How the NYT Handles Mistakes: NYC Murder Edition
The problem isn't just egregiously wrong headlines: it's a basic indifference to responsibly presenting crime narratives.
Let’s work backwards from the big picture problem: the NYT will occasionally edit an article when they’ve gotten something wrong, but: (1) without always acknowledging the correction, (2) while often leaving other errors intact in a way that suggests they aren’t terribly interested in meaningfully revisiting or rethinking their coverage.
I reported on this last year1, when the NYT conflated a single NYPD crime metric (shootings) with violent crime in general, leading to an alarmist piece about “a spike in violence unlike anything [NYC] has seen in decades” when violent crime stats were actually down citywide and when it was hard to imagine a more sensitive subject that particular summer. While they ended up making changes, they did so via unmarked edits while telling me that a proper corrections notice “would not serve the reader”.
Well, they’re at it again, on more or less the same subject. While they get credit for a partial correction note this time around, they failed to demonstrate any curiosity about whether the piece still had other mistakes—which it very much did.
We reward corrections. See something wrong, misleading, or unfair? Say something. It helps!
The Headline
NYC teacher and criminal reform advocate Tom Robinson flagged a tweet to me on Friday night, from someone on the NYT’s Metro desk (also the lead reporter here):
Note that the original headline here was far more specific: “After Murders ‘Doubled Overnight,’ the N.Y.P.D. Is Solving Fewer Cases”.
Tom rightly thought this felt off, and looked up the official NYPD stats, which clarified that murders had not in fact doubled, either overnight or over any other recent period. They’re actually down 2% vs. 2020 (though up ~45% from pre-COVID).
But we can see how this particular mistake happened. Quoting from the article body:
“The increase in shootings, that’s got to have a negative impact on clearance rates,” said Peter Moskos, a professor at John Jay College of Criminal Justice. “Their caseload literally doubled overnight. The odds are never great. That’s the unfortunate truth.”
Ok, so an obvious parsing error here. Moskos was talking about caseloads (mostly driven by a doubling in non-fatal shootings), which the NYT’s headline writer misread as “murders”. And that’s the issue that the NYT went on to correct yesterday AM:
Correction: Nov. 27, 2021
An earlier version of this article’s headline incorrectly characterized a quotation by a criminologist on a sudden increase in gun violence in New York City that began in 2020. He said detectives’ caseloads “doubled overnight,” not murders.
But while I give them vague applause for doing one right thing here2, they clearly didn’t really re-review the piece to look for other errors.
Before we get to the rest though, I want to be emphatic that this mistake wasn’t purely accidental. Note the headline they used in the print version: “A Grieving Family Waits as Cases Go Unsolved”. That’s a bit more appropriate3 for what the piece was: a human interest story about a family’s wait for justice, with a spartan level of background context about why more cases than usual were going unsolved.
While it turns out that they’d also misinterpreted the percentage of unsolved cases (!), the digital headline would have been irresponsible even if they gotten that part right: city murders didn’t double, their increase started some 20 months ago, families caught in this particular situation are few in absolute numbers, and the rise in shootings/murders is only partially relevant to clearance rates today. This is why I consider it more a motivated mistake: it can’t be an honest mistake when your intent is to maximize traffic. If you reward headline editors for supplying, in effect, the most sensational headlines possible, their mistakes will err in a predictable direction, and it will be your fault as an institution, whether you admit it or not.4
Anyway, let’s move on to the rest of the article.
Beyond the Headline
Imagine you were a Metro reporter asked to tell this family’s story. You sit with them in their pain, and you walk away determined to call attention to their plight. But as you connect their particular case to the larger phenomenon (i.e., an apparent decrease in murder clearances), you owe it to your readers to anchor onto real-world data:
The actual change in clearances
What those clearance and murder rates look like in context, both nationally and relative to recent local trendlines
You can see where this going:
The clearance rates were both misstated and sensationally framed (the journalist wrote on Twitter that they “plunged”)
The only other context supplied in the article was a non-specific claim that “shootings and homicides” had “increased” from pre-pandemic times. (In the above tweet, they glossed this as “violent crime rates spiked”.)
