Bryan Caplan echoes my own pragmatic/consequentialist arguments for pacifism :

P1: The immediate costs of war are clearly awful.
P2: The long-run benefits of war are highly uncertain.
P3: For a war to be morally justified, its long-run benefits have to be substantially larger than its short-run costs.
C: [B]efore you kill innocent people, you should be reasonably sure that your action will have very good consequences.

Tyler Cowen responds with a series of counterexamples:

There is not enough consideration of specific times and place.  Had England been pacifist in 1914, that might have yielded a better outcome.  Had England been pacifist in 1939, likely not.  Switzerland has done better for itself, and likely for the world, by being ready to fight back.  Pacifism today could quite possibly doom Taiwan, Israel, large parts of India (from both Pakistan and internal dissent), any government threatened by civil war (who would end up ruling Saudi Arabia and how quickly?), and I predict we would see a larger-scale African tyrant arise, gobbling up non-resisting pacifist neighbors.  Would China request the vassalage of any countries, besides Taiwan that is?  Would Russia “request” Georgia and the Baltics?  Would West Germany have survived?

Interestingly, John Dewey famously called the World Wars wrong using a similar standard, supporting the first and opposing the second. Decision-making under uncertainty is hard! Still, I think Cowen is being a bit unfair here. This parade of horribles could likely be run through Caplan’s Common Sense Pacifism test: we should ask for each whether there was “reasonable certainty” that the war would have “very good consequences.” Cowen claims a commonsensically pacifist world would be a worse world, but I think he’s attacking an absolute pacifism, not Caplan’s common sense version. Some now-well-regarded wars would fail the test, and the current geopolitical world would be differently aligned, but I don’t think Caplan is arguing that Georgia ought to give in to Russia.  For one thing, I think self-defense against the incursion of neighbors is likely going to pass this test (as opposed to “self-defense” like the “preemptive” invasion of Iraq.) Take India: does its use of violence to repel Pakistan pass Caplan’s test? Arguably, yes: the hostilities along the border are relatively cheap and harm few civilians, but these measures have substantial long-run economic and political benefits. (Although note Amartya Sen’s arguments against India developing nuclear weapons.)

UPDATE: Another way of reading Caplan is that the burden of proof ought to fall on the warmongers rather than the pacifists, and that the standard of proof ought to be “reasonable certainty” of “very good consequences.” This standard may or may not already underwrite the various Just War theorists, but perhaps Caplan is simply noting that many of the older standards and precedents may have changed in light of new ways of making war that increase the costs and thus raise the estimation required for “very good consequences.”