Thursday, July 13, 2006

Keen And Ormerod, Econophysics, And The Distribution Of Income

This humble blog is all about documenting fallacies in mainstream economics. Very little of what I have to say is new, and some have tried to popularize some of these criticisms. I think, in particular of Paul Ormerod (1994) and Steve Keen (2001). Ormerod and Keen have some similarities. Both:
  • Have written popular books (Ormerod 1994 and Keen 2001) explaining that neoclassical economics is incorrect
  • Mention Lipsey and Lancaster's theory of the second best
  • Emphasize that mainstream economists studying the theory of General Equilibrium have demonstrated the theoretical untenability of neoclassical economics
  • Recommend an approach to economics based on the mathematics of complexity theory
  • Publish original criticisms of mainstream economics first in their popular books (Ormerod 1999 and Keen 2001) and then later in a physics journal (e.g., Keen and Standish (2006) and Ormerod and Mounfield (2000))

They differ in that they express different opinions on Piero Sraffa in their popular books. Keen (2001) praises Sraffa, while Ormerod (1999) snarks.

I find that Ormerod and Keen have teamed up with two other authors for a recent paper (Gallegati et al. 2006). This paper contains an approving reference to Sraffa. A major theme of this paper is to question whether the evidence for power laws, in particular in the tails of certain distributions of certain economic variables, is as definite as seems to be claimed sometimes in the econophysics literature.

  • Gallegati, Mauro, Steve Keen, Thomas Lux, and Paul Ormerod (2006). "Worrying Trends in Econophysics", Physica A
  • Keen, Steve (2001). Debunking Economics: The Naked Emperor of the Social Sciences, Zed Books
  • Keen, Steve and Russell Standish (2006). "Profit Maximization, Industry Structure, and Competition: A Critique of Neoclassical Theory", Physica A
  • Ormerod, Paul (1994). The Death of Economics, London: Faber & Faber
  • Ormerod, Paul (1999). Butterfly Economics: A New General Theory of Social and Economic Behavior, Pantheon
  • Ormerod, Paul and Craig Mounfield (2000). "Random Matrix Theory and the Failure of Macro-Economic Forecasts", Physica A


Anonymous said...

Hi Robert.

I'd like you to know that I found your blog really interesting, even if I have never commented till now.

Since my ignorance (in economics and in english), I don't understand you at all, but I'm learning every day, reading the online docs you're indicating.

BTW, I'd like to ask you what's your opinion about a guaranteed income (alias basic income) for all, and/or a guaranteed wage for unemployed people only.

What if the State will give, let's say, 500$/month to *everyone* out there? And what if the State gives, let's say, the 70% of their last wage to unemployed people?

Some economists says that wages will fall down, unemployment and prices will grow up, and so on.

I think that even according to the LTV, some of these claim could be true. I've seen you made some thoughts about the minimum wage in the past, but I think that the basic income is something different.

What can you say about this?

Thank you very much!

ps. please take all the time you need to answer (well, if you *want* to answer), there's no hurry. :)

radek said...

Seriously, I don't know about Ormerod, but Keen is an all out hack. He's a pretty skillful hack in that he mixes some legtimate criticisms with outright bullshit. As far as I can tell he's just interested in tooting his own horn and mostly appeals to folks who don't bother reading him closely enough or are ignorant of the material.

The stuff's he's written on the SDM theorem or Cournot, or the whole P /= MR is just plain emberassing. His webpage evokes a mixture of pity and disgust in me.

You loose some serious credibility points by pushing him.

Robert Vienneau said...

Giovanni, thanks for the comments. I like the idea of a guaranteed income, at least in OECD countries. But this blog is not much about policy analysis. At the level of abstraction of presentations of reswitching or capital-reversing, one cannot legitimately deduce specific policy proposals.

Radek, but enough about your feelings. Can you formulate an argument? I think my post covers a lot of material in a non-judgemental way. The Sonnenschein-Mantel-Debreu results are huge. Popularizers have a challenge. I am in the acknowledgements for Keen's book because I had some comments on aspects of his presentation. As for Keen's novel critique of the Marshallian theory of the firm, you want to read Keen and Standish (2006). I think Keen's uncovery of Stigler (1957) is impressive, and I find Stigler's results disquieting. I worry that some of Keen and Standish's simulation results may be an artifact of implementation details. But to substantiate these worries, I would need to analyze the results of a formal experimental design on this simulation.

dsquared said...

on the general subject of econophysicists and their power laws.

radek said...

