The Macroeconomic Policy Outlook: Q3 2026

On the whole, the 2020s has been a disastrous decade for UK productivity growth. But an under-discussed revival appears to be underway. While the much-maligned Labour Force Survey’s headline number is showing no such pickup, more trustworthy measures of productivity are growing more strongly. Administrative tax data suggests that output per hour grew by 1.1 per cent a year over the two years to Q2 2026 – not just a sharp turnaround from the 0.7 per cent a year fall in the two years before, but also faster than the 0.7 per cent growth rate seen in the late 2010s. It is also broadly in line with the OBR’s revised medium-term assumption for UK productivity growth, bringing the OBR in line with reality arguably for the first time ever. This edition of the Macroeconomic Policy Outlook asks whether the recent productivity pickup is genuine, and what might be driving it.

We find that the improvement is real and that it is broad: 12 of the UK’s 19 main sectors contributed to the pickup, so this is not one corner of the economy flattering the aggregate. Some have put the pickup down to low-productivity sectors like retail and hospitality shrinking, or to employers shedding low-paid jobs as employer National Insurance and the minimum wage rose. We find no support for either. Almost all of the pickup has come from faster productivity growth within sectors, not from the reallocation of hours between them. For example, the share of employees working in hospitality – often held up as a low-paid sector being hard hit by higher employment costs – is no lower today (6.6 per cent in Q2 2026) than it was in the late 2010s, and sits only 0.3 percentage points below its brief post-Covid peak (in Q4 2022). Digging down, there is also scant evidence that this within-sector growth is being driven by low-paid jobs simply falling away. Looking across around 400 occupations, growth in below-median-paid roles was 2 percentage points stronger in the year to April 2025 than growth in jobs above median pay. Whatever is raising productivity is doing it with the same workers, in much the same jobs, in much the same sectors. And nor can it be explained by capital deepening: investment has been flat at 19 per cent of GDP since 2023, and it would take a sustained or sharp rise in investment to move productivity this much.

So we have what looks like total factor productivity (TFP) growth, the secret sauce of economic growth that all but vanished after the financial crisis. It is probably too early to chalk up an apparent TFP revival to AI. Although nearly half of UK businesses are using AI in some form, it does not yet seem deeply embedded across the economy: only around one-in-twenty firms say they are using it extensively. For now, the revival of productivity looks as mysterious as its fall. But it looks genuine: workers across the economy are producing more, rather than the least productive parts of the economy simply shrinking.

Measured properly, productivity growth has picked up in the past two years

Labour productivity – the amount of output the average worker produces in an hour – is the keystone to economic growth and improving living standards. Strong productivity growth before the financial crisis delivered consistent real-wage gains for British workers, and its subsequent slowdown was the biggest contributor to the period of wage stagnation that followed. On the whole, the 2020s have been no better. Repeated shocks – Covid, Brexit and the war in Ukraine – have left Britain facing another level hit to productivity. In Q2 2026, the ONS’s headline measure of output per hour worked was around 5 per cent lower than the level implied by the OBR’s final pre-Covid forecast – equating to lost output of around £150 billion a year, or around £4,500 a year per worker.

But underneath that disappointing overall picture there is an undercurrent of recovery, if you know where to look. The Labour Force Survey (LFS) – which underpins the ONS’s headline productivity figures – has been marred with issues since the pandemic (although it has improved recently, and now more closely aligns with other employment estimates). In response, the Resolution Foundation has created an alternative series to estimate UK employment, which can be used to measure productivity too. Instead of relying on the LFS’s count of workers, we measure the number of employees in administrative records from HMRC’s PAYE Real Time Information (RTI), and the number of self-employed people from self-assessment tax returns.[1] Because RTI is a near-census rather than a survey, it isn’t affected by the falling response rates that have distorted the LFS. Multiplying by average hours per worker gives us an estimate for total hours worked (and output per hour) that is largely based on reliable administrative data. However, as we still rely on the LFS to measure average hours per worker, our alternative productivity estimate only corrects for who is working, not how long they work for.[2]

The two productivity measures both imply barely any productivity growth since the pandemic but they produce notably different trends in the most recent years. According to the LFS, productivity has been slowly falling for some time: over the two years to Q2 2026, LFS-based output per hour fell by an average of 0.2 per cent a year. In the two years prior to that it fell by 0.1 per cent a year.[3] However, using our preferred admin-based estimates, the same two periods show a swing from an average fall of 0.7 per cent a year to growth of 1.1 per cent a year. Both measures use the same output series from the National Accounts, so the difference comes down to how much labour was used to produce it. In 2025, the LFS recorded employment growth of 1.7 per cent while the admin data recorded a fall of 0.1 per cent – an implausible gap that is more likely to reflect survey improvements rather than genuine employment growth. The ONS takes the same view: since May 2026 it has advised users to focus on its administrative-data estimates as the best guide to how productivity is changing.

