RE: Reply to me and paul [VERY LONG]

[ecwfm@xxxxxxxxxxxxxxxxxxx (bill mitchell)]

THIS IS A LONG POST.

Paul: This is a difficult subject to reply to. It can range over a technical
discussion to a fairly esoteric philosophical discussion about absolutes,
truth, and existence and what is knowledge.

So here goes in not any specific order, a sort of non stationary response.

1) my initial retort to Paul was based on his assertion that CI analysis was
the "latest con" from econometrics. Con is an insulting word and conjures up
images of slippery dick types wanting to rip people off. That seemed an
unfortunate term to use.

I do not discard any knowledge. every bit helps. i clearly cannot be certain
what knowledge is. why should we believe that b/c it is a subject which has
been misused by the less informed, less well intentioned, plain crooks
whatever, in the past and probably still is, that it is always going to be
misused. a chain saw can rip a natural forest apart in no time if some
cretinous woodchipper is rampant, but can also cut some plantation trees
quickly for timber, more quickly than a hand axe.

Econometrics is a tool which can provide some additional knowledge if it is
in the hands of the right people. It does not test theories. How can we do
that. It tests econometric hypotheses. How they relate to economic theories
is a good question. It is no more religious to make these relations than it
is to believe in the economic theories in the first place.

economic theory has no exclusive or independent way of establishing its
veracity. like econometric inference it is just another source of knowledge.
But of-course, we like to say it is a prior knowledge and a good
econometrician should be guided by it.

But even that is only part of the story. Economic theory can say nothing
about dynamics. It can tell us nothing much about adjustment processes. It
can not distinguish between low frequency and higher frequency data in a
time series spectrum. econometrics can do those things. and they are useful
as additional sources of knowledge to do them.

I am not like Keynes (in his debate with tinbergen) who claimed he knew what
all the numbers were anyway without any statistical work. typically modest
claim from JMK. which leads me to the next point.

2) PK economists presumably are about policy action. Policies spend money
and allocate resources. There are at any time lots of different policies. I
would think that it is only practical to be able to provide policy makers
with some numbers about the policies, their maybe effects, and that sort of
thing. I quite frankly cannot see any way a PK economist can say the world
is uncertain and therefore stats is a messy game, leave well alone.

Who is going to come up with the numbers to decide which policy and how much
to spend? Econometric analysis is very useful for scaling and measuring
things like this. After all i cannot see any other alternative.

3) paul says:

I am not as ar
>Ah! My! if only CI researchers were as noble as Bill Nitchell who avoid
>talking about their results as revealing any "discovery or truth" . Bill
>says lets separate theory from practice and the practitioners admit that
>"any econometric analysis would [not] suggest the praxis is 100 per cent".
>But then Bill that impliess that it could be zero%! Where beteeen zero and
>100% would you put it?
>And on what statistical significance test would you rest your quantitative
>estimate of praxis?

well we all accept rules of the game in our lives. We accept that things
exist without knowing them. we accept grammatical conventions without
knowing why. we accept communication conventions - like on this list we use
First names and not our academic titles. so we also accept standards where
knowledge is deemed to be useful.

in stats, there is a well developed set of principles upon which inference
is considered sound. I know paul what you will say. It will be about ergodic
and non ergodic processes. Well i know about them.

but again i raise the issue that you declined to follow up a few weeks ago
about habitual uniformity in our behaviour. My claim is that the customs and
habits which dominate our daily behaviour also dominate our weekly and
quarterly and so on behaviour. Business firms use rules of thumb, so do
individuals.

This in no way disputes the essential claim which binds us on the same list
(i think) that uncertainty is endemic and there is no way of "knowing the
future". of-course there is not. but it is a fair bet that in the next short
period things will be highly persistent due to customs, institutions and habits.

of-course, also, a kobe earthquake can hit, or some oil sheiks decide they
haven't enough mercedes or bmws, and these events will interrupt the time
series we observe (imperfectly due to measurement problems). so what?
no theory or practice can cope with them. econometric models will be
interrupted by these events. so what?  i will do a stability test and i will
find my model is interrupted, cannot forecast out of sample. etc so what?

but the history of this century at least suggests to me that between these
shocks, things are stable (within the meaning given to it by
econometricians), and there is a basis for short period forecasts into that
mire of uncertainty based on models which reflect the rules of thumb
behaviour of the economic institutions.

what other game is there is town?

