CIRJE-F-1074 "Testing the Validity of Non-Parametric Value Estimates in Treasury Bill Auctions Using Top-Up Auction Data"
Author Name

Marszalec, Daniel

Date December 2017
Full Paper PDF file 
Remarks Revised in May 2018

This note uses data from top-up auctions to test the validity of value functions I derive using the Hortaçsu and McAdams (2010) methodology in my paper on Polish Treasury Bills (Marszalec, 2017). The testing procedure assumes that bidders have the same value function across both auctions; in this setting, since the top-up phase price is fixed, a presence of the top-up bid can be used to pin down the position where the value function ought to lie. The test I propose rejects in over 70% of the bidding data when a top-up bid is observed, indicating that a bias may occur in the estimation method that does not model top-up auctions explicitly. The current note doesn’t find a bound on the magnitude of the bias - but finding such a bound for both the non-parametric models of Hortaçsu and McAdams (2010) as well as the semi-parametric model of Fevrier et al. (2004)is now work in progress.