September 6, 2013
-
Some Anvils Need to be Dropped
JDN 2456534 PDT 15:58.
A review of Debunking Economics: The Naked Emperor Dethroned?
by Steve Keen.
The basic message Keen is trying to
send is a vitally important one: Neoclassical economics is failing.
Models based around rational agents and static equilibrium simply
fail to represent the real world, and and as a result give
policymakers a false sense of security against economic crisis.
The way Keen delivers this message is
by avalanche: Page after page, chapter after chapter, he tears apart
neoclassical economics piece by piece. His goal, indeed, is not to
show that the theory is wrong—refuting even a few assumptions
or equations would do that—but rather to show that it is rotten
to the core, that virtually every assumption and every equation is
defective.
And this, I think, is Keen’s greatest
failing. Some of the walls he tries to tear down are stronger than he
imagines them to be, and this draws attention away from the very real
gaps in the walls of the citadel.
He also has this weird obsession with
"what they originally meant"; he clearly knows a great deal
about the history of economics and economists, and so I’m inclined to
think he’s basically right about what Keynes, Marx, Von Neumann, etc.
originally meant; but so what? These men were brilliant, and many of
their insights are useful, but they were still wrong about a lot of
things. Even Darwin and Einstein made mistakes, and Marx was no
Darwin.
The first two chapters are
introductory material, in which he appeals to the reader (in a rather
melodramatic fashion) to be the change that economics needs.
In chapter 3, he examines the
theory of consumer behavior, and in particular focuses on several
mathematical results showing that preferences cannot be aggregated.
This is of course absolutely correct; the notion that we can get the
preferences of society by simply adding up the preferences of
individuals is profoundly naïve, and indeed the whole point of
Arrow’s
Theorem is that you can’t
do that. This certainly does damage neoclassical economics severely,
since the whole "New Keynesian synthesis" of modern
macroeconomics is based upon aggregated demand functions that are
simply the sum of individual demand functions.
The problem can actually be
avoided by accepting something classical economists did but
neoclassicists fear: cardinal
utility. If utility
isn’t just an ordering, but actually a measurable value, then
aggregation becomes much easier. Indeed, as I often like to point
out, Arrow’s "Impossibility" Theorem simply assumes a
priori that you’re not allowed
to do range voting, even though range voting is in fact a real system
that real institutions (like Amazon.com and much psychology and
sociology research!) use with great success. Include range voting as
an option, and the "impossible" conditions Arrow presents
turn out to be actually quite easy to satisfy. They still don’t
guarantee a negative-slope demand curve, but so much the worse for
negative-slope demand curves, because we know for a fact that some
goods increase in aggregate demand when their price rises. (Gold,
oil, diamonds…) Mathematically that system is unstable? Yes, yes it
is; and so are markets for these commodities. That’s an
empirically valid prediction.
Now, the original reason
cardinal utility was abandoned was that it appears unmeasurable; what
are the units? How do you detect utility? Well… actually it’s not
that hard. How many papercuts would you be willing to endure for a
delicious meal? How many electric shocks would you accept to get an
HDTV? If those examples sound silly, how about these: How many hours
would you be willing to work, in order to buy a nicer car? How long a
commute are you willing to bear to save $100 on monthly rent? These
are questions that we not only can answer, we do and
must answer in our
daily lives. If your answers
aren’t precise, chalk that up to the inherent uncertainty in the
universe. At the very least you should be able to place some bounds
on your answer: The car is not worth less than 100 hours of work
(spread over a year), nor is it worth more than 100,000. I’d
definitely accept another 10 minutes to save $100, and I definitely
wouldn’t accept another 2 hours.
Chapter 4 is the chapter Keen
says drew the most criticism in the first edition, and I can see why:
It’s clearly wrong. He
tries to argue that there is no difference in pricing between
competitive markets and monopolies, based on the fact that the
aggregate total of firms would maximize their profits if they acted
like a monopoly.
Well, yeah, of course they
would; they’d be effectively a monopoly. The
whole point of perfect competition is that the firms can’t collude
effectively, so they make their decisions independently. With that
assumption in mind, you can show mathematically that they’ll charge a
much lower price than the monopoly would. In effect, they are trying
to exploit their competitors, and being exploited in return.
