On efficient simulation in dynamic models

Ways of improving the efficiency of Monte-Carlo (MC) techniques are studied for dynamic models. Such models cause the conventional Antithetic Variate (AV) technique to fail, and will be proved to reduce the benefit from using Control Variates with nearly nonstationary series. This paper suggests modifications of the two conventional variance reduction techniques to enhance their efficiency. New classes of AVs are also proposed. Methods of reordering innovations are found to do less well than others which rely on changing some signs in the spirit of the traditional AV. Numerical and analytical calculations are given to investigate the features of the proposed techniques.