anticipated decreases in volatility, with slower mean reversion, asymmetric relationship
with returns and positive adjustment to changes in realized volatility.
The onset of financial crises does not have the effect of shifting the implied
volatility process from one regime to another for long protracted periods. For the Japanese
market however, major events such as the burst of the asset bubble and the Asian financial
crises seem to trigger the regime characterized by expectations of significant increments in
volatility, stronger mean reversion and negative correlation with changes in realized
volatility. The immediate reversal to the pre-crisis regime, featuring expectations of
decreasing volatility and strong mean reversion, suggests that the impact of financial crises
on the dynamics of volatility expectations is short-lived. With respect to the US market, it
increases the likelihood of regimes characterized by expectations of decreasing volatility,
significant mean reversion, stronger asymmetric impact of news and more significant
adjustment process.
Evidence of regime shifts in volatility expectations has some implications for
risk-hedging, policymaking and future research in financial economics. It offers new
avenues for research on such important issues as to whether market consensus expectations
are consistent with rational expectations and whether financial crises may be induced by
regimes consistent with self-fulfilling expectations and speculative bubbles. Empirical
studies may also shed light on the relationship between the leverage effects and the speed of
mean reversion or length of the memory process across regimes. Future research in
behavioral finance can benefit from tests of the stochastic properties of investors’ attitudes
towards risk, regime-dependencies in investor confidence and irrational exuberance. From
the risk-hedging perspective, regime shifts imply nonlinear serial dependence in volatility
expectations. It is thus important to examine to what extent the quantification of risk
exposure is affected, the estimation of dynamic hedge ratios is further complicated, and the
17