Fallible estimates + Competition = Winner's Curse

Some time ago Steve McConnell and I had an interesting debate, via his blog.  I suggested that when we combine estimates such as those we have in software (which have high uncertainty early in the project) with a competitive market containing price-sensitive customers; then market forces conspire to bias customers' choices towards those supliers who have accidentally under estimated.

It turns out the concept has a name, "The Winner's Curse", and it's been known to economists for over 200 years!  

I've updated my original post with details and a link to an excellent article from the Simula Research Lab in Norway.