Earnings Announcements, Realized Volatility and its Components

I suggest and apply methodology for analyzing a consistent and non-parametric estimator of the integrated variance, the realized volatility, around earnings announcements. Modeling expected realized volatility as the out-of-sample forecast from a HAR model, I define the abnormal realized volatility as the difference between the actual and expected realized volatility. I find that the realized volatility is abnormally high throughout the event period, rejecting the semi-efficient market hypothesis. I then decompose the realized volatility into a continuous- and jump component to analyze what drives the volatility increase at the earnings announcement. The jump component of volatility is only abnormally high on the day of the earnings announcement and the subsequent day, capturing the burst of information. If one accepts the jump component of volatility as the best measure of the information effect, my results are then in line with prior literature, suggesting that the semi-efficient market hypothesis can not be rejected.