Pat Gunning wrote:   
> Regarding Alan Isaac=E2=80=99s query, the logic of choice is as   
> applicable to making predictions as it is to understanding   
> past behavior.   
  
  
The contrast in my post was between the proposal that economics is a "theory of choice,"
which baffles me, and other proposals that I find somewhat easier to understand, such as
Michael's proposal that economics is the "study of constrained choice [including] the
constraints and the production of constraints."  And so I asked for "a prediction that
emerges from an economic model that is due to this 'theory of choice' independently of the
additional structure (e.g., constraints) within the specific model chosen to represent
that theory."  You have not offered this (yet).
  
I also suggested that those who propose that economics is a "theory of choice" (fine
sounding words) really mean something simpler (and not so fine sounding): "that if we
combine a folk understanding of human psychology (including very metaphysical beliefs
about agency and deep presumptions about preferences) with enough concrete particulars
about the situations in which humans are embedded then we get better at making guesses
about behavioral outcomes."  You have not indicated how economics is more than this, and
of course Alexander Rosenberg has argued from this starting point to the conclusion that
economics is an empirical dead end.  (I am not quite so pessimistic, but saying why is a
long conversation.)
  
I do not think shifting talk from a "theory of choice" to a "logic of choice" buys you
anything here, although it is certainly easier to understand how much of "economic theory"
is an exercise in logic than to understand how it is a theory of human behavior.
  
  
Pat Gunning wrote:   
> To make predictions, one builds images, or models, of the   
> future choices people will make under conditions that she   
> presumes will exist in the future.   
  
  
Actually we make predictions of human behavior in lots of ways that are very important,
quite successful, and not nearly so formal.  (Consider walking down the sidewalk, mailing
a package, crossing a busy street, or driving.)  In contrast, the literature I cited shows
that general economic models are able to derive essentially no predictions whatsoever.  As
we impose structure (in the form of constraints, special utility functions, representative
agents, steady state equilibria, constant market clearing, etc.) we slowly increase the
ability of a model to offer specific predictions.  Concretely, economics gains empirical
content the *further* we get from the pure logic of choice.
  
  
Pat Gunning wrote:   
> Economists build models of choices; it is what   
> distinguishes them as a species.   
  
  
As an empirical matter, I claim this is wrong.  Of course you may mean this as a criterion
of demarcation, a way of asserting e.g. that person running vector autoregressions is
actually not an economist, even if that person and everyone around them refers to them
that way.  Or perhaps it is a way of denying the Keynes was an economist.  Etc. If so, we
might (some other time) discuss the usefulness of such a criterion.
  
  
Pat Gunning wrote:   
> It may be impossible to "test a theory?"   
  
This does not correspond with how I understand the word 'theory'.  It also seems a
reversal on your part, since you claimed earlier that the logic of choice is "applicable"
(whatever that means) to making predictions.
  
Pat Gunning wrote:   
> There are no empirical laws in economics. There are only   
> logical ones derived from assumptions about the conditions   
> that people perceive and hypotheses about the choices they   
> make under those conditions.   
  
I am inclined to agree, but the profession is not.  E.g., consider the "law of demand." Of
course even this "law" is not implied by the "logic of choice", so it is purely an
empirical matter.  (And I am not even taking up the substantial issue of preference
stability.)  And so if economists are to make predictions about demand, this will be more
of a statistical exercise---anticipating the recurrence of past patterns in the data---
than an effort to test a choice-theoretic economic theory.  And it is always worth
remembering the nice demonstration of Becker (1962 JPE) that this law *can* appear as a
prediction---in an economic model where the logic of choice is essentially missing.
  
So, since you gesture toward Gossen, I propose that we reach this point wondering exactly
what bequest the "ingenious idiot" left economic science in his second law.
  
Cheers, Alan Isaac