Keen does point out something
rather important: The slope of the marginal demand curve for a
competitive firm and the slope of the demand curve for a monopoly are
exactly the same. That
is indeed exactly right; since the firms act independently, producing
one more unit leads to an increase in the total production of, you
guessed it, one unit—and thus drives down the price at which
all units can be sold. The claim that the marginal demand curve is
"effectively horizontal" as many neoclassicists assert is
simply untrue; one more unit of production is one more unit of
production.
What Keen fails to realize is
that the assumption of "effectively horizontal" isn’t
necessary to show that competition drives prices to equilibrium. I
actually worked out the full algebra for the price charged and
quantity produced by an n-oligopoly,
and sure enough, as n goes
to infinity, the price and quantity converge to the equilibrium.
Thus, Keen tried to argue that the
basic mathematics of neoclassical economics are wrong, and failed
miserably; the math works just fine, thank you. Where Keen should
have focused is on the absurdity of perfect competition as a
real-world phenomenon: It requires not only an infinite number of
consumers and firms (where at least we can get pretty good
approximations with millions of customers and thousands of firms),
but also perfect information, zero externalities, and perfect
neoclassical rationality—which is to say, the behavior of an
omniscient psychopath. Neoclassical economics does not model the
behavior of human beings; it models the behavior of amoral gods.
Similarly, Keen tries to make
the bizarre argument that profit is not maximized by setting marginal
revenue equal to marginal cost. That’s not something neoclassicists
assumed; that’s a calculus theorem. P
= R – C. max P → dP/dq = 0 = dR/dq – dC/dq →
dR/dq = dC/dq.
He claims that this disappears
if you include time as a variable; no, it does not. The math gets
more complicated (I probably already glazed some eyes with the
above), but when you solve that Lagrangian you still get marginal
revenue equal to marginal cost. If you can produce something for less
than you can sell it for, you do so; and you sell it, and you make
money. If it costs more to make than you could sell it for, you don’t
make it. And the point where you change your mind (assuming things
are continuous, which is a pretty good approximation for large
corporations) is the point at which the price you sell it for is
exactly equal to what it cost you to make it! He babbles some
nonsense about how you want to maximize the rate of change
of profit, but that’s simply not
true; you want the rate of change to be zero, because
that means you’ve hit a local maximum of the function.
Once again, Keen could simply
have pointed out that firms are not optimal
profit-maximizers, and it’s not clear that it would be a good thing
if they were. Nor is it realistic to separate the market into
individual pieces and solve the "partial equilibrium", as
though raising wages for auto workers wouldn’t affect the price of
real estate in Detroit, or an increase in the price of oil wouldn’t
change the cost of shipping from China.
Chapter 5 is better. Keen rightly
points out that economies of scale are the norm, not the exception;
and the fact that neoclassical models can’t handle them is so much
the worse for neoclassical economics. He compares the strange "short
run" in which firms somehow can change their labor usage without
changing their capital consumption with the empirical reality of
business, that buying more capital is often easier (and certainly not
systematically more difficult) than hiring new talent. Which takes
more effort and time, do you suppose: Buying a new copy machine, or
hiring a new secretary? Installing a new computer, or finding a new
programmer? Neoclassicists treat hiring as if it simply involved
pulling someone off the street, and capital purchase as if it
involved building a new factory from raw materials.
In chapter 6, the topic is labor
and how wages need not equal marginal productivity; he points out
that this result is derived from many unrealistic assumptions—most
of all that capitalists have no more market power than laborers, and
that people decide how much to work as a trade-off between money and
leisure. In fact, even with that
assumption, it’s possible to derive the result that workers won’t
vary their hours based on pay rates (a "vertical labor supply
curve")—and here’s the best part: To get that result, all
you need to do is add the extra assumption of homothetic
preferences, which is
normally something neoclassicists assume all the time. It’s a totally
unrealistic assumption of course; but isn’t it interesting how they
suddenly abandon the assumption when it leads to a conclusion they
don’t like?
On the other hand, Keen spends a
lot of time in this chapter criticizing the idea of a "social
welfare function"
that takes only GDP as input and outputs a utility score for the
whole society, and that really doesn’t seem to belong in a chapter
about labor markets. He is of course absolutely right that this is
ludicrous; the distribution of that GDP is vital to deciding how just
and happy our society will be, and there are other factors as well.
