True, as Ph.D. students soon discover, narrow specialization is the only way to complete a dissertation, to get a job teaching in a university economics department, and to get tenure. But I think of those stages as akin to basic training in the Army or to 30-hour shifts in medical residencies. After one has achieved the prize (tenure), one can work on whatever one wants to work on.
Il a donc écrit sur une variété importante de thèmes, dont les taux de change, la croissance économique, le commerce, l'environnement... Il blogue aussi régulièrement sur différents sujets d'actu ici. Il m'est parfois arrivé d'être critique sur certains de ses articles ici ou là, mais j'aime bien son style et d'une façon générale sa vision historique des événements, qui, condensée dans cette longue (mais savoureuse) citation clôture ce bref portrait:
"In the 1980s, it became fashionable to claim that the real exchange rate followed a random walk, because statistical tests were unable to reject that null hypothesis at conventional significance levels. (Analogous claims were made about all sorts of variables in macroeconomics and finance.) But these tests were typically run on a few decades of data. I argued that one would not expect such limited data sets to offer enough power to reject the random walk even if mean reversion were the right answer. Economists had forgotten the lesson from introductory econometrics: “failure to reject the null hypothesis does not entitle you to assert that the null hypothesis is necessarily true.”
More provocatively (in “Zen and the Art of Modern Macroeconometrics”), I alleged that economists had subtly redefined the rules for a specific reason: it was too hard in macroeconomics to find statistically significant relationships. It is much easier to fail to find significant relationships. It hardly takes any work at all. But the affirmation “my research supports the hypothesis that the exchange rate follows a random walk” sounds much more respectable and publishable than “I have been studying exchange rates statistically for a year and have absolutely nothing to say about what makes them move.”
If one is in pursuit of the right answer, one needs to cast the net wider, to encompass a century-long time series, or a panel of countries. On a priori grounds, that is how much data it should take, before the test will have the requisite power. Sure enough, when one did that, one could reject a random walk in the real exchange rate, and find mean reversion.
Many have taken to using the “black swan problem” to mean a highly unlikely event, as the sub-prime mortgage crisis of 2007-08 is interpreted to have been. The way I would prefer to define it is when an event is considered virtually impossible by those whose frame of reference is limited in time span and geographical area, but that is well within the probability distribution for those whose data set includes other countries and other centuries (or those who make appropriate use of a priori theory, as with those irrational numbers). Analysts don’t cast the net widely enough. They can’t imagine that terrorists might inflict mass casualties by bringing down a buildings (New York, 2001) or that housing prices might fall in dollar terms (US, 2007) or that an advanced economy might suffer a loss of confidence in its debt (Greece, 2010). “I haven’t observed such a thing in the past, so it won’t happen in the future.”
These things had happened before, but mostly in times and places far away.
What do “black swans” have to do with it? An Englishman in the 19th century who encountered a black swan for the first time might have considered it a “7-standard deviation event,” even though one could have learned of their existence from ornithology books (Black swans had been discovered in