• Share this page to Facebook
  • Share this page to Twitter
  • Share this page to Google+
Methodology changes to the Retail Trade Survey

This page describes changes we've made to the methodology used in the Retail Trade Survey from the September 2017 quarter.

Why we made changes to the Retail Trade Survey

We changed the methodology in the Retail Trade Survey to:

  • make greater use of administrative data sources 
  • reduce number of respondents by 87 percent 
  • implement a consistent methodology across economic statistics 
  • improve the quality of the published series, particularly at regional level, by removing sampling error 
  • make more data available and easier to access at a lower level, such as data for industry by region

The effects of the changes will be seen in Retail Trade Survey: September 2017 quarter, to be published 23 November 2017

Summary of changes

Under the previous design, we surveyed all the large businesses in each industry, plus a sample of medium- and small-sized businesses. We supplemented this with modelled tax data for the smaller businesses.

Under the new design, we use administrative data (goods and services tax (GST) data, sourced from Inland Revenue) wherever possible, and supplement this by surveying only the largest and most complex businesses. With the new design, we reduced the number of businesses being surveyed by 87 percent and eliminated most of the small- and medium-sized businesses from the survey entirely. Figure 1 compares the current and new designs.

Figure 1

Diagram showing that the new methodology will make greater use of administrative data (GST) than the current design.

The methodology changes improved the quality of the series we publish, particularly at sub-national level. This is because we effectively have a full coverage of all businesses within an industry, rather than relying upon a smaller sample to represent the entire population.

The new design lets us produce sales data for each of New Zealand's 16 regions. Up until now, we have only been able to release data for 6 regional groupings.

Stats NZ has already implemented this new methodology in two other surveys:Economic Survey of Manufacturing: June 2017 quarter and Wholesale Trade Survey: June 2017 quarter. The changes made to the Retail Trade Survey will more closely align it with these other sub-annual financial collections, and flow through into a new income-based measure of quarterly gross domestic product (GDP).

See Business Data Collection – initial data release for more information.

Scope and coverage

The scope and coverage of the Retail Trade Survey has changed under the new design.

Population of interest – Retail Trade Survey

The target population is all kind-of-activity units (KAUs) operating in New Zealand classified on Stats NZ’s Business Register (BR) to Australian and New Zealand Standard Industrial Classification 2006 (ANZSIC06) Division G – Retail Trade, and ANZSIC06 Division H – Accommodation and Food Services.

Previously the target population was all geographic units (GEOs) operating in New Zealand classified on Stats NZ’s Business Frame to ANZSIC06 Division G – Retail Trade, and ANZSIC06 Division H – Accommodation and Food Services.

We moved to a KAU-based collection from a GEO-based collection to align with our other sub-annual financial collections including the Economic Survey of Manufacturing and the Wholesale Trade Survey. The Annual Enterprise Survey is also collected on a KAU basis.

Although the Retail Trade Survey will be collected on a KAU basis from the September 2017 quarter onwards, Stats NZ will continue to collect GEO or store-based sales information from the largest businesses. This will ensure we can continue to provide regional and other geographic-based retail statistics.

Additional variables

The Retail Trade Survey used to collect sales and stocks only. Under the new methodology retail trade has been incorporated into the relatively new business data collection, a quarterly financial collection covering most industries in the economy and with an expanded range of variables.

In addition to sales and stocks, retail trade businesses will also provide information on purchases, salaries and wages, and net profit (EBIT) from the September 2017 quarter onwards. These new variables will be used as flow measurements in our new balance sheets, and a new income measure of quarterly GDP.

See Business Data Collection – initial data release for more information on the Business Data Collection.

Regional information

The new methodology, with its greater reliance on administrative data, allows for more accurate provision of data at lower geographic levels. This is due to the full coverage nature of the new design.

We will release retail trade data at the regional council level (16 regions) from the September 2017 quarter. The regional council series will replace the six retail regions published under the previous Retail Trade Survey design. These regional series will be backcast to the June 2011 quarter using the new methodology.

