How are seasonal adjustment, calendar day adjustment, and trend adjustment documented in Data Buffet series metadata, calendar, and news articles?
In series "description" metadata
Each Data Buffet time series is accompanied by two textual metadata attributes: description and source. The last segment of the description metadatum is a parenthesized phrase (the unit-descriptor) that represents the series measurement, usually in symbols that we have standardized across Data Buffet. The seasonal adjustment is listed after the comma.
- Single adjustment:
- NSA = not seasonally adjusted
- SA = seasonally adjusted
- CDA = calendar day adjusted
- WDA = working day adjusted
- Trend = trend component
- Compound adjustment:
- CDASA = calendar day adjusted and seasonally adjusted
- WDASA = working day adjusted and seasonally adjusted
Seasonal adjustment is commonly paired with annualization, a combination that uses the symbol "SAAR" (read as "seasonally adjusted at an annualized rate") and similar.
Because series with low frequencies (semiannual, annual and censal) cannot, by definition, exhibit seasonal effects, this clause is omitted from their unit-descriptor.
In series "source" metadata
If a source does not publish an SA variant, we may choose to construct our own (this is one kind of historical supplement we provide). To denote our role, we append a secondary source citation that reads "Moody's Analytics Adjusted."
In the catalog
A source may publish the same measurement in several analytic variants. In the catalog we order them with the "most useful" first, that being:
- Trend
- Seasonally adjusted
- Not seasonally adjusted
If we produce an SA supplement, the catalog header will read "Seasonally adjusted by Moody's Analytics."
Related series have mnemonics that are systematically similar (e.g., the presence/absence of the "X·" prefix in U.S. concept codes) but not 100% predictable. Consequently, to select variants we recommend that you browse the catalog.
Some historical datasets report both NSA and SA variants, but are incompletely parallel; that is, some indicators are unpaired. When the raw series does not exhibit seasonality, the source may deem an SA variant to be superfluous. The U.S. BLS detailed producer price index (PPI) is known for this.
In Data Buffet News
When we add a new dataset, we may construct the SA supplement at the same time, and will mention it in the "New Data" article. For example:
If we construct it later, we will post a news article with a title of the form "New Data ... SA" with an explanation of our motivation, response, and specific procedure. For some datasets, adjustment is one of several improvements we apply to a single series or a family of related series.
We prefer to republish source data verbatim. If the source subsequently starts publishing its own SA data, that will render our supplement superfluous, so we will post a "Data Change ... Source" or "Data Note ... Source" article.
See also
Updates to this article
- 22 Dec 2015 - Initial version.
- 12 Feb 2016
- 6 Feb 2024 - Added SAAR, updated cross-reference links.