Time-consistent evaluation of credit risk with contagion (Q2667125)

In this paper, the authors present an intensity approach to credit risk. The defaults of companies are modelled by a point process \((P_t)_{t \geq 0}\), which is of the form \[ P_t = \sum_{k=1}^{N_t} \xi_k. \] Here \((N_t)_{t \geq 0}\) is a point process with intensity process \((\lambda_t)_{t \geq 0}\), and \(\xi_1,\xi_2,\ldots\) are independent identically distributed random variables. It is assumed that the intensity process \((\lambda_t)_{t \geq 0}\) satisfies the stochastic differential equation \[ \text{d} \lambda_t = \kappa(\theta - \lambda_t) \text{d}t + \eta \text{d}P_t + \sqrt{\lambda_t} \sigma \text{d}W_t, \] where \((W_t)_{t \geq 0}\) is a standard Brownian motion, \(\kappa,\theta,\eta > 0\) and \(\sigma \geq 0\) are constants, and concerning the initial value it is assumed that \(\lambda_0 \in [\theta,+\infty[\). The authors derive concrete expression for the expectations \[ \mathbb{E}[\lambda_t], \quad t \geq 0, \] and for the variances \[ \mathbb{V}\text{ar}(\lambda_t), \quad t \geq 0. \] Furthermore, they derive expressions for the Laplace transform. Namely, introducing for any \(r \in \mathbb{R}\) the process \((P_t^r)_{t \geq 0}\) as \[ P_t^r := \int_0^t e^{-rs} \text{d}P_s, \] and introducing the process \((\Lambda_t)_{t \geq 0}\) as \[ \Lambda_t := \int_0^t \lambda_u \text{d}u, \] they show that \[ \mathbb{E}[\exp \{ -v_1(N_T - N_t) - v_2(P_T^r - P_t^r) - v_2(\Lambda_T - \Lambda_t) \} | \mathcal{F}_t] = \exp \{ A(v,t,T) + \lambda_t B(v,t,T) \} \] for all \(0 \leq t \leq T\) and \(v \in \mathbb{R}^3\), where \(A\) and \(B\) satisfy a system of ordinary differential equations. The authors also deal with time-consistent evaluations. More precisely, they derive the time-consistent evaluation function derived from the variance principle, from the standard-deviation principle, from the mean value principle, and from the exponential principle. Numerical applications are presented as well.