A bigger budget and larger team would enable Statistics South Africa (Stats SA) to introduce technology that would improve the accuracy and frequency of inflation data, as well as do more surveys.
The data Stats SA tracks directly informs monetary policy, wage negotiations and aids it in understanding consumer behaviour. More, and better, data would enable real-time consumer behaviour shifts, improve the accuracy of inflation measures and provide deeper sector insights, such as into telecoms.
Patrick Kelly, chief director of price statistics at Stats SA, says one area the agency is exploring is the use of scanner data from retailer point-of-sale systems to capture real-time inflation data.
Weighing the data
Kelly explains this would involve using barcode-level sales data provided by retailers, typically weekly, which would give the agency visibility into products sold across outlets and weighting them by actual sales volumes.
“Scanner” technology would allow Stats SA to track shifts in consumer behaviour – such as switching from one bread brand to a cheaper alternative – when purchasing decisions are made, rather than waiting for basket reweighting, which currently happens every five years.
“That's the advantage of the scanner data: you'll be able to pick up those shifts. And so, you would see then that, overall, people are paying perhaps a bit less for bread or a bit more.”
Kelly adds that the UK’s Office for National Statistics introduced grocery scanner data into inflation data last month after a five-year project involving around 40 dedicated staff. StatsSA has two to three people working on it part-time, alongside other responsibilities, he explains.
“We have received positive feedback from retailers in relation to this. Sadly, due to the financial constraints that we are under, we do not have staff that we can dedicate to this work and so progress is slow.”
From clipboards to tablets
Kelly notes that, currently, Stats SA has a dedicated team of people who price between 60 000 and 70 000 items across 25 hubs across the country, capturing the selling price of the exact same item from the exact same outlet in successive months.
“It wasn't that long ago that we used clipboards, but about five years ago or so, we switched to using tablets,” says Kelly.
Not all items are surveyed monthly, with rentals captured quarterly and school fees annually in March; timeframes Kelly says are reflective of how often and when each basket item would increase in cost. Additional surveys are triggered if there’s a significant price change outside the normal cycle.
The changes in selling prices, captured as the Consumer Price Index each month, is part of the data that the Monetary Policy Committee (MPC) uses to inform monetary policy, such as interest rates.
The South African Reserve Bank’s MPC next meets on 28 May, with markets expecting either a 25-basis point interest rate increase at that meeting or the following one in July.
The CPI is published with a one-month lag, which Kelly says is the international standard. “It is designed to reflect changes in prices of consumer goods and services over the whole month.”
Risky data
However, incorporating real-time information from point-of-sale devices would pose some challenges, says Kelly. One technical issue is classification because different retailers use different naming conventions for the same product, he explains.
Stats SA has begun applying machine learning to standardise inputs, says Kelly. “That is one area that we started to try to apply some kind of AI methodologies.”
The agency is collaborating with Microsoft on experimental projects in these areas. Documents from the National Budget indicate that the agency has received R343 billion for the current fiscal year to use technology to enable the production of statistics, as well as broader public communications.
Kelly points out that AI and big data introduce variables that are harder to manage when the goal is consistent, publishable monthly data. He notes that traditional survey methodologies are controllable and enable scientific surveys.
“When you start to talk about AI and big data, it’s more experimental – you have to figure out how you’re going to use it. The data can change. You don’t have control over the data source. You can't manage that risk.”
Underfunding cost
Stats SA aims to release at least 258 publications per year over the medium-term related to socio-economic statistics, including those on gross domestic product, price indexes and the labour market.
The 2026 National Budget allocated almost R3 billion to Stats SA for the current financial year, with R208.4 million set aside over three years for IT infrastructure and operational shortfalls. “Our money woes are well publicised, and everyone seems to think it's very important that we get sorted out,” says Kelly.
Kelly cites examples of surveys of services sectors such as telecoms, business services, legal services and road freight as among those that it can’t currently undertake, all of which would also provide useful economic insights. “Our surveys of new sectors in the economy are weak because we don’t have money to run new surveys.”
National Treasury documents also show the agency aims to partner with other government entities to determine if other data, such as from the Companies and Intellectual Property Commission and National Treasury, can be incorporated into its publications, which would reduce sample sizes and costs. This work will be carried out in the Economic Statistics programme, which is allocated R995.9 million over the next three years.
Despite a general lack of funding, Kelly says Stats SA has managed to maintain its “product offering” and its quality of work. However, the lack of budget means the agency is “very constrained” in terms of development, which would then enable additional surveys each month.
National Treasury was not immediately available to provide comment.
Share
