The finite sample properties of the GARCH option pricing model

The authors explore the finite sample properties of the generalized autoregressive conditional heteroscedasticity (GARCH) option pricing model proposed by S. L. Heston and S. Nandi (2000). Simulation results show that the maximum likelihood estimators of the GARCH process may contain substantial estimation biases, even when samples as large as 3,000 observations are used. However, it was found that these biases cause significant mispricings only for short-term, out-of-the-money options. It is shown that, given an adequate estimation sample, this bias can be reduced considerably by employing the jackknife resampling method.