Michael Simmons has narrated this article for you to listen to.
There were cheers in the Treasury last month as the nation’s statisticians discovered a spare £3 billion down the back of the sofa. The Office for National Statistics (ONS) admitted that VAT receipts had been under-reported by £1 billion last year and £2 billion this year.
The newly found cash will go some way to filling the Chancellor’s £30 billion fiscal black hole, but not everyone was celebrating. Just 150 miles west along the M4, at the home of the ONS in Newport, Wales, the mood was grim. The embattled agency was splashed across the papers for the wrong reasons. This error was just the latest in a long line of data disasters. That’s because the quality of Britain’s official data, like much of British life, has deteriorated to the point of being dangerously unreliable. Ministers, officials and policymakers are governing in the dark.
This was the last thing the ONS needed. Staff morale had only just begun to recover after boss Sir Ian Diamond retired in May, citing health issues. His ‘Medici prince’ manner had grated on ministers and it was accepted that new leadership was needed.
Progress, however, has been slow. Staff complain in online forums that the management has little analytical experience and is instead fixated on tech gimmicks. One insider says: ‘Senior leaders make decisions with no operational understanding and no thought for impact.’ Another adds: ‘There are lazy people at all grades who feel entitled and have no idea what work looks like outside the civil service.’ Meanwhile, senior ONS staff have acknowledged that people who raised issues were ignored and in many cases were ‘not treated very well’.
I’m told that the management is obsessed with ‘the next shiny thing’ rather than the dull fundamentals. Sir Robert Devereux, a former Whitehall permanent secretary, reached much the same conclusion in his review of the ONS this summer. There was a ‘commendable interest’ in innovation, he said, but it had led to the ‘deprioritising of the less exciting but crucial task of delivering core economic statistics of sufficient quality to guide decision–making’.
Some trace the rot back to the ONS’s relocation to Newport, an early levelling-up experiment that drove out much of its experienced London workforce. But the real collapse began during the pandemic. The ONS scaled back knocking on doors and chatting to people. It’s an expensive and time-consuming process, but it’s also the best way to gather information. When the ONS switched to internet surveys and phone calls, it found far fewer people answered. The Labour Force Survey – which underpins all official jobs and unemployment data – has never fully recovered. A decade ago, more than half of the households approached took part; during lockdown the rate fell to 27 per cent and has only crept back to a third. The result is that the sample size has collapsed and the data is far more error-prone. Read the Labour Force Survey and there have never been as many Brits in employment; use tax data and Rachel Reeves has destroyed more than 100,000 jobs. Both can’t be right.
The figures for one industry had to be suppressed after the sample size fell to just five people
Such discrepancies make governing almost impossible. In March it emerged that the figures for one industry had to be suppressed after the sample size fell to just five people. Retail data has gone haywire too: the ONS postponed its August report for two weeks because it was so obviously wrong, which Andrew Sentance, the former Bank of England policymaker, called ‘a totally excessive delay’.
GDP figures are so volatile that it can be claimed Liz Truss both shrank and grew the economy, depending on which set of numbers you use. Inflation was overstated in May because of a data glitch, while an error earlier in the year wiped £2.3 trillion off estimates of household wealth. The ONS now struggles to produce reliable numbers on jobs, inflation, trade and retail – the basics of a functioning economy.
The ONS is far from being the only place where Britain’s data infrastructure is crumbing. Between 2018 and 2023, the Office for Statistics Regulation stripped 17 government datasets of their ‘official statistics’ status. Last year alone it suspended 15 and this year it’s already up to 14. Nearly 85 per cent of economists polled by Reuters say they no longer trust Britain’s data, and the quality is getting worse, not better.
Things have become so bad that the Bank of England – one of the ONS’s biggest ‘customers’, in civil service parlance – has started to build its own data sources. Private consultants smell an opportunity to cash in: Capital Economics now publishes an unemployment metric ‘to provide a clearer read’ on the labour market.
