Abstract

If an asset is perfectly liquid, one can trade the quantity one desires, immediately, without moving the market price. When liquidity is less than perfect, a trader must sacrifice on one or more of these three dimensions, perhaps trading a different quantity, over time and/or at a worse price. Previous work has generally focused on the price and time dimensions, assuming that a trader always trades his desired quantity or, similarly, that he does not have access to substitute assets to fill some or all of his trading demand. In this dissertation I examine liquidity from theoretical and empirical perspectives, emphasizing the quantity dimension of liquidity and its interaction with the time and price dimensions.

In the first chapter, Laurie Simon Hodrick and I develop a theoretical model of liquidity incorporating all three dimensions. Our main innovation is to allow all investors to choose whether and how much to trade. This change alters many of our classic adverse selection intuitions. For example, we find that bid-ask spreads may actually be decreasing in the significance of informed traders when the uninformed investor's preferences are the more binding.

In the second chapter, I use foreign-exchange transaction data to examine whether the quantity dimension is more binding at some times than at others. I find evidence suggestive that the quantity dimension binds more at the end of calendar quarters as investors refine the precision of their trades. I also find that the price impact of trades is greater when the quantity dimension is more binding.

In the third chapter, I investigate the determinants of relative repo specialness in the U.S. Treasury market. Repo specialness is a price indication of how strongly traders want to short a particular issue rather than a close substitute. In this respect it reflects how strongly the quantity dimension binds. My analysis shows how relative repo specialness changes with the auction cycle, hedging demand and factors relating to the potential for short squeezes.