Thursday, March 19, 2009

Leftist Econophysics

There is a branch of econophysics that has attracted serious attention from some Marxist and other leftist economists. This is the series of studies of income and wealth distributions. Traditionally income has been thought to reflect lognormal distributions, which can arise endogenously from random processes from an initially equal distribution. In fact, labor income appears to follow such a pattern. However, wealth appears to follow the power law distributions of financial asset returns, unsurprisingly, that have fatter tails than lognormal distributions, that is, greater inquality with more people far off into the upper end of the distribution. Econophysics studies of this began around a decade ago, appearing in places like Physica A and European Physical Journal B by people like Levy and Solomon and also Dragulescu and Yakovenko, although John Angle showed some of the things in the early 1990s in the Journal of Mathematical Sociology.

Then it was figured out by people like Yakovenko and some others that income distribution looks lognormal for most of its lower portion, where it is determined by labor outcomes, but that its upper portion looks more like a more unequal power law, a la wealth and financial market returns. The class nature of this then began attracting the attention of some leftist economists such as Allin Cottrill and Paul Cockshott, who have now coauthored a book, Classical Econophysics, with Gregory John Michaelson, Ian P. Wright, and Victor Yakovenko. More of these studies, some done by economists such as Mauro Gallegati and some of his coauthors, including some physicists, have appeared in the 2005 book, Econophysics of Wealth Distributions, ed. by Chatterjee, Yagarladda, and Chakrabarti, Milan: Springer.

9 comments:

Jack said...

Sorry, but this comment, a question really, has nothing to do with the content of the post. Why
when I click the Read more notation
does nothing but "hidden chonclusion here" show on the screen? This has happened on othe threads here. If there is no additional text why have the Read more notation?

rosserjb@jmu.edu said...

I do not know the answer. There is a "further" something when one posts, and maybe one has to delete that.

YouNotSneaky! said...

Hmm, I didn't know that this fact - income distribution is log normal with Pareto at the top - was discovered by econophysicists. I thought it was a widely known fact established pretty well shortly after Pareto.

I've also wondered how all those kernel estimates of income distributions which suggest multi peaked distributions (particularly for Latin America) where you can actually see distinct "classes", fit into this.

rosserjb@jmu.edu said...

YNS,

Pareto claimed that the entire income distribution followed a power law, indeed that is what he invented the distribution to explain. The coefficient he found he claimed (wrongly) to be a universal law, and of course he fit this in with this theory of the "circulation of the elites."

Later the lognormal became the standard view in much of the literature, although there were always some doubters, such as Champernowne. However, I am reasonably certain that this idea of the lower part being lognormal and the very top part being power law is a recent idea due to these econophysicists.

Anonymous said...

my own view is that actually what yakovenoko showed something quite diffrent. lognormal is old. hOW anyone could be an economist (cept for (aprox) 'you not snitching'} without noticing this is beyond me.

rosserjb@jmu.edu said...

media,

Lower part is Boltzmann-Gibbs, which is Gaussian. Upper part is power law. One major issue is where is connecting point from one to the other, looks to be maybe in the mid-90 percentile or thereabouts.

Much of the older literatured did focus specifically on the lognormal, and there is also this theoretical lit, around for a long time, that shows how the lognormal can arise ramdomly from an initially equal distribution, out of, well, call it "luck."

Anonymous said...

yeah, boltzmann. that suggests the lower classes are 'zero intelligence' traders, which explains their income and is explained by the social darwinists. (alan wilson's gravity model more or less said the same thing, in 1969, though it didnt specify money.)

in a sense the MB distn is a gaussian, but with a constraint, so it is and it isn't. also, note its 'one sided' without a central mode. and in a sense the lognormal is a (multiplicative) gaussian too, so in a different set of coordinates ('preferential attachement' or positive feedback the lognormal is just another maximum entropy solution) . (and often power laws are not that distinguishable (especially by just looking at the curves) from lognormals, or even the messier versions like fitted beta distributions.)

there is actually an application of the MB distn to income from the 40's or 50's in the econ lit but i forget where---maybe it was champerowne. i actually looked at some old tables from the 20's on income distribution among working people, and yakovenko was right---it was negative exponential (MB), not Pareto. i was amazed nobody had noticed this---you have to seperate wage from capital income.

Anonymous said...

p.s. the 'mb' distn actually should be called boltzmann (exponential) to avoid confusion. 'money is energy' (not just kinetic energy, as in the mb case---i guess you can add a potential there too, eg p=exp-mg/kt, for dust in a gravitational field (see feynman's lectures)).

rosserjb@jmu.edu said...

Champernowne.