatal encounters cluster by age and sex (risk peaks between ages ~20–35), so analysts compute age‑ and sex‑specific rates or lifetime risk to avoid misleading comparisons that conflate differing age structures across racial groups. The PNAS study used life‑table methods to convert observed yearly risks into a synthetic lifetime risk for cohorts of 100,000 and presented age‑specific curves by race and sex [2].
3. Adjustments for exposure and behavior: competing interpretations
Some scholars argue raw per‑capita rates obscure “exposure” differences — how often groups encounter police — and differences in offending rates or contextual factors. Fryer and others explicitly note one must account for differential exposure to police and race‑specific crime participation rates when interpreting racial disparities in police use of force [5] [3]. That leads to two competing interpretations in the literature: one treats per‑capita disparities as evidence of systemic differences in policing; the other emphasizes selection effects and varying exposure that could explain part of the gap [3] [5].
4. Data sources and reliability problems: numerator issues
Researchers rely on media and NGO databases because official reporting is incomplete. The Washington Post’s database logs on‑duty police shootings since 2015 and is widely used, but independent projects (Mapping Police Violence, The Guardian’s The Counted) have different inclusion rules; totals differ year to year and race sometimes is unknown or unrecorded, creating uncertainty in the numerator [4] [6] [7]. WorldPopulationReview and others explicitly note “unknown ethnicity” entries can rise and distort per‑race totals [8].
5. Denominator choices matter: census vintage and subgroup definitions
The population denominator can be total population, sex‑and‑age subgroups, or cohort adjustments. Mapping Police Violence uses 2020 Decennial Census race/ethnicity population data for its rate calculations, while other analyses may use annual estimates; choice of census year and race definitions affects the resulting per‑capita rate [7] [9].
6. Scaling and presentation: per 100,000 vs. per million vs. lifetime risk
Different publications present rates at different scales. Statista reproduces Washington Post rates per million; some policy reports prefer per 100,000 for familiarity; academic studies like the PNAS paper convert annual rates into lifetime risk for a synthetic cohort of 100,000 to communicate cumulative risk across a life course [1] [2]. Readers should compare like with like.
7. What the major studies conclude and where they disagree
Population‑adjusted rates consistently show non‑white groups die at higher per‑capita rates than whites in many datasets, but the magnitude and causal interpretation vary. The Post and Mapping Police Violence show higher per‑capita rates for Black and Hispanic people in many years [4] [9]. Fryer and other economists caution that after accounting for exposure and other controls the racial gap can shrink or change, while public health style life‑table work emphasizes lifetime disparities that remain substantial for men of color [3] [2].
8. Limitations and what reporting rarely mentions
Available sources document numerator and denominator choices, age/sex adjustments, and exposure debates, but they do not uniformly settle causal explanations; some reports note police departments resist sharing data, increasing uncertainty [8] [4]. Explicit estimates of how much of the disparity is due to exposure vs. bias are not consistently reported across these sources; available sources do not mention a single agreed percentage attribution between causes.
9. Practical checklist for readers and analysts
When you see a “per‑100,000 by race” number, check: which database supplied the deaths (Post, Mapping Police Violence, state reports); what year and census population were used for the denominator; whether rates are age/sex‑standardized or lifetime risk; and whether authors adjusted for police exposure or crime rates—differences in any of these choices explain much of the variation across studies [4] [7] [2].
Sources cited above: Washington Post database and reproduced Statista rates [1] [4], PNAS lifetime‑risk methods [2], Fryer/Harvard notes on exposure adjustments [3] [5], Mapping Police Violence and population denominator practices [7] [9], and commentary on data gaps [8].