How the ONS rewrote Britain’s productivity statistics, and why the government’s economic strategy has a rather awkward problem.
Nine years ago I accidentally lied to the Office for National Statistics.
I was taking part in the Labour Force Survey. My working hours varied enormously, the questionnaire wasn’t very good at coping with this, and eventually I got bored with repeatedly explaining it. So when they asked whether my hours had changed, I started saying no.
In 2017 I wrote a blog arguing that this was a problem. The way we measured work hadn’t kept pace with the way people actually worked.
It turns out I had a point.
Last month, the Office for National Statistics published a paper proposing substantial revisions to the way it measures productivity. This might not sound particularly exciting, but the consequences are remarkable. It turns out that for much of the last two decades, British workers were considerably more productive than we thought.
Between 2009 and 2019, productivity grew by an average of 1.3% a year, rather than the 0.7% previously reported. That’s almost twice as fast.
The reason isn’t that we’ve suddenly discovered Britain was producing more goods and services. It’s that we’ve discovered people were working fewer hours than we thought.
Which is pretty much what I was complaining about nine years ago.
The productivity puzzle
Productivity is one of those economic concepts that sounds complicated but is actually quite straightforward. It measures how much we produce for each hour of work. If a factory produces 100 widgets in 100 hours, productivity is one widget per hour. If it buys a new machine and produces 150 widgets in the same time, productivity has increased by 50%.
This matters because, in the long run, increasing productivity is one of the principal ways economies become wealthier. If we can produce more goods and services with the same amount of labour, there is more wealth available to pay higher wages, increase profits, reduce prices or fund public services.
“Productivity isn’t everything, but, in the long run, it is almost everything. A country’s ability to improve its standard of living over time depends almost entirely on its ability to raise its output per worker.” Paul Krugman Nobel Prize for Economics
Before the financial crisis, Britain had a reasonably respectable record of improving productivity. Between 1997 and 2007, output per hour increased by an average of 2.1% a year. Then something apparently went badly wrong. Between 2009 and 2019, productivity growth averaged just 0.7% annually, and economists have spent much of the last two decades trying to explain what happened.
There have been plenty of theories. British businesses don’t invest enough, our education system doesn’t produce workers with the right skills, our infrastructure is inadequate, and British managers aren’t as good as their counterparts elsewhere. Some economists have blamed the growth of low-productivity service industries, while others have pointed to the financial crisis and the subsequent weakness of investment. I pointed out repeatedly that there are lots of customers who prefer inefficient small batch products over efficient mass production; craft gin, real ale, artisanal bakers, farmhouse cheeses, handspun fabrics, bespoke suits.
There is probably some truth in several of these explanations, but none has provided a particularly convincing account of why productivity growth appeared to collapse so dramatically. The productivity puzzle became one of the defining economic problems of modern Britain, with successive governments developing policies intended to solve it and economists producing countless papers explaining why it had happened.
The problem is that a substantial part of the puzzle now appears to have been caused by the way we measured working hours.
Under the ONS’s proposed new methodology, productivity growth between 1997 and 2007 was 2.0% rather than 2.1%, which is hardly a dramatic revision. But between 2009 and 2019, the revised figure is 1.3% rather than 0.7%. The cumulative effect is substantial. Under the existing methodology, productivity in 2024 was 34.0% higher than in 1997. Under the new methodology, it was 40.7% higher.
The productivity slowdown hasn’t disappeared, and growth of 1.3% is still considerably worse than the 2.0% achieved before the financial crisis. Nevertheless, it is a rather different problem from the one economists and politicians have been discussing for the last twenty years.
What makes the revision particularly interesting is that the ONS hasn’t discovered that we were producing more goods and services than previously thought. The estimates of economic output haven’t changed as a result of the new productivity methodology. What has changed is the estimate of how many hours people were working to produce that output.
The revised figures also make relatively little difference to productivity measured per worker or per job. The substantial difference is in productivity per hour, because the ONS now believes that average working hours continued to decline after the financial crisis, rather than remaining broadly stable as its previous estimates suggested.
In other words, the statistics were giving us a misleading picture of how much work people were actually doing.
The Labour Force Survey
When I wrote my original article in March 2017, I was setting up Durham Distillery, having spent most of my previous working life running organisations in the public sector. I had been randomly selected to participate in the Labour Force Survey, which was then, and remains, one of the principal sources of information about employment and unemployment in Britain.
