Using bank transactions data to measure household spending


As the COVID-19 pandemic gathered pace early last year, the ABS secured access to additional data sources to assist in producing our statistics, including from selected banks which provided de-identified data in an aggregated format. Through this period of rapid change, the banks provided monthly data to assist with the compilation of household final consumption expenditure (HFCE) estimates, helping enhance the quality of Australia’s national accounts statistics. Banks data continues to provide a rich and valuable data source to support and supplement our estimates, and also offers opportunities for more frequent estimates.

A monthly indicator of household spending

The ABS has been working with selected banks to explore the possibility of producing a new monthly indicator of consumer spending. The banks are providing data for transactions recorded either from merchants, who have card terminals or eCommerce terminals with that bank, or customers who have cards with that bank. Other variables differ across the banks, with different levels of geography and differences in time series length. Data supplied by banks is aggregated at either merchant category, retail category or industry class.

The ABS first assigned the various banks’ spending categories to the international Classification of Individual Consumption According to Purpose (COICOP) used in the Australian national accounts. Some bank spending categories are excluded, for example spending categories that are defined as capital rather than consumption in the national accounts.

This monthly indicator is still in the early stages of development, although we are already seeing improvements in the methodology for supplementing the banks data and in compilation processes. This spotlight provides an overview of developments to date.

Alignment with quarterly national accounts estimates

To demonstrate alignment with quarterly national accounts HFCE, the monthly indicator series was aggregated to a quarterly frequency. At this stage, 7 out of 12 categories of expenditure have demonstrated good, or reasonably good, alignment with quarterly national accounts HFCE estimates, while the remaining 5 categories require improvement. Figures 1 – 5 illustrate some of the categories with good alignment.

There are some spending categories for which banks data is not suitable for measuring household spending. For example, rent for housing, water and sewerage charges, electricity, gas and other fuels, and purchase of vehicles. These are not well represented by banks card transactions data as households rarely use credit or debit cards to pay for these types of expenditure and they are not usually recorded at a card terminal. To estimate monthly household spending for these categories, the ABS is looking for data that better represents this spending, such as banks data from other payment channels, including BPAY, and data from alternative sources.

The national accounts measurement of some spending categories doesn’t conceptually align with banks transactions data. For example, household expenditure for insurance and other financial services and gambling is a margin between spending and return whereas banks transactions data represents total spending. Additionally, the spending category of education services is represented in banks transaction data on a cash basis rather than an accrual basis with large spikes at the beginning of the first and the second semesters. This reflects when payments were made rather than when the education services were consumed.

Once the monthly indicator has been further developed, the ABS will publish a full quality assessment in late 2021. The ABS appreciates the support of the data providers which has enabled these important statistics to be produced.

Back to top of the page