When measured properly, annual growth in output per hour has averaged 1.1 per cent in the two years of data since Q3 2024, a sharp turnaround from an average fall of 0.7 per cent a year in the two years prior, but also considerably faster than the 0.7 per cent a year seen in the late 2010s (between 2016 to 2019). Figure 1 shows three measures of productivity: all using the same output series, with employment measured by administrative data, the LFS and the ONS’s Workforce Jobs (WFJ). The LFS has muddied the waters, showing productivity growth of −0.2 per cent a year over the past two years. But the two more reliable measures both point clearly the other way, at 1.1 per cent a year on Workforce Jobs and our admin-based estimate. To be clear, productivity is still far below its pre-pandemic path. But the most reliable data we have suggests a clear change in the direction of travel. Notably, after the OBR revised downward its medium-term assumption for UK productivity growth to around 1 per cent, its outlook is now in line with reality for arguably the first time ever. In the rest of this Outlook, we unpick what has driven the recent rebound.

Figure 1: Measured properly, productivity growth has picked up since late 2024

The pickup is bigger than the cyclical behaviour of productivity growth would suggest

In an accounting sense, stronger productivity growth is down to flatlining hours, not faster output growth. Figure 2 sets the four-quarter growth of GDP against that of hours worked over the past decade. Over the two years to Q2 2026, GDP grew by 1.3 per cent a year while hours worked were essentially flat (growing by only 0.2 per cent a year). In the two years prior to Q2 2024, GDP growth was a little slower, at 0.9 per cent a year, but hours worked grew much faster – at an average rate of 1.6 per cent a year. Three-quarters (77 per cent) of the pickup in productivity growth between those two periods comes from slower hours growth. The contrast with the late 2010s is starker. Average four-quarter GDP growth over the past two years (1.3 per cent) has been below the average rate in Q1 2016-Q4 2019 (2.0 per cent). So the higher rate of productivity growth in the recent period is entirely due to weaker growth in hours worked (0.2 per cent in Q3 2024-Q2 2026, versus 1.3 per cent rate in Q1 2016-Q4 2019).

Figure 2: Faster productivity growth has been mainly driven by flatlining hours – not faster GDP growth

One might think that a productivity rebound is inevitable as the labour market cools, and economic theory suggests reasons why this is would be the case. With a slow-moving capital stock, a sharp fall in hours growth leads to a mechanical rise in capital per worker. Compositional effects also play a role. For an individual worker, the marginal hours trimmed when business is quiet are typically less productive than average. And firms tend to shed their least productive workers first when the job market enters a downturn. In the US – with its hyper-flexible labour market – these channels mean that productivity growth is negatively correlated with the labour market cycle. There, when jobs and hours growth falls, productivity growth predictably rises. But the UK is not the US: between 1972 and 2019, UK productivity growth was positively correlated with employment growth and only slightly negatively correlated with hours growth (the sign reversal is because hours adjust by more than jobs through the cycle).[4] Given this historic correlation, the 1.4 percentage point slowdown in hours growth between the two years to Q2 2024 and the two years to Q2 2026 would be expected to raise productivity growth by just 0.3 percentage points – explaining only a sixth of 1.9 percentage point improvement that we have seen.

So, if we look through the fog of Britain’s labour market statistics, there has been a much-needed improvement in UK productivity growth in the past two years. And it is larger than the usual cyclical pattern would suggest. But what is driving it? There are two broad possibilities: either the recent labour market slowdown has fallen hardest on lower paid workers – with jobs disappearing in lower productivity sectors and occupations – or workers have become more productive in general. The first would show up either as a reallocation of workers between sectors, or as a change in the mix of jobs within sectors. Yet it appears to be neither.

