Seasonal Adjustment of Weekly Trade Data

Abstract: The main objective of this paper is to equip the trade policy analyst with an appropriate method of seasonally adjusting trade data with weekly observations. To that end, a structural time series model containing a trend, seasonal and irregular component is specified. The seasonal component is represented by a time-varying periodic spline. Casting the model in state-space form enables time-varying parameters as well as use of the powerful Kalman filter for trend estimation. The resulting trend can then be interpreted as a seasonally adjusted series. A simulation exercise shows that the correct trend is identified with an average absolute error of 0.4 percent. An application to Swedish imports during 2017-2021 shows that the model produces a reasonable trend estimate when applied in 'real-time' and that its application is preferred to smoothing the series using a simple moving average.