Predictability of Bond Risk Premia with an Affine Term Structure Model

PREDICTABILITY OF BOND RISK PREMIA WITH AN AFFINE TERM STRUCTURE MODEL by SIBEL KORKMAZ Adviser: Professor Liuren Wu This thesis focuses on zero-coupon bond risk premia. In chapter 1 first I summarize the literature that defines expectation hypothesis, time variation in bond risk premia, well known macroeconomic and variety of technical indicators which predict bond return forecasts. Then, I continue with introducing a high dimensional affine dynamic term structure model and unscented Kalman filtering as its estimation technique.In chapter 2, I explore predictive capacity of the estimated latent states of this term structure model. I constructed a single return forecasting factor by using these states. I report comparison of well known excess bond return forecasting factors’ predictive performance with the one I construct. I found that this single factor has superior performance in predicting excess bond returns and future macro activity especially for the sample period which includes the Great Recession. In chapter 3, I develop a two step estimation technique to extract market price of risk from time dimension of zero-coupon bond prices. I also provide robustness results in this chapter.