Empirical bias in intraday volatility measures

Intraday volatility measures have recently become the norm in risk measurement and forecasting. This article empirically investigates the unbiasedness of three of these measures over four different datasets. We find that the three measures are significantly biased and that the bias can have either sign.

► This short article questions the empirical unbiasedness of three realized volatility measures. ► We investigate the realized variance, the bipower variation and the median realized variance. ► We run a simple test for bias in these volatility measures using four different financial datasets, ► We find that these measures are empirically biased, despite their underlying theoretical properties. ► We find that this finding is robust to sample splitting.