Stop comparing the totals and start comparing the actual people. The answer flips.
So far this series has compared the databases' totals. Now let's compare the actual people. We can, because Mapping Police Violence does the matching for us: on each of its cases, it notes whether the Washington Post had that person, and whether Fatal Encounters had them too. So you can line the three lists up and see exactly who overlaps, with no guesswork.[src]
You might expect a mess: rival projects, totally different people. It's the opposite. In every year the three lists overlap, more than 98 percent of the people on MPV's list are also on the Post's or Fatal Encounters'. The databases mostly know about the same deaths. Whatever their totals do at the edges, they are looking at nearly the same set of names.
Take out the cases the Post shares and you are left with about a hundred a year that only the wider lists caught. Sort those by how the person died and the gap makes sense at a glance: roughly fifty a year were killed some way other than a gunshot, exactly the deaths the Post's shooting-only rule was never going to catch. The famous "disagreement" between the databases is mostly those fifty people, one definition apart.
Now the hard question: how many people did none of the lists catch? There is a classic way to estimate that. To count fish in a pond you can't drain, you net some, tag them, and throw them back. Later you net again. If a fifth of the second batch is already tagged, then your first net probably caught about a fifth of the pond, so you can work backward to the total, including the fish no net ever touched.
Run that math on the Post and Fatal Encounters for 2019 and it spits out 1,092, give or take almost nothing. That "give or take almost nothing" is the giveaway that something is wrong.
To find the people nobody counted, you need a net that has nothing to do with the others. There is one: the government's own death certificates. Every death in America gets recorded, with a cause. A team of researchers writing in The Lancet compared those official records against the civilian databases and found the certificates had quietly missed 55.5 percent of deaths from police violence, 17,100 of an estimated 30,800 over nearly forty years, filed under some other cause.[src]
Put it together and the series lands. The same study estimated about 1,190 real police-violence deaths for 2019. The civilian trackers recorded around 1,100. They were close. The counts that came in far too low were the government's: the FBI's tally and the official death records. The undercount everyone worried about after Ferguson is real, and it is the government's. A newspaper, a nonprofit, and a volunteer turned out to be the fix.
One more thing the matching earns us. Once you know the lists are seeing nearly the same deaths, their year-by-year totals stop being rival claims and become one shared record. Here it is, as far back as year-by-year counting exists, with each project's own rule left intact.
| Year | MPV all killings | Post shootings only | Guardian all killings |
|---|---|---|---|
| 2013 | 1,075 | · | · |
| 2014 | 1,039 | · | · |
| 2015 | 1,099 | 995 | 1,146 |
| 2016 | 1,065 | 959 | 1,093 |
| 2017 | 1,091 | 984 | · |
| 2018 | 1,138 | 992 | · |
| 2019 | 1,099 | 993 | · |
| 2020 | 1,161 | 1,021 | · |
| 2021 | 1,148 | 1,050 | · |
| 2022 | 1,204 | 1,097 | · |
| 2023 | 1,248 | 1,164 | · |
| 2024 | 1,270 | 1,175 | · |
The table is also an answer to the question "how far back can we go?" A clean national count per year begins in 2013, when Mapping Police Violence's records start. Fatal Encounters reaches deeper, more than 28,000 records back to January 2000, but as a running roster rather than published annual totals. Before that there is no count to consult at all, only the death-certificate reconstruction from the last section, which the Lancet team modeled back to 1980. The record is short because, until about a decade ago, nobody was keeping one. That is itself the finding this series keeps arriving at.
Read down the columns and the direction is unmistakable. The Post's count held in a narrow band just under a thousand for its first five years, 995 in 2015 to 993 in 2019, then rose every single year after: 1,021, then 1,050, 1,097, 1,164, and finally 1,175 in 2024, the highest it ever recorded. MPV, with the wider rule, tells the same story, from 1,075 in 2013 to 1,270 in 2024, its last three years each a new record. Two projects, two definitions, one direction: up roughly 18 percent across the decade.
Start with what the rise is not. It is not the counting getting better. These are the same instruments applying the same rules across the whole span, and this essay just showed they agree on the actual people to within two percent. When two rulers with different markings both read longer, the thing that grew is the thing being measured.
The clearest explanation in the data is where the killings moved. MPV's own year-end analysis finds killings declining in the biggest cities, many of which rewrote their use-of-force rules over the past decade, while rising in suburban and rural areas, which are policed by thousands of smaller agencies where less changed.[src] The national total is just the sum of those two curves, and so far the rise outside the big cities has outweighed the decline inside them.
The longer arc says this is not new. The Lancet team's forty-year reconstruction puts the death rate from police violence at 0.25 per 100,000 in the 1980s and 0.34 per 100,000 in the 2010s, an increase of about 38 percent across four decades.[src] The drift of the 2020s continues a long climb; it is not a sudden break.