Re: Simulation

This is the "Cournot was wrong!" thing, right? (The link to the paper doesn't work). If it's the same thing that Keen had on his website (which isn't there anymore) then it's easy - it's just the particular behavioral strategy, which looks like tit-for-tat, sneaked in in the computer program. Which is why Keen is a charlatan as I expect he knows better.

In fact, when this thing was on his website he explicitly stated something like "I don't rely here on standard arguments from theory of repeated games, like tit-for-tat strategies, or evolutionary game theory". Which is a total lie. He does exactly that.

So the point of the simulation is to show that the market price doesn't converge to the competitive price as number of firms, N, goes to infinity. Standard models say this non-convergence would happen if firms manage to collude. Evolutionary theory would say that since in a Cournot game there are no strictly dominated strategies (internally) then there's some set of behavioral rules for the firms which will get you pretty much any outcome you want. Keen's result is a combination of these.

In fact it's not necassary to run the simulation or spend too much time on the code - except to unearth the assumption that Keen doesn't tell you about. The assumption is this: if your profit increased last period and you took action A, then take action A this period (regardless of what other firm is doing), if it decreased then take action not-A (he's got it set up so these are the only choices + a stopping rule)(note also that there is NO strategic interaction here whatsoever). So you can actually solve this in the quantities and set up a system of difference equations and then solve the sucker. But this isn't necassary either. Knowing how the Cournot game (particularly the fact that it's a game in strategic substitutes - i.e. the fact that reaction functions are downward sloping, i.e. profits decrease in other agents' choices) and a bit of intuition lets you figure out that this has to converge to the monopoly solution regardless of N - because the behavioral rules for firms are chosen so that they MUST end up colluding.

So let's just think of 2 firms - any other N works the same - and it helps to sort of picture in your mind the standard Cournot diagram, with the two downward sloping reaction functions. The Cournot quantity is where they cross, the monopoly quantity is to the SW of it.

Suppose both firms are NE of the monopoly solution, Qm, and they both decrease (increase) their production. Then, since they're NE of Qm, revenue increases more (less) than costs and both make higher (lower) profits. So next period both will continue to (will choose to) decrease their production until they reach Qm.

Alternatively suppose that, still to NE of Qm, one firm increases its q and the other decreases it. In this case it's obvious that the change in profits for the two firms will be in opposite directions (doesn't matter which firm wins and which looses) - this means that in the next period they'll start moving in the SAME direction since one which gained will continue it's behavior, and the one which lost will change to it. Once you have both doing the same thing they're either moving SW towards Qm, or moving NE in which case see previous paragraph. So basically this looks like tit-for-tat except it's even stronger thant tit-for-tat, since deviations from it get reversed quickly.

The case when both firms start to the SW of Qm is just the reverse of above. It's a bit less intuitive if you're to the NW or SE of Qm but if you just think about it a little you can see that Qm is basically a "sink" of the system of the difference equations - you're always gonna converge to it. Thinking about it a bit more you'll also realize that this is completely independent of the number of firms***.

So. Keen sneaks in an assumption and then claims he's proven that the Cournot theorem is wrong. Bullshit. As any fool knows the Cournot theorem is a result for a one shot game. In repeated games, Cournot or otherwise, you run into the Folk Theorem, a crazy multiplicity of equilibria and anything can happen. Parallel arguments apply in evolutionary game theory. This is all 'standard theory'. Let me repeat, my undergrad students understand this! Keen's done nothing new except making a self-congratulory, false claim. Which is pretty much the substance of all his "work". The fact that he tries to peddle this off as original research is both dishonest and embarrassing.

*** There is an interesting question here actually, from point of view of evolutionary game theory which Keen completly ignores, and that's how the speed of convergence varies with N. I played around a bit with it when I wanted to figure out the charlatany of this particular example and though I didn't prove it I'm pretty sure you get a dynamic-evolutionary-sort-of version of the Cournot Theorem. Essentially, if you start of at the Cournot solution (for the given N)(or alternatively start everbody at some common point independent of N) then let the sucker go in time, the higher the N the longer it will take to converge to Qm. In other words, markets with lots of firms will start closer to the competitive solution and it will take them longer to diverge from it, with all the usual welfare implications that follow.

