The UK Office of National Statistics has been under a lot of pressure—sometimes even mockery—in recent years for the continued deterioration of its statistics quality. The Labour Force Survey, which produces readings like the unemployment rate, is the most notable example. It has been plagued by an extremely low response rate.
On March 25, though, the ONS completed a significant milestone which should receive a round of applause. The ONS successfully switched to retailer-supplied scanner data as the source of most of its grocery price calculations as they published the report for February CPI inflation figures. Scanner data, sometimes called point-of-sale price data, are the digital records of how much consumers actually paid at the checkout counter in the supermarkets.
This is a huge contrast to the conventional way of CPI data sourcing: sending a data collector to the physical stores to find the sample product and write down their prices. In a general sense, it is a major advancement in data collection: The sampling period can be wider; the statistical agency can acquire the quantity purchased data, on top of the prices; and, most important of all, they are real prices paid out of pocket by real humans—those are records of real-life commercial activity, instead of a list of prices of representative products chosen by statisticians.
Of course, the ONS is not the first statistical agency to adopt grocery scanner data in its national inflation statistics. Far from it. Just take an example closer to us: Statistics Canada has been implementing scanner data in the grocery price index since 2018.
As shopping experiences in advanced economies have long been fully digitalized—checkout staff uses a barcode scanner and the POS system to tally the total price, and customers commonly use Apple Pay or credit cards to pay—we should expect that most statistical agencies are adopting this kind of statistical enhancement project.
Interestingly, though, in the US—the world’s most important economy—its inflation statistics enhancement progress has surprisingly lagged.
BLS said scanner data are ‘too expensive’
To be clear, the agency responsible for CPI calculations, the Bureau of Labor Statistics (BLS), has achieved some meaningful progress in adopting the so-called alternative collection methods in its inflation data compilation. For example, it uses transaction data provided by J.D. Power to calculate the index for new cars and new trucks and it uses non-survey-based data sources to calculate the motor fuel index, among other recent improvement efforts.
The absence of a grocery scanner data project in the US’s CPI calculation is extra intriguing given the fact that the BLS had actually looked into implementing this data source. According to a 2022 draft paper written by three BLS economists—Crystal G. Konny, Brendan K. Williams and David M. Friedman—the BLS in 2012 attempted to test incorporating scanner data into the CPI Food at Home categories.
While BLS researchers studied technical methods to handle the data incorporation (some of them are detailed in this working paper), the statistical agency in the end halted the project. The reason?
“Ultimately, BLS found that it was less expensive to collect data in stores than to pay for Nielsen Scantrack real-time data and the geo-graphic and outlet detail needed to support the monthly CPI.” The BLS acquired data from data aggregator Nielsen Scantrack for the test, but these data were so costly that the agency would rather continue using in-person price collections.
This is not what we expected from POS data at all. We thought cost efficiency is also one of its advantages. More interestingly, why StatCan and ONS can do that if scanner data are so pricey?
The UK and Canada don’t have to pay for the data
To find out, we contacted both statistical agencies. Here is the reply from ONS:
We do not pay for the data, retailers provide it voluntarily, which helps to keep overall costs down.
The data are free? What about StatCan? Here is their reply:
We can confirm that all scanner data used in the CPI program is provided voluntarily by our data providers at no cost.
So, on the one hand we have BLS complaining scanner data are expensive to purchase, and on the other, statistical agencies in other countries can get similar data without even paying?
Let’s dive into it a little bit deeper. How does the ONS get its scanner data? “We currently receive data directly from individual retailers. This approach enables us to discuss data issues and quality with retailers directly, giving us a much better understanding of the underlying data,” they told us in an email.
The ONS, in their January explainer, said that their current scanner data setup covers around 50% of the UK grocery market, giving them “approximately 300 million price points derived from sales of over a billion units of products per month, collected directly from supermarket scanners in-store and online.”
They explained to us that dealing directly with the supermarket corporations can reduce the risk of relying on a single aggregated feed from one provider, like Nielsen Scantrack mentioned by the BLS above. “[W]hile an issue with an individual retailer would reduce scanner data coverage, the failure of a single aggregated feed could prevent us from producing meaningful statistics for that part of the basket,” the ONS representative added.
Why are corporations willingly cooperating with statistical agencies? Well, a disclosure by StatCan gives us some hints. On this page, the Canadian statistical agency clearly stated that “[r]etailer Point of Sale data” are requested under “the authority of the Statistics Act” and that “[t]his is a mandatory request for data.”
So, in the case of Canada, the request for scanner data is, in a way, an offer which the supermarkets cannot refuse!
We asked StatCan about the use of such a mandatory request. They explained, “[u]nder the Statistics Act, the agency has the authority to acquire data from businesses at no cost to ensure comprehensive national statistics.”
“In practice, however, our data acquisition is rooted in collaborative partnerships built on mutual trust and shared public benefits,” the StatCan representative added, “By prioritizing voluntary cooperation and collaboration, we aim to create value for all stakeholders while upholding the highest standards of efficiency and respecting the role businesses can play in contributing to broader economic insights.”
Remember, as we cited above, StatCan confirmed all the Canadian grocers provide their data “voluntarily.”
So, does the ONS have similar authority? The UK statistics-collecting agency pointed us to the ONS data acquisition policy, highlighting this passage:
“The ONS will consult with, and consider the views of, data suppliers before issuing a statutory notice or requesting access to data. We commit to exploring collaborative solutions and negotiated data arrangements in preference to issuing requests or notices to enable this access.”
As the policy page explained, the legal provision for the ONS to acquire data was “enhanced by the introduction of the Digital Economy Act in 2017.”
The 2017 Act inserted the power for the UK Statistics Authority, the overarching government department that houses the ONS, to, “by notice in writing,” require data disclosure from UK corporations.
Which is why, also as cited above, “retailers provide it voluntarily, which helps to keep overall costs down.”
Why can’t the BLS get free data?
So, we assume the reason BLS thinks scanner data is expensive is because they don’t have similar authority? Admittedly, we are not experts in US statistics authority and that is not information that is easy to confirm.
Lucky for us, Erika McEntarfer, former BLS commissioner who was unfortunately fired by Donald Trump last year after a jobs report showed unfavorable readings and now a Tad and Dianne Taube Policy Fellow at Stanford Institute for Economic Policy Research, was willing to have a quick chat with us on this issue.
“BLS does get commercial data directly from companies for some items – new vehicles, gas, cell phone services, prescription drugs. They have been expanding use of alternative data but are not yet where they would like to be – mostly because they are constrained by cost and staffing,” McEntarfer explained. For example, the BLS has to engage in extensive research before they can integrate the data into the US statistical system.
Also, as we have been wondering, “data often isn’t free,” McEntarfer added.
So why statistical agencies in other countries don’t have to pay, but the BLS does?
“The US has no legislation that allows statistical agencies access to scanner data for free,” McEntarfer confirmed to us.
There we have it. US statistical agencies don’t have the same legal power as their counterparts in the UK and Canada to make a mandatory request for data. It may be the reason that US companies are less likely to voluntarily provide data to the BLS for free.
The BLS operates without the legislative authority that lets the ONS and StatCan “encourage” grocery chains to share their checkout records — which means it has to pay for what its counterparts receive for free. Why the US has never passed similar legislation — whether that is a matter of political culture, corporate lobbying, or simply a statistical agency that has never had the clout to push for it — is a question we hope to return to.
If you are an expert in the field or you work in US statistical agencies and have something you would like to share, please feel free to contact us. We would like to report more on this issue.
… liked this!
… liked this!