« Sarah asks about age and productivity | Main | The More Loving One »
Making Knightian uncertainty operational
Tell us how "Knightian uncertainty can be made operational"...
Knightian uncertainty is not usually important when you are playing the first twenty (or is it these days thirty?) moves of a Ruy Lopez in grandmaster chess. When Knightian uncertainty matters, we should observe market participants investing more in opportunities for serendipitous discovery. This might, for instance, mean buying new books on a lark, traveling randomly around the world in search of insight, and in general mimicking wunderkind Ben Casnocha.
Do you think Knightian uncertainty is important in labor markets? If so, go out and hire some people on the basis of rumors.
Norton A. Myers's new and excellent Happy Accidents: Serendipity in Modern Medical Breakthroughs: When Scientists Find What They're NOT Looking For is one of the best books I know of on science.
#18 in a series of 50.
Posted by Tyler Cowen on March 10, 2007 at 07:28 AM in Economics | Permalink
Comments
My question is whether,with Knightian uncertainty, you should pursue some sort of minimax or precautionary principle? I don't believe these principles make much logical sense but they seem to approximate how we do actually behave in these situations. I wonder if there is some normative basis for them.
Posted by: harry clarke at Mar 10, 2007 9:45:14 AM
...
"Entering a new school for the first time, he says, he would seek out the library and then close his eyes and spin around with his hand outstretched. He would read every book on the shelf that his finger ended up pointing to, one a day -- forcing himself to take interest in something completely new and to devour a topic in depth."
Soul Proprietor
http://www.fastcompany.com/magazine/37/tyler.html
Posted by: anon at Mar 10, 2007 10:44:12 AM
Use financial options or insurance products to hedge against volatility
Use (a portfolio of) real option(s) to hedge against Knightian uncertainty
Posted by: Yan Li at Mar 10, 2007 11:14:34 AM
When Knight described the inherent uncertainty of business decisions, and how they are nonquantifiable, clearly on one level this is absolutely correct. But what metric do we apply to these decisions? Is it a function of volatility, if so, is in linear in variance, or perhaps nonlinear in some explicit way. If so, how can we produce a test that Knightian uncertainty exists, such as whether it demands a risk premium (for Knight, the reason business profits exist)? The scope of business decisions you suggests, seems a very qualitative metric, which is very Knightian, but also, very difficult to quantify.
Your solution is not so much a metric, but a strategy. Is it cost effective? That's not obvious, search takes time. Lots of grad students read omnivorously and just keep reading to no effect other than becoming insanely educated. Min-Max approaches are generally predicated on an assumption about everything having super fat tails, but are these cost effective?
Posted by: eric at Mar 10, 2007 12:52:04 PM
Actually, Keynes's treatment of uncertainty, which overlaps Knight's in
many ways, and appeared in print in his Treatise on Probability in the
same year as Knight's, 1921, is more sophisticated, with greater
gradations, and hence greater likelihood of being quantified part of
the time. Keynes allows for four degrees of it from classical known
probability to pure, unquantifiable uncertainty, whereas Knight only
had two categories, quantifiable risk and unquantifiable uncertainty.
Regarding Casnocha, he seems to imply that tenure is a problem for
universities, and that they should think about getting rid of it in
the future, although he does not come right out and say that. This,
of course, presumes that university administrators are actually competent
to guide university research, which I would question.
Posted by: Barkley Rosser at Mar 10, 2007 12:58:29 PM
We all start in uncertainty; with learning and experience, we move to risk. So in a world of uncertainty (e.g., exploring a new planet) one probably does better staying in one little corner and learning the odds. That contradicts the idea of pursuing serendipity and random trips. Sorry. (And did I really say "probably" in the second sentence?)
Posted by: Bill Conerly at Mar 12, 2007 1:12:20 AM
I am amazed that nobody is aware that attempts at succesful formal representations of Knightian unceratinty have existed for a while. Schmeidler developed the axiomatic basis for maximizing Choquet utlity, which is the appropriate utility to be used when probabiblities are non-additive (which in turn is tied to the idea of imperfect knowledge of state space.) Subsequently, Mukerji have shown that maximizing Choquet utility is the "procedurally" rational thing to do when faced with "uncertainty." Ghirardato has some further developments on that. There are many interesting implications: (1) why people generally do no prefer indexed wage or debt contracts (2) too much or too little trading.
Posted by: srinivas at Mar 12, 2007 11:28:06 PM
srinivas,
I did not lay out the intermediate cases described by Keynes in his
Treatise on Probability, but the Schmeidler solution is approximately
equivalent to one of them. When one is dealing with truly true
Knightian/Keynesian "uncertainty," mere non-additivity or "incomplete
knowledge" is not the problem. One is dealing with, as Donald Rumsfeld
so succinctly put it, "unknown unknowns."
Posted by: Barkley Rosser at Mar 13, 2007 2:10:07 PM
I have posted a different response, here.
Posted by: sammler at Apr 5, 2007 7:58:02 AM
大家好，我是臺灣人，從臺灣一個人搬家來到美國，環境很陌生，感覺很孤單。以前在臺灣幾家知名的徵信社工作過，我是一個優秀的徵信工作者，希望早點找到適合自己的工作。希望通過貴站，認識更多的朋友。
Posted by: 謝文豪 at Apr 2, 2008 2:03:06 AM
Could you indicate scientific papers where knightian uncertainty (subjective one)were operationalized?
Thanks, Marta
Posted by: Marta at May 31, 2009 5:01:20 PM