But I don’t see how it’s necessary to assume a social welfare
function in order to argue that wages equal marginal productivity.
Chapter 7 is a mixed bag; some
of Keen’s arguments are cogent, while others seem weak. He spends a
lot of time on Sraffa’s esoteric argument that capital should be
priced as "dated labor" discounted at the average
risk-adjusted rate of profit. As far as I can tell no actual capital
is ever priced this way, certainly not consciously; nor is it clear
to me that this would be a normatively good way of pricing it. It
seems to me that capital should be priced at the point where marginal
value equals marginal cost, just like anything else; and no CEO is
concerned with the average risk-adjusted rate of profit for the
economy, but rather for the amount (not
rate, amount) of
profit he himself can make on any given activity. To see that it is a
question of amount and not rate, which would you rather do: A, buy
something for $0.01 you could sell for $1, or B, buy something for
$10,000 you could sell for $15,000? The former way you make a 9900%
profit… of $0.99. The latter way you make a 50% profit, but it’s
$5,000. If you could repeat them both infinitely, you’d choose the
former; but you can’t, so why does that matter? Given the choice
between A or B, done once, you choose B.
If you were trying to assess the
value of the capital
this way, that’s wrong as well; its value is the amount of utility it
can provide by what it does, including the labor it saves, the
products it makes, and any joy it produces directly.
Fortunately, Keen is right in
his basic message of this chapter, which is that aggregating all
"capital" based on their price is basically nonsensical.
Except in a very narrow sense indeed, 10 books for $50 each and 2
computers for $1000 each is not $2500 of "stuff", and you
cannot calculate a rate of profit based on the value of the "stuff".
Prices vary as markets vary; you can’t take the price as some kind of
absolute measure of the good’s worth. Different capital goods are
used for different things, in different amounts, and depreciate at
different rates. The Cambridge
Capital Controversy
established all this… at which point neoclassicists simply ignored
this fact and went on aggregating capital in the same way.
Now, this might be an acceptable
simplification for some purposes; after all, there’s no way we could
possibly include every single type of good in our model. But you
should keep that in mind when you speak of a nation’s
"capital-to-labor ratio"; what does that mean exactly? If
it means anything at all, it’s clearly not about
price. A post-scarcity society in which everyone has automated
systems that provide them with anything they need is one in which the
marginal cost—and hence price—of all capital goods is
effectively zero. The price of labor might also be zero, or rather
labor decisions are not made on a wage basis; so do we count the
number of people who work, or what they are paid (nothing)? By price,
the capital-to-labor ratio would be 0/0, undefined; yet in a real
sense, it’s obvious that their capital-to-labor ratio is enormously
higher than ours. Conversely, if you are stranded in a desert and
someone offers to sell you a bottle of water for $1000 or a car for
$200,000, you’re spending an awful lot on capital and nothing on
labor; but we would not say that your capital-to-labor ratio is high.
Chapter 8 is my favorite chapter of
the book. Indeed, it could have been the entire book; most of the
rest seems superfluous. Chapter 8 is about assumptions and their role
in scientific theory. Keen distinguishes between three different
types of assumptions: negligibility assumptions, which are
safe (e.g. ignoring friction in plotting the trajectory of
spacecraft); heuristic assumptions, which are undesirable but
sometimes necessary (e.g. special relativity ignoring acceleration);
and domain assumptions, which are dangerous (e.g. neoclassical
economics assuming that markets are efficient). He thoroughly
demolishes Milton Friedman’s pseudoscientific argument that
"assumptions don’t matter" and "the more significant
the theory, the more unrealistic the assumptions".
Yes, this is where neoclassical
economics fails: its ludicrous assumptions that humans are rational,
information is perfect, externalities don’t exist, debt doesn’t
matter, prices adjust instantaneously, preferences are homothetic,
production has constant returns to scale, expectations are
rational… none of these assumptions are even remotely true, they
fundamentally alter the character of economic models, and often they
don’t even make the mathematics simpler. They merely serve to feed
the ideology that the free market is magic and government
intervention is evil.
