The Italian Continuous Time Model : Results of the Nonlinear Estimation

Continuous-time econometric models (i.e., models specified as systems of stochastic differential equations) are by now well established both in theory and in practice. Several macroeconometric continuous-time models exist for various countries. Among these the Italian continuous-time model developed by Gandolfo and Padoan over the years is in its fifth version. It is a medium-term disequilibrium model of real and financial accumulation of an open economy capable of generating both long-run growth and cycles. It has been successfully used for various purposes, amongst which the determination of the exchange rate (where it has been capable of outperforming the random walk in out-of-sample forecasting experiments). Up to now, the estimates of the parameters have been obtained using a linearized version of the model. Although for simulation purposes the original nonlinear version can be (and has actually been) used, it would be highly desirable to obtain the parameter estimates through a nonlinear estimation method. Such a method has been made available recently by Clifford R. Wymer (1993a,b) through his ESCONA program. The purpose of this paper is to apply such a method to our Italian model, and to compare the results with those of the linear approximation