Saturday, June 03, 2017

Scientist Responds to Claim that John Lott is Responsible for Mass Murder

A writer at Pacific Standard, Peter Moskowitz, assigned a measure of responsibility for a few of the latest mass public shootings ( Orlando Pulse Jihadi attack, Columbine, Sandy Hook, Fort Hood Jihadi attack, Navy Yard shooting, Aurora Theater shooting, and Charleston Church shooting) to John Lott. Why?

It appears to be because John Lott conducts research that shows that more legally armed citizens reduce violent crime, and that gun free zones attract mass shootings.

 Moskowitz admits that he is ill prepared to judge Lott's work. From
I will not be able to debunk Lott here and now. I am not an academic. I—and 99 percent of people, I'd venture to guess—am not as good with statistics as Lott. What I can tell you is what the people who do have that skillset say. There are many people who agree with Lott—especially in the fields of criminology and economics. But it appears the majority of researchers who work in the field say Lott's wrong: that his analyses are misleading, that they skew data to favor certain outcomes, and that his research methods don't stand up to scrutiny. If they did, his critics say, Lott would still be published in academic journals, or doing his research out of a university instead of a non-profit called the Crime Prevention Research Center.
Every person is faced with the problem of confirmation bias when determining how to evaluate various claims. People selectively prefer claims that confirm their perception on disputed subjects. Scientists are trained to fight this tendency. It was with interest that I read another scientist' evaluation of Lott's work. It was posted in a series of comments on Mosqowitz' article.

 The commenter with the screen name of Dan N agreed to the publishing of his comments. Daniel sent me an updated copy. The arguments and data presented are essentially the same:

This article is written in reaction to a piece published on June 1, 2017 by Peter Moskowitz entitled “Inside the mind of America’s Favorite Gun Researcher”. To give the reader some perspective, I never fired a firearm until I was 36 years old and did a year of research before bringing one into my house. I was firmly on the fence about owning a gun until I did this research. I took it seriously and did not give much credence to opinions that were not thoroughly supported by statistics. I have a Ph.D. in a science that requires me to be at conversant, however not fluent, in statistics. I take the frustrating stance that people need to validate what they read. If the statistics are too confusing, as Moskowitz says, then you need to learn some basic statistics, not just trust what someone else tells you.

It is clear that Moskowitz of has not thoroughly read the source material. I have read Lott's updated "More Guns, Less Crime" (3rd edition, 2013) and some of Hemenway’s reports and other leading anti-gun reports. 
I cannot comment on the original 1997 publication; however Lott has updated it extensively with new data and did additional permutations in response to the criticisms of Hemenway and others. In some of Lott’s responses to criticisms he added additional permutations, the results are more dramatic. In other permutations, it's less. 
Lott gives a tremendous amount of detail on how he handles the data sets. None of it appears outlandish. It is mostly for mathematical purposes. 
For example, he can't compare a county gaining right to carry in 2006 that had 0 murders in 2005 and 1 in 2006, as that is essentially an infinite increase. Likewise, 1 murder in 2005 and 0 in 2006 it would be a 100% decrease. He admits that the two things that do increase are property crimes and aggravated assaults. 
Such changes are rationale and justifiable, but this type of data filtering is exactly what Lott is often criticized for. For anyone who has read the actual source material, Lott makes no obscene claims. Effects are quite modest. However, a modest 5% difference in homicide rate is all the difference in the world to someone able to live to tell about it.

Mr. Moskowitz made some errors in his critique. For example, he mentions the white urban men owning guns and the permits while urban minorities are often the perpetrators and victims of crimes, but this criticism does not hold water. 
Lott polled over 3,000 counties. The trends he calculates most of his work on are within counties, not within states. He goes to extreme length to specify this, and it is something the author has miscategorized. 
As far as his critics are concerned, Ayres and Donohue published a piece entitled “Shooting down the more guns, less crime hypothesis.” Despite their criticisms of Lott’s data analysis, they have faults of their own. For example, the authors do not use defined periods before and after the passage of concealed carry laws and compare that to violent crime rates as Lott does. Instead, states are lumped together in the same time frame where one state could have passed RTC early and another late. Furthermore, anti-gun papers rarely conclude that guns make crime rates any worse when viewed at a national level. At best they claim that Lott's numbers are more dramatic than reality. For a better idea of errors, please refer to Lott’s own work as his dissection of methodological errors is superior to anything I’ll write here.

