Kang, Boo-Sung (2006-04). Empirical study on the Korean treasury auction focusing on the revenue comparison in multiple versus single price auction. Doctoral Dissertation.
Thesis
This dissertation pursues to find an answer empirically to the question of the revenue ranking between the multiple price auction and the single price auction. I also attempt to get empirical clues in terms of the efficiency ranking between the two. Under the assumptions of symmetric bidders and private independent value (PIV), I derive the optimal bidding conditions for both auction formats. Following the structural model estimation approach, I estimate the underlying distribution of market clearing price using the nonparametric resampling strategy and recover the biddersA? unknown true valuations corresponding to each observed bid point. With these estimated valuations of the bidders, I calculate what the upper bound of the revenue would have been under the Vickery auction to perform the counterfactual revenue comparison with the actual revenue. I find that, ex-post, the multiple price auction yields more revenue to the Korean Treasury than the alternative. I also investigate the efficiency ranking by comparing the number of bids switched and the amount of surplus change which would occur when the bidders are assumed to report their true valuations as their bids. I find that the multiple price auction is also superior to the alternative in efficiency which supports the current theoretical prediction. Finally, I investigate the robustness of my model and empirical results by relaxing the previous assumptions. I, first, extend the model and estimation to the case of asymmetric bidders where the bidders are divided into two groups based on their size. It shows that the model and estimation framework are still valid and that the empirical findings are very similar to the symmetric case. I also test for the presence of common value (CV) component in the biddersA? valuation function. I propose the simple regression model adopting the idea of the policy experimental approach. I obtain quite an inconclusive result in general but find some evidence supporting PIV for relatively higher bid prices while supporting CV for lower bid prices.
