Title: From previous-tick to pre-averaging: Spectra of equidistant transformations for unevenly spaced high-frequency data
Authors: Vitali Alexeev - University of Technology Sydney (Australia) [presenting]
Katja Ignatieva - University of New South Wales Sydney (Australia)
Jun Chen - UNSW Sydney (Australia)
Abstract: To convert tick-by-tick data into equidistant series, the Exponential Moving Average (EMA) sampling scheme is proposed. EMA is a parametric generalisation of the two popular methods: previous tick and pre-averaging. In essence, the proposed scheme is a spectrum of schemes that span between these two extremes. By varying a degree of the smoothing parameter, the scheme is capable of focusing on only the latest observations or favour more pronounced averaging to reduce microstructure noise. When computing the realised variance (RV) and assessing its convergence to the integrated variance (IV), existing methods have their drawbacks. Simulation study and empirical analysis demonstrate that at ultra-high sampling frequencies (10s, 20s and 30s), the EMA scheme collapses to the pre-averaging. In contrast, for lower frequencies (30-min or lower), one can rely on the previous tick sampling scheme. For frequencies ranging from 1-min to 10-min, the EMA sampling scheme must be employed to achieve reliable RV estimates.