The reallocation of workers between sectors continues to have a limited impact on aggregate productivity

One way to raise aggregate labour productivity is to move workers from less productive sectors to more productive ones, and the long-run decline in exactly this sort of reallocation is a key driver of Britain’s economic stagnation. So, it matters whether the recent pickup in productivity can be attributed to it. Indeed, some have suggested that bringing about this sort of reallocation was a deliberate aim of the Starmer government’s recent legislation (although this has been explicitly denied by those involved). The policies in question are the April 2025 increases to employer National Insurance contributions (NICs), the National Living Wage (NLW) and the minimum wage (NMW), all of which raised the cost of employing low-paid workers. Our own analysis put the combined effect of these policies at a 14 per cent rise in the cost of employing a part-time minimum-wage worker, which could result in around 80,000 fewer jobs.

Whether deliberate or incidental, these changes haven’t affected the distribution of workers across sectors of the economy at the scale needed to explain the recent rise in productivity growth. Figure 3 decomposes four-quarter growth in output per hour into productivity growth within 19 high-level industries and the reallocation of hours between them. It shows that reallocation has contributed almost nothing in any recent period, outside the lockdown-induced upheaval in 2020 to 2022. Compared to an average within-industry contribution of 1.1 percentage points over the past two years, the reallocation term is a rounding error.

At this level of aggregation at least, the industrial structure of employment has barely moved, directly contradicting the idea that Britain’s productivity rebound has been driven by shrinking low-paid sectors. Hospitality, a sector often said to be a particular victim of higher employment costs, accounted for 6.6 per cent of employee jobs in Q2 2026, the same share as in 2016-2019 and just 0.3 percentage points below its peak in the immediate post-Covid rebound (6.9 per cent in Q4 2022). Although there have been larger moves elsewhere. Comparing the two years to Q2 2026 with the late-2010s, the share of jobs in wholesale and retail fell by 1.5 percentage points, the largest fall of any sector. But this reflects a long-term decline due to the rise of online retail, not a recent spate of job losses. Wholesale and retail accounted for 14.8 per cent of jobs over 2016-2019, 13.6 per cent over Q3 2022-Q2 2024 and 13.3 per cent over the past two years. Four-fifths of the 1.5 percentage point fall happened before the past two years.  The largest change in any sector’s proportion of jobs over the whole decade was health, up 1.7 percentage points, again reflecting a long-term trend of expanding healthcare provision. But, in general, sectors of this size do not often grow or shrink their workforce fast enough to explain a sharp turnaround in productivity growth, as seen in the past two years.

Figure 3: Reallocation of workers between industries hasn’t driven UK-wide productivity growth

The productivity pickup is broad-based across industries

Instead of a reallocation of workers between sectors, the recent productivity bounce has been driven by higher average productivity growth within them. So which sectors are performing well?

When moving from the aggregate to individual sectors, we switch our measure of productivity from output per hour to output per job. This is because average hours (taken from the LFS) are measured with considerable uncertainty at the sectoral level and cloud the sectoral picture more than they clarify it.[5] But there is unavoidable uncertainty in the numerator of sectoral productivity, namely gross value added (GVA). Sectoral GVA is particularly uncertain in the recent past, as initial ONS estimates use turnover as a proxy for GVA (at the time of writing, this applies to estimates from Q1 2024 onwards).[6] Nonetheless, averaging over two years should give a reasonable early indication of which parts of the economy are driving Britain’s productivity turnaround.

Figure 4 shows each sector’s average contribution to four-quarter growth in UK output per job across the three periods. The improvement is broad: 12 of the 19 sectors made a positive contribution over the past two years, and 12 contributed more than they did in the two years before. Information and communication (ICT) is the largest single contributor, at 0.5 percentage points a year. At first glance this looks like a clear AI story. Sectoral output grew at a blistering pace of 2.7 per cent in the second quarter of 2026, mainly driven by growth in computer programming activities. However, a fast-growing ICT sector is not new. It has long been a bright spot in the UK economy. The sector contributed 0.4 percentage points a year to aggregate productivity growth in 2016-2019, well before AI existed in any commercial form. Therefore, it is worth resisting the temptation to read its recent strength as purely an AI effect.

