When I took over the Labour Market Analysis team at the Office for National Statistics (ONS) around 2008, I remember asking colleagues why we were producing a particular article.
The answer was simple. Because it had to go in the quarterly publication.
But I had a rather different question. Who was actually going to read it?
We were producing enormous publications, almost like books, without necessarily asking whether the information was making any difference.
And that mattered because taxpayers were funding the work.
I subsequently helped lead a programme within the ONS looking at the impact of official statistics, something I’ve always considered fundamental to their purpose.
Nearly two decades later, I still believe we’re not asking that question often enough.
For me, statistics have two fundamental purposes: to help us make better-informed decisions and to give the public the evidence to hold those in power to account.
Everything else should flow from those principles.
Producing Statistics Isn’t Enough
During my career, I’ve sat in meetings in Whitehall with senior ministers who struggled to grasp the data being discussed. I’ve also worked in the BBC newsroom and seen how journalists can struggle to interpret numbers.
And these are people making decisions or communicating information that affects millions of lives.
That illustrates a fundamental problem. Producing accurate statistics is only half the job.
If the people making decisions cannot understand them, or the public cannot interpret what they’re being told, the value of those statistics is diminished.
We don’t necessarily need more statistics. We need statistics that people can understand, trust and use.
What Happens When the Data Isn’t Reliable?
There’s another challenge, one that goes to the heart of public confidence in official statistics.
The quality of the underlying data.
Take Britain’s Labour Force Survey, which underpins our estimates of employment, unemployment and economic inactivity.
It’s a survey I worked with extensively during my time at the ONS.
Yet the latest published figures for April to June 2026 show a response rate in Great Britain of just 22.8%, excluding imputed cases.
That’s fewer than one in four eligible households responding across the survey waves.
Statisticians can use weighting and other techniques to compensate for missing responses. But those adjustments cannot automatically eliminate the risk of bias, particularly if the people who don’t respond differ systematically from those who do.
A low response rate doesn’t make every estimate wrong. It does, however, make understanding the limitations increasingly important.
That’s why, when examining changes in employee numbers, I now place considerable weight on administrative data collected through HMRC’s PAYE system.
These records provide a much more comprehensive measure of payrolled employment than a household survey. They don’t replace the Labour Force Survey for measuring unemployment, self-employment or economic inactivity.
Indeed, recent ONS research supports PAYE data as the most reliable current measure of employee numbers.
If we’re going to make important decisions using statistics, we must have confidence in the evidence underpinning them.
Can Government Statistics Be Truly Independent?
Then there’s the question of independence.
Britain has a Government Statistical Service, with statisticians working across different government departments.
These statisticians operate under professional standards designed to protect their independence, and many do excellent work.
But I’ve long questioned whether the current arrangements go far enough.
Should the same government departments responsible for delivering policies also be responsible for producing and publishing the statistics used to judge their performance?
Consider the controversy surrounding crime statistics broken down by nationality.
In June 2026, the Information Commissioner’s Office ruled that the Ministry of Justice had failed to justify withholding requested conviction figures under the cost exemption it cited. The Commissioner also found that the personal-information exemption did not apply to counts of five or more.
The Ministry of Justice has appealed that decision, with the government confirming in September that proceedings remained ongoing.
There are legitimate considerations around privacy, disclosure risks and data quality. But this case raises wider questions about public access to politically sensitive information.
Statistics funded by taxpayers should be available for public scrutiny wherever lawful and appropriate.
I believe there’s a strong case for greater separation between government departments responsible for policy and the organisations responsible for publishing the evidence used to assess it.
A more centralised, visibly independent statistical system deserves serious consideration.
Not because government statisticians are necessarily politically influenced, but because the independence of official statistics must be demonstrable, not simply assumed.
Statistics Must Serve the Public
We spend substantial sums of public money collecting, analysing and publishing statistics.
That investment should deliver something meaningful.
We need reliable data, transparent methods, genuinely independent publication and a willingness to acknowledge when things go wrong.
But above all, we need to remember why we’re producing statistics in the first place.
Not to fill government reports. Not to provide convenient headlines for politicians. And certainly not to create the illusion that complex political arguments can be settled with a single percentage.
Statistics should help us make better decisions and hold those in power to account.
If we’re not achieving those two things, something needs to change.
Discussing This at the Battle of Ideas
Tomorrow, Saturday 17 October, I’ll be speaking at the Battle of Ideas Festival in Westminster, joining a panel discussing Lies, damned lies… let’s talk about stats.
We’ll be examining trust in official statistics, political manipulation of figures, independence and whether the evidence we rely upon is still fit for purpose.
It’s a debate that matters far beyond the world of statistics.
Because without reliable evidence, we’re left with little more than competing opinions.
✍️ Jamie Jenkins
Stats Jamie | Stats, Facts & Opinions
If you found this analysis useful, please share it and subscribe for more.
📲 Follow me here for more daily updates:
A little while ago you did a report with a comparison of the education standards between England and Wales and it reminded me that many politicians don't want statistics which go against their agenda. Katherine Birbalsingh did a piece about the row which erupted when Michael Gove started his education reforms. After your post I looked it up and it seemed to me that Gove has been vindicated in his reforms. England is now in the OECD top ten for 15 year olds and there have even been headlines in the Guardian no less to the effect that England has bucked the international trend. England's education rating have been rising at a time when most of the world's has been falling. So given that Gove's reforms have succeeded why is the present government changing school education again. I can see that there might be scope for further improvement but I don't think that's why they are doing it. I think they are doing it for ideological reasons and not on real evidence.
Another example is Net Zero. We know that the drive to electrification and renewables is putting up electricity bills but the government puts the blame elsewhere. They are ignoring real data. They don't want the truth to get in the way of their preferred policy.
I completely agree. I know someone who works for an agency which collects and compiles statistics about education. I've asked him who looks at the data and he doesn't know. I'm sure that the people who look at the data are often only looking for statistics which support the current agenda. I'm currently reading a book called "Wrong number" by Aaron Brown who is a statistician and poker player. He describes the many ways that published statistics are misinterpreted or just plain wrong. It's a very recent book and the examples are mostly about the US but is fascinating anyway. I got Aaron's name when he appeared on Triggernometry with Konstantin Kisin.