Bryan Caplan  

Is "Contagion" the Result of Statistical Discrimination?

The Coase Theorem and Divorce;... Folk Songs...

Yes or no: Is financial "contagion" the result of statistical discrimination? I.e., is it just a case where all financial institutions of a given type get judged by the average quality of their type?

Comments and Sharing

COMMENTS (4 to date)
Maniakes writes:

If I said instead that when a financial institution fails, it causes people to reassess theirs estimates of the inherent risk level of that institution's business model and thus reprice institutions with similar business models, would you consider that a yes?

JD writes:
Douglas Colkitt writes:


In any modern markets where derivatives dominate as soon as one player goes under than everyone else is vulnerable. If Bear Stearns collapses and can't honor their swaps then any counter-party to Bear is all of a sudden holding a bunch of worthless swaps on their books. This means that their financial position has completely changed, for example if they had an overall neutral exposure to corporate credit but with BS a positive corporate credit exposure, now instead of being hedged they are holding a position.

This increases their capital use and consequently they will have to liquidate or re-hedge to get back within their capital limits. This causes massive market disruptions.

Tom Grey writes:

I think Maniakes' first comment on repricing / using new facts to re-evaluate prior assumptions, is a key issue in contagion.

With respect to the risk of any new financial instrument, like junk bonds (80s), internet stocks (90s), or securitized mortgages (2000s), there is no "law of law numbers" to allow statistics enough time be statistically significant. So decision makers are using Bayesian a prior probability functions, which are rapidly and heavily influenced by new facts.

Perhaps most bank bonus-chasers didn't realize the huge "house prices always go up" assumption built into their risk models, but all the most "successful" high-volume investors acted as if they had this assumption.

So I think "no" is the answer. A general false assumption about an underlying asset (essentially no discrimination!) isn't quite the same, tho it's close, to statistical discrimination.

Comments for this entry have been closed
Return to top