Logout / Access Other products Drop Down Arrow
Get live help Monday-Friday from 7AM-7PM ET (12PM-12PM GMT)  •  Contact Us
Check out our new FAQ section!
RSS Feed
TitleUsing Data Buffet: Do forecasts contain history?
AuthorPhillip Thorne
Question

Do Moody's Analytics forecasted time series contain historical values?

Answer

Yes, with caveats.

Every Data Buffet forecasted time series has two parts: a historical segment (populated from historical data) and a forecasted segment (populated from a model).  The historical segment is based on the driver, but is not necessarily identical to it, for reasons described below.  Our forecasted time series are optimized for purposes of forecasting, and the historical segment provides context, but is not a substitute for the driver; rather, if you need the exact reported values, as quickly as possible, then use the historical driver directly.

(These explanations assume a quarterly forecast with a single monthly historical driver series.)

Frequency.  If the driver is non-quarterly, we frequency-transform it to populate the historical segment.  This transformation is irreversible; i.e., any original volatility cannot be recreated.  (You can't "unscramble the egg.")  To properly compare the historical and forecast series, you must replicate our process, i.e., convert the historical to quarterly.

Period-filling.  We extend the historical segment when a full quarter of historical data are available.  If the driver is monthly, we wait until we have all three months that constitute a calendar quarter.  Hence, the historical segment may at times seem to be "missing data."

Timing.  Our forecasting process is based on models, and all series in a model are updated in unison.  To be integrated, the driver (a) must be published before the model and run, and (b) will not be integrated until the model is run.  For example, if the driver is published on January 1, it is not integrated until the model is run on January 15; conversely, if the driver if published on January 16, it has "missed the boat."

Finite processing time.  If the driver is published on Thursday, and the "last updated" value on a forecast series is Friday, shouldn't the new value be integrated?  Not if the model has been in-process since Wednesday.

The date of the last period in the historical segment is the last historical point, displayed in Mnemonic 411 and as an optional basket field.  This metadatum is emitted by many, but not all, of our forecast models.  It is used by the "highlight forecast segment" feature of the Data Buffet Chart module, and to mark forecasted periods (in red color) in the View module.