Bryan Caplan  

When Do Hypotheticals Cover Their Cost?

Tyler Cowen versus Frederic Ba... It's a Busy Week in the Carden...
Suppose I asked, "Where would you buy steaks if you only shopped at stores starting with the letter Q?"  A few people would wrack their brains for an answer.  But most would dismiss the question: "That will never happen, so who cares?!"

Economically speaking, the popular reaction seems perfectly sensible.  The cost of devising a contingency plan has little to do with the probability of the contingency.  The expected benefits of devising a contingency plan, in contrast, heavily depend on the probability of the contingency.  So when someone confronts you with an extremely unlikely hypothetical, spurning the question is usually the prudent course. 

This seems like an awkward conclusion for me.  I habitually propose remote hypotheticals.  If you could save either one American or x foreigners, how big would x have to be before you'd save the foreigners?  If a million farmers were stuck in Antarctica, would allowing them to migrate to a more favorable climate enrich non-Antarcticans?  If six trillion Nazis were sadistically savoring the Holocaust on interstellar television, would the Holocaust still be wrong? 

When thinkers refuse to engage such hypotheticals, I tend to see them as evasive or anti-intellectual.  But couldn't they just as easily be prudent people who value their time too much to devise contingency plans for these extraordinarily remote contingencies?

No.  Hypotheticals serve two radically different functions.  Devising practical contingency plans is one such function.  The other function, however, is to achieve intellectual clarity in a complex world.  Pondering a hypothetical is fruitful as long as it serves one of these two functions.

"Where would you buy steaks if you only shopped at stores starting with the letter Q?" is indeed a useless hypothetical.*  The scenario will never occur, and it fails to illuminate any broader issue. 

My other hypotheticals are even less likely to occur.  But each shines a spotlight on a big question.  "If you could save either one American or x foreigners, how big would x have to be before you'd save the foreigners?" provides a clean measure of the respondent's nationalism.  "If a million farmers were stuck in Antarctica, would allowing them to migrate to a more favorable climate enrich non-Antarcticans?" shows how immigration can enrich non-immigrants.  "If six trillion Nazis were sadistically favoring the Holocaust on interstellar television, would the Holocaust still be wrong?" exposes the moral absurdity of Kaldor-Hicks efficiency maximization

Each of these topics is important.  Why not stay close to the real world when we analyze them?  We typically do.  Unfortunately, the real world is so complicated that thinkers who stay close to the real world keep getting bogged down in side issues.  Quality hypotheticals push the debate forward by stripping side issues away and getting to the heart of the matter.  As my favorite philosopher Mike Huemer patiently explains:
Some will still want to know, reasonably enough, why thought experiments are useful... Why can we not learn at least as well through the consideration of actual or at least realistic examples? Briefly, the reason is that hypothetical thought experiments provide a means for conceptual controls that often cannot be reproduced in reality. Or, in other words, they provide a way of mentally isolating a causal, explanatory, or logical factor for examination on its own which normally, in the real world, cannot be isolated, and to do so while still discussing a concrete situation.
Five-year-olds have been known to barrage their parents with tiresome hypotheticals.  "Where will my birthday party be if there's a fire at Pump It Up?"  "Then we'll do laser tag."  "What if the laser tag place burns down too?"  And so on.  The lesson to draw is not that hypotheticals are immature, but that five-years-olds pose immature hypotheticals.  Instead of teaching kids to make all their questions realistic, we should teach them to make their fantastic questions insightful.  Yes, some hypothetical questions fail to cover their costs.  But unless you entertain hypothetical questions, you can't even enter the idea business in the first place.

* Except in this post, where it's a useful example of a useless hypothetical.

Comments and Sharing

COMMENTS (19 to date)
andy writes:

I am trying to evade implausible hypotheticals because they are usually extreme positions. And I think it was someone at econlog who argued that at the extremes no ideology works.

