| Summary: | 碩士 === 國立臺灣大學 === 經濟學研究所 === 92 === We extend the nonstationary nonlinear heteroskedasticity (NNH) model proposed
by Park (2002) to construct a new volatility model, namely,
the asymmetric NNH (ANNH) model,
in which a particular nonlinear transformation of a standardized random walk is
used to depict the evolution of volatility.
Compared with the NNH model,
the ANNH model is able to capture asymmetric reactions to good and bad shocks
and exhibits long memory in volatility while maintaining finite unconditional
variance.
In addition,
this model also improves on efficiency in parameter estimation.
As the NNH model, the ANNH model also exhibits properties of volatility clustering
and leptokurtosis,
which are usually observed in economical and financial time series data.
Simulation results show that if we use time series generated by the ANNH
model to fit a GARCH(1,1) model,
the sum of parameters alpha_1 and beta_1 will be close to 1,
likes the IGARCH(1,1) model,
which does not have finite second and fourth moment.
|
|---|
