The Lincoln Plawg

Sunday, February 9, 2003 11:50 a.m.

Random insight into War Party warped thinking

[This piece is for the moment here on account of the main blog being down.]

Josh Chafetz at Oxblog has a laudatory piece on an article by Victor Davis Hanson which castigates the moaning minnies and dismal jimmies who are warning of the possibility that an Iraq war might not be a cakewalk.

In support, Chafetz has the following:

The military always tries to figure out the worst case scenario casualty figures, and then opponents of military action jump all over those figures. But that's really stupid -- no one should ever plan anything based simply on the worst case scenario. The only sensible way to plan is based on a calculus: consider each potential outcome, weight it by the likelihood of its actually occurring, and add them. Then do the same on the other side of the balance. If the weighted calculus of casualties is worse than the weighted calculus of not fighting, then don't fight. If vice-versa, then do. Obviously, no one knows what the numbers or weights are -- we can only make guesses, and we can argue about the assumptions underlying those guesses. But it's just plain foolish to pretend that only the worst case scenario should be taken into account.

First and last clauses, I don't dispute. The rest, as the product of so distinguished an academic institution, appears to your more humbly educated correspondent to leave a good deal to be desired.

The 'Chafetz method' fails to take into account the vital question of the utility of the various estimates of casualties. A notion of which even a mathematical near-illiterate like me has heard!

Simple case: on the Chafetz method, a 100% chance of 1,000 US KIAs is equivalent to a 1% chance of 100,000 [1]. It just plain isn't so!

Looked at from the perspective of US politicians looking to save their own asses, as much as from that of the loved ones of the prospective dead US soldiers, the numbers do not figure. They assume (against all evidence) that the parties interested [1] are risk-neutral. Whereas I believe (and have certainly seen no evidence to the contrary) that they are risk-averse. No doubt, the way in which, say, a military commander weighs up the risk of action will be different to that of the parents of one of his men.

But I hypothesise that they would all believe that a 1% chance of 100,000 deaths was, by a factor of several, worse than a 100% chance of 1,000 [2].

There are, of course, two components to this: the consequence-adjusted risk may be greater than the multiply-through risk as an objective matter. For instance, suppose I have a net worth of $1m; a bookie offers me odds of 10-1 against for a $100 bet that a fair coin will land heads. I take it: the multiply-through risk of the bet is $50 (there's a 50% chance of the coin landing tails), and the potential reward is much greater. No-brainer.

Suppose he offers the same odds, but only for a $1m bet. According to the Chaftez method, it's still a no-brainer - the relative multiply-through risks haven't changed. But, in this case, if I lose, I'm wiped out. No sane man would take that bet in those circumstances - the percentages are the same, but the consequences of losing are too great even if the bookie was offering 1,000 to 1.

But then, there's also the psychological factor. I'm no more an expert in psychology than mathematics. But, on the evidence that I have, people may, depending on the circumstances, assign a risk to an event which is higher or lower than the consequence-adjusted risk.

Example: transport. The risk of dying in a car accident is much greater than dying in a plane or train. Yet the perceived risk is lower. The explanation I've heard given is that, in a car, you, or someone you know, is responsible for your safety. Whereas with a plane or train, it's strangers.

(A similar way of thinking applies to the risk from immunisation of children.)

On war casualties, my guess would be that there is some kind of scale, based on past experience, that guides the average Joe on these matters. Perhaps at three levels: (1) World War 2, (2) Korea/Vietnam, and (3) Gulf War. Failing a fiasco such as that in Somalia, the casualties in a war in category (3) are below the political radar. The psychological difference between (3) and (2) is Lyndon Johnson's March 1968 shall not seek and will not accept broadcast [4].

Which leads on to the point that, in the real world, these risks tend to be lumpy, rather than the smooth, continuous function posited by the Chafetz method. We are in the realm of the price point and the tipping point. One side of a line (of varying width) is OK, the other is not-OK.

It's a binary question: just like whether George Bush should be dumped as GOP candidate in 2004. Because the ultimate outcome is binary (he can't be half-dumped!), the risk is assessed as binary[5].

To put it bluntly, a thousand US dead in Iraq would be compatible with Bush's retaining a substantial chance of reelection; 100,000 would not be.

And, whereas Blair's protestations of being happy to stand alone might be credible, I don't figure Bush for political martyrdom.

Now, as I say, I'm no expert. But it seems to me that there is a genuine complexity here which it serves the interest of the War Party to skate over.

I'm sure that these issues have all been extensively dealt with in the academic literature. I am, as ever, happy to yield to a balance of cogent evidence that I have, in fact, got the whole thing utterly wrong. Bring on the evidence......

Call the risks in each case as so calculated multiply-through risks. (I'm sure there is a technical term, but...)
Call the value placed on this risk the consequence-adjusted risk.
Most in the US would, I suspect, know that Vietnam KIAs were about 50,000: how many are as confident in estimating Gulf War KIAs?
And Korea cost Truman a chance at reelection in 1952.
Or ternary: the range of Bush poll ratings below which he will be dumped; above which he will not; and an indeterminate third band in between. Binary or ternary, the risk function is certainly not continuous.