Feature Article—Revisions to retail trade data
The Australian Bureau of Statistics (
ABS
) has announced revisions to the Retail Business
Survey (RBS), which provides the information for its regular Retail Trade
publication. This feature discusses the changes and the implications for the
Monthly statistical bulletin.
What is
changing?
The RBS is used to estimate the value of the turnover
of retail businesses in Australia. The
ABS
reports that three changes will be made to the RBS. These changes include removing
some sectors from the survey, reducing the survey sample size and changing the
sampling method.
The RBS will no longer include the hotels and licensed
clubs sector and the selected services industries sector. These exclusions
reduce the size of the ‘completely enumerated’ sector from 730 to 640
businesses, and sampled businesses from 3214 to 2775—a reduction of 13 per cent
in businesses surveyed. (The enumerated sector includes major retailers such as
department stores, and the sampled sector includes all other retail businesses).
The survey size will also be reduced for the businesses
that remain in the survey. The completed
enumerated sector size will fall from 640 to 500 businesses, due to a
reclassification of businesses through size. For the sampled sector the survey sample will be reduced from 2775 to 2702
businesses. The combined reduction in
survey sample size will be around six per cent.
The monthly activity is also affected by the change in
sampling methodology, not just the sample reduction. Typically, the RBS would
survey businesses on a monthly basis. That is, before the change, 3944
businesses would have been surveyed each month. After the change, the survey will be run monthly; however, the survey
will include only 1401 businesses. This figure includes the completely
enumerated sector’s 500 businesses, each month, and only 901 sample sector
businesses. For the latter sector, the monthly target is derived from sampling
901 businesses per month, over a three month period, so that 2702 businesses
are sampled per quarter.
Why does it
matter?
A survey measures responses to questions from a sample
of participants within a population. The intention is to estimate statistics
from the sample, such as average monthly turnover, which can be used to
estimate, or infer, the approximate values of a population as a whole.
Surveys contain sampling errors which occur because they
do not cover an entire population, as in a census. While understanding the cause
of, and reasons for, the errors is not critical it is important to understand
that as the sample size decreases, the errors from the sampling exercise
increase. Therefore, reducing the scope
and sample of the RBS mean the retail trade data are less accurate.
The
ABS
provides
details of the RBS sampling error by publishing the relative standard errors (
RSE
) for the retail trade data. According to ABS’ estimates, the reduced
sample size will cause an increase in the monthly RSEs of between 30 per cent
and 100 per cent, depending on the sector or State being considered.
What will be
available?
As a consequence of the discussed changes the retail
trade information will be split into two components. The first component will be a monthly report ‘Retail
Trade Trends’. This report will provide trend estimates of broad industry
groups and State totals, but limited industry level details. Additional
information will be provided, such as seasonally adjusted data, but is
considered too volatile due to the smaller sample size and associated higher
RSEs.
The second component will be a quarterly publication
‘Retail Trade Quarterly Indicators’. This publication will be more reliable
than the monthly report. It will include original, seasonally adjusted and
trend details for industry and sub-industry data, at the Australian and State
levels. The publication will also include chain volume estimates to provide
inflation adjusted turnover.
What is the
expected impact on the data?
The immediate impact of the changes will be to lower
the value of sales activity reported, compared to the previous series. The
impact of the sample size and methodological change will be to reduce the
reliability of retail trade data.
The ABS has estimated that the revisions would have
seen a 0.3 per cent smaller measure of retail trade for June 2008 data. To see
how the series compare over time, the graph below shows the pre and post change
seasonally adjusted series from April 1982 (based on ABS backcasts) to June
2008. It is clear that total retail trade is now lower.
Monthly trade
data under different methods
Monthly statistical
bulletin Table 4.1
Monthly statistical bulletin Table 4.1 reports
seasonally adjusted monthly retail sales data. As a consequence of the changes,
the monthly data will be replaced by quarterly seasonally adjusted retail sales
data, once the quarterly series begins being published in November 2008. When the quarterly ABS report is published, this
bulletin will incorporate all backcast data for consistency, and any further
changes will be advised.
Adrian
Makeham-Kirchner
|
