Exact calibration of structural models via time-change

Battle Count: 60

The paper is primarily relevant to credit risk modeling and derivatives pricing rather than direct quantitative trading strategies. However, it has significant implications for: (1) CDS pricing and hedging, which are traded instruments; (2) CVA calculations affecting trading desk P&L; (3) structural model calibration for equity-credit correlation strategies; (4) the AT1P model is used in practice for CDS calibration. The exact calibration methodology could improve model risk management for trading desks dealing with credit derivatives. The relevance is moderate as it addresses a foundational modeling/calibration problem rather than a direct trading signal or execution strategy.

For the Brownian motion case, implementation is straightforward: the implied clock has a closed-form expression (Theorem 1, equation 7) involving only the inverse normal CDF. For the drifted Brownian motion and AT1P cases, numerical inversion of the latent survival curve is required, but this is described as computationally efficient. The main complexity lies in: (1) correctly implementing the time-change for Monte Carlo simulation (exact Gaussian-increment scheme); (2) handling the potential explosion of the clock rate near t=0; (3) managing multiple time-changed processes with different clocks in joint models; (4) ensuring the regularity conditions on G and the latent survival curve are met. The theoretical framework is elegant but practical implementation requires careful numerical treatment.

The paper provides complete mathematical derivations, closed-form expressions for the Brownian case (Theorem 1, Corollary 1), and explicit formulas for the AT1P connection. Numerical examples include Monte Carlo simulation parameters (N=1000 trajectories, time step dt=0.01, T=20). The methodology is fully specified with definitions, lemmas, and theorems. However, no code repository is provided. The piecewise constant hazard rate parameters are referenced from Table 2 of Brigo et al. [5], requiring access to that external source.