Abstract
In this article we investigate the efficiency of least squares (l.s.) in estimating a stable pattern. A model is proposed for an economic time series, which, after filtering, is a suitable basis for our investigation. The covariance structure of the disturbances is assumed known and several possible generating processes for the disturbances are considered. We contrast the variances of the l.s. estimates with those from the best linear unbiased (b.l.u.) procedure for four sample sizes and several possible “trend” removing filters.