A method for identifying high outliers in construction contract auctions

Abstract

Construction contract auctions are characterized by (1) a heavy emphasis on the lowest bid as it is that which usually determines the winner of the auction, (2) anticipated high outliers because of the presence of non‐competitive bids, (3) very small samples, and (4) uncertainty of the appropriate underlying density function model of the bids. This paper describes a method for simultaneously identifying outliers and density function by systematically identifying and removing candidate (high) outliers and examining the composite goodness‐of‐fit of the resulting reduced samples with censored normal and lognormal density function. The special importance of the lowest bid value in this context is utilized in the goodness‐of‐fit test by the probability of the lowest bid recorded for each auction as a lowest order statistic. Six different identification strategies are tested empirically by application, both independently and in pooled form, to eight sets of auction data gathered from around the world. The results indicate the most conservative identification strategy to be a multiple of the auction standard deviation assuming a lognormal composite density. Surprisingly, the normal density alternative was the second most conservative solution. The method is also used to evaluate some methods used in practice and to identify potential improvements.