Chapter 9 is about the difference
between statics and dynamics; it has always bothered me how
economists use "comparative statics" instead of actually
trying to work out dynamics. This is how you get things like "the
classical model" in which wages adjust to money supply changes
instantly and "the Keynesian model" in which they don’t
adjust at all. Obviously, wages do adjust to money supply changes,
but equally obviously it doesn’t happen immediately or all at once. A
dynamic model could very easily take this into account—dW/dt—but
comparative static models are forced to take an all-or-nothing
approach that produces ludicrous results either way.
Keen also uses this opportunity to
discuss complex systems and chaotic dynamics, which is also something
that is clearly desperately needed in economics. I think this is an
area I myself would do well to study more: I know very little about
the methods for solving complex nonlinear systems, and my first
thought would be to run huge computer simulations, which would
probably work but seems rather inelegant.
Chapter 10 is also a good one; it’s
about the greatest singular failing in the history of economics,
which is the failure of neoclassical economics to predict, prevent,
or even significantly mitigate the Second Depression. True, it wasn’t
as bad as the Great Depression, but back then economic science was
still quite new and computer simulations didn’t even exist. To make
the same fundamental mistakes 80 years later is simply unforgivable.
But of course they didn’t; the only
ones who got even close were the ones who are more sympathetic to
non-neoclassical theories, like Stiglitz and Krugman. Neoclassical
models predict smooth, efficient equilibrium; they simply can’t cope
with the idea of a sudden crash. And so, blinded by their own
ideology, neoclassicists didn’t believe a crash was possible even as
it had already begun. Honestly, the highest single priority of
macroeconomic theory should be to predict and prevent depressions,
shouldn’t it? But that is one thing that neoclassical economics is
simply incapable of doing.
Chapter 11 isn’t bad either; it’s
about financial markets, and how a massive system of everyone
second-guessing everyone else can lead to chaos and absurdity with no
foundation in the real world. This honestly is obvious to anyone who
looks (and it was obvious to Keynes), but somehow there are still
economists—a lot of them—who believe in the "efficient
market hypothesis".
Personally, I’m not sure the financial
markets as we know them should exist. Yes, we need savings and
checking accounts. Yes, we need loans to finance investment (new
factories, student loans) and large consumer purchases (houses,
cars). And… that’s about it, frankly. All these other layers of
finance really aren’t necessary as far as I can tell, and right now
what they do is sap wealth from everything else and make the whole
system more unstable. I really don’t see why we couldn’t have a fully
nationalized banking system that would perform only these basic
functions and nothing else. Nevada effectively has a nationalized
banking system, and not only is Nevada doing rather well, it actually
fared better in the crisis than any other US state. The basic
arguments for why free markets work (which are already limited) break
down almost completely for financial markets; there’s no comparative
advantage, no gains from trade, no mutual benefit; the whole system
is zero-sum and overwhelmed by asymmetric information. There are some
pretty good arguments for why auto manufacturing and computer
hardware are better off being privatized; but the fundamental basis
for those arguments breaks down completely when you try to apply it
to financial markets.
Alternatively, we could just regulate
much better, like in the Canadian system, which is almost universally
agreed to be the most stable and efficient banking system in the
world.
Chapter 12 is on the Great Depression.
Keen is absolutely right about two things: We need an empirically
validated theory of what causes depressions, and neoclassical
economics is not equipped to provide such a theory. Where he loses me
is on his particular Minsky-based theory of "debt deflation";
deflation was not a serious issue in 2009, and it is quite easy to
correct—raise minimum wage and print more money. Yes, the Great
Depression involved a lot of deflation, but that doesn’t seem to be
what caused the worst problems.
Debt, on the other hand, obviously is
important—but exactly how it causes depressions remains
unclear. Keen’s model proposes that aggregate demand is equal to GDP
plus the rate of change in debt; this doesn’t make much sense, since
GDP includes all spending on final goods—if that debt is being
spent on goods, it will already be counted. (And if it isn’t, what’s
it for?)
Chapter 13 is on Keen’s own
predictions of the Second Depression. I will give him credit where it
is due: He was one the first to predict the financial crisis, and did
so with precision and mathematical rigor. He didn’t just say "A
depression is coming"; he pointed to specific problems in the
housing market and the financial system.
