A Markov-Switching Equilibrium Correction Model for Intraday Futures and Stock Index Returns

A considerable literature in market microstructure has analyzed the information flows between stock index futures and spot markets. Most of those studies estimate deviations from the cost-of-carry model within the framework of vector equilibrium correction models (VECM). The typical finding is that futures prices lead spot prices and are the primary source of price discovery. Purely linear models can, however, lead to fallacious or at least incomplete inference in the presence of significant nonlinearities in the return generating process. Recent research has reported evidence for nonlinearity in the distribution of stock market returns. According to this literature, their empirical distribution can be characterized by a mixture of normal distributions whose dependence is well described by a hidden Markov chain. This thesis contributes to the former field by allowing for Markovian regime switches in the cointegrated system. The empirical analysis is carried out using high-frequency data for the German and Swiss markets, i.e. two closely interrelated markets which differ substantially in terms of liquidity. This thesis consists of three major parts. In the first part, an MS-VECM is estimated for each market and tested against the linear VECM. In both cases, the linear model is strongly rejected. The Markovian chain consists of three regimes, which can be well described in terms of volatility. Price discovery differs from regime to regime, but the overall evidence is consistent with the well-documented leading role of futures markets. The MS-VECM provides additional insights into the dynamics of price discovery. Interestingly, shocks are absorbed more rapidly in regimes of high volatility. A possible explanation is provided, based on trading activity. Intraday volatility is shown to be associated with the volume of trading. Heavy trading reveals more information per unit of time and thus improves index arbitrage and informational.