Bryan Caplan  

Amazing Dan Klein

Benefits Bleg... How U.S. Drug Enforcement in A...
I'm puzzled by Dan Klein's recent guest post.  I trust his empirics.  Indeed, I prize them.  But what do they show?  As far as I can tell, Dan's empirics show that a sizable minority of economists are small-l libertarians, but most economists aren't.  By that standard, yes, economists don't enjoy an amazing consensus. 

Yet why is that the only relevant benchmark for amazing consensus? 

Here's an alternative: Economists across the political spectrum accept many empirical claims that non-economists habitually deny.  They recognize the economic benefits of downsizing.  They agree that rent control causes housing shortages.  They think a market in human kidneys would save lives.  Amazing, I say!

If Dan wants to stick to policy positions, there's still ample ground for amazement.  If Dan gave his survey to the U.S. public, there's good reason to believe that the typical economist would seem very libertarian by comparison.  Suppose the median economist were more libertarian than 90% of the U.S. public.  Would that be an amazing consensus in Dan's book?  It sure is in mine.

I'm as disappointed in our profession's statism as Dan is.  Maybe more so.  But there's a big difference between saying, "Economists don't enjoy an amazing consensus on the issues that most concern me" and "Economists don't enjoy an amazing consensus at all."

COMMENTS (21 to date)
Vipul Naik writes:

I think there are differing definitions of consensus at play here. To take an extreme example, suppose group A is the majority group, whose positions on an issue are distributed in a Bell curve. Group B is a minority group whose positions are also distributed in a Bell curve whose mean is 4 SD (where the SD measure is as per Group A) to the right of Group A's mean, and with a standard deviation twice that of Group A.

Relative to Group A, then, Group B enjoys an amazing consensus in the sense of mostly agreeing to positions far to the right of Group A (in fact, about 97.7% of Group B is to the right of the median for Group A, if I'm reading the bell curve numbers correctly). But viewed in terms of the extent of intra-group disagreement, Group B has much larger intra-group disagreement than Group A, so enjoys less consensus.

Of course, the example of economists and the general public is far less extreme, but I think your differences with Dan Klein are more on issues of what it means to have a non-mainstream consensus.

Chris Koresko writes:

@Vipul Naik: I agree. It would have been useful to define a numerical consensus metric. I'd suggest a simple variance would work for these data.

But I think there's another, more fundamental problem: If I read correctly, the range of options in the survey was essentially from "keep existing policy" to "change to a much more interventionist policy". But presumably some respondents would have preferred a less interventionist policy, and their answers would be clustered on the edge of the graph, making the variance look artificially low (i.e., biasing the consensus metric upward).

One of the basic rules of measurement is to make sure your measurement apparatus is capable of recording the full signal range of interest. Or if that's not practical, then commit yourself to the effort of accounting for the inevitable distortion of the signal during the data analysis stage.

Daniel Klein writes:

@Chris Koresko: I don't think the upward-ratcheting nature of the reforms of the questions of the survey I singled out makes a difference. The four other surveys show similar results -- frequent lack of single-peakedness among the responses to a question. Frequent lack of consensus.

@Bryan: I need to ponder your response, but I think the comments above have a point. The meaning of "consensus" seems to be getting fuzzy.

@Daniel Kuehn (who commented at my post): I disdain talk of "positive v. normative." You should, too. Such disdain is wise. Such disdain is enlightened.

Daniel Kuehn writes:


Mixing positive and normative is barbaric. You should agree. Path to enlightenment is their segregation. Nothing good comes from treating preferences and values like objective truth that all ought to obey. There's plenty of latitude for disputing truly heinous subjective values without squeezing liberalism down to some miniscule inflexible libertarian creed.

Daniel Kuehn writes:


david writes:

If you had time-travelled to 2006 and quizzed economists on whether they thought buying government bonds en masse was a good idea for a country at a zero-bounded nominal interest rate, you'd have received a lot more agreement amongst the conservative economists than you would have five years later. Friedman loudly supported it, after all.

Partisanship makes people more careful about what positions they support. Economists have a consensus on kidney markets because kidney markets are not, in fact, a viable issue in public debate. If it ever became an actually-contested hot-button issue, the consensus amongst economists would rapidly collapse.

