Three-Toed Sloth

Books to Read While the Algae Grow in Your Fur, April 2010

Jean-Guy Prévost, A Total Science: Statistics in Liberal and Fascist Italy

A history of the Italian statistical community (or, as he prefers, "field") from around 1900 through the fall of Fascism, with a brief glance at the immediate post-war era. This is not about the history of statistical technique, but about the development of statistics as an autonomous academic discipline, with pretensions of in fact being the key discipline for all empirical investigation, especially into social and biological matters. So we get a lot about university positions, internal disputes (as Prévost says, one mark of a field is precisely that there are recurring internal arguments, with well-worn positions), how methodology came to be seen as more important than either mathematical theory or applications, the conflict with political economy, etc. Naturally, this extends to looking at how the statistical establishment eagerly sought to serve the Fascist state, proposals for "corporatist" and "totalitarian" statistics, and the elaboration of Fascist ideology by leading statisticians, relying on their self-presentation as polymaths. (Tukey's line about how "The best thing about being a statistician is that you get to play in everyone else's backyard" assumes a new significance when you imagine it being uttered by a blackshirt.) In all this, including the last, Gini is the central figure; quite honestly he should have been purged after the war, but somehow escaped justice.

Some familiarity with the history of both Fascism and of the intellectual content is presupposed. If you have that, and are willing to tolerate a minimal (almost homeopathic) dose of Bourdieu, this provides a lot of interesting, if unhappy, food for thought.

(Prévost makes it clear how Gini's work on measuring inequality was in the tradition launched by Pareto's laws of income and wealth distribution. Cantelli was a friend and collaborator of Gini's, and the Glivenko-Cantelli theorem is the kind of result which would guarantee non-parametric consistency of estimated Gini coefficients from sample data. Was this what motivated Cantelli?)

Robert B. Reich, Supercapitalism: The Transformation of Business, Democracy, and Everyday Life

This is aiming to be something like The Affluent Society or The New Industrial State for modern times; it does a pretty good job. Basically, his argument is that Galbraith was more or less right about how the economy worked during the post-WWII golden age of capitalism: large, autonomous, oligopolistic firms more interested in continued steady growth, exploiting economies of scale, than anything else. JKG's mistake was in thinking this regime would continue. Reich sees Galbraithian capitalism as being upset not so much through deliberate political action in the 1980s as through new technologies in the 1970s, especially improvements in logistics, communications and information technology, which made it possible and efficient to replace the vertically-integrated firm with global supply networks, and to replace investment financed out of retained earnings with global financial markets. (As Reich points out in some detail, all of the key technologies, from container shipping through microelectronics and the Internet, were devised by the military-industrial-university complex to fight the Cold War; sowing the dragon's teeth, as it were.) Deregulation, to Reich's way of thinking, was more a consequence than a cause — the legal superstructure accommodating changes in the forces of production, though he doesn't use such language. The result, he says, is a system more responsive to consumer demand and to investors, but where most of the population sees no gains from economic growth, inequality soars, countervailing power evaporates, security is steadily eroded, and the primary check on the political influence of corporations is the opposing commercial interests of other corporations. (He also has an ingenious argument as to why decreasing regulation led to increasing lobbying.) This he calls "supercapitalism"; I dislike the term and will avoid it.

The way the system is set up, he says, the people running corporations simply have no choice but to do whatever they can to maximize profit in the short term; if they won't, they will shortly be replace by those who will. Calls for corporate social responsibility, still less trying to shame or pressure individual corporations, therefore misses the point. The goal, rather, has to be to change the laws under which all corporations must act, ultimately, to neuter corporations politically, and creating a non-corporate social safety net. (The idea that health insurance, for instance, should be provided by one's employer is just nuts.) Something he does not adequately address, though, is that laws and regulations must be enforced, which is hard to do when one of the two parties regards them as necessarily illegitimate...

