Seasonal Adjustment

Seasonal adjustment is the statistical removal of patterns that repeat at the same point every year, so that consecutive periods of a series can be compared. Holiday retail activity, hiring in agriculture and construction, and the effect of the school year on employment all move a series for reasons that carry no information about the underlying trend; the adjustment estimates those recurring effects and takes them out. The adjustment is an estimate, and consequences follow from that. Seasonal factors are re-estimated as new observations arrive, so an adjusted figure for a past month can change even when the unadjusted number does not. Unusual events — a shifted holiday, a strike, a disrupted year the model still carries — distort the factors for some time afterwards. Agencies publish both the adjusted and the unadjusted series and describe the method used, which is the only way to tell an adjustment artefact from a change in activity.