Affine Term Structure Models Approaching The Zero-Lower Bound

This thesis reviews the affine term structure model class as proposed by Duffie and Kan (1996). Since prior research is scarce on evaluating data including the zero-lower bound, I contribute by investigating a recent U.S. Treasury data set. I estimate one-, two-, and three-factor Vasicek (1977) and Cox et al. (1985) models using the Kalman filter approach of De Jong (2000) in an empirical study while including the zero-lower bound. I also perform a simulation study for the three-factor models under a zero-lower bound environment by lowering the short rate and the volatility. I provide evidence that the three-factor Vasicek (1977) model obtains the best fit and captures the stylized facts whereas the Cox et al. (1985) only captures the yield curve's level and slope. This evidence is less apparent on data including the zero-lower bound. In the simulation study, the Cox et al. (1985) model is more accurate than the Vasicek (1977) model on CIR data whereas the performance is close on Vasicek data.