An intervention for UK statistics - but will it be enough?
The ONS must change to provide the good data that government and the public need
April Fools’ Day is not the most fortunate date for the birth of an organisation wanting to be taken seriously.
Nevertheless, it was on this date in 1996 that the UK’s Office for National Statistics (ONS) was brought into being by the merger of the existing bodies which had, in various guises, been collecting statistics for the nation since 1837.
To mark the occasion, staff planted a tree in the grounds of the organisation’s Southport headquarters, the unattractively-named Smedley Hydro building which had once been a Victorian health spa and, later, the operational centre for the population register and ID card system which ran throughout World War II.
I do wonder if that tree is still there. By now it would be tall, at 29 years old, and firmly rooted in place. Each year its bark will have thickened against the sea wind and new growth will have expanded its canopy, keeping the rest of the plant in protective shade.
The ONS itself, meanwhile, is looking less robust as it comes up to its 30th birthday. Bearing more resemblance to the human cohort of its same age, it is neither fully Millennial nor Gen-Z and youthful self-assurance is wearing off to expose a core of anxiety. For the most part it resembles a successful adult and yet at times it can, directionally and financially, feel like two kids in a trench coat.
Last week, the Devereux Review, which had been commissioned to look into ONS’s performance, delivered a cutting assessment of an organisation with “deep-seated” problems with its governance and operations. While it acknowledged that some recent problems have resulted from it being harder to collect certain types of data in the modern age, the ONS itself was ultimately the one held responsible, with Devereux writing that:
“most of the well-publicised problems with core economic statistics are the consequence of ONS’s own performance. And that performance is affected by certain cultural issues.”
So how bad really are the problems, and is this the beginning of a turnaround?
High-impact mistakes and slow decline
It’s important to realise that statistics are not produced just to provide colour and shading for people wanting to broadly understand a topic area. The exact, often highly specific numbers count - and they need to be precise.
Economic statistics, for example, directly feed into processes for making decisions about government spending and investment, not to mention being crucially important to the Bank of England and the finance and business sectors. Yet the ONS’s core economic statistics have been borderline unusable in recent years, due to a combination of low accuracy and a high rate of errors.
In early 2025, the ONS had to apologise after noticing mistakes in two of its main price indexes that mean it has likely miscalculated economic growth over the past two years. Not even two months later it had to admit that it had overestimated inflation and had to revise the highly influential published figure, amid a chorus of criticism from economists.
The Labour Force Survey, which is the only means of capturing what people are doing - or not doing - in terms of livelihoods across the whole of the economy has been allowed to crumble into inaccuracy. The problem is that not enough people are responding to the survey to make it reliable - a mere 15% of people asked in 2023. Some analysts claim that the survey may have led to over a million workers being misclassified as not working (see chart below) which, if true, could mean years of policy directed at getting people back to work after the pandemic may have been misguided.
Migration statistics have also had to be repeatedly revised in recent years, as the ONS was forced by the pandemic to scrap its old survey-based method for counting migrants but without having an alternative method fully in place. Refining the new method has meant many of the migration figures published in recent years have been concluded to be wrong, and then revised, only a few months later.
The ONS initially estimated net migration in 2022 at around 606,000 people which was already the highest on record. Later the same year this was revised to 764,000 and ultimately, after four revisions, the current figure stands at 872,000 or 44% higher than the original one. The new method also meant EU migration between 2012 and 2018 had to be quietly revised upwards by 75%. We were planning Brexit with a very poor understanding of how many EU nationals were in the UK.
ONS also damaged its own credibility by introducing a gender identity question to the 2021 England and Wales census which, it later became clear, was misunderstood by a lot of people. Newham in East London became the area with the largest transgender population simply by virtue of it having the most beginner-level English speakers. A defensive attitude when the figures were initially questioned is also considered not to have helped.
Problems with morale at the organisation have also been well-documented. Some say the joi de vivre was lost for good already in 2006 when the ONS headquarters relocated to Newport in South Wales, with an outpost in Titchfield. Recent data shows a story of employees having felt more confident in their leadership during the pandemic and this trust falling away almost completely in recent years.
The turnaround starts now
I’m not here to heap criticism on the ONS. I’ve worked closely with them for many years and admire the work of all the individual colleagues I’ve encountered there. I back them completely - but it’s also clear that all is not well.
The Devereux review seems honest and sensible, and its recommendations - which include splitting the current job of National Statistician into one role of running the business and another focused on conceptual leadership - will surely make a welcome difference.
What I wonder is whether they go far enough.
The ONS has already accepted Devereux’s recommendations and has committed a further £10 million towards fixing economic and population statistics. And they’ve made a point of thanking the vital frontline workers who, for quite measly pay, face the increasingly hard task of persuading members of the public to answer intrusive questions for a distant and vague public good.
Considering the ONS’s overall budget of around £360 million - £63 million of which went towards carrying out surveys in 2023/24 - another £10 million may not be enough of a lifeline. If the Financial Times won’t mind me sharing it from behind a paywall, take a look at this shocking chart which shows the response rate to the main household survey in the UK, EU countries, and the US. Yes, some other countries are seeing response rates decline, but the UK is in a special category of panic stations.
In 2023, the response rate to the UK’s main household survey was 15%, while in France it was 81%, in the USA it was 68% and even in Denmark - which had the worst rate of EU countries - it was still 37% (these latest figures are not shown in the chart above). Something is going on in - I wish I could say what - and, until that is identified and fixed, sticking plasters are only going to do so much.
I’d love it if someone could tell me if that ONS birthday tree is still there and how it looks. I think the office is now used by NHS Digital.
And I will be watching closely to see what happens now in light of the Devereux review (and last year’s Lievesley Review which looked at similar issues). Will a bit of pruning and care be enough, or do we need root and branch reform?