How do trends since 2013 (or the last decade) differ across databases in rates of police killings by race and location?

1. Different definitions, different totals — why databases don’t match

Databases diverge first on definitions: The Washington Post restricts its feed to people fatally shot by on‑duty police beginning in 2015, while Mapping Police Violence counts any killing by police (including chokeholds, tasers, off‑duty officers and non‑gun deaths) starting in 2013; Fatal Encounters uses an even broader inclusion of all deaths during interactions with police with no requirement of culpability ^{[1] [2] [6]}. Those methodological choices explain why MPV typically records higher totals than the Post and why Fatal Encounters reaches further back in time and captures different event types ^{[2] [6]}.

2. Trend divergence since 2013 — rising, steady, or declining depending on the source

When analysts look at different windows and definitions they reach different trend narratives: MPV shows a large, multi‑year series from 2013 onward and reported more than 1,200 killings in 2024 alone, implying recent increases ^[3]. The Washington Post’s shooting‑only series since 2015 averages roughly 1,000 fatal shootings per year and reports totals consistent with a persistent, high baseline of police shootings ^[1]. Long‑run academic work (e.g., The Lancet review) also suggests official records undercount many deaths, so trends based on death‑certificate data can understate long‑term numbers ^[6].

3. Race and location: consistent disparities, contested explanations

Multiple projects and peer‑reviewed studies find Black Americans are disproportionately killed by police: MPV’s headline finding — Black people are about three times more likely to be killed than White people — appears in MPV materials and related analyses ^{[7] [8]}. Public‑health research links fatal shootings to neighborhood social‑vulnerability indexes and shows strong interactions of race, place, and socioeconomic status, implying geography and poverty explain some but not all racial gaps ^{[4] [9]}. Brookings and other scholars caution that controlling for socioeconomic context changes the size of race coefficients, arguing class‑and‑place factors can mediate measured racial disparities — a competing interpretation to MPV’s emphasis on systemic racism ^{[5] [4]}.

4. Geography: pockets of high deadliness and ZIP‑code concentration

Several analyses highlight that killings concentrate in particular places and departments. Risk‑adjusted studies show the deadliest local police departments kill many times more often than the least deadly even after accounting for contextual risk; Johns Hopkins and other researchers show zip‑code level disparities tied to social vulnerability and racial composition ^{[10] [4] [11]}. MPV’s state and department tools allow comparisons across states and localities and are frequently used to show where per‑capita rates are highest ^{[12] [2]}.

5. Data quality, missingness, and why policy debates hinge on methodology

Scholars warn that federal reporting has historically undercounted fatal police violence; The Post began its tracker after finding FBI data omitted many cases, and the Lancet found the National Vital Statistics System underreported police‑related deaths by over half for some periods ^{[1] [6]}. MPV and Fatal Encounters attempt to fill gaps through media, public records, and triangulation, but each has trade‑offs: broader scope increases coverage but also requires more adjudication of cause and race coding ^{[2] [13] [6]}.

6. What this means for interpretation and action

Comparing databases shows agreement on two points: many people die in police encounters annually, and racial and place‑based disparities exist ^{[1] [2] [4]}. They disagree on magnitude, which cases to include, and how much socioeconomic context explains disparities ^{[2] [5]}. For policymakers and researchers, the practical implication is that any reform or evaluation must state which dataset and definition it uses and consider cross‑referencing MPV, The Washington Post, and academic datasets to triangulate both counts and drivers ^{[1] [2] [4]}.

Limitations: available sources here do not provide a single harmonized time‑series across all databases, so precise numeric comparisons by race and place over 2013–2024 require assembling and standardizing the original datasets (not found in current reporting) ^{[2] [1]}.