We present a new methodology for estimating time-varying conditional skewne
ss. Our model allows for changing means and variances, uses a maximum Likel
ihood framework with instruments, and assumes a non-central t distribution.
We apply this method to daily, weekly, and monthly stock returns, and find
that conditional skewness is important. In particular, we show that the ev
idence of asymmetric variance is consistent with conditional skewness. Incl
usion of conditional skewness also impacts the persistence in conditional v
ariance.