Revenue Maximizing Auctions When Types Are Multidimensional
<-- Return to the list
Date: 03-28-2006
Start Time:
1:00pm
End Time: 2:00pm
Speaker: Rakesh Vohra, Northwestern University
Location: Mudd 303
Abstract
A title that is both a mouthful and opaque. Consider a monopolist selling a limited supply to a group of buyers whose preferences are private information to them. A buyers preferences can be encoded by a vector, called their type. It is usual to assume that they type is single number, representing perhaps a buyers monetary value for an object. However there are many applications where it is more reasonable for a buyers type to be multidimensional. For example, in a procurement setting, a buyers type might be their marginal cost of production as well as their capacity.
In this talk I will describe a network flow interpretation for the problem of finding a revenue maximizing incentive compatible auction when types are multidimensional (this part is based on joint work with Alexey Malakhov). I will argue the utility of such an interpretation with a "hometown" example (there are more), motivated by a model of Vulcano, van Ryzin, and Maglaras. In that model they consider the problem of a monopolist selling a fixed quantity over a finite number of periods to buyers with unit demands and private valuations who cannot postpone their purchases. In the presentation I will extend the model to allow the buyers to have multiunit demands (which will be private information). In addition the buyers can be strategic about when they choose to reveal their presence to the seller (this part is based on joint work with Mallesh Pai).