Selecting the winner in Texas school milk procurement auctions: Theory and evidence

This first part of this dissertation models how auctioneers in Texas school milk procurement auctions select the winning bidders. Because the bids are multidimensional, the auctioneer has to design a scoring rule which assigns weights to the various bid components to create a summary measure of each bid. The scoring rule used by the auctioneer to compare bids is not public information and I estimate it by means of a conditional logit model. For this purpose, I assume that the auctioneer uses a “naive” scoring rule—as opposed to the theoretically optimal rule—because he can not commit to a rule which distorts his preferences. I find most notably that non-bid characteristics have a strong impact on the winning probability: incumbents in a school district fare much better in the auctions than their non-incumbent competitors. I propose and test two different hypotheses why it may be rational for an auctioneer to favor the incumbent.;The second part of the dissertation analyzes the effect of the scoring rule on the strategic behavior of the bidders. I estimate bidding strategies by regressing the bids on a variety of cost related variables as well as incumbency status. My analysis includes winning bids only because there is evidence that a bidding ring at the auctions used a phantom bidding scheme to disguise its presence; as a result, non-winning bids have no meaningful interpretation. I control for the selection bias which arises when only winning bids are used by means of a two-step procedure. The results show, most importantly, that incumbents in large districts react to being preferentially treated and submit substantially less competitive winning bids than non-incumbents.