We present a new methodology for estimating time-varying conditional skewness. Our model allows for changing means and variances, uses a maximum likelihood framework with instruments, and assumes a non-central <italic>t</italic> distribution. We apply this method to daily, weekly, and monthly stock returns, and...