« Voting Cycles: A Game Theory Problem | Main | Cascades in Affirmative Action; Sander Data »

November 12, 2004

Marsh and McLennan Insurance Brokers: Common Agency

I gave my options paper at Georgia State this week at their
department of risk (a neat idea for a department!), and
talked with some people about the Marsh and McLennan scandal. Marsh is a very large
insurance broker, which companies such as Delta Airlines
hires to find the best insurance deal for them. Marsh was
engaged in fraud, it seems, but another practice, more
common and perhaps defensible, was that it took commissions
from both sides of the transaction-- from the client, Delta,
and from the insurance company that got Delta's business.
Moreover, the commission from the successful insurance
company was based on the ex post profitability of its
contract with Delta, I was told.

Could there be an efficiency reason for this "common
agency" problem-- in which the agent, Marsh, tries to
satisfy two principals, Delta and the insurance company?
Maybe. Our first thought is that this is simple corruption--
that Marsh is supposed to be acting just on behalf of Delta,
but secretly takes bribes from the insurance company. But
can we imagine a situation in which the "kickbacks" or
"commissions" to the insurance company are known to Delta,
but Delta still wants to hire Marsh?

Here is a possibility. Suppose Marsh's function is to
warrant that an insurance customer is a customer worth
having--that it has no hidden costs for the insurance
company. When Marsh says that a customer is a "good
customer", the insurance company gives the customer a low
price for insurance, but asks Marsh to back up its claim by
accepting a financial penalty if the customer turns out to
be a "bad customer", by agreeing to take 10% of the profits
from the insurance contract. If the customer is bad, that
10% amounts to nothing; if the customer is good, Marsh gets
some money. Marsh would then accept only good customers,
and good customers would agree to this, because it is a way
they can prove they are good to insurance companies.

I don't know enough about Marsh's particular situation
to know if this fits it, and formal modelling might show up
some inconsistency in my story, but it has at least slight

Posted by erasmuse at November 12, 2004 02:34 PM

Trackback Pings

TrackBack URL for this entry: http://www.rasmusen.org/mt-new/mt-tb.cgi/299


Post a comment

Remember Me?

(you may use HTML tags for style)