Ariel's JEL article

[Itzhak Gilboa]

Hi Ariel (and all),

Thanks for writing and circulating this.  As usual, it’s very interesting, and, as usual, I couldn’t agree more with some statements, nor less with others J

Trying to distinguish between these, I found it helpful for myself to highlight two distinctions:

·         Descriptive/normative:

Statements about academic activity can be descriptive or normative (and sometimes both).  This isn’t very surprising if we think of academic activity as a social phenomenon, and of the philosophy/sociology of science as a social science.  Both descriptive and normative claims can be first-order, about reality itself, or second-order, about how academics think about it, but that’s probably of secondary importance.

In the case of the statement “economic models are stories” I think that many would agree with the descriptive claim that economic models are sometimes just like stories, but would not endorse it as a normative claim.  I, for one, feel uncomfortable with accepting the state of affairs you describe and judging economic models by the same standards we judge literature.  I think that our field enjoys a lot of resources (for example, it’s much easier to make a living as an economist than as a writer), and that people outside it expect our models to have a positive impact on people’s lives that's more immediate/concrete/material than that of good stories.

·         Scientific and meta-scientific results:

Your distinction between results that discuss specific models and those that are about families of models is very important.  I usually think about a related distinction: between results that are supposed to say something about a certain reality and results that are supposed to say something about the way economists work.  We can use the terms “scientific” for the former and “meta-scientific” for the latter, though both are pompous.  For want of better terms I’ll use these for now.

Whether a result is scientific or meta-scientific is a matter of interpretation, and usually doesn’t follow directly from the math (as is the case with the distinction between descriptive and normative results).  For example, a result about the existence of equilibrium could be read as a predictive statement about the long-run behavior of a system (in particular if it is supported by a convergence result), and it could be interpreted as saying “Let’s use a solution concept that’s at least guaranteed to exist.”  Axiomatizations and impossibility results are often much more powerful as meta-scientific than as scientific statements.

Often, scientific statements are about a specific model, whereas meta-scientific ones – about classes of models.  But the two distinctions don’t overlap precisely.  For example, if I analyze an insurance problem with a CARA utility function and then generalize it to any concave function, I can get a result about a class of models, in a sense, but it still aims to discuss a particular reality.  By contrast, if Rosenthal computes the backward induction in the Centipede game he is probably trying to tell us something about the solution concept we use, though he’s dealing with a very specific example.  

I agree that, by and large, fancy math is much more justified for meta-scientific results than for scientific ones.  Partly, this is due to the fact that assumptions never really hold in reality.  This is a problem when one takes them very seriously in generating predictions, but it is much less of a problem for the rhetorical arguments of meta-scientific results.  Partly, this might be due to the fact that scientific results, such as computation of equilibria, often implicitly assume that the agents can do the math that we can (or can’t) do, whereas no such claim is made in the meta-scientific interpretation.

Anyway, I suppose that on some issues we started out with different priors and will continue to disagree J

Best

Tzachi

On Fri, Mar 10, 2017 at 11:49 PM, Ariel Rubinstein <ariel.ru...@gmail.com> wrote:

I am clearly jetlagged...  3 of u correcred me....

I meant an article in Journal of Economic Literature (not JEPerspectives..)

http://arielrubinstein.tau.ac.il/papers/100.pdf

Ariel

--------------------------------------------------------------------
Ariel Rubinstein
Homepage:  http://arielrubinstein.tau.ac.il/
Cell: +972-544-235347
Skype: arielrubinstein
Contact information: http://arielrubinstein.tau.ac.il/contact.html
---------------------------------------------------------------------

--
You received this message because you are subscribed to the Google Groups "decision_theory_forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to decision_theory_forum+unsub...@googlegroups.com.
To post to this group, send email to decision_theory_forum@googlegroups.com.
Visit this group at https://groups.google.com/group/decision_theory_forum.
For more options, visit https://groups.google.com/d/optout.

[Attila Ambrus]

Hi Tzachi, Regarding your last point, there is of course Milton
Friedman's pool (billiard) player analogy. As a textbook later put it:
"Professional pool champion Corey Deuel may not know all the formal
laws of Newtonian physics, but the quality of his play suggests that
he has a deep understanding of them." It varies with the setting how
convincing the point is.
Regarding the broader topic, I wonder if anyone knows if economists
self-reflect less or more than other social scientists. Discussions
like what economics is/should be and what the role and value of of
economic modeling is often heat up, most recently by Ariel's
fascinating paper. I don't know though how frequent similar
discussions are in political science, anthropology, etc. In many of
those disciplines there is no dominant mainstream methodological
framework and as in economics these days, and the power hierarchy is
much more blurred than in economics where there is a broad agreement
about the rankings of departments, journals, etc. Does that imply more
debates about what the discipline should be about or the opposite -
different schools of thoughts just do their own things and don't get
into debates with the other ones?
Best, Attila

>> email to decision_theory_...@googlegroups.com.

