CHAPTER 3: CONCEPTS, SOURCES AND METHODS
3.1 This chapter describes the retail trade margins concept as well as the data sources and methods employed to construct PPIs for the output of the retail trade industry.
THE RETAIL TRADE MARGINS CONCEPT
3.2 The SNA describes wholesale and retail margins as follows:
“Although wholesalers and retailers actually buy and sell goods, the goods purchased are not treated as part of their intermediate consumption when they are resold with only minimal processing such as grading, cleaning, packaging, etc. Wholesalers and retailers are treated as supplying services to their customers by storing and displaying a selection of goods in convenient locations and making them easily available for customers to buy. Their output is measured by the total value of the trade margins realized on the goods they purchase for resale. A trade margin is defined as the difference between the actual or imputed price realised on a good purchased for resale and the price that would have to be paid by the distributor to replace the good at the time it is sold or otherwise disposed of. The margins realised on some goods may be negative if their prices have to be marked down. They must also be negative on goods that are never sold because, for example, they go to waste or are stolen.
"The standard formula for measuring output has to be modified for wholesalers or retailers by deducting from the value of the goods sold or otherwise used the value of the goods that would need to be purchased to replace them. The latter includes the additional goods needed to make good recurrent losses due to normal wastage, theft or accidental damage. In practice, the output of a wholesaler or retailer is given by the following identity:
the value of output = the value of sales,
plus the value of goods purchased for resale and used for intermediate consumption, compensation of employees, etc.,
minus the value of goods purchased for resale,
plus the value of additions to inventories of goods for resale,
minus the value of goods withdrawn from inventories of
goods for resale,
minus the value of recurrent losses due to normal rates of
wastage, theft or accidental damage.” (Endnote 16)
3.3 Or put more simply, Output = Sales – Cost of Goods Sold.
MEASURING PRICE CHANGE
3.4 A price index is a measure of the proportionate, or percentage, changes in a set of prices over time. In constructing a price index to be used to deflate observed transactions the objective is to measure pure price change which abstracts from any change in the quality of the products that are priced. This is referred to as pricing to constant quality. The starting point is to measure price change for specific individual products to constant quality and then to weight these measures together in terms of their relative economic importance with respect to one another to arrive at an aggregate or summary measure of price change. These objectives present a number of practical challenges when producing RTPIs.
PRICING BASIS FOR THE COMPILATION OF A RTPI (Endnote 17)
3.5 The unobservable retail trade margins price is derived as the difference between the price at which the product is sold and the current replacement cost, to the retailer, of the product sold.
3.6 While collecting information on the prices at which products are sold is relatively straightforward, the collection of information on the replacement cost of goods sold is less so, as in concept they should be valued at their replacement cost at the time of sale. This is difficult to achieve in practice. Although most businesses are able to report the cost of goods sold, the pricing basis is almost universally based on the purchase price rather than the current replacement cost. The extent to which a measure based on purchase price will deviate from the conceptually preferred measure will depend on the length of time goods are held in inventory and the rate at which their purchase price changes.
3.7 Typically, price indexes are based on prices for a sample of very narrowly specified products (e.g. 1 kilogram of Granny Smith apples). The use of narrowly specified products is not necessary for the purpose of measuring retail trade margin prices. In a typical price index the selection of narrowly defined products for pricing is designed to achieve the objective of pricing to constant quality. The quality of the products produced or purchased can be seen to be embodied in their physical characteristics, so preserving these characteristics over time serves to ensure that measures of price change are based on comparisons of like with like. The argument for following this approach is less compelling when the objective of the RTPIs is to measure the price of the distribution service. While there must exist a retail trade margin for each product (positive or negative), it is less clear that the distribution service itself can be meaningfully disaggregated to this level. The view of the ABS is that the quality of the distribution service is more closely related to the range of similar products provided for sale. In other words, the distribution service associated with the provision of fresh fruit and vegetables as a whole is a more meaningful concept than the distribution service associated with Granny Smith apples. This view would also appear to align better with the pricing practices of businesses which tend to set selling prices of individual products with the objective of maximising a retail trade margin across a range of products.
3.8 This 'range of products' view underpins the RTPIs developed by the ABS. Accordingly, these indexes are based on sales and cost of goods sold data collected for categories of products rather than for specific products. The problem then becomes how to define the various 'product categories'. If the level of product aggregation is too broad, the price measure is likely to be overly influenced by any shifts in sales between products that have naturally different retail trade margins. For example, the measured retail trade margin for a product grouping that includes both dairy products and fresh fruit and vegetables is likely to vary depending on the relative value of sales of dairy products compared to fresh fruit and vegetables – that is, the measured aggregate retail trade margin price could vary from period to period due to changes in the relative volume of sales rather than any change in individual retail trade margins. An index displaying this characteristic is referred to in the price index literature as suffering from compositional shift. The challenge is to clearly define product categories in terms of product coverage while minimising the risk of compositional shift. The ABS has selected the Supply and Use Product Classification (SUPC), used in compiling the S–U tables which underpin the annual Australian National Accounts, for this purpose. The product groups are not specific to the retail trade industry, but represent product groups across the broader economy.
3.9 Therefore, aggregate retail margins data for the product groups represented by each SUPC are collected from retailers. These aggregate data reflect the price of the retailer’s service.
PRICING TO A CONSTANT QUALITY (Endnote 18)
3.10 The quality of the distribution service provided by a retailer is difficult to measure. Any attempt to measure the distribution service needs to consider characteristics such as outlet opening hours, numbers of checkouts, floor space, general ambience, temperament of staff, ease of parking, range of products on offer, proximity to other stores etc. While these characteristics may not lend themselves to ready measurement, it is clear that they are linked to the specific outlet providing the products. Therefore, although it may not be possible to make explicit adjustments for any changes in the quality of the service, steps can be taken to minimise or control for quality change.
3.11 Given that the quality of the service can be considered to be unique to each outlet, it follows that the measurement of retail trade margins prices is best done at the outlet level. In the case of multi–location businesses, aggregated data for products represented by selected SUPCs are requested for the selected outlets in each state and territory in which the business operates.
3.12 To assist in identifying any changes in outlet specific quality characteristics, the ABS maintains a close relationship with all data providers. When a quality change is identified at an outlet then the results for that outlet are adjusted to accommodate for the change in quality. Such adjustment includes the imputation of retail trade margins for that outlet from other similar outlets in the sample.
DATA SOURCES AND COLLECTION
3.13 The ABS assessed the availability and suitability of various data sources for the production of RTPIs. This included a search for administrative data as well as utilising data currently collected by the ABS from retail trade businesses. It was determined that the detailed data required to produce the RTPIs was not available and that a separate data collection would need to be implemented. A Retail Trade Margins Survey (RTMS) was then developed.
3.14 The data items collected are:
SCOPE AND COVERAGE
- Business level total Sales and COGS, by state/territory of operation, for product groups as classified by SUPC;
- For each business, outlet level total Sales and COGS, for each of a selection of outlets within each state/territory of operation, for product groups as classified by SUPC.
Retail trade margins data are collected by the ABS for those SUPCs with the highest contribution (Endnote 19)
to total retail trade margins, and where data collection is feasible. A sample of businesses are selected from those classified to ANZSIC06 Division G Retail Trade, i.e. “those units engaged in the purchase and on–selling of goods to the general public (Endnote 20)
. Individual businesses are selected for the survey on the basis of market influence and the individual products they sell.
The RTMS collects retail trade sales and COGS data from, primarily, large retail businesses. A partial coverage collection approach is used where large businesses represent the retail trade margins of all sizes of retail business. Medium sized businesses are included only where industry concentration is not sufficient to exclude them and/or when the retail trade margins of large businesses are not sufficiently representative of those of medium sized businesses. The RTMS excludes small businesses, except where they are a part of a franchise operation.
Where possible, businesses selected in the RTMS were enrolled via personal visit. This is an active strategy to address provider concerns relating to the confidentiality of retail trade margins data. This strategy ensures a high level of cooperation from providers and assists in the control of data quality. The ABS has implemented an electronic data collection using a web form for this survey.
The most significant challenge faced by the ABS during the enrolment phase for the RTMS was obtaining historical data back to the March Quarter 2008. These data are required for linking the current RTPIs with those indexes created in the experimental program conducted from the December Quarter 2003 to the December Quarter 2007.
The ABS has initially limited data collection to 27 SUPCs out of a total of 119 SUPCs corresponding with ANZSIC06. The ABS will investigate extending the collection of data on SUPCs beyond current coverage and will do so where retail trade margins share is significant and data collection is feasible.
METHODS USED FOR THE COMPILATION OF A RTPI
Index Number Formulae
Various methods are available to compile price indexes. While NSOs may prefer conceptually superior index formulae, such as the Fisher index, decisions about index compilation methods need to consider key issues such as index compilation costs, data availability, provider burden, timeliness, and end use/purpose.
Following a review of the original RTPI and consideration of the various index compilation methods available, the ABS determined that the RTPI would be constructed using a Lowe index formula (Endnote 21)
with weights updated and indexes chained annually. The Lowe class of price index is defined as measuring the proportionate change between two periods in the total value for a specific fixed basket of goods. The ABS, like most NSOs, makes use of the Lowe index approach to compile their PPIs and CPIs.
The formula for the Lowe index is presented below. Let there be n products in the basket with prices Pi
and quantities Qi
, and let the two periods being compared be 0 and t while b is the quantity reference period. The Lowe index, PLo, is defined as:
The RTPI Construction Process (Endnote 22)
A price index is constructed by measuring the proportionate changes in the prices of many products over time and combining these measures using weights to form aggregate indexes. For the RTPIs, the aggregation process is in three stages.
In the first stage, final margin prices and final margin price relatives are calculated from reported outlet sales and COGS data against SUPCs. The final margin price relatives are then aggregated to form the state/territory level SUPC indexes.
In a second stage of processing, national SUPC indexes are constructed by aggregating the state/territory SUPC indexes.
The third and final stage of processing combines the national SUPC indexes to produce the national RTPI. The weights used during aggregation in each of the three stages of processing are updated on an annual basis.
Stage 1: Aggregating Over Outlets to Form SUPC Specific State/Territory Level Indexes
Stage 1 involves the creation of state/territory indexes by SUPC. These state/territory level SUPC indexes are calculated by first determining the final price (dollar retail trade margin) for the SUPCs at the outlet level. This is achieved by applying the current period percentage retail trade margin by outlet to a reference period ‘preserved volume measure of sales’ as explained below.
The ‘preserved volume measure of sales’ is determined by adjusting the reference period sales data by the movement in the component of the CPI which most closely matches the product coverage of the index SUPC. The purpose of this transformation is to maintain the reference volume of sales from which current period retail trade margins data are calculated. The outcome of this process is the determination of the dollar retail trade margins in period t based on the sale volumes in period 0.
A final margin price relative at the outlet level is obtained from the ratio of the final margin price in the current period, t, to the final margin price in the price reference period, 0. The SUPC state/territory index is then calculated as a weighted sum of these outlet price relatives.
The outlet weight at an SUPC level is calculated by multiplying the outlet’s contribution to the outlet sample retail trade margin for that SUPC, by the business’s contribution to the state/territory retail trade margin for that SUPC.
Stage 2 and 3: Calculating National SUPC Indexes and a RTPI
Stage 2 involves the creation of national margin indexes representing the products grouped by SUPCs. These national SUPC indexes are created by weighted aggregation of the state/territory level SUPC indexes. Weights used at this stage are ‘SUPC retail trade margin shares by state/territory’.
Stage 3 involves the creation of the national RTPI by weighted aggregation of the national SUPC indexes. Weights used in this stage are unpublished national SUPC retail trade margin shares. SUPC retail trade margin shares by state/territory are determined from unpublished national SUPC retail trade margin values obtained from the national S–U tables. National S–U tables are compiled on an annual basis.
See Appendix 1 for a detailed explanation of the compilation process for the RTPIs.
16 2008 SNA, p113, [6.146 – 6.147]17 This section is largely a reproduction of text found in cat. no. 6402.0 published on 20 Feb 2007.18 This section is largely a reproduction of text found in cat. no. 6402.0 published on 20 Feb 2007.19 Weights are based on the unpublished 2008–09 S–U tables.20 ANZSIC 2006, cat.no. 1292.0, p244.21 The original RTMPI 2002 to 2007 used a Fisher index formula in stages 1 and 2 of the index construction process. Owing to compilation costs and data availability the new RTPI construction process uses a Lowe index formula.22 A detailed description of the RTPIs construction process can be found in Appendix 1 of this publication.