Paul Sand writes:

I've been considering a hypothetical to pose to an open-borders advocate. Pointing to a random recent headline from Iraq (doesn't matter much which, they're all pretty bad), ask: "How about these guys? You want to let them walk across the border if they feel the urge to do so?"

That's a remote hypothetical. I had been thinking it too remote to actually pose, but after reading Bryan's post, I'm less sure of that. Maybe the answer would illuminate?

ThomasH writes:

I like hypothetical, too. They can illuminate cost-benefit trade offs.

I recall a few years ago when the Texas legislature was considering a requirement for showing a picture ID when voting, I was frustrated by what seemed to me refusal of the defenders of the rule to consider it in a proper cost-benefit framework as one would wish to do for any other government regulation. I proposed the hypothetical: how many otherwise eligible voters would they be willing to see not vote in order to prevent one fraudulent vote from being case? I was trying to separate out the empirical questions of how effective the measure would be in preventing fraudulent voting and how "effective" it would be in discouraging eligible voters from the issue of how to value of the votes so affected.

Not surprisingly, no one would answer.

Handle writes:

"When thinkers refuse to engage such hypotheticals, I tend to see them as evasive or anti-intellectual. But couldn't they just as easily be prudent people who value their time too much to devise contingency plans for these extraordinarily remote contingencies?"

Nice try framing it as a false choice between anti-intellectualism and time-based CBA. But there are other sensible reasons for judging that the consideration of a particularly posed hypothetical would be nonproductive.

When you are posing a hypothetical, you are usually trying to probe your interlocutor's fundamental values and priors, their instincts and intuitions, and also their decision-making process. You could do this to illustrate a point, demonstrate an inconsistency, contradiction or disagreement, or logically explore the conclusions and consequences that derive from the following of certain principles.

However, some people have decision-making processes which are based on principles of pragmatism and reasonableness. Reasonableness could be defined as being the opposite of unreasonable - that is - ideologically dogmatic. In other words, it takes all relevant and material factors into account and when coming to an overall conclusion, and none of those factors is completely inflexible or inelastic to certain extreme shifts in the underlying facts (a good definition of dogmatism).

The point is, context matters, and that's why people get 'stuck' in side-issues, because side-issues might affect the resulting answer and flip it from one position to another.

But hypotheticals, in their attempt at intellectual sterilization and to conduct a 'controlled' thought-experiment, attempt to remove all considerations of the context, as if it is irrelevant, which it never is, and which is also foreign to most human reasoning, moral and practical. It's justifiable to say, "It depends on a lot of other factors, and your hypothetical is unrealistically restrictive and flawed".

Especially in moral hypotheticals, backstories matter, but people who present hypotheticals pretend that they shouldn't ever have to tell any backstories, as if they couldn't possibly affect the final decision, or if they shouldn't. That's already artificially constraining the range of possible answers to one's favored domain, and it's perfectly reasonable to say, "You're rigging the game, I'd prefer not to play."

ThomasH writes:

One necessary element often left out of hypotheticals of the form how many Y equals Y is the failure to specify the margin that changes in X and Y are affecting and or the margins implicit in the non-hypothetical counterpart.

regularjoeski writes:

Hypothetically how many hypotheticals are you willing to consider when empirically they never occur. Contingency planning is planning for actual possible events. They may have a low probability of occurring but tend to have catastrophic consequences if not planned for, ie fire drills, data recovery, invasion by armed groups, etc. Hypotheticals are just games people play.

Philo writes:

Well put.

Your initial question is, indeed, a useless hypothetical. Still, for some of us the cost of coming up with an accurate answer to it is near zero. My answer: "Nowhere; I'm a vegetarian." My cost: approximately the value of my best alternative use of the ten seconds it took me to listen to and answer your frivolous question.

Handle writes:

By the way, I should note that it's somewhat ironic that the coiner of the term, "Social Desirability Bias" wouldn't note that moral hypotheticals are often purposefully designed by their presenters to try and trap their respondent and force them into an artificial corner in which they'd have to take a public stance which seems to conflate their view with opposing the dogmatic and absolutist version of the socially desirable position.

