Heterogeneous Markov-Switching GARCH Models for U.S. Tourism Active Stock Trading

This paper tests the benefits of using heterogeneous Markov-Switching GARCH (MS-GARCH) models for active trading of tourism (leisure and entertainment) stocks by performing a weekly backtest of the 36 combinations of two-regime MS-GARCH models, given their regime-specific marginal probability (Gaussian and Student-t). Their regime-specific variance model (time-fixed, symmetric GARCH, or asymmetric EGARCH), and by assuming a two-regime context with a low (high)-volatility regime s = 1 (s = 2), the results suggest that using an MS-GARCH model with a Student-t pdf and a symmetric GARCH variance in s = 1, and a Gaussian pdf with a time-fixed variance in s = 2, leads to a better performance than a buy-and-hold strategy (with a compound annual growth rate, or CAGR, of 10.0716% and an annualized Sharpe ratio of 5.0891). This performance reflects the impact of a 0.1% trading fee per traded amount and a 10% tax. This result suggests that, only in the short term, MS-GARCH models are useful for active trading in tourism stocks by portfolio managers and could be used to forecast high-volatility episodes in such companies, which are prone to price declines during sanitary, geopolitical, or consumer-sentiment crises. Despite this in-sample result, it is important to highlight that the results do not hold in the long term, as tested for randomness in the backtest results (data snooping), and that further improvements must be made to the algorithm to generate a significant overperformance of the trading strategy.