Volatility and High-Frequency Data
KERNEL
NOISE
MARKOV
GARCH
OTHER
High-Frequency Data and Estimators of Volatility
High-frequency price data contain valuable information about the underlying process. A challenging aspect of estimating volatility from high-frequency data is noise that conceals the efficient price. In my reasearch I have studied the empirical features of the noise and developed accurate estimators of volatility.Realized Kernel
Realised kernel estimators are designed to estimate volatility in the context of noisy high-frequency data, such as tick-by-tick transaction prices. Key results:- » A general theory for realised kernels is established, including a feasible limit theory. The class of realised kernels includes an efficient estimator of the quadratic variation.
- » The multivariate realised kernel is guarenteed to be PSD and robust to noisy and non-synchronous high-frequency data.
- » Combining subsampling with realised kernels results in less efficient estimators.
- » Challenging empirical features are discussed and analysed.
