Inside statistics: Retail trade data are used to monitor pulse of economy
August 12, 2004
By Pali Lehohla
Last week Statistics SA announced that the value of retail trade sales at constant (2000) prices for the three months to May increased 8.4 percent, compared with the same period a year ago.
Similarly, retail trade sales, at constant (2000) prices for the period January to May increased 9.4 percent compared with the same period in 2003.
Furthermore, retail trade sales at constant (2000) prices for May 2004 increased 10.3 percent compared with May 2003.
What is the importance of this? What are data from the survey of retail trade sales used for, and how are they compiled?
Why has Stats SA revised this data, and might they be revised again?
The results of the monthly retail trade sales survey are used to compile estimates of gross domestic product and its components, which are used in monitoring the state of the economy and in the formulation of economic policy.
These statistics are used in the analysis of comparative business and industry performance.
Retail trade includes the resale (sale without transformation) of new and used goods and products to the public for household use.
A retailer is defined as an enterprise deriving more than 50 percent of its turnover from sales of goods to the public for household use.
The survey covers retail enterprises - butchers, general dealers, bottle stores, dealers in clothing, footwear and textiles, dealers in furniture and household requisites, bookstores and stationers, jewellers, chemists, dealers in miscellaneous goods, and repairers of personal and household goods.
It is conducted by mail on a monthly basis.
Questionnaires are sent to a sample of about 3 000 enterprises from a population of about 28 000 enterprises. The value of sales is obtained monthly from the sample of 3 000 enterprises.
The retail industry is divided into four groups. All large enterprises (group one), which comprise about 11 percent of the enterprises in the current sample, are completely enumerated.
Simple random sampling is applied to group two (medium size), and to groups three and four, which are small enterprises. The total value of sales of the large enterprises is added to the weighted totals of groups two, three and four to reflect the total value.
In January 2003 Stats SA introduced a new retail sales survey using an upgraded business register.
The previous retail trade sales survey and the current one were run simultaneously from January 2003 to December 2003.
This was to ensure that the new survey was bedded in properly and to provide a set of factors that could be applied to the old series to revise previous estimates back to January 1998.
The new survey produced estimates of a significantly higher level and exhibited a distinctively different seasonal pattern.
Because this new seasonal pattern had only been established for a 12-month period, Stats SA was reluctant to introduce this new seasonal pattern into the old series.
Instead, it chose to use the January 2003 difference between the old and new series to back-cast the previous series.
Stats SA has revised the old series, taking into consideration the seasonal pattern established by the new collection (even though only 17 data points are available and 36 are usually required).
The effect of this change in methodology resulted in the revised series exhibiting the same year-on-year movements as the old series, but with month-on-month movements reflecting the seasonal pattern exhibited by the new series.
As more information becomes available, it may be necessary to modify this seasonal pattern.
Estimates from January 2003 onwards have not been affected by this revision, although January to March 2004 data have been revised.
Even though the current survey uses Stats SA's new business register as its sampling frame, it is now about two years old and, as a result, may not reflect structural changes that may have taken place over that period.
Measures are under way to construct a new frame from the business register with its reference period being the end of June 2004.
Pali Lehohla is statistician-general and head of Stats SA
|
|