As these back-series have been created using the new methodology, they have not been edited and processed as part of our normal statistical production. We may make some small improvements to these series at a later date. The backcast regional series will not exactly sum to the New Zealand retail trade totals as the official New Zealand retail trade series is backcast using industry (and not regional) values and is designed to preserve previously published industry movements.

Methodology changes

The series for the Retail Trade Survey is now produced using GST data wherever possible. We have done extensive work over the last few years on GST data and have established that it is a reliable measure of activity in retail businesses, except for the largest and most complex businesses.

We developed robust methods of transforming the data, which is submitted at different frequencies, to a quarterly frequency. In addition, we developed methods of detecting and removing sales and purchases of large capital items, which can at times occur in the GST data. These are not part of the conceptual measure of sales required for national accounts purposes.

We supplement the GST data for each of the series with survey data for large and complex businesses which meet the following criteria:

  • a $50 million significance rule – if an enterprise, or group of enterprises linked by ownership, has an annual GST turnover of more than $50 million 
  • a 3 percent industry dominance rule – if an enterprise makes more than a 3 percent contribution to annual total income for an industry 
  • all enterprises that have a significant level of activity across multiple industries.

More detail on the GST data assessment and methodology changes is available on request from info@stats.govt.nz.

Sales

Under the new design, we will only survey the largest and most complex businesses. For the remaining businesses, we will apply a statistical model to their GST data to determine sales.

GST sales for each non-surveyed business in the population is subject to modelling to:

  • transform the GST data to a quarterly frequency; and 
  • apportion data between members of GST groups.

Standardising the GST reference period to quarterly

Inland Revenue collects GST data as part of administering New Zealand’s taxation system. It is not primarily designed to produce economic statistics.

We developed methods to transform the GST data, which can be submitted monthly, two-monthly, or six-monthly, to a quarterly frequency.

Sales for monthly returns are added together to get their quarterly sales. For two-monthly and six-monthly GST filers, some modelling and forward-casting is required to estimate a quarterly value.

Example 1 – For a business that files GST two-monthly, we need to estimate their quarterly sales based on just two months’ data. To do this, we apply a modelling factor to their two-monthly sales, to estimate their quarterly sales.

Example 2 – For a business that files GST six-monthly, we need to estimate their quarterly sales without any data from the period of interest. In this case, we disaggregate their sales into three two-monthly portions, and then apply a forward-casting factor to estimate their most recent sales. We then convert this two-monthly estimate into a quarterly estimate, as above.

Accounting for GST groups

Where one business reports GST on behalf of a number of other businesses that are linked by ownership (GST groups); the GST data must be apportioned between the different businesses.

For GST groups, data from the Inland Revenue’s employer monthly schedule is used as follows:

  • simple groups (where one group member has more than 80 percent of the group’s employees) – all sales are assigned to the dominant group member 
  • complex groups (where each individual group member has fewer than 80 percent of the group’s employees) – GST sales are apportioned between group members in proportion with their share of the group's employees.

Stocks

Quarterly stock (inventory) measures are used in the production of GDP. In the previous Retail Trade Survey design, stocks are primarily measured from data collected in the postal survey.

Under the new design, we will continue to collect stocks data for large and complex businesses. However, there is no quarterly stocks data available for other businesses from administrative sources.

We have a range of methods to estimate stocks for businesses that are not surveyed. Different methods are better suited to different conditions depending on the size of the industry and the contribution of the surveyed units. The aim is to use the best method available for each industry.

