The 2020 murder surge wasn't about Covid. By Matthew Yglesias

www.slowboring.com

Matthew Yglesias

31 Oct 2023

12 - 15 minutes

Paul Krugman wrote a column last week on public perceptions of crime versus crime realities with some analogies to and implications for public perceptions of the economy.

I’m in agreement with probably 80 percent of what he writes. But I think the remaining 20 percent exemplifies why liberals have a hard time securing the public’s trust on this issue, which exacerbates the misperceptions that Krugman is nervous about. The stylized facts about crime he is working with are that murder (and non-fatal shootings, but murder is the best-measured offense) went up a lot in 2020 and up a bit more in 2021. We then had a decent murder drop in 2022 and another one in 2023 that is now on track to leave the 2023 murder rate lower than the 2020 rate. Dark Brandon reversed the Trump crime wave.

So what happened in 2020?

I think most people are aware that George Floyd was arrested and then killed by a police officer in Minneapolis while several of the officer’s colleagues stood around and watched. That touched off a massive and multi-dimensional social upheaval about race and racial equality that had a particular locus of concern around questions of policing and criminal justice. And then crime spiraled out of control.

Or as Krugman puts it:

    Unlike the somewhat mysterious decline in crime in previous decades, this crime wave wasn’t too hard to explain. The Covid-19 pandemic led to a lot of isolation and disruption, plus a lot of psychological stress, making it plausible that some Americans became disconnected from the social bonds that usually keep most of us law-abiding.

In other words, he not only thinks the 2020 crime wave had nothing to do with Floyd and the post-Floyd reaction, he thinks this is so obviously the case that he doesn’t even need to argue about it. The causes of the 2022-2023 murder decline, according to Krugman, are obvious — the virus went into remission. The only question is why don’t people realize murder is down.

My view is that there are, actually, a lot of valid and unanswered questions about why murders spiked in 2020, and almost all of those questions center around Floyd and the Floyd fallout. The reason there’s a fair amount of mystery is that it’s challenging to pin down exactly what about the Floyd fallout was responsible for the large increase. There were, in fact, a handful of cities that took steps to defund their police departments, but most places didn’t do this, and the crime increase was very widespread. Similarly, the handful of “progressive prosecutors” scattered around the country are not nearly numerous enough to explain the broad national trend. But I do think there’s evidence that it had something to do with Floyd and the post-Floyd fallout.

Failing to recognize that is bad across multiple dimensions. One is that it’s substantively important to try to understand exactly what went wrong and how we can do better. But the other is that progressive discomfort with acknowledging the facts here speaks to some of the broader epistemic issues in mainstream left of center politics.

Figuring out causality is always hard, because our evidence is almost always correlational and (as people on the internet are happy to tell you) correlation is not causation. So one obvious task when a hypothesis is based on a time-series correlation (crime went up during Covid disruptions) is to check for a cross-sectional correlation. After all, Covid was an international phenomenon. So if Covid disruptions caused murder to rise, we’d expect to see murder up everywhere.

But that’s not really what we see. The United States had a larger percentage increase in murders than any other developed country, and that’s from a higher base of murders.

To be clear, this relatively large percentage increase in Germany was from a base murder rate that was about half the murder rate of the safest American states — not half the murder rate of the United States, half the murder rate of Maine.

And in lots of large European countries, homicides actually fell in 2020.

If you look at the two-year span, the United States is an even bigger outlier. I think that supports the hypothesis that European countries were experiencing random variation, while the United States experienced a much more meaningful shift.

This is not to say that Covid was totally unrelated to the increase in crime in the United States. Alex MacGillis reported in 2021 from Philadelphia’s murder surge that the whole court system had stopped functioning:

    The court system stopped processing new and existing cases almost entirely. [District Attorney Larry] Krasner said in an interview with ProPublica that he questioned the extent of the shutdown and unsuccessfully urged the courts to hold more hearings remotely, even if actual trials had to wait for a resumption of in-person proceedings. “I was a little disappointed in the courts’ response,” he said. “We could have done better and we could have done more.”

That’s bad. But “let’s not process criminal cases” seems like a non-obvious response to Covid. So while this insight helps us understand why Philadelphia’s surge was especially severe, it doesn’t really explain the broader trend. I think the real impact of Covid is that it helped transform the post-Floyd protests into a much bigger deal than they otherwise might have been.

Although murder is falling nationally in 2023, it’s rising in two large cities.

One is DC, where I live, and the other is Memphis. Among crime policy people I talk to, the DC situation is considered really interesting and a little bit vexingly hard to explain, but the Memphis situation — though very sad — is easier to explain. In Memphis, police officers brutally beat Tyre Nichols after a traffic stop and foot chase and he died as a result of his injuries. This incident was in many ways one of the most obvious, least controversial police misconduct cases of all time. Both the victim and all of the officers involved were Black, so there wasn’t a huge outcry about individual-level racism. And we didn’t have a racist counter-mobilization with people arguing that Nichols got what he deserved or — as in the Floyd case — that his death was somehow unrelated to the misconduct. It was just an egregious example of misconduct and basically everyone agreed that’s what it was.

