Iraq and the sunk cost fallacy

Is the sunk cost fallacy really a fallacy? That's the question raised by this:

Demands for the withdrawal of troops from Iraq intensified as the death toll of British soldiers hit 100.

My immediate reaction was: this is illogical. The matter of how many soldiers have died in Iraq already is irrelevant to the question of whether troops should stay in the country. Logically, that should depend only upon the future costs and benefits of them staying. The costs we've already incurred - the sunk costs - are irrelevant. Nothing we can do now will bring those soldiers back. To think otherwise is to commit the sunk cost fallacy.
However, several  things make me think my first reaction was too glib.
1. Sunk costs can matter, to the extent that they help us predict future costs. A rational anti-war protestor can argue that the deathtoll so far is an argument for withdrawal insofar as it shows that staying in Iraq will prove to be costly.   
2. Honoring sunk costs can be a form of pre-commitment. It may be irrational in isolation, but it's part of a rational strategy overall. As Tyler Cowen said:

We follow through on sunk costs as a kind of self-discipline, to prevent ourselves from initiating too many stupid undertakings in the future. If you self-signal that you will follow through on your commitments, you will be more careful in accepting commitments in the first place.

3. Acting apparently irrationally can deter future terrorists and aggressors. In signalling that we will incur any cost to defeat insurgents/terrorists - even a greater cost than the isolated benefit would warrant - we show that we cannot be influenced by terrorism. That should deter terrorist acts, just as no-one picks a fight with the local psycho. Again, what looks irrational as an isolated act might be rational as part of a wider strategy.
4. Instrumental rationality is not the only form of rationality, as Robert Nozick argued in his best book, many acts have symbolic meaning - they signal who we are. Staying in Iraq - honoring the sunk cost of the 100 deaths - symbolizes that we are a country that seeks to give meaning to the deaths of our soldiers, and that doesn't give up when the going gets tough.
I don't know whether these considerations are a case for withdrawal or not. To me, that's not important. What is important is that our - or at least my - snap judgments about what's rational or not are sometimes inadequate.

Comments

[kharris]

Getting past the question of whether issues of war and peace are ever decided based on logic, let's think about this further. Is this sunk cost rule the only issue we need to address in assessing the logic of counting up our dead when deciding whether to stay or leave?

Each war is new. It is not a repetition of prior wars. We learn about the cost of winning our objectives are we go. Perhaps we are learning that our objectives are not as attainable as we thought. Certainly, relative to early blustering about open arms and flowers, or about a few dead-enders, or about turning the corner, or about strong nationalistic tendencies, what we have learned is that stabilizing Iraq is going to be brutally expensive, for Iraqis and occupiers, if it is possible at all.

So maybe counting up our dead, so far, is a pretty important exercise in cost benefit analysis for this war, or any war.

Point 3 here needs to be chewed over, as well. In saying we might want to act in apparent irrational manner now, to convince our enemies we will do so in the future, is to say that we ought to let more US soldiers die than this particular project seems to merit. And more Iraqis, too. Getting Iraq the way our leaders want it is going to cost more US and Iraqi lives than they bargained for. More than Iraqis agreed to. Easy to say when your life, and the life of your kid, are not on the line.

Nixon is thought to have liked the notion of appearing irrational to his enemies, but he failed to win his war, and that loss seems to have fostered backlash in the US against international intervention.

It is also not clear that grinding away in a country not associated with terrorst acts against the US, while the Taliban reassert themselves in Afghanistan, and bin Laden runs around making tapes at will, is going to deter terrorism against the US.

Your statement treats this as a problem in decision making and nothing more. It isn't important to you, you write, whether these arguments support withdrawal. What is important is to realize that snap judgements sometimes prove wrong. I'm willing to bet that most folks who read this already knew that snap judgements could be wrong. I want to ask what the point of sharpening one's judgement skills is, if we don't then use those skills to try to make better judgements about issues just like the one presented here.

[dsquared]

I don't think it's true that "demand for troops to withdraw intensified". The same bunch of usual suspects on both sides made their typical talking points, but this time they got in the papers because a round number is as good a peg as any to hang a story on (this point [c] Thomas Schelling).

The real problem in Iraq is the "non-existent objective function fallacy". We are apparently going to stay in Iraq "until we have seen it through" or "until the job is done", without any objective criteria being set for what would constitute having seen it through or doing the job. We've even explicitly rejected any suggested standards like "until the Iraqis ask us to leave" or "until we have x number of domestic troops reaching x standard of preparedness". This was always the problem from the moment we went in; if you have no idea what you are doing, you are unlikely to do it properly.

[dearieme]

1. "Sunk costs can matter, to the extent that they help us predict future costs." But a better predictor would nullify that.
2. "Honoring sunk costs can be a form of pre-commitment. It may be irrational in isolation, but it's part of a rational strategy overall." With presidents of the calibre of Clinton and Bush the Younger, this is no more than a pious hope.
3. "Acting apparently irrationally can deter future terrorists and aggressors. In signalling that we will incur any cost to defeat insurgents/terrorists - even a greater cost than the isolated benefit would warrant - we show that we cannot be influenced by terrorism." From the UK point of view, the IRA-placating Blair can't deter anyone. And the record of US support for the IRA - terrorists against the USA's main ally - makes that argument a joke in poor taste.
4. "Instrumental rationality is not the only form of rationality, as Robert Nozick argued in his best book, many acts have symbolic meaning - they signal who we are. Staying in Iraq - honoring the sunk cost of the 100 deaths - symbolizes that we are a country that seeks to give meaning to the deaths of our soldiers, and that doesn't give up when the going gets tough." See the IRA comment above. Or read Fromkin's "Peace to End All Peace": we could have honoured the dead of the (British) Indian Army in Iraq, in and after the First World War, by behaving more rationally this time round i.e. by declining to join W in his rash and foolish adventure.

[AJE]

Rationality as defined by immediate material self-interest (which is pretty much the neoclassical position) is indeed flawed, and we should expect behavioural - and anecdotal - evidence to falsify it.

As an Austrian, my instincts are for a "purposefulness" which is very similar to rationality but includes the subjectivist assumption that people have different utility functions - hence it's hard to define peoples actions as irrational.

Vernon Smith's Nobel lecture (http://nobelprize.org/economics/laureates/2002/smith-lecture.pdf) very nicely states the case for "ecological rationality", and you should all check it out. To me, it clears up these sorts of confusions/uneasiness.

[rjw]

Seems to me that saying that people may behave differently just because they have different utility functions is both true and monumentally unhelpful. It could also be true that they have the same utility functions but one acts rationally and the other does not. Now - distinguishing the cases is surely tricky. But the conceptual distinction is surely helpful. If nothing else the framework of the rational actor model (preferences, constrains, behaviour) allows us to explore the ways in which a normative model of rationality can be subverted.

[rjw]

'constraints' not 'constrains' of course...

[david]

Amazon says the perfect partner for this Nozick is Freakonomics. Yikes.