Sunk Costs

by Brian on March 15, 2004

I had thought that the idea that the sunk costs fallacy is really a fallcy was as close a thing as there was to a consensus amongst philosophers. But now I see that “Tom Kelly”: has a paper forthcoming (in __No{u^}s__) saying that “honouring sunk costs can be rational”: Like a few other people in the New England area, whenever I think of the sunk costs fallacy I think of the Red Sox continuing to play “Tony Clark”: long after he showed he wouldn’t justify his $5,000,000 salary, so I’m not exactly positively disposed to the fallacy. But Tom’s paper makes several interesting points, even if it doesn’t do anything to redeem the Sox management __circa__ 2002.

Strictly speaking, Tom isn’t as much concerned to defend the rationality of honouring sunk costs as the possible rationality of actions that are usually so described. He quibbles a bit in various places throughout the paper about whether there are as many clear experimental cases of the sunk costs fallacy, but let’s assume for now these are instances of the fallacy.

The main idea (roughly stated) is that since the value of an action is partially determined by what happens in the future (just like “the value of an organism”: our current actions can be sometimes justified by the redemptive value they confer on past actions. It’s an interesting idea, though I’m not sure how much it should matter in practice. For one thing, I think we need a more comprehensive theory than Tom offers here about which past actions are worth honouring. (I imagine Tom has such a theory but space constraints kept it out of the _No{u^}s_ paper.) It clearly isn’t worth redeeming an off-season waiver claim by running Tony Clark out there every day when it’s really unlikely he’ll hit the ball out of the infield, let alone out of the park. Tom often refers to the kind of actions that are worthy of redemption as ‘sacrifices’, and I wonder if there’s more to be said about what makes those actions redemption worthy.

Tom also notes that honouring sunk costs, or at least being perceived to do so, can have game-theoretic advantages in certain situations. I’m less impressed by this as an argument for the rationality of such actions. (And Tom doesn’t lean on it particularly.) In some games of Chicken, the best thing to do is to unbolt the steering wheel and throw it out the window. The situations where it is best to honour sunk costs remind me of those games of Chicken. When it works, it’s a neat stunt, but it doesn’t take much for circumstances to change and then it becomes a really really _bad_ strategy.

Obviously there’s a lot more to the paper than I can indicate in a blog post, so if you’re interested in these topics, er, “read the whole thing”:



rea 03.15.04 at 6:58 pm

Why on earth did the Red Sox contract to pay Tony Clark $5 million, after he was waived by the Detroit Tigers, of all teams? Why didn’t they instead hang on to Scott Hatteburg, for example?


Brian Weatherson 03.15.04 at 8:13 pm

He wasn’t waived outright by the Tigers. His contract was offered up to anyone who was interested, and given his pre-2002 performance it looked like a good deal for $5 mill. I remember being excited by it at the time.

But wow did it turn out badly. I think it was mostly an injured back. From casual observation he still played decent D, and there was no suggestion he was giving less than 100% effort, but he just had no bat power.

I remember Bill Simmons writing about how when Clark came up to bat he’d cheer for a strikeout because the only other live alternative was grounding into a double play. It was that bad.


John Quiggin 03.15.04 at 8:40 pm

Brian, I would have thought there was something close to a 1-1 mapping between arguments for/against paying attention to sunk costs and arguments for/against consequentialism. Since there are lots of people opposed to consequentialism, there should be lots in favour of honouring sunk costs.

Turning to Kelly, from your summary, he really seems to be saying that in many cases, costs that appear to be sunk are not, for example, because they are in invesment in reputational capital. Economists have been all over this kind of argument, taking all possible views, for a couple of decades – the locus classicus is the chain store paradox.


Brian Weatherson 03.15.04 at 11:37 pm

Tom’s not just making the point about reputation. He notes that, and says it can explain some of the experimental data, but he really means to be defending the stronger claim that it can be OK to do X because it makes some past action Y make sense.

I was worrying when I wrote that about whethe all arguments against consequentialism (understood as a decision theoretic rather than ethical doctrine) were arguments for the non-fallaciousness of sunk costs reasoning. I wasn’t entirely convinced because I thought there was a difference between (a) choosing a course of action and sticking to it through a decision tree and (b) using sunk costs reasoning to make a decision at later stages. Many anti-consequentialists recommend (a) at various times, but few recommend (b). Few enough that there’s a consensus against sunk costs reasoning _by philosophy standards_. Since we all never agree on anything, these are very _very_ low.


Matt Weiner 03.16.04 at 12:09 am

Wouldn’t a lot of arguments for one-boxing in the Newcomb problem translate into arguments for the sunk cost fallacy? Perhaps one-boxing is a marginal position as well.

Incidentally, I find this link a more perspicuous explanation of the chain-store paradox; go to #2. It seems a lot like a case David Gauthier discusses in “Assure and Threaten.” I’m not sure whether following through on an unsuccessful threat counts as an example of the sunk cost fallacy–the decision to make a threat isn’t the same as the decision to make a threat and follow through on it.


rea 03.16.04 at 6:43 pm

“His contract was offered up to anyone who was interested, and given his pre-2002 performance it looked like a good deal for $5 mill. I remember being excited by it at the time.

“But wow did it turn out badly. I think it was mostly an injured back.”

Brian, what I as a Tiger fan know (and you Red Sox guys evidently failed to note)is that he had major back problems in ’00 and ’01. That, of course, was why the Tigers got rid of him. The days when he was arguably worth $5 million were 3 years in the past when the Sox decided to acquire him.

Comments on this entry are closed.