Mises: On Subjectivism, Measurement, and Aggregates

[From Mike Acree (2001.01.18.1551 PST)]

Nearly a year ago Fred Nickols suggested that Ludwig von Mises' _Human
Action: A Treatise on Economics_ might be worth looking at from a PCT point
of view. He was struck by statements in the opening pages about human
action as purposeful behavior, impelled by some sense of "uneasiness" (which
could be read as error). It has taken me this long to read the book and
write up a few impressions. I don't propose to discuss here Mises' theory
of economics per se, but rather his approach to economics--his
metaeconomics, as it were. There are several interrelated features of his
approach that set his theory sharply apart from prevailing approaches to
economics, in just the same ways that PCT is set apart from prevailing
theories of psychology. These include subjectivism, prescission from
measurement, and a dynamic analysis of individuals rather than a static
analysis of aggregates. They also include, notably, a pragmatic, value-free
approach to the subject. Mises contends that reasoning from natural law or
right is idle; the only question is what works for what ends.

Though Mises doesn't discuss the history of measurement theory, it is
interesting to note that the modern debate on measurement was instigated by
developments in economics. The rapid monetization of the economies of
Western Europe in the 11th and 12th centuries, as I discussed in my talk to
the 1994 conference in Durango, led to the emergence of the price system.
This appears to have been the origin of the idea that everything could be
measured, just because everything was found to have its price--even
salvation, through the sale of indulgences--without prices being determined
by anyone in particular. So far as I know, the psychologists and physicists
participating in the measurement debate in the early decades of the 20th
century were unaware that they were recapitulating the same debates at the
universities of Oxford and Paris 600 years earlier. The debate over the
meaningfulness of psychological measurement was never won. A new
technology--the Likert scale--simply allowed questions of meaning to be
swept under the carpet. Just ask people to assign a number to their
experience and most of them will comply. The rest is (supposedly
unassailable) mathematics. Questions of meaning arose only several decades
later, in relation to racial differences in intelligence. (Terman's group
at Stanford found they had to revise their original version to bring women's
IQs _down_ to the level of men's--everybody knew there couldn't be a gender
difference, at least in that direction. On the other hand, racial
differences, being in the expected direction, were allowed to stay.)

Mises and his younger brother Richard (who at least used to be better known)
stood on opposite sides of the measurement debate, most explicitly on the
measurement of probabilities. Richard famously tried to justify frequency
interpretations of probabilities like that of Al Gore's winning the White
House in 2004. Ludwig saw these as qualitative and subjective; there wasn't
any obvious class in which to embed them to derive a relative frequency,
Richard's strained attempts notwithstanding.

In economics, Mises (Ludwig, henceforth) argues that values, the basis of
human action, are intrinsically subjective and incapable of numerical
measurement. It is often supposed that prices measure values, but Mises
demurs. Exchanges, which determine prices, express preferences. If I pay
$30 for a truffle, it's because I _prefer_ the truffle to the $30, at this
moment, and the grocer _prefers_ the $30. The relation is not one, as is
commonly assumed, of indifference. The values in question are not
susceptible of more than ordinal measurement.

In addition to subjectivity, measurement is frustrated by heterogeneity and
change. Labor is obviously heterogeneous, but buyers and sellers, Mises
also points out, distinguish qualities and grades of peas, making index
number methods idle. Moreover, he argues, there are no constant relations
between economic elements to be expressed in equations. If a 10% rise in
the supply of potatoes in Colorado is followed by an 8% decrease in price,
we can't generalize this relation to other times and places. All prices we
know are past prices; statistics can deal only with historical data, not
understanding. Mathematical economists treat equilibrium as a real rather
than a limiting notion; equilibrium is in fact never reached--every change
brings new changes in its train. Their differential equations represent
only a superficial analogy to market process.

For all of these reasons the quantitative analysis of aggregate data is
specious. We actually know nothing of the shape of supply and demand
curves, only a temporary intersection point. Mises also argues that the
concept of national income or wealth is meaningless. The wealth of an
individual or group of individuals can be calculated in terms of how much
money their property would bring in sale on the market. But the whole
country (whatever exactly that would mean) can't be put on the market at
once. Rick (2000.03.06.1340) asserted, in response to my preliminary post
on Mises, that the analysis of aggregates was indeed the object of
economics, just as virtually all textbooks of psychological research assert
that the object of psychological inquiry is aggregates--populations. Mises
might say that starting with that assumption begs the question. Both Mises'
Austrian economics and PCT are capable of comprehending aggregates, but they
do so in terms of individual behavior. At least that is my impression of
PCT models of the movement of flocks and crowds.