Ok, so let’s ground ourselves on real numbers, starting with violent crime5. If we extrapolate 2021 YTD totals evenly to year’s end, we’ll end with 38,344 violent crimes in NYC. Now let’s put that figure next to equivalent totals from the previous six years, with 2021’s relative increase vs. each year in brackets:
2015: 38,991 (-1.7%)
2016: 38,120 (+0.6%)
2017: 35,749 (+7.3%)
2018: 35,210 (+8.9%)
2019: 36,144 (+6.1%)
2020: 35,573 (+7.8%)
Violent crime in NYC is so bad right now that it’s higher than any year since the sophomore year of the de Blasio administration, or about when Furious 7 came out!
But, ok, sure, it has risen a bit relative to the years in between. That said, 2017-2020 were unusually calm years6 in a remarkably unviolent era, which is why a difference of only ~350 violent crimes per million people moves the year-over-years percentages up by 7-8%. It’s not nothing. But it’s more a modest reversion than a “spike”, and only really notable in a world where crime is already historically and comparatively low. Even with the 2020-2021 increase in murders (about 45% vs 2017-2019 lows, compared to a national increase of 30%), your new chance of being murdered in NYC (5.6 per 100k) is roughly the same as in the notorious criminal hotbed of Colorado Springs, lower than in Omaha or Denver, and about 1/10th your odds in St. Louis or Baltimore.
Anyway, moving on to clearance rates. Our reporter says in both their tweet and the body of their article that the NYPD was clearing 90% of murders, either “pre-COVID” or “in the years before the pandemic”. Well, about that. Here are the NYC-wide murder clearance rates for 2019: 7
Q4 2019: 86.2%
Q3 2019: 52.9%
Q2 2019: 67.2%
Q1 2019: 74.7%*
A simple average here gives us a baseline of 70.3% (the NYT says elsewhere that the adjusted number was 67%), against a national benchmark of just 61%. This seems intuitively reasonable: NYC is a well-funded city with a low homicide rate. But if I were to ask what you’d expect this rate to drop to if: (1) both shootings and homicides increased, mostly among gangs, (2) new legislation de-anonymized informants8, (3) half of locals started wearing camera-blinding masks, (4) your police force experienced unprecedented levels of sick leave9, (5) the courts all got severely backed up with COVID adjustments, you’d probably guess a lot less than 60%! 10 It’s a pretty good result in context, and hardly worth any alarmism.
Unless you have newspapers to sell, I guess.
PS
It’s worth noting that there’s a quirk to the Unified Crime Reporting formula for the clearance numbers, where the real 2020 number was likely 2-3 points higher (dropping from 67% to something like 62% or 63%). Briefly, the system uses a cleared/committed ratio based on what new things happen each quarter, vs. matching backward to tell us how many homicides from any past quarter ever got cleared. So if the rate of murders increases while the time to solve them stays stable, your clearance rate will drop even if you clear them at the same actual rate—because of that weird lag.
To simplify, imagine 100 murders in year one, increasing by 50% to 150 in year two. Now imagine that every murder takes a year to solve. Even if you solve 100% of them, your clearance rate for year two will only be 66.7% (i.e., 100 clearances from year one that happen in year two, against 150 murders that happen in year two). This is why multi-year snapshots are important. While we can’t calculate exact relevance here without knowing how long the NYPD has been taking to clear murders, we’d expect for at least 2020 to have been artificially high by a few points.
PPS
As for those who might argue “ah yes but shootings did double from pre-COVID, and that’s something to be alarmed about”, I’d say two things: (1) we’re still only talking about +750 shootings a year in a city of 8.4 million people, with 50% of overall shootings apparently being gang-related, (2) this is happening to other cities with similar compositional makeups, and isn’t specific to NYC or the NYPD. We’re just in a really weird micro-era triggered by a once-a-century pandemic.
Is the increase still a problem worth discussing in measured terms? Of course! But the measured part matters! This is hardly Gotham. Shootings are mostly non-random, and more broadly-targeted violent crimes (rape, robbery, assault) in NYC and most other major American cities are still within single digits of all-time lows. While the existence of any violent crime anywhere is always tragic and deserving of our thoughtful response, we’re still doing remarkably well all things considered, and any necessary adjustments can be approached better without all the sensationalism.
