Finding the likely response of a likely prediction

Capital spectator list the methods used to generate predictions of the ADP employment report due.in a few days.  But the ADP data is already impulse response corrected for seasons, so you want to go back to the raw data, then perform predictions incorporating the models own detection of seasonal patters.  Maybe these researchers do that, but they should put up boldface assumption, 'Using Raw Numbers' If they are not going to raw data, then they have a big problem with interference of the two filters

Methods:

ARIMA: An autoregressive integrated moving average model that analyzes the historical record of ADP’s estimate of private payrolls in R via the “forecast”package.
ES: An exponential smoothing model that analyzes the historical record of ADP’s estimate of private payrolls in R via the “forecast” package.