This dissertation is composed of two essays concerning the role of sub-second high-frequency trading in regards to strategic trading behavior and price discovery.The first essay shows how sub-second periodicities in stock trading activity affect volatility and liquidity: Many more trades and quote updates arrive within the first 100 milliseconds than during the rest of a second. These periodicities most likely originate from algorithms that trade predictably by repeating instructions in loops with round start times and round time increments. This seemingly irrational behavior serves as synchronization mechanism for other investors. I find that during the beginning of a second, when trading activity is high, volatility increases but liquidity remains unaffected. These results are consistent with the Back and Pedersen (1998) model of strategic trading but are hard to reconcile with the predictions of Admati and Pfleiderer (1988). Overall I find that—perhaps surprisingly—trade clustering does not reduce trading costs.The second essay analyzes the price discovery process of gold by using high-frequency price series of three commonly traded gold investment products. I find that first: modern markets disseminate new gold pricing information in less than one hundred milliseconds, which shows that high-frequency data is important for price discovery in modern markets. Sampling intervals of one second or longer have lost most of the crucial timing information necessary to establish price leadership and price contributions. My second finding is that gold future contracts lead the price discovery process of gold, closely followed by a physical gold based exchange traded fund—prices of an exchange traded fund based on gold mining stocks do not provide much informational content, but track the price of gold closely. This implies that institutional investors, who predominantly trade gold future contracts, lead the price discovery process; retail investors (trading ETFs) do not contribute much to price discovery and act as consumers.
Read