>> To post to this group, send email to

>> decision_t...@googlegroups.com.

>> Visit this group at https://groups.google.com/group/decision_theory_forum.
>> For more options, visit https://groups.google.com/d/optout.
>
>

> --
> You received this message because you are subscribed to the Google Groups
> "decision_theory_forum" group.
> To unsubscribe from this group and stop receiving emails from it, send an

> email to decision_theory_...@googlegroups.com.
> To post to this group, send email to decision_t...@googlegroups.com.

[DUBRA Juan]

Regarding Attila's question, I know there is a much bigger debate in political science between "old school" and "economic like" analysis.

A standard argument of the old school (people who sometimes do detailed comparative analyses of two political systems) is that if one adopts the Friedman type position, one may be able to predict, but we fail to understand **why** things happen.

I am sure there are people in this forum who can expand on this point. But I think that discussion is probably more important there than the current debate in economics (which I find very interesting).

To Ariel: I read your piece, and I hoped to find something about the distinction you recently made in one of our conversations between a theory and a model.

Best regards

Juan

[Itzhak Gilboa]

Hi, 

Attila: indeed, the merits and limitations of the "as if" justification have been discussed at length, even on this forum.  I am not terribly convinced by analogies to motor tasks.  Evolution prepared us to walk on two, catch objects, and avoid rotten food.  It hasn't had the time to prepare us to make optimal decisions in elections or financial markets, even if we theorists knew what these "optimal" decisions were supposed to be.  

(Btw, since Friedman suggested this analogy our Billiard player learned a bunch of new tricks.  I think this defense of the "as if" story [!] is weaker today than it was in 1953.)

Re other fields: from the little I know, it seems that our case poses a sociological puzzle more than others do.  There seems to be a consensus among economists about how to do what we do and what to teach graduate students (more than in other social sciences) and yet it's not obvious what our models mean.  There's a lot of interest in this phenomenon among philosophers of science as well.  (See, for instance, the special issue of the Journal of Economic Methodology in 2009, and recent books on the topic of models in economics.)  So, from the very little I know, I think we can feel pretty special :)

Best 

Tzachi 

>> send an email to decision_theory_forum+unsub...@googlegroups.com.

>> To post to this group, send email to

>> decision_theory_forum@googlegroups.com.

>> Visit this group at https://groups.google.com/group/decision_theory_forum.
>> For more options, visit https://groups.google.com/d/optout.
>
>
> --
> You received this message because you are subscribed to the Google
> Groups "decision_theory_forum" group.
> To unsubscribe from this group and stop receiving emails from it, send

> an email to decision_theory_forum+unsub...@googlegroups.com.
> To post to this group, send email to decision_theory_forum@googlegroups.com.

> Visit this group at https://groups.google.com/group/decision_theory_forum.
> For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "decision_theory_forum" group.

To unsubscribe from this group and stop receiving emails from it, send an email to decision_theory_forum+unsub...@googlegroups.com.
To post to this group, send email to decision_theory_forum@googlegroups.com.

Visit this group at https://groups.google.com/group/decision_theory_forum.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "decision_theory_forum" group.

To unsubscribe from this group and stop receiving emails from it, send an email to decision_theory_forum+unsub...@googlegroups.com.
To post to this group, send email to decision_theory_forum@googlegroups.com.

[Attila Ambrus]

Hi Tzachi, It would indeed be silly to think that evolution lead us to
equilibrium behavior on financial markets. But I think people who try
to make the case for the as if argument empirically have a different
story in mind, that people who acquire expertise in certain decision
situations intuitively learn to play in a way that's consistent with
game theoretical predictions, without thinking in formal game
theoretic terms. I have the Chiapport, Levitt and Groseclose paper on
soccer penalty kicks and the Levitt, List and Sadoff paper on
professional chess players and backward induction in laboratory games.
Best, Attila

>>> send an email to decision_theory_...@googlegroups.com.

>>> To post to this group, send email to

>>> decision_t...@googlegroups.com.

>>> Visit this group at
>>> https://groups.google.com/group/decision_theory_forum.
>>> For more options, visit https://groups.google.com/d/optout.
>>
>>
>> --
>> You received this message because you are subscribed to the Google
>> Groups "decision_theory_forum" group.
>> To unsubscribe from this group and stop receiving emails from it, send

>> an email to decision_theory_...@googlegroups.com.