Though also see higher-profile examples like how they handled the Twitter hack.
Discounted by the fact that a claim so bold should have been at least cursorily fact-checked with a 30-second google search. That these mistakes make it through editing in the first place signify a problem that late modest corrections can’t come close to solving.
Though still pretty bad, as the clearance rates are still wrong, and the story is still really muddled about the point it’s trying to make about said rates. Is it about the NYPD failing to adapt to the challenges posed by COVID? Is it about the costs of the new discovery law (see footnote 8)? If either/both, that should be clear. The NYPD has had 18 months since shootings escalated, and almost 30 since the law changed. Either they should be doing more, or what they’re doing is fine and it’s just an unusual set of conditions in which we have to grade on a curve (the national murder clearance rate dropped from 61% to 54%), in which case what’s the actual point of the article? The NYT seemingly couldn’t decide, and just went alarmist without actually helping the reader understand whether the NYPD are doing a good job or not, or what could/should be changed to improve clearances.
Also everyone who chimes in with “ah yes but journalists don’t write their own headlines and thus aren’t responsible” is part of the problem. Every writer is responsible for how their words are framed and used, and journalists are hardly powerless to correct serious headline mistakes. If they can’t get something like this fixed, they need to find a new employer. If they just don’t care enough to try, that’s indicative of their personal ethics.
There’s some confusion around what’s officially included under the “violent crimes” umbrella in any given jurisdiction. Official NYPD stats include 7 top-line “index” metrics based on the Uniform Crime Reporting standard, the top four of which are violent crimes (murder, rape, robbery, felony assault). But their reports also include some below-the-line numbers on other crimes, including shootings (separated by incidents and number of total victims). Unhelpfully, it isn’t obvious whether this category is double-counting murders and felony assaults. I’m assuming it is, and am thus just using the top numbers. Though including or not including shootings doesn’t change the totals much anyway, as the raw numbers here (1,410 incidents so far in 2021) are only ~4% of the ~34k other violent crimes over that same period. (There’s also a bit of normal / minor conflation in all crime reporting between “homicides” and “murders”, as different jurisdictions may define manslaughter as one but not the other. The NYPD’s main category here is “murders and non-negligible manslaughter”.)
Or there was a change in this span to the category definitions, or to reporting rates, or to some other more nuanced variable. This would take some real reporting to tease out.
The Bronx itself had clearance rates well over 100% for quarters 1 and 4 in 2019 because of the UCR quirk outlined in the PS. But it’s otherwise still strange that the Bronx was such an outlier, with much higher clearance rates than surrounding boroughs that year. It could be that their homicides are more easily solved, it could be that they have unusually gifted detectives, it could be they have more homicide-department manpower, or it could be that a murder committed in one borough can be cleared in another based on the residence of the murderer. I couldn’t find any good commentary that cleared up said mystery, and it wasn’t super relevant here anyway: because even if the reporter meant to ground their 90% clearance rate claim on Bronx 2019 stats specifically, her article says “in the city” and her tweet indexes on the NYPD as a whole. And that’s just objectively wrong.
This kicked in in 2019, but likely had a disproportionate impact in 2020-2021, as the eyewitnesses/informants most likely to be scared off by the new rule were/are those testifying against gangs. If this NYT editorial is right that roughly half of the post-COVID shooting increase came from gangs, we’d expect a clearance drop of maybe a point or two.
While this stat is from 2020, I’m taking it as intuitive that 2021 will have seen a far-above-baseline number of sick days too, if maybe fewer than 2020.
The 60% estimate is from the main NYT article, which isn’t specific about which NYPD disclosure they got it from. The public data I found cuts off in Q4 2020.
Terrific article, Jeremy. I'd also love an analysis that looks at the articles or angles NYT publishes vs. those others print. I know there are already folks like AllSides.com show liberal, central, and conservative articles on a given topic. What I wonder about is the balance of pro-progressive articles vs. articles that are critical of progressives. Though, I'd bet that would be a tough analysis to make.