In this paper we consider joint estimation of objective and risk-neutral parameters for the autoregressive gamma volatility option pricing model. The model is fitted on sample of 10 years of daily observations on S&P 500 returns, realized variances and options, collected in a large unbalanced panel. Maximum likelihood estimation is implemented using a sequential version of the Efficient Importance sampling algorithm to handle latent factors. Our method provide a computationally fast and reliable solution to estimate the historical and the risk neutral distributions implied by an option pricing model with multiple volatility components.