by Peer Van Tassel
This paper identities systematic patterns within the trading day by analyzing high frequency data from a market index and nine individual stocks. Empirical results expand on the previously documented U-shape in intraday equity volatility by implementing non-parametric statistics to test for patterns in the jump and diffusive components of volatility. Additional results indicate that a recently developed non-parametric jump detection scheme may under-report the number of returns flagged as statistically significant jumps in the middle of the day while exaggerating the number of statistically significant jumps in the early morning and late afternoon. The paper concludes by investigating whether incorporating the observed patterns into a historical forecasting model can improve performance.
Advisor: George Tauchen