[IWE] Galbraith's OpEd on what should be done about the Bailout.

[Ben Tilly]

On 9/25/08, Carl Forde <carl.forde@gmail.com> wrote:
> On Thu, Sep 25, 2008 at 9:28 AM, Ben Tilly <btilly@gmail.com> wrote:
[...]
> Ben & co., you might be interested in this essay on the limitations of
> statistical methods. In particular of financial models.
> http://www.edge.org/3rd_culture/taleb08/taleb08_index.html

Thanks, I will forward it to a friend on Wall St.

> Basically, there are areas where statistical methods aren't useful, yet they
> are still applied and the results acted on as if they provided meaningful
> and reliable information. The argument is that the financial industry is
> built on such models.

Oh, absolutely.  And it bugged me to no end when I worked there.

Take securitization.  In securitization you take a big bundle of
loans, which is a stream of money promised far into the future, then
slice it by time.  The first tranche pays off first, then the second,
then the third.  Defaults hit the last tranche, then the second to
last, and so on to the first.  Clearly the first is much safer than
the last.  These tranches are broken up into bonds and sold.  (This is
called securitization because it turns loans into securities that can
be traded.)

How much safer?

Various models were used to judge that.  And ratings were issued.  But
none of them took into account that the performance of different loans
from all over the country could be strongly correlated.  With the
result that a lot of people bought bonds that were backed by known
risky mortgages that they were told were safe, which weren't safe at
all.  But it gets worse.  The result of these loans failing was
foreclosures that killed housing prices that has caused people who had
traditional, safer, mortages to foreclose.

Obviously the models were wrong.  Rather predictably so as I said
years ago, and the author of your article apparently was as well.

Cheers,
Ben