I have limited real auction data which I am trying to see if I can do any empirical analysis. The data consists of auction type (i.e., forward or reverse), protocol (e.g., Vickrey, Dutch, English, etc.), winning bid value, 2nd highest bid, length of auction and number of bidders. The dataset is large (~500+ for each protocol). I do not have bid-level information (e.g., how the bids are distributed for each auction) which really limits what I can do with the data.
As you see, the problem is that I do not have bid-level data (e.g., how each bidders were bidding). I had two ideas originally:
- Determine whether bidders are bidding according to the Bayesian Nash strategy but I don't think it is possible to establish a distribution of valuations since I do not have bid-level data.
- An analysis of the Revenue Equivalence Theorem by examining the actual seller revenue for each protocol versus the expected revenue but not sure if this is possible.
Any suggestions or suggestions of past studies will be greatly appreciated!