Staff Working Paper No. 1,061
By Charlotte Grace
This paper provides a method for comparing the performance of different models of bidding behaviour. It uses data on participants’ bids but does not require data on their values. I find that a model of ‘truthful bidding’ – bidding one’s true value for liquidity – outperforms a conventional model in which bidders shade their bids to maximise their expected surpluses, in the Bank of England’s uniform-price divisible-good liquidity auctions. I provide two possible explanations for this result. First, when bidders are sufficiently risk averse, optimal strategies in the conventional model approximate truthful bidding. For the conventional model, I develop new identifying conditions which allow for risk aversion. I find that the degree of risk aversion required for truthful bidding to be approximately optimal is consistent with that found in studies that are the most similar to my setting. Second, the optimal strategy can be complicated. Truthful bidding is preferable, even for risk neutral bidders, if the cost of calculating what would otherwise be the optimal strategy exceeds around 5% of bidder surplus.