A Comparative Analysis: A Piecewise Approach To Modeling the Co-movement of Treasury Bond Yields for Different Country Groups

Abstract

This thesis aims to find an appropriate and functional model for co-movement of interest rates and to subsequently investigate how the characteristics of the model vary as we vary the types of countries considered by the model. By varying the relationship between the country being modeled and those in the set, we compare the quality and accuracy of the forecasts generated by the model. We aim to adapt and apply that a model that will forecast how yields in different countries will move in relation to one another, accounting for the time differences between trading hours. By doing this for several groups of countries, where the trade relationships between the countries vary, this thesis aims to explore how the forecasts from model for comovement vary. We compare three groups of countries: developed countries in a Free Trade Agreement (FTA), developing countries in a FTA, and a set of countries that are not linked by a common FTA. We employ a piecewise optimal ARMA-GARCH model that is segmented by time zones to accommodate for the different hours during which international markets are active. The results of the forecasting using this model indicate that it produces the most accurate forecasts for developed countries in a FTA, and the least accurate predictions for countries not in a FTA. With the help of Caramazza et al.’s work, we see that one possible reason for the disparity in prediction error stems from the level of economic integration (caused by the volume of trade) among the countries in the group. According to the aforementioned authors, trade financially links economies thereby making it more likely for one country to feel the effects of a change in another’s economic state. This property explains why the most accurate model for co-movement consist of developed countries in a FTA where trade volumes are the highest. This explanation in fact falls hand-in-hand with the conclusions made by Calderon in his paper.