Affine Term Structure Models with Martingale Estimating Functions

Affine term structure models (ATSM) are used to model bond prices and yields. In this thesis, I propose to estimate the term structure parameters by martingale estimating functions (MEF). I analyze the feasibility of this method in a simulation study to better understand the behavior of the objective function with respect to parameters, as well as the actual convergence to the true population parameters using different search algorithms. MEF works very well in the simulated environment but requires a rescaling of the optimal MEF weights for the most robust results. The method is also applied to an empirical dataset where MEF provide good parameter estimates for the one- and three-factor settings and can directly compete with the minimum chi-squared estimation (MCSE) and the Kalman filter.