Predictive power of the term structure

This PhD thesis analyzes the predictive power entailed in the term structure of interest rates and interest rate differentials. It consists of three key chapters based on three research papers. The first research paper titled ’Forecasting the Term Structure of Interest Rates near the Zero Bound - a New Era?’ investigates the forecasting performance of popular dynamic factor models of the yield curve after the global financial crisis (GFC). This time period is characterized by a low and nonvolatile interest rate environment in most major economies with short rates close to the zero bound. We focus on two popular factor models which exploit the information contained in the cross-section of the term structure – the dynamic Nelson-Siegel model and regressions on principal components – to show that subsequent to the GFC both models are significantly outperformed by a random walk no-change forecast. Especially for short and medium term yields, the random walk is up to ten times more accurate. Interestingly, these results are not picked up by traditional global forecast evaluation metrics. We further show that combining forecasts mitigates model uncertainty and improves the disappointing forecasting accuracy especially after the GFC. The research work ’Factors of the Term Structure of Sovereign Yield Spreads’ investigates the term structure of sovereign yield spreads for five advanced economies against the US and provides novel insights into the key drivers of the spread term structure. We show that the spread term structure dynamics are driven by three latent factors which can be labeled as spread level, slope and curvature, similar to common interpretations found in the yield curve literature. We further show that these estimated spread factors have predictive power for exchange rate movements and excess returns above the predictability of an uncovered interest rate parity approach. As the yield curve contains information about expected future economic conditions, we conjecture that these yield spread factors reflect expected macroeconomic differentials which in turn drive exchange rates. Using the information content of yield spread curves may thus be a promising approach to improve the forecasting accuracy of exchange rate models. The third research paper titled ’Exchange Rates and Unobservable Fundamentals : A New Approach to Out-of Sample Forecasting’ builds on the key findings of the previous paper and suggests applying the empirical sovereign yield spread level and slope to forecast exchange rates out-of-sample. Traditional exchange rate models are usually based on differences in observable macroeconomic fundamentals such as output and inflation. However, while being well grounded in economic theory, these fundamental models have a rather poor out-of sample forecasting record. This empirical failure may be a result of the overly restrictive choice of macroeconomic fundamentals. We thus apply the empirical sovereign yield spread level and slope as unobservable proxies of the market’s expectations for current and future fundamentals. Our approach outperforms traditional exchange rate models in forecasting accuracy and profitability for all applied forecasting evaluation metrics. It is also superior to a random walk in terms of direction of change forecasts and profitability of the forecasts.