>> To post to this group, send email to

>> decision_t...@googlegroups.com.

>> Visit this group at https://groups.google.com/group/decision_theory_forum.
>> For more options, visit https://groups.google.com/d/optout.
>
> --
> You received this message because you are subscribed to the Google Groups
> "decision_theory_forum" group.
> To unsubscribe from this group and stop receiving emails from it, send an

> email to decision_theory_...@googlegroups.com.
> To post to this group, send email to decision_t...@googlegroups.com.

> Visit this group at https://groups.google.com/group/decision_theory_forum.
> For more options, visit https://groups.google.com/d/optout.
>
> --
> You received this message because you are subscribed to the Google Groups
> "decision_theory_forum" group.
> To unsubscribe from this group and stop receiving emails from it, send an

> email to decision_theory_...@googlegroups.com.
> To post to this group, send email to decision_t...@googlegroups.com.

> Visit this group at https://groups.google.com/group/decision_theory_forum.
> For more options, visit https://groups.google.com/d/optout.
>
>
> --
> You received this message because you are subscribed to the Google Groups
> "decision_theory_forum" group.
> To unsubscribe from this group and stop receiving emails from it, send an

> email to decision_theory_...@googlegroups.com.
> To post to this group, send email to decision_t...@googlegroups.com.

[Spiegler, Ran]

The Levitt-List-Sadoff paper shows that the chess masters know how to play optimally in a zero-sum, race to 100 game. In contrast, in the centipede game, they play sensibly and don't do backwards induction. Race to 100 is just like chess - a complete-info, sequential-move zero-sum game - except that it's much simpler for an experienced chess player to reason about. I don't see how this supports in any way Friedman's argument.

Ran Spiegler

School of Economics, Tel Aviv University

&

Department of Economics, University College London

 

URL: http://www.tau.ac.il/~rani

---

From: decision_t...@googlegroups.com <decision_t...@googlegroups.com> on behalf of Attila Ambrus <aambr...@gmail.com>
Sent: 17 March 2017 17:28

[Attila Ambrus]

I didn't mean to say that these papers prove the relevance of the Friedman argument, just wanted to say that authors who try to do something like that have different channels in mind than evolution. I am not a fanboy of the Friedman argument, but also don't think that it is completely absurd.

Sent from my iPhone

[Attila Ambrus]

Btw, happy birthday, Ran!:)

[Attila Ambrus]

Actually, I take it back. Maybe they have a social evolution story in mind, as opposed to some more conscious learning.

Sent from my iPhone

On Mar 17, 2017, at 11:45 AM, Spiegler, Ran <r.spi...@ucl.ac.uk> wrote:

[Levin, Dan]

Why are the playing such a game?  My recollection is that the game of chess has a winning strategy so “fully rational” players just need to flip a coin to determine  who is the winner.

 

Since I almost never seen a non-trivial game-theory prediction survives an experimental test, it implies that Friedman/Popper paradigm of “as if” is useless in what we do, and for much of social science.

 

(As the famous joke we tell: “In physics one good falsifying experiment kills the theory where as in economics one lousy supporting experiment keeps that theory alive forever.”  I cannot recall too many died.)

 

Our models are either for insights (A-R) or have Normative appeal  (I-G), both are important and useful.  Both imply that assumptions (axioms) are important.   We study very complicated beings with complicated and heterogeneous (in my view) behavior. But we insist of asking tough questioned that are hard to answer (unlike physics).  The best we can do is keeping a healthy dialog among theoretical, empirical and experimental work. Keep  “wrong” stories alive as long as we don’t have better one.  

 

Have a wonderful WE

 

dan

_______________________________________________

Dan Levin, Professor                  E-mail: levi...@osu.edu
Department of Economics          Office: (614)-688-4239
The Ohio State University           FAX: (614)-292-3906
443B Arps Hall
1945 North High Street
Columbus, Ohio 43210-1172
             Web: http://www.econ.ohio-state.edu/levin/

 

[Attila Ambrus]

I think the Friedman argument was supposed to apply to professional billiard players, not to random undergrads asked to play billiards in the lab. I think typical lab experiments don't speak to the concrete question being discussed here.

Sent from my iPhone

[Levin, Dan]

The “as if” is more general it means roughly “what matters is if the predictions works –test the predictions don’t worry about the assumptions. 

 

Experimental or empirical testing is the main elements of any science so I simply don’t understand your comments.  Are undergrad not part of the population we theorize about? Do you think expers behave differently in the lab (in general)?  (They don’t).

