Treasury Auctions, Uniform or Discriminatory?: An Agent-Based Approach

This study explores the use of the agent based computational economics (ACE) technique to address the question of how a Treasury should auction its securities. In particular, this study explores whether a Treasury should use a discriminatory-price rule. Buyers are modeled as profit seekers that are capable of submitting strategic bids via reinforcement learning. The buyers' profits are determined by auction prices and ex-post competitive resale prices. Experimental designs focus on four treatment varibles: (1) the buyers' learning representation; (2) market structures; (3) volatility of security prices in the secondary market; and (4) relative capacity (RCAP). Experimental findings show that security price volatility in the secondary market has little effect on market outcomes. However, market outcomes are sensitive to market structures, RCAP, and the buyers' learning representation. The two different auction rules result in different, persistent, systematically patterned market outcomes. Moreover, these findings help to explain why discrepancies have arisen among previous Treasury auction studies.