EYE ON THE NEWS
Stat Crimes Matter
How researchers try to obscure the existence of the Ferguson effect
February 22, 2016
A group of criminologists has purported to answer the question: “Was there a Ferguson effect on crime rates in large U.S. cities?” The “Ferguson effect” refers to the phenomenon of police officers backing off from proactive policing in response to the anti-cop Black Lives Matter movement, with a resulting rise in violent crime. The criminologists answer their own question with a minutely qualified “No.” In fact, their analysis resoundingly confirms the existence of the Ferguson effect.
Anyone not well-versed in “discontinuous growth models,” “empirical Bayes predictions,” the “Bonferroni correction,” and “Nakagawa’s hypotheticals” will have to take on faith a great deal of the recent paperpublished in the Journal of Criminal Justice. The authors, four professors led by sociologist David Pyrooz of the University of Colorado Boulder, created a complex econometric model that analyzed monthly rates of change in crime rates in 81 U.S. cities with populations of 200,000 or more. The other 24 cities in that size cohort were not included in the study due to lack of crime data.
The researchers found that in the 12 months before Michael Brown was shot in Ferguson, Missouri—the event that catalyzed the Black Lives Matter movement—major felony crime, averaged across all 81 cities, was going down. In the 12 months after Brown was shot, that aggregate drop in crime slowed down considerably. But that deceleration of the crime drop was not large enough to be deemed statistically significant, say the criminologists. Therefore, they conclude, “there is no systematic evidence of a Ferguson Effect on aggregate crime rates throughout the large U.S. cities . . . in this study.”
But the existence of a Ferguson effect does not depend on its operating uniformly across the country in cities with very different demographics. When the researchers disaggregated crime trends by city, they found that the variance among those individual city trends had tripled after Ferguson. That is, before the Brown shooting, individual cities’ crime rates tended to move downward together; after Ferguson, their crime rates were all over the map. Some cities had sharp increases in aggregate crime, while others continued their downward trajectory. The variance in homicide trends was even greater—nearly six times as large after Ferguson. And what cities had the largest post-Ferguson homicide surges? Precisely those that the Ferguson effect would predict: cities with high black populations, low white populations, and high preexisting rates of violent crime.
A virulent anti-cop protest movement dedicated to the proposition that murderous, racist cops are the biggest threat facing young black men today will have its biggest impact on policing in black neighborhoods. It is in these neighborhoods that cops will face the most hostility from residents steeped in the Black Lives Matter ideology and where cops will most worry that, if an encounter with a civilian goes awry, they will become the latest racist officer-of-the-week on CNN. It is in black neighborhoods, in other words, where proactive policing—making pedestrian stops, enforcing quality-of-life public order laws—will be most inhibited. And given the already high rates of violent crime in black neighborhoods, any drop-off in policing is going to unleash even more crime, since it is in these high-crime neighborhoods where informal social controls have most disintegrated and where cops alone stand between law-abiding residents and anarchy. Even if the Black Lives Matter movement inhibited proactive policing uniformly in cities across the country, a place like Scottsdale, Arizona, say, will suffer less of an impact if cops back off, because the police are not as essential there to maintaining order as they are in Baltimore and St. Louis.
The researchers are unwilling, however, to accept the implication of their findings. They grudgingly admit that “the data offer preliminary support for a Ferguson Effect on homicide rates in a few select cities in the United States”—those cities, according to their model, are Baltimore, St. Louis, New Orleans, Washington, D.C., Newark, Milwaukee, Rochester, Detroit, Oakland, Richmond, Cincinnati, Fort Wayne, and Baton Rouge—but then they backpedal furiously. (Cities that barely missed making the “statistically significant” cut include Kansas City, Indianapolis, Oklahoma City, and Chicago.) What’s important about those cities, they claim, is that “they had much higher crime rates before Ferguson.” Those higher crime rates, they say, “in turn may have primed [those cities] for increases in crime.”
That conclusion is groundless. The proactive policing revolution that began in the 1990s had its greatest effect on high-crime cities; crime went down dramatically in neighborhoods that had been written off as ungovernable. If cities with a “higher proportion of black residents, lower socioeconomic status, and more police per capita,” in the authors’ words, were primed for a crime increase, and if those factors “lead to questions that may inhibit any ability to attribute crime increases specifically to the Ferguson Effect,” the authors need to explain how those cities experienced a crime drop in the first place. Moreover, if the authors think that high-black, high-crime cities were due for a crime increase regardless of changes in policing and a worsening in resident attitudes toward law enforcement, they didn’t alert us to such a reversal ahead of the fact.
In a separate analysis, the authors disaggregated the seven felonies included in the FBI’s crime index and tracked the movement of each felony averaged across all 81 cities. Robbery registered a statistically significant upward surge in monthly rates: before Ferguson, the aggregate robbery rate was dropping; after Ferguson, the rate reversed course, rising enough to be considered statistically significant. The criminologists conclude that “changes in robbery rates constitute the lone exception to a spurious Ferguson Effect,” but demur from speculating why that may be. Perhaps it is because robbery and drive-by shootings are the quintessential violent street crimes, both committed disproportionately by blacks. If police are making fewer street stops, thus deterring gun-carrying less, a rising robbery rate is not contrary to what the Ferguson effect would predict. (Shootings are not captured in the FBI data used by the researchers, so their pre- and post-Ferguson trajectories are not easily available.)
A few analysts have pointed out that the paper’s dismissal of a more widespread Ferguson effect rests on arbitrary statistical conventions. Fordham law professor John Pfaff notes that the rate of change in the aggregate violent crime rate rose tenfold after Ferguson. That increase was not deemed “statistically significant,” however, because it missed falling within the conventional statistical confidence interval by .02 crimes per 100,000 residents per month. The confidence interval tells you how certain you can be that the events being measured actually happened or were not the products of random chance. Statistical conventions deem a data distribution statistically significant only if there is not more than a 5 percent chance that the data points were arrived at in error or that the distribution curve mapping those data points would have occurred randomly. Had the increase in the rate of change in violent crime increase been .02 crimes per 100,000 per month higher, the authors would likely have had to change their conclusion regarding a “spurious” Ferguson effect. As it is, the existing tenfold increase in the rate of change has only a 12 percent chance of being a mirage—that is, the product of incorrect crime data, say, or of a random distribution of events, according to Manhattan Institute fellow Scott Winship. And the aggregate increase in the homicide rate of change, which the authors dismiss as “statistically insignificant,” has less than an 11 percent chance of being a random occurrence, according to Winship. Concludes Pfaff about the Pyrooz study: “So [the] claim of ‘no Ferguson Effect’ is built on little more than a century-old arbitrary line that arbitrarily balances 2 core error costs.”
The authors are doing nothing untoward in resting their conclusions on statistical conventions. But a lay reader may conflate their finding of “no statistically significant effect” with no effect at all and will likely not understand how narrowly a tenfold rise in the rate of change in violent crime missed being deemed statistically significant—if that lay reader even grasps the change at all from the paper’s tables.
The Pyrooz article will undoubtedly become a standard artillery piece on the activist and academic left. You would think that the fact that the Ferguson effect has been most pronounced in black areas would be cause for concern among those who claim to represent black interests against a sea of racism and oppression. In 2015, homicides in the 50 largest cities rose nearly 17 percent, “the greatest increase in lethal violence in a quarter century,”according to the Washington Post. The overwhelming majority of those additional victims were black. But the furious attempt to deny the Ferguson effect shows yet again that black lives seem to matter only when they are taken by police officers.
Heather Mac Donald is a contributing editor of City Journal and author of the forthcoming The War on Cops (Encounter Books