Using Seasonally Adjusted and Unadjusted Data

[ac_1]

Hi Tom,

My aim is to forecast CPI index and infl (calculated from the CPI index), single equation, using BOXJENK and LINREG for:
(1) Seasonally Adjusted series, and
(2) Unadjusted Data
And generate forecasts for both (1) and (2):
(3) medium-term dynamic, and
(4) one-step ahead static

Here says, https://www.bls.gov/cpi/seasonal-adjust ... d-data.htm . An interpretation could be to use just the Seasonally Adjusted series for medium-term dynamic, and use just the Unadjusted Data for one-step ahead static. Is that reasonable?

Also,
(a) The numbers which make the press/news (I think), as the data are released, are for the Unadjusted Data. I suppose because "The unadjusted data are of primary interest to consumers concerned about the prices they actually pay."
(b) But, https://www.bls.gov/news.release/cpi.nr0.htm focuses mostly on Seasonally adjusted data, although is for both.


Additionally, I would like to use just the one series i.e. the unadjusted data (availability with type of data varies for the other Major's) to generate all four sets of forecasts (3) and (4); eight sets of forecasts if I double-up for core inflation.

Something like the following:
Unadjusted Data (CPI index and infl) -> Seasonally adjust (somehow) -> model & forecasts
Unadjusted Data (CPI index and infl) -> Seasonally adjust (somehow) -> model & forecasts -> Reverse Seasonally adjust forecasts to back-out Unadjusted forecasts.

How to proceed?

Thanks,
Amarjit

[TomDoan]

I don't know about the U.K. but basically every "headline" series released in the U.S. (whether by BLS, Census Bureau or Federal Reserve) is seasonally adjusted. In some cases you have to work very hard to find seasonally unadjusted data. (NSA GDP, for instance, isn't routinely published).