Modeling and Forecasting Stock Return Volatility and the Term Structure of Interest Rates

This dissertation consists of a collection of studies on two topics: stock return volatility and the term structure of interest rates.

Part A consists of three studies and contributes to the literature that focuses on the modeling and forecasting of financial market volatility. In this part we first of all discuss how to apply CUSUM tests to identify structural changes in the level of volatility. The main focus of part A is, however, on the use of high-frequency intraday return data to measure the volatility of individual asset eturns as well as the correlations between asset returns. A nonlinear long-memory model for realized volatility is developed which is shown to accurately forecast future volatility. Furthermore, we show that daily covariance matrix estimates based on intraday return data are of economic significance to an investor. We investigate what the optimal intraday sampling frequency is for constructing estimates of the daily covariance matrix and we find that the optimal frequency is substantially lower than the commonly used 5-minute frequency.

Part B consists of two studies and investigates the modeling and forecasting of the term structure of interest rates.
In the first study we examine the class of Nelson-Siegel models for their in-sample fit and out-of-sample forecasting performance. We show that a four-factor model has a good performance in both areas.
In the second study we analyze the forecasting performance of a panel of term structure models. We show that the performance varies substantially across models and subperiods. To mitigate model uncertainty we therefore analyze forecast combination techniques and we find that combined forecasts are consistently accurate over time.