UC San Diego

In Chapter 1, I find that stock characteristics do predict a stock's time-varying liquidity beta, i.e. its sensitivity to market, with the effect varying according to the assumed holding period using data on 30 small, medium, and large cap stocks between 1997 and 2002. I also find that liquidity is a priced factor for stock return even after controlling for market and stock measures of risk such as estimates of market volatility and stock level volatility. In order to mitigate problems arising from a small panel, I also test the returns model with ARCH errors on a larger sample of 2000 stocks. Chapter 2 accounts for endogenous liquidity in a standard asset pricing model. Loss of liquidity, especially during times of crises, needs to be incorporated into models of financial assets so as to forecast returns correctly. To identify the effect of endogenous stock liquidity on stock returns, an instrument is constructed from the NYSE, AMEX, and NASDAQ's decimalization program, which shrunk tick sizes from one-sixteenth of a dollar to one-hundredth of a dollar. Decimalization led to an increase in liquidity by allowing for narrower bid-ask spreads. Using daily price and quote data on U.S. common stocks, I find that as stocks becomes more illiquid, their future expected returns increase. In Chapter 3, I propose two related measures for algorithmic trading constructed from the Disclosure of Order Execution Statistics data in order to study the effect of algorithmic trading on stock liquidity. The first is the average time taken to fill an order once it arrives in the market, known as fill time, and the second is the proportion of orders executed within ten seconds of order arrival at the NYSE. Since the decision to use algorithms in trading is an endogenous one, I use the NYSE's introduction of autoquotes in 2003 to identify the causal effect of algorithmic trading on a stock's liquidity. Using IV estimation, I find that a one second decrease in fill time narrows spreads by two basis points