If I have time I'll write up why Keen's full of shit on the SMD theorem (actually what he's talking about is the Gorman Form Representation theorem), which, while huge, don't say what a lot of people think it says.

BTW, I have no problem with valid criticisms of 'standard theory' (I try to avoid the term 'neoclassical' as these days it essentially just means 'the kind of economics I don't like' - Cournot is definetly not neoclassical by the usual definitions) and I appreciated and learned quite a bit with your Saffrian/intermediate production examples on international trade. But Keen's different. He's just plain ol' bunk.

Anonymous said...


I'll just refer to this, at the bottom in the Maths section:

First, before making any specific criticism, it's good to note the dishonest rhetorical trick Keen engages in. He lays out (a total misrepresentation) the 'standard argument', claims he's gonna disprove it, then goes on to prove that the square root of 2 is an irrational number. WTF? Ok, ok, so he claims that he's just illustrating what a proof-by-contradiction is and that one can prove by contradiction that it's impossible to aggregate preferences, errrr, utility. Of course the 'sqr2 is irrational' is a real, rigorous proof. The 'can't agregate utility' argument is not rigorous and doesn't prove crap. But by including the 'sqr2 is irrational' proof Keen makes himself look like a mathematically minded, technically competent, knows-what-he's-talking-about, type of guy. Like those faith healers who drop in some medical terms to boost their credibility with the ignorant.

I'll put the specific criticism in a seperate comment.

radek said...

Re: SMD 2
(Previous comment by me, BTW)

First off, SMD is not about aggregating individual utilities in the sense that Keen's talking about. What SMD says is that the excess demand function inherits only certain properties of individual demand functions, and that in turn these properties are not enough to guarantee that the excess demand function is monotonic. Hence it may have more than one fixed point - i.e. there may be multiple price vectors which clear the markets.

So SMD just says that even if individual's demands are 'well-behaved' you may still get multiple equilibria. The conditions which Keen mentions are not the 'SMD conditions' (whatever that means) they are the sufficient (not nec.) conditions for a Gorman Form Representative consumer to exist. Now it's true that if (but not only if, not by far) these conditions are satisfied then uniqueness is guaranteed. But Gorman Form is a different theorem and here Keen's just confused or making shit up.

So how 'huge' are SMD results - the fact that one may have multiple equilibria. Well, here I think there's often a good bit of intellectual dishonesty among critics of GE. On one hand they criticise GE for being 'unrealistic' on the other they wave SMD theorem as proof that GE is 'untenable' (again, whatever that means). But seriously, why should you expect the Real World to have a unique equilibrium? Particularly in the presence of wealth effect, which are the driving force behind SMD theorem, and the whole point of doing GE in the first place? Personally I think the possibility of multiple equilibria is a REALISTIC feature of general equilibrium theory and a point in favor of it rather than against it.

One would be in trouble if the theorem predicted a continuum of equilibria, or that the First Theorem no longer applied but that's not the case. One can still do local comperative statics and all them multiple equilibria are still Pareto efficient.

So that's about SMD. Next comment about Keen's double misrepresentation - claim that the Gorman Form theorem implies that you can only aggregate utility when there's only one consumer and one good

Russell Standish said...

I just want to comment on some of Radek's statements. It is not a question of the Cournot analysis being a "one-shot" analysis, versus repeated game play. Everything is conducted in a continuous market setting.

The Cournot result assumes that an individual firm's output decisions have no effect on market price in the limit as the number of firms goes to infinity. This is a sort of independence assumption.

However, unfortunately this assumption is wrong for perfectly rational agents, as perfectly rational agents will react instantly to market change, and will always act the same way in the samer circumstances. The problem is this leads to a syncronised behaviour that drives the system away from the Cournot equilbrium. Such syncronous effects have been seen in the real world, for instance when computerised share traders first came on the market, the software trading agents locked into a syncronised pattern that caused a market collapse.

The Cournot equilibrium can be obtained, but only by breaking up this syncronous behaviour. Two methods we looked at for doing this was to introduce irrationality into the agents' behaviour, and also to introduce laziness (not all agents will react instantly to changes in the market). More details is available in arXiv:nlin.AO/0411006

Of course we know that real markets do not have perfectly rational agents. It then becomes an empirical question as to whether the market settles on the Cournot, the Keen or some other equilibrium entirely (the lazy agent simulations tended to settle on a value in between Cournot and Keen, depending on how many agents were updating their output decisions).