SANTA FEDERICO, ANDREA PETRELLI, JUN ZHANG, AND VIVEK KAPOOR
We employ here a latent variable framework, which accounts for default correlation by introducing common factors among the variables themselves. Two semianalytical approaches are described—namely Fourier transform and direct calculation of the convolution integral via a recursive method. We present the formulation in a loss continuum framework and later we briefly describe the mapping of relevant quantities after having discredited the loss support for numerical integration.
Define the latent variable Xi = f(w)Zi, the normal standardized asset return Zi has a systematic component and an idiosyncratic one (Z i = √p Z + √1−ρ ε i), w is ...