[iguy]

There is a neat paper I can't find which runs the numbers. So from memory:

For about 1000 years, we know that west-european states executed about 1% of each generation. Assume it's the most violent 1%, and that violence is 50% heritable. Plugging these into standard formulas for animal breeding, and I believe the answer is that this explains about half the decline in the murder rate over this period.

Obviously there are some giant caveats here. But it's super-interesting that a back-of-the-envelope calculation gets you the right order of magnitude.

I'm dubious though of wars doing much. The big 20th C meat-grinder was largely conscripts. And anyway I'm not sure how correlated with blood-thirst most recruiting was. Whereas getting yourself hanged for murder in peacetime was pretty tightly correlated.