This paper employs the autoregressive conditional jump intensity model, incorporating a forecast error, to investigate the relationships between the changes in the implied volatility and the relevant determinants in the Taiwan market. We further apply the orthogonality test to explore forecast error and content of information. The empirical results show that the changes in the implied volatility are affected by the contemporaneous returns, lagged returns, lagged changes in the implied volatility, contemporaneous daily changes in the realized volatility and lagged forecast error. At the same time, these results are consistent with the behavioral explanations of Hibbert et al. [Journal of Banking and Finance (2008) vol. 32, pp. 2254–2266]. We also demonstrate that the implied volatility in Taiwan seems to be an efficient forecast of future realized volatility during the whole of the sample period, except for the financial crisis period.