« China Op-Ed of the day | Main | Best economics paper of 1958 »
What you don't know that you don't know
Bryan Caplan writes:
A common Austrian slogan is that "Neoclassical economists study only cases where people know that they don't know; we study cases where people don't know that they don't know."
He demands a good example in support of the Austrian view. I would cite the arrival of the Spaniards during the time of the Aztecs and the subsequent conquest (see also Fabio's comment on Bryan's post). This was not literally an unimaginable event, since it had (in modified form) been foretold by Nahua prophecy, but still the Aztecs had no ability to respond effectively, given their prevailing mental frameworks.
More generally, look at the implied volatility embedded in options prices. Is it forecasting how much volatility is really out there? Here's one possible quantitative measure: if you have futures on options, take the measure of "surprise" in implied volatility (the change in implied volatility not forecast by the futures price for the options contract) as the relevant measure of "what you didn't know that you didn't know." I'm not saying this figure is large, or even necessarily positive on average, but I do think it is a meaningful concept. Arnold Kling responds to Bryan as well. Here is my previous post on this topic.
Posted by Tyler Cowen on June 17, 2008 at 10:18 AM in Philosophy | Permalink
Comments
The Big Bang
Posted by: aje at Jun 17, 2008 11:24:07 AM
Nitpicking, but I don't like his Trojan horse example from Virgil's Aeneid.
It's a great poem, but it was written by a Roman at least 700 years later than Homer's account of the Trojan war. Greek myths are packed full of inconsistencies and implausible errors (Cronus mistaking a stone wrapped in a blanket for his baby son for instance). Those don't seem to have troubled the Greeks at all. The Romans had a more literal attitude: I suspect Virgil added the Laocoon stuff since the stupidity bugged him. Also he believed the Trojans were the Romans ancestors, and didn't want them to look like idiots.
Posted by: TheophileEscargot at Jun 17, 2008 11:48:35 AM
The request was not for events anticipated but not in much detail - that is most every event. The request was for events for which people didn't even have a category to describe them.
Posted by: Robin Hanson at Jun 17, 2008 11:52:14 AM
How about relying on behavioral economics? Confirmation bias generates self-deception. People refuse to know what they don't know. Hierarchies can re-inforce that.
In addition, as Ernst Fehr shows, strategic complementarities, can lead people to imitate others' ignorance (or confirmation bias). Personally, I bet liquidity constraints can also reinforce the process in financial markets.
Examples: The housing bubble. The French keeping troops on the Maginot line even when intelligence clearly showed the Germans flanking to the North.
Posted by: Keith at Jun 17, 2008 12:15:21 PM
For fully unanticipated, there is the impact of disease on the Aztecs and other native americans. Nobody on either side anticipated the devastation that would result when the Spaniards and others brought over European diseases. A 75% population loss was beyond their expectation, and vastly exceeded to worst prior experiences with famine and war.
Posted by: rjh at Jun 17, 2008 1:13:09 PM
I am really, really, not an expert here, but here´s how I understand this problem:
Right now, we know there is a possibility that the world economy will go into recession, but we don't know the exact probability of that event, or how big it would be, or which nations would be hit the hardest. But we know we have to solve these puzzles. These would be known unknowns.
On the other hand, it is conceivable, but completely outside our horizon, that economic crisis will lead to World War III. If that happens, we will not have problems measuring the Betas in our models, the whole set of variables will have been wrongly picked. This sort of data that screws up the whole exercise would be the unknown unknowns.
Two additional comments:
1) Unknowns are not known or unknown by virtue of their specific nature. Unknown unknows may become known (and, maybe, known unknowns may be forgotten).
2) While I admit the concept makes sense, I really can't figure how one could study this kind of thing empirically.
Posted by: NPTO at Jun 17, 2008 1:27:11 PM
"The request was not for events anticipated but not in much detail - that is most every event. The request was for events for which people didn't even have a category to describe them."
You mean like the (heretofore unknown) 5th horseman of the apocalypse?
That sort of thing?
Posted by: meter at Jun 17, 2008 1:55:43 PM
NPTO,
I think one could use set theory to think about these sort of things. I imagine that we could think of the set of all possible events and then visualize the set of "known unknown" events as a subset of that set, and the set of "known" events as a subset of the known unknowns. Then, one could see the "unknown unknown" events as "all events" - "known unknowns" (where - is the set difference operator).
Now, of course, we have to determine the size of "all events". While, in some cases, "all events" might be infinite and uncountable, I'd suspect that there are cases where that set is countable due to constraints imposed by physical and/or human laws. For example, it is physically impossible for temperature to go below 0 Kelvin. It is a law of physics that gravity is an attractive force. Working with natural constraints like these, I think it ought to be possible in principle to determine the size of the overall event space, allowing for the generation of useful probabilities.
Of course, it is possible to overconstrain the overall set by choosing constraints that exclude events that are possible but highly unlikely. Also, the person choosing the constraints is likely to bring in their own personal and institutional selection biases. However, I still think its a useful way of approaching the problem of dealing with things that you don't know you don't know.