I am not sure I have represented Mises fairly in this methodological
synopsis; there are some points, as I've expressed them here, that I would
have trouble defending. In his rejection of mathematical analysis, Mises
was nevertheless more extreme than many economists of his Austrian school.
PCT, so far as I know, has taken no correspondingly strong stand on the
meaningfulness of measuring psychological constructs like anxiety or
self-esteem or intelligence, presumably because there was no compelling need
to. The more important task was to build a positive theory.

Historically it looks to me as though philosophically skeptical approaches
like Mises' are strategically doomed. The technologies of psychological
measurement and statistical analysis have created a huge body of work which
serves powerfully as its own self-justification. Doubts about the
meaningfulness or utility of the whole enterprise simply will not be
entertained as reasonable questions. Similarly with the huge business of
mathematical economics. I think serious questions can indeed be raised
about what understanding has been achieved in either case. On the other
hand, there is also reason to believe that attacks on or criticisms of these
massive projects is a waste of time. They supply employment for a multitude
of academics (like myself), who constitute a powerfully vested interest
group. I don't believe most of this work, at least in psychology, would
ever have been supported by the private sector; but its continued funding by
government looks secure, barring a general economic collapse. Given the
relative inertia and monolithic quality of the science bureaucracy, I think
it will be awhile before most of us are doing more useful work.
Psychologists and economists have been so long accustomed to _playing
scientist_, as David Bakan put it--and getting away with it--that real
scientific work--for example, with PCT--will look unattractively difficult.

Mike

[From Bill Powers (2001.01.21.0422 MST)]

Mike Acree (2001.01.18.1551 PST)--

Nearly a year ago Fred Nickols suggested that Ludwig von Mises' _Human
Action: A Treatise on Economics_ might be worth looking at from a PCT point
of view. He was struck by statements in the opening pages about human
action as purposeful behavior, impelled by some sense of "uneasiness" (which
could be read as error). It has taken me this long to read the book and
write up a few impressions.

Glad you did, although I'm not erudite enough to catch all of your
allusions. I was disappointed to discover that "prescission" was a
misspelling, at least according to my dictionary which omits it. It just
_looks_ like a useful word.
I guess you meant precision but had a couple of surplus s's to get rid of.

Measurement, in PCT terms, is expressing one perception in terms of
another. This came up in connection with the Weber-Fechner and Stevens
scales in which logarithmic or power scales were used to relate different
perceptions. As I saw it, what was missing was an objective scale against
which all subjective scales could be measured. If all perceptions contained
an unknown scaling function relating them to the objective reality, this
fact would never be discovered using the accepted techniques.

It's been assumed that a numerical estimate is a standard that can be
applied across modalities as an absolute measure of the magnitude of a
subjective impression. But exactly the same problem exists: all you have to
do is ask "how big is 3?" to see the problem. If one unit of one sensation
is deemed to be as large as 3 units of another, in terms of laboratory
measurement on a physical scale, all we get from that is the ratio of
magnitudes of the sensations in units of the size of a number. The "actual"
size of those units remains unknown. All psychophysics can ever give us is
relative magnitudes: the principle is that of _relative_ measurement, as we
could call it.

... the price system.
This appears to have been the origin of the idea that everything could be
measured, just because everything was found to have its price--even
salvation, through the sale of indulgences--without prices being determined
by anyone in particular.

Prices can be _set_ by one person, but a _workable_ price can be set only
by two people -- _any two_ rather than _anyone_. The workable price is that
which results in the the greatest decrease in error in the system of two
people. If the total error in both people would be increased by the
transaction, it does not occur: no deal. I think that might be a good
PCT-based principle of economic transactions.

There's another principle behind a price system which I think is taken for
granted: it is that price is independent of the individuals who are
involved in a transaction. That is, if I charge $10 to cut Charlie's hair,
I cannot charge Joe, next in line, either $8 or $12. Even if I charge Joe
$9.99, Charlie is going to demand a refund of a penny.

