[From Bill Powers (2009.09.05.1230 MDT)]
Rick Marken (2009.09.05.0950)–
RM: You don’t have to do
anything fancy to improve your chances of observing a statistic (r) that
will allow you to reject the null; just increase the sample
size.
BP: But if your second sample is simply an exact copy of the first sample
(i.e., not new observations independent of the first set, but just the
first set written down again) does that count as increasing the sample
size? That’s what I think Martin L. is talking about – though I’m
exaggerating his point here by saying that the two sample sets are
identical rather than just correlated by being partly determined by some
common extraneous influences like seasonal variations.
RM: So an r of .2 might be
reported as significant at the .00001 level in one report and only at the
.05 level in another but now it would also have to be reported that this
correlation means that one variable accounts for only 4% of the variance
in the other in both cases!
BP: I like that. Good example. But I already know the difference between
significance and importance.
BP: Martin L. is saying some things about data that raise my
eyebrows. If a theory doesn’t predict what actually happens, how can it
be considered right? That is starting to sound theological.
My point as well!
BP: Right, the statistics are being used correctly in Runkel’s
“casting nets” mode. However, if a policy that economic theory
predicts will improve the economy in the aggregate has a 33% chance of
making it worse, should we put that policy into practice?
Wait, you switched from data to theory. I presume you meant to say: if a
policy that economic data predicts will improve the economy in the
aggregate.
I don’t think data predict anything. They just describe observations. A
occurs, then B occurs. If your theory is that A causes B, you predict
that making A occur again will cause B to occur. But if your theory is
that some unknown C is causing both A and B, you will not make that
prediction, using the same data. The behavioral illusion,
remember?
Whatever, I’m not using my data
to recommend policy. I’m using my data to show that there is no basis for
economists to recommend against a policy of raising taxes as recession.
BP: Yes, and you can use the same data to show there is no basis for
supporting a policy for raising taxes. There is no basis for drawing any
conclusion. If you want to raise taxes, or lower taxes, it can’t be
because of what the data tell you.
RM: Since no one has a good
model of how the economy works, all policy makers are shooting in the
dark. But there seems to be a consensus among economists, based
apparently on nothing but theology, that one of the possible shots we
might take – increasing taxes, especially on the upper brackets – is
recessionary. All I’m trying to say is that these people are full of some
serious bull. Raising taxes may not help. But there is absolutely no
empirical basis for saying that raising taxes will hurt.
BP: Nor that it will help, as you said. I agree that economists base
their recommendations on inadequate evidence. How does that show their
recommendations are manure? Or that yours aren’t?
BP: The reason that arguments about economics get so passionate is
that the arguments on all sides are so ineffective. That’s good control
theory.
Sounds like a tautology rather than control theory.
Not at all. If your action is only weakly connected to the variable
you’re trying to control, you will have to produce a lot of output to
have the effect you want – and if that doesn’t work, you will raise the
output gain some more.
An increase in the vehemence of an argument shows that the outputs are
being increased because they are not having the desired effect. You then
have to make them even more vehement, or give up the argument. Excessive
heat in an argument shows that the arguments are failing.
Rm: I suspect that even a
perfect model will not keep some people from ignoring the results. If the
model shows, for example, as I think it will, that increases in wealth
discrepancy lead to decreases in economic performance, the people who are
against any form of wealth redistribution because it’s
“socialist” will be out in the street with their guns drawn
just as fast as they are now.
BP: I don’t believe that increases in wealth discrepancy lead to
decreases in economic performance. I don’t believe the opposite, either.
What data led you to your conclusion? How do you know what the model will
show? And if you have already make up your mind about that, shouldn’t you
recuse yourself from the model-building effort? What if you begin to
suspect that the model is going to show something else?
That’s good control
theory; people will protect their perceptions from disturbance. If people
want to perceive “capitalism red in tooth and claw” as the
ideal economic system then that’s what they are going to see, facts and
models be damned. I think it’s great to build an economic model that
predicts the data perfectly. But I predict, based on control theory, that
when we have that model, it will be to the “free marketers” of
the world like water off a ducks back.
So all the people who believe in free markets are prejudiced and
impervious to reason? You anticipate universal stupidity from
free-marketers. I predict that you will stick to that expectation no
matter what happens, which makes you just like them, just another control
system controlling for predetermined truths. You’ve declared your
position – doesn’t that mean you have to defend it now?
BP earlier: After the reasoning fails, the yelling starts, and then
the guns come out.
RM: My experience dealing with ideologues is that the reasoning (and
certainly the consideration of the data) never really starts.
BP: What makes you different from them, except for the particular
ideology you’re expressing?
BP earlier: As far as I’m concerned, I won’t know the truth about
economics until I see a working model that correctly represents what has
already happened and correctly predicts what will happen if various
policies are put into effect.
Me too!
