Welfare and the unexpected tyranny of government statistics

How the government’s obsession with tiny numbers has crippled decision making (and undermined the ONS).

Don’t forget to check out The Abundance Agenda, my podcast with James O’Malley, which looked at the crisis at the ONS in a recent episode.

---

There is no ‘correct’ number for ‘GDP’ or ‘unemployment rate’ or even ‘the population of Britain’. There are estimates of these things made by skilled people, but they are impossible to know or measure beyond a certain level of accuracy, and even their definitions are ultimately subjective. Should crime count toward productivity? Is someone unemployed or choosing to be inactive? What you see is your brain’s interpretation of the world, an extrapolation from incomplete and messy data; and the same applies to the British economy. Our exact GDP is not just unknown, but unknowable, so the false precision with which it is discussed is deeply misleading.

GDP fell ‘more than expected’ in April, according to Reuters. They had forecast a 0.1% decline, but the ‘real’ drop - and by real they meant an early estimate from the ONS - was 0.3%. Dramatic stuff, which neglected to mention that the estimate has a typical error of 0.2 percentage points¹. Presumably because it would destroy the entire premise of the story.

Still, this is just journalism. No government would be stupid enough to base serious decisions about the future of the country on statistical noise, right? Right?

Photo by Maxim Hopman on Unsplash

Last week, the government passed the Universal Credit and Personal Independence Payment Bill, which we’re calling the ‘Welfare Bill’ because who has the time to type all that. It was supposed to save around £5bn, in order to preserve the government’s fiscal headroom. That headroom is the difference between the amount the government plans to spend in the future, and a maximum ‘safe’ number set by the government’s fiscal rules, which change regularly in a sort of elaborate courting ritual² between successive Chancellors and the markets who buy government debt.

The government spends around £1,300bn each year. Its headroom at the Spring statement was supposedly £9.9bn, which sounds like a lot but is basically a rounding error - roughly 0.6% of the budget. Let’s bring this down to human scale: say your budget is £2,600 per month. Your fiscal headroom would be £18.80. If your energy bill goes up a bit, or you have a surprise expense, or you make even a small error estimating your grocery spending, your budget is blown out of the water and you’re in a debt spiral. That sounds like a scary and unsustainable way to run a budget, because it is.

It gets worse. Like all the other numbers we’re talking about, that headroom is an estimate - a forecast built on a whole load of other forecasts, including GDP, which we can only measure to a certain degree of accuracy, and which forecasters typically overestimate by about 0.9 percentage points over a 5-year forecast. That’s still far more accurate than estimates of future government income and spending, which are basically always wrong.

All of which is to say that guessing headroom to even the nearest five or ten billion would be a major achievement, a fiscal 147 in Snooker terms. Insisting that it is precisely £9.9bn is the statistical equivalent of your Dad saying he could take a point off Serena Williams in a tennis match³, the hubristic prediction of a charlatan, drunk in charge of a spreadsheet. To then set budgets that require that level of precision to avoid things like a national financial meltdown from happening is a whole new level of bonkers.

The Welfare Bill is a bad bill. It’s bad on so many levels that it’s like a Russian doll of bad, or a particularly bad onion where each layer uncovers a new level of fetid badness. It’s bad regardless of your political beliefs. It’s obviously bad if you don’t want to save money on welfare, but even if you do it makes no serious attempt to address the reasons why welfare bills have been rising significantly since the pandemic. Professor Aditya Goenka, Chair in Economics at the University of Birmingham, put it better than I can:

“It did not address the fundamental issue of why an increasing share of the population is receiving these benefits: is it the case that increased illness payments are substituting for one of the least generous unemployment benefits amongst comparators; does it reflect the aging population who become disabled before they are eligible for National Insurance; are they a reflection of the mental and physical costs of Covid; or are they a reflection of the declining health outcomes in the UK population relative to G7 countries?”

All that boring systems thinking is left for others. Instead, the bill attempts to micromanage the budget to make just enough savings to meet the government’s fiscal rules. Since most of the cost-saving measures were gutted by MPs anyway it fails to do that; but even if it had come through Parliament unscathed, the flap of a gnat’s testicle could have wiped out those savings in a heartbeat.

Last week on the podcast we talked about the recent troubles at the Office for National Statistics. The ONS emerged from the dying days of the Tory government in 1996⁴, a legacy of John Major’s passion for open government. He saw open statistics as a vital component of a healthy democracy, a tool for administration but also a way for voters to hold governments to account. What Major couldn’t have appreciated at the time was how awfully successful this would become; how central his mathematical project would be to political discourse.

Thirty years later, Westminster coverage has degenerated into a perma-running soap opera of who’s up or down this week, and the data produced by the ONS has become a convenient way to keep score. If GDP is up 0.1% it’s a triumph; down 0.2% and it’s trouble for Rachel Reeves. In our desperation for drama, even the tiniest bits of noise are inflated to huge significance.

