Hacker News Reader

You don't need to invoke social narratives and consistent behaviors to justify honoring sunk costs.

Sunk costs are often a predictor of a developed position.

For example, you spend 20 years in advancing in field and then worry that you might not like it that much anymore.

If you avoid leaving the field purely because of the sunk time, you are honoring a sunk cost.

If you avoid leaving the field because your 20 years have made you good enough at it that your pay is good and your job stability is assured... and meanwhile you would have a hard time getting hired in a new preferred field or keeping a job there because your expected pay would be unaligned with your experience there (see also: ageism)... then that is just playing the hand you are currently holding.

Usually your current position isn't completely clear. You don't know precisely to what degree 20 years experience have made you better suited to your past field than another.

Under the reasonable assumption that your past costs weren't entirely wasted and aren't perfectly fungible to different moves it is reasonable to use your past costs as a predictor of your current position. 20 year experience in a field is probably worth something, and probably worth a lot more than 5 years.

This is a good point and it was going to make it in my original post!

I was going to say that the $1,000,000 price (as opposed to $100,000) tag on a piece of land will tend to be indicative of the value of the piece of land, assuming you weren't in a drunken stupor when you paid a million bucks for it. In other words, most decisions in life aren't bets made in a vacuum.

The inference though is that the million dollar land _was_ worth that much because it was next to a (now-condemned) shopping centre, and the cheaper land was next to nothing; the new subdivision means the values have changed dramatically. Extreme example, but the point is the price paid is irrelevant if circumstances render it so

It is indicative of the value of the land but says very little about whether you should or in what way you should choose to develop either plot. The surrounding economy, zoning and geography will have far more to say about that.

This is true. However, suppose that all you knew about two plots of land was that one was bought for $1m and one was bought for $100k. Then, someone asks you which one you should develop.

Buying land isn't like betting on black. Land has a history, markets are (more or less) efficient, and buyers are usually rational. The sunken cost fallacy suggests the answer is a coin-toss, but in the real world, the correct answer would (usually) be the expensive plot.

I think this comes down to correlation versus causation.

The fact that someone (perhaps yourself) valued a property highly recently is correlated with that property having a high present value. But it does not cause it.

It is possible the million dollar purchase was a mistake, or that something has changed and it's now worth far less. Valuing a property highly may cause you to have paid a lot of money for it, and that might be a reason to assess the current value highly. But that's very, very different from valuing a properly highly because you paid a lot of money for it.

For sunk costs, you are supposed to consider your all options including the sunk cost project.

For your example your options might look like:

* Stick with industry, 0 year lead time, no cost, possible sadness, low risk

* Slight change, 2 year lead time, $20,000, moderate happiness, medium risk

* Vast change, 10 year lead time, $100,000, unknown happiness, high risk

The fallacy would be giving the first option some sort of financial value because you spend time and money on it in the past. You are just supposed to look at your options looking forward only.

> You are just supposed to look at your options looking forward only.

In most realistic scenarios your estimations of the payoff matrix has a significant uncertainty. It isn't just that there is risk, but your estimation of the risk is uncertain as well (as well as your estimation of your estimation, and so on).

When reasoning under uncertainty we can usually achieve significant benefits from regularizing the decision.

"Do what everyone else is doing", "Keep doing what I was already doing", and "Do what is most consistent with my past investments" are time tested highly effective regularizers. When we are trying to rationalize taking actions that defy billions of years of evolved heuristics for reasoning under uncertainty we call the first 'bandwagon fallacy', the second 'status quo bias', and the third 'sunk cost fallacy'.

There is often a fine line between rational decision making and rationalization. Awareness of the ways that people sometimes make errors in their decisions can be useful, but one should take care to avoid using a little bit of knowledge to come up with specious justifications for poor choices.

Much of the time I see the word 'fallacy' used it sure seems to be sophistry. When a reasoned position is better you can just state why its better outright and the justification will stand up on its own merit without any invocation of a named fallacy.

