Auto-Regression Auctions

This paper proposes two auto-regression auction formats namely, simple autoregression and multiple autoregression auctions. Both auto-regression auctions use the valuable information available in both past auctions and current auction to predict the start price, recommended price and reserve price of the current auction’s object which allow auctioneers to maximize the current auction’s revenue. In simple autoregression auctions (SARA), an autoregression moving average model ARMA(p,q) is used to predict the start price of the current auction object based on a time series of start bids placed in past auctions, and predict the recommended object price of the current auction based on the time series of bids placed by all bidders during the current auction. The reserve price that can be accepted by the seller of the object in the current auction is predicted using ARMA(p,q) based on the highest bids placed by the winners of past auctions. In multiple auto-regression auctions (MARA), the auction runs in multiple rounds where the start price of the first round in the current auction is predicted based on a time series of start bids placed in past auctions using ARMA(p,q) model. The temporary recommended price of the auction object in each round of the current auction is predicted using ARMA(p,q) model based on a time series of all bids placed in the same round during the current auction. The temporary reserve price of the auction object in each round of the current auction is calculated based on a time series of bids placed in the same round of past auctions. The start price of each round is based on the reserve price of the auction object in the previous round and on a time series of past start bids in the same round of past auctions. At the end of the current auction whether it is SARA or MARA, the winner with the highest bid will pay the average of the predicted final recommended price and predicted final reserve price of the current auction object if the final recommended price is within the interval of the final reserve price and seller price. Keywords: Auctions, Bids, Time Series, Auto-regression Models, Moving Average, Open Auctions, Closed Auctions