Saturday, November 5, 2016

Trying to Dissect the Center for American Progress Gun Report (America Under Fire)

The Center for American Progress, a progressive and liberal think tank, recently released a report "America Under Fire" that claimed a strong correlation between state gun law strength and reduced levels of gun violence. Their main correlation chart is shown below. To get this chart, CAP researchers tallied up ten variables on gun violence, took the average, and plotted it against gun law strength (an admittedly qualitative and subjective measure) as defined by the Law Center to Prevent Gun Violence.

From what I can tell so far, most of this correlation can be traced to two variables, suicides and gun traces to where crime guns were originally sold. The rest of the correlations are very low in significance.

The ten variables used by CAP are:
Rate of overall gun deaths per every 100,000 people, 2005-2014
Rate of gun suicides per every 100,000 people, 2005-2014
Rate of gun homicides per every 100,000 people, 2005-2014
Rate of fatal gun accidents per every 1 million people, 2005-2014
Rate of mass shootings per every 1 million people, 2006-2015
Rate of intimate partner gun homicides of women per every 1 million women, 2005-2014
Rate of gun deaths among people younger than age 21 per every 100,000 people younger than age 21, 2005-2014
Rate of law enforcement officers feloniously killed with a firearm per every 1 million people, 2005-2014
Rate of fatal shootings by police per every 1 million people, 2015-2016
Crime gun export rates per every 100,000 people, 2010-2015

Here is their chart, widely reported in the popular press,
For starters, the linear correlation is a piss poor fit to the data points

 I had several problems with this study. First, the report had a very thin section on methodology. If one looks at the ten variables, one sees they vary widely in frequency from several parts per hundred thousand (suicides range from about 14 to 2 per 100,000 population in various states and homicides from about 10 to 0.5) while other categories are measured in less than a part per million (mass shootings) so may occur rarely. Police shootings were only reported for one year. It would appear to me that all of these variables are weighted equally even when they are rare or poorly known events and this is done for subjective and politicized reasons. We know that gun crime and violence are not homogeneous on the state level. Illinois, for example, has a few homicides per 100k but some parts of Chicago have rates, given current trends, likely exceeding an order of magnitude higher than that while other parts have rates approaching zero (see figure). Crime must be looked at on a scale defined by the problem, not by arbitrary units convenient to advocates. Finally, the authors give lip service to correlation not being causation but nonetheless implicitly state that it is by basing recommendations for gun control on their study.
Homicide by police district, from Wikipedia

So with these caveats, I tried to look at simpler questions in the CAP study to see if strong gun laws make us safer. Here are some charts. Click on them to see in full scale. To make these, I used the fatality rates in the CAP report, converted the Smart Gun Laws rankings of A to F (where F is a weak set of laws and A is a strong set, obviously subjective and defined by a gun control advocacy organization) to a zero to 4 ranking on a plus vs. minus scale, i.e. A- = 3.7, B+ = 3.3, etc, and used a Kalesan study of gun ownership as reported in Business Insider for gun ownership rates.

Looking at homicides, homicides show no correlation with Smart Gun Law rankings or frequency of gun ownership (data taken from the CAP report except where otherwise noted). I'm not surprised at the lack of a gun correlation because with over 300 million guns in America and about 10000 gun homicides per year (round numbers), one would expect about one in 30,000 guns to be used in a homicide if each gun only caused one homicide. Such small numbers are lost in the weeds. The lack of a correlation with gun law strictness (as defined by LCPGV) suggests these laws are ineffective at curbing gun homicide.  Further, one would surmise that restrictive gun laws might stop criminals from buying guns on the legitimate market but as I recently posted on this blog, a survey of jailed criminals, when asked in a recent study,  said they obtained guns on the black market or stole them.

Law Center gun law strictness ranking vs homicide rate
Gun ownership in a state vs. homicide rate
Looking at gun ownership and laws, one sees less gun ownership in states with strong gun laws. Not sure this means people are deterred from buying guns by the hassle and red tape of gun laws or if it is easier to pass onerous gun laws in states where fewer own guns. New York State, for example.
Law Center To Prevent Gun Violence rankings vs. gun ownership
 There is a correlation between suicide rate and gun ownership and therefore, restrictive gun laws. Suicide by gun is pretty much guaranteed, so if one is suicidal and armed, there is a pretty good chance of success should the urge overwhelm someone. As Dr. Dan Nolan said on, there are states like Alaska where gun ownership is high but so are factors contributing to suicide (seasonal affective disorder and a strong asymmetry of male vs. female population) so the tendency to suicide may be high and there is a gun present to get it done. But clearly gun ownership does not control suicide as Alaska and Hawaii have high ownership rates (62 and 45 percent, respectively) and differ by close to an order of magnitude in suicide rate (14 and 2 per 100k population, respectively). For that matter, Japan's suicide rate is 50% higher than ours in the U.S. in a virtually gun-free society. I don't know much about Alaska but worked in Honolulu for 14 years. Hawaii has its social ills but also has a strong sense of ohana which may give people in despair resources other than a bullet to the brain, thus ohana, as well as a latitude at the tropics rather than nearer the North Pole and thus without huge seasonal swings in daylight, may have more control over suicides than gun availability.
Gun ownership per state vs. suicide
As far as other variables, I have not looked at all of them but did some preliminary calculations on how one could come up with a strong correlation by lumping all these variables together, since one is counting the same thing over again in some cases (total gun deaths, suicides and homicides) and weighting extremely rare events equally with more common ones. For example, there is  a low correlation between gun ownership and domestic homicide  of women.
Gun ownership vs. domestic homicides

