International Law

Backtracking from an Uncertain Past

Image via Wikipedia

In relation to several posts I have written about Charlie Munger's use of inversion and preparing for highly unlikely events, a reader writes:

"As an engineer/applied mathematician trained in control theory, I think I see where you are coming from.

This is the common practice, e.g., if you were designing an aircraft autopilot system - you'd define a lot of scenarios and work out what can go wrong.

However, there are two issues in markets: (1) How do you know which events to backtrack from?

It makes perfect sense that people know how they might have avoided a catastrophe after it has happened because you have a single definite event to backtrack from.

How do you make a list of such events before they happen so you can protect against all of them? (2) If you take the view that you should just protect against the absolute worst case, then of course you shouldn't invest at all!

Is that what we really want?"

There a several excellent points here, and it won't surprise anyone when I say that this is a problem which is not a calculation.

It would be both simple and wrong to say backtrack from the worst case scenario, because anchored in the present our ability to predict a worst case scenario is going to be attenuated by how much we think that future is going to be like the past - a past we don't know that much about anyway, but think that we have a distribution of past events upon which to forecast.

However, the technique that I am advocating requires you to pick some uncertain event that looks highly unlikely now.  Jump to that future and work your way back.

For example, if you think that the newspaper industry is finally about to die, then pick the event that the New York Times goes into bankruptcy.

Mike Hishorn's piece in the Atlantic, on a possible New York Times bankruptcy, is an example of this technique.

"But what if the old media dies much more quickly? What if a hurricane comes along and obliterates the dunes entirely? Specifically, what if The New York Times goes out of business--like, this May?"

Is he right?  I don't know, but his analysis technique from the future through the uncertain past is interesting.

The subprime mortgage market blew up because people told themselves that housing prices were regional and that they had never seen a correlated downturn in housing prices - and so the CDO's they sold were diversified by geography.

Again, the worst case and probabilistically unlikely even is a sudden correlation of housing prices.  Assume it happened.  How could these previous uncorrelated price movements all move together?  It cannot be too hard to figure out that the answer must be related to what happened to Long Term Capital.

This technique is not likely to give rise to calculations, but it is helpful to counter anchoring on model prediction.

Thanks, again to the reader who raised this point.  I hope that I have made some progress on this problem.

Full post as published by Misleading Advertising Law on March 11, 2009 (boomark / email).