Why are outlier events not a reason to take action?

Question

What is meant by "politicizing" an event and how to avoid it?

In an answer there, someone recommends that one should not use outlier events to recommend action on gun control policy. He says specifically that since use of automatic weapons in mass shootings is rare, this is not a reason to make a policy change on that matter.

But, uhm... why? Is this not straight up normalcy bias?

That is, just because "normally" people, up until now, do not use automatic weapons for mass shootings, then that must mean they'll never do it, hence it'll never be a problem, hence all policy changes are moot?

Indeed, the point of a policy change is to PRECEDE a time where such attacks become more regular, such that we can prevent such a time in the first place.

Answer 1

Outliers

Outliers are, by definition, contrary to the totality of our experience. In quantitative data, that's what makes an observation an outlier - it is sufficiently different from the rest of the data that it becomes labelled abnormal.

You wouldn't want to use an abnormal occurrence to create public policy. That would mean you targeted your policy to a situation that you know doesn't normally reflect the problem.

Let's re-use the example from the linked answer, which claims:

this was the only mass murder in US history that used automatic guns

Assuming this is true, if you wanted to create a public policy to limit the risk of mass murder, then you would want to base your policy on what is generally true of mass murders. Outlier events necessarily do not reflect what is generally true, so they aren't very useful.

Rare Events

Now, perhaps what you are describing is more of a rare event, rather than outlier. Rare events can be amazingly informative observations.

A rare event is just an event which does not occur very often. Rare events are not outliers, because they are entirely normal in their own domain. For example, for the purposes of modelling a single insurance policy a tornado is a rare event (the odds of any piece of property being destroyed by a tornado is exceedingly rare). It isn't an outlier because it is expected that some property is destroyed by tornadoes (that is, the rare-event conforms to the totality of our experience).

To re-use the same example, if your goal was to create a public policy which reduced the risk of mass murder using automatic weapons then you would want to consider past experience of mass murders which use automatic weapons (even though they don't reflect the general truth of mass murder). These events are rare, so you would probably want to focus your policy on the few past events you can find.

Answer 2

The idea that outlier events shouldn't be used is based on the concept that you don't want to make general rules based on very specific events. For instance, imagine trying to write laws that cover Enron. There were many interlocking pieces, and in isolation any one of them may have a legal and justifiable use. That's why the government didn't try to prove any given act was illegal, but instead tried to wrap the whole operation up as a fraud.

I could try to explain, by wikipedia covers the more common phrasing Hard cases make bad law

(I'm not writing about the specific event or whether the term was properly used. I'm explaining what the phrase means)

Answer 3

Indeed, the point of a policy change is to PRECEDE a time where such attacks become more regular, such that we can prevent such a time in the first place.

But you also say,

In an answer there, someone recommends that one should not use outlier events to recommend action on gun control policy.

By their nature, outlier events happen in the past. You can't prevent past events by making changes in the present, only future events.

The basic argument is that we shouldn't use one single event (the outlier) to determine policy. Instead, we should base our policies on a sober evaluation of the possibilities of events. More people died in the 2016 Nice Attack, which used a truck as a weapon. By outlier logic, we should ban all trucks. After all, trucks are much easier to obtain than automatic weapons.

One way of evaluating the possibility of future events is to look at what events have occurred in the past. Since 1934, access to fully automatic weapons has been quite limited. Even more so since 1986. It is noteworthy that this is the first time such a weapon has ever been used in such a crime. We don't know what went wrong in the system such that this person obtained a weapon that is both expensive and difficult to get. Presumably we will know more later.

There is no reason to think that this is the first of many attacks at this point. There are quite a few reasons to think that this was an exceptional occurrence that is unlikely to be repeated. Of course, it is possible that the investigation will uncover additional information that contradicts that view. If so, we should evaluate it when it is available.

Answer 4

Outlier events, by definition, can't be managed by changes to the normal process. Normal process will have impact on things that are withing the "normal" statistical boundaries.

Just about any business process improvement model (TQM, Six Sigma, etc) is based on this concept. If an event falls outside of the norms for a process or system, then you look at the special circumstances that led to it, and deal with that event on an individual basis.

On the other hand, paying specific, individual attention to an event that is outside what is considered acceptable, but inside the normal boundaries of a process is also a waste of time, because those events will continue to occur until the process is changed/improved/tightened.

