Golara Zafari

Title: Three Essays in Financial Econometrics
Date: February 5th, 2026
Time: 10:00am
Location: LIB 2020/ Zoom
Supervised by: Jean-François Bégin

Abstract:

In the context of market risk, modelling volatility and parameter estimation for volatility models are essential for capturing asset dynamics. In contrast, credit risk analysis requires robust scenario generation to evaluate potential default events and portfolio losses. Motivated by their significance, this thesis offers three contributions. The first contribution is the development of a deterministic filtering and estimation methodology for a broad class of affine multifactor stochastic volatility jump-diffusion models. This filtering approach leverages a marginalized discrete nonlinear filter. By conditioning on the jump components, the procedure employs Kalman-like recursions to manage the states that appear linear and Gaussian in the model’s dynamics, while jumps are addressed by constructing a grid of likely values. The efficacy and practical applications of this methodology are demonstrated through extensive simulation and empirical studies. The second contribution is proposing a new class of affine discrete-time multifactor stochastic volatility model with jumps. Unlike GARCH-type models with predictable conditional variance, the proposed model enables volatility to be genuinely stochastic and evolve independently of the returns. The physical and risk-neutral model dynamics are linked through a pricing kernel that disentangles various risk factors. The risk premiums associated with different moments of the return distribution are investigated. The last contribution is proposing a scenario-generating model for default intensities and recovery rates based on key credit risk determinants. In this framework, the time-series dynamics of credit risk determinant factors are presented and are linked to firm-specific default intensities and recovery rates—derived from market prices of American out-of-the-money put options and credit default swap premiums—through random forest. Analyzing data, the dynamics of determinant factors and their influence on default intensity and recovery rate are explored. Moreover, the applications of the framework in credit risk management applications, such as market-consistent expected and unexpected loss calculation, are evaluated.

Keywords: credit default swap; credit risk; cumulant; default intensity; GARCH; jumpdiffusion stochastic volatility; option; parameter estimation; pricing kernel; recovery rate