Instead, it is services that are driving the pickup, after having fallen furthest. Financial and insurance activities moved from -0.7 to +0.01 percentage points. And the two sectors most often blamed for flattering the figures by laying off workers, have instead helped drive the pickup by becoming more productive. The wholesale and retail sector has seen its productivity contribution increase from -0.2 to +0.2 percentage points, while hospitality went from -0.2 percentage points to zero. Some of this reflects a recovery from a period of acute weakness rather than a boom. Financial and insurance output fell by 8.3 per cent over 2023 and 2024, before growing 0.2 per cent in 2025, leaving it 9.8 per cent below its 2016-2019 average. It has stopped shrinking rather than truly growing.

Figure 4: The pick-up in productivity growth is broad

Nor is there a shaking-out of low-paid workers within sectors

So far, for data reasons, we have based our analysis on 19 high-level sectors of the UK economy. But for an economy of 34 million workers and over 5 million firms, those sectoral groupings are necessarily broad. Manufacturing alone spans shipbuilding and pharmaceuticals. So it may be that productivity growth is being boosted by a fall in low-productivity jobs within sectors rather than a shrinking of low-productivity sectors themselves.

If this were happening, it would probably show up in the occupational mix of employees. Occupational classifications are absent from the most reliable administrative labour market data that we have used so far, but they are captured at a very granular level in the Annual Survey of Hours and Earnings (ASHE), which measures wages and hours worked for more than 400 four-digit occupational groups. These are narrow categories: chefs and kitchen assistants are separate occupations, as are financial managers and financial accounts managers.

If sectoral productivity had been boosted by low-paid workers being shed, we would expect to see employment growth slow most for low-paid occupations. In fact, the opposite was true. Take the year to April 2025, during which productivity grew by around 1.7 per cent and which covered the announcement and the first month of the implementation of minimum wage and employer NICs changes.[7] The number of jobs grew by 3.4 per cent in occupations paying below the median in 2024, against 1.4 per cent in those paying above it – a gap of 2 percentage points in favour of lower-paid occupations.

Figure 5 shows what the broader composition of employment growth across occupations meant for pay. The shifting pattern of hours worked across occupations contributed -0.1 percentage points to real hourly pay growth in the year to April 2025 – that is, work shifted slightly towards lower-paying occupations over the year – against an average contribution of +0.7 percentage points over the preceding three years. Essentially all the 2.3 per cent real pay growth in 2025 came from higher pay within occupations rather than from a change in the mix of occupations.

There is little evidence, then, that the UK’s productivity bounce has been driven by a shake-out of lower-paid jobs (at least in the early stages up to April 2025). That is not to say that increasing employer NICs and the NMW had no labour market impact – just that any such effect is too small to show up here. That’s not necessarily surprising when their combined impact has been estimated at around 80,000 jobs – a mere 0.2 per cent of total employment.

It turns out that neither the reallocation of workers between 19 sectors nor the changing occupational mix within them has contributed anything material to the pickup. Whatever is raising productivity is doing it with the same workers, in much the same jobs, in much the same sectors.

Figure 5: There’s no evidence of a shakeout of low-paid occupations within industries

Capital Deepening Also can’t explain the recent productivity rebound

If a shakeout of low-paid workers does not appear to be driving Britain’s productivity rebound, what is? One place to look is capital deepening, or the amount of kit that workers have to work with. Along with falling dynamism, this has been another important drag on UK economic growth for decades. And it is an issue that Labour has taken seriously since coming into power, with former Chancellor Rachel Reeves cancelling the large cuts planned to public investment announced by the previous government.

But that only spared Britain a future drag on productivity rather than raising growth in the here and now. Labour’s plans set public sector net investment at 2.7 per cent of GDP in 2025-26, only marginally above its level in the previous two years (2.4 per cent in 2023-24 and 2.6 per cent in 2024-25) – levels largely set by the previous Conservative government. With a net capital stock worth over £6 trillion, or some 200 per cent of GDP, such a marginal uplift in public investment cannot explain the size of the productivity rebound.

Looking beyond public investment alone, overall economy-wide investment has also been flat in recent years: total gross fixed capital formation (GFCF) was 19 per cent of GDP in both 2023 and 2025.[8] For capital deepening to explain a pickup in productivity growth, a sharp uptick in the level of investment is needed. This happened in the mid-2010s. As investment recovered from the financial crisis, so did the ONS’s estimate of capital’s contribution to market-sector productivity growth. Between 2010 and 2016, investment rose by 2 percentage points of GDP, and capital deepening went from a 0.4 percentage point drag on productivity growth to a 1.0 percentage point boost. This shows how sustained higher investment is essential for Britain’s long-run economic success. But it cannot explain the past two years.

