Sometimes we know we don't know

Being involved in our neighborhood association for the past few years has brought home to me just how poor some of data collection processes are in terms of some quite serious issues.  Crime is the easy example.  We have crime data from the Atlanta Police Department which shows increased crime but at a relatively low level.  Why?  Because they only show outcomes resulting in an arrest or a report of some obvious crime (car stolen, for example).

Then we have sources like Citizen (reports of incidents reported on the police dispatch system.)  Or Neighbors which are resident generated incident reports.  These numbers are much higher than APD's partly because it includes unverified crime but primarily because many people are reluctant to report to the police but are more comfortable reporting it on a social media app.  Also, APDs protocols preclude many crimes being reported.  For example, I am out on a morning walk and see a car with its window smashed in and robbed.  I call it in to APD but they won't dispatch on it because I am neither the vehicle owner nor did I witness the window smashing.  A crime has been committed but it is not in the numbers.

Then there is NextDoor, often dominated by Karens but still pretty useful because it covers things that slip through the APD cracks.  A homeless man stands on a corner, lunging and shouting at passersby.  He slips away before the police arrive.  No crime to report because the police weren't there to see it.  But there was an attempted crime non-the-less.  

So if you are interested in the real world prevalence of crime, what numbers do you use?  Do a survey?  APD numbers?  Citizen?  Neighbors?  NextDoor?  A synthesis of all of them together?

None of them is reliable or complete on its own.  Yet without good crime numbers, how do we allocate and manage scarce policing manpower?

Many social issues have the Achille's heel.  The problem is real.  We want to solve it.  The causal factors are ambiguous.  We don't have a holistic measure either of dimensions of the issue nor the process.  

This issue of consequential social issue undermine by poor measurement mechanisms is explored in There Are Far More Defensive Gun Uses Than Murders in America. Here's Why You Rarely Hear of Them. by John R. Lott Jr..

We have a reasonably good handle on the number deaths by murder and deaths by suicide per year.  What we are missing are the number of attempted crimes?  How many crimes were attempted but not completed?  And of those which were attempted, how many were abandoned based on victim response (physical attack, knife, gun, etc.)?  

There is a philosophical argument for citizen ownership of guns as represented in the Second Amendment and that is to ensure that the individual citizen can and does represent a constraint on government.  You can accept or reject the argument, but it is a valid argument.

But there is a pragmatic argument as well.  Does gun ownership reduce attempted crime?  Here is where the data is critical and so often absent.

These are just a few of the nearly 1,000 instances reported by the media so far this year in which gun owners have stopped mass shootings and other murderous acts, saving countless lives. And crime experts say such high-profile cases represent only a small fraction of the instances in which guns are used defensively. But the data are unclear, for a number of reasons, and this has political ramifications because it seems to undercut the claims of gun rights advocates that they need to possess firearms for personal protection -- an issue now before the Supreme Court.

Americans who look only at the daily headlines would be surprised to learn that, according to academic estimates, defensive gun uses — including instances when guns are simply shown to deter a crime — are four to five times more common than gun crimes, and far more frequent than the fewer than 20,000 murders each year, with or without a gun. But even when they prevent mass public shootings, defensive uses rarely get national news coverage. Those living in major news markets such as New York City, Washington, D.C., and Los Angeles are unlikely to hear of such stories.

Lott then addresses other issues in the measurement of victim deterrence of criminals.  He then addresses what is known.  

The U.S. Department of Justice’s National Crime Victimization Survey indicates that around 100,000 defensive gun uses occur each year -- an estimate that, though it may seem like a lot, is actually much lower than 17 other surveys. They find between 760,000 defensive handgun uses and 3.6 million defensive uses of any type of gun per year, with an average of about 2 million.

The difference between these surveys arises from the screening questions. The National Crime Victimization Survey first asks a person if they have been a victim of a crime. Only respondents who answer “yes” are asked if they have ever used a gun defensively.

In contrast, the other surveys screen respondents by asking if they have been threatened with violence. That produces more self-acknowledged defensive gun users, since someone who successfully brandished a gun is less likely to self-characterize as a crime victim. Survey data indicate that in 95% of cases when people use guns defensively, they merely show the gun to make the criminal back off. Such defensive gun uses rarely make the news, though a few do.

About 75% of all (~20,000) murders per year are committed with guns.  Obviously a number we would want to reduce.

But how?  The obvious, instinctual response is gun control.  Keep guns out of the hands of citizens.  

But H.L. Mencken's 1917 observation kicks in.

There is always a well-known solution to every human problem — neat, plausible, and wrong.

Gun control, aside from its constitutionality and 2nd Amendment implications, is a solution, but is it a viable solution.

Does it reduce murders?  The evidence is unclear in the US whether that is the case.  What is rarely addressed is what Lott is addressing here.  Guns in the hands of criminals are used to commit murders.  But guns in the hands of law abiding citizens avert crime and death as well.

If a well armed populace suffers 15,000 gun murders a year, we do have to set that in the context of the number of murders which they may be used to avert.  That is hard to know.  Using the 2 million number used above, one might argue that guns in the hands of responsible citizens helps reduce murders.  But by how much?  If citizens use guns 2 million times a year to deter a crime being committed, how many of those might have been murders?  We don't know.  If all of them were averted murder attempts, then 2 million.  If 10%, 200,000.  If 1%, then 20,000.  

We don't have the data to be confident in the number.  It is not implausible that it is 1% or more, in which case, armed citizens may actually be an effective means of reducing gun deaths.  

But we don't know.  Just as with Covid-19 and just as with most other consequential social issues, our data definitions and data collection mechanisms are simply not up to answering the questions we most want answered.  Arguments then revert to quasi-religious convictions rather than actual evidence.