The only way for this principle _not_ to hold, I would guess, is for people
to agree not to reveal the real price for which they bought something. When
you think about it, how could there be any bargaining if each person knew
exactly what each other person paid for a given item? The only way would be
to obscure the actual price by embedding the particular good in a crowd of
other tangible and intangible goods or services: Yes, I charged Ahmed only
3 shekels for that camel, but the camel was old and also sick, and Ahmed
also agreed to rake my garden next week. I am charging you only 5 shekels
for this young healthy camel, and my garden doesn't need raking again.

Now that I elaborate on that a bit, I can see that the concept of price is
an illusion anyway. There is a certainly an average price for a 1998
Oldsmobile 88: you find it by adding up what people paid for such cars and
dividing by the number of cars sold. But when you start analyzing the price
into base price, MSRP, price at various ARPs over various payment times,
price with and without accessories and extras, and This Week Only prices,
you find practically as many prices as transactions. Knowing "the price" of
this car, you can't deduce how much any given purchaser paid. Same old
principle: you can't deduce individual characteristics from mass measures.

In economics, Mises (Ludwig, henceforth) argues that values, the basis of
human action, are intrinsically subjective and incapable of numerical
measurement. It is often supposed that prices measure values, but Mises
demurs. Exchanges, which determine prices, express preferences. If I pay
$30 for a truffle, it's because I _prefer_ the truffle to the $30, at this
moment, and the grocer _prefers_ the $30. The relation is not one, as is
commonly assumed, of indifference. The values in question are not
susceptible of more than ordinal measurement.

Which is to say, relative measurement. Ordinal measurement only lets you
sort things by size; it doesn't tell you even the relative sizes of the
intervals between 3rd and 4th on the one hand, and 9th and 10th on the
other hand.

In addition to subjectivity, measurement is frustrated by heterogeneity and
change. Labor is obviously heterogeneous, but buyers and sellers, Mises
also points out, distinguish qualities and grades of peas, making index
number methods idle. Moreover, he argues, there are no constant relations
between economic elements to be expressed in equations. If a 10% rise in
the supply of potatoes in Colorado is followed by an 8% decrease in price,
we can't generalize this relation to other times and places. All prices we
know are past prices; statistics can deal only with historical data, not
understanding. Mathematical economists treat equilibrium as a real rather
than a limiting notion; equilibrium is in fact never reached--every change
brings new changes in its train. Their differential equations represent
only a superficial analogy to market process.

I see I have been anticipated. However, the ensuing argument is weak,
because it can be applied to anything. There are no constant relations
among different sound pitches; a 10% rise in sound pitch from an oboe is
followed by an 8% decrease in pitch, but we can't generalizes this relation
to other times and places. All we know are past pitches; statistics can
deal only with historical data, not understanding.

You can see the falsity of this argument when applied to the case of the
oboe. All that our inability to predict the oboe's pitches shows is that
we're using the wrong model. If you know that the oboe player is playing
his part of "Rhapsody in Blue," you can predict very well how the pitch
will change at various places in the score, today, next year, or any year
after that, in Chicago or Philadelphia. Mises is really declaring system
analysis to be impossible, and that is simply wrong.

For all of these reasons the quantitative analysis of aggregate data is
specious. We actually know nothing of the shape of supply and demand
curves, only a temporary intersection point. Mises also argues that the
concept of national income or wealth is meaningless. The wealth of an
individual or group of individuals can be calculated in terms of how much
money their property would bring in sale on the market. But the whole
country (whatever exactly that would mean) can't be put on the market at
once.

What Mises is missing here is the idea of a working model. To construct and
then validate a working model, you need past data. Once the model can
explain past volume and price changes, you can test it as a predictor of
future volume and price changes. There is nothing meaningless about this.
Of course in constructing a model one may find that certain concepts are
useless: "wealth" could be one of them. But one can't generalize from this
to conclude that all analysis is useless.

Rick (2000.03.06.1340) asserted, in response to my preliminary post
on Mises, that the analysis of aggregates was indeed the object of
economics, just as virtually all textbooks of psychological research assert
that the object of psychological inquiry is aggregates--populations. Mises
might say that starting with that assumption begs the question. Both Mises'
Austrian economics and PCT are capable of comprehending aggregates, but they
do so in terms of individual behavior. At least that is my impression of
PCT models of the movement of flocks and crowds.