Man I cannot agree at all. I've personally seen some majorly bad decisions made on the backs of those fallacies.

And I've see it work out brilliantly

I've seen some people get killed after leaving their homes. Better stay inside all the time.

You can't make bias go away just by speaking its name.

Maybe calling it sunk cost heuristic would be better? Acknowledging that most of the time it does work, and then explaining why in this particular situation it does not?

> The fallacy would be giving the first option some sort of financial value because you spend time and money on it in the past.

But when we have sunk costs, we really have a past investment that we have an unknown return upon-- and we're deciding whether to abandon that investment.

If we've made a substantial investment, it can make sense to have a bias towards avoiding actions that definitely invalidate that investment-- a bias towards inaction.

It's rare that a position presents itself where it is completely clear what the future value of different tracks is worth so clearly.

In this case you're merely using history to make a relevant choice about the future. You're not honoring the sunk cost because of the investment, you're just using the lessons your learnt while making this investment.

You are basically just playing with the words, but you also seem to contradict yourself: > you're just using the lessons your learnt while making this investment

That's exactly what honoring sunk costs mean. Making the investment means the cost factor. And using the lessons your learnt means honoring.

It seems to me like one simply should incorporate social costs into the calculation.

If your wife was looking forward to the concert and doesn't give one whit about modern economic theories like sunk cost, her disappoinment at not following through with attending the concert tonight is not a sunk cost, that's a future cost which you should weigh.

I think you're conflating what the paper author calls "binding" and "betting" -- "binding-type" decisions (like the one in the concert example) seem to be candidates for honoring sunken costs. Betting, on the other hand, does not.

No. I understand exactly what the author means. The paper is quite simple and clear. I just think it's a meaningless distinction. Rather than try to create a two tiered decision making process where some decisions should include sunk costs and others shouldn't, you should have a single tiered system that includes social costs.

You don't need to think in sunk cost terms to value plausible deniability. You should view losing credibility with your superiors and peers as a current cost that runs into the future.

If for example you suddenly realize your thesis paper is junk and are faced with decision to push forward or start anew it's not sunk cost to consider that your academic superiors might see this as flailing about and that you should count that as a cost of changing direction. That's just a social cost and not a time or materials cost.

That's being victim of your peers falling for the sunk cost fallacy

No. It's realizing your peers can induce a cost to you in the future whether they are acting from a rational PoV or not and incorporating that cost to your calculation.

Yep, that’s a future cost of the decision to not go to the concert, not a sunken cost like the cost of the ticket. It’s not relevant to the discussion of sunken costs.

The point is the same cost can be sunken and future cost too. It's not binary, and your implied false dichotomy (either a future cost or a sunken cost) is wrong

The dichotomy does seem to apply perfectly well. The cost of the ticket is sunken. The social cost of skipping the concert is not.

imagine every instant in your life is an infinitesimally small point on a timeline (or every millisecond if something concrete is easier). the present is t0.

an event occurs (buy ticket) at tX. according to "ignore sunk costs", at tX+delta you should annihilate all influence from any event at tX or before. therefore that you have a concert ticket is, in a weird way, almost immaterial. it enables a future decision (to "freely" go to a concert, as you would any other activity) but, my interpretation is that you only have present and future after every instant moment.

you may say: well what about eliminating future opportunities? e.g. i spent my last $250 on concert ticket so can't go to hawaii. that is in part of a matrix i would call "regret" or counterfactuals (not sure if there's an economic/scientific term here), which involves thinking about how past decisions affected future ones.

I have implied no false dichotomy. I haven't said one must accept either or. I've said simply one can account for the costs implied by the OP without rejecting the sunk cost falacy.

If it helps you to think in sunk costs that's probably fine. I don't care. My point was purely that one can incorporate the kinds of concerns the OP was proposing without rejecting the "sunk cost falacy".