[Kagel, John]

For a survey comparing professionals with students in the lab you might have a look at

Fréchette, Guillaume R. (2015), Laboratory Experiments: Professionals versus Students, in Handbook of Experimental Economic Methodology, Guillaume R. Fréchette and Andrew Schotter (editors), Oxford University Press, pp: 360-390.

 

My own experience is that the best thing a lab exp can do with professionals is to point out important practical elements the standard theory leaves out. So for the winner’s curse in construction industry – experienced bidders recognizing a mistake in their bids are typically allowed to withdraw without penalty (inexperienced bidders view the WC as “entry fee” for getting into a new branch of the industry).  In off shore oil lease auctions bidders specialize on certain geological formations that they have experience with (so can better interpret seismic readings) + by specializing it reduces the number of competitors which provides some relief from the winner’s curse. 

 

Dyer and Kagel "Bidding in Common Value Auctions: How the Commercial Construction Industry Corrects for the Winner's Curse,"Management Science, Vol.42 (1996).

 

There is also evidence that volunteers in lab exps are typically “smarter” (higher SAT scores and HS class rank) than the av collegel student so that to the extents they are “unrepresentative” they are smarter than average!

 

Casari et al. “Selection Bias, Demographic Effects and Ability Effects in Common Value Auction Experiments,” American Economic Review vol 97, 2007.

 

 

 

 

 

John H. Kagel

University Chaired Professor of Economics and

Director of the Economics Laboratory

Ohio State University

614-292-4812

614-292-4192 (Fax)

http://econ.ohio-state.edu/kagel/

[Attila Ambrus]

Hi John, many thanks for the references! Some of them I knew of, but I
haven't read them so far.
I guess also comparison between inexperienced and experienced subjects
in lab experiments (where experienced means that the subject already
participated in a very similar decision situation a bunch of the
times, either in the same session or a previous one) is trying to get
at the same issue. In general, it is clear that experimental
economists are aware of the issues concerning external validity of lab
experimental results, and they are thinking hard about coming up with
clever ways to address these concerns.
I was aware of the fact that lab experiment subjects are not a
representative sample of undergrad students. I used the wording random
undergrad students in my email because if you indeed wanted your
subjects to play billiards, your subject pool would be random in the
sense that I have no clear prior which way the sample would be biased.
Best regards, Attila

[Ola Mahmoud]

Hi -- I apologize for perhaps taking this post off-track, but I have two follow-up questions on tangentially related topics:

- Evolution may indeed not have had the time to "explicitly" prepare us for making optimal decisions in financial markets, but decision heuristics we use today may be the result of cognitive adaptation. In other words, we may all be applying strategies that generally increase the likelihood of solving a problem that our ancestors (and perhaps other species) routinely faced. The Gamblers' fallacy, risk aversion, and even diversification in portfolio choice could be the result of evolutionary adaptation. Whether or not such solutions are optimal, I am not sure. So how would both the as-if-argument proponents and its opponents respond to this view, which is promoted by cognitive and evolutionary psychologists? 

- Are there any serious and insightful meta-scientific formalizations of the relationship between economic models and the reality to which they apply?

(I am asking for admittedly selfish reasons, as I am working on related topics.)

Thanks,

Ola

>>>> send an email to decision_theory_forum+unsub...@googlegroups.com.

>>>> To post to this group, send email to
>>>> decision_t...@googlegroups.com.
>>>> Visit this group at
>>>> https://groups.google.com/group/decision_theory_forum.
>>>> For more options, visit https://groups.google.com/d/optout.
>>>
>>>
>>> --
>>> You received this message because you are subscribed to the Google
>>> Groups "decision_theory_forum" group.
>>> To unsubscribe from this group and stop receiving emails from it, send

>>> an email to decision_theory_forum+unsub...@googlegroups.com.

>>> To post to this group, send email to
>>> decision_t...@googlegroups.com.
>>> Visit this group at
>>> https://groups.google.com/group/decision_theory_forum.
>>> For more options, visit https://groups.google.com/d/optout.
>>
>> --
>> You received this message because you are subscribed to the Google Groups
>> "decision_theory_forum" group.
>> To unsubscribe from this group and stop receiving emails from it, send an

>> email to decision_theory_forum+unsub...@googlegroups.com.

>> To post to this group, send email to
>> decision_t...@googlegroups.com.
>> Visit this group at https://groups.google.com/group/decision_theory_forum.
>> For more options, visit https://groups.google.com/d/optout.
>>
>> --
>> You received this message because you are subscribed to the Google Groups
>> "decision_theory_forum" group.
>> To unsubscribe from this group and stop receiving emails from it, send an

>> email to decision_theory_forum+unsub...@googlegroups.com.

