Bryan Caplan  

Coercive Priors

The art of the possible... Recession Bet...
I enjoyed scoffing at the 30% of Americans who want to bomb the fictional kingdom of Agrabah.  But for balance, note that multiple surveys - and a Penn and Teller episode - have found plenty of Americans who want to ban water - as long as you call it "dihydrogen monoxide."  A national representative survey could easily find 30% support for such a ban. 

By construction, these policy proposals have no evidence in their favor.  Why then would anyone support them?  Prior probabilities.  Some people have the prior, "If we're discussing the bombing of a Muslim country, bomb it."  Some people have the prior, "If we're discussing the banning of a chemical, ban it."  And as a counter-survey points out, 44% of Democrats favor admitting refugees from the fictional kingdom of Agrabah.

What lesson should we draw?  If you're a strict consequentialist, no lesson at all.  If however you think there's a moral presumption against coercion, widespread coercive priors are a symptom of moral corruption.  Decent people say, "After careful examination of the facts and consideration of our options, we regretfully conclude that coercion is the only viable way to prevent great harm," not "Coercion - hell yes!"

COMMENTS (17 to date)
Thomas writes:

If "decent people" conclude that coercion is the right thing, "after careful examination of the facts and consideration of ... options," what's to regret? This strikes me as a good example of consequentialist thinking. Your use of the term "strict consequentialist" is mere hair-splitting.

pongogogo writes:

The responses also demonstrate that some political polls are meaningless, and we shouldn't overweight their conclusions.

Phil writes:

what these sorts of polls tell me, is that if you have a cause you care about promoting, you should be careful about how you politicize it

once partisans get their hands on a particular idea, most people stop considering the idea on its merits and simply pattern match whether the idea is one of the ideas that their side is supposed to like

Matt Moore writes:

People draw conclusions based on the existence of a survey. They don't tend to consider the possibility that the survey is a trick. They mentally link Agrabah with IS (because media), assume it is an IS stronghold and answer yes. Which is a valid policy position, although not one I agree with.

That is the, prior is in fact 'if we are discussing a country in the context of bombing, it is an IS stronghold, and the real question is whether bombing is effective or not.

Ditto water. 'If we are discussing banning a chemical, I am really being asked where I fall on the environmental vs industrial spectrum'

The fact of the survey is important context.

Greg G writes:

"What lesson should we draw?"

You should draw the lesson that surveys are an extremely poor tool for drawing conclusions that just happen to justify your priors.

I majored in Sociology at a top Liberal Arts college. I have almost nothing to say in defense of the discipline of Sociology. Except for one thing. It's a great way to learn to be very skeptical of anyone drawing any significant conclusions from a survey.

It is astonishingly easy to tweak the desired results with a subtle change in the framing of the questions. It is even easier to do so when you are as misleading as possible as was done with the description of dihydrogen monoxide.

It was described as an acid which can cause severe burns. Anyone constructing a survey question in good faith would realize readers will assume this means at room temperature. All materials can be heated to the point where the could cause severe burns. The purpose in describing it the way they described it was precisely to elicit from people a result they could not have gotten with a more honest survey question.

I am sorry to say this, but this is one of the worst posts I have ever seen on this site. That is the lesson I have drawn. There are any number of good ways to argue for more libertarian policies. This was not one of them.

B K writes:

As a proponent of open borders, I would imagine that you, Bryan, would also be a supporter of admitting refugees from the fictional kingdom of Agrabah...

Chris Koresko writes:

It might be worth thinking about responding to one of these polls as a close analogy to voting. The person being polled probably didn't spend much time preparing to answer the questions and probably has little incentive to do a good job at it.

Asking a question about bombing a fictional kingdom in the context of today's news insinuates that this kingdom is a stronghold of one or more enemies; willingness to support bombing it is a proxy for taking a more aggressive stand against them. Likewise, accepting immigrants is a proxy for supporting a more aggressively humanitarian approach to dealing with the suffering created by those same enemies.

The first poll was presumably conducted to embarrass Republicans by revealing their ignorance, and the second to do the same thing to Democrats.

What conclusions can be drawn? That many Americans are willing to support a much more activist foreign policy aimed at reducing the suffering of people abroad preventing the violence from being brought home. That many Americans (rationally) haven't invested a lot of time into learing the details of the situation over which they individually have little control. And that pollsters are trying to manipulate public opinion through deceit.

