Inference in dynamic models containing 'surprise' variables (Q1822190)

Some new results on calculating moving average representation (MAR) coefficients and their limiting distribution from estimated vector ARMA processes are presented. The technique is applied to the problem of estimating the coefficients of unanticipated or 'surprise' variables in a single equation for a multi-period expectations horizon. The method naturally conditions the expectations on all past values of the process and avoids the necessity of using two-step regression procedures and adjusting the resulting standard errors.