On-Line Detection of Changes in the Shape of Intraday Volatility Curves

We devise an on-line detector for temporal instability in the shape of average intraday volatility curves under a general semimartingale setup for the price-volatility dynamics. We adopt a block-based strategy to estimate volatility nonparametrically from the intraday observations over local time windows with asymptotically shrinking size. Our detector then tracks sequential changes in running means of the intraday volatility curve estimates. Asymptotic size and power properties of the detector follow from a weak form invariance principle, which is established under the strong mixing condition aligned with our semimartingale setup. Simulation and empirical results demonstrate good finite-sample performance of the proposed detection method.