But if we count homicides and domestic homicides of women per 100,000 we get no correlation as domestic homicide of women is a rarer event compared to homicide.
homicides plus domestic homicides,both per 100k, vs. gun ownership in a state

If we weight domestic homicides by 10x as the study authors may have done, i.e., add domestic violence of women gun homicide in parts per million as an equal variable to homicides in parts per 100k and add suicides, we get this, which is a correlation approaching that reported by CAP.
Combined fatality (homicides, 10x domestic homicide, plus suicide) vs gun law rank

I think its worth continuing to try to figure out what CAP did but  in spite of a request for information, it has not been forthcoming. I think CAP proved that strong gun laws are present in states with fewer guns and therefore fewer gun suicides. The study also confirmed that stronger gun laws don't prevent gun homicide, which could potentially be a good proxy for gun crime.
Unemployment v Suicide. From Huffington Post
Guns in the US from Azreal et al, preprint (2016)
Suicides are a major problem but a mental health, economic, and social problem, not a crime problem. Studies that purport to make Americans safer from gun violence need to point out the difference and direct crime reduction at crime and suicide reduction at suicide. Furthermore, cutting edge work by Andrew Papachristos and other sociologists gives us a better idea of who is committing violent crime and why. As Professor Papachristos' web site tells us, "...Most recently, Papachristos was awarded an NSF Early CAREER award to examine how violence spreads through high-risk social networks in four cities. He is also currently involved in the evaluation and implementation of several violence reduction strategies, most notably the Project Safe Neighborhoods and the Group Violence Reduction Strategy in Chicago..."  Such studies and directed efforts, including the New Mexicans To Prevent Gun Violence work with youth in Santa Fe in pledging to renounce gun violence, need to be integrated into violence prevention strategies rather than simply looking at suspiciously defined correlation analysis that may tell us nothing about causation or prevention.

Added over the weekend. Here are some more plots out of the CAP report. Sources of data in are in the CAP report. Most of these correlations are quite weak.

Traces of crime guns to origin of sale

Law Enforcement shootings of civilians

Law enforcement officers killed in action

CAP uses a USA Today source on mass shootings, in which 4 or more people are shot regardless of reason. So this conflates gang shootings, drug shootouts, family violence, and random acts

I summed up all of the above variables to get this correlation.

Anti Trafficking laws aimed at so called gun exporting states (see graph above) are worth looking at if these are written wisely rather than with the typical birdshot approach; the CAP report shows a strong correlation between gun laws and sources of guns traced to crime. Maybe there is room for compromise here in enacting a national standard for gun sales tracking to cut down on trafficking. To some degree, the paranoia on a national registry and tracking system is well deserved but its not clear to me how one cuts down on illicit sales without some ability to figure out how guns get from W. Virginia to New York. But first we need a better estimation of how many guns are trafficked between state lines vs. how many are stolen or acquired by other means. Let's not have a solution in search of a  problem.

But some current "common sense laws" being proposed are nothing of the sort. The Bloomberg-written ballot initiatives voted on in Maine and Nevada could, with some provisions, increase gun violence. For example, suicide is the major cause of gun deaths. Right now if a close friend told me he was suicidal and asked me to lock up his guns, I could do so. With the provisions of those initiatives, he would have to legally transfer ownership of them to me via a Federally licensed gun dealer (FFL) or it would be a criminal offense.  Then, when his urge passed, I would have to legally transfer the guns back to him at an FFL. I suspect such provisions would strongly discourage such cooperative, potentially life saving initiatives. Go figure. These are the sorts of provisions offered by those who know little about the firearms community that build walls rather than break them down.

Finally, from Dr. John Lott

"The Times incorrectly describes a study as showing that gun control laws reduce violence (“Gun-Control Groups Push Growing Evidence That Laws Reduce Violence,” October 11th). But the study — by the Center for American Progress — never examines how rates of violence change before and after gun laws are adopted. The study simply compares violence rates across states. Yet, states have a lot of differences beyond what gun control laws are on their books. The Center for American Progress does not account for any factors such as differences in law enforcement (e.g., arrest rates), demographics, and income. For example, rural states have higher suicide rates partially because the male/female ratio is so out of balance.
More useful academic research follows states over time to see how rates of violence change with the adoption of different laws. These changes are then compared to the states that did not change their laws."

John R. Lott, Jr., Ph.D.
Crime Prevention Research Center

Lott makes suggestions which are quite good but even comparative before/after studies are frought with difficulty as economic, social, and demographic changes are occurring, i.e., there is no real controlled experiment. As Lott would probably agree if I could ask him, there is no such thing as a properly controlled experiment as far as changes in law are concerned. Stay tuned.

No comments: