CHAPTER 7 SAMPLING
7.1 The Australian Producer Prices Indexes (PPIs) measure the change in the prices of products (goods and services) as they leave the place of production or as they enter the production process. The International Trade Prices Indexes (ITPIs) measure the change in the prices paid for imported products and the prices received for exported products.
7.2 The volume and complexity of the available transactions from which to obtain prices means that it is not possible to collect prices from every provider and for every product or to take into account every price at which products are sold. Consequently, it is necessary to adopt a sampling approach to obtain transaction prices for representative products from selected businesses.
7.3 The selection of a sampling approach to produce PPIs and ITPIs is based on criteria that include available resources and data, as well as consideration of survey sample error.
7.4 Probability sampling is the selection of a sample of producers and products from a population of industrial activity in which each producer and product has a known chance of selection. Under this approach, all producers and products have a known probability (chance) of inclusion in the sample. The key benefits of probability sampling are that sampling design controls for sampling error and allows for its measurement. There are, however, disadvantages associated with this approach.
7.5 The use of probability sampling requires identification of all units (e.g. producers and products) in all industries of the economy that are in-scope of the price index (known as the sampling universe). This requirement translates into the need for an up to date sampling frame of products. The practical difficulties in satisfying this requirement mean that there are high costs in the design, implementation and ongoing administration of probability samples for price index purposes.
7.6 Non-probability sampling is known as judgmental or purposive sampling, or expert choice, and samples are chosen by experts to be representative. In a price indexes context this involves index compilers selecting producers and products from which to obtain prices using available information on the relative importance of individual producers and products.
7.7 A key benefit of non-probability sampling is that it can be used where the sample population is not known. However, it is not possible to produce a measure of sampling error for indexes compiled from non-probability samples. It is generally accepted that price indexes are an area of statistics where the risks in not using a probability sample are relatively low, as the diversity of price change charged by various producers over time is usually small(footnote 1) .
THE ABS SAMPLING APPROACH
7.8 Non-probability sampling is used by the Australian Bureau of Statistics (ABS) to compile the PPIs and ITPIs(footnote 2) . This is primarily due to the lack of available data to undertake a probability sampling approach.
7.9 The non-probability sampling approach uses available data to build a picture of the overall market for a particular index. It is used to determine who the potential providers are, their relative importance, what particular products they sell, who they sell to, and their pricing policies. A range of information sources are used in the selection of products and corresponding businesses. These include market reports, industry associations, ABS industry census data, other ABS surveys, and discussions with potential providers. This information is used to develop a comprehensive understanding of the market. The index compiler uses this information to make appropriate judgments in the sample selection process.
7.10 The effectiveness of the non-probability sampling approach depends on the index compiler’s ability to construct product samples that produce price movements that are representative of price movements of all in-scope products. This is achieved by:
Selecting individual products to be priced
- sampling products to represent the price movements for all the products which come within the scope of the particular price index
- sampling providers to represent all the suppliers/users of the selected products. In general, the aim is to cover businesses which account for a high proportion of sales or purchases of the products in the index
- sampling products from each provider to represent the whole product range within the selected product group
- obtaining prices for each sampled product which best represent the price movements of all transactions in the selected product group.
7.11 The choice of individual products to be priced is made in consultation with providers to the Survey of Producer Prices
. This process ensures that the sampled products are clearly identified and described, that they are representative of the price behaviour of the products they represent and confirms that the product specifications will be available for pricing in future periods. The representivity of the products in the price basket
is regularly reassessed. This process is described further in Chapter 11.
7.12 It is not always possible to sample products for pricing purposes. On occasion, model pricing is used. For example, while many construction projects differ markedly in their design, there may be common elements, such as construction materials used. In these circumstances, the provider may provide regular re-pricing of a model, whose price will be expected to vary in line with market changes in prices for these construction materials. See Chapter 8 for more information on pricing specifications.
1 For example, Dalén, J (1998) ‘‘Studies on the Comparability of Consumer Price Indices’’, in International Statistical Review, Vol. 66, No. 1, pp. 83–113 and de Haan, J and E. Opperdoes, and C. Schut (1999) ‘‘Item Selection in the Consumer Price Index: Cut–off Versus Probability Sampling’’, in Survey Methodology, Vol. 25, No. 1, pp. 31–41. <back
2 Non–probability sampling is also used in the Consumer Price Index (CPI). <back