It’s hard to tell what is causing faster productivity growth – but that is a good problem to have

With reallocation and capital deepening ruled out, we are left with what looks like one of the most mysterious quantities in economics: total factor productivity (TFP) growth. In theory, TFP is the efficiency with which labour and capital are combined to produce output, but in practice it is measured as a growth-accounting residual. Moses Abramovitz, writing in 1956, called it a “measure of our ignorance about the causes of economic growth”. While we are sure that a slowdown in TFP growth played a key role in the global economic stagnation that followed the financial crisis, we are less sure about the fundamental reasons why it slowed.

And if the recent rebound does turn out to be driven by TFP growth, we may never fully understand why it has returned. Chalking it up to the AI rollout seems premature. It’s true that the prevalence of AI use has expanded rapidly: excluding micro-SMEs with fewer than 10 employees, the proportion of UK businesses using AI has doubled in the past three years, from 23 per cent in September-October 2023 to 46 per cent in June 2026. But only one-in-ten (9.9 per cent) of these AI-using businesses – or around one-in-twenty (4.6 per cent) of all businesses – are currently using it extensively. UK businesses are certainly positive about the future impact of AI on productivity, but less so on its impact to date. Three-fifths (62 per cent) say it will have a positive impact on revenue per employee in the next three years, versus less than a fifth (17 per cent) who say it has had a positive impact over the past three years.

For now, the revival of productivity growth looks as mysterious as its fall. But an unexplained improvement is a far better problem to have than the unexplained decline that followed the financial crisis. And the good news is not just that it looks real, but that it looks broad: workers across the economy are producing more, rather than the least productive parts of the economy simply shrinking.

[1] Self-employment is the smaller and the weaker half of this measure. According to HMRC’s self-assessment tax data, there are around 3.8 million self-employed people in the UK, or 11 per cent of all workers. Although this number is precise, it is lagged, with the latest data available from 2023-24. For the most recent quarters the self-employment count is carried forward from the latest annual reading.

[2] Average hours worked have not changed substantially in the post-Covid period, so we don’t believe this weakest part of our productivity measure is driving the recent rebound. According to the LFS, the average worker logged 31.9 hours per week in the two years to Q2 2026, compared to an average of 31.8 hours in the two years prior — a difference of around six minutes a week. As a result, the productivity rebound looks similar when measured by output per job rather than output per hour, with growth averaging 1.3 per cent a year in the two years to Q2 2026 versus an average fall of 0.6 per cent in the two years before.

[3] Throughout, we refer to average four-quarter productivity growth in different time periods.

[4] Based on correlations between four-quarter growth in GDP per hour worked and hours worked, and GDP per worker and employment between Q1 1972 and Q4 2019.

[5] Our sectoral jobs measure combines RTI payrolled employees with WFJ self-employment, each scaled so that the all-sector total matches RF’s admin-based employment estimate. For more details, see Annex 1 of S Pittaway, Yanked away: Accounting for the post-pandemic productivity divergence between Britain and America, Resolution Foundation, April 2025, https://doi.org/10.63492/DMXN5237.

[6] On 20 August 2026, the ONS published some initial industry-level data from the Blue Book 2026, which uses supply and use tables to estimate sectoral GVA in 2024. As they have not yet been incorporated into the published National Accounts, we have not incorporated them into our analysis. But an initial assessment suggests that they do not change our main conclusions that the productivity rebound is broad-based – occurring in 12 of the 19 sectors – and that information and communication services are the largest contributor to aggregate productivity growth.

[7] The 2025 wave of the ASHE covers the pay period that includes 30 April 2025. Changes to the NLW and employer NICs that were announced in the October 2024 Budget were, respectively, implemented on 1 April 2025 and 6 April 2025. So the 2025 ASHE captures their impacts in the first month of implementation, plus any anticipatory effect, but not impacts that occurred later.

[8] Real investment has grown a little faster than real GDP over that same period, but that is mostly down to relative price moves. Since 2016, investment goods have become around 4 per cent cheaper, relative to GDP.