Davidson's Seminar: Reponse to 3rd Day response

[6155GUASTELL@xxxxxxxxxxxxxxx]

From: Paul Davidson <pdavidso@xxxxxxxxxxxxxxxxxx>

>Stephen's examples confuse  epist. uncertainty with questions
>involving whether agents' utility functions are the same as
>classical theorists presume, i.e., do agents maximize actuarial
>values? For example Stephen's category of overweighting uncertainty
>-- has nothing to do with uncertainty- IT INVOLVES PEOPLES
>PROBABILISTIC RISK AVERSION PROPENSITY.

I raised the K-T issues initially in response to the remark you
had somewhere in the paper about limits to rationality, which appeared
confined to computational limitations, and, it appears, knowledge
of what to compute. Cognitive psyc has uncovered several other
limites to rationality which I itemized.

I also suspected that the cognitive explanations are necessary to
explain WHY, or at least HOW uncertainty is translated into
liquidity/consumption/investment patterns. Once the HOWs and
WHYs are straightened out, then global/macro policies might
be effective at forestalling perverse consequences such as
one nation that doesn't save enough (USA by most standards) and
another that doesn't spend enough (Japan per recent discussions here).
It may not be so important for forecasting future patterns based
on present habits, demographics, other macro indicators, and so forth.

>When the agents KNOW
>the probability associated with gain and the probability associated
>with loss, and are willing to "forego a portion of" the actuarial
>gain, means that their utility function does not search solely for
>maximizing expected returns. It has nothing to do with UNCERTAINTY.

It most certainly IS uncertainty (pardon the pun). The agents
just find a way to dump the uncertainty on somebody else, e.g.,
an insurance company who is in the business of doing that sort
of thing. In other sorts of contracts, agents externalize
risks in the same manner as they would externalize costs; it's
a popular sport.

Should probably add that most insurance buyers don't have a
grip on the actual probability of bad things happening that
the insurance companies have. Their insurance buying habits are
pretty consistent with K&T explanation of what happens when
overwieghting certainty and reflect get together.

>In fact, Stephen implicitly recognizes this in his "reflecting
>effect" -- which is "not the case of perfect rationality" where by
>perfect rationality, Stephen must mean the classical maximization
>of return. Stephen specifically says "the odds [probabilistic risk]
>of loss are overweighted in the evaluation of an investment
>decision".

Yes.

>In other cases, Stephen is merely suggesting the problem is short-
>run epis. uncertainty where subjective probabilities do not equal
>objective probabilities. For example Stephen's cases of
>"overestimating probabilities" and "underestimating probabilities"
>are merely cases where estimated subjective probabilities differ
>from the objective probabilities associated with an ergodic
>reality.

This part is OK too.

>Stephen is too enamored with Kahneman-Twersky to see that they
>misuse language. In all the experiments of K-T, the success and
>failure (or payoffs) associated with different choices are
>predetermined by the experimenters.

I don't think K&T are misusing language at all; see above.
What I think Paul is targeting is the difference between
probabilities and fuzzy probabilities, the latter being more
uncertain because we don't know what the odds of an event
really are. There is a new line of research going on that
is addressing this limitation of the K&T work.

>Even if they tell the decision-
>makers the "odds" so that no epist. uncertainty exists (assuming
>the decision maker believes the experimenter), the decision-makers
>often do not reveal preferences that classical utility maximization
>claim -- rather hey fall into one of the just discussed categories.

>In conclusion, Stephen is just wrong when he claims the K-T
>"verified" experiments provide  a foundation for a hybrid theory.

I still thing it's at least part of a foundation. Will agree
that the cement is a little wet.

>ALL the K-T "verifications" show is that the classical case is a
>special case that requires additional restrictive assumptions
>(including what is being maximized) to reach its "hard-headed"
>"scientific" conclusions.

>Stephen also goes into a long discussion of control operators of
>chaos theory. As I have already said several times on this net,
>control theory requires Liouville's Theorem which is also the
>starting point for ergodic theorems.

If what you're saying is that the controller must be at least
as complex as the thing it's trying to control, then we don't
have a problem on this point.


>Finally Stephen uses languge usually associated with Bill (when
>reasoned discourse fails) to denigrate my example of the Univ. of
>Wisconsin study of the aged consumption patterns. Had Stephen read

Actually, the word "shit" referred to the stuff that people buy.
Madison avenue has been trying to get us to buy stuff we don't
need for years. Or telling us we have needs that we didn't think
we had yesterday.

>the paper he would have seen that this was a  cross-sectional
>studiy of thousands of households. It showed that at ANY income
>level including the very poor (after adjusting for all sorts of
>socioeconomic variables) the marginal propensity to consume was
>inversely related to age.  As such, this study  has nothing to do
>with Stephen's belief that the aged are satiated  ( or in his
>earthy language:"They just don't need to buy any more shit at their
>stage of life").

I don't have a problem with the finding that consumption is
inversely related to age. The point I was trying to make is
you can't go from that finding to a CAUSAL statement that lower
consumption was a result of concern over future health costs.
There are also other factors contributing to their lower
consumption.

Yet another is a different kind of uncertainty
having to do with living on a fixed income on the one hand
and expecting the cost of living to rise on the other, which
it always does. Perhaps we need to separate "uncertainty"
from "certainty that the future will be grim." This issue
does speak to the point earlier about knowing probabilities
and not knowing them.

--Stephen Guastello