SMU SOE Seminar Series (August 6, 2026): Intraday Volatility Dynamics

TOPIC:

ABSTRACT

Return-based spot volatility estimates are noisy and prone to biases. This paper develops inference for the autocorrelation of intraday volatility changes that is robust to sampling errors in volatility estimates and biases due to jumps and microstructure noise. Our procedure builds on two results: First, we show that the limiting autocorrelation function of increments of a continuous-time process over shrinking horizons is parametrically determined and must coincide with that of fractional Gaussian noise. Second, we characterize the infill asymptotic distribution of realized autocovariances of spot variance estimators. The two results combine to yield a feasible generalized method of moments estimator of the autocorrelation of high-frequency volatility increments. In an empirical application to SPY transaction data, we document widespread short-term predictability of intraday changes in spot volatility. Finally, we discuss the implications for standard estimators based on high-frequency data, both in theory and through simulations.

Keywords: High-Frequency Data, Realized Autocovariance, Regular Variation, Rough Volatility, Smoothness Parameter.