Wednesday, April 12, 2017

Looking at fake news about insanely high homicide rates in U.S. cities

Univision's web site has a report claiming that four U.S. cities are among those with the highest homicide rates in the world. The statistics are based on FBI reports, but are computed by some group in Mexico despite the insistent and express warning by the FBI against the procedure followed:
The data user is, therefore, cautioned against comparing statistical data of individual reporting units from cities, counties, metropolitan areas, states, or colleges or universities solely on the basis on their population coverage or student enrollment. (The emphasis is as presented in the original).
Why? Well for one reason, the original source of homicide data is often a coroner's office; but these are frequently county offices. City boundaries are not the same as county ones, and even if the city's police department is the one reporting the figures, these may well be from a broader area.

But even if totally "accurate," using population statistics doesn't work. A city of 300,000 could easily have a million people in it every weekday. And local political subdivisions in the United States are often arbitrary and fragmented.

In Atlanta, where I live, there are --depending on who is counting-- between five and thirty-some counties and scores of cities. These cross county lines, counties divide neighborhoods and your house might be partly in one county, partly in another. Talking about crime rates might be meaningful or might be a discussion involving a completely arbitrary area created a century or more ago for gerrymandering purposes or to exclude Blacks and that we have inherited even though it simply has nothing to do with the real world today.

A Mexican group used to that country's system of municipalities might not understand this, but a journalist in the United States needs to.

This is why the FBI only computes homicide rates for standard metropolitan statistical areas and NOT individual cities or counties, and warns that even state rates may not be meaningful (because major metro areas cross state lines).
I can't imagine that the writer even bothered to check the original source of the information, which took me about three minutes. If he would have done so, in addition to seeing the strongly worded warning not to do what his source was doing, he would have seen that the statistics for the supposedly most dangerous city, St Louis, MO, made absolutely no sense if handled in this way. By that measure, the city of St Louis itself supposedly had a homicide rate of 59.3 in 2015, making it one of the most dangerous in the world. But the metro area had an overall rate of 10.5, a bit high but not extraordinary, and if you calculate the rate for the metro area outside the city limits only, you get 4.3, even lower than the U.S. national average of 4.9 per 100,000 population. It makes no sense that the murder rate drops more than 90% by crossing the street to get outside the city's jurisdiction. I know that data journalism is really hot right now, but it is no substitute for common sense.

No comments:

Post a Comment