If the Reuters comments system wasn't so annoying I wouldn't have to bother you lot with these ... also from Felix's site, Alan Greenspan:
Since the devastating Japanese earthquake and, earlier, the global financial tsunami, governments have been pressed to guarantee their populations against virtually all the risks exposed by those extremely low probability events. But should they? Guarantees require the building up of a buffer of idle resources that are not otherwise engaged in the production of goods and services. They are employed only if, and when, the crisis emerges.
An earthquake, in Japan, is not an "extremely low probability event". Like the eventual collapse of the San Andreas Fault (which will not be a "Black Swan", when it happens, btw), it is pretty much a probability one event. It is a low frequency event, but that is not the same thing at all. The words "only if, and" should be deleted from the last sentence quoted above, which makes the whole thing correct and shows you why social guarantees against inevitable but difficult-to-plan-for events are an obviously good idea, which is why literally every literate society ever has had them as a top priority. Alan Greenspan presumably hasn't got life assurance because seriously, death, how many of those are you likely to go through in your life?
Tangentially related to this, and to the post below, the original Twitter rumour that motivated that discussion was one to the effect that "Piers Morgan has been sacked by CNN". I would say that this is in the class of statements that were false when made, but which are basically true statements waiting for reality to catch up with them. If the newspapers ever mistakenly print my death, I plan to write to the obits editor concerned and reassure him that "History will vindicate you one day".
An earthquake, in Japan, is not an "extremely low probability event". Like the eventual collapse of the San Andreas Fault (which will not be a "Black Swan", when it happens, btw), it is pretty much a probability one event. It is a low frequency event, but that is not the same thing at all.
ReplyDeleteWhat, are you a bayesian now? Or perhaps, taking a Keynesian/subjectivist view of probability? If you were a frequentist, you would admit the 1-1 correspondence between frequency and probability. And of course, a frequentist view of probability makes a world of sense in the real world, as Greenspan's whole approach to the economy in advance of the GFC makes clear.
I know engineers have come in for some bad press in the last ten years or so, what with their proclivity for terrorism. But I was in Tokyo last month and I was relieved to think that engineers designed the skyscrapers rather than financial market analysts/journalists/traders, who presumably wouldn't have bothered making them earthquake proof.
ReplyDeleteIt seems to me that Alan Greenspan is shockingly ignorant of the Kelly criterion.
ReplyDeleteOT: this seems to echo d^2's point on Frank Field's use of "thinking the unthinkable" (ie "it is unthinkable because nobody in their right mind could possibly think it, not because it is original and exciting").
ReplyDeletejohn b: I got as far as the following sentence: "To
ReplyDeletetest this assertion, we have projected the implications of continuing to run primary deficits at the 2009-10 level".
This translates as "I am either an incompetent fool or an immoral political hack". Seriously, you pick the bottom of the worst recession since the war and decide that's a typical year to project forward? That's like someone saying "one hour ago it was almost completely dark. Now it's bright and sunny. Projecting this rate of increase forward, the world will catch fire by teatime."
If you were a frequentist, you would admit the 1-1 correspondence between frequency and probability
ReplyDeleteNo, I think this is an acceptable use of language - you don't have to take any particular view of the philosophy of probability to agree that there are infrequent but inevitable events; I'm just saying that while "There will be an earthquake in Japan during short time period t" is low probability, "There will be an earthquake in Japan within the next hundred years" is very high probability.
"Seriously, you pick the bottom of the worst recession since the war and decide that's a typical year to project forward?"
ReplyDeleteIs that where all the global warming denialists picked up the habit, or vice versa?
Not even wrong.
ReplyDeleteI'm just saying that while "There will be an earthquake in Japan during short time period t" is low probability, "There will be an earthquake in Japan within the next hundred years" is very high probability.
ReplyDeleteIt depends on how short your time period "t" is.
There have been 7 earthquakes greater than magnitude 8 in the past 100 years, 19 greater than magnitude 7, and 24 that were 6.5 or greater and caused significant deaths or damage. There have been 11 large earthquakes since 2000. This isn't a "once in five decades event."
Now, not all of these earthquakes are in populated areas, and any one particular local would have a lower probability of experiencing damaging shaking. But we are still talking about a fairly high probability event.