1346.0.55.003 - Interpreting Time Series: Are you being misled by the Seasons, 2012  
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 11/12/2012  First Issue
   Page tools: Print Print Page Print all pages in this productPrint All

REVISIONS TO TIME SERIES ESTIMATES

WHAT ARE REVISIONS?

Revisions are minor improvements to original, seasonally adjusted and trend estimates that are made as further information about a time series becomes available. Revisions are implemented so that we produce the most appropriate original, seasonally adjusted and trend estimates at any particular point in time.


KEY POINT:

Revisions are improvements to the original, seasonally adjusted and trend estimates.


WHEN WILL ESTIMATES OF THE ORIGINAL SERIES BE REVISED?

For some series, original estimates are routinely revised as better quality source data becomes available. For most series, revisions to original estimates are rare, and will be noted in ABS publications. Isolated points have been corrected where flawed data or processes have escaped normal quality checks. Sometimes there is a change in the way data is collected, or the scope or definitions used in a survey change; in this case, the original series may need revision so that it is consistent over time. Revisions to the original estimates will also lead to revisions in the seasonally adjusted and trend estimates.


KEY POINT:

Revisions to the original estimates will lead to revisions in the seasonally adjusted and trend estimates.


WHEN WILL SEASONALLY ADJUSTED ESTIMATES BE REVISED?

Seasonally adjusted estimates will generally face several years of small revisions as more original estimates of the same month or quarter become available. These revisions usually become even smaller over time, and can be expected for up to 7 years after an initial estimate is made.

The largest revisions to the seasonally adjusted estimates usually occur in each of the first three years after our initial estimate is made and after this time we can make a relatively stable estimate of the seasonal pattern. This occurs because changes to the seasonal pattern become most apparent when we see more original estimates of the same month or quarter.

Figure 4: Short term visitor arrivals to Australia from Japan, Seasonally Adjusted Estimates

Figure 4: Short term visitor arrivals to Australia from Japan, Seasonally Adjusted Estimates

Figure 4 shows seasonally adjusted estimates for June, July and August 1990. The different estimates produced in different years shows how the seasonally adjusted estimates change as additional information becomes available.


KEY POINT:

Revisions are improvements to the seasonally adjusted estimates as seasonal patterns in the original estimates become clearer over time.


WHEN WILL TREND ESTIMATES BE REVISED?

Trend estimates also face minor levels of revision as more data becomes available. This occurs as both the seasonal pattern and the underlying direction of the series become clearer. Usually the largest revisions occur in the three time periods following the initial trend estimate. For example September 2005 trend estimates will normally face revision in October 2005 and be relatively stable after December 2005.

Figure 5: Short term visitor arrivals to Australia from Japan, Trend Estimates

Figure 5: Short term visitor arrivals to Australia from Japan, Trend Estimates

Figure 5 shows that trend estimates for September 1990 are revised as more original estimates become available. The size of these revisions becomes smaller over time. We have selected an unusually drastic example for clarity.


KEY POINT:

Revisions are improvements to the trend estimates as seasonal patterns and the underlying level of the original estimates become clearer over time.