>> To post to this group, send email to
>> decision_t...@googlegroups.com.
>> Visit this group at https://groups.google.com/group/decision_theory_forum.
>> For more options, visit https://groups.google.com/d/optout.
>>
>>
>> --
>> You received this message because you are subscribed to the Google Groups
>> "decision_theory_forum" group.
>> To unsubscribe from this group and stop receiving emails from it, send an

>> email to decision_theory_forum+unsub...@googlegroups.com.

>> To post to this group, send email to
>> decision_t...@googlegroups.com.
>> Visit this group at https://groups.google.com/group/decision_theory_forum.
>> For more options, visit https://groups.google.com/d/optout.
>
> --
> You received this message because you are subscribed to the Google Groups
> "decision_theory_forum" group.
> To unsubscribe from this group and stop receiving emails from it, send an

> email to decision_theory_forum+unsub...@googlegroups.com.

> To post to this group, send email to decision_t...@googlegroups.com.
> Visit this group at https://groups.google.com/group/decision_theory_forum.
> For more options, visit https://groups.google.com/d/optout.
>
> --
> You received this message because you are subscribed to the Google Groups
> "decision_theory_forum" group.
> To unsubscribe from this group and stop receiving emails from it, send an

> email to decision_theory_forum+unsub...@googlegroups.com.

> To post to this group, send email to decision_t...@googlegroups.com.
> Visit this group at https://groups.google.com/group/decision_theory_forum.
> For more options, visit https://groups.google.com/d/optout.
>
> --
> You received this message because you are subscribed to the Google Groups
> "decision_theory_forum" group.
> To unsubscribe from this group and stop receiving emails from it, send an

> email to decision_theory_forum+unsub...@googlegroups.com.

> To post to this group, send email to decision_t...@googlegroups.com.
> Visit this group at https://groups.google.com/group/decision_theory_forum.
> For more options, visit https://groups.google.com/d/optout.
>
> --
> You received this message because you are subscribed to the Google Groups
> "decision_theory_forum" group.
> To unsubscribe from this group and stop receiving emails from it, send an

> email to decision_theory_forum+unsub...@googlegroups.com.

> To post to this group, send email to decision_t...@googlegroups.com.
> Visit this group at https://groups.google.com/group/decision_theory_forum.
> For more options, visit https://groups.google.com/d/optout.
>
> --
> You received this message because you are subscribed to the Google Groups
> "decision_theory_forum" group.
> To unsubscribe from this group and stop receiving emails from it, send an

> email to decision_theory_forum+unsub...@googlegroups.com.

> To post to this group, send email to decision_t...@googlegroups.com.
> Visit this group at https://groups.google.com/group/decision_theory_forum.
> For more options, visit https://groups.google.com/d/optout.
>
> --
> You received this message because you are subscribed to the Google Groups
> "decision_theory_forum" group.
> To unsubscribe from this group and stop receiving emails from it, send an

> email to decision_theory_forum+unsub...@googlegroups.com.

> To post to this group, send email to decision_t...@googlegroups.com.
> Visit this group at https://groups.google.com/group/decision_theory_forum.
> For more options, visit https://groups.google.com/d/optout.
>
> --
> You received this message because you are subscribed to the Google Groups
> "decision_theory_forum" group.
> To unsubscribe from this group and stop receiving emails from it, send an

> email to decision_theory_forum+unsub...@googlegroups.com.

[Itzhak Gilboa]

Hi Ola (and all),

Only briefly on your first point: I don't think anyone challenges the evolutionary stories -- neither the biological one nor the learning one.  The question is mostly how far we take them.  For example, I think we all agree that people care about material resources and about their social rank, and that these are innate preferences.  The divergence would be when we ask whether people evolved to behave as if they had a Bayesian prior over a universal type space or something of the like.

Btw, Bayes's argument for the existence of God (which necessitated a prior) is basically the Intelligent Design argument.  I therefore wouldn't be surprised if some of the devout Bayesians on this forum don't believe in evolution.  (Just a joke.  Some of my best friends are devout Bayesians.)

A minor serious comment: I'm not sure that the Gambler's Fallacy is in the same category as the rest.  I think it's a reasoning error that is mostly committed by people who have some vague idea, but obviously not perfect understanding, of the Law of Large Numbers.  But I don't think this phenomenon would be selected as a result of evolutionary forces and I doubt it existed before J. Bernoulli became so popular.

Best

Tzachi