Title: Portfolio diversification in the spectral domain
Authors: Martin Hronec - Faculty of Social Sciences, Charles University in Prague (Czech Republic) [presenting]
Jozef Barunik - UTIA AV CR vvi (Czech Republic)
Abstract: When investors' risk preferences differ across time horizons, the diversified portfolio needs to be immune not only against shocks aggregated across time horizons but also at specific time horizons. We apply spectral analysis into diversification-based portfolio selection models. By replacing the covariance matrix estimates with the cross-spectrum based ones, we restrict the optimization problems in these models to the desired frequency band, allowing an investor to target specific time horizons. Further, we generalize for an investor facing risk constraints at different frequencies by including the shape of the cross-spectrum into the optimization problem. We provide several numerical and empirical examples that show investors may benefit by considering not only diversification across aggregate risk sources but also across different frequencies.