Challenging Golden Standards in Ewma Smoothing Parameter Calibration Based on Realized Covariance Measures

In this paper, we investigate the calibration of the smoothing parameter in an exponentially weighted moving average (EWMA) for realized covariance matrices. Although it is the crucial determinant in steering the dynamics of the EWMA, little attention is drawn on its calibration in many empirical applications with regard to predictive accuracy. Whereas the popular risk metrics calibration of JPMorgan Reuters (1996) has been proven to be successful in many applications that are based on daily or monthly return data, we demonstrate that the golden standard degree of smoothing is not transferrable to applications with realized (co)variances which convey a substantially more informative signal. Moreover, we propose a data-driven heuristic for the calibration of λ and demonstrate in our empirical application its superiority over ad-hoc choices for λ particularly in multivariate settings