So, criticisms: (1) As I hinted, I think Reich underplays the role of ideology and political action, in favor of technological developments and market forces. It would be interesting to try to synthesize this with Krugman's take in Conscience of a Liberal. (2) There are some bits where the economics is a bit odd. For instance, economies of scale are certainly important in information production, just as in making steel. (Cf.) Arguably though the sheer magnitude of the fixed costs, and the time-scale, has shrunk, and that would be enough for Reich's argument. Also, profits decline as industries become more competitive, falling to the cost of capital plus the cost of the entrepreneur's time.

(Picked up after someone, I forget who, pointed me at Lessig's review.)

Nunzio DeFilippis, Christina Weir and Christopher J. Mitten, Past Lies

Brandon Graham, King City

Michael Alan Nelson, Emma Rios and Cris Peter, Hexed

Mark Waid and Minck Oosterveer, The Unknown

Brian Michael Bendis and Michael Avon Oeming, Powers: 1 (Who Killed Retro Girl?), 2 (Roleplay)

Robert M. Solow, Monopolistic Competition and Macroeconomic Theory

"Monopolistic competition" is the slightly oxymoronic name for the situation where there are a number of goods which are all more or less close substitutes for each other, but each good has a monopoly producer. It can arise in a number of ways, from legal restrictions (e.g., copyright on particular pieces of software) or from increasing returns to scale. (Successful branding convinces consumers that basically identical commodities are really different, and so creates monopolistic competition.) In monopolistic competition, firms have some control over their prices, but to maximize profits they need to forecast quantitative demand. The theory is quite well-established microeconomics, having begun its real development in the 1930s with Chamberlin and Robinson, and is a standard part of industrial organization (Cabral's textbook has an especially nice treatment).

This extremely short (88 pages including the index) book consists of Solow pointing out that once you admit monopolistic competition is not just possible but actually common, a lot of the conclusions of macroeconomic models which rest on the idea of perfect competition in all markets evaporate, and one is led to Keynesian conclusions, even if one assumes that everyone in the economy is a perfectly foresightful utility maximizer. In particular, the way is opened for the existence of multiple equilibria: low-output equilibria in which everyone correctly forecasts that there will not be a lot of demand, so they produce little, pay little, and buy little, and high-output equilibria in which everyone correctly forecasts high demand, produces a lot, pays a lot and buys a lot. Everyone prefers the high-level equilibrium to the low, but that doesn't mean they'll manage to coordinate on it. Solow takes this insight, and related ones, and does what he does best, namely build and solve elegant little models of the resulting macroeconomy. He is quite open about these being toy models, and that in some places he has to stipulate some macro-level relations which he doesn't directly derive from the micro assumptions. (But, though he doesn't mention this, the same is true of the usual representative-agent macro models which purport to be aggregations of perfect competition.) The results are not strictly in line with every detail of the General Theory, but are clearly closely related, and make a lot of sense.

This book is very enjoyable, if you have any taste for elegant economic modeling, though alas the price of the actual physical artifact (twenty six cents per page in paperback) is insane. But I've ranted about this before.

Richard A. Berk, Statistical Learning from a Regression Perspective

A gentle introduction to modern nonparametric regression and classification, for people who are comfortable with running linear and logistic regressions, and curious about data mining and/or machine learning. After a brief review of regression (following the lines laid down in his earlier book), Berk covers smoothing (especially with splines), additive models, classification and regression trees, bagging, random forests, boosting, and support vector machines. There are many real-data examples and exercises, all done in R, and all of them I think from the social sciences, with a certain emphasis on his own field of criminology.

Berk relies very heavily on The Elements of Statistical Learning as an authority, and one might think of this as a simplified presentation of the key parts of that book, for social scientists, or advanced undergraduates in statistics — I used it as a supplementary text in my data mining class last fall, and would happily do so again.

Books to Read While the Algae Grow in Your Fur; Scientifiction and Fantastica; Pleasures of Detection, Portraits of Crime; The Dismal Science; Enigmas of Chance; Writing for Antiquity; The Continuing Crises; The Running-Dogs of Reaction