I'm not sure I'm following. In the example, there's a sunk cost of dressing up and setting aside time. Then, other than sunk cost, there's the probable future cost of reduced cooperation as well as cost in rapport and morale. I don't see those things as sunk cost because the decision to go or not go can influence those costs.

The idea is this: the "rational" thing would be to sell the tickets for $500 a pop. However, even if there's just a probability of cooperation and social rapport suffering†, it wouldn't be irrational to honor the sunk cost and still attend the concert.

In the paper, the "Camping Rainstorm" example is similar in spirit. Instead of the protagonist suffering what the author calls a "diachronic misfortune," maybe in having no plans for the rest of the night, maybe in suffering a cooperation loss vis-a-vis the spouse, etc., they honor the sunken cost.

† I'd probably argue it's guaranteed.

I think you and the blog post author are making the same mistake. You're just not putting a dollar value on these social costs. If the blog author valued their partner's feeling they would enumerate upsetting them as a cost and it wouldn't be so clear that you're earning $500. But rather you're earning $500 less the cost to your personal relationship.

This is a similar mistake to failing to price other intangibles like risk.

It’s an odd mistake to make, since it implicitly recognizes that someone may value the experience of seeing the concert more than $500, but fails to recognize that someone may value their close social relationships more than $500.

It’s not rational to choose receiving $500 and having your wife and friends be upset at you, unless you value the $500 more than having your wife and friends be on good terms with you. That’s a tautology, of course, but apparently it needs to be said.

This is the most correct answer I’ve seen in this thread. Not all costs are dollar prices.

Pissing off your spouse is a pertinent new cost of bailing from the concert, not a sunk cost. The ticket price is sunk, because you paid it in the past and can not change it.

If the ticket is fully refundable then it is also not a sunk cost. The land example also has potentially the same issue if the land can be sold for the purchase price. If the land can be sold but at a 20% discount then the sunk cost is only the 20%.

Of course the world is complex. This stuff makes sense unless it doesn’t. That $10,000 piece of land may require capital budget to utilize that you don’t have, for example.

I ran into the concert example when it was announced a couple of years ago that David Wright would take the field for the N.Y. Mets for the last time. My $12 tickets were suddenly worth $300 or more. Rationally, it would make sense to pocket a few thousand bucks. But the experience had a certain value, and I wouldn’t be able to get my kids and nieces and nephews together, in NY, etc for a year or more. So we had an amazing time and missed an economic coup.

The way I see it, you didn't miss the coup. You got a $300 experience for $12.

I'm monetary terms, you got the exact same outcome as if you'd have sold the tickets.

> but not honoring them might hinder current (or future) cooperation.

Then you're not honoring sunk cost. You're honoring current or future costs, just like you should do.

> In Seth's concert example, this is particularly relevant: your spouse might be a huge Springsteen fan, for instance.

Your example misses the whole point of OP's demonstration of why sunk costs make no sense. OP's demonstration focuses on a single metric (cash), a single focused investment (time spent searching for the tickets) and a single critical change that exposes how the sunken cost fallacy negatively impacts decision-making (a random person offers you 10x the cash for your tickets).

The example illustrates that the amount of resources invested in the project in the past should not be a factor in the decision-making process because it leads to poor, irrational decisions. That's the core message of the sunken cost fallacy.

Your example fails to reproduce the problem because it switches the focus to an entirely different metric (utility) and frames the problem in a way that the new metric is actually negatively impacted by the change, thus the rationale decision is to not sell. That says nothing about the sunken cost fallacy because poor decisions are not alternatives to good decisions.

> These are all technically "sunken costs"

No, they're not, they're additional factors to be considered in determining the value of the tickets to you, as compared to the value of the $500 you could get for them. The article ignores such factors, presumably for simplicity of exposition, but that doesn't mean they're not relevant. The correct rule is still to ignore sunk costs, because those are in the past; but the correct rule by no means says you should ignore relevant future impacts of your choice (like your spouse being upset because you sold the concert tickets) just because they aren't monetary.