I wasn’t particularly impressed by the experience. The questionnaire was enormously long, the interviews took a considerable amount of time, and I had to repeat the exercise several times. The questions seemed to have been designed around a rather conventional understanding of employment, in which people worked regular hours for a single employer and could provide reasonably straightforward answers about what they had been doing during the previous week.
Unfortunately, my working life wasn’t remotely like that. Some weeks I was travelling around Scotland visiting distilleries, looking at commercial properties and trying to establish a new business. Other weeks I was waiting for HMRC to send me a letter, and there was relatively little I could usefully do until it arrived. I was still working on establishing the business, but the amount of time I spent doing anything that could reasonably be described as work varied enormously.
The questionnaire wasn’t particularly good at accommodating this, and explaining the situation became increasingly tedious. Eventually, I started telling the interviewer that my hours were the same as the previous week, whether that was true or not. I wasn’t attempting to manipulate the unemployment statistics, I simply wanted to finish the interview.
As I explained in my original article, I had encountered similar problems with other government surveys. I once became sufficiently irritated with a manufacturing survey to tell them that all my employees were goats, which I suspect did relatively little to improve the quality of Britain’s industrial statistics.
My concern wasn’t simply that the questionnaire was irritating. It was that the Labour Force Survey seemed increasingly ill-equipped to describe the way people actually worked. By 2017, self-employment had increased substantially, zero-hours contracts had become commonplace, and the gig economy was expanding rapidly. Increasing numbers of people were combining several jobs, working irregular hours or making themselves available for work without necessarily being paid for all the time they were available.
Consider someone working as a delivery driver. They might spend forty hours a week available to accept deliveries but only receive enough work to occupy twenty of those hours. Depending on how they interpreted the questions, they might report themselves as working full-time, even though they were only earning money for half that time.
Someone on a zero-hours contract might work forty hours one week and ten the next, while someone who had become self-employed because they couldn’t find conventional employment might be working very long hours trying to establish a business without earning much money at all.
All of these people could legitimately describe themselves as employed, but that didn’t necessarily mean they had enough paid work to earn a living.
In my 2017 article, Unemployments Rising in the Chigley End of Town, I argued that the distinction between being employed, actually working and being available for work was becoming increasingly important. I also suspected that the statistics were overstating employment and understating the extent of underemployment, particularly among people who were technically self-employed but had insufficient work.
Nine years later, the ONS has effectively acknowledged an important part of that criticism. I was right.
Its new methodology combines information from household surveys, employer surveys and administrative records, rather than relying so heavily on the Labour Force Survey’s estimates of working hours. It also makes explicit adjustments for annual leave, sickness, bank holidays, overtime and other factors that affect the amount of time people actually spend working.
The result is a rather different picture of working hours, particularly in the period following the financial crisis. Under the previous methodology, the increase in total hours worked was driven almost entirely by an increase in the number of jobs, with average hours per job contributing relatively little. The revised figures suggest that average hours continued to decline, partly offsetting the increase in employment.
To be fair to the ONS, this doesn’t prove that everything I wrote in 2017 was correct. The new figures concern the measurement of working hours rather than simply the number of people employed, and they don’t establish that gig economy workers were routinely being incorrectly classified as full-time employees.
Nor has the ONS suddenly discovered the problem. Concerns about the reliability of the Labour Force Survey have been growing for some time, particularly as response rates have fallen, and the organisation has been working on alternative measures for several years.
Nevertheless, the underlying point I made in 2017 was that our system for measuring employment and working hours hadn’t kept pace with changes in the way people worked. The ONS has now produced evidence that the way it measured working hours was sufficiently inaccurate to change our understanding of one of Britain’s most important economic problems.
I don’t suppose I’ll be receiving a letter of apology, but I am rather pleased with myself.
The economy isn’t behaving as expected
There is another reason why these revisions are particularly interesting. Productivity is at the centre of the government’s economic strategy, and some of the latest economic statistics suggest that something rather unusual may be happening.
For the last couple of years, economic commentators have been remarkably pessimistic about Britain’s prospects. There have been repeated predictions of stagnation, warnings about the consequences of higher taxes and employment costs, and considerable concern about the impact of international uncertainty on business confidence.
