Did fifteen inches of grading, crushed some dreams, and now there's time for some rambling, motivated by the recent tragedy in Connecticut and BD's question "How do you regulate crazy".
A few points to start off with:
1. The difference between a country where mass shootings are rare, and one where they are (relatively) common, boils down to whether 1 person in 10,000,000 snaps, or 10 in 10,000,000 snaps. Even in someplace supposedly ultraviolent - Juarez, for example - the newsworthy killings are carried out a tiny number of outliers, perhaps 1 person in 100,000.
2. On the other hand, when measuring social, medical, or criminal events that are common, like voting, heart attacks, or ordinary one-on-one murders, it is extremely difficult to measure the effect of contributing factors to an accuracy of even 1%. Simply taking larger and larger samples doesn't help, the problem is the the large number of uncontrollable factors involved.
3. The fact that one can't measure the contributing factors very accurately doesn't mean they aren't there, and significant. For example, take suicide rates versus time:
Those trends are not statistical noise. Given the huge number of events, and the huge population pool, statistical noise should not even be visible on the scale of the graph. Something was definitely happening, and of course one can correlate it with the economy, war, etc. But a social scientist would be very hard pressed to predict whether a given future change in the economy or a proposed war would cause a 5% versus a 6% change in the suicide rate. There are just too many variables.
So, someone snaps, and happens to have at hand the right weapon and the right target, and just as importantly, they happen to snap in just that rare, precise way that removes their inhibitions on violence but leaves intact their ability to formulate a plan and follow through on it.
Naturally we want to do something to prevent the next one. And maybe we should, but my issue comes up when the proposed solution to an outlier problem is a general rule that will apply to everyone, or to some very large group that happens to include the outliers of interest.
I'll stay off guns for now because the issue has too much baggage attached. Let's say, instead, that the proposed rule is something relatively mild: on the standardized academic tests that every 8th grader takes, we'll mix in a dozen or so questions about feelings of anger, shame, etc. No follow up, no mandatory treatment, no required intervention, no reassignment. Just an "awareness" measure so the principal will know if any kids scored 9+ out of 12, and can be on the lookout for other warning signs. The proposers are optimistic and think it might preempt one shooter out of 10. Actually, it doesn't matter what the details of the plan are. The plan could be to change the lunch trays from purple to pink. If you do this - or anything really - to 20,000,000 school kids, you can't be sure that it doesn't cause (for example) a 0.1% lifetime increase in some other relatively common, less newsworthy form of violence, e.g. suicide. The mechanism could be that the test questions bring successfully suppressed emotions back to the surface, or lead to subtle changes in how the teacher treats the whole class, or that pink becomes associated with shooting in people's minds. Or it could equally well be that the test questions and the colored trays actually decrease the suicide rate. The point is, you don't know and an effect that size can't be measured.
But a change of 0.1% in the suicide rate is 25 deaths a year, enough to completely swamp the 10% chance your test questions might stop a 20 victim shooting. If your intent was to minimize violent deaths, you don't know whether you helped or hurt. And no amount of record keeping or stats will tell you.
Does this mean do nothing? Not at all. When the proposed action only applies to the outliers, or to a small group that includes them, the unmeasurable effects scale down as well. For example, if one changes the rules on how police respond to an in-progress shooting , the rules only apply rarely, and even a 10% chance that the change might make the cops shoot the wrong person might be worth it if you judge these things solely by body count.
Does this mean never address problems through general laws? Not that either. If a generally applicable gun law (for example) is likely to have an effect on regular murders, of which we have 10's of thousands, the expected benefit may be many times the size of any unmeasurable consequence, again only judging by body count.
My point is that it almost never makes sense to address a problem affecting 20 or even 200 people by creating a new law or process for 300,000,000 people, no matter how innocuous it seems.
And now for a bit of partisan ranting. This habit of smacking outliers with a rule affecting everyone is not limited to crime, and it seems to be a tendency particularly of the left. School shooters are the issue of the week, but blanket policy change seems to be the preferred solution to every problem. Something on the order of 1% of the population freeloads at the emergency room, and this justifies regulating the health care of every person in America. Some tiny fraction of 1% makes meth out of cold medicine, and everyone is limited to buying only a few days supply.
In general, the Precautionary Principle, applied to wrongdoing, is not compatible with a free society. But even if one cares nothing for liberty, when the problem being addressed is tiny compared to the unmeasurable consequences of the solution, it doesn't even make sense numerically.