Accuracy and reliability

Overall accuracy

The general assessment is that the quality of the statistics is high, with a few reservations. Using the time series for long-term analysis can be problematic. This is because of the way the retail trade index is designed and calculated, where not all structural changes in the population and sample are reflected in the figures (e.g. changes in the activity codes causing units to enter or leave the retail trade sector), which is why chaining is used in the calculation of growth rates for the value indices (see data compilation).

Moreover, there will typically be a slight underestimation of the volume index in periods where consumers are buying less expensive goods and in low-priced stores, and correspondingly there will be a slight overestimation in periods where consumers choose high-end goods and stores. This is caused by the use of CPI sub-indices as deflators for calculating the volume indices. The quality of the volume indices is generally very dependent on the composition of these deflators used in the conversion from value index to volume index. In periods with changing consumption patterns, the deflation, and hence the volume indices, are less accurate, because CPI does not reflect all the changes.

A provisional analysis shows that the lack of consideration of the above-mentioned substitution effects have had an effect on the volume index in recent years causing an underestimation of the yearly growth rates of approximately 0.2-0.4 percentage points.

To state the overall accuracy, one must distinguish between the accuracy of the estimated totals and the accuracy if the monthly growth rates, which is the prime focus of these statistics.

The overall uncertainty of the total index is estimated to be less than 1 per cent. On commodity group level, the uncertainty of the group Food and other convenience goods is about the same, whereas for Clothing etc. it can be as high as 3 per cent and for Other consumer goods up to 2 per cent.

The accuracy of the monthly growth rate is generally very high. For the total index, the uncertainty is estimated to be less than 0.2 percentage points., while it can be a little higher on commodity group level. In a few cases, non-response from highly essential enterprises can cause slightly lower accuracy in the first publication of figures for a particular month.

Sampling error

Sampling error, is measured by the CV-values for the estimated turnover totals behind the retail trade index. For the total was on average 0.2 pct. this year. This is below the accepted threshold of 10 pct. and therefore, considered acceptable. At the subgroup level, the sampling error can be somewhat higher. For the largest industries it is normally below 2 percent, for medium-sized industries it is less than 5 percent., and for the small industries it can be as high as 10 per cent., but only in very unusual cases it is higher than this. The average for all industries this year was 1.79 pct.

Non-sampling error

The overall accuracy is affected by sampling error and non-sampling error combined. Non-sampling error includes measurement error and non-response error.

The overall uncertainty of the monthly total retail trade index was 0.69 percent in 2021. On commodity group level, the uncertainty of the group Food and other convenience goods is about the same, whereas for Clothing etc. it can be up to 10 per cent and for other consumer goods up to 5 per cent. Approximately half of this uncertainty is due to non-sampling error such as bankruptcies or changes in sector classification.

Quality management

Statistics Denmark follows the recommendations on organisation and management of quality given in the Code of Practice for European Statistics (CoP) and the implementation guidelines given in the Quality Assurance Framework of the European Statistical System (QAF). A Working Group on Quality and a central quality assurance function have been established to continuously carry through control of products and processes.

Quality assurance

Statistics Denmark follows the principles in the Code of Practice for European Statistics (CoP) and uses the Quality Assurance Framework of the European Statistical System (QAF) for the implementation of the principles. This involves continuous decentralized and central control of products and processes based on documentation following international standards. The central quality assurance function reports to the Working Group on Quality. Reports include suggestions for improvement that are assessed, decided and subsequently implemented.

Quality assessment

The general assessment is that the quality of the statistics is high, with a few reservations. Using the time series for long-term analysis can be problematic. This is because of the way the retail trade index is designed and calculated, where not all structural changes in the population and sample are reflected in the figures (e.g. changes in the activity codes causing units to enter or leave the retail trade sector), which is why chaining is used in the calculation of growth rates for the value indices (see data compilation).

Moreover, there will typically be a slight underestimation of the volume index in periods where consumers are buying less expensive goods and in low-priced stores, and correspondingly there will be an overestimation in periods where consumers choose high-end goods and stores. This is caused by the use of CPI sub-indices as deflators for calculating the volume indices. The quality of the volume indices is generally very dependent on the composition of these deflators. In periods with changing consumption patterns, the deflation, and hence the volume indices, are less accurate, because CPI does not reflect all the constant changes in the consumption pattern.

A provisional analysis shows that the lack of consideration of the above-mentioned substitution effects in recent years have had a combined effect on the volume index causing an underestimation of the yearly growth rates of approximately 0.2-0.4 percent points.

To state the overall accuracy one must distinguish between the accuracy of the estimated totals and the accuracy of the monthly growth rates, which is the prime focus of these statistics.

The overall uncertainty of the total index is estimated to be less than 1 per cent. On commodity group level, the uncertainty of the group Food and other convenience goods is about the same, whereas for Clothing etc. it can be up to 3 per cent and for Other consumer goods up to 2 per cent.

The accuracy of the monthly growth rate is very high. For the total index the uncertainty is estimated to be less than 0.2 per cent, while it can be a little higher on commodity group level.

In a few cases, non-response from highly essential enterprises (units) can cause slightly higher uncertainty in the first publication of figures for a particular month.

Data revision - policy

Statistics Denmark revises published figures in accordance with the Revision Policy for Statistics Denmark. The common procedures and principles of the Revision Policy are for some statistics supplemented by a specific revision practice.

Data revision practice

The first published figures for a given month are published 22-28 days after the end of the reference period and only include figures on commodity groups (not industries). On industry level, the first publication is one month later. At the same time, the first published figures on commodity groups are revised. Hence, these figures are published 52-58 days after the end of the reference period.

One month later, the second and final revision of the figures is made on commodity group level, and the figures on industry level are revised for the first and only time. Hence, the final figures are published 72-78 days after the end of the reference month.

In case of major changes in methodology or in case of errors, the figures can be revised further back. In 2021, there were the grossing up model did not handle missing observations that resulted from restrictions against COVID-19 well. Therefore, in the fall of 2021 the whole year were recalculated using improved grossing up methods. These revisions were significant for individual branches but not retail trade as a whole. This type of revision is rare; there have been no other major revisions in the past 5 years.

Size of revisions

The accuracy of the monthly growth rate is very high. For the final version of the total index, the uncertainty as measured by relative mean absolute revision (R-Mar) was 0.041 percent this year. While the mean absolute revision (MAR) for the final version of the index was a little higher with a value of 0.320 percent. Here too, approximately half of this uncertainty is due to non-sampling error.

The first published figures on commodity groups are revised twice in connection with the publication of the figures for the two following months. The revisions are generally small, especially bearing in mind the very short production time.

Measured on the monthly growth rate of the seasonally adjusted volume index, the first revision resulted in an R-Mar average of 0.037 percent this year. While the MAR revisions were 0.286 percent points for the first revision.

Measured on the monthly growth rate of the unadjusted volume index, the first revision is normally between 0.1 and 0.3 percent points, primarily caused by data submitted at a late stage and to a lesser degree corrections of timely submitted data.

The second revision is normally below 1 percent.

On the commodity group level, the revisions can in general be twice the size of the revisions to the total index. Revisions on industry level are normally quite small due to the high response rate at the time of first publication.