Introduction

When it comes to seasonal adjustment, the most widely used statistical method is without a doubt that implemented in the Census X-l 1 software. Developed at the US Bureau of the Census in the 1950’s and 1960’s, this computer program has undergone numerous modifications and improvements, leading especially to the X-11-AR1MA software packages in 1975 and 1988 (Dagum [19, 20]) and X-12-ARIMA (the first beta version of which is dated 1998, Findley et al. [23]). While these software packages integrate, to varying degrees, parametric methods, and especially the AR IMA models popularized by Box and Jenkins [9], they remain in essence very close to the initial X-11 method, and it is this “core” that will interest us here.