Bootstrap inference about integrated volatility (in Russian)

We extend the work of Goncalves & Meddahi (2009) who suggest using the iid and wild bootstrap for realized volatility instead of the asymptotic approach in order to estimate integrated volatility. We propose the block bootstrap and GARCH residual bootstrap approaches motivated by the persistence of the intraday term structure of returns. Using Monte Carlo simulations we show that the block bootstrap is more accurate for a low intraday frequency, more robust and valid. Another result is that the GARCH bootstrap outperforms others when the data imply strong persistence in conditional heteroskedasticity. It also demonstrates good inference on simulated data along the baseline model with a high frequency. However, the GARCH bootstrap is more computationally costly and less robust than the others.