Robust estimation of intraweek periodicity in volatility and jump detection

Abstract

Opening, lunch and closing of financial markets induce a periodic component in the volatility of high-frequency returns. We show that price jumps cause a large bias in the classical periodicity estimators and propose robust alternatives. We find that accounting for periodicity greatly improves the accuracy of intraday jump detection methods. It increases the power to detect the relatively small jumps occurring at times for which volatility is periodically low and reduces the number of spurious jump detections at times of periodically high volatility. We use the series of detected jumps to estimate robustly the long memory parameter of the squared EUR/USD, GBP/ USD and YEN/USD returns.

Macroannouncements, bond auctions and rating actions in the European government bond spreads

“…where |r t,i | is the absolute value of return on day t and time-interval i defined as p t,i −p t,i−1 and σ t is the bipower volatility of day t. [19] show that the filtered jump test statistics increases the accuracy of intraday jump detection methods. To obtain an estimate of the periodicity component s t,i ,…”

“…we implement the [19] robust estimation technique based on the Truncated Maximum Likelihood (TML) estimator. The parametric specification we adopt for s t,i is similar to the one proposed by [20] but here we include even government bond auctions and rating actions as follows:…”

“…Average yield and bid-to-cover are collected just for auctions concerning 10-year bonds. Boudt et al (2011), defined in (3) and (4), at the 5% significance level as well as the average absolute size of jumps. Panels B-D of the table report a preliminary analysis of the degree of association between jumps and macroannouncements, government bond auctions and rating actions by applying the z-test to compare the frequency of jumps occurrence around the event in analysis with respect to no-event situation.…”

“…***, **, and * denote 1%, 5% and 10% significance level, respectively. Boudt et al (2011), defined in (3) and (4), at the 5% significance level. Moreover, in order to identify a sufficient number of cojumps for further analysis, we define a cojump if two or more jumps occured in a 15-minute time window.…”

“…The dependent variable is the contemporaneous Cojump defined in (5) based on jumps identified according to Lee and Mykland (2008) test corrected by the intraday periodicity of the volatility as proposed by Boudt et al (2011) and defined in (3) and (4). The interecept of the logit model is corrected by (10) as suggested by the prior correction approach.…”