The problem is that the economy has stubbornly refused to perform as badly as the forecasts suggested.
| Indicator | Latest figure |
|---|---|
| GDP growth, Q1 2026 | +0.6% |
| GDP growth, Q2 2026 | +0.5% |
| GDP growth, year to Q2 | +1.4% |
| GDP per head, year to Q2 | +1.2% |
| Business investment, Q2 | +1.8% |
| Business investment, annual | +5.2% |
In November 2023, the Office for Budget Responsibility forecast that the economy would grow by just 0.7% in 2024. It revised that figure to 0.8% in March 2024, but the economy actually grew by 1.1%. In March 2025, the OBR forecast growth of 1.0% for that year, whereas the latest figures suggest that the economy grew by 1.2%.
This doesn’t mean that the forecasts have always been too pessimistic. In October 2024, the OBR was forecasting growth of 2.0% for 2025, which turned out to be considerably too optimistic. Nevertheless, the economy has repeatedly performed better than some of the gloomier predictions suggested, and the latest figures for 2026 are also reasonably encouraging.
GDP increased by 0.6% in the first quarter of this year and another 0.5% in the second. By the end of June, the economy was 1.4% larger than a year earlier, while GDP per head had increased by 1.2%. The latest monthly figures also suggest that the third quarter started reasonably strongly, with GDP increasing by 0.4% in July.
| Year | Forecast | Actual performance | Assessment |
|---|---|---|---|
| 2024 | OBR November 2023: 0.7% | 1.1% | Beat forecast |
| 2024 | OBR March 2024: 0.8% | 1.1% | Beat forecast |
| 2025 | OBR March 2025: 1.0% | 1.2% | Beat forecast |
| 2025 | OBR October 2024: 2.0% | 1.2% | Missed forecast |
| 2026 | OBR March 2026: 1.1% | 0.6% in Q1, 0.5% in Q2 | Strong first half; full-year result unknown |
None of this amounts to an economic miracle. Britain remains a relatively slow-growing economy, and the recovery in living standards since the financial crisis has been disappointing. Nevertheless, it is difficult to reconcile the actual figures with some of the more apocalyptic commentary about the state of the British economy. We are now one of the fastest growing major economies.
Perhaps more interestingly, business investment has been surprisingly buoyant. In the second quarter of 2026, investment increased by 1.8%, taking it to 5.2% above its level a year earlier. This is particularly striking given the uncertainty surrounding international trade, energy prices, interest rates and the wider global economy.
The largest contribution to the quarterly increase came from investment in buildings and structures, so it would be premature to conclude that businesses are suddenly spending enormous amounts of money on artificial intelligence. Nevertheless, there is some evidence of increasing interest in automation and AI, and the Bank of England’s business intelligence has identified tentative improvements in investment intentions.
There are still plenty of reasons for caution. Borrowing costs remain relatively high, consumer confidence is subdued, and businesses continue to report uncertainty about demand. But investment is one of the principal ways productivity improves, and the latest figures suggest that British businesses are continuing to invest despite these difficulties.
There is another potentially encouraging development, because wages are also continuing to rise. In the three months to July 2026, average regular earnings increased by 3.5% compared with the previous year, while total earnings, including bonuses, increased by 3.9%. Inflation has absorbed much of these increases, but real regular earnings still rose by 0.6% using the CPIH measure of inflation, or 0.8% using CPI.
These aren’t spectacular improvements, but they do represent a modest increase in purchasing power, and in normal circumstances rising wages might suggest that businesses were experiencing increasing demand for labour. When employers need more workers, they have to compete to recruit and retain them, which tends to put upward pressure on pay.
The difficulty is that the employment statistics appear to be telling us something rather different.
| Indicator | Position |
|---|---|
| Unemployment | 4.9% |
| Employment rate | 75.1% |
| Payrolled employment | Down 0.5% annually |
| Job vacancies | 702,000 |
| Regular wage growth | Approximately 3.5% |
| Consumer price inflation | 3.1% |
Unemployment has increased to 4.9%, around 0.2 percentage points higher than a year earlier, while the latest payroll figures suggest that the number of employees fell by 0.5% in the year to August. Job vacancies have also fallen to 702,000, which, apart from the disruption caused by the pandemic, is a level last seen in 2014.
So we have an economy in which output is increasing, business investment is rising and wages are growing, but demand for labour appears to be weakening. This is an interesting combination, because we would normally expect stronger economic activity to increase employment rather than reduce it.