When people are able to mentally simulate the social consequences and see the trap laid out before them and say, "No thanks," it's not correct to accuse them of intellectual laziness, insincerity, or time-based cost-benefit-analysis.

For instance, trying to equate one's opponent's political preferences as being equivalent to supporting torture or slavery are the rhetorical analogs of violating Godwin's Law.

Indeed, you went there, "If six trillion Nazis were sadistically favoring the Holocaust on interstellar television, would the Holocaust still be wrong?" If you disagree with me politically, then - yadda yadda - the same as Hitler.

If anything, your example should demonstrate the absurdity or even impossibility of certain kinds of moral reasoning in extremely unfamiliar and unrealistic scenarios and linking them to the ordinary and familiar context, and yet this is precisely the kind of situation a moral hypothetical tries to exploit as if the conclusions resulting therefrom were not subject to the same absurdities.

Bostonian writes:

"If you could save either one American or x foreigners, how big would x have to be before you'd save the foreigners?"

This is not a remote hypothetical. Money that we spend on foreign aid, for example to provide antiretrovirals to combat AIDS, could also be spent on Medicare and Medicaid. The low reimbursement rates of Medicaid makes it harder for recipients to find providers and probably causes some deaths. To stay in power, politicians cannot admit that they are making money-vs-life trade-offs.

Arthur B. writes:

[Comment removed for improper coding.--Econlib Ed.]

Answering the second commenter's hypothetical question: I think it's better for the ISIS people to move to an area where they're surrounded by as many people as possible who can return fire.

Emily writes:

If I only shopped at stores beginning with "Q", I'd buy steaks at QFC (Quality Food Centers) -- a NW grocery chain.

Lanet writes:

Misc-- Sometimes hypotheticals are not hypothetical at all. Ex: If you are 100% certain the US DoD (DARPA...NSA-CIA/J-helpers) hacks into people's minds, and you tell people this, what would be the reaction by most folks? If people are not privy to what is technologically possible, even the (non-assissinated) "experts" in the field will refuse to believe it's *currently* possible.

Read: "The Matrix Deciphered" by (ex CIA) Dr. Robert Duncan.

TMC writes:

ThomasH, at least 1000.
If those 1000 folks are too lazy to pull out an ID we are better off without their votes. I'd says your hypothetical shows the Texas law would have more than one benefit.

Julien Couvreur writes:


You argue that hypotheticals are useful to simplify and control the question, but they may omit relevant aspects of the real situation.
That is fine, the hypothetical remains helpful to tease out these factors.

You may say: I would answer "X" this hypothetical, but I would answer "Y" in this other situation, because of a difference "Z".
By forcing you to explain what is that relevant difference, the hypothetical helps the discussion progress.
If you are unable to identify and justify a relevant difference, then the hypothetical reveals a contradiction.

ThomasH writes:


That's the kind of answer that I was hoping for from by interlocutors at the time. From there one might move on to think about why the trade off has the value it has: is it just the relative moral horror at the discouraged or fraudulent vote (so there is nothing more to discuss on that dimension) or does it depend on something else such as an unstated estimate about how the typical discouraged or fraudulent voter would vote.

Or if the parties had estimates of the ratios of discouraged voter was far more and less than 1000:1 then the discussion could move on to those estimates.

In either case it would have been a productive hypothetical

Tracy W writes:

Personally my suspicion with an unlikely hypothetical is that the proposer is planning to use my answer to the hypothetical out of context. Not everyone on the Internet is an honest debater.

Michael Waldman writes:

I think this hints at a central disagreement about how one should go about answering big political questions. People (such as myself) who get irked at hypotheticals like these, believe it is not productive to reason about big political questions from first principles. The core of the matter is understanding the endless facts on the ground and the interplay among them. If we're 80% aligned on what is "good" and what is "bad," that's a hell of a lot closer than we'll ever get trying to predict outcomes of policies. So why waste time on that last 20%?

