(Bloomberg) -- The same kind of survey failures that have left the UK unsure about the number of people in Britain’s workforce are now raising doubts about the size of its economy.
The Office for National Statistics has seen a collapse in responses to a key household questionnaire that it relies on to produce growth estimates, prompting it to take remedial action. The so-called Living Costs and Food Survey — used in GDP readings such as those for the third quarter due to be released on Monday — helps the British statistics agency determine people’s incomes and what they’re spending their money on.
The latest data shows that barely one-fifth of the forms are being completed and returned, mirroring a shortfall in the ONS’s monthly jobs survey, which policymakers say no longer provides a reliable snapshot of the country’s labor market. In the case of the growth report, the drop-off means an increased risk of sudden and retrospective changes to gross domestic product readings, potentially feeding bad decisions by political leaders and officials.
“There’s a bigger chance of revisions, [and] there’s a bigger chance of policy errors as a result because policy is made in real time,” said Greg Thwaites, research director at the Resolution Foundation. “If there’s a problem with the LCF survey and the consumption data, that will have a small effect on the GDP estimate, but not a huge one.”
The decline shows that the crisis of confidence that has rocked the ONS’s labor report could spread to other indicators used by the Bank of England, the Treasury and others. In slow-growing Britain, small data revisions can make the difference in public perceptions about whether Prime Minister Keir Starmer is failing or succeeding in his promise to rebuild the economy.
Employment Minister Alison McGovern told Bloomberg last week that the government was already using alternative sources for jobs data, warning it is “too big for us to wait” for reliable ONS figures. BOE officials have also expressed dismay over the inaccurate figures that have clouded their view of the labor market at a crucial moment for interest rates.
Deploying AI
An ONS spokesperson said in a statement that it was deploying artificial intelligence to streamline data collection in the LCF survey and, from the first quarter of 2025, relying more on card spending to supplement responses. Still, the agency said that the survey remained an “important source of household spending data” and that efforts to boost the number of questionnaires have increased the number of responses by more than a third.
While response rates for such economic surveys have been declining for decades, the trend has accelerated since 2020, possibly due to changes in behavior and data collection during the Covid-19 pandemic. In the case of the LCF, the response rate was just 22% last year. That compares with roughly 40% in 2020 and 70% three decades ago.
Concerns over the accuracy of the LCF survey during the pandemic led the ONS to withhold some tables. The low share of households answering means biases may remain in the data, such as one group, like younger people who might not check their mail as frequently, being underrepresented.
“People are much more reluctant and also people are just time scarce,” said Huw Dixon, professor of economics at Cardiff University. “They’ll use data which they have from other sources, such as tax returns.”
The ONS uses three approaches to measure GDP: Output, expenditure and income. Monthly estimates are produced using only the output data, while later figures are derived by averaging output, expenditure and income data from the LCF to create a more complete reading.
The ONS has mentioned issues with LCF response rates in recent GDP data releases, including when it made a sharp, retrospective upgrade to 2022 growth from 4.3% to 4.8%. Britain has performed better since the pandemic than originally thought, undercutting some of the criticism levelled at the now-former Conservative government.
In explaining the revision, the ONS highlighted large differences in the growth readings derived from each of the three methods, which produced GDP growth readings ranging from 3.1% to 5.3%. Problems with the LCF were also mentioned in the latest quarterly growth release, although the output approach takes the lead for the two most recent quarters of data.
“The low response rate on the LCF isn’t great of course, but it’s only really a problem if it’s a biased non-response rate,” said Thwaites of the Resolution Foundation.
©2024 Bloomberg L.P.