This paper derives a method for constructing the likelihood function of a general class of linearized dynamic general equilibrium models that does not require the application of the Kalman filter. The method easily handles models in which variables are observed with error.