• Summary
• Downloads
• Explanatory Notes
• Related Information
• Past Releases

METHODS FOR MAKING USE OF TRANSACTIONS DATA IN THE CPI

The Australian Consumer Price Index (CPI) is a robust measure of household inflation. The ABS produces the CPI each quarter using price quotations sourced via personal visits, online and telephone collection, and from administrative data sources (ABS, 2016a). One type of administrative data used is transaction records captured by retailers at the point of sale (transactions data).

Since 2014, prices from transactions data have accounted for about 25% of the weight of the CPI. At present, the ABS uses each retailer’s transactions data to source prices for a sample of products that were formerly collected via personal visits to the corresponding retailer. This approach reduces the cost of (in-person) price collection. The prices derived from transactions data are also more representative of the prices actually paid by consumers than price quotes collected via one or a few personal visits.

The ABS has been assessing methods for maximising our use of the information on transactions datasets to enhance the CPI. Our goals have been to expand the product sample to incorporate all of the products that appear on these datasets, and to use the revenue information on these datasets to weight products in accordance with their economic importance. This information yields a more complete picture of the products purchased by consumers, and of how consumers substitute between products as prices and preferences change over time.

This information is also more dynamic than the samples traditionally used in the CPI, in the sense that products may appear and disappear rapidly, and product revenues may change dramatically from one period to the next due to sales. In this context, the bilateral index (estimation) methods traditionally used in the CPI which compare prices between two periods can yield implausible results. Ivancic, Diewert and Fox (2011) suggest using multilateral index methods instead, which simultaneously compare prices over a window of more than two periods.

Over the last few years, the ABS has conducted extensive testing of a range of multilateral index methods. As there is no universally endorsed “best” multilateral method, we have developed a framework to guide this assessment. We have also assessed further methodological changes necessitated by the adoption of multilateral methods, including methods for splicing multilateral comparisons onto previously published index levels to extend the index, and minor changes to the index structure. To resolve methodological issues, we have collaborated with international experts and with our counterparts at other statistical offices, and consulted with key stakeholders. Through this assessment, we have found the GEKS-Törnqvist multilateral method and the “mean splice” method for extending the index to be most suitable for our purposes—for more details, see (ABS, 2017).

The ABS plans to use multilateral methods to make greater use of transactions data in the production of the official CPI from the December quarter 2017 onwards.

References
Consumer Price Index: Concepts, Sources and Methods.
cat. no. 6461.0.
Australian Bureau of Statistics, Canberra

Making Greater Use of Transactions Data to Compile the Consumer Price Index.
cat. no. 6401.0.60.003.
Australian Bureau of Statistics, Canberra.

Holt, M., Webster, M & Pham, H (2017) An Implementation Plan to Maximise the Use of Transactions Data in the CPI.
cat. no. 6401.0.60.004. ABS, Canberra.

Ivancic, L., Diewert, E. W. & Fox, K. J. (2011) Scanner data, time aggregation and the construction of price indexes.
Journal of Econometrics, 161, 24-35.