Chapter 14 is Keen’s monetary model of
capitalism. Once again, Keen is absolutely right that we need
something like this—we need a comprehensive model of the
macroeconomy that includes monetary and financial systems. Our
current models are woefully inadequate—they massively
oversimplify money, and they typically don’t include debt at all. But
once again, Keen’s specific choice is less compelling; he bases it
around double-entry accounting with three sectors: Firms, consumers,
and bankers. It doesn’t seem completely wrong, but it’s also far from
complete. Still, it’s better than nothing—which is basically
what we have right now.
Chapter 15 is on the instability and
inefficiency in the stock market. Here is where chaos theory really
shines; if we are ever going to understand the dynamics of such a
complex system of feedbacks, it’s going to be using the tools of
chaos theory.
Right now, it’s mostly a sketch; it’s
certainly no comprehensive theory. But Keen freely admits this; he
more uses the chapter to express hopes for the future of chaos-based
economics, and also to demolish any vestiges of belief the reader
might have in the "efficient market hypothesis".
Chapter 16 is about mathematics, and
how it’s still vitally important, but must be used correctly. I
couldn’t agree more; one thing that never ceases to aggravate me is
people who try to argue that the basic methods of science, or math,
or even logic are flawed simply because they haven’t solved
some particular problem. Even worse is when they simply won’t accept
the solution—"Science has nothing to say about God!"
Yes, actually it does: It says he doesn’t exist.
Chapter 17 is about Marxism and why it
fails, particularly how the labor theory of value collapses.
Obviously the labor theory of value is dumb; something does not
become valuable simply because people have worked on it, and
something doesn’t stop being useful just because it was easy to make.
Yet Keen feels such a need to reject
neoclassical economics that he can’t even accept the utility theory
of value, even though it is obviously correct. Instead he tries to
come up with some alternative theory based on the supposedly
fundamental difference between use-value and exchange-value—when
that distinction has already been captured in the standard
distinction between utility versus price.
He mocks the neoclassical
theory as the "subjective theory of value", which is at
best a misleading way of putting it; the subjectivity involved is the
kind of "subjectivity" involved in saying that pain hurts
and orgasms feel good. It is the "subjectivity" involved in
saying that chairs are for sitting on and apples are for eating. One
could perhaps imagine a lifeform that felt differently, but we are
not that lifeform.
This is also something that moral
scientists have trouble explaining to everyone else: Morality is only
"subjective" in the sense that it depends on the
experiences of sentient beings. That doesn’t mean anything goes. It’s
not "rape is wrong if you think it is"; it’s "rape is
wrong because people don’t like being raped". Likewise, it is
objectively the case that goods have subjective value to
people, and that’s how our economy works.
Chapter 18 briefly summarizes some of
the alternative approaches: Austrian economics, evolutionary
economics, Post-Keynesian economics, Sraffian economics, and
econophysics. He rightly concludes that Austrian economics is a dead
end for the same reasons as neoclassical, and that Sraffian economics
doesn’t go far enough; but he is oddly skeptical of evolutionary
economics and econophysics, and gives more credit to
Post-Keynesianism than I think it really deserves. (Especially since
"Post-Keynesian" is a lot like "post-modern"; it
doesn’t describe what you’re doing so much as what you’re not doing
anymore.)
Oddly he does not include cognitive or
behavioral economics as one of the alternative approaches, even
though it plainly is. Indeed, he takes an oddly dismissive view of
cognitive economics, seeing us as apparently just working out a
series of ad hoc examples of minor irrationality that
ultimately will have no meaning for real-world economics.
And I suppose this is true, at least
of cognitive economics as currently practiced; but it is a field in
its infancy—it basically didn’t exist until about 15 years ago.
Give us a few more years, and you will see that it is a radical
paradigm shift as significant as evolutionary economics or
econophysics, and clearly far more so than Post-Keynesian or
Sraffian.
All the way throughout the book, Keen
seems intent on showing that neoclassical economics is wrong, wrong,
wrong, about everything, in every possible way. But
this was not necessary; reversed
stupidity
is not intelligence. Neoclassical economics does not get
everything wrong; in fact, it probably gets more things right
than the general folk notions most people have about economics. I
doubt most people realize that sales taxes create deadweight loss,
for example, or that the money supply is largely created by bank
loans and isn’t backed by any commodity; yet these are things that
neoclassicists definitely do understand quite well. Neoclassical
economics is wrong not at the core, but at the margins; but isn’t it
neoclassicists most of all who taught us that margins are everything?
