Volatility Estimation in the Era of High-Frequency Finance

Abstract

Over the last two decades, ultra-high frequency (or tick-by-tick) transaction data has become increasingly available. This surge of high-frequency finance data has brought disruptive revolution that makes modeling asset prices as continuous-time processes more possible than ever before. This is because we can now witness market microstructures and stock market volatility over tiny time intervals. This chapter reviews some general frameworks like realized volatility (RV) in estimating the latent volatility and their recent developments in the era of high-frequency finance. New empirical facts are presented to help lay the foundation for creating intraday volatility models that can overcome noise interferences in high-frequency finance data. These facts also help explain some stock market anomalies like volatility jumps and flash crashes, which favor intraday RV over the traditionally used daily RV as a reliable physical measure of market risk.