Empirical investigation of currency and term structure of interest rates with nonlinear asset pricing models and cointegration.

Abstract

The purpose of this study is to extend the empirical research on the term structure dynamics for international interest rates and exchange rates using non-affine asset pricing models. There is very limited empirical work on term structure characteristics of international economies especially using structured asset pricing models. Furthermore, these models generally fail to explain the forward premium puzzle, the tendency of high interest rate currencies to appreciate, with an exchange rate risk premium which is negatively correlated with and has a higher volatility than the expected depreciation of the currency. These properties are also referred to as Fama's conditions for the explanation of the puzzle. The dynamic model of this research does explain the puzzle with such a risk premium. Specifically, I investigate modified quadratic autoregressive independent variable nominal term structure models and the dynamics of their state variables in an international setting using a nonlinear non-Gaussian estimation method and the extended Kalman filter method. I use this non-affine term structure model to investigate, empirically, the term structures and bond premiums for the US, Germany, and the UK. I also investigate the exchange rate dynamics in a two-country setting and explain the forward premium puzzle using Fama's conditions. The multi-factor term structure model for each country consists of amplification coefficients reflecting the relative impact of the factors on the local interest rate and economy. Using these scaling coefficients and the influence of state variables in each economy, I demonstrate the limited diversification in bond portfolios and illustrate the international diversification puzzle, which is the tendency to invest in domestic assets, although an international portfolio has less risk and higher returns. The model is extended with correlated state variables, and the empirical performance of this modified model is examined using extended Kalman filter estimation methodology. This version of quadratic models proved to have better empirical properties and is suggested as the basis for future empirical studies of international fixed income securities markets. Finally in the last section I move away from asset pricing models and consider a more traditional empirical model. I examine the empirical relationship between short-term interest rates and very long term interest rates of government bonds with exchange rates. I examine cointegration issues in the time series data and run regressions accordingly. I also give a theoretical explanation of the forward premium puzzle consistent with the data.