Nonetheless, not only did crime rise, experts were totally unsurprised that it rose.

And yet, even though this has become well-known and not that controversial among researchers (the controversial part is exactly why these high-profile misconduct cases lead to higher crime), my sense is that a lot of generally well-informed political observers still don’t know this.

One important piece of statistical evidence comes from a June 2020 study by Tanaya Devi and Roland Fryer of pre-Floyd incidents. They find that in general, Department of Justice “pattern-or-practice” investigations of police departments for racial bias lead to better outcomes, including “a statistically significant reduction in homicides and total crime.” But there is an important exception: When such an investigation is preceded by what they term a viral misconduct incident, there is “a large and statistically significant increase in homicides and total crime.” They attribute about 900 homicides (this is pre-Floyd) to the aftermath of such viral incidents.

Progressives probably recall that after Michael Brown’s death there was a spike in homicides in the St. Louis area, which led the St. Louis police chief to blame the “Ferguson Effect.” Richard Rosenfeld, a well-known criminologist who also happens to live in St. Louis, initially offered a skeptical take on that hypothesis and it became a bit of an ideological battle zone. But I think a lot of liberals who didn’t continue to follow the issue don’t realize that Rosenfeld changed his mind about this as more information accrued. There was also a pronounced rise in crime after the Freddie Gray case in Baltimore.

I think that piece of knowledge — that when a city has a viral police misconduct issue, crime rises in the aftermath — goes a long way toward explaining why we had a national surge of shootings in 2020 and 2021.

George Floyd’s death was not so different from some of these other instances, but it did involve a racial dynamic that drives the maximum amount of attention. And perhaps even more importantly, it happened in the context of Spring 2020 Covid shutdowns, which boosted rally turnout. Actual legal restrictions on Americans’ conduct were not that strict in the United States compared to those in many European and Asian countries. But progressive Americans in particular were under pretty intense social pressure to act as if the country was experiencing strict lockdown orders. One of the few real exceptions, one that was validated and valorized in progressive circles, was attending racial justice protests.

Of course, it is hard to say for sure. If you look at the time series trend of 2020 murders, you see two things that can be interpreted both ways:

    The increase in homicides did not occur in April 2020 when lockdowns began; murder was up a little relative to April 2019, but that’s because 2019 had an unusually peaceful April. If you compare to April 2018 or April 2017, there was nothing special happening. The quantitatively enormous spike was post-Floyd.

    May 2020 really was deadlier than a typical May. Not by the kind of huge margins that we saw later in the summer, but it was definitely an outlier, so the phrase “the increase in homicides started before Floyd’s death” is accurate. 

John Roman from the University of Chicago made what I think is the best effort to do what Krugman didn’t do and really make the case for why we should see this as a Covid effect rather than a Floyd effect. He argues that, especially in the summer, we saw “dense clusters of young men stuck at home with little to do carrying the burden of past trauma, knowing those with whom they hold deep grudges are close by and home too.” Roman acknowledges the Ferguson Effect research, but says this is different because those findings are always local and Covid was national.

Still, I do think that the evidence as a whole supports Floyd:

    Covid happened everywhere and the murder spike is not evident in foreign countries.

    The murder spike occurred nationally, and was more severe in big cities than in small towns or rural areas. Big cities like Dallas got hit hard even if they were in red states that had lax Covid non-pharmaceutical interventions.

    The timing of the murder surge is a little ambiguous, but it doesn’t line up with the start of stay-at-home orders, and it wasn’t alleviated by the relaxation of policy over time. Roman’s argument about summer would make more sense if the crime surge were limited to cold cities with strong seasonality (Chicago), but it happened in Phoenix and Atlanta, too. 

All you need to get to “the Ferguson effect, but national” is the belief that there was a lot more national media attention paid to police and policing in the wake of Floyd’s death than in the wake of other viral incidents. And that is clearly true.

I bring this all up not to argue with Krugman or to dunk on the left but to say that there is a pretty serious problem here that we, as a society, need to try to understand. We’re over 2,000 words at this point, so I’m not going to delve into the debate over exactly why high-profile misconduct cases lead to a surge in serious crime.

But I will note that one version of this argument, one that I associate with Heather MacDonald, essentially invokes the Ferguson Effect to suggest that nobody should ever complain about police misconduct.

If you don’t like that idea — and I don’t think that you should — then you do need to admit that something or other has gone wrong with the way we are collectively processing these events. Improving the overall quality of policing continues to be an important issue, and I think problems related to the Ferguson Effect are part of the reason that, realistically, better police forces are going to be more expensive not cheaper. We need to think harder about how to get more people, including people with different political beliefs and different preferences around urbanism, interested in careers in law enforcement. That ought to include efforts modeled on Teach for America to recruit a new cohort of talented idealists.

Reasonable people can disagree about these ideas, and I think the debate about the exact causal pathways behind the correlations we are talking about here remains unsettled.

My point for now is that the first step in solving a problem is to acknowledge that it exists. If we talk ourselves into the view that the crime surge and subsequent waning was only about Covid so we don’t need to worry about it anymore, the country is going to be caught flat-footed the next time something goes wrong in a way that’s high profile enough to attract national attention.