“Bury the lede” is an old newspaper trade term for not disclosing the interesting part of the story until so late in the article that few people read it. Paragraph #1, Senator Bedfellow gave money to a shelter for battered women; paragraph 48, Senator Bedfellow was charged with wife beating in 2006.

It is a propaganda tactic, and it is used too often in gun policy research. In this case concerning suicides in California, the stunt was aided and abetted by the California government.

Woe, thy name is Wintemute

Long ago, a doctor with zero criminology credentials started publishing on gun control topics. His early research was, in a word, hilarious. One chunk of his intellectual effluvium used “observational evidence” about people at gun shows in two states. Generalized, the assumption the doctor made was that if Bill asked his friend Ted about a firearm for sale at the gun show, and Ted nodded and Bill paid for it, then this was a strawman sale and Ted was a prohibited person.

The more likely case, that Ted was more of an expert and Bill was relying on his guidance, appears to have never penetrated the doctor’s cranium. This led his paper to receiving almost as much derision as another doctor’s paper about a gun in the home being 43 times more likely to kill a resident (that bit of criminology malpractice once had two separate entries at JunkScience.com).

So, when the name Garen Wintemute appears on a paper, it needs a vivisection (the paper, not the doctor).

The claim

The paper ¹ asserts owning a handgun is associated with a higher rate of firearm suicide, before skating off onto even thinner ice.

Thanks to the political biases of the State of California (AKA the State of Emergency), Wintemute and company has received tax money to conduct gun violence research, and the politicians pretty much get the results they want. In this case Wintemute had access to data most people don’t, including details on the owners of registered guns (alert to civil rights attorneys: there may be a privacy case here), voter registration roles with identifying data, and more.

To their credit, the approach of the research was not entirely unsound. Simplified, they merged voter registration data (which provided street address for geolocation) with handgun purchasing and sales records (to know who allegedly did and did not have access to a handgun) and cause-of-death data to see if buying a gun led to killing oneself with it. We will get to some of the data and methodology inequities later. But we’ll note in passing that if the data, proxies and confounding variables had been more robust, this might have been a good approach.

The age obfuscation

Did Wintemute and company not care about old people?

In the paper, they noted the median age of firearm owners, but not the median age or geography of people who commit suicide with a gun. Both are well known data-skewing phenomena, but the “researchers” were conspicuously reluctant to disclose their data on these topics.

Guess we’ll have to do it for them.

click for larger version

click for a larger version

Thanks to the Centers for Disease Control, California suicide stats are available to anyone (you don’t have to be a tax funded propaganda organization working with insider government data). They also allow anyone to get some rather interesting breakdowns. This chart shows the mode of suicide for Californians (in the same range of years as the study on which we are performing an autopsy). As with national trends, we see that firearm suicide is heavily skewed to older people. And like the rest of the country, it is skewed toward old, white men living outside of metro areas.

This is critical to the study at hand. Recall that they used voter registration data to triangulate gun ownership, and in their paper admitted that this added up to only 61% of adult Californians. This is fairly imprecise on its own but gets much worse when you look at the individual counties (which may explain why the authors  opted out of reporting geographical differences).

Here’s an example. The counties with the highest voter registration rate (coastal Marin, 85% registration rate, primarily Democrat) and the lowest (rural Tulare, 54%, largely Republican). The population density, culture differences and other demographics (including likely pre-existing gun ownership rates) are so vastly different that the lack of reporting about these differences is suspicious, given that the authors geocoded the entire study and could have noted variations in suicide rates (Marin’s suicide rate is 15.4 dead per 100,000 population; Tulare’s is 9.4, which discounts the nationally common trend of old, rural men being more likely to kill themselves).

Why is this seeming distraction important? Let us tell you.

Immediacy vs reality

The much beloved former mayor of Alameda, California shot himself dead to avoid a protracted battle with cancer. He was 67. Older white man, prime suicide candidate and likely (unconfirmed) owned the gun in question for a long time before his tumors appeared.

His suicide was not spontaneous. It was the result of his cancer spreading and likely becoming terminal. He thought through his options – a clear-headed decision, if you will. Most people do so, even the chronically depressed. Sundry psych studies report that many (perhaps most) suicidal people think about killing themselves for months or even years before committing the act. This is an important dig at the study under our microscope because of “buy to die” cases and the excluded others.

click for larger version

In the study, one table (seven pages past the headline conclusions) shows the amount of time from the acquisition of a handgun to using a handgun for suicide. We do see that in terms of “people-years” (the authors did not provide the raw data) many people do buy a gun to kill themselves.

But there is zero accounting for how long before the handgun purchase these people had been thinking about suicide. This gets very important as the authors appear to have removed from the data people who owned handguns before the beginning of the study period (“excluded cohort members who had acquired one or more handguns before coming under observation during the study period”).

Let’s take the case of old men, including California mayors with metastasized cancer. Older people – who have lost spouses, friends, children, who may not have enough money to survive retirement, who may have painful debilitations or terminal diseases – may decide to die. Some of them may opt to use a gun and need to buy one for the purpose. Others, likely in the rural areas, likely already have one or more guns and may have had them long before a doctor probed their personal gun ownership data.

So, the exclusion of people who owned guns before the study period, but did not purchase one during it, artificially skews the data into the “spontaneous suicide” camps. How much flatter the bars in the previous chart might be is completely unknown, thanks to the intentional exclusion. As people age odds are they don’t buy many new guns but may already own several. Thus, this most suicide-prone group of people were mistakenly dropped out of the mix in this study for being old and on fixed incomes.

The lack of exploration of these well-known issues – the age, mental health, and metro status of the victim – makes their absence from this study highly suspicious, especially since the authors claim that they “geocoded residential addresses and then assigned them to census tracts,” and that they calculated age clusters for handgun ownership, but not handgun suicides. What is not revealed is likely interesting to voters (which may have been the motivation to obscure critical insights).

But wait, there’s more!

As odd as this non-reporting of data is, there are many awkward and silly aspects to this study.

On the silly side is their statement, “Whether a suicide attempt is fatal depends heavily on the lethality of the method used.” As we showed in our original international review, people around the world find many lethal methods for ending their existence. Rope is just as lethal as bullets, a fact that Wintemute and his acolytes fail to grasp (and let’s hope these doctors have a better grasp on viruses and cancer, lest we all die under their “care”).

One problem is the weighting (we’ll skip the obvious jokes about doctors and waiting rooms). Because their core indexing data (voter registrations) only covered 61% of California adults, and because voter registration is highly variable from county to county, they were forced to statistically weight populations. This is not insanely bad, but it points to an area where the source data (garbage in) may lead to suspect results (garbage out).

Lastly, the authors seem to have pretended to use control proxies to test for confounding issue (other variables that might explain data trends). In particular, they used non-suicide lung cancer, endocarditis, and alcoholic liver disease to test for associations with smoking, injecting drugs, or drinking their livers into submission (indicators of other bad lifestyle choices that occasionally lead to suicide). But as our cancer-ridden mayor shows, there is a disconnect whereby people who don’t abuse themselves may develop these diseases anyway and decide to die from suicide instead of the illness.

What not up, Doc?

How people commit suicide has been proven meaningless. In the year of the original Gun Facts international review, we found no correlation between successful suicides and firearm availability. Because of the seemingly intentional skewing of the originating data, even this study fails to prove a real correlation between purchasing handguns and killing oneself (and even if there was an unpolluted association, “substitution of means” has been proven, thus net suicides would be unchanged).

If doctors committing statistical malpractice actually cared about people, they would focus on what causes people to want to die, and defuse that bomb – instead of picking just one out of hundreds of ways suicidal people exit.