Here at the University of Oregon, one of our specialties is developing models where agents in the macroeconomy don't have rational expectations, instead they learn about the economy over time. Of course, these models need to be taken to the data to see if people do actually learn in the way the models predict. But if the data sets contain too many "Very Serious People", the tests will surely fail. They learn nothing from experience:
The China Debt Fizzle, by Paul Krugman: Remember the dire threat posed by our financial dependence on China? A few years ago it was all over the media, generally stated not as a hypothesis but as a fact. Obviously, terrible things would happen if China stopped buying our debt, or worse yet, started to sell off its holdings. Interest rates would soar and the U.S economy would plunge, right? Indeed, that great monetary expert Admiral Mullen was widely quoted as declaring that debt was our biggest security threat. Anyone who suggested that we didn’t actually need to worry about a China selloff was considered weird and irresponsible.
Well, don’t tell anyone, but the much-feared event is happening now. As China tries to prop up the yuan in the face of capital flight, it’s selling lots of U.S. debt; so are other emerging markets. And the effect on U.S. interest rates so far has been … nothing.
Who could have predicted such a thing? Well,... anyone who seriously thought through the economics of the situation ... quickly realized that the whole China-debt scare story was nonsense. But as I said, this wasn’t even reported as a debate; the threat of Chinese debt holdings was reported as fact.
And of course those who got this completely wrong have learned nothing from the experience.