Posted by: quanticle at Jun 17, 2008 2:38:37 PM
I don't think that Robin identifies the problem correctly. Ex post there is always a category you can put an event under, if only the general category of "unexpected event." The relevant question is whether fuzziness of categories leads to systematic mistakes. Maybe it does or doesn't but I think that way of posing the problem makes perfect sense and it is not the kind of meaningless nonsense that Bryan wants to attribute to the Austrian theory. That said, the Austrian theory can be charged with insufficient clarity on what the concept of radical uncertainty means.
Posted by: Tyler Cowen at Jun 17, 2008 2:48:47 PM
The next major scandal.
I have an entire section on my blog devoted to unintended consequences. Of course, not all unintended consequences are unanticipated. But the fact that counterproductive policies are routinely adopted in the face of such anticipation, often driven by headlines that no one could have predicted, could itself be viewed as a humbling example of second-order uncertainty.
Posted by: M. Hodak at Jun 17, 2008 3:26:18 PM
Other than the timing of events, there is nothing that comes to mind that has outright surprised me. That includes 9/11, the flooding of New Orleans, or any of the recent bursting bubbles (dotcom, real estate, and soon-to-be commodoties).
I'd add the decline of a certain empire, which I know won't be a popular sentiment in this particular blog.
Posted by: meter at Jun 17, 2008 3:59:24 PM
Er, commodities.
Posted by: meter at Jun 17, 2008 4:19:50 PM
I've been following this discussion this morning, and in my own research came across a simple example that I haven't seen mentioned here yet. A young child does not know that she does not know calculus -- an unknown unknown.
Posted by: JD at Jun 17, 2008 4:23:48 PM
I've been following this discussion this morning, and in my own research came across a simple example that I haven't seen mentioned here yet. A young child does not know that she does not know calculus -- an unknown unknown.
Posted by: JD at Jun 17, 2008 4:23:52 PM
Stock picking vs investing in an ETF. Stock pickers don't know that they don't know, otherwise they would pick the ETF every time.
Posted by: Russell Nelson at Jun 18, 2008 2:10:51 AM
How about this one: An Austrian economist knows that he doesn't know anything that would make him employable. He doesn't know that he doesn't know economics :)
Posted by: steve at Jun 18, 2008 4:49:13 PM
"we study cases where people don't know that they don't know."
How pure do we have to be in our "unknown unknowns"? September 11 2001 was eminently predictable - I was shocked, but not surprised - and Taleb's famous example, the black swan, was hardly even surprising, if thought about rationally. Quantifiable versus unquantifiable, or insurable versus uninsurable, would be a far more useful distinction, and is where the Austrians can usefully critique the neoclassicals.
Most ideas of unknown unknowns focus on the negatives - war, famine, plague, terror, which are all significantly frequent not to count - yet the most obvious and significant example is positive. Until approx. 1800, world GDP had never grown by more than a fraction of a percent per year. That GDP per capita would then grow at 1.5% for 200 years, while population grew at over 1%, would have seemed exceptionally unlikely, far more so than any amount of war etc. Go back a further 300 years, to when European GDP per capita had been static for 1000 years, and the idea of such profound long term growth would probably have seemed impossible. If there had been any economists around in 1500, I guess they would have been convinced of the impossibility of all but the slightest long term growth; after all, conditions might fluctuate, but everyone knew that while populations might grow slightly, people never got any richer, on average - and they would have had several thousand years of statistics* to prove it. They could easily have statistically "proven" the negative serial correlation of income per capita, over most all of recorded history.
*Assuming, counterfactually, that they had such statistics and the ability to interpret them.
Posted by: Timothy at Jun 18, 2008 5:44:27 PM
How about: all the answers that nobody posts here? By definition.
Posted by: Jordan at Jun 18, 2008 10:01:05 PM
Couldn't you just look at the implied vol on an option on an option?
Posted by: Kyle at Jun 18, 2008 10:45:02 PM
my impression is that "the unknown unkowns" should be factors which significantly impact the economic activity, but have not been studied in an organized fashion.
Since I am not an expert on Austrian econ, I don't have a good example for it, but i have examples from the larger world of econ/finance:
*) when Shleifer &al publish their paper on "Law and Finance", there was not much attention paid to the way the legal framework (in the world of finance) influences the development of the financial sector.
*) institutional econ: there used to be a lot of talk about growth without taking the institutional framework into account, now it is not the case anymore.
i don't think economics goes towards predicting events, as towards discovering principles that govern the economic development.
As for the implied volatility - it is a well known problem, for people working with derivatives. Volatility is a good example of a problem which has been around for a long time, known, and still without a satisfactory framework developed (despite active research from all sides).
Posted by: lb at Jun 19, 2008 4:38:58 AM
λ=h/p → P=h/λ (The story of de Broglie hypothesis)
plus my experience of learning this hypothesis
Posted by: T. at Jun 19, 2008 12:37:55 PM
We are professional cheap kids shoes online store .We provide Nike shoes cheap ,Nike air jordan,cheap kids clothing
cheap womens clothing,Nike shoes online store,Nike shoes cheap,Nike shoes for kids,air jordan sneakers,nike air jordan,air jordans,brand shoe,shoes brand, dropship Nike shoes , brand clothes,brand clothes ,shoes brand,cheap kids clothing for cheap wholesale Air Jordans shoes for kids & women.
Posted by: air jordan sneakers at Jul 10, 2009 2:33:32 AM
