|Page tools: Print Page Print All RSS Search this Product|
Reviewing the Impact of Major Data Revisions to the Seasonal Adjustment of the Building and Construction Collection
Earlier this year, the Building Approvals collection implemented extensive revisions to their Original estimates from July 2001 onwards due to a combination of:
· the introduction of the new Australian Statistical Geography Standard1
· the introduction of the revised ABS Functional Classification of Buildings2
· a major audit of data provision, which identified a number of reporting issues and systematic reporting anomalies in some constituencies.
The implementation of almost 12 years of revisions called for the seasonal adjustment settings for the 129 affected series to be reassessed by the Time Series Analysis (TSA) section. This process was further complicated by the work having to be completed in an extremely short time frame between the revised data being loaded and the next publication being created, with TSA analysts having only 4 days to complete the initial work, and then another 4 days for further data checking and validation.
When reviewing the seasonal adjustment settings for a series, an analyst identifies any ‘unusual’ behaviour which is adversely affecting the adjustment of the series. The most common unusual behaviour identified in Building Approvals series is the occurrence of extreme values, where the series value for a particular period is significantly larger than the underlying level of the series. This is generally caused by one (or more) very large ‘jobs’ being approved in that period, where a job is a single building approval application, and very large jobs tend to be large apartment buildings (residential) or major infrastructure such as hospitals (non-residential). Where causes (i.e. specific jobs) have been identified for extreme values, we ‘correct’ the value by removing the impact of the identified job from the Original data, performing the seasonal adjustment process, and then replacing the identified job back into the Seasonally Adjusted series. This way, the seasonal adjustment process is no longer adversely affected by the extreme value, but the impact of the very large job can still be seen in the published Seasonally Adjusted series.
In the process of reviewing the seasonal adjustment settings for the affected series, TSA analysts took care to review the data from July 2001 onwards to identify any extreme values. Any identified extremes were then investigated to determine if they were:
· a previously identified extreme with an appropriate correction
· a previously identified extreme with an inaccurate correction (i.e. where the value of the job has been revised)
· a previously unidentified extreme.
Analysts also considered if any existing extreme value corrections were no longer deemed necessary, and could therefore be removed.
The process was a collaborative exercise, with TSA staff working closely with the Building & Construction Section to ensure the highest quality Seasonally Adjusted and Trend estimates were produced.
The Building Activity collection was also revised over the same period, as its frame is drawn from the Building Approvals collection. Again, TSA and the Building & Construction Section successfully worked collaboratively to review the seasonal adjustment for the collection.
This project work was both highly complex and very well coordinated, with all TSA staff required to be engaged. The project outcomes were positive, both in technical terms and in terms of the experience gained by key TSA staff.
The revised Building Approvals estimates were released in ABS cat. no. 8731.0 on 4th March 2013 and the revised Building Activity estimates in ABS cat. no. 8755.0 on 29th May 2013 and ABS cat. no. 8752.0 on 17th July 2013.
1Australian Statistical Geography Standard (ASGS), 2011 Edition (cat. no. 1270.0.55.001)
2ABS Functional Classification of Buildings 1999 (Revision 2011) (cat. no. 1268.0.55.001)
These documents will be presented in a new window.