Self-recommending. But here is an excerpt.
Most empirical data do not come from experiments but from non-experimental sources such as surveys and routinely collected information. Along with Chris Doucouliagos and Tom Stanley, my research center examined 6,700 empirical studies encompassing 159 topics. We found that there is probably substantial bias in much of this literature. For example, the value of a statistical life, which measures how much people are willing to pay to reduce their risk of death, appears to have been exaggerated by a factor of eight. On average, the strength of the results may have been exaggerated by a factor of two. In a third of the studies, by a factor of four.
But overall, he is more upbeat than I am on economics as a science.
Pointer from Mark Thoma.
Aren’t we a little bit of Thaler reality here? The cost of death is hard to define as it is:
1) A such very rare event that easily change per person depending upon their mood and situation. I like to hike and Geocache (basically a tupperware treasure hunt) and live by desert hills that have rattlesnakes. So my value of death is ridiculous low in that situation to hike the hills and search for tupperware containers.
2) The chances of death can calculated at .00001% but really be .00002% which is really seems insignificant but put in cost per death changes your calculations a lot.
From John: Credit rating agencies and highly paid gurus are largely selling products of little or no value to the gullible.
If this were true about credit rating agencies, it would be true that banks which do not do credit checks would have similar default rates as those that do use the rating. I’m 99.9% certain that this is not true.
On the other hand, a bank doing its own quick credit check might well be “as good as” the credit rating based loan approval rate.
Unsurprisingly, this hyperbolic non-scientific claim comes in a note asking for more and better science — much as con men are often quick to warn you of … other con men.
If economics WAS a science, which could be accurately used to predict stock market prices thru a probability that a stock price would increase or decrease in some limited term, then the resulting stock market prices would follow the random walk.
Because any econ law which can be used to make money WILL be used, and this use changes the reality so that the previously useful law becomes invalid.
This was cleverly shown in Back to Future 2, where an almanac of horse race winners from the present was used in the past to help the bad guy make money — until his money making changed his present enough that the former future was invalid.
Only economic laws which can’t be used to make money will remain valid. That’s a big limit on econ.