A0264
Title: Spectral factors for functional data
Authors: Maria Grith - Erasmus University Rotterdam (Netherlands) [presenting]
Abstract: A stylized fact has emerged that volatility, skewness, kurtosis, and term structure factors effectively forecast future implied volatility surfaces. These surfaces reflect investors' anticipated market conditions at various points in the future. A novel approach is proposed for disentangling risk factors that capture fluctuations across cycles of different lengths. Specifically, we adopt a double orthogonal decomposition of the implied volatility surfaces in the time and space domains. Our method allows us to estimate frequency-specific risk factors that are spectral counterparts of those commonly identified in the existing literature. These spectral factors offer valuable insights into the behavior of various investors operating under distinct market conditions and may reduce dimensions in the factor space in an economically-meaningful way. In addition, our approach demonstrates the potential to improve the accuracy of forecasting implied volatility curves relative to the traditional methods.
