In this paper we consider how well seasonal adjustment methods (X-11 and ARIMA model-based), and certain of their variations, satisfy one objective of seasonal adjustment: facilitating short-term forecasting of nonseasonal movements in time series. We do this via an empirical study using a number of seasonal time series of major U.S. economic aggregates. For these series we examine how forecast accuracy is affected by the following choices: (1) alternative choices of simple models for forecasting the seasonally adjusted series (or trend estimates); (2) use of seasonally adjusted series versus trend estimates; (3) use of time series of unrevised versus time series of revised seasonally adjusted data; (4) use of X-11 versus ARIMA model-based adjustment; and (5) use of seasonally adjusted data in forecasting versus directly forecasting the unadjusted series.