A recent article on gun homicide in the US could not have done a better job of telling us nothing while convincing uncritical readers that they were learning something (thanks to Paul Bergen for pointing it out, albeit for a slightly different reason). It was based on a MMWR report (the publication which I previous described as basically being the US CDC’s research blog) which provided some of the missing information, but that did not solve all the problems.
The basic message is (from the news story):
According to the CDC, 25,423 murders by gunfire took place in the United States in 2006 through 2007 — the years of the most recent available statistics.
That is pretty horrific, and I am not just talking about how long it takes to compile such simple statistics. (Note that the MMWR report was also about suicide which had equally troubling statistics.) In case that was not clear enough, an easy way to put this in perspective is to compare it to other countries, but neither the news report nor the MMWR bothered to tell us that, so the simple and obviously relevant observation that this number is ridiculously high for a state with a functioning government and that fights its wars elsewhere.
The news went on to quote a CDC employee:
Among these deaths, the rate of firearm homicides was higher in inner cities than in other parts of cities and higher than the murder rate of the country as a whole, Dahlberg said. People living in 50 of the largest cities, in fact, accounted for 67 percent of all firearm homicides.
Should we be impressed by that? Perhaps it is most useful, for Unhealthful News purposes, to observe that the reporter seems to be telling us to be impressed, and most readers probably blindly comply. But do you even know what portion of the population lives in the 50 largest cities? Do readers stop to ask if it might be as high as 67%? I suspect it is that high for most every other country in the world (there might be a handful of exceptions). It turns out to not be quite that much for the U.S., though I suspect few readers would know. And yet they are asked to be impressed by a statement that should only be impressive if they know that additional fact. The MMWR report actually did provide the missing information, as you might expect:
The firearm homicide rate in the 50 largest MSAs collectively was 5.2 per 100,000 persons per year, and 66% of these MSAs (33 of 50) had rates exceeding the national rate of 4.2. The central cities within these MSAs collectively had an annual all-ages firearm homicide rate of 9.7, and 86% of these cities (48 of 56 cities with reportable all-ages firearm homicide statistics) had rates exceeding those of their MSAs.
Now we know that the major metropolitan area rates were somewhat higher than the national average and their central cities were more than twice the average. That is potentially interesting to the reader. Too bad the health reporters tell us statistics that sound impressive without understanding what information is needed for them to be genuinely impressive. The news story went on:
In addition, children and teens aged 10 to 19 in these areas — more than 85 percent of them male — accounted for 73 percent of all firearm homicides in that age group, Dahlberg noted.
Oh, wow. It is even worse for urban youth. Oh, wait. This just means that the elevated rate for youth in the cities barely exceeded the average across all ages. So what? Despite them playing this up, it really tells us nothing about how bad a problem there is. Something like this is probably true for the UK, with its very low gun violence. As for 85% of them being male, that is a meaningful statistic because we happen to all know that about half the population is male, but it is also not very impressive because we know other things about males already (indeed, it is kind of surprising it is that low).
The game of telling the reader she should be impressed by a meaningless observation can also be found in more subtle implicit claims too. Going back to the second quote, is it surprising that the rate is higher in inner cities than elsewhere in cities, and that in turn is higher than outside of cities? Consider that question even without reference to socioeconomics, the location of drug markets, culture, and such, but just in terms of the simple arithmetic. Should we not expect that as population density increases, gun violence would increase? Notice that the statistics CDC reported are in terms of events per number of people. This linear measure is about right for most of what CDC studies, like cancer incidence. But what if we were collecting statistics on, say, “number of conversations with people you were not scheduled to see that day”? In that case, we would expect the rate to increase faster than linearly with the number of people somewhere. Each additional person in a network (a real social network of people who actually see each other) creates more potential chance meetings if there are already more people in the network. For the present case, you can think of each new person in a socially-linked population as one additional potential perpetrator of violence against anyone already there as well as one additional potential victim of any violent person already there.
Translating those exact statements into a mathematical function most likely overstates the rate of increase with population density. But the implicit claim, that violence should only increase linearly with population increases, undoubtedly understates is. So should we think that the higher rates in denser populations are particularly disturbing or just inevitable? It is really not clear.
Finally, this being a US government report, they cannot simply convey some scientifically useful information and shut up. The research was a useful description of the basic descriptive statistics, but that is all. It could not even tell us that more interventions are warranted, let alone anything about particular intervention. But they just could not pass up a chance to turn tangential information into a chance to rationalize their preferred policies. Naturally the government called for sensible European-style gun control, which would have the added benefit of ending the flow of guns used in the ongoing near-genocide in Mexico and Central America. Hahaha – yes, of course, I am kidding. The MMWR report and the interview talked of various programs designed to get young people to not resort to violence, but since the US government encourages the gun culture, the message that people, not guns, are to blame was clear.
Naturally, the press pointed out this omission, and challenged the government spokesperson about it.
Yes, of course, I was just kidding about that too.