To make this concrete, I'll consider your first hypothetical.

"If you could save either one American or x foreigners, how big would x have to be before you'd save the foreigners?"

Ok, well suppose I believe that foreign lives are of approximately equal dignity and worth as American lives, but it doesn't work out well in practice for people to focus on saving foreigners. How am I supposed to answer the question then? Do I pretend that the hypothetical exists in a vacuum, where we are not saving them in any particular way, not at any particular cost, not as part of a repeated policy, not impacting the expectations of fellow Americans, but just saving them, and then the world continues on as if nothing had happened? If I say that's not realistic, I'm "evading the hypothetical" ... but the complications that make it unrealistic are at the very core of why I believe what I believe! Or I could answer the hypothetical as hypothetically as possible, and then say, "But that obviously doesn't apply to immigration," and you'd say why not ... and then what did the hypothetical accomplish? We're right back into those "side issues" which are actually the crux of the matter.

Daniel Wiener writes:

I largely agree with Handle's analysis. Hypotheticals such as Bryan Caplan posits are often used as "gotcha" rhetorical tricks. If someone answers "7" or "1 out of 34" to one of these questions, the response isn't necessarily to probe the reasoning behind that number but to attack it as arbitrary. Why "7" instead of "8"? Why "34" instead of "20" or "50"? That kind of precision is indeed silly, and yet the positer of the hypothetical doesn't want to accept an answer such as "a lot" or "some" depending on may other factors.

But most answers fall in the latter category. To use another more common hypothetical, William Blackstone's formulation asserted that "It is better that ten guilty persons escape than that one innocent suffer." Sometimes one will hear variations of it, using 1 in 100 rather than 1 in 10. Isn't that rather arbitrary? How do we determine exactly what that number should be, how do we measure it, and what is the principle behind it (if any)? What it is really trying to express is the inherent uncertainty of assessing guilt, and a moral desire to err on the side of caution. There is always going to be a gray area involving factors such as limited human knowledge, flaws in the judicial system, mistakes, imperfect or corrupt human beings, etc. Trying to assign numbers or percentages can at best mark out the extreme limits. If the ratio at one extreme was one guilty person to one innocent person, that would imply only a 50% accuracy in our legal/judicial/forensic system, no better than a coin flip, which most would deem unacceptable. Whereas a ratio at the other extreme of 10,000 guilt people to one innocent one would imply a 99.99% accuracy requirement, which most would agree is unrealistic. So people use other factors to help fill in the gray area between those extremes. For example, technology can factor in, so that a generation of jurors who've been raised on TV programs like CSI and know about DNA will now look for a higher level of evidential certainty than they previously might have, just because it's feasible. But they still can't assign an exact percentage to "beyond a reasonable doubt" and there's no way to verify such a percentage anyway.

As Handle said, stripping out the backstory and other factors doesn't necessarily clarify a hypothetical, but rather can make it either meaningless or a debating weapon. No wonder people refuse to play that game. The question of saving one American versus X foreigners has little meaning without context. If all are total strangers, I may be indifferent. Whereas if it was a question of saving a loved one versus X strangers (whether American or foreign), I would say that X would have to be extremely large (although I still wouldn't try to assign an exact number). It matters who the individuals are, why they are in danger, what the cost to me is, etc. Without that additional information I can't answer the original question. Perhaps that's the purpose of stripping it down to "American" vs "foreigner", to try to assess nationalism, but I can't make such a determination in a vacuum. Perhaps "Americanism" might factor in slightly or somewhat, so that nationalism isn't totally precluded, but other factors are far more important. Hence I and most other people would indeed refuse to play that game, as it could so easily be used to misrepresent our real views rather than illuminate the underlying issues.

Comments for this entry have been closed
Return to top