From: Rakesh Bhandari (bhandari@BERKELEY.EDU)
Date: Wed Sep 15 2004 - 18:16:11 EDT
At 3:03 PM -0700 9/15/04, Ian Wright wrote: >Hi Jerry > >Regarding tendencies to narrow the wage dispersion vs. the emprical >fact of wage inequalities. > >This is one of those cases where the empirics are hugely >overdetermined (e.g., Howard's "avalanche" of causal process occurring >one on top of another). The factors you mention that generate wage >inequalities are efficacious. But it all depends on the level of >abstraction. > >Even if the factors you mention were not efficacious, there are good >reasons to still expect pronounced income inequality. That's because >the economy is like a huge cocktail shaker in constant motion. It >never rests, and does not have a deterministic equilibrium. The income >distribution of the lower 90% or so of people can be fitted to >an exponential (there are other fits, but not important here). This is >the maximum entropy distribution under the constraint of money >conservation. In other words, the income distribution can be >considered the most disordely, most mixed up, distribution of income, >assuming only that money is conserved. > >Only in simple mechanical systems with very few degrees of freedom >should we expect the empirics to be adequately characterised by >equilibrium values rather than equilibrium distributions. > >So given that we lower our expectations about the empirics, we can >then ask whether the wage distribution in capitalism displays a >tendency to narrow. My guess is that it will stretch and contract >depending on all kinds of local contingencies, but will do so within >certain limits. > >But at a higher level of abstraction I think the empirical data does >indicate a historical tendency for intra-class wage equalisation >amongst workers. The lower (predominately wage) end of the income >distribution is unimodal. It has a single peak. It is not multi-modal. >Why is it unimodal and not multi-modal? > >If we sampled the heights of people that visited a web-site during a >month, then the histogram of heights would in all likelihood be >bimodal. That's because the population is split fifity-fifty into men >and women, and men are generally taller than women. The existence of a >bimodal distribution indicates the population can be split into two >types, and that members of each type share common properties that >affect the measured variable. In contrast, the existence of a unimodal >distribution indicates that the population cannot be sorted into >different types of significant statistical weight at this level of >abstraction. In his study of the history of statistics, inspired by Pierre's Bourdieu's sociology of knowledge, Alain Desrosieres has uncovered the first use of statistical analysis for racially classificatory argument--Adolphe Bertillion's famous study of conscripts from the department of Doubs: Instead of being Gaussian and unimodal in appearance, as elsewhere, the distribution of the heights of soliders from this department offered two curious humps (bimodal distribution). From this Bertillon deduced that this strange form resulted from the superimposition of two normal distributions, and the population of the Doubs was a mixture of two distinct ethnic groups: Burgundians and Celts. Thus the form of the curve served as a classificatory argument. See discussion in Alain Desrosieres, The Politics of Large Numbers: A History of Statistical Reason (Harvard), p. 96 > >I submit that if people were not qualitatively equal in their >productive capabilities then a part of the working population within a >single labour market would historically have been left behind. That >would manifest as a bi- or multi-modal income distribution, >representing a process of separation and breakdown of wage >homogenisation. But that is not what we find. Instead we find a >unimodal wage distribution. > >The change in wages over time is like swarming behaviour: some get >ahead, some are at the back, but the swarm moves together and does not >split. The fact that it does not split requires an explanation. > >Overlaid on the unimodality are the kind of factors you mention, >including wage inequality due to discrimination; the formation of >coalitions, such as unions; or the effect of political movements, such >as those Anders mentioned, which may more or less consciously reflect >the objective equality. > >-Ian.
This archive was generated by hypermail 2.1.5 : Thu Sep 16 2004 - 00:00:03 EDT