evidence from Nishina, Maghrebi and Kim (2006) of higher out-of sample forecasting
power for the S&P 500 and Nikkei 225 implied volatility indices.1
The informational content of implied volatility has been traditionally examined
using conventional regression analysis and GARCH modelling. The present study examines
the nonlinear dynamics of the relationship between implied volatility and realized volatility
using Markov regime-switching models introduced by Hamilton (1989). The economic
motivation for modelling the dynamics of implied volatility with regime-switching
processes lies in the changing likelihood for market volatility to fluctuate between regimes
of higher versus lower volatility and/or slower against faster mean reversion. Much like the
return-generating process, the dynamics of anticipated volatility can be also governed by
different regimes characteristic of periods of bearish or bullish markets and associated with
the troughs and peaks of economic cycles.
In contrast, the literature on implied volatility provides little empirical evidence
based on regime-switching dynamics. This study constitutes, to the best knowledge of the
authors, the first attempt to test for the existence of Markov regime switches in anticipations
of stock market volatility.2 The regime switching approach is rather widely applied to
account for structural breaks in a range of financial variables including interest rates, equity
1 There is evidence from Jorion (1995) for instance that implied volatility provides efficient
estimates of future volatility in currency markets as well.
2 Guo and Wohar (2006) identify regime changes in stock market volatility implied by S&P
100 and S&P 500 options prices using Bai and Perron (1998) method to determine
structural breaks in the mean level of market volatility. Their approach differs however
from Markov regime-switching modelling proposed by Hamilton (1989), which is applied
in the present paper. Furthermore, this model allows for regime-dependent speed of
adjustment and tests for leverage effects and the impact of past realized volatility.