| The use of diffusion models has been widespread in finance and economics for the last thirty years. The parameter estimation for such models hence becomes a crucial issue. It is not easy to develop an estimation method that is both accurate and computationally efficient because of the mathematical nature of the diffusion model. Numerical stability problems keep maximum likelihood estimation from becoming an efficient choice for multivariate diffusion models with lots of parameters.; By contrast, we develop new estimators---Quadratic Variation Estimator (QVE) for diffusion models. Numerical experiments and consistency proofs reveal the accuracy of the estimators for many diffusion models of interest in finance. Moreover, it is computationally efficient. The same method can be applied to a wider class of univariate diffusion models and many multivariate diffusion models, in particular, multi-asset model and stochastic volatility models. Asymptotic normality for one important estimator is considered. |