There are several possible explanations, and the wage figures are probably a useful place to start. The headline increase in wages may be giving a rather misleading impression of the strength of demand for labour. Some of the increase reflects earlier pay settlements and attempts to compensate employees for inflation, rather than employers competing to recruit additional workers. The timings of public sector pay settlements has inflated some of the numbers. Real wage growth is also relatively modest, particularly given the substantial loss of purchasing power experienced during the preceding period of high inflation.
Businesses may also be reluctant to recruit because of higher employment costs, uncertainty about future demand and the continuing effects of relatively high interest rates. None of these explanations requires any improvement in productivity, and it would be a mistake to assume that the recent increase in unemployment has been caused by automation or artificial intelligence. This is partly the result of Government policies – increasing labour costs and reducing immigration are all designed to force businesses to move away from low wage business models and invest. It is striking that immigration is falling sharply and unemployment is rising.
Nevertheless, there is another possibility worth considering, particularly in the light of the latest investment figures. Businesses may be beginning to increase output without needing to employ more people, which is, after all, precisely what increasing productivity is supposed to achieve.
Be careful what you wish for
The government’s economic strategy places considerable emphasis on improving productivity through higher investment, better infrastructure, technological innovation and the adoption of artificial intelligence. The argument is that more productive businesses will generate faster economic growth, allowing wages and living standards to increase.
There is nothing particularly controversial about this. Over the long term, productivity growth is one of the principal determinants of rising prosperity, and Britain’s relatively disappointing performance since the financial crisis has undoubtedly contributed to the weakness of real wage growth.
But there is a potential contradiction in the government’s approach, because the whole point of increasing productivity is that we can produce more goods and services with fewer hours of labour.
Consider a business producing £100 million worth of goods and services using five million hours of labour. If it increases productivity by 10%, it can produce the same output using approximately 9% fewer hours. Alternatively, it could maintain its existing workforce and increase output by 10%, assuming there is sufficient demand for the additional production.
Both outcomes represent an improvement in productivity, but they could have very different consequences for employment.
If the business uses its improved productivity to expand, it might create additional jobs, increase wages and generate more economic activity elsewhere. If it uses the improvement to reduce costs, it might employ fewer people while maintaining the same output. It could also reduce working hours without reducing pay, allowing employees to enjoy more leisure time while maintaining their living standards.
Which of these outcomes occurs depends on rather more than the productivity figures themselves. It depends on demand for the business’s products, the competitive environment, the availability of workers, the way businesses distribute the benefits of increased productivity and the wider condition of the economy.
Historically, technological improvements have often displaced workers in particular industries while creating employment elsewhere. The mechanisation of agriculture dramatically reduced the number of people needed to produce food, but industrialisation created new opportunities. More recently, computers eliminated enormous amounts of clerical and administrative work while creating entirely new industries and occupations.
There is no particular reason to assume that artificial intelligence will necessarily be different, although there are some reasons to think that the transition could be unusually disruptive. Unlike many previous technological developments, AI has the potential to automate tasks undertaken by relatively well-paid professional employees, including some of the work currently performed by accountants, lawyers, software developers and administrative staff.
It may also allow businesses to expand their output without increasing employment at anything like the rate they would previously have required. A business that once needed ten additional employees to handle an increase in demand might discover that it can achieve much the same result by investing in new technology and employing only two or three.
This could be excellent news for the business and for its remaining employees, particularly if some of the productivity gains are reflected in higher wages. It could also benefit consumers through lower prices and the wider economy through increased investment and profits.
But it would be rather less encouraging for the people who would previously have been recruited to undertake the additional work.
There is also an interesting statistical consequence. If businesses use new technology to reduce employment while increasing the wages of the employees they retain, average earnings could increase at the same time as unemployment rises. The economy would become more productive, wages would increase and GDP might continue to grow, but a growing number of people could find themselves unable to obtain sufficient paid work.
| Indicator | Latest evidence | Possible interpretation |
|---|---|---|
| GDP | +0.5% in Q2 2026, following +0.6% in Q1 | Economic activity is expanding |
| Business investment | +5.2% annually | Businesses are investing despite uncertainty |
| Nominal wages | +3.5% annually | Pay continues to increase |
| Real wages | +0.6% annually (CPIH-adjusted) | Living standards are improving modestly |
| Unemployment | 4.9%, up 0.2 percentage points annually | Labour market weakening |
| Payrolled employment | Down 0.5% annually | Businesses employing fewer people |
| Vacancies | 702,000 | Recruitment demand remains subdued |
This isn’t necessarily what is happening in Britain at the moment. The evidence isn’t strong enough to establish a connection between increasing investment, improving productivity and the recent weakening of the labour market. But the combination of relatively resilient economic growth, increasing business investment, rising average wages and falling employment is sufficiently unusual to deserve closer examination.
