Approximation of Arbitrary Distributed Lag Structures by a Modified Polynomial Lag: An Extension

Abstract

Since a uniform approximation of distributed lag models by representatives of some subfamily of the family of all possible lag structures guarantees neither a reasonable standard error of forecast in large samples nor a good approximation of the true long-run response and mean lag, it is shown that the modified polynomial lags fulfill a stronger approximation property. A possible simplification of the estimation procedure for these models is mentioned.