Sampling

7.1 To achieve the conceptual objective of measuring pure price changes over time, regular monitoring of the prices of goods and services acquired by the Consumer Price Index (CPI) population group is necessary. However, it is not possible in practice to price every single type or variety of good and service purchased by the CPI population group at each collection cycle. The ABS overcomes this practical problem by using purposive sampling procedures, where representative sets of goods and services are selected for regular pricing. Similarly, it is not practical to observe the prices of the selected goods and services in all retail outlets selling these items to the CPI population group. Again, the ABS uses purposive sampling to select a representative sample of outlets at which to price the selected items in each collection cycle.

7.2 With the availability of transactions (scanner) data, it is possible to include all items in some measures of price change. As of the December quarter 2017, the ABS has introduced new index methods, known as multilateral methods, for 28 ECs in the CPI. These methods enable a census of products to be used from transactions datasets and the weighting of prices at the product level, enhancing the quality of the CPI. For further detail on these methods, see Use of transactions data in the Australian CPI of this manual.

7.3 However, for other items in the CPI a sampling approach is required.

7.4 The goods and services included in the CPI pricing samples are selected carefully to represent the range of types and varieties of goods and services bought by the CPI population group. Selection is made only after obtaining detailed information about the buying habits of the CPI population group, such as which varieties and brands of products are the largest selling types or which packaging sizes are most commonly purchased. This process involves extensive consultations with, for instance, retailers, manufacturers, importers, government authorities, and professional and trade associations. In selecting the items to be priced, the following factors are taken into consideration. 

• The importance of the expenditure class relative to the total CPI. In general, the more important the expenditure class, the larger the number of items priced.
• The degree of homogeneity in the range of goods or services covered by the expenditure class. The more homogeneous the range, the fewer the number of price indicators required.
• The extent to which the various products covered by an expenditure class are subject to different influences and cost pressures which are likely to result in disparate movements in prices.
• The likelihood of the particular type of good or service continuing to be available on the market for a reasonable period of time. In general, it is preferable to price the same specific items for a reasonable length of time rather than having to change price indicators regularly when particular goods or services appear and then disappear after only a short time on the market.
• The extent to which the item can be defined and described clearly and unambiguously to ensure that the selected goods or services can be priced to constant quality over time. For example, in pricing confectionery it is likely that packaged, brand name chocolates would be easier to price to constant quality over time than loose chocolates with no identifying brand name.
• The availability of transactions data provides detailed item revenue data to enable evidence based selection of the highest revenue items. It also enables more timely replacement of items with falling revenues to newer items with increasing revenues.

7.5 After the items to be priced have been selected, detailed specifications are prepared to ensure that all staff involved in price collection and compiling the CPI have exactly the same understanding of which particular items are to be priced. For most goods, it is a straightforward matter of describing their characteristics. These may include brand name, material of composition, model number, style, size, and type of packaging.

7.6 It is generally more difficult to specify service items adequately because both quantity and quality are harder to describe. In addition, more detailed descriptions are usually required in the specifications for services in comparison to those for goods. For example, the specification for a can of tomato soup may consist of only two characteristics: the brand name, and the weight of the can. However, the specification for a travel service such as a bus fare would have three characteristics: the concessional status of the traveller (e.g. adult, student, child, pensioner); the specific bus route including the origin of the journey and its destination; and the time of the journey (e.g. peak or off-peak).

7.7 The most representative items for each city are selected for pricing. In many instances, items are available nationally, and are thus included in the sample for each of the capital cities. In these cases, items are described in sufficient detail to ensure that price collection staff are able to locate them. These items can be readily and clearly defined by characteristics such as make, model, and size as a specification for use nationally.

7.8 Where items aren't representative (or available) nationally, a generic description is provided in sufficient detail to ensure that ABS staff will be able to locate a matching item. The item selected should be representative for the particular city and its characteristics align to the generic description provided.

7.9 Consumers purchase the goods and services priced in the CPI from a wide variety of retail outlets. Examples of these outlets include supermarkets, department stores, hotels, motor vehicle dealerships, doctors’ surgeries, electricity and gas shopfronts, travel agencies, schools, and child care centres. For every item selected for pricing, the main types of outlets from which the CPI population group buys the items need to be identified so that the ABS can select representative samples of these outlets.

7.10 In selecting outlets for inclusion in samples for the CPI, the following factors are taken into account.

• The importance of the expenditure class relative to others in the CPI. In general, the more important the item is (i.e. the larger the expenditure weight), the larger the sample.
• The number of suppliers of the good or service in the city concerned. Generally, the larger the number of suppliers, the larger the sample. In some cases, however, there may be only one supplier, such as a city council or transport authority.
• The degree of dispersion in prices among outlets. Where the expected dispersion in prices is large, the sample should be large too. For example, a large sample of fruit and vegetables outlets is usually needed. However, with newspapers, a small sample is sufficient because standard prices are generally adhered to.
• The geographical spread of outlets. As far as possible, the samples are selected to cover the main areas in which households from the CPI population group are known to make their purchases.
• The ownership of retail chains. Large retail chains frequently have an Australia-wide or state-wide pricing policy. In these cases, pricing one outlet in the chain would be considered sufficient to obtain a representative estimate of price movement for that chain.
• The availability of transactions data. Transactions data allows for full coverage of all stores within a city's metropolitan area to determine the average price paid for an item for a given month or quarter.

7.11 The samples of respondents are reviewed regularly to ensure that they remain representative of the CPI population group's sources for purchases. Events such as company takeovers, new retailers entering the market, existing chain organisations opening new outlets, or new shopping complexes opening up can all lead to the need to change the samples of respondents so that they continue to be representative of the CPI population group's purchases. Changes to the sample of respondents or specifications are carried out using the splicing process discussed in Price Index Theory of this manual.