The benchmark case corresponds to an auction with N=5 potential candidates. Firms' construction costs are randomly drawn from a uniform distribution Ii?U [100; 1,000], and demand beliefs from an independent distribution Qi e?U [10; 30]. The contract-term is set at T=30 years for the fixedterm auctions. Price is fixed at P=0.75 for the case of maximum payment auction; and payment at Z=100, for the price auction. With these values, the expected range of total revenue for firms is [225, 21 675]. For the average type of firm, with cost Ii = 550, this implies that profits may vary between [- 325, 125], so the project is quite risky and only attractive to firms with low costs. Table 1 shows the outcomes of each type of auction after performing 5,000 simulations, at each of which five bidders were randomly chosen regarding both their cost types and beliefs on future demand levels. Bids from all firms are computed and evaluated to determine the auction's winner. The auction's outcome is then compared with the actual ranking of firms' costs to calculate the probability of selection error. [ INSERT

Table 1. Average outcomes from auctions (Case with no maintenance costs) 'Winner's real \"'inner's 'Winner's Probabilityr of construction expected expected Probability of contract cost El'ots demand selection error renegotiation Price auction 314.9 132.0 24.2 41.2% 33.5% Payment auction 291.5 141.7 23.5 35.5% 33.9% LPVR \"dim 243.3 150.3 19.9 o o Ill expected values 250 0 2|] 0