[dirtboy]

Suppose, for example, that CCW significantly lowers crime if it is already high, but that its effect on crime diminishes as the crime rate goes down.

I would believe that CCW would have exactly such an effect.

[traditionalist]

Yes, well, you are probably right. I suppose that's going to be the next step in CCW research: coming up with an appropriate non-linear specification. If you've studied econometrics, however, you will know that such specifications are very tricky and often lead to suspect results because there really is no way to know that your specification is the right one.

Here's my guess as to what will happen: some specifications will show that CCW decreases crime, others will show that it increases it, others will show that it has no effect, and all will be equally plausible.

That's why I think Gary Kleck has the right approach to the gun control debate. Trying to figure out whether gun rights (CCW as well as others) on net increase or decrease crime is a fruitless excercise, so arguments should instead be based on principle.