4) Paul challenges my use of plausibility:

>But my gosh Monetarism is also "plausbile". And Lucas would
>argue that Rational expectations" is a plausible (and the only scientific)
>way (and ingenious as well) of organizing the way the analyst thinks
>about economic processes. Yet Lucas in his 1981 JEL response to Tobin
>admits that rational expectations models are "patently false".

It was not plausible to me at all. The econometric evidence was appalling
and the less formal analysis also exposed vital flaws.

I only need to correlate quits with the business cycle, a low level
econometric task to see that they are overwhelmingly procyclical, instead of
the essential Monetarist-New Classical-RATEX requirement that they be
countercyclical. That simple operation then leads to more in depth
econometric and logical analysis, to dismiss the plausibility argument.

if we were to rely just on theoretical juxtaposition then the problem of
observational equivalence is endemic is distinguishing involuntary un. from
Keynes from a Pigovian-Classical view of un.

both theorise that the real wage will be lower when employment is higher.

5. Barkley came in and said:

>f course, the real issue here which has been raised
>by Ebert, Davidson, et al, is what does finding (or not
<finding) any CI mean?  The answer may be that it is more
>useful (Popperianly) for _rejecting_ hypothesized equilibrium
>or other linking relationships if no CI at any difference can
>be found.  It is not at all clear that finding a CI "proves"
>anything anymore than finding nice values for various
>statistics in an OLS regression "proves" anything.  Why
>are CI findings any less likely to be "spurious"?  Correlation
>is still not necessarily causation, in spite of Granger's
>highly misleading use of the term "causality."
>     To get at "causation," one must have theory, of one
>sort or another, both deductive and inductive (yes, Paul,
>I agree with you on that one).

As i said barkley (welcome back BTW) CI may just be interpreted as a finding
of statistical regularities or proportionalities between a set of time
series which has persisted for an extended period. it does not prove
anything. what is proof? econometrics does not seek to prove anything,
except if you are want to use the term to the derivation of properties of
estimators or test stats. but that is not the sense you raise the term. CI
findings are less likely to be spurious, b/c spurious is a technical term
and CI overcomes the technical deficiency.

what i say here is that of-course, the econometrics only fits into a wider
picture. it is only one piece of information we can use to make up our world
view, and pray (as it is religion) that it is right.

granger did not misuse the term causation. we all knew that it meant
correlation. regression relationships are always in one sense meagre
correlations. to think of them as causal we put in our theories
(prejudices). but i am more comfortable when my prejudices are also
correlated (:-))

6. paul again:

"You may wish to read L.J. Christiano and M. Eichenbaum's
 article "Unit Roots in Real GNP: Do we Know Do We Care?" in CARNEGIE-
ROCHESTER CONFERENCE SERIES ON PUBLIC POLICY, Vol. 32, Spring 1990, editor
Allen Meltzer. (Bill Mitchell please note). The use of such statistical
techniques is questionable,-- see the minisymposium on hysteresis especially
articles by Rod Cross, P. Davidson. D. Katzner in JOURNAL OF POST KEYNESIAN
ECONOMICS Spring 1993, vol 15."

I have read them all and the response is so what?

Paul again:
"Also there is a questionable theoretical issue-- namely are there
long calendar duration structural relationships (i.e., immutable economic
laws) in any nonergodic system."

is it theoretical or empirical or a bit of both. how do you know that all
the system is nonergodic and non CI in the conceptual sense. I do not b/c i
do ot know the future and only vaguely understand and observe the past.

why is it less valid to assume proportionately and feedback processes to
generate this proportionately, than it is too assume infinite variance
processes dominate economic systems. casually speaking, the latter do not
seem over 300 years or so to have been to infinite.

8. stuffed. time to get back to real world.

Kind regards
bill






**************************************************************************
 William F. Mitchell            Telephone: +61-49-215027      .-_|\
 Department of Economics                   +61-49-705133     /     \
 The University of Newcastle    Fax:       +61-49-216919     \.--._/*<--
 Callaghan   NSW  2308                                            v
 Australia                      Email : ecwfm@xxxxxxxxxxxxxxxxxxx

 WWW Home Page: http://econ-www.newcastle.edu.au/~bill/billyhp.html
**************************************************************************