Guessing the direction it will collapse in is easy: survey only the economists who already specialize in that field. The consensus in opposition of rent control is much weaker amongst housing economists, in the same way the consensus toward expansive regulation of market failures is weaker amongst public choice economists.

Daniel Kuehn writes:

Normative questions like the perspective on kidney markets can come in different flavors too. It's one thing to say "I don't think people should find this reprehensible - I wish they didn't think that way", and it's another thing to say "Ignore their impassioned views - make them live with what they consider to be reprehensible - it's good for them".

Sometimes we are willing to make both statements. Sometimes we are only willing to make the former statement. Sometimes we find it reprehensible ourselves because of some sense of the sacredness of the body without actually denying the economic point about the efficiency of the kidney market.

All of this is subjective and it seems very dangerous and illiberal to pretend otherwise.

What we can say safely and liberally is what sort of outcomes we might expect from a kidney market.

Seth writes:

Economists seem to enjoy an amazing consensus on things that don't matter much. Or, don't matter much, now.

I'd be interested in learning more about why Dan Klein disdains talk of positive v. normative.

Daniel Klein writes:

@Seth and Daniel:

I hope someday to write about "positive v. normative" talk.

Let me just say that when I hear "p v. n" talk, it is not as though I regard the talk as meaningless (nor that I disdain the speaker). When I hear such talk, I read meaning into it. But, whichever meaning I read into it, there are always better terms than "p v. n."

For example, in some instances of "p v. n," the best meaning is reserved v. outspoken. In some instances it is noncontroversial v. controversial. In some it is conventional v. unconventional.

Meanwhile, once you confine "p v. n" to any one of the common meanings, you will find that many of the general statements people make about the matter of "p v. n" do not hold water.

Daniel Klein writes:

I should add, another common meaning is precise/accurate v. loose/vague.

Daniel Kuehn writes:

Could you give an example where those better terms apply? I can't think of a single instance where that would clarify the issue.

My worry is that blurring the distinction (which seems to me to be pretty concrete, although perhaps I'm fooling myself) has been a classic way for people with a certain set of values or a certain ideology to smuggle that in with a veneer of science.

Science is not the be-all-end-all, but to the extent that you can wrap it up with other methods of persuasion I think that's dangerous.

But to get back to my question - what is an example? Take fiscal policy. What government spending does to the economy is a very hard, very imprecise, very contingent question. But the distinguishing difference between "what does fiscal policy do?" and "is fiscal policy a good idea?" is pretty stark. The former question has nothing to do with other things we're worried about (fairness, human dignity, responsible government, competing public priorities, etc.). The latter question has everything to do with those issues, and it's informed by the positive question.

None of your other words seem to get at that distinction that "p v. n" is getting at, and since that's an important distinction to think about, it's no wonder most people agree keeping positive and normative claims distinct in our minds is crucial.

Daniel Kuehn writes:

As far as your point about "enlightened", I genuinely don't get this. When you mix "is" assertions and "ought" assertions what you get is an undifferentiated assertion you are making for others that is highly arbitrary. That doesn't sound enlightened or liberal to me at all.

Seth writes:

@D. Klein - Thanks. Perhaps another distinction is less prone to bias and more prone to bias?

thuddmonkey writes:

@ D. Klein - please do write about that (or follow up further on this thread). Like D. Kuehn, I think of "p v. n" as equivalent to "is v. ought", and that seems a crucial distinction. If I'm wrong about something in there, I could significantly improve my thinking. An example would be helpful.

Maybe you are not opposed to distinguishing "is v. ought", but think that is not what people who talk of "p v. n" are usually doing? That is, you are opposed to the _misuse_ of "p v. n" framing?

Daniel Klein writes:

Hi people,

I don't think there is much to "is" vs. "ought."

If I say: "You owe X 10 dollars", is that is or ought?

Meanwhile, look at this:

If I say: "You were schooled to capitalize the first word in the sentence", would that be is or ought?

Meanwhile, look at this:

Which notes derivation from the Swedish skola, which is the infinitive of the Swedish skulle, which would be translated as should. Meanwhile, skola is also the Swedish word for school.

But, etymology aside, I think it is easy to rephrase "is" statements into "should/ought" statements, and vice versa.