Stephen Bush has talked a lot in recent months about how the current generation of politicians and journalists seem heavily influenced by video game culture,⁵ to the point where they come to view these data products like the stats in ‘SimCountry⁶’. But there’s a fundamental problem with this mentality - computer games provide omniscience. The computer can accurately quantify every datum in a simulation and tell you precisely how many people live in your virtual city, how much money they earn, the exact employment rate, and anything else you care to know.

This data simply doesn’t exist for a real-life economy, there are no ‘correct’ numbers to be had. Instead we have imperfect estimates based on imperfect observations of messy systems.. They may be very good, but they are still only estimates, and in recent years there have been serious problems with several of them.

If you’re interested we delved into more detail in the podcast, but here’s a quick summary. Firstly, the job of the ONS is harder and getting harder still. To measure employment for example they need to conduct real-world surveys, visiting people in their homes to ask them questions about their current status, but in common with polling companies they are finding this more and more difficult. People in 2025 have video doorbells and are working on Zoom calls and are loathe to open the door to a random man in a suit. By 2023, response rates had fallen to a catastrophic 17%, rendering the data all but useless.

This is a mighty challenge, but it’s not the biggest cause of the organisation’s woes. The ONS has also been hamstrung by poor leadership. You don’t simply come up with an algorithm for inflation or GDP and then copy the number into a press release every month. These measurements require constant research, recalibration, maintenance and improvement. Senior leaders failed to understand that this was the primary purpose of the ONS, and failed to provide adequate attention to or investment in these tasks, distracted by shiny baubles elsewhere.

An example of this is the Producer Price Index, which (very roughly) looks at ‘factory gate’ or wholesale prices for goods and services, and which was being calculated by code which turned out not to do what was expected. That index is being repaired in the coming months, but its numbers were fed into the models that estimate GDP. It’s a similar story with trade statistics, where there were issues with the underlying software configuration. Staff repeatedly raised warnings, but senior leaders didn’t want to hear it⁷. Fortunately, that leadership is being replaced, but it will take probably a year or two to bring things back up to scratch.

But I think there’s another, more fundamental problem. We - the government but also journalists - have become obsessed with these numbers, attaching a weight to them that they - and the ONS - cannot bear. The publication of monthly GDP estimates to one decimal place implies a level of accuracy and precision that they just cannot provide, yet despite that we try to divine our future in their tiniest variation. Yes these statistics are important and they should be as accurate as possible, but the understanding we have built around them has become increasingly brittle and fragile.

As a thought experiment, there are some simple (if perhaps unworkable) ways to reduce the level of precision to something the maths can actually support: round estimates of GDP to the nearest 0.25%, or publish a range, or better yet resolve the whole thing to a simple traffic light system. Coding anything between -0.5 and +0.5 as amber would be a far better reflection of Britain’s stagnation in recent years than the pretence that +0.1 for two quarters is fine while -0.1 is a recession. Consumers of these numbers would be furious, but they should take a hard look in the mirror and ask: if their downstream systems, processes and calculations are so dependent on a level of precision that can’t actually be reached, then how good are they, really?

That is a defining question for the government. You cannot micromanage policy based on statistical noise. The more you try to do this, the more you’ll find that, in fact, the noise is managing you. If the government wants to actually govern, and not just endlessly tweak small numbers, it needs to restore its headroom to something sensible. That means it is going to have to make big decisions that it should have made a year ago, and it needs a clear, coherent vision to guide those decisions.

The good news - and I mean this unironically - is that there are no great options on the table: tax rises⁸, radical levels of building, gutting public services, taking on the gerontocracy or borrowing more… all of these carry a steep price, all could be politically fatal. This is brilliant if you’re in charge and should be incredibly exciting and liberating, because here’s the thing: when all the options are bad, there are no bad options. You’re free to just do whatever you think will make the biggest difference. Which is surely what you came into politics to do.

So the question then is: what does Keir Starmer’s government actually believe? Does it really have the appetite to make these big-picture decisions, or is it easier for Britain’s micromanagers-in-chief to just stare at the graph like amateur day-traders, tweaking numbers for the next four years?

---

Jabs for the rich, moralising for the poor

1. I’ve made a real effort to distinguish between percentages and percentage points in this article, which means I’ve almost certainly screwed up somewhere, and I look forward to the inevitable comments pointing this out.

2. I imagine something like a bee dance.

3. This example included in the sure knowledge that people (men) will turn up in my mentions claiming they could take a point of Serena Williams.

4. As mentioned on the pod, I was slightly shocked that anything much got done in 1996, and assumed such a technocratic policy was a New Labour innovation, so well done to Major on this at least.

5. I was going to include some examples, but Bluesky is nearly impossible to search effectively, so if anyone has some I’ll be happy to insert them.

6. I would play this game. I’ve tried Democracy but I find it a bit lacking in depth or the sense that the economy is a real living thing you. My favourite game of all time is the colonial-era Europa Universalis IV, which it’s fair to say would be a problematic influence for policy-makers.

7. To put it mildly - more on this on the pod.

8. I’m not a massive fan of tax rises for reasons I’ve set out previously, but if they are necessary they need to be accompanied with a clear vision for how that money will be spent and why it will be different from previous massive cash injections.