The Department of Justice released a report yesterday looking at possible explanations for why murder rose in America last year before ultimately concluding that the data just isn’t good enough yet to know for sure. It’s certainly worth a read and provides interesting insight into a few aspects (namely whether drugs or returning prisoners contributed to more murders) that I hadn’t considered. But one aspect in particular has been bothering me.

The report focuses attention on 56 cities with a particular examination of ten cities that accounted for roughly two-thirds of the overall murder increase. Per the report:

The top ten cities not only produced two-thirds of the big-city homicide increase in 2015, they also experienced a far larger percentage increase than the sample as a whole. The percentage increases in the top ten ranged from 90.5 percent in Cleveland to 12.9 percent in Philadelphia. The average homicide increase over 2014 in the top ten was 33.3 percent, compared with a 16.8-percent rise for the sample as a whole.

The largest difference between the top ten and other cities in the sample, however, is their race/ethnic composition. The top ten have larger black populations and smaller Hispanic populations than the other cities.

The 10 cities are Baltimore, Chicago, Houston, Milwaukee, Cleveland, Washington DC, Nashville, Philadelphia, Kansas City and St. Louis. Looking at the problem in this way can lead readers to believe the rise in murder last year was isolated to 10 cities with large black populations.

Of course Cleveland, Nashville, Kansas City and St. Louis all rank outside in the top 15 of American cities in terms of size of their black population, two cities with the largest black populations in the country (Detroit and Memphis) saw drops in murder in 2015, and Houston, Nashville and Kansas City are all under 30 percent black. Also, Oklahoma City (tied for 10th with +29 murders and a small black population) is left off the list. But that’s neither here nor there.

Taken together these cities make up over 60 percent of the total change in murder which isn’t insignificant.

The real problem with analyzing the raw murder change  is that cities with larger black populations tended to have bigger murder totals to begin with. This is shown on the below scatter plot of US cities by 2014 murders (X-axis) and black population according to 2014 Census estimates (Y-axis).

Graph 1So it’s not inherently unusual that cities with larger black populations would have bigger rises in murder, they already had more murders to begin with for the most part! And that’s what we see when we chart the change in murder in 2015 with black population. The relationship isn’t all that strong (r – .34), but cities with more black people tended to have more bigger raw rises in murders in 2015.

Graph 2

The real question, however, is whether these cities had more murders in 2015 because of a substantial change that was isolated just to these cities and not cities with smaller black populations, or whether looking at the raw numbers hides pretty substantial — albeit smaller raw —  rises in murder in cities with smaller black populations.

Turns out it’s the latter.

When the percent rather than raw change in murder from 2014 to 2015 is graphed along with the black population we see there’s no relationship whatsoever. Cities with smaller black populations were just as likely to have big percentage increases in murder as cities with larger black populations.

Baltimore, Cleveland, Washington, DC and Milwaukee all have large black populations and had really bad years in 2015, but Orlando, Anchorage, Fort Wayne, Anchorage and Denver all have relatively small black populations and had just as bad, if not worse, years in terms If yof percent change in murder.

Graph 3

If you don’t like using percent change you can also look at the standard deviation change which shows just how unusual the change in murder is for each city. The below graph shows a similar lack of correlation between the unusualness of the change in murder and a city’s black population.

Graph 5

In fact, 13 of the 17 cities in the US over 250,000 people that had a 50 percent or more rise in murder last year are also under 30 percent black (Milwaukee, Cleveland, Baltimore and Washington, DC being the exceptions).

As the DOJ paper notes, last year’s national rise in murder was very real. But as I’ve tried to show it was also very national. And people thinking it was isolated can lead to dangerous takes like this one by CityLab.

These 10 also account for two-thirds of the total rise in homicides from 2014-2015, so any talk about a new wave of violent crime should probably be applied specifically to just this subset of cities. Final numbers collected and crunched by the FBI later this year will give a more complete picture of the nature and distribution of homicides, as well as whether similar rises were found in smaller cities.

So, the rise in killings is indisputable at this point, even if they occurred in just a few cities.

Last year’s rise was most assuredly not limited to just a few cities with large black populations. The rise in Baltimore was unprecedented, but so was Omaha going from 32 murders in 2014 to 50 in 2015 or Louisville going from 56 to 81 murders.

Something happened with murder in the United States though it is certainly too soon to tell whether it was a one year blip or the start of a larger trend. isolating the change to only a few cities obfuscates how serious the change is in lots of cities and creates a false narrative of what happened last year.