Manufacturing examples - Company "A" has technical specifications on a ball bearing sizes for their widgets. They are having problems where wrong-sized balls are being installed and product is returned. Their manufacturing process is not well-designed, so even though these are considered "fail" events, they are part of the normal process, so seeking out and firing the workers on duty when the wrong-sized balls were made or installed would not address the problem. They need to look at the entire process and correct it.

Company "B" makes the same widgets with the same specifications. Anything within 1/4" of the diameter specification is acceptable. Their process is so well-designed that their normal process is to within 1/8" of the spec. In this case, company "B" can identify and deal with individual issues (supplier put wrong barcodes on boxes causing 3/16" variance) where they identify an outlier and deal with it, even though it is within acceptable specification and does not cause any kind of failure.

I only pointed that out to illustrate that "problem" <> outlier, and "no problem" <> "normal," necessarily.

Just because an event is spectacular, unusual, odd or infrequent does not necessarily make it an "outlier" event. It may, indeed, be something that could easily occur repeatedly within the system, as set up. If that's the case, it isn't technically an outlier.

Rolling this all back into the discussion of recent news events - if the events were enabled by actions that, for the most part, were allowed or enabled by the normal system that is in place, then it's not an outlier. This would include, say, where a law is in place to stop something, but there are other laws that create loopholes, gaps, or hamper enforcement of those laws, because that process would allow for a certain amount of expected circumvention of the letter of the law.

If the events are caused by actions that fall outside of what should be possible in the normal process, and a number of exceptional circumstances come into play, then trying to "fix" the normal process that did not cause the problem will not address that problem or help in preventing future occurrences.

Answer 5

Addressing an outlier event can be expensive, both in terms of money and political capital. Is it worth the money, and the legislative time, and the favor trading that has to go on, to enact legislation to avoid a recurrence? Is it worth that expense, when considering the lives that could be saved if that money and political capital were expended in other needy areas, such as relief efforts or boosting education and economic opportunities in economically depressed areas? There's only so much money and political capital available.

Let's consider two such 'outliers':

In the aftermath of the Oklahoma City bombing, which killed over 150 people, regulations were put in place to track large purchases of ammonium nitrate, the material used in the truck bomb. Probably a wise move, there hasn't been a large scale bombing in the US since then.

And now, the Las Vegas shooting, which was a worst case scenario: a huge crowd, and one person with automatic weapons. Early reports are that the shooter used 'bump stocks'... this is a stock on a semiautomatic rifle with a spring that lets the gun shake back and forth. Hold it just right, your finger bumps into the trigger, and it acts like a fully automatic weapon.

Of what use is a 'bump stock'? Absolutely none, other than cheap thrills. The result will be terribly inaccurate, which isn't a factor if you're firing into a large crowd, but otherwise, utterly useless. In this particular case, it shouldn't be a problem to outlaw 'bump stocks', and 'trigger cranks', a crank device that works a rifle trigger rapidly... also legal at this time.

On a larger scale, the US does have an issue with the proliferation of dirt cheap assault style rifles. Today, one can buy a mass produced AR-15 rifle for $500, about the same price of an original Colt AR-15 rifle, back in 1972. Way too cheap, and that particular weapon has now been used in four recent mass shootings: Sandy Hook, Orlando, San Bernadino, and now Las Vegas.

The problem there is - any attempt to restrict them just fuels more purchases - sales of assault rifles soared in the Obama years, out of just fear of restrictions, not attempts to actually implement restrictions. Personally, I think future manufacture should be stopped. That would send the price of existing ones way high, making them harder to obtain. But, this has to be done soon, if it is to be done at all.

And that won't cover the other danger exposed by this tragedy: The effects of stampeding a large crowd of people.

Note that the casualties from Las Vegas haven't been separated into gunshot deaths, and deaths resulting from the huge crowd panicking and trampling people. Consider the Who concert, where a mistake by officials at the stadium resulted in eleven people dying by being trampled. And that wasn't even a panicked crowd, it was just a huge crowd funneled into too small a space. Four or five bombs, set off one after another, would have the same effect on a large crowd, so firearms restrictions won't stop that from happening again, and the outlier of stampeding a large crowd still exists, as well as the terrorist value in preying on large crowds.

That danger still exists. And there is no simple solution for that. Sadly, the tech that makes our lives better, also makes building remote control bombs easier, from remote detonators to drones to carry small bombs into a large event.