This note in TPM contains links to two fascinating articles about economics.
The first is an examination of the ideological basis of modern economics and its absurd claims to mathematical basis
http://www.prospect-magazine.co.uk/article_details.php?id=10683
The second is a defense of conventional modern economics by Paul Krugman
http://krugman.blogs.nytimes.com/2009/03/26/ive-got-keynes-right-here/
Kaletsky notes that Keynes did not do mathematical modeling and Krugman responds by quoting some mathematicish passage from Keynes. But mathematical framing is not the same as mathematical modeling and the difference can be illustrated from Krugman’s own misleading critique of Geithner’s toxic assets auction – where Krugman uses an expected value argument (vey is mir).
Expected value is just a fancy way of saying “average value over the very long haul” or more precisely “the value that one gets closer and closer to by taking more and more samples”. What Krugman does to explain the horror of Geithner’s plan is to make an example of an auction of “assets” which will either be worth $150 or $50 and have an expected value of $100. But it’s ridiculous to talk about expected value of an auction of perhaps 100 pools of shares – and in fact, using the term, smuggles in the very view of finance that sent AIG and Moody’s down the hole.
What Keynes did was use mathematical terms to be precise about what he was studying. What Krugman is doing is asserting that real-world properties can be derived from a statistical model. These are utterly different things.
The standard caution of using mathematical models is the story of the drunk who looks for his lost bottle under the lamppost because that’s where the light is. Economists, dependent on their toolbag of probability measures and models are not immune from the lure of looking where their model shows them, no matter where the object of the search may really be. Krugman is assuming that a potential purchaser of the assets (1) knows what the average value is over the set of assets and (2) is confident that he/she can purchase enough assets at the same price to assure that average value of the whole set is the average value of the purchased assets (if the expected value is 100 and you buy 10 assets that are all 50, you are in a sad place), and (3) have ZERO transaction costs and (4) the set of assets is opaque – there is no way to learn more about them than the average value.
Imagine what happens if TWO investors try the same model and make identical bids over the who set of assets. If they randomly win, then neither of them has purchased enough assets for the model to work – remember the expected value is at the limit. If you are told that a bag contains 50 gold bars and 50 copies of atlas shrugged, then when you only get to buy 50 of them, you have a chance of getting no gold bars a all! Assigning a value of 1000 to gold bars and $0 to the Ayn Rand books, the expected value over the entire set is $500. But only if you and, say, 3 other competitors each use this clever trick and bid $500 100 times, and each of you ends up with 25 items, the odds are strong that at least one of you will be weeping.
As a final note: the trick that Krugman and others assume that people will use to game the system depends both on the existence of a reliable valuation model better than anyone can get by inspecting the assets and that each of the famous non-recourse loans is tied to a single asset. The second assumption would indicate that the system is designed to be gamed – but if you look at Geithner’s proposal you can see that the loans use POOLS as collateral. That is, you cannot go to a pool of Citibank loans and buy 1000, on the theory that 501 will be winners, 499 will be losers and the losses on loans will be limited to the 499. The FDIC will take its loan out of your 501 winners too.