Three Essays on Short-Term Interest Rate Futures.

Abstract

This dissertation examines the relationship between U.S. and Eurodollar interest rates by using daily U.S. Treasury bill and Eurodollar futures. Weak evidence of cointegration is found. The VAR and error correction models do not give better forecast performance than the naive model. Other evidence, particularly the simultaneous equations model, suggests that the hypothesis of contemporaneous relationships is not rejected. Further analysis of the deviations from the cointegrating relationship shows that the Treasury bill and Eurodollar futures are fractionally cointegrated after the 1987 stock market crash. Some preliminary statistics seem to support the hypothesis that these futures interest rates share the same volatility process, which follow a GARCH process. However, this hypothesis is rejected by the common volatility test. A bivariate EGARCH model which allows for asymmetric volatility influence of the TED spread, as well as that of the domestic market, is used to analyze the volatility spillovers between markets. Results show that the lagged TED spread change is the driving force of the volatility process. This dissertation also studies the international transmission of identical Eurodollar futures contracts traded on three exchanges, the IMM, SIMEX, and LIFFE. An approach of variance decomposition and impulse response functions exploring the common factor in the cointegration system is employed. It is shown that the markets are extremely efficient on a daily basis such that each market, when it is trading, impounds all the information that will affect other markets, and rides on the common stochastic trend. The significant results of volatility spillovers among markets suggest that certain market dynamics lead to a continuation of volatility. Particularly, the volatility spillover mechanism may be influenced by the U.S. stock markets. In addition, using a Monte Carlo approach, this dissertation investigates whether the cointegration and fractional cointegration results reported are biased by the GARCH innovations. The size of fractional cointegration tests is evinced to be less distorted by the GARCH effects.