The Bank is also guilty of its own data debacles. Recently it quietly altered its survey of inflation expectations from 3.5 to 3.4 per cent – a seemingly trivial change worth billions – without issuing a correction until embarrassed by the Financial Times coverage. Getting this stuff right really does matter. A misread on wage growth, for instance, could lead the Bank to cut or raise interest rates, costing jobs or stoking inflation.
While the quality of data is a big source of frustration, ‘the patchwork of data systems across government and the way civil servants present data’ is a bigger issue, explains one official. ‘Literally the prime minister doesn’t know the real numbers, it’s often just concealed from them,’ George Osborne revealed. In one Whitehall basement, vital pensions data still sits on magnetic tape – readable but not updateable. The system is so decrepit that information on one of the state’s biggest liabilities has been ‘basically unusable’ for years.
In England alone, GPs register five million more patients than the ONS says are alive
When data is lost, there is the potential for dire consequences. An HMRC data feed which powers Universal Credit payments is so error-prone as many as one in four in-work claimants could have been underpaid, overpaid or not paid at all, according to figures seen by The Spectator. Officially, the government says there is an error rate of less than 1 per cent. Some insiders tell me they suspect people may have been wrongfully prosecuted for benefit fraud, in what could turn out to be a scandal equivalent to the Post Office’s if action isn’t taken.
Immigration statistics are in a similar state of chaos. Britain cannot say how many people live here: in England alone, GPs register five million more patients than the ONS says are alive. Depending on the definition – ‘foreign nationals’ or ‘born overseas’ – foreigners are either more or less likely than natives to commit crime. Which is it? Nobody knows. Collecting data on visa overstays has proved so difficult that the Home Office ‘just gave up’, a former special adviser tells me. I asked one senior Reform figure if they were worried that the quality of population data would make it difficult to implement their policies in government. ‘Frankly, we’re fucking terrified. We’ve got a Heath Robinson system when we should have Tron,’ he said.
The Treasury is seething at the Office for Budget Responsibility, which tries to predict the effects of government policies. The OBR has chosen this Budget to revise down its estimates of productivity growth, which it claims have been overly optimistic for years. That decision has cost the Chancellor tens of billions of pounds. While Whitehall obsesses over this revision, the Bank of England quietly went much further. In May it downgraded its productivity growth forecast to just 0.7 per cent in 2027 – a catastrophic figure. Yet if you use modelling based on another dataset entirely – HMRC payroll – it looks like the OBR has in fact been underestimating productivity. These are three different parts of government, using different figures and each telling a totally different story.
Whichever part of the state you look at, the data upon which decision-making relies is failing. The government can no longer see where it’s going. The ONS, whose job it is to record what’s actually happening, insists it is getting its act together. Around 100 people have been freed up from unnecessary data analysis to focus on the economic fundamentals. Health and crime data is being scaled back, with civil servants claiming the government needs to pay more if it wants the figures.
The ONS says it aims to copy airlines and the nuclear industry on transparency. It wants to create what’s known as ‘psychological safety’, where errors can be raised internally without fear of career repercussions.
Aside from cultural changes, another big shift across government is encouraging a greater reliance on ‘administrative’ as opposed to survey data. That is data collected for another purpose: a database of people with a driving licence, for example. This administrative data can be more accurate but it is not without risk: using it means trusting non-statisticians. The £3 billion revision to borrowing figures was caused by HMRC reeling in more tax but forgetting to pass that information on to the ONS. Until the state remembers how to count, and agrees on the best methods for doing so, Britain will remain a country governed by guesswork.
An ONS spokesperson said:
Under its new leadership the ONS is executing with urgency an improvement plan for economic statistics and focusing resources firmly on the key outputs. We recognise fully the need to rebuild trust and will continue to be open and transparent as we report on our plan and in dealing with issues as they emerge. We have also just announced that we will work with a number of external bodies to assure better the quality of data they supply to us.
Watch Michael Simmons’s new show, Reality Check, which makes the data-driven case on the stories behind the headlines
Comments