A structural time series model is estimated for U.S. West Coast oil imports on the basis of seasonally unadjusted and seasonally adjusted data. The estimated model is used to generate sets of out-of-sample forecasts, with those generated from seasonally adjusted data re-seasonalised by adding the seasonal factors derived from the X-11 procedure. The quantitative measures of forecasting accuracy of the models do not show any improvement by switching from unadjusted to adjusted data. It is concluded that seas onal adjustment is not as detrimental to forecasting accuracy as it is often claimed.