Saddlepoint approximations in the frequency domain

Saddlepoint techniques provide accurate, higher order, small sample approximations to the distribution of estimators and test statistics. However, except for a few simple models, these approximations are not available in the framework of stationary time series. We contribute to fill this gap by developing new saddlepoint approximations for frequency domain statistics. Under short or long range serial dependence, we show how to derive and implement our saddlepoint techniques for two relevant classes of statistics: ratio statistics and Whittle’s estimator. A Monte Carlo study for different models illustrates the theory and compares the new approximations with those obtained by first order asymptotic theory and the bootstrap. An example based on data about the European Central Bank assets motivates and concludes the paper.