Please remember that Lott has done 20,000 regression analysis in his reports. Yes, some of them will be wrong. If you are reporting at 95% confidence (meaning a 5% error rate, common in statistics) you are looking at 1,000 of his 20,000 measurements wrong. Cherry picking the studies which do not pan out over the years, as is frequently done to discredit Lott, does not disprove the data set in its totality. 
Using the Ayres paper again as a reference, they criticize Lott’s use of arrest rates as a mechanism of judging crime levels. While it is true that this is not perfect, Lott uses multiple comparisons and collectively concludes the benefits of gun ownership from the totality of evidence, not just single regressions. I've personally downloaded and done my own longitudinal studies from publicly available FBI reports ( I've come to the same conclusion as Lott with independent data collection.

I publish heavily in peer reviewed journals, some of the top journals in the world in fact (nothing to do with criminology). I know editors of these journals. That world is, unfortunately, extremely liberal and highly politicized. I have viewed the Ivory Tower of Academia from within and without, and I assure you it can be an ugly place. It is so liberal and politicized, my wife has insisted I do not bring up my hobby of trap shooting around academics because she thinks it will negatively impact my career. In fact, my name here is a nom de plume. While I cannot comment on the specifics of Lott’s exit from academia, it does not surprise me in the least.

And now for a quick peek inside the world of statistics. It will delve a bit off topic, but the point is to illuminate how the author can pull the wool over your eyes. Part of what I do for a living as a scientist is to analyze data sets. If I ever published data in the manner these gun control groups academics did, my career would be over. It can be trivial for me to debunk them and is a guilty pleasure to do so. 
Few, if any, do longitudinal studies. By this I mean that they do not study a region before or after an event (passage of RTC). Far to many are taking data sets from a single year. A particularly egregious error was made by Kalesan and Galea published in the very prestigious Lancet last year. They claim that ballistic fingerprinting/microstamping to be one of the most efficacious laws on the books. However, California simply passed a law mandating microstamping but not a single firearm is commercially available capable of microstamping. Maryland’s ballistic fingerprinting program was a colossal waste of taxpayer’s money and police efforts failing to solve a single case in 15 years. So how can Kalesan and Galea find them to be such influential laws? 
It simply means that there are variables beyond that law that dictate crime rates in those states. Many arguments disintegrate once suicide is separated from homicide. Having a best friend that killed himself with his father’s gun, I find this dishonesty especially infuriating. For example, one cannot compare the high gun ownership of Alaska with the low ownership of Hawaii and claim any causality with suicide. Many in Alaska suffer from depression due to the lack of sunlight for several months and there are far fewer women than men, Hawaii suffers from neither of these problems. I cannot recall anyone, other than Lott, distinguish legal gun ownership from illegal gun ownership despite the fact that as much as 80% of homicides are gang and drug related. I've critiqued numerous top studies, and I've never found any that come close to the comprehensiveness of Lott's.

Hopefully this encourages whoever reads this to dig into primary data for themselves and not to put blind trust into pixels on a screen, print on a page, or a talking head on a screen.

I have read Lott's books, and quite a few of his papers. I am a trained scientist, (retired out of the DOD). I find the statistical methods used intimidating, but understandable.

One thing stands out: No one is credibly claiming that violent crime is increased by the increase in people legally carrying firearms. You see this reflected in the arguments used to oppose gun reform legislation. Instead of people claiming that a reform of restrictions on carry will increase murder and robbery, you see more and more of the argument shifted to the claim that a reform of gun law will "not make us safer". Some make the unsurprising claim that more guns will lead to more suicides committed with guns.

It is not a persuasive argument. Our society is based on the model that if it is not prohibited, it is allowed, instead of the totalitarian model that if it is not allowed, it is prohibited. While the rate of suicide with guns has risen, it has risen less than the rate of all suicides.

None of the people pushing for more controls are using the county level of data to analyze the effects of guns on crime. My suspicion is when data is analyzed at the level of counties, the anti-rights researchers are not getting the results that they want.

It is credible that Dan N is not interested in having his understanding of the debate over the effects of legal firearms carry published in his name. His wife fears for his career. It is a real fear, based on real potential consequences.

I was downgraded in my career for a number of years, based on my willingness to push for less restrictions on firearms use. Two separate supervisors both told me that was the reason I was downgraded, in separate conversations, years apart. They were sympathetic, and appreciated that I did not hold them responsible. As I recall, one of them volunteered the information, the other responded to a carefully worded neutral question.

As a white male who had no reasonable claim to membership in a specially protected group, I judged it best to forgo legal challenges.

There can be a price for integrity and/or activism. It is wise to pick your battles and chose your favored field of action. I do not regret my decisions.

Consider what Peter Moskowitz is preaching. Because John Lott has published work that John reasonably believes to be true, John Lott is judged to be complicit in mass murder. It is hard to see the causal connection. It seems that John Lott's real "crime" is disagreeing with Peter Moskowitz, and challenging his world view. 

©2017 by Dean Weingarten: Permission to share is granted when this notice and link are included.

Gun Watch

No comments: