In his New York Sunday column Paul Krugman launches a blistering attack on mainframe macro (mostly based on representative agents with rational expectations who maximize over an infinite future). John Cochrane launches an equally blistering response rhetorically beginning with:
“Krugman, says, most basically, that everything everyone has done in his field since the mid 1960s is a complete waste of time. Everything that fills its academic journals, is taught in its PhD programs, presented at its conferences, summarized in its graduate textbooks, and rewarded with the accolades a profession can bestow, including multiple Nobel prizes, is totally wrong.”
No doubt, there are quite a number of economists, not to mention non-economists, that would nod at least in partial agreement with the above. As a result of the crisis, such views have manifested themselves in numerous critical articles in the press and in a chorus of critical voices (including the English queen!) demanding that standard textbooks in economics should be re-written. Also, a number of economists have triumphantly declared ‘I told you so, Keynes is right after all!’ But isn’t the latter just a foreseeable reaction by those whose papers were previously rejected by editors of top economic journals and whose proposals were largely ignored by policy makers? Loosely interpreted this seems more or less what Cochrane says in his response to Krugman’s critique. So, has Cochrane a valid point in saying that Krugman has misled New York Times readers with his outdated Keynesian insights?
I believe this is a question of considerable interest both among economists and the general public and have tried to address it based on my experience as an empirical econometrician over a period over roughly 25 years. Let me say it right from the beginning: when we let the data speak freely (in the sense of not constraining the data according to one’s economic prior without first testing such restrictions) they mostly tell a very different story than the one found in standard textbooks. No doubt, our fast changing reality is infinitely more rich and complicated than the models which are discussed there. This does not necessarily mean that these models are irrelevant; it only means that abstract theory models and the empirical reality are two very different entities. Therefore, it did not come as a big surprise that very few observers came close to predicting the scale of the global financial and economics crisis that erupted in August 2007, and that those who did predict a massive correction in financial markets, with the associated impact on the real economy, typically relied on anecdotal evidence and rule-of-thumb indicators that are far less sophisticated than the models employed by central banks and leading academics (Colander et al. 2009).
What lessons can we draw from the widespread failure of central bank modelling and forecasting prior to the crisis? Which considerations need to be built into future models to improve their predictive power? Cochrane essentially argues that the efficient market hypothesis tells us that there is no need to do anything. This is just what one should expect to happen in a market economy every now and then: “it is fun to say we didn’t see the crisis coming, but the central empirical prediction of the efficient markets hypothesis is precisely that nobody can tell where markets are going – neither benevolent government bureaucrats, nor crafty hedge-fund managers, nor ivory-tower academics. This is probably the best-tested proposition in all the social sciences.”
But is this really so? The efficient market hypothesis says that nobody should be able to systematically make money by predicting future outcomes. This has generally been interpreted to mean that financial prices should behave like a random walk and this hypothesis has certainly been tested many times. The problem is that the univariate random walk hypothesis is all too simple as a test of the EMH. For example, if we assume for a moment that stock prices in general should be affected by the level of savings and investments in the economy, it would be reasonable to assume that stock prices would have a common component. In that case some linear combination would very likely be cointegrated and, hence, testing the random walk hypotheses in a multivariate setting would reject the hypothesis (which is what we find). This, however, does not as such imply a rejection of the EMH. If everybody reacts on the common information (though not according to the Rational Expectations Hypothesis) the market may not be able to systematically make money. Over the long run prices would converge to a level that could, for example, be consistent with a Tobin’s q level or a constant price/earnings ratio, and would therefore in a broad sense be predictable. But, even though the EMH might be right over the very long run, over the medium run of say 5-10 years, unregulated financial markets are likely to produce persistent swings in asset prices (Frydman and Goldberg, 2009) which are basically inconsistent with the following two interrelated assumptions:
1. Financial markets tend to drive prices towards their fundamental equilibrium values
2. Financial markets pricing behaviour is influenced by fundamentals in the real economy, but it does not change the fundamentals.
While most rational expectations models are implicitly or explicitly based on the above assumptions, empirical evidence (when letting data speak freely) suggest that financial markets drive prices away from fundamentals for extended periods of time with strong effect on the real economy. This is what George Soros has named reflexivity between the financial and real sector of the economy. Since it undermines the idea of unregulated financial markets as a guarantee for efficient capital allocation, reflexivity is an important feature of free financial markets that need to be included in our economic models.
Could empirical analysis have signalled the looming crisis? Already many years before the bubble burst, the relative house-consumption price index exhibited very pronounced nonstationary behaviour. Such movements of house prices away from ordinary consumer prices was (primarily) facilitated by low price inflation and interest rates. Prior to the bursting of the bubble, the house – consumption price ratio increased almost exponentially signalling that house prices were moving far away from their sustainable level, given by the level of inflation and interest rates. This, in my view, would have been the absolutely last moment for the authorities to react to prevent the subsequent disaster. But, the empirical result that the extraordinary high level of house prices were sustainable only by the extraordinary low levels of nominal interest rates and inflation rates should have been a reason for concern much earlier.
Has beauty been mistaken for truth? Assume for a moment that economists would agree that a minimum requirement for a theory model to be called empirically relevant is that it would be able to explain basic empirical facts as formulated in terms of the pulling and pushing forces estimated from a correctly specified and adequately structured VAR model. In such a case I would not hesitate to answer the above question in the affirmative. Economic data, when allowed to speak freely, often tell a very different story than being told by most Walrasian type of models. If anything, the stories data tell have much more a flavour of Keynesianism (though not New Keynesianism!) than Neoclassicism (or other recent isms). But when this is said, I doubt whether it is a good idea to ask which school is right and which is wrong. In some periods, one school seems more relevant, in other periods, it is another one. Quite often data suggest mechanisms which do not fit into any school. But, I am convinced that relying on beautiful but grossly oversimplified models as a description of our economic reality may easily prevent us from seeing what we really need to see. For example, I believe information about looming crisis was there in sufficiently good time to have helped us see the coming disaster, had we chosen to look carefully.
To conclude: I believe many economists were indeed blinded by the beauty of their theory models and in my view Krugman was right to point this out to the readers of NY Times.