ScottA writes:

Complicated answer to a simple problem; if you ask people the answer to a question they don't know, but imply that it's important, they'll answer it nearly all the time.

Priors are the only potential thing to get out of those questions, but not sure that it's that worrisome. People are more reasonable than pollsters give them credit for; if there's a serious question about attacking a country, it's fair to assume that there's a good reason. The respondents don't know what the reason is, but see the first paragraph. Same issue in the water example. There are groups of respondents who assume this means that there's a threat to America, who assume that there are human rights issues, etc. We have no idea what any given person's reason for answering one way or another is. Maybe it's moral priors, but maybe it's just a different assumption about what the guy in the call center means by this weird question.

Zack writes:

Look a little closer at that PPP poll. For Republican voters, the Agrabah question was number 38 in a 41 question poll. It was also immediately preceded by several questions about muslims,9/11, and terrorism.

For Democratic voters, it was question 6 in a 10 question poll, and none of the preceding questions were about Islam or terrorism.

I think this poll says a lot more about PPP than it does about voters from either party.

Jeff writes:

Isn't the more obvious conclusion that universal suffrage democracy is a nutty system of governance?

Nathan W writes:

I think it's a lot more ignorant to support bombing a place that doesn't exist than to support accepting refugees from that place.

At least in the case of accepting refugees, the respondent is provided with the nugget of information that there is a refugee. Not just "should we bomb A?" and respond "yes" to whatever A happens to be.

The educational system should encourage people to be more willing to say "I don't know" when they don't know.

jon writes:

This tweet -

They oppose bombing 'it' 36/19, while GOP supports bombing 'it' 30/13
- from the polling company does not mean what they think.

The implication they are trying to put across is that more Republicans than Democrats are uninformed because more of them supported bombing a fake country. But wouldn't you get similar results from a poll about a real country? In other words, it's not necessarily true that more Democrats realized that there is no such place as Agrabah, it's just that fewer democrats would support bombing any country, real or fake.

Jim Glass writes:

What lesson should we draw? If you're a strict consequentialist, no lesson at all. If however you think there's a moral presumption against coercion, widespread coercive priors are a symptom of moral corruption.

Corruption implies an initial natural state of purity. This is itself a very common cognitive error which leads the bad mistake of concluding things are getting worse when in fact they are getting measurably better compared the *actual* initial natural state, which in almost all social dimensions was an unhappy one by our standards indeed.

This error is very commonly seen in discussions of social/political economics/behavior. Libertarians certainly are no more immune to it than anyone else.

Decent people say, "After careful examination of the facts and consideration of our options...

Entirely decent people take huge shortcuts in their thinking all the time, committing such consequent errors. Everybody does. Nobody in the world has the capacity to 'conduct a careful examination of the facts and consideration of all the options' on every subject they might ever be asked a poll question about -- sexual behavior, middle east politics, health policy, justice system policies, whatever ... And for an expert specializing in such a field to then cherry-pick a poll relating to it and personally slur the average-person respondents for answering so superficially and glibly ... why, it's just not decent!

I suppose one could say that anyone who hasn't conducted a careful examination of the facts and considered of all the options relating to a poll question just shouldn't answer it, except by saying "I dunno." But that would produce a 99% no-response rate.

And perhaps the greatest, most obvious cognitive failure of human beings is, as Kahneman says, that even though 100% of the population knows absolutely or near nothing about 95+% of everything, everybody has an opinion about everything, and values it.

But hey, that's reality and is what political systems have to work with. And there is nothing at all "corrupted" in it. It is the natural initial condition.

Jesse C writes:

Having three young daughters, I know all about Agrabah and fully support bombing it. What is there to scoff at?

G. White writes:

Nice/keen post, Jim Glass.

I used to enjoy your "Grinch" posts from the sci.econ days.

Ben Kennedy writes:

This isn't what is going on at all... people hear "should we bomb Agrabah", then assume it has something to do with ISIS, then answer the question "should we bomb ISIS". The conclusion is that people think ISIS should be bombed, not fictional cities or any Muslim-sounding city

Comments for this entry have been closed
Return to top