This dissertation studies two features of high-value auctions that are not explicitly captured by the standard models in the auction theory literature. The first is that bidders in auctions for valuable assets sometimes have binding budget constraints. Standard models of auctions assume that bidders can submit any bid up to their valuation (or willingness to pay). An existing literature has developed models where bidders may face binding budget constraints and from these models has concluded that the presence of budget constraints has important implications for the relative performance of different auction formats, and as a consequence argues that the presence of budget constraints should be an important factor used in choosing an auction format.

Chapters 2 and 3 develop and study a model of budget constraints where the budget constraint is chosen explicitly in the model in response to a principal-agent problem between each bidder and a corresponding principal. In previous literature, the budget constraint is assumed to be given by some exogenous procedure, and hence is not affected by changes in the auction rules. The model presented here, however, allows the choice of budget constraint to depend on the auction rules, and the main result of Chapter 2 shows that allowing for this effect leads to outcomes that closely resemble the classic results from the auction literature without budget constraints.

Chapter 3 investigates the theoretical predictions of Chapter 2 in an experiment involving undergraduate students at the University of Maryland. The experiment is designed to evaluate the decisions made by the subjects acting as the person responsible for deciding on a budget for the bidder. We perform treatments where the bidding behavior is simulated by computerized agents and ones where half the subjects in each session play the role of the bidder. Our results indicate that the subjects take the auction rules into account when deciding on their respective bidder's budget, and the direction of the response in the data agrees with the theoretical predictions.

Chapter 4 studies a separate feature of high-value auctions that is not captured by the standard auction models. That is, the bidders in the auction may have valuations for the auctioned item that depend on the the identities of the other winning bidders. If the auction determines the structure of the market the bidders will compete in after the auction, the bidders' values for the items will be affected by who participates in that market. The typical notion of efficiency in the auction literature corresponds to maximization of producer surplus in this model, but the auctioneer may also be concerned with total surplus in this environment. The main results show that these two notions of efficiency do not agree in this model, and that a sequential auction favors maximization of producer surplus, while a sealed-bid auction can favor maximization of total surplus. The key distinction between the two is that the sequential auction is assumed to reveal the identity of early winners to the later winners, while the sealed-bid auction reveals no information to the participants until the auction concludes.