So, no, is versus ought doesn't cut ice with me. I regard it as more modern dogma (Booth, Modern Dogma and the Rhetoric of Assent 1974).

david writes:

It seems a little strange to blame modern dogma and cite centuries-old Swedish etymology.

The usual modern attack on the is-ought problem is that discourse always buries 'oughts' of some sort, which renders the is-ought problem a theoretical curiosity in practical applications, not that all invocations of the problem are trivial.

Your choice of example is telling: questions of debt are in fact obviously linked to an underlying premises on which the ownership is generated. Those premises are, despite the best efforts of American-libertarian authors, not self-evident and most certainly not universal. So the point that you need oughts to imply other oughts becomes relevant again.

Todd Kuipers writes:

Kuehn is relatively clear - is/are vs. ought/should are very different, e.g.:
- bears ought to stop mauling people
- bears are not mauling people

Former normative, latter positive. Former, possibly a nice idea, latter measurable. Mixing up the former for the latter mauls facts in favour of value statements (often of undefinable morality) - very different things. Politicians and savvy advocates do this to great effect, dangerous when combined with coercion.

"If I say: "You were schooled to capitalize the first word in the sentence", would that be is or ought?"
- Clearly positive
If I say: "you should be schooled to capitalize the first word in the sentence."
- Clearly normative

Former, measurable. Latter, a value statement, likely one that would cause me to provide an unkind verbal rejoinder.

Daniel Kuehn writes:

Daniel Klein -
This is getting silly. Sure language is variable and complex. "Owe" can be used to describe a specific contractual arrangement and it can be used to describe an ought obligation.

Let me ask you this - some people want to "sanction" Iran in the sense of putting financial burdens on them, while others want to "sanction" Iran in the sense of supporting what they are doing.

Do we conclude that there is nothing really substantive to the idea in international relations of enemies and allies?

Of course not. It's not even a serious contention, and it's certainly not something that citing "sanction"/"sanction" provides a plausible defense for.

Daniel Kuehn writes:

To resolve your "owe" sentence all I have to do is ask the speaker "do you mean owe in the sense of being in financial debt or do you mean owe as somehow under a moral obligation". Bingo - problem solved. It's just a quirk of language.

You cannot resolve is and ought statements about fiscal policy that way.

Which brings me back to my question - can you give us an example of a normative/positive distinction in economics that is better thought of with one of your distinctions?

thuddmonkey writes:

@ Dan Klein - thanks. So far, though, it still seems important to be able to identify "ought" statements and notice the value judgments implicit within them.

The etymology is interesting, but unconvincing; at best, it shows that people have historically been sloppy about making the distinction, which doesn't mean there _isn't_ an important distinction.

The "owe" statement is ambiguous because English is ambiguous - you might have meant "... according to such and such a system of accounting and ownership", which would be an "is". Or you might have meant "... and so you would be morally derelict not to pay me", which would be an "ought".

In the hypothetical, probably you meant both, but I might object to either part. Being able to say *which* part seems vital, since if we are to reach agreement, the two objections should be pursued very differently - one by a mere retallying, the other by attacking or justifying the system of accounting and ownership. Hence the importance of is v. ought.

Brian writes:

While I agree with the majority here that the positive/normative distinction is necessary and good, I also think Dan Klein has a point that is not being fully considered. He says

"I think it is easy to rephrase "is" statements into "should/ought" statements, and vice versa."

He's right. The seemingly normative statement "bears ought to stop mauling people" is really saying "I don't like it when bears maul people," which is a positive statement.

Likewise, the clearcut positive statement "bears are not mauling people," by the very act of being mentioned, is carrying normative implications. After all, why focus on the people-mauling activity? Why not state the equally positive statements "bears are not flying today" or "bears are not singing opera"?

The reality is that oughts (normative) are always backed by CLAIMED facts (positive). But how can the oughts be justified without testing the associated facts? And why do facts matter if they don't carry any implications (oughts)?

E.g. "We ought to ban assault rifles. A ban would reduce the number of people killed." The normative statement stands or falls on the validity of the claimed positive one.

Failure to recognize that many value statements are unsupportable is the source of much grief and gridlock in the public sphere.

Comments for this entry have been closed
Return to top