ARE VOLATILITY EXPECTATIONS CHARACTERIZED BY REGIME SHIFTS? EVIDENCE FROM IMPLIED VOLATILITY INDICES

function compared to the previous Model 1 for both markets. For the US market, two
regimes of implied volatility are identified, one with longer memory and negative
correlation with stock market returns (regime 1) and another with shorter memory and
positive correlation with returns (regime 2). Again, there is no evidence of a negative
relationship between implied volatility and stock market returns for the Japanese market.
The regime of low expected volatility is characterized by shorter memory and positive
correlation with returns.

Testing for the asymmetric impact of news on market volatility should take into
account both the sign and magnitude of returns. Thus, Model 3 describes the level of
implied volatility as a function of past return levels as well as squared innovations, while
excluding autoregressive terms. There is evidence of significant leverage effects since both
the sign and magnitude of returns are likely to affect volatility expectations. However, the
negative relationship with returns is not found to be regime-dependent in both markets since
the null of equal β parameters across regimes cannot be rejected. Given the
positiveγcoefficients, shocks to the return-generating process, are irrespective of their sign,
conducive to expectations of higher volatility. Thus, the evidence of leverage effects is
found to be regime-dependent in both markets.

The estimation of Model 4, which is comparable to Model 3 but inclusive of
autoregressive terms, reveals the existence of two regimes featuring equal intercepts and
positive serial correlation for the US market. The two regimes differ only with respect to the
relationship of volatility expectations with market returns. The presence of leverage effects
is manifested by the significance of bothβandγparameters, with the level of expected
volatility in regime 1 being less responsive to variations in market returns. In contrast, the
estimation results for the Japanese market suggest that drift terms and autoregressive
parameters are regime-dependent. Regime 1 is identified with expectations of low volatility,

12

<<   12   13   14   15   16   17   18   >>

More intriguing information