It also raises a more fundamental question about what we actually want from economic growth.
There is a tendency to treat increasing GDP and productivity as objectives in themselves, but presumably the reason we want a more productive economy is that it should allow people to enjoy higher living standards. If technological improvements mean that we can produce the same amount of goods and services while working fewer hours, that ought to be a good thing.
Indeed, one of the more remarkable features of economic development over the last two centuries has been the enormous reduction in the amount of time people spend working. The working week has shortened, paid holidays have become commonplace, and retirement has become a realistic prospect for large sections of the population.
There is no obvious reason why further increases in productivity shouldn’t allow this process to continue. If we can produce more with less labour, we could choose to distribute some of the benefits through shorter working hours rather than simply increasing output.
The difficulty is that shorter working hours are only an improvement in living standards if people can afford them. Someone who moves from working forty hours to thirty-two while maintaining their earnings is probably better off. Someone whose hours are reduced from forty to twenty because their employer no longer needs them may be considerably worse off, even if the business has become more productive.
Similarly, a society in which most people work fewer hours while maintaining their incomes might be considerably wealthier in any meaningful sense than one in which a smaller number of highly productive workers earn substantial salaries while everyone else struggles to find employment.
The distribution of the benefits of productivity growth therefore matters at least as much as the rate of growth itself. It is perfectly possible for an economy to become more productive without everyone becoming better off, particularly if the benefits are concentrated among business owners and those employees whose skills remain in demand.
And this is where the argument returns to the Labour Force Survey.
Nine years later
When I wrote my original article in 2017, my concern was that the employment statistics were increasingly poor at describing the circumstances of people whose working lives didn’t fit the traditional pattern of regular, paid employment.
Someone working in the gig economy might be technically employed while spending much of their time waiting for work. Someone on a zero-hours contract might have a job but insufficient hours to earn a living. Someone who had become self-employed because they couldn’t find conventional employment might be counted as employed even though their earnings were negligible.
All of these people could appear in the employment statistics, but the fact that they were employed told us relatively little about whether they had enough work or sufficient income.
The ONS’s latest productivity revisions don’t prove every aspect of the argument I made nine years ago, but they do demonstrate how substantially our understanding of the economy can change when we measure working hours differently. A problem that economists have spent much of the last two decades trying to explain now appears considerably smaller, not because the economy has changed, but because the way we measure it has improved.
At the same time, the government is pursuing an economic strategy intended to increase productivity through investment, automation and artificial intelligence. There are some encouraging signs that the economy is performing better than the more pessimistic forecasts suggested, with investment increasing and real wages beginning to recover.
But unemployment is also rising, and demand for labour appears to be weakening, raising interesting questions about the consequences of improving productivity in an economy where most people still depend on selling their labour to earn a living.
If the government’s strategy succeeds, we could find ourselves producing considerably more goods and services while working fewer hours. In principle, that ought to be an excellent outcome, allowing people to enjoy higher incomes, more leisure time and better living standards. But there is nothing inevitable about the benefits being distributed in that way. We could equally find ourselves with an economy in which ahighly productive workers earn more money, while increasing numbers of people struggle to find enough paid work.
Which is why the distinction I was making in 2017 remains important. Knowing how many people are employed is useful, but it doesn’t necessarily tell us how much work they are doing, how much they are earning or whether they have sufficient income to live on.
Nine years ago, I argued that the official statistics weren’t adequately capturing these differences. Last month, the ONS published evidence that its estimates of working hours had been sufficiently inaccurate to distort our understanding of Britain’s productivity performance.
I am rather pleased to have been proved right, even if it has taken nine years.
But there is a more important point. We have spent much of the last two decades worrying that Britain wasn’t producing enough for every hour worked. We may now be entering a period in which the bigger problem is ensuring that everyone has enough paid work, or receives a sufficient share of the benefits of higher productivity, to maintain a decent standard of living.
The government wants us to produce more while working fewer hours, which sounds like a perfectly sensible ambition. The difficulty is that most of us still need to earn enough money to pay the bills, and producing more with fewer hours of labour doesn’t necessarily mean that everyone will be better off.
It would be rather unfortunate if we finally solved Britain’s productivity puzzle, only to discover that we had created a rather different problem in its place.