BP: Wait a minute, “Me too” what? You’re speaking as if you
know what the model will show us, but what if it shows that you’re wrong
and the other side is right about some things? If you take a position
now, isn’t that going to make it hard to evaluate the model correctly? I
was saying I don’t know the truth, and that I am committing myself to the
model-based approach which I’m trying to do without hoping or wishing for
it to turn out any particular way. You say “Me, too” but it
doesn’t sound as if you’re doing that.
BP earlier: So far I haven’t seen or been able to devise such a
model, so all the confident pronouncements about economic truths that I
hear and read appear hollow to me.
RM: I agree. But we are living in the real world here and we have to try
to act in ways that make things better. Since we have no model, I
recommend (for the time being) the e. coli approach that I described
earlier; if a policy seems not to be working (and it’s clear that
virtually all the policies implemented over the last 8 years are not
working)
BP: It’s not clear to me. I can think of lots of things I didn’t like
about the last 8 years, but those are just my own reactions and have
nothing to do with Truth. There were lots of other people who were quite
happy with what was going on, with equally little reason except for their
own personal preferences. Whose preferences count the most? You see
errors and want to reorganize. What about others who don’t see errors,
and don’t want to reorganize?
RM: then change it and see how
things go. If the change seems to make things better then stay with it;
if not, change again. I presented the data on taxes and recession to show
that there is no empirical basis for eliminating tax increases on the
wealthy as a policy option, even during a recession.
BP: But you haven’t shown the opposite, that we should maintain taxes on
the wealthy. Just showing that an argument is baseless doesn’t
automatically mean that its opposite must be true. Reorganization is
based on error, so before you reorganize you have to determine that there
is an error. If you just reorganize because you don’t know what to do
next, you could be throwing perfectly good control systems away. You
still might get somewhere, because there’s always the chance that you’re
right. But reorganization is slow and inefficient, and it’s always better
to have a systematic way of solving problems; in fact, finding those ways
is supposed to be one of the outcomes of reorganization.
BP earlier: And so do the criticisms of economic theories, because
the people arguing on your side don’t have a working model, either.
RM: I don’t have a side. Again, I’ve been trying to be up front about the
fact that I have no better idea of how the economy works than anyone
else. All I am doing is presenting data that shows that there is NO
EVIDENCE that a policy of increasing the upper marginal tax rate is
recessionary; so there is no reason to avoid trying this policy at this
time, especially since we apparently could use some revenue to balance
the budget.
BP: But haven’t you just stated which side you’re on? You say there’s no
reason not to try it, but is that a reason to try it? There are a lot of
people saying that they can think of reasons not to try it. Do we just
brush them away? It doesn’t sound to me as if you don’t have a
side.
BP earlier: Nobody has a
verified working model; that’s why it’s a waste of time to argue. If
anyone wants to know how the economy works, work on the model. Otherwise
the arguments pro and con are built on sand.
RM: But I’m not arguing about how the economy works. I’m arguing about
one little thing: that the data does not seem to support the idea that we
should not raise taxes during a recession.
BP: Nor does it support the idea that we should raise taxes during a
recession. You’re trying to make knowledge out of ignorance. If you can’t
prove something is wrong, it must be right?
RM: I have no idea whether
raising taxes will make things better or worse. I’m just saying that I
see NO EVIDENCE that this policy has ever been recessionary. I’m waiting
to hear from those who believe that taxes are recessionary, why they
think so. So far, I’ve heard only that this belief is based on
theory. Which, of course, is like saying “it’s because I say
so”.
BP: And there you are saying we should try raising taxes, or keep them
raised, because you say so. Your willingness to proceed as if raising
taxes is not recessionary is no more justified by the data than the other
side’s reluctance to raise taxes because doing so would be
recessionary.
All my objections here have to do with clinging to previous arguments and
positions instead of abandoning them and trying to develop a model which
will give us some basis for drawing conclusions. Wouldn’t it be enough
just to know that we are working our way toward a real basis for making
decisions, and a solid understanding of how economies work? If a project
like that were under way, couldn’t we simply drop whatever half-baked
ideas we had been having and look forward to finding out the
truth?
I don’t know if you remember the days before the first moon landing.
There were furious debates about what the astronauts would encounter.
There was one camp which argued that the first step into the surface of
another world would end up with the man (or the landing module)
disappearing into a fine dust layer several kilometers deep, never to be
seen again. There were arguments about finding lava or volcanos,
arguments about just about everything anyone had imagined about the moon.
And all this only about a day or two before the astronauts would actually
get there and find out what the conditions actually were. I simply
couldn’t imagine why anyone was arguing at all, when all they had to do
was wait a few days to see the answers.
If I thought a model was really going to be developed, and this isn’t
quite an empty dream yet, it wouldn’t matter to me what opinions about
the economy anyone had, including my own. I’m a little discouraged, as I
was the last time, when arguments start to develop before the model is
even started. This pretty much tells me that the people doing the arguing
are not really going to try to help develop the model, or that if they
did, they would be so prejudiced that they wouldn’t contribute much
anyway.
I wish we could just start working on the model, not worrying about where
it will lead but just trying to get it right. That’s all we really need
to do.
Best,
Bill P.