The three methods we use to estimate the value of stocks for a given industry are:

  • Benchmark to annuals: We use this approach where the surveyed businesses capture quarterly change in stocks (from one quarter to the next quarter), but do not accurately capture the level of stocks. We obtain stock levels by ‘rating up’ the aggregate stock series using data from the Annual Enterprise Survey.
  • Model from annuals: We use this approach for some smaller industries where the stock levels remain relatively consistent over time. We use a model to calculate the sub-annual stock estimates – using quarterly GST sales or purchases in combination with stocks obtained from the Annual Enterprise Survey. The model is updated on an annual basis when new data becomes available.
  • Stratified random sample: In industries where these methods are not suitable, we use a stratified random sample survey. We send a stocks-only questionnaire to sampled businesses. We do not use this method for any industries in Retail Trade, although we use it for other industries in the sub-annual financial collection.

Table 1 details the method used for each industry:

Table 1  

 Method used to estimate value of stocks, by industry
Industry Method 
 G1110 Motor vehicle and parts  Benchmark to annuals
 G1120 Fuel  Benchmark to annuals
 G1210 Supermarket and grocery stores  Benchmark to annuals
 G1221 Specialised food  Model from annuals
 G1222 Liquor  Model from annuals
 G1311 Furniture floor coverings houseware textiles  Benchmark to annuals
 G1312 Electrical and electronic goods  Benchmark to annuals
 G1313 Hardware building and garden supplies  Benchmark to annuals
 G1321 Recreational goods  Benchmark to annuals
 G1322 Clothing footwear and accessories  Benchmark to annuals
 G1330 Department stores  Benchmark to annuals
 G1340 Pharmaceutical and other store based retailing  Benchmark to annuals
 G1350 Non store and commission based retailing  Benchmark to annuals
 H2110 Accommodation  Benchmark to annuals
 H2120 Food and beverage services  Model from annuals

Linking data from the new methodology with historic series

To ensure continuity, we will link data produced using the new methodology with previous retail trade data using the June 2017 quarter as the linking period. Each published industry has been backcast with previously published percentage movements maintained. This has resulted in some level changes at the published industry level. As we have maintained the changes at the industry level there are some small changes at the total retail trade levels.

We also took the opportunity to make some revisions to three industries where recent movements have been inconsistent with the movements shown when applying the new methodology. This resulted in some changes to recently published industry movements. The industries where this occurred were:

  • Supermarket and grocery stores – comparisons between the old and new methodology suggested that published movements over the last two years have been slightly weaker than they should have been. We increased the movements from September 2015 to June 2017 to match the rate of increase indicated by the new methodology. Figure 2 shows the series under the new design compared with the previously published series.

Figure 2

 

Table 2 shows the changes to previously published movements for supermarkets and grocery stores as a result of the changes made.

Table 2 

 Changes to previously published movements for supermarkets and grocery stores
 Series

 Supermarket and grocery stores
Actual retail sales values

% change from same quarter last year

 Supermarket and grocery stores
Actual retail sales values

% change from previous quarter

Originally published
14 August 2017

 New design

 Originally published
14 August 2017

New design 
 June 2017 quarter

0.4

 2.4

-0.1 

 0.5

 March 2017 quarter

 0.4

 2.4

-1.3 

-0.9 

 December 2016 quarter

 1.1

 3.2

-0.7 

 -0.6

 September 2016 quarter

 1.7

 3.8

 0.1

0.9 

 June 2016 quarter

 2.0

3.6 

 0.7

1.3 

 March 2016 quarter

 2.2

3.2 

0.5 

 0.7

 December 2015 quarter

2.0 

2.5

0.4 

1.0 

 September 2015 quarter

2.8 

 2.8

0.4 

0.6 

 Source: Stats NZ
  • Liquor retailing – the contribution from the relatively small tax portion of the current Retail Trade Survey for the June 2017 quarter was unusually strong. We made a correction, and the June 2017 data was republished. 
  • Electrical and electronic goods retailing – the contribution from the relatively small tax portion of the current Retail Trade Survey for the June 2017 quarter was unusually strong. We made a correction, and the June 2017 data was republished.

These revisions will be incorporated into GDP statistics in Gross Domestic Product: September 2017 quarter, to be published 21 December 2017.

