Evolutionary learning in auctions

Classical auction theory relies heavily on the assumption of perfectly rational bidders that play according to their Nash Equilibrium Strategies. An alternative model is developed in this Thesis. It develops a framework for analyzing evolutionary learning in repeated auctions - bidders' strategies are influenced by the outcome of the past rounds in the game. Scientifically, the Thesis takes a complementary approach: On the one hand it develops mathematical techniques to quantify the outcome of learning in auctions. On the other hand it uses computational simulation. The simulation is first checked against the mathematical results and then used to assess the outcome of more complex setups that can't be analyzed mathematically. The main result of the Thesis is the instability of the NE in first price auctions, third and higher price auctions and double auctions under best response when compared to second price auctions.