Yes, I think that's the right impression. There are ways of comprehending
aggregates that are meaningful, and others that are meaningless. The
meaningless ways are those that simply deal with numbers, lacking any model
to suggest why aggregate order is found. The phase of the Moon and the
price of cheese may indeed correlate significantly, but without some reason
for this correlation, this is a useless fact. In the Crowd program, on the
other hand, you might very well see a correlation between turning radius
and distance between moving objects, and in fact this correlation
necessarily would follow from the way each control system is organized: if
you didn't see it, something would be wrong with the data or your way of
recording it.

Historically it looks to me as though philosophically skeptical approaches
like Mises' are strategically doomed. The technologies of psychological
measurement and statistical analysis have created a huge body of work which
serves powerfully as its own self-justification. Doubts about the
meaningfulness or utility of the whole enterprise simply will not be
entertained as reasonable questions.

If you treat problems of acceptance as laws of nature you might as well cut
your throat right now. If the way things are determines the way they will
be, why bother trying to change the way things are? Why even bother to
describe the problem, unless you just enjoy complaining? I'm unsympathetic
with gloom and doom pronouncements about the impossibility of change.
They're self-fulfilling: to accept them is to work actively against change.

Given the
relative inertia and monolithic quality of the science bureaucracy, I think
it will be awhile before most of us are doing more useful work.
Psychologists and economists have been so long accustomed to _playing
scientist_, as David Bakan put it--and getting away with it--that real
scientific work--for example, with PCT--will look unattractively difficult.

I know, its much easier to sit in a relaxed position and analyze things to
death than it is actually to set up and carry out experiments which are
always accompanied by painful surprises. But so what? That just separates
the men from the boys, or if that seems sexist, the sheep from the goats.
Some of the experiments with PCT have been done, though the number of
people who understand what they tell us is still small. A lot more remain
to be done, but they will be done some day and then a few more people will
understand a few more things about control in living organisms. The rest of
the yammering crowd can just go play with themselves, which is what they
are mainly doing anyhow. Why should we care about that? The people we
should care about are those who have grasped the experiments, learned to
set up their own, and are doing something to further our understanding.
It's nice that there are people sympathetic to the cause who are willing to
support it and spread the word. But what really matters is that there are
some people who don't care how hard it is to do real science, and who
therefore go ahead and do it instead of making self-fulfilling prophesies.

Best,

Bill P.

[From Mike Acree (2001.01.22.1043 PST)]

Bill Powers (2001.01.21.0422 MST)--

Thank you very much for a splendid reply. In truth I was not necessarily
hoping to start a thread--more to end one, belatedly--but I'm gratified that
my post gave you the occasion for some further elaboration on these issues.
I agree wholeheartedly with everything you said in the entire post. What
you say about measurement is consistent with what I had read or inferred
from your writings; what you say about prices, as well as measurement, is,
for better for worse, I think, very much what Mises said or would say.

The criticism you make of Mises is also exactly the same one that occurred
to me: that Mises appears to be declaring analysis impossible. I'm certain
he didn't believe that, but I haven't run across an acknowledgment of the
difficulty, either. I don't see it as a fundamental flaw in his theory, but
more a matter of his having insufficiently qualified some of his assertions.
That was the main thing that made me uncertain I was representing him
fairly.

I'm delighted to be taken to task for my pessimism, as two other people did
in private responses. I do think of myself as a cheerful cynic; but my
original point was to be just that I wouldn't expect it to be very effective
if, say, you spent most of your time critically analyzing mainstream
approaches to psychological research, as opposed to developing your own.
Your own balance seems to me an excellent one, even as I marvel at your
patience with critical analysis of behaviorism. But I can see that I got
carried away. In another forum, I recently wrote, addressing pessimists:
"Pessimism is hard to argue against; the world is a hell of a place, for
sure. But it is simply stupid, on the other hand, to argue for it. We can
rarely be sure how much we can accomplish until we have tried. What on
earth do we stand to gain by arguing that action is futile? The only
possible pay-off in arguing in advance for our own impotence is just the
satisfaction of proving ourselves right."

Thanks again.
Mike