Seasonal adjustment

We produced seasonally adjusted and trend series using the X-13ARIMA-SEATS package developed by the U.S. Census Bureau. As we have maintained the previously published industry movements for retail trade data, except as noted for specific industries, there is minimal change to the seasonally adjusted movements over time.

Data comparability over time

Figures 3–5 show the comparison between previously published data for retail trade and the series using the new design.
Please note these backcast series are released for information, and are based on data available at the time of compilation. Because we are still receiving and analysing data for the September 2017 quarter, there may be some changes. Therefore, treat the series presented in these graphs as provisional.

Figure 3

 

Figure 4

Figure 5

Timing change

Due to the increased reliance on administrative data the release dates for the Retail Trade Survey have been adjusted slightly. Retail trade releases will be approximately one week later than they were under the previous methodology. Retail Trade Survey: September 2017 quarter will be published on 23 November 2017.

New time series families

We created new families and time series identifiers for the data produced using our new methodology. The previous families will no longer be updated, and will be labelled as discontinued on Infoshare. The previous series will still be accessible by selecting the ‘show discontinued’ option from Infoshare.

Retail Trade Survey

The Infoshare families for the new data are:

  • sales and stocks by industry, in current and constant prices (SAFC) (Qtly–Mar/Jun/Sep/Dec) 
  • percentage changes, total and core sales, in current and constant prices (SAFC) 
  • sales by region in current prices (SAFC) 
  • sales per head of population (SAFC).

The new time series identifiers (with examples) are detailed below.

TSM series info

Sales and stocks by industry, in current and constant prices (SAFC)

Series description: Quarterly sales in current and constant prices
New series pattern: RTTQ.SF[19A-V][1][CK][AST]
Example series description: Total sales in current prices, actual
Example series pattern: RTTQ.SF91CA

Series description: Annual sales in current and constant prices
New series pattern: RTTA.SF[19A-V][1][CK][A]
Example series description: Core sales in current prices, seasonally adjusted Example series pattern: RTTQ.SF11CS

Series description: Quarterly stocks in current prices
New series pattern: RTTQ.SF[19A-V][9][C][AST]
Example series description: Total stocks in current prices, actual
Example series pattern: RTTQ.SF99CA

Percentage changes, total and core sales, in current and constant prices (SAFC)

Series description: Percentage change from same period previous year, actual sales New series pattern: RTTQ.SF[19][CK][A][A]C
Example series description: Total sales in current prices, actual, % change from same period previous year
Example series pattern: RTTQ.SF9CAAC

Series description: Percentage change from previous quarter, seasonally adjusted sales
New series pattern: RTTQ.SF[19][CK][S][P]C
Example series description: Total sales in constant prices, seasonally adjusted, % change from previous quarter
Example series pattern: RTTQ.SF9KSPC

Sales by region in current prices (SAFC)

Series description: Actual sales by region, total and core industries
New series pattern: RTT_.SF[19]R[A-PXYZ][C][A]
Example series description: Core sales for Auckland in current prices, actual, annual (year ended March)
Example series pattern: RTTA.SF1RBCA

Series description: Sadj/Trend total sales by region
New series pattern: RTTQ.SF[9]R[A-PXYZ][C][ST]
Example series description: Total sales for Nelson in current prices, seasonally adjusted
Example series pattern: RTTQ.SF9RKCS

Sales per head of population (SAFC)

Series description: Sales per head of population in current and constant prices
New series pattern: RTT_.SFPP[CK][AST]
Example series description: Total sales per head of population in current prices, trend, quarterly
Example series pattern: RTTQ.SFPPCT

Contact

For more technical information, please contact:

Citation

Stats NZ (2017). Methodology changes to Retail Trade Survey. Retrieved from www.stats.govt.nz.

ISBN: 978-0-908350-23-0

Published 30 October 2017

  • Share this page to Facebook
  • Share this page to Twitter
  • Share this page to Google+
Top
  • Share this page to Facebook
  • Share this page to Twitter
  • Share this page to Google+