On Rationality and War


War is a negative sum game: the victors often are worse off than they would have been if the war could have been avoided, and even when in some sense they gain, almost invariably they gain less than the losers lose. The expected value of engaging in war is sharply negative. Moreover, prospects in war are famously clouded by a fog. No one can be confident of victory, historical examples are legion of aggressors who were confident they would triumph, and yet went down to defeat. One can rattle off dozens without effort. It follows, essentially as a theorem, that the substantial majority of historic decisions to war have been irrational. The question is, why are leaders, and nations, so irrational, and what can be done about it.

Decision makers consistently overestimate the expected cost of choosing peace and also consistently underestimate the expected cost of war. The gains of war over peace typically depend on contrafactuals. If we don't fight this war, something bad will happen. But in cases where one can check, these fears are often exaggerated. For example, we fought the Iraq war in large measure to prevent Saddam from acquiring nuclear weapons, yet that turned out not to be a prospect. We fought the Vietnam war in large measure because of fears of the domino effect, but that turned out much less far reaching, and much less important strategically than feared, even when we went down to a vastly greater defeat than could have been imagined.*. Had we walked away in 1964, or even simply left precipitously in 1968 or 9, its hard to imagine our enemies could have been more emboldened than in 1973, yet few dominoes fell, and to little strategic import in the context of the cold war. In retrospect the conceivable gains, had we won, were not comparable to the investment we made.

The case for war routinely cites appeasement, backing down from war, as promoting aggression. But almost the only historical example, where avoiding a war evidently incurred substantial cost, is Chamberlain and Hitler.

(Please supply other examples, if you can.)
By contrast, examples where war turned out badly are legion.

The costs of war are routinely radically underestimated. Would Bush have fought in Iraq if he'd known the war would cost a trillion $? Would we have fought in Vietnam if we'd expected it would suck up 9% of GDP for years ( citation ombwatch.org) and kill 58,000 Americans? But even these figures don't capture the costs, because there are numerous hidden costs. A big part of the problem is that economists have been remiss in not addressing the wider costs of war, for example its impact on the economy.

Every sustained war the US has ever fought, since 1776, has featured serious inflation. This is hardly surprising, we print money to pay for war, creating no goods for the money to chase. But the effects of the printing and the inflation spread beyond the war. And the economic impact alone frequently far overshadows any plausible strategic gain.

The Iraq war, for example, has featured a run up in the price of gold by well over a factor of three. The Fed printed money in part to pay for the war, holding interest rates too low, and the housing/banking crisis resulted. Its unfair to lay the blame for this entirely on the war, but it is also unrealistic to dismiss the connection. (Its also an interesting coincidence that the last banking/housing crisis followed the last Iraq war.).

Nixon was elected in 1968, and chose to pursue Peace with Honor, as opposed to rapid withdrawal. From the 1968 election, through May 1970, the Dow dropped 36%. The economic effects expanded well into the 1970's, as inflation proved hard to reign in. In the strategic context of our overall conflict with Soviet Union, the economic costs, even just of continuing to pursue victory in Vietnam after 1968, were huge, and as discussed above, we know in retrospect that the potential strategic gains, had we won, were small. We eventually won the cold war mainly by outproducing the USSR, so its plausible this blunder extended the cold war a decade. And note, the USSR repeated the same error in Afghanistan.

This does not exhaust hidden costs. Wars suck up political capital and divide the nation. Bush, for example, felt constrained not to veto Republican spending bills. Would he have chosen war if he'd known it would swallow his Presidency and bring down his party, even if he knew it would have the various benefits it could be argued to have had, and (to be optimistic) that he'd ultimately win it?

Wars impose opportunity cost. If Bush hadn't surrendered his political capital, he might have accomplished something else. He went to war to keep nukes out of the hands of terrorist states, yet the preoocupation with Iraq has perhaps tied his hands in Iran, leading toward exactly that which was feared.

Yet, such is the nature of irrationality, that conservatives argue to this day the Iraq war, and even the Vietnam war, were worthwhile. Bret Stephen's 7/8/08 column in the WSJ lauded Nixon's Secret Plan which kept us in the war, and conservative friends of mine defend it to this day.

For such stubborn people, let me reemphasize my opening paragraph. People will argue about any given war, essentially making up contrafactuals that justify it. One can't say for sure what would have happened had one not fought. But you should be suspicious of this, because it is essentially a theorem, that averaged over all wars, those contrafactuals don't usually happen-- the sum of results of all wars is obviously substantially negative.

Even celebrated war choices like Lincoln's in the civil war are quite questionable. Would Lincoln have chosen war if he'd known 700,000 people would die and the massive economic cost? And he could easily have lost: for example, all it would have taken would be to lose the next election, which was a narrow thing, and hardly predictable. Wouldn't that have turned out much much worse than never fighting? Did he correctly take the expected cost of that possibility into account in his decision process? Might it not have been, if he had chosen peace, that slavery would vanish shortly after in the natural scheme of things, as it did everywhere else in the world? It's even conceivable the North and South would have reunited, much stronger and richer than they were in the event. We'll never know. The purported benefits depend on contrafactuals, and are uncertain, but many of the costs are evident and massive.

Where does this bias to war come from?

Where does this bias to war come from? Perhaps it is instinctual. It seems likely people were evolved for war. For one thing, Chimps fight wars in much the same way as hunter-gatherers. And you can see the selection for war-seeking in action among the Yanamomo: males who have killed in war have 2.5 times as many wives and three times as many children as men who haven't. In fact, as people have been domesticated, ie over the last few thousands of years, we've become much less warlike. If we killed each other in war at the same rate now as the Yanamomo, 2 Billion people would have died in warfare the 20th century, instead of merely several tens of millions.[N. Wade, Before the Dawn, pp140-152]

A big part of it, I suspect, is simply the general deficiencies of our mental computations. People don't intuitively understand negative sum environments. They never take the fact that they are participating in a negative sum game sufficiently into account, whether in war or in the lottery or in proposing government solutions to problems (for example, they dramatically underconsider the dead weight of taxation). In general, people have a big problem with the seen and the unseen. One sees that the Iraq war increased liberty in Iraq, but we forget the cost of something like a million lifetimes of forced labor ($1 trillion taxes) in the US.

People also routinely underestimate the cost and duration of all projects. Perhaps we are evolved for this: or we'd be too discouraged to start important ones. This is probably positive for constructive, positive sum activities. When this bias meets negative sum environment, however, bad things happen.

Also, although the "fog of war" is a famous concept, that decision makers generally try to take into account, people's thought processes are inherently built to try to understand the world in terms of causal models-- we are built to believe we can predict things that we can not, even, in laboratory experiments, things that unbeknownst to us are truly random. So we are too confident about the benefits of war. Likewise, in evaluating the alternative, peace, we are much too confident. We never consider that the possibility that the problem we are worried about (eg Saddam's nukes, or very hypothetically, the south's cessession) might cure itself.

Another big part of it, as discussed above, is that economists have just not adequately considered the economic impact of war. We have run serious inflation in every sustained war we've ever fought since 1776. A war sucks off resources, produces nothing, sucks up the President's political capital, and the economy goes into the tank. Presidents, and voters, considering war need to hear advice from economists: if you start this war, and it drags on as many do, you will have inflation and and a severe impact on the economy. Some wars justify incurring those costs, but are the costs of backing off from this war really so big as to qualify? Purely in strategic terms relative to war itself, the economic consequences often outweigh the potential gains.

Mitigating Factors

Finally, two mitigating factors against the picture that people irrationally overestimate the case for war.

(1) Part of the observation may be explained as a selection effect on the data. One might suspect there are numerous wars that were never fought, where people estimated accurately or even underestimated. The cases we see, are only the fraction where decision makers overestimated the case.

(2) Engaging in a few negative value wars might be rational as a bluff or deterrent. Wars are rarely positive expected value, but a nation that was perceived to only engaged in positive expected value wars could be subject to being pushed around.

My answer to these is, they but don't color the picture much. For the first, it doesn't relieve us of the goal improving our decision process in the wars that occur. For the second, it's not even clear how a nation's engaging in wars affects the perception of its being a patsy. The immediate aftermath of our engagement in Vietnam, for example, was clearly that we were much less likely to respond to aggression, and our adversaries could not have been fooled about this. Almost surely our engagement in Iraq has made us less likely or able to respond to provocations elsewhere, and our adversaries perceive this too. And, as mentioned above, with the exception of Chamberlain and Hitler, its hard to come up with convincing historical examples of wars anyone should have fought but didn't. And both Chamberlain's inability to deter Hitler (and reluctance to fight) and our inability to deter Iran from pursuing nukes (and reluctance to fight about it) might on the contrary be laid to exhaustion from previous choices to war. So arguably, engaging in wars is often not only not a deterrent to other aggressors, but on the contrary an active source of encouragement to them.

The Moral of the Story: a Remedial Methodology

The situation is reminiscent of the famous Oil Well Lease Problem, which runs as follows. An oil lease is auctioned. 10 companies each send their scientists to look it over, and form the best assessment of the prospects. They each then take their estimate of the value of the lease, cut it by some fraction, say 20% or whatever, to represent their cost of capital etc, a fraction determined by the best expert opinion of the risk, opportunity cost, etc, and put in their bid.

Result: the great majority of the time, whoever wins the auction, loses money.

Reason: the amount of oil to be extracted, the price it will be sold at, etc., is not exactly predictable, so opinions differ. Whichever oil company overvalues the lease the most, wins the auction.

Thus it is necessary for them to adjust their best estimates by a large fudge factor. They were wrong to think the best available opinions, however scientifically based, were right. Conditioned on winning the auction, the best scientific opinions were strongly biased upwards.

Likewise, given the persistent historical undervaluing of the costs of war and overvaluing of its benefits, decision makers need to apply a much larger fudge factor to their decisions than they have been, if they want to behave rationally.

*Arguably, the US retreat from Vietnam in 1973 was also irrational. I haven't studied the issue, but it seems very plausible that the costs going forward here may have likely been much less than the gains from continuing to support the Vietnamese. This assertion doesn't affect the observed fact that the strategic costs of withdrawal turned out to be less than had been feared, but does provide another possible example where a more aggressive policy was warranted, and the fact that this occurred is another warning about the fog of war: decisions to engage in a war need to take into account the possibility that some years later, having almost won the war at huge expense, domestic decision processes will simply decide to abandon it.


Eric Baum
Last modified: Mon Jan 5 21:01:48 EST 2009