Modeling the Effect of NASDAQ and Index Futures Introduction on the
Volatility of NSE Nifty Using GARCH Methodology
Student Name: K. Kiran Kumar
The first problem of interest was to
study causal issues of mean returns and return volatilities between NSE, India
and NASDAQ, USA and then empirically model this dynamics using univariate
ARMA-GARCH, two-stage ARMA-GARCH and multivariate ARMA-GARCH techniques. The
study found both linear and non-linear Granger causality running
unidirectionally from USA to India with respect to both mean returns and their
volatilities. Though two-stage ARMA-GARCH models, which are special cases of
multivariate ARMA-GARCH models, are advocated in the literature for such equity
market inter-linkages study with markets having non-overlapping trading hours,
as in the case of NSE and NASDAQ, this study found that a much simpler
univariate ARMA-GARCH model is equally capable of capturing the same dynamics
with fewer number of clearly interpretable parameters. The in-sample validation
and out-of-sample forecast performance did not significantly improve after
applying the multivariate ARMA-GARCH technique either, which yielded
qualitatively similar results. The major findings of this study was that the
mean overnight returns of NSE Nifty are getting affected by both the daytime
returns of NSE Nifty and NASDAQ composite of the previous day, while its
volatility seems to be affected only by the NASDAQ. For the next problem, the
daily returns of NSE Nifty were first subjected to an informal graphical and
formal Bayesian and likelihood based change-point analysis to confirm that
indeed there is a shift in its volatility post Futures introduction. Then this
shift was modeled using a switching ARMA-GARCH model. The study found that
recent news is being assimilated in the volatility at